2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93
33 * Ethernet address resolution protocol.
35 * add "inuse/lock" bit (or ref. count) along with valid bit
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
43 #include <sys/param.h>
44 #include <sys/kernel.h>
46 #include <sys/queue.h>
47 #include <sys/sysctl.h>
48 #include <sys/systm.h>
50 #include <sys/malloc.h>
52 #include <sys/rmlock.h>
53 #include <sys/socket.h>
54 #include <sys/syslog.h>
57 #include <net/if_var.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/netisr.h>
61 #include <net/ethernet.h>
62 #include <net/route.h>
65 #include <netinet/in.h>
66 #include <netinet/in_fib.h>
67 #include <netinet/in_var.h>
68 #include <net/if_llatbl.h>
69 #include <netinet/if_ether.h>
71 #include <netinet/ip_carp.h>
74 #include <security/mac/mac_framework.h>
76 #define SIN(s) ((const struct sockaddr_in *)(s))
78 static struct timeval arp_lastlog;
79 static int arp_curpps;
80 static int arp_maxpps = 1;
82 /* Simple ARP state machine */
83 enum arp_llinfo_state {
84 ARP_LLINFO_INCOMPLETE = 0, /* No LLE data */
85 ARP_LLINFO_REACHABLE, /* LLE is valid */
86 ARP_LLINFO_VERIFY, /* LLE is valid, need refresh */
87 ARP_LLINFO_DELETED, /* LLE is deleted */
90 SYSCTL_DECL(_net_link_ether);
91 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
92 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
95 static VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20
97 static VNET_DEFINE(int, arp_maxtries) = 5;
98 static VNET_DEFINE(int, arp_proxyall) = 0;
99 static VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for
101 static VNET_DEFINE(int, arpt_rexmit) = 1; /* retransmit arp entries, sec*/
102 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */
103 VNET_PCPUSTAT_SYSINIT(arpstat);
106 VNET_PCPUSTAT_SYSUNINIT(arpstat);
109 static VNET_DEFINE(int, arp_maxhold) = 1;
111 #define V_arpt_keep VNET(arpt_keep)
112 #define V_arpt_down VNET(arpt_down)
113 #define V_arpt_rexmit VNET(arpt_rexmit)
114 #define V_arp_maxtries VNET(arp_maxtries)
115 #define V_arp_proxyall VNET(arp_proxyall)
116 #define V_arp_maxhold VNET(arp_maxhold)
118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW,
119 &VNET_NAME(arpt_keep), 0,
120 "ARP entry lifetime in seconds");
121 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW,
122 &VNET_NAME(arp_maxtries), 0,
123 "ARP resolution attempts before returning error");
124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW,
125 &VNET_NAME(arp_proxyall), 0,
126 "Enable proxy ARP for all suitable requests");
127 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW,
128 &VNET_NAME(arpt_down), 0,
129 "Incomplete ARP entry lifetime in seconds");
130 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
131 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
132 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW,
133 &VNET_NAME(arp_maxhold), 0,
134 "Number of packets to hold per ARP entry");
135 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
136 CTLFLAG_RW, &arp_maxpps, 0,
137 "Maximum number of remotely triggered ARP messages that can be "
138 "logged per second");
140 #define ARP_LOG(pri, ...) do { \
141 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \
142 log((pri), "arp: " __VA_ARGS__); \
146 static void arp_init(void);
147 static void arpintr(struct mbuf *);
148 static void arptimer(void *);
150 static void in_arpinput(struct mbuf *);
153 static void arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr,
154 struct ifnet *ifp, int bridged, struct llentry *la);
155 static void arp_mark_lle_reachable(struct llentry *la);
156 static void arp_iflladdr(void *arg __unused, struct ifnet *ifp);
158 static eventhandler_tag iflladdr_tag;
160 static const struct netisr_handler arp_nh = {
162 .nh_handler = arpintr,
163 .nh_proto = NETISR_ARP,
164 .nh_policy = NETISR_POLICY_SOURCE,
168 * Timeout routine. Age arp_tab entries periodically.
