2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
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 University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93
35 * Ethernet address resolution protocol.
37 * add "inuse/lock" bit (or ref. count) along with valid bit
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
45 #include <sys/param.h>
46 #include <sys/kernel.h>
48 #include <sys/queue.h>
49 #include <sys/sysctl.h>
50 #include <sys/systm.h>
52 #include <sys/malloc.h>
54 #include <sys/rmlock.h>
55 #include <sys/socket.h>
56 #include <sys/syslog.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/if_types.h>
62 #include <net/netisr.h>
63 #include <net/ethernet.h>
64 #include <net/route.h>
67 #include <netinet/in.h>
68 #include <netinet/in_fib.h>
69 #include <netinet/in_var.h>
70 #include <net/if_llatbl.h>
71 #include <netinet/if_ether.h>
73 #include <netinet/ip_carp.h>
76 #include <security/mac/mac_framework.h>
78 #define SIN(s) ((const struct sockaddr_in *)(s))
80 static struct timeval arp_lastlog;
81 static int arp_curpps;
82 static int arp_maxpps = 1;
84 /* Simple ARP state machine */
85 enum arp_llinfo_state {
86 ARP_LLINFO_INCOMPLETE = 0, /* No LLE data */
87 ARP_LLINFO_REACHABLE, /* LLE is valid */
88 ARP_LLINFO_VERIFY, /* LLE is valid, need refresh */
89 ARP_LLINFO_DELETED, /* LLE is deleted */
92 SYSCTL_DECL(_net_link_ether);
93 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
94 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
97 VNET_DEFINE_STATIC(int, arpt_keep) = (20*60); /* once resolved, good for 20
99 VNET_DEFINE_STATIC(int, arp_maxtries) = 5;
100 VNET_DEFINE_STATIC(int, arp_proxyall) = 0;
101 VNET_DEFINE_STATIC(int, arpt_down) = 20; /* keep incomplete entries for
103 VNET_DEFINE_STATIC(int, arpt_rexmit) = 1; /* retransmit arp entries, sec*/
104 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */
105 VNET_PCPUSTAT_SYSINIT(arpstat);
108 VNET_PCPUSTAT_SYSUNINIT(arpstat);
111 VNET_DEFINE_STATIC(int, arp_maxhold) = 1;
113 #define V_arpt_keep VNET(arpt_keep)
114 #define V_arpt_down VNET(arpt_down)
115 #define V_arpt_rexmit VNET(arpt_rexmit)
116 #define V_arp_maxtries VNET(arp_maxtries)
117 #define V_arp_proxyall VNET(arp_proxyall)
118 #define V_arp_maxhold VNET(arp_maxhold)
120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW,
121 &VNET_NAME(arpt_keep), 0,
122 "ARP entry lifetime in seconds");
123 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW,
124 &VNET_NAME(arp_maxtries), 0,
125 "ARP resolution attempts before returning error");
126 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW,
127 &VNET_NAME(arp_proxyall), 0,
128 "Enable proxy ARP for all suitable requests");
129 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW,
130 &VNET_NAME(arpt_down), 0,
131 "Incomplete ARP entry lifetime in seconds");
132 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
133 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
134 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW,
135 &VNET_NAME(arp_maxhold), 0,
136 "Number of packets to hold per ARP entry");
137 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
138 CTLFLAG_RW, &arp_maxpps, 0,
139 "Maximum number of remotely triggered ARP messages that can be "
140 "logged per second");
143 * Due to the exponential backoff algorithm used for the interval between GARP
144 * retransmissions, the maximum number of retransmissions is limited for
145 * sanity. This limit corresponds to a maximum interval between retransmissions
146 * of 2^16 seconds ~= 18 hours.
148 * Making this limit more dynamic is more complicated than worthwhile,
149 * especially since sending out GARPs spaced days apart would be of little
150 * use. A maximum dynamic limit would look something like:
152 * const int max = fls(INT_MAX / hz) - 1;
154 #define MAX_GARP_RETRANSMITS 16
155 static int sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS);
156 static int garp_rexmit_count = 0; /* GARP retransmission setting. */
158 SYSCTL_PROC(_net_link_ether_inet, OID_AUTO, garp_rexmit_count,
159 CTLTYPE_INT|CTLFLAG_RW|CTLFLAG_MPSAFE,
160 &garp_rexmit_count, 0, sysctl_garp_rexmit, "I",
161 "Number of times to retransmit GARP packets;"
162 " 0 to disable, maximum of 16");
164 #define ARP_LOG(pri, ...) do { \
165 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \
166 log((pri), "arp: " __VA_ARGS__); \
170 static void arpintr(struct mbuf *);
171 static void arptimer(void *);
173 static void in_arpinput(struct mbuf *);
176 static void arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr,
177 struct ifnet *ifp, int bridged, struct llentry *la);
178 static void arp_mark_lle_reachable(struct llentry *la);
179 static void arp_iflladdr(void *arg __unused, struct ifnet *ifp);
181 static eventhandler_tag iflladdr_tag;
183 static const struct netisr_handler arp_nh = {
185 .nh_handler = arpintr,
186 .nh_proto = NETISR_ARP,
187 .nh_policy = NETISR_POLICY_SOURCE,
191 * Timeout routine. Age arp_tab entries periodically.
