1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93
30 */
31
32 /*
33 * Ethernet address resolution protocol.
34 * TODO:
35 * add "inuse/lock" bit (or ref. count) along with valid bit
36 */
37
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40
41 #include "opt_inet.h"
42
43 #include <sys/param.h>
44 #include <sys/kernel.h>
45 #include <sys/queue.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/malloc.h>
50 #include <sys/proc.h>
51 #include <sys/socket.h>
52 #include <sys/syslog.h>
53
54 #include <net/if.h>
55 #include <net/if_dl.h>
56 #include <net/if_types.h>
57 #include <net/netisr.h>
58 #include <net/if_llc.h>
59 #include <net/ethernet.h>
60 #include <net/route.h>
61 #include <net/vnet.h>
62
63 #include <netinet/in.h>
64 #include <netinet/in_var.h>
65 #include <net/if_llatbl.h>
66 #include <netinet/if_ether.h>
67 #ifdef INET
68 #include <netinet/ip_carp.h>
69 #endif
70
71 #include <net/if_arc.h>
72 #include <net/iso88025.h>
73
74 #include <security/mac/mac_framework.h>
75
76 #define SIN(s) ((const struct sockaddr_in *)(s))
77 #define SDL(s) ((struct sockaddr_dl *)s)
78
79 SYSCTL_DECL(_net_link_ether);
80 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, "");
81 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, "");
82
83 /* timer values */
84 static VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20
85 * minutes */
86 static VNET_DEFINE(int, arp_maxtries) = 5;
87 VNET_DEFINE(int, useloopback) = 1; /* use loopback interface for
88 * local traffic */
89 static VNET_DEFINE(int, arp_proxyall) = 0;
90 static VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for
91 * 20 seconds */
92 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */
93 VNET_PCPUSTAT_SYSINIT(arpstat);
94
95 #ifdef VIMAGE
96 VNET_PCPUSTAT_SYSUNINIT(arpstat);
97 #endif /* VIMAGE */
98
99 static VNET_DEFINE(int, arp_maxhold) = 1;
100
101 #define V_arpt_keep VNET(arpt_keep)
102 #define V_arpt_down VNET(arpt_down)
103 #define V_arp_maxtries VNET(arp_maxtries)
104 #define V_arp_proxyall VNET(arp_proxyall)
105 #define V_arp_maxhold VNET(arp_maxhold)
106
107 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW,
108 &VNET_NAME(arpt_keep), 0,
109 "ARP entry lifetime in seconds");
110 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW,
111 &VNET_NAME(arp_maxtries), 0,
112 "ARP resolution attempts before returning error");
113 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW,
114 &VNET_NAME(useloopback), 0,
115 "Use the loopback interface for local traffic");
116 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW,
117 &VNET_NAME(arp_proxyall), 0,
118 "Enable proxy ARP for all suitable requests");
119 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW,
120 &VNET_NAME(arpt_down), 0,
121 "Incomplete ARP entry lifetime in seconds");
122 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat,
123 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)");
124 SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW,
125 &VNET_NAME(arp_maxhold), 0,
126 "Number of packets to hold per ARP entry");
127
128 /*
129 * Due to the exponential backoff algorithm used for the interval between GARP
130 * retransmissions, the maximum number of retransmissions is limited for
131 * sanity. This limit corresponds to a maximum interval between retransmissions
132 * of 2^16 seconds ~= 18 hours.
133 *
134 * Making this limit more dynamic is more complicated than worthwhile,
135 * especially since sending out GARPs spaced days apart would be of little
136 * use. A maximum dynamic limit would look something like:
137 *
138 * const int max = fls(INT_MAX / hz) - 1;
139 */
140 #define MAX_GARP_RETRANSMITS 16
141 static int sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS);
142 static int garp_rexmit_count = 0; /* GARP retransmission setting. */
143
144 SYSCTL_PROC(_net_link_ether_inet, OID_AUTO, garp_rexmit_count,
145 CTLTYPE_INT|CTLFLAG_RW|CTLFLAG_MPSAFE,
146 &garp_rexmit_count, 0, sysctl_garp_rexmit, "I",
147 "Number of times to retransmit GARP packets;"
148 " 0 to disable, maximum of 16");
149
150 static void arp_init(void);
151 static void arpintr(struct mbuf *);
152 static void arptimer(void *);
153 #ifdef INET
154 static void in_arpinput(struct mbuf *);
155 #endif
156
157 static const struct netisr_handler arp_nh = {
158 .nh_name = "arp",
159 .nh_handler = arpintr,
160 .nh_proto = NETISR_ARP,
161 .nh_policy = NETISR_POLICY_SOURCE,
162 };
163
164 #ifdef AF_INET
165 /*
166 * called by in_ifscrub to remove entry from the table when
167 * the interface goes away
168 */
169 void
170 arp_ifscrub(struct ifnet *ifp, uint32_t addr)
171 {
172 struct sockaddr_in addr4;
173
174 bzero((void *)&addr4, sizeof(addr4));
175 addr4.sin_len = sizeof(addr4);
176 addr4.sin_family = AF_INET;
177 addr4.sin_addr.s_addr = addr;
178 IF_AFDATA_WLOCK(ifp);
179 lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR),
180 (struct sockaddr *)&addr4);
181 IF_AFDATA_WUNLOCK(ifp);
182 }
183 #endif
184
185 /*
186 * Timeout routine. Age arp_tab entries periodically.