173 struct llentry *lle = (struct llentry *)arg;
177 if (lle->la_flags & LLE_STATIC) {
181 if (callout_pending(&lle->lle_timer)) {
183 * Here we are a bit odd here in the treatment of
184 * active/pending. If the pending bit is set, it got
185 * rescheduled before I ran. The active
186 * bit we ignore, since if it was stopped
187 * in ll_tablefree() and was currently running
188 * it would have return 0 so the code would
189 * not have deleted it since the callout could
190 * not be stopped so we want to go through
191 * with the delete here now. If the callout
192 * was restarted, the pending bit will be back on and
193 * we just want to bail since the callout_reset would
194 * return 1 and our reference would have been removed
195 * by arpresolve() below.
200 ifp = lle->lle_tbl->llt_ifp;
201 CURVNET_SET(ifp->if_vnet);
203 switch (lle->ln_state) {
204 case ARP_LLINFO_REACHABLE:
207 * Expiration time is approaching.
208 * Let's try to refresh entry if it is still
211 * Set r_skip_req to get feedback from
212 * fast path. Change state and re-schedule
218 lle->ln_state = ARP_LLINFO_VERIFY;
219 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
223 case ARP_LLINFO_VERIFY:
225 r_skip_req = lle->r_skip_req;
228 if (r_skip_req == 0 && lle->la_preempt > 0) {
229 /* Entry was used, issue refresh request */
231 dst = lle->r_l3addr.addr4;
233 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
235 arprequest(ifp, NULL, &dst, NULL);
239 /* Nothing happened. Reschedule if not too late */
240 if (lle->la_expire > time_uptime) {
241 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
247 case ARP_LLINFO_INCOMPLETE:
248 case ARP_LLINFO_DELETED:
252 if ((lle->la_flags & LLE_DELETED) == 0) {
255 if (lle->la_flags & LLE_VALID)
256 evt = LLENTRY_EXPIRED;
258 evt = LLENTRY_TIMEDOUT;
259 EVENTHANDLER_INVOKE(lle_event, lle, evt);
262 callout_stop(&lle->lle_timer);
264 /* XXX: LOR avoidance. We still have ref on lle. */
269 /* Guard against race with other llentry_free(). */
270 if (lle->la_flags & LLE_LINKED) {
272 lltable_unlink_entry(lle->lle_tbl, lle);
274 IF_AFDATA_UNLOCK(ifp);
276 size_t pkts_dropped = llentry_free(lle);
278 ARPSTAT_ADD(dropped, pkts_dropped);
279 ARPSTAT_INC(timeouts);
285 * Stores link-layer header for @ifp in format suitable for if_output()
286 * into buffer @buf. Resulting header length is stored in @bufsize.
288 * Returns 0 on success.
291 arp_fillheader(struct ifnet *ifp, struct arphdr *ah, int bcast, u_char *buf,
294 struct if_encap_req ereq;
297 bzero(buf, *bufsize);
298 bzero(&ereq, sizeof(ereq));
300 ereq.bufsize = *bufsize;
301 ereq.rtype = IFENCAP_LL;
302 ereq.family = AF_ARP;
303 ereq.lladdr = ar_tha(ah);
304 ereq.hdata = (u_char *)ah;
306 ereq.flags = IFENCAP_FLAG_BROADCAST;
307 error = ifp->if_requestencap(ifp, &ereq);
309 *bufsize = ereq.bufsize;
316 * Broadcast an ARP request. Caller specifies:
317 * - arp header source ip address
318 * - arp header target ip address
319 * - arp header source ethernet address
322 arprequest(struct ifnet *ifp, const struct in_addr *sip,
323 const struct in_addr *tip, u_char *enaddr)
328 u_char *carpaddr = NULL;
329 uint8_t linkhdr[LLE_MAX_LINKHDR];
336 * The caller did not supply a source address, try to find
337 * a compatible one among those assigned to this interface.