196 struct llentry *lle = (struct llentry *)arg;
200 if (lle->la_flags & LLE_STATIC) {
204 if (callout_pending(&lle->lle_timer)) {
206 * Here we are a bit odd here in the treatment of
207 * active/pending. If the pending bit is set, it got
208 * rescheduled before I ran. The active
209 * bit we ignore, since if it was stopped
210 * in ll_tablefree() and was currently running
211 * it would have return 0 so the code would
212 * not have deleted it since the callout could
213 * not be stopped so we want to go through
214 * with the delete here now. If the callout
215 * was restarted, the pending bit will be back on and
216 * we just want to bail since the callout_reset would
217 * return 1 and our reference would have been removed
218 * by arpresolve() below.
223 ifp = lle->lle_tbl->llt_ifp;
224 CURVNET_SET(ifp->if_vnet);
226 switch (lle->ln_state) {
227 case ARP_LLINFO_REACHABLE:
230 * Expiration time is approaching.
231 * Let's try to refresh entry if it is still
234 * Set r_skip_req to get feedback from
235 * fast path. Change state and re-schedule
241 lle->ln_state = ARP_LLINFO_VERIFY;
242 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
246 case ARP_LLINFO_VERIFY:
248 r_skip_req = lle->r_skip_req;
251 if (r_skip_req == 0 && lle->la_preempt > 0) {
252 /* Entry was used, issue refresh request */
254 dst = lle->r_l3addr.addr4;
256 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
258 arprequest(ifp, NULL, &dst, NULL);
262 /* Nothing happened. Reschedule if not too late */
263 if (lle->la_expire > time_uptime) {
264 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit);
270 case ARP_LLINFO_INCOMPLETE:
271 case ARP_LLINFO_DELETED:
275 if ((lle->la_flags & LLE_DELETED) == 0) {
278 if (lle->la_flags & LLE_VALID)
279 evt = LLENTRY_EXPIRED;
281 evt = LLENTRY_TIMEDOUT;
282 EVENTHANDLER_INVOKE(lle_event, lle, evt);
285 callout_stop(&lle->lle_timer);
287 /* XXX: LOR avoidance. We still have ref on lle. */
292 /* Guard against race with other llentry_free(). */
293 if (lle->la_flags & LLE_LINKED) {
295 lltable_unlink_entry(lle->lle_tbl, lle);
297 IF_AFDATA_UNLOCK(ifp);
299 size_t pkts_dropped = llentry_free(lle);
301 ARPSTAT_ADD(dropped, pkts_dropped);
302 ARPSTAT_INC(timeouts);
308 * Stores link-layer header for @ifp in format suitable for if_output()
309 * into buffer @buf. Resulting header length is stored in @bufsize.
311 * Returns 0 on success.
314 arp_fillheader(struct ifnet *ifp, struct arphdr *ah, int bcast, u_char *buf,
317 struct if_encap_req ereq;
320 bzero(buf, *bufsize);
321 bzero(&ereq, sizeof(ereq));
323 ereq.bufsize = *bufsize;
324 ereq.rtype = IFENCAP_LL;
325 ereq.family = AF_ARP;
326 ereq.lladdr = ar_tha(ah);
327 ereq.hdata = (u_char *)ah;
329 ereq.flags = IFENCAP_FLAG_BROADCAST;
330 error = ifp->if_requestencap(ifp, &ereq);
332 *bufsize = ereq.bufsize;
339 * Broadcast an ARP request. Caller specifies:
340 * - arp header source ip address
341 * - arp header target ip address
342 * - arp header source ethernet address
345 arprequest(struct ifnet *ifp, const struct in_addr *sip,
346 const struct in_addr *tip, u_char *enaddr)
351 u_char *carpaddr = NULL;
352 uint8_t linkhdr[LLE_MAX_LINKHDR];
359 * The caller did not supply a source address, try to find
360 * a compatible one among those assigned to this interface.