187 */
188 static void
189 arptimer(void *arg)
190 {
191 struct llentry *lle = (struct llentry *)arg;
192 struct ifnet *ifp;
193
194 if (lle->la_flags & LLE_STATIC) {
195 return;
196 }
197 LLE_WLOCK(lle);
198 if (callout_pending(&lle->la_timer)) {
199 /*
200 * Here we are a bit odd here in the treatment of
201 * active/pending. If the pending bit is set, it got
202 * rescheduled before I ran. The active
203 * bit we ignore, since if it was stopped
204 * in ll_tablefree() and was currently running
205 * it would have return 0 so the code would
206 * not have deleted it since the callout could
207 * not be stopped so we want to go through
208 * with the delete here now. If the callout
209 * was restarted, the pending bit will be back on and
210 * we just want to bail since the callout_reset would
211 * return 1 and our reference would have been removed
212 * by arpresolve() below.
213 */
214 LLE_WUNLOCK(lle);
215 return;
216 }
217 ifp = lle->lle_tbl->llt_ifp;
218 CURVNET_SET(ifp->if_vnet);
219
220 if ((lle->la_flags & LLE_DELETED) == 0) {
221 int evt;
222
223 if (lle->la_flags & LLE_VALID)
224 evt = LLENTRY_EXPIRED;
225 else
226 evt = LLENTRY_TIMEDOUT;
227 EVENTHANDLER_INVOKE(lle_event, lle, evt);
228 }
229
230 callout_stop(&lle->la_timer);
231
232 /* XXX: LOR avoidance. We still have ref on lle. */
233 LLE_WUNLOCK(lle);
234 IF_AFDATA_LOCK(ifp);
235 LLE_WLOCK(lle);
236
237 /* Guard against race with other llentry_free(). */
238 if (lle->la_flags & LLE_LINKED) {
239 size_t pkts_dropped;
240
241 LLE_REMREF(lle);
242 pkts_dropped = llentry_free(lle);
243 ARPSTAT_ADD(dropped, pkts_dropped);
244 } else
245 LLE_FREE_LOCKED(lle);
246
247 IF_AFDATA_UNLOCK(ifp);
248
249 ARPSTAT_INC(timeouts);
250
251 CURVNET_RESTORE();
252 }
253
254 /*
255 * Broadcast an ARP request. Caller specifies:
256 * - arp header source ip address
257 * - arp header target ip address
258 * - arp header source ethernet address
259 */
260 void
261 arprequest(struct ifnet *ifp, const struct in_addr *sip,
262 const struct in_addr *tip, u_char *enaddr)
263 {
264 struct mbuf *m;
265 struct arphdr *ah;
266 struct sockaddr sa;
267 u_char *carpaddr = NULL;
268
269 if (sip == NULL) {
270 /*
271 * The caller did not supply a source address, try to find
272 * a compatible one among those assigned to this interface.
273 */
274 struct ifaddr *ifa;
275
276 IF_ADDR_RLOCK(ifp);
277 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
278 if (ifa->ifa_addr->sa_family != AF_INET)
279 continue;
280
281 if (ifa->ifa_carp) {
282 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0)
283 continue;
284 sip = &IA_SIN(ifa)->sin_addr;
285 } else {
286 carpaddr = NULL;
287 sip = &IA_SIN(ifa)->sin_addr;
288 }
289
290 if (0 == ((sip->s_addr ^ tip->s_addr) &
291 IA_MASKSIN(ifa)->sin_addr.s_addr))
292 break; /* found it. */
293 }
294 IF_ADDR_RUNLOCK(ifp);
295 if (sip == NULL) {
296 printf("%s: cannot find matching address\n", __func__);
297 return;
298 }
299 }
300 if (enaddr == NULL)
301 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp);
302
303 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL)
304 return;
305 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) +
306 2*ifp->if_data.ifi_addrlen;
307 m->m_pkthdr.len = m->m_len;
308 MH_ALIGN(m, m->m_len);
309 ah = mtod(m, struct arphdr *);
310 bzero((caddr_t)ah, m->m_len);
311 #ifdef MAC
312 mac_netinet_arp_send(ifp, m);
313 #endif
314 ah->ar_pro = htons(ETHERTYPE_IP);
315 ah->ar_hln = ifp->if_addrlen; /* hardware address length */
316 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */
317 ah->ar_op = htons(ARPOP_REQUEST);
318 bcopy(enaddr, ar_sha(ah), ah->ar_hln);
319 bcopy(sip, ar_spa(ah), ah->ar_pln);
320 bcopy(tip, ar_tpa(ah), ah->ar_pln);
321 sa.sa_family = AF_ARP;
322 sa.sa_len = 2;
323 m->m_flags |= M_BCAST;
324 m_clrprotoflags(m); /* Avoid confusing lower layers. */
325 (*ifp->if_output)(ifp, m, &sa, NULL);
326 ARPSTAT_INC(txrequests);
327 }
328
329 /*
330 * Resolve an IP address into an ethernet address.
331 * On input:
332 * ifp is the interface we use
333 * rt0 is the route to the final destination (possibly useless)
334 * m is the mbuf. May be NULL if we don't have a packet.
335 * dst is