342 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
343 if (ifa->ifa_addr->sa_family != AF_INET)
347 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
349 sip = &IA_SIN(ifa)->sin_addr;
352 sip = &IA_SIN(ifa)->sin_addr;
355 if (0 == ((sip->s_addr ^ tip->s_addr) &
356 IA_MASKSIN(ifa)->sin_addr.s_addr))
357 break; /* found it. */
359 IF_ADDR_RUNLOCK(ifp);
361 printf("%s: cannot find matching address\n", __func__);
366 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
368 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
370 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
372 m->m_pkthdr.len = m->m_len;
373 M_ALIGN(m, m->m_len);
374 ah = mtod(m, struct arphdr *);
375 bzero((caddr_t)ah, m->m_len);
377 mac_netinet_arp_send(ifp, m);
379 ah->ar_pro = htons(ETHERTYPE_IP);
380 ah->ar_hln = ifp->if_addrlen; /* hardware address length */
381 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */
382 ah->ar_op = htons(ARPOP_REQUEST);
383 bcopy(enaddr, ar_sha(ah), ah->ar_hln);
384 bcopy(sip, ar_spa(ah), ah->ar_pln);
385 bcopy(tip, ar_tpa(ah), ah->ar_pln);
386 sa.sa_family = AF_ARP;
389 /* Calculate link header for sending frame */
390 bzero(&ro, sizeof(ro));
391 linkhdrsize = sizeof(linkhdr);
392 error = arp_fillheader(ifp, ah, 1, linkhdr, &linkhdrsize);
393 if (error != 0 && error != EAFNOSUPPORT) {
394 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
395 if_name(ifp), error);
399 ro.ro_prepend = linkhdr;
400 ro.ro_plen = linkhdrsize;
403 m->m_flags |= M_BCAST;
404 m_clrprotoflags(m); /* Avoid confusing lower layers. */
405 (*ifp->if_output)(ifp, m, &sa, &ro);
406 ARPSTAT_INC(txrequests);
411 * Resolve an IP address into an ethernet address - heavy version.
412 * Used internally by arpresolve().
413 * We have already checked than we can't use existing lle without
414 * modification so we have to acquire LLE_EXCLUSIVE lle lock.
416 * On success, desten and flags are filled in and the function returns 0;
417 * If the packet must be held pending resolution, we return EWOULDBLOCK
418 * On other errors, we return the corresponding error code.
419 * Note that m_freem() handles NULL.
422 arpresolve_full(struct ifnet *ifp, int is_gw, int flags, struct mbuf *m,
423 const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
425 struct llentry *la = NULL, *la_tmp;
426 struct mbuf *curr = NULL;
427 struct mbuf *next = NULL;
435 if ((flags & LLE_CREATE) == 0) {
436 IF_AFDATA_RLOCK(ifp);
437 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
438 IF_AFDATA_RUNLOCK(ifp);
440 if (la == NULL && (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) {
441 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
444 "arpresolve: can't allocate llinfo for %s on %s\n",
445 inet_ntoa(SIN(dst)->sin_addr), if_name(ifp));
450 IF_AFDATA_WLOCK(ifp);
452 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
453 /* Prefer ANY existing lle over newly-created one */
455 lltable_link_entry(LLTABLE(ifp), la);
456 IF_AFDATA_WUNLOCK(ifp);
457 if (la_tmp != NULL) {
458 lltable_free_entry(LLTABLE(ifp), la);
467 if ((la->la_flags & LLE_VALID) &&
468 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
469 if (flags & LLE_ADDRONLY) {
470 lladdr = la->ll_addr;
471 ll_len = ifp->if_addrlen;
473 lladdr = la->r_linkdata;
474 ll_len = la->r_hdrlen;
476 bcopy(lladdr, desten, ll_len);
478 /* Check if we have feedback request from arptimer() */
479 if (la->r_skip_req != 0) {
481 la->r_skip_req = 0; /* Notify that entry was used */
485 *pflags = la->la_flags & (LLE_VALID|LLE_IFADDR);
490 renew = (la->la_asked == 0 || la->la_expire != time_uptime);
492 * There is an arptab entry, but no ethernet address
493 * response yet. Add the mbuf to the list, dropping
494 * the oldest packet if we have exceeded the system
498 if (la->la_numheld >= V_arp_maxhold) {
499 if (la->la_hold != NULL) {
500 next = la->la_hold->m_nextpkt;
501 m_freem(la->la_hold);
504 ARPSTAT_INC(dropped);
507 if (la->la_hold != NULL) {
509 while (curr->m_nextpkt != NULL)
510 curr = curr->m_nextpkt;
517 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
518 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
519 * if we have already sent arp_maxtries ARP requests. Retransmit the
520 * ARP request, but not faster than one request per second.