365 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
366 if (ifa->ifa_addr->sa_family != AF_INET)
370 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
372 sip = &IA_SIN(ifa)->sin_addr;
375 sip = &IA_SIN(ifa)->sin_addr;
378 if (0 == ((sip->s_addr ^ tip->s_addr) &
379 IA_MASKSIN(ifa)->sin_addr.s_addr))
380 break; /* found it. */
382 IF_ADDR_RUNLOCK(ifp);
384 printf("%s: cannot find matching address\n", __func__);
389 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
391 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
393 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) +
395 m->m_pkthdr.len = m->m_len;
396 M_ALIGN(m, m->m_len);
397 ah = mtod(m, struct arphdr *);
398 bzero((caddr_t)ah, m->m_len);
400 mac_netinet_arp_send(ifp, m);
402 ah->ar_pro = htons(ETHERTYPE_IP);
403 ah->ar_hln = ifp->if_addrlen; /* hardware address length */
404 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */
405 ah->ar_op = htons(ARPOP_REQUEST);
406 bcopy(enaddr, ar_sha(ah), ah->ar_hln);
407 bcopy(sip, ar_spa(ah), ah->ar_pln);
408 bcopy(tip, ar_tpa(ah), ah->ar_pln);
409 sa.sa_family = AF_ARP;
412 /* Calculate link header for sending frame */
413 bzero(&ro, sizeof(ro));
414 linkhdrsize = sizeof(linkhdr);
415 error = arp_fillheader(ifp, ah, 1, linkhdr, &linkhdrsize);
416 if (error != 0 && error != EAFNOSUPPORT) {
417 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
418 if_name(ifp), error);
422 ro.ro_prepend = linkhdr;
423 ro.ro_plen = linkhdrsize;
426 m->m_flags |= M_BCAST;
427 m_clrprotoflags(m); /* Avoid confusing lower layers. */
428 (*ifp->if_output)(ifp, m, &sa, &ro);
429 ARPSTAT_INC(txrequests);
434 * Resolve an IP address into an ethernet address - heavy version.
435 * Used internally by arpresolve().
436 * We have already checked than we can't use existing lle without
437 * modification so we have to acquire LLE_EXCLUSIVE lle lock.
439 * On success, desten and flags are filled in and the function returns 0;
440 * If the packet must be held pending resolution, we return EWOULDBLOCK
441 * On other errors, we return the corresponding error code.
442 * Note that m_freem() handles NULL.
445 arpresolve_full(struct ifnet *ifp, int is_gw, int flags, struct mbuf *m,
446 const struct sockaddr *dst, u_char *desten, uint32_t *pflags,
447 struct llentry **plle)
449 struct llentry *la = NULL, *la_tmp;
450 struct mbuf *curr = NULL;
451 struct mbuf *next = NULL;
461 if ((flags & LLE_CREATE) == 0) {
462 IF_AFDATA_RLOCK(ifp);
463 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
464 IF_AFDATA_RUNLOCK(ifp);
466 if (la == NULL && (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) {
467 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
469 char addrbuf[INET_ADDRSTRLEN];
472 "arpresolve: can't allocate llinfo for %s on %s\n",
473 inet_ntoa_r(SIN(dst)->sin_addr, addrbuf),
479 IF_AFDATA_WLOCK(ifp);
481 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
482 /* Prefer ANY existing lle over newly-created one */
484 lltable_link_entry(LLTABLE(ifp), la);
485 IF_AFDATA_WUNLOCK(ifp);
486 if (la_tmp != NULL) {
487 lltable_free_entry(LLTABLE(ifp), la);
496 if ((la->la_flags & LLE_VALID) &&
497 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
498 if (flags & LLE_ADDRONLY) {
499 lladdr = la->ll_addr;
500 ll_len = ifp->if_addrlen;
502 lladdr = la->r_linkdata;
503 ll_len = la->r_hdrlen;
505 bcopy(lladdr, desten, ll_len);
507 /* Notify LLE code that the entry was used by datapath */
508 llentry_mark_used(la);
510 *pflags = la->la_flags & (LLE_VALID|LLE_IFADDR);
519 renew = (la->la_asked == 0 || la->la_expire != time_uptime);
521 * There is an arptab entry, but no ethernet address
522 * response yet. Add the mbuf to the list, dropping
523 * the oldest packet if we have exceeded the system
527 if (la->la_numheld >= V_arp_maxhold) {
528 if (la->la_hold != NULL) {
529 next = la->la_hold->m_nextpkt;
530 m_freem(la->la_hold);
533 ARPSTAT_INC(dropped);
536 if (la->la_hold != NULL) {
538 while (curr->m_nextpkt != NULL)
539 curr = curr->m_nextpkt;
546 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
547 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
548 * if we have already sent arp_maxtries ARP requests. Retransmit the
549 * ARP request, but not faster than one request per second.
551 if (la->la_asked < V_arp_maxtries)
552 error = EWOULDBLOCK; /* First request. */
554 error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN;
560 la->la_expire = time_uptime;
561 canceled = callout_reset(&la->lle_timer, hz * V_arpt_down,
567 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
576 * Resolve an IP address into an ethernet address.
579 arpresolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
580 char *desten, uint32_t *pflags, struct llentry **plle)
584 flags |= LLE_ADDRONLY;
585 error = arpresolve_full(ifp, 0, flags, NULL, dst, desten, pflags, plle);
591 * Lookups link header based on an IP address.