522 if (la->la_asked < V_arp_maxtries)
523 error = EWOULDBLOCK; /* First request. */
525 error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN;
531 la->la_expire = time_uptime;
532 canceled = callout_reset(&la->lle_timer, hz * V_arpt_down,
538 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
547 * Resolve an IP address into an ethernet address.
550 arpresolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
551 char *desten, uint32_t *pflags)
555 flags |= LLE_ADDRONLY;
556 error = arpresolve_full(ifp, 0, flags, NULL, dst, desten, pflags);
562 * Lookups link header based on an IP address.
564 * ifp is the interface we use
565 * is_gw != 0 if @dst represents gateway to some destination
566 * m is the mbuf. May be NULL if we don't have a packet.
567 * dst is the next hop,
568 * desten is the storage to put LL header.
569 * flags returns subset of lle flags: LLE_VALID | LLE_IFADDR
571 * On success, full/partial link header and flags are filled in and
572 * the function returns 0.
573 * If the packet must be held pending resolution, we return EWOULDBLOCK
574 * On other errors, we return the corresponding error code.
575 * Note that m_freem() handles NULL.
578 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
579 const struct sockaddr *dst, u_char *desten, uint32_t *pflags)
581 struct llentry *la = 0;
587 if (m->m_flags & M_BCAST) {
590 ifp->if_broadcastaddr, ifp->if_addrlen);
593 if (m->m_flags & M_MCAST) {
595 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
600 IF_AFDATA_RLOCK(ifp);
601 la = lla_lookup(LLTABLE(ifp), LLE_UNLOCKED, dst);
602 if (la != NULL && (la->r_flags & RLLE_VALID) != 0) {
603 /* Entry found, let's copy lle info */
604 bcopy(la->r_linkdata, desten, la->r_hdrlen);
606 *pflags = LLE_VALID | (la->r_flags & RLLE_IFADDR);
607 /* Check if we have feedback request from arptimer() */
608 if (la->r_skip_req != 0) {
610 la->r_skip_req = 0; /* Notify that entry was used */
613 IF_AFDATA_RUNLOCK(ifp);
616 IF_AFDATA_RUNLOCK(ifp);
618 return (arpresolve_full(ifp, is_gw, la == NULL ? LLE_CREATE : 0, m, dst,
623 * Common length and type checks are done here,
624 * then the protocol-specific routine is called.
627 arpintr(struct mbuf *m)
634 ifp = m->m_pkthdr.rcvif;
636 if (m->m_len < sizeof(struct arphdr) &&
637 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
638 ARP_LOG(LOG_NOTICE, "packet with short header received on %s\n",
642 ar = mtod(m, struct arphdr *);
644 /* Check if length is sufficient */
645 if (m->m_len < arphdr_len(ar)) {
646 m = m_pullup(m, arphdr_len(ar));
648 ARP_LOG(LOG_NOTICE, "short packet received on %s\n",
652 ar = mtod(m, struct arphdr *);
657 switch (ntohs(ar->ar_hrd)) {
659 hlen = ETHER_ADDR_LEN; /* RFC 826 */
663 hlen = 6; /* RFC 1390, FDDI_ADDR_LEN */
667 hlen = 1; /* RFC 1201, ARC_ADDR_LEN */
670 case ARPHRD_INFINIBAND:
671 hlen = 20; /* RFC 4391, INFINIBAND_ALEN */
672 layer = "infiniband";
674 case ARPHRD_IEEE1394:
675 hlen = 0; /* SHALL be 16 */ /* RFC 2734 */
679 * Restrict too long hardware addresses.