593 * ifp is the interface we use
594 * is_gw != 0 if @dst represents gateway to some destination
595 * m is the mbuf. May be NULL if we don't have a packet.
596 * dst is the next hop,
597 * desten is the storage to put LL header.
598 * flags returns subset of lle flags: LLE_VALID | LLE_IFADDR
600 * On success, full/partial link header and flags are filled in and
601 * the function returns 0.
602 * If the packet must be held pending resolution, we return EWOULDBLOCK
603 * On other errors, we return the corresponding error code.
604 * Note that m_freem() handles NULL.
607 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
608 const struct sockaddr *dst, u_char *desten, uint32_t *pflags,
609 struct llentry **plle)
611 struct llentry *la = NULL;
619 if (m->m_flags & M_BCAST) {
622 ifp->if_broadcastaddr, ifp->if_addrlen);
625 if (m->m_flags & M_MCAST) {
627 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
632 IF_AFDATA_RLOCK(ifp);
633 la = lla_lookup(LLTABLE(ifp), plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, dst);
634 if (la != NULL && (la->r_flags & RLLE_VALID) != 0) {
635 /* Entry found, let's copy lle info */
636 bcopy(la->r_linkdata, desten, la->r_hdrlen);
638 *pflags = LLE_VALID | (la->r_flags & RLLE_IFADDR);
639 /* Notify the LLE handling code that the entry was used. */
640 llentry_mark_used(la);
646 IF_AFDATA_RUNLOCK(ifp);
651 IF_AFDATA_RUNLOCK(ifp);
653 return (arpresolve_full(ifp, is_gw, la == NULL ? LLE_CREATE : 0, m, dst,
654 desten, pflags, plle));
658 * Common length and type checks are done here,
659 * then the protocol-specific routine is called.
662 arpintr(struct mbuf *m)
669 ifp = m->m_pkthdr.rcvif;
671 if (m->m_len < sizeof(struct arphdr) &&
672 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
673 ARP_LOG(LOG_NOTICE, "packet with short header received on %s\n",
677 ar = mtod(m, struct arphdr *);
679 /* Check if length is sufficient */
680 if (m->m_len < arphdr_len(ar)) {
681 m = m_pullup(m, arphdr_len(ar));
683 ARP_LOG(LOG_NOTICE, "short packet received on %s\n",
687 ar = mtod(m, struct arphdr *);
692 switch (ntohs(ar->ar_hrd)) {
694 hlen = ETHER_ADDR_LEN; /* RFC 826 */
697 case ARPHRD_INFINIBAND:
698 hlen = 20; /* RFC 4391, INFINIBAND_ALEN */
699 layer = "infiniband";
701 case ARPHRD_IEEE1394:
702 hlen = 0; /* SHALL be 16 */ /* RFC 2734 */
706 * Restrict too long hardware addresses.
707 * Currently we are capable of handling 20-byte
708 * addresses ( sizeof(lle->ll_addr) )
710 if (ar->ar_hln >= 20)
715 "packet with unknown hardware format 0x%02d received on "
716 "%s\n", ntohs(ar->ar_hrd), if_name(ifp));
721 if (hlen != 0 && hlen != ar->ar_hln) {
723 "packet with invalid %s address length %d received on %s\n",
724 layer, ar->ar_hln, if_name(ifp));
729 ARPSTAT_INC(received);
730 switch (ntohs(ar->ar_pro)) {
742 * ARP for Internet protocols on 10 Mb/s Ethernet.
743 * Algorithm is that given in RFC 826.
744 * In addition, a sanity check is performed on the sender
745 * protocol address, to catch impersonators.
746 * We no longer handle negotiations for use of trailer protocol:
747 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
748 * along with IP replies if we wanted trailers sent to us,
749 * and also sent them in response to IP replies.
750 * This allowed either end to announce the desire to receive
752 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
753 * but formerly didn't normally send requests.
755 static int log_arp_wrong_iface = 1;
756 static int log_arp_movements = 1;
757 static int log_arp_permanent_modify = 1;
758 static int allow_multicast = 0;
760 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
761 &log_arp_wrong_iface, 0,
762 "log arp packets arriving on the wrong interface");
763 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
764 &log_arp_movements, 0,
765 "log arp replies from MACs different than the one in the cache");
766 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
767 &log_arp_permanent_modify, 0,
768 "log arp replies from MACs different than the one in the permanent arp entry");
769 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
770 &allow_multicast, 0, "accept multicast addresses");
773 in_arpinput(struct mbuf *m)
775 struct rm_priotracker in_ifa_tracker;
777 struct ifnet *ifp = m->m_pkthdr.rcvif;
778 struct llentry *la = NULL, *la_tmp;
780 struct in_ifaddr *ia;
782 struct in_addr isaddr, itaddr, myaddr;
783 u_int8_t *enaddr = NULL;
785 int bridged = 0, is_bridge = 0;
787 struct sockaddr_in sin;
788 struct sockaddr *dst;
789 struct nhop4_basic nh4;
790 uint8_t linkhdr[LLE_MAX_LINKHDR];
795 char addrbuf[INET_ADDRSTRLEN];
797 sin.sin_len = sizeof(struct sockaddr_in);
798 sin.sin_family = AF_INET;
799 sin.sin_addr.s_addr = 0;
803 if (ifp->if_type == IFT_BRIDGE)
807 * We already have checked that mbuf contains enough contiguous data
808 * to hold entire arp message according to the arp header.
810 ah = mtod(m, struct arphdr *);
813 * ARP is only for IPv4 so we can reject packets with
814 * a protocol length not equal to an IPv4 address.
816 if (ah->ar_pln != sizeof(struct in_addr)) {
817 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
818 sizeof(struct in_addr));
822 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
823 ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
824 ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
828 op = ntohs(ah->ar_op);
829 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
830 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
832 if (op == ARPOP_REPLY)
833 ARPSTAT_INC(rxreplies);
836 * For a bridge, we want to check the address irrespective
837 * of the receive interface. (This will change slightly
838 * when we have clusters of interfaces).
840 IN_IFADDR_RLOCK(&in_ifa_tracker);
841 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
842 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
843 ia->ia_ifp == ifp) &&
844 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
845 (ia->ia_ifa.ifa_carp == NULL ||
846 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
847 ifa_ref(&ia->ia_ifa);
848 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
852 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
853 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
854 ia->ia_ifp == ifp) &&
855 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
856 ifa_ref(&ia->ia_ifa);
857 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
861 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \
862 (ia->ia_ifp->if_bridge == ifp->if_softc && \
863 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \
864 addr == ia->ia_addr.sin_addr.s_addr)
866 * Check the case when bridge shares its MAC address with
867 * some of its children, so packets are claimed by bridge
868 * itself (bridge_input() does it first), but they are really
869 * meant to be destined to the bridge member.
872 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
873 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
874 ifa_ref(&ia->ia_ifa);
876 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
881 #undef BDG_MEMBER_MATCHES_ARP
882 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
885 * No match, use the first inet address on the receive interface
886 * as a dummy address for the rest of the function.
889 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
890 if (ifa->ifa_addr->sa_family == AF_INET &&
891 (ifa->ifa_carp == NULL ||
892 (*carp_iamatch_p)(ifa, &enaddr))) {
895 IF_ADDR_RUNLOCK(ifp);
898 IF_ADDR_RUNLOCK(ifp);
901 * If bridging, fall back to using any inet address.
903 IN_IFADDR_RLOCK(&in_ifa_tracker);
904 if (!bridged || (ia = CK_STAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
905 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
908 ifa_ref(&ia->ia_ifa);
909 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
912 enaddr = (u_int8_t *)IF_LLADDR(ifp);
913 carped = (ia->ia_ifa.ifa_carp != NULL);
914 myaddr = ia->ia_addr.sin_addr;
915 ifa_free(&ia->ia_ifa);
916 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
917 goto drop; /* it's from me, ignore it. */
918 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
919 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
920 "%s!\n", inet_ntoa_r(isaddr, addrbuf));
924 if (ifp->if_addrlen != ah->ar_hln) {
925 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
926 "i/f %d (ignored)\n", ifp->if_addrlen,
927 (u_char *) ar_sha(ah), ":", ah->ar_hln,
933 * Warn if another host is using the same IP address, but only if the
934 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
935 * case we suppress the warning to avoid false positive complaints of
936 * potential misconfiguration.
938 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
939 myaddr.s_addr != 0) {
940 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
941 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
942 inet_ntoa_r(isaddr, addrbuf), ifp->if_xname);
947 if (ifp->if_flags & IFF_STATICARP)
950 bzero(&sin, sizeof(sin));
951 sin.sin_len = sizeof(struct sockaddr_in);
952 sin.sin_family = AF_INET;
953 sin.sin_addr = isaddr;
954 dst = (struct sockaddr *)&sin;
955 IF_AFDATA_RLOCK(ifp);
956 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
957 IF_AFDATA_RUNLOCK(ifp);
959 arp_check_update_lle(ah, isaddr, ifp, bridged, la);
960 else if (itaddr.s_addr == myaddr.s_addr) {
962 * Request/reply to our address, but no lle exists yet.
963 * Calculate full link prepend to use in lle.
965 linkhdrsize = sizeof(linkhdr);
966 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
967 &linkhdrsize, &lladdr_off) != 0)
970 /* Allocate new entry */
971 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst);
975 * lle creation may fail if source address belongs
976 * to non-directly connected subnet. However, we
977 * will try to answer the request instead of dropping
982 lltable_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
985 IF_AFDATA_WLOCK(ifp);
987 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
990 * Check if lle still does not exists.