680 * Currently we are capable of handling 20-byte
681 * addresses ( sizeof(lle->ll_addr) )
683 if (ar->ar_hln >= 20)
688 "packet with unknown hardware format 0x%02d received on "
689 "%s\n", ntohs(ar->ar_hrd), if_name(ifp));
694 if (hlen != 0 && hlen != ar->ar_hln) {
696 "packet with invalid %s address length %d received on %s\n",
697 layer, ar->ar_hln, if_name(ifp));
702 ARPSTAT_INC(received);
703 switch (ntohs(ar->ar_pro)) {
715 * ARP for Internet protocols on 10 Mb/s Ethernet.
716 * Algorithm is that given in RFC 826.
717 * In addition, a sanity check is performed on the sender
718 * protocol address, to catch impersonators.
719 * We no longer handle negotiations for use of trailer protocol:
720 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
721 * along with IP replies if we wanted trailers sent to us,
722 * and also sent them in response to IP replies.
723 * This allowed either end to announce the desire to receive
725 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
726 * but formerly didn't normally send requests.
728 static int log_arp_wrong_iface = 1;
729 static int log_arp_movements = 1;
730 static int log_arp_permanent_modify = 1;
731 static int allow_multicast = 0;
733 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
734 &log_arp_wrong_iface, 0,
735 "log arp packets arriving on the wrong interface");
736 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
737 &log_arp_movements, 0,
738 "log arp replies from MACs different than the one in the cache");
739 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
740 &log_arp_permanent_modify, 0,
741 "log arp replies from MACs different than the one in the permanent arp entry");
742 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
743 &allow_multicast, 0, "accept multicast addresses");
746 in_arpinput(struct mbuf *m)
748 struct rm_priotracker in_ifa_tracker;
750 struct ifnet *ifp = m->m_pkthdr.rcvif;
751 struct llentry *la = NULL, *la_tmp;
753 struct in_ifaddr *ia;
755 struct in_addr isaddr, itaddr, myaddr;
756 u_int8_t *enaddr = NULL;
758 int bridged = 0, is_bridge = 0;
760 struct sockaddr_in sin;
761 struct sockaddr *dst;
762 struct nhop4_basic nh4;
763 uint8_t linkhdr[LLE_MAX_LINKHDR];
769 sin.sin_len = sizeof(struct sockaddr_in);
770 sin.sin_family = AF_INET;
771 sin.sin_addr.s_addr = 0;
775 if (ifp->if_type == IFT_BRIDGE)
779 * We already have checked that mbuf contains enough contiguous data
780 * to hold entire arp message according to the arp header.
782 ah = mtod(m, struct arphdr *);
785 * ARP is only for IPv4 so we can reject packets with
786 * a protocol length not equal to an IPv4 address.
788 if (ah->ar_pln != sizeof(struct in_addr)) {
789 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
790 sizeof(struct in_addr));
794 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
795 ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
796 ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
800 op = ntohs(ah->ar_op);
801 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
802 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
804 if (op == ARPOP_REPLY)
805 ARPSTAT_INC(rxreplies);
808 * For a bridge, we want to check the address irrespective
809 * of the receive interface. (This will change slightly
810 * when we have clusters of interfaces).
812 IN_IFADDR_RLOCK(&in_ifa_tracker);
813 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
814 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
815 ia->ia_ifp == ifp) &&
816 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
817 (ia->ia_ifa.ifa_carp == NULL ||
818 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
819 ifa_ref(&ia->ia_ifa);
820 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
824 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
825 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
826 ia->ia_ifp == ifp) &&
827 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
828 ifa_ref(&ia->ia_ifa);
829 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
833 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \
834 (ia->ia_ifp->if_bridge == ifp->if_softc && \
835 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \
836 addr == ia->ia_addr.sin_addr.s_addr)
838 * Check the case when bridge shares its MAC address with
839 * some of its children, so packets are claimed by bridge
840 * itself (bridge_input() does it first), but they are really
841 * meant to be destined to the bridge member.
844 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
845 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
846 ifa_ref(&ia->ia_ifa);
848 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
853 #undef BDG_MEMBER_MATCHES_ARP
854 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
857 * No match, use the first inet address on the receive interface
858 * as a dummy address for the rest of the function.