991 * If it does, that means that we either
992 * 1) have configured it explicitly, via
993 * 1a) 'arp -s' static entry or
994 * 1b) interface address static record
996 * 2) it was the result of sending first packet to-host
998 * 3) it was another arp reply packet we handled in
1001 * In all cases except 3) we definitely need to prefer
1002 * existing lle. For the sake of simplicity, prefer any
1003 * existing lle over newly-create one.
1006 lltable_link_entry(LLTABLE(ifp), la);
1007 IF_AFDATA_WUNLOCK(ifp);
1009 if (la_tmp == NULL) {
1010 arp_mark_lle_reachable(la);
1013 /* Free newly-create entry and handle packet */
1014 lltable_free_entry(LLTABLE(ifp), la);
1017 arp_check_update_lle(ah, isaddr, ifp, bridged, la);
1018 /* arp_check_update_lle() returns @la unlocked */
1023 if (op != ARPOP_REQUEST)
1025 ARPSTAT_INC(rxrequests);
1027 if (itaddr.s_addr == myaddr.s_addr) {
1028 /* Shortcut.. the receiving interface is the target. */
1029 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1030 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1032 struct llentry *lle = NULL;
1034 sin.sin_addr = itaddr;
1035 IF_AFDATA_RLOCK(ifp);
1036 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
1037 IF_AFDATA_RUNLOCK(ifp);
1039 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
1040 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1041 (void)memcpy(ar_sha(ah), lle->ll_addr, ah->ar_hln);
1048 if (!V_arp_proxyall)
1051 /* XXX MRT use table 0 for arp reply */
1052 if (fib4_lookup_nh_basic(0, itaddr, 0, 0, &nh4) != 0)
1056 * Don't send proxies for nodes on the same interface
1057 * as this one came out of, or we'll get into a fight
1058 * over who claims what Ether address.
1060 if (nh4.nh_ifp == ifp)
1063 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
1064 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
1067 * Also check that the node which sent the ARP packet
1068 * is on the interface we expect it to be on. This
1069 * avoids ARP chaos if an interface is connected to the
1073 /* XXX MRT use table 0 for arp checks */
1074 if (fib4_lookup_nh_basic(0, isaddr, 0, 0, &nh4) != 0)
1076 if (nh4.nh_ifp != ifp) {
1077 ARP_LOG(LOG_INFO, "proxy: ignoring request"
1078 " from %s via %s\n",
1079 inet_ntoa_r(isaddr, addrbuf),
1085 printf("arp: proxying for %s\n",
1086 inet_ntoa_r(itaddr, addrbuf));
1091 if (itaddr.s_addr == myaddr.s_addr &&
1092 IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
1093 /* RFC 3927 link-local IPv4; always reply by broadcast. */
1094 #ifdef DEBUG_LINKLOCAL
1095 printf("arp: sending reply for link-local addr %s\n",
1096 inet_ntoa_r(itaddr, addrbuf));
1098 m->m_flags |= M_BCAST;
1099 m->m_flags &= ~M_MCAST;
1101 /* default behaviour; never reply by broadcast. */
1102 m->m_flags &= ~(M_BCAST|M_MCAST);
1104 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
1105 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
1106 ah->ar_op = htons(ARPOP_REPLY);
1107 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
1108 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
1109 m->m_pkthdr.len = m->m_len;
1110 m->m_pkthdr.rcvif = NULL;
1111 sa.sa_family = AF_ARP;
1114 /* Calculate link header for sending frame */
1115 bzero(&ro, sizeof(ro));
1116 linkhdrsize = sizeof(linkhdr);
1117 error = arp_fillheader(ifp, ah, 0, linkhdr, &linkhdrsize);
1120 * arp_fillheader() may fail due to lack of support inside encap request
1121 * routing. This is not necessary an error, AF_ARP can/should be handled
1124 if (error != 0 && error != EAFNOSUPPORT) {
1125 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n",
1126 if_name(ifp), error);
1130 ro.ro_prepend = linkhdr;
1131 ro.ro_plen = linkhdrsize;
1134 m_clrprotoflags(m); /* Avoid confusing lower layers. */
1135 (*ifp->if_output)(ifp, m, &sa, &ro);
1136 ARPSTAT_INC(txreplies);
1145 * Checks received arp data against existing @la.
1146 * Updates lle state/performs notification if necessary.
1149 arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, struct ifnet *ifp,
1150 int bridged, struct llentry *la)
1153 struct mbuf *m_hold, *m_hold_next;
1154 uint8_t linkhdr[LLE_MAX_LINKHDR];
1157 char addrbuf[INET_ADDRSTRLEN];
1159 LLE_WLOCK_ASSERT(la);
1161 /* the following is not an error when doing bridging */
1162 if (!bridged && la->lle_tbl->llt_ifp != ifp) {
1163 if (log_arp_wrong_iface)
1164 ARP_LOG(LOG_WARNING, "%s is on %s "
1165 "but got reply from %*D on %s\n",
1166 inet_ntoa_r(isaddr, addrbuf),
1167 la->lle_tbl->llt_ifp->if_xname,
1168 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1173 if ((la->la_flags & LLE_VALID) &&
1174 bcmp(ar_sha(ah), la->ll_addr, ifp->if_addrlen)) {
1175 if (la->la_flags & LLE_STATIC) {
1177 if (log_arp_permanent_modify)
1179 "%*D attempts to modify "
1180 "permanent entry for %s on %s\n",
1182 (u_char *)ar_sha(ah), ":",
1183 inet_ntoa_r(isaddr, addrbuf),
1187 if (log_arp_movements) {
1188 ARP_LOG(LOG_INFO, "%s moved from %*D "
1190 inet_ntoa_r(isaddr, addrbuf),
1192 (u_char *)la->ll_addr, ":",
1193 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
1198 /* Calculate full link prepend to use in lle */
1199 linkhdrsize = sizeof(linkhdr);
1200 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr,
1201 &linkhdrsize, &lladdr_off) != 0)
1204 /* Check if something has changed */
1205 if (memcmp(la->r_linkdata, linkhdr, linkhdrsize) != 0 ||
1206 (la->la_flags & LLE_VALID) == 0) {
1207 /* Try to perform LLE update */
1208 if (lltable_try_set_entry_addr(ifp, la, linkhdr, linkhdrsize,
1212 /* Clear fast path feedback request if set */
1216 arp_mark_lle_reachable(la);
1219 * The packets are all freed within the call to the output
1222 * NB: The lock MUST be released before the call to the
1225 if (la->la_hold != NULL) {
1226 m_hold = la->la_hold;
1229 lltable_fill_sa_entry(la, &sa);
1231 for (; m_hold != NULL; m_hold = m_hold_next) {
1232 m_hold_next = m_hold->m_nextpkt;
1233 m_hold->m_nextpkt = NULL;
1234 /* Avoid confusing lower layers. */
1235 m_clrprotoflags(m_hold);
1236 (*ifp->if_output)(ifp, m_hold, &sa, NULL);
1243 arp_mark_lle_reachable(struct llentry *la)
1245 int canceled, wtime;
1247 LLE_WLOCK_ASSERT(la);
1249 la->ln_state = ARP_LLINFO_REACHABLE;
1250 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
1252 if (!(la->la_flags & LLE_STATIC)) {
1254 la->la_expire = time_uptime + V_arpt_keep;
1255 wtime = V_arpt_keep - V_arp_maxtries * V_arpt_rexmit;
1257 wtime = V_arpt_keep;
1258 canceled = callout_reset(&la->lle_timer,
1259 hz * wtime, arptimer, la);
1264 la->la_preempt = V_arp_maxtries;
1268 * Add permanent link-layer record for given interface address.
1270 static __noinline void
1271 arp_add_ifa_lle(struct ifnet *ifp, const struct sockaddr *dst)
1273 struct llentry *lle, *lle_tmp;
1276 * Interface address LLE record is considered static
1277 * because kernel code relies on LLE_STATIC flag to check
1278 * if these entries can be rewriten by arp updates.
1280 lle = lltable_alloc_entry(LLTABLE(ifp), LLE_IFADDR | LLE_STATIC, dst);
1282 log(LOG_INFO, "arp_ifinit: cannot create arp "
1283 "entry for interface address\n");
1287 IF_AFDATA_WLOCK(ifp);
1289 /* Unlink any entry if exists */
1290 lle_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst);
1291 if (lle_tmp != NULL)
1292 lltable_unlink_entry(LLTABLE(ifp), lle_tmp);
1294 lltable_link_entry(LLTABLE(ifp), lle);
1295 IF_AFDATA_WUNLOCK(ifp);
1297 if (lle_tmp != NULL)
1298 EVENTHANDLER_INVOKE(lle_event, lle_tmp, LLENTRY_EXPIRED);
1300 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED);
1302 if (lle_tmp != NULL)
1303 lltable_free_entry(LLTABLE(ifp), lle_tmp);
1307 * Handle the garp_rexmit_count. Like sysctl_handle_int(), but limits the range
1311 sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS)
1314 int rexmit_count = *(int *)arg1;
1316 error = sysctl_handle_int(oidp, &rexmit_count, 0, req);
1318 /* Enforce limits on any new value that may have been set. */
1319 if (!error && req->newptr) {
1320 /* A new value was set. */
1321 if (rexmit_count < 0) {
1323 } else if (rexmit_count > MAX_GARP_RETRANSMITS) {
1324 rexmit_count = MAX_GARP_RETRANSMITS;
1326 *(int *)arg1 = rexmit_count;
1333 * Retransmit a Gratuitous ARP (GARP) and, if necessary, schedule a callout to
1334 * retransmit it again. A pending callout owns a reference to the ifa.