861 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
862 if (ifa->ifa_addr->sa_family == AF_INET &&
863 (ifa->ifa_carp == NULL ||
864 (*carp_iamatch_p)(ifa, &enaddr))) {
867 IF_ADDR_RUNLOCK(ifp);
870 IF_ADDR_RUNLOCK(ifp);
873 * If bridging, fall back to using any inet address.
875 IN_IFADDR_RLOCK(&in_ifa_tracker);
876 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
877 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
880 ifa_ref(&ia->ia_ifa);
881 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
884 enaddr = (u_int8_t *)IF_LLADDR(ifp);
885 carped = (ia->ia_ifa.ifa_carp != NULL);
886 myaddr = ia->ia_addr.sin_addr;
887 ifa_free(&ia->ia_ifa);
888 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
889 goto drop; /* it's from me, ignore it. */
890 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
891 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
892 "%s!\n", inet_ntoa(isaddr));
896 if (ifp->if_addrlen != ah->ar_hln) {
897 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
898 "i/f %d (ignored)\n", ifp->if_addrlen,
899 (u_char *) ar_sha(ah), ":", ah->ar_hln,
905 * Warn if another host is using the same IP address, but only if the
906 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
907 * case we suppress the warning to avoid false positive complaints of
908 * potential misconfiguration.
910 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
911 myaddr.s_addr != 0) {
912 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
913 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
914 inet_ntoa(isaddr), ifp->if_xname);
919 if (ifp->if_flags & IFF_STATICARP)
922 bzero(&sin, sizeof(sin));
923 sin.sin_len = sizeof(struct sockaddr_in);
924 sin.sin_family = AF_INET;
925 sin.sin_addr = isaddr;
926 dst = (struct sockaddr *)&sin;
927 IF_AFDATA_RLOCK(ifp);
928 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
929 IF_AFDATA_RUNLOCK(ifp);
931 arp_check_update_lle(ah, isaddr, ifp, bridged, la);
932 else if (itaddr.s_addr == myaddr.s_addr) {
934 * Request/reply to our address, but no lle exists yet.
935 * Calculate full link prepend to use in lle.
937 linkhdrsize = sizeof(linkhdr);
938 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
939 &linkhdrsize, &lladdr_off) != 0)
942 /* Allocate new entry */
943 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
947 * lle creation may fail if source address belongs
948 * to non-directly connected subnet. However, we
949 * will try to answer the request instead of dropping
954 lltable_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
957 IF_AFDATA_WLOCK(ifp);
959 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
962 * Check if lle still does not exists.
963 * If it does, that means that we either
964 * 1) have configured it explicitly, via
965 * 1a) 'arp -s' static entry or
966 * 1b) interface address static record
968 * 2) it was the result of sending first packet to-host
970 * 3) it was another arp reply packet we handled in
973 * In all cases except 3) we definitely need to prefer
974 * existing lle. For the sake of simplicity, prefer any
975 * existing lle over newly-create one.
978 lltable_link_entry(LLTABLE(ifp), la);
979 IF_AFDATA_WUNLOCK(ifp);
981 if (la_tmp == NULL) {
982 arp_mark_lle_reachable(la);
985 /* Free newly-create entry and handle packet */
986 lltable_free_entry(LLTABLE(ifp), la);
989 arp_check_update_lle(ah, isaddr, ifp, bridged, la);
990 /* arp_check_update_lle() returns @la unlocked */
995 if (op != ARPOP_REQUEST)
997 ARPSTAT_INC(rxrequests);
999 if (itaddr.s_addr == myaddr.s_addr) {
1000 /* Shortcut.. the receiving interface is the target. */
1001 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1002 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1004 struct llentry *lle = NULL;
1006 sin.sin_addr = itaddr;
1007 IF_AFDATA_RLOCK(ifp);
1008 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1009 IF_AFDATA_RUNLOCK(ifp);
1011 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1012 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1013 (void)memcpy(ar_sha(ah), lle->ll_addr, ah->ar_hln);
1020 if (!V_arp_proxyall)
1023 /* XXX MRT use table 0 for arp reply */
1024 if (fib4_lookup_nh_basic(0, itaddr, 0, 0, &nh4) != 0)
1028 * Don't send proxies for nodes on the same interface
1029 * as this one came out of, or we'll get into a fight
1030 * over who claims what Ether address.