1337 garp_rexmit(void *arg)
1339 struct in_ifaddr *ia = arg;
1341 if (callout_pending(&ia->ia_garp_timer) ||
1342 !callout_active(&ia->ia_garp_timer)) {
1343 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp);
1344 ifa_free(&ia->ia_ifa);
1349 * Drop lock while the ARP request is generated.
1351 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp);
1353 arprequest(ia->ia_ifa.ifa_ifp, &IA_SIN(ia)->sin_addr,
1354 &IA_SIN(ia)->sin_addr, IF_LLADDR(ia->ia_ifa.ifa_ifp));
1357 * Increment the count of retransmissions. If the count has reached the
1358 * maximum value, stop sending the GARP packets. Otherwise, schedule
1359 * the callout to retransmit another GARP packet.
1361 ++ia->ia_garp_count;
1362 if (ia->ia_garp_count >= garp_rexmit_count) {
1363 ifa_free(&ia->ia_ifa);
1366 IF_ADDR_WLOCK(ia->ia_ifa.ifa_ifp);
1367 rescheduled = callout_reset(&ia->ia_garp_timer,
1368 (1 << ia->ia_garp_count) * hz,
1370 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp);
1372 ifa_free(&ia->ia_ifa);
1378 * Start the GARP retransmit timer.
1380 * A single GARP is always transmitted when an IPv4 address is added
1381 * to an interface and that is usually sufficient. However, in some
1382 * circumstances, such as when a shared address is passed between
1383 * cluster nodes, this single GARP may occasionally be dropped or
1384 * lost. This can lead to neighbors on the network link working with a
1385 * stale ARP cache and sending packets destined for that address to
1386 * the node that previously owned the address, which may not respond.
1388 * To avoid this situation, GARP retransmits can be enabled by setting
1389 * the net.link.ether.inet.garp_rexmit_count sysctl to a value greater
1390 * than zero. The setting represents the maximum number of
1391 * retransmissions. The interval between retransmissions is calculated
1392 * using an exponential backoff algorithm, doubling each time, so the
1393 * retransmission intervals are: {1, 2, 4, 8, 16, ...} (seconds).
1396 garp_timer_start(struct ifaddr *ifa)
1398 struct in_ifaddr *ia = (struct in_ifaddr *) ifa;
1400 IF_ADDR_WLOCK(ia->ia_ifa.ifa_ifp);
1401 ia->ia_garp_count = 0;
1402 if (callout_reset(&ia->ia_garp_timer, (1 << ia->ia_garp_count) * hz,
1403 garp_rexmit, ia) == 0) {
1406 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp);
1410 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1412 const struct sockaddr_in *dst_in;
1413 const struct sockaddr *dst;
1415 if (ifa->ifa_carp != NULL)
1418 dst = ifa->ifa_addr;
1419 dst_in = (const struct sockaddr_in *)dst;
1421 if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY)
1423 arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp));
1424 if (garp_rexmit_count > 0) {
1425 garp_timer_start(ifa);
1428 arp_add_ifa_lle(ifp, dst);
1432 arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr)
1435 if (ntohl(addr.s_addr) != INADDR_ANY)
1436 arprequest(ifp, &addr, &addr, enaddr);
1440 * Sends gratuitous ARPs for each ifaddr to notify other
1441 * nodes about the address change.
1443 static __noinline void
1444 arp_handle_ifllchange(struct ifnet *ifp)
1448 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1449 if (ifa->ifa_addr->sa_family == AF_INET)
1450 arp_ifinit(ifp, ifa);
1455 * A handler for interface link layer address change event.
1458 arp_iflladdr(void *arg __unused, struct ifnet *ifp)
1461 lltable_update_ifaddr(LLTABLE(ifp));
1463 if ((ifp->if_flags & IFF_UP) != 0)
1464 arp_handle_ifllchange(ifp);
1471 if (IS_DEFAULT_VNET(curvnet)) {
1472 netisr_register(&arp_nh);
1473 iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event,
1474 arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
1478 netisr_register_vnet(&arp_nh);
1481 VNET_SYSINIT(vnet_arp_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
1486 * We have to unregister ARP along with IP otherwise we risk doing INADDR_HASH
1487 * lookups after destroying the hash. Ideally this would go on SI_ORDER_3.5.
1490 vnet_arp_destroy(__unused void *arg)
1493 netisr_unregister_vnet(&arp_nh);
1495 VNET_SYSUNINIT(vnet_arp_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD,
1496 vnet_arp_destroy, NULL);