1032 if (nh4.nh_ifp == ifp)
1035 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1036 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1039 * Also check that the node which sent the ARP packet
1040 * is on the interface we expect it to be on. This
1041 * avoids ARP chaos if an interface is connected to the
1045 /* XXX MRT use table 0 for arp checks */
1046 if (fib4_lookup_nh_basic(0, isaddr, 0, 0, &nh4) != 0)
1048 if (nh4.nh_ifp != ifp) {
1049 ARP_LOG(LOG_INFO, "proxy: ignoring request"
1050 " from %s via %s\n",
1051 inet_ntoa(isaddr), ifp->if_xname);
1056 printf("arp: proxying for %s\n", inet_ntoa(itaddr));
1061 if (itaddr.s_addr == myaddr.s_addr &&
1062 IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
1063 /* RFC 3927 link-local IPv4; always reply by broadcast. */
1064 #ifdef DEBUG_LINKLOCAL
1065 printf("arp: sending reply for link-local addr %s\n",
1068 m->m_flags |= M_BCAST;
1069 m->m_flags &= ~M_MCAST;
1071 /* default behaviour; never reply by broadcast. */
1072 m->m_flags &= ~(M_BCAST|M_MCAST);
1074 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1075 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1076 ah->ar_op = htons(ARPOP_REPLY);
1077 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1078 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1079 m->m_pkthdr.len = m->m_len;
1080 m->m_pkthdr.rcvif = NULL;
1081 sa.sa_family = AF_ARP;
1084 /* Calculate link header for sending frame */
1085 bzero(&ro, sizeof(ro));
1086 linkhdrsize = sizeof(linkhdr);
1087 error = arp_fillheader(ifp, ah, 0, linkhdr, &linkhdrsize);
1090 * arp_fillheader() may fail due to lack of support inside encap request
1091 * routing. This is not necessary an error, AF_ARP can/should be handled
1094 if (error != 0 && error != EAFNOSUPPORT) {
1095 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
1096 if_name(ifp), error);
1100 ro.ro_prepend = linkhdr;
1101 ro.ro_plen = linkhdrsize;
1104 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1105 (*ifp->if_output)(ifp, m, &sa, &ro);
1106 ARPSTAT_INC(txreplies);
1115 * Checks received arp data against existing @la.
1116 * Updates lle state/performs notification if necessary.
1119 arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, struct ifnet *ifp,
1120 int bridged, struct llentry *la)
1123 struct mbuf *m_hold, *m_hold_next;
1124 uint8_t linkhdr[LLE_MAX_LINKHDR];
1128 LLE_WLOCK_ASSERT(la);
1130 /* the following is not an error when doing bridging */
1131 if (!bridged && la->lle_tbl->llt_ifp != ifp) {
1132 if (log_arp_wrong_iface)
1133 ARP_LOG(LOG_WARNING, "%s is on %s "
1134 "but got reply from %*D on %s\n",
1136 la->lle_tbl->llt_ifp->if_xname,
1137 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1142 if ((la->la_flags & LLE_VALID) &&
1143 bcmp(ar_sha(ah), la->ll_addr, ifp->if_addrlen)) {
1144 if (la->la_flags & LLE_STATIC) {
1146 if (log_arp_permanent_modify)
1148 "%*D attempts to modify "
1149 "permanent entry for %s on %s\n",
1151 (u_char *)ar_sha(ah), ":",
1152 inet_ntoa(isaddr), ifp->if_xname);
1155 if (log_arp_movements) {
1156 ARP_LOG(LOG_INFO, "%s moved from %*D "
1160 (u_char *)&la->ll_addr, ":",
1161 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1166 /* Calculate full link prepend to use in lle */
1167 linkhdrsize = sizeof(linkhdr);
1168 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
1169 &linkhdrsize, &lladdr_off) != 0)
1172 /* Check if something has changed */
1173 if (memcmp(la->r_linkdata, linkhdr, linkhdrsize) != 0 ||
1174 (la->la_flags & LLE_VALID) == 0) {
1175 /* Try to perform LLE update */
1176 if (lltable_try_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
1180 /* Clear fast path feedback request if set */
1184 arp_mark_lle_reachable(la);
1187 * The packets are all freed within the call to the output
1190 * NB: The lock MUST be released before the call to the
1193 if (la->la_hold != NULL) {
1194 m_hold = la->la_hold;
1197 lltable_fill_sa_entry(la, &sa);
1199 for (; m_hold != NULL; m_hold = m_hold_next) {
1200 m_hold_next = m_hold->m_nextpkt;
1201 m_hold->m_nextpkt = NULL;
1202 /* Avoid confusing lower layers. */
1203 m_clrprotoflags(m_hold);
1204 (*ifp->if_output)(ifp, m_hold, &sa, NULL);
1211 arp_mark_lle_reachable(struct llentry *la)
1213 int canceled, wtime;
1215 LLE_WLOCK_ASSERT(la);
1217 la->ln_state = ARP_LLINFO_REACHABLE;
1218 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
1220 if (!(la->la_flags & LLE_STATIC)) {
1222 la->la_expire = time_uptime + V_arpt_keep;
1223 wtime = V_arpt_keep - V_arp_maxtries * V_arpt_rexmit;
1225 wtime = V_arpt_keep;
1226 canceled = callout_reset(&la->lle_timer,
1227 hz * wtime, arptimer, la);
1232 la->la_preempt = V_arp_maxtries;
1236 * Add pernament link-layer record for given interface address.
1238 static __noinline void
1239 arp_add_ifa_lle(struct ifnet *ifp, const struct sockaddr *dst)
1241 struct llentry *lle, *lle_tmp;
1244 * Interface address LLE record is considered static
1245 * because kernel code relies on LLE_STATIC flag to check
1246 * if these entries can be rewriten by arp updates.
1248 lle = lltable_alloc_entry(LLTABLE(ifp), LLE_IFADDR | LLE_STATIC, dst);
1250 log(LOG_INFO, "arp_ifinit: cannot create arp "
1251 "entry for interface address\n");
1255 IF_AFDATA_WLOCK(ifp);
1257 /* Unlink any entry if exists */
1258 lle_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
1259 if (lle_tmp != NULL)
1260 lltable_unlink_entry(LLTABLE(ifp), lle_tmp);
1262 lltable_link_entry(LLTABLE(ifp), lle);
1263 IF_AFDATA_WUNLOCK(ifp);
1265 if (lle_tmp != NULL)
1266 EVENTHANDLER_INVOKE(lle_event, lle_tmp, LLENTRY_EXPIRED);
1268 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED);
1270 if (lle_tmp != NULL)
1271 lltable_free_entry(LLTABLE(ifp), lle_tmp);
1275 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1277 const struct sockaddr_in *dst_in;
1278 const struct sockaddr *dst;
1280 if (ifa->ifa_carp != NULL)
1283 dst = ifa->ifa_addr;
1284 dst_in = (const struct sockaddr_in *)dst;
1286 if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY)
1288 arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp));
1290 arp_add_ifa_lle(ifp, dst);
1294 arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr)
1297 if (ntohl(addr.s_addr) != INADDR_ANY)
1298 arprequest(ifp, &addr, &addr, enaddr);
1302 * Sends gratuitous ARPs for each ifaddr to notify other
1303 * nodes about the address change.
1305 static __noinline void
1306 arp_handle_ifllchange(struct ifnet *ifp)
1310 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1311 if (ifa->ifa_addr->sa_family == AF_INET)
1312 arp_ifinit(ifp, ifa);
1317 * A handler for interface link layer address change event.
1320 arp_iflladdr(void *arg __unused, struct ifnet *ifp)
1323 lltable_update_ifaddr(LLTABLE(ifp));
1325 if ((ifp->if_flags & IFF_UP) != 0)
1326 arp_handle_ifllchange(ifp);
1333 netisr_register(&arp_nh);
1334 if (IS_DEFAULT_VNET(curvnet))
1335 iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event,
1336 arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
1338 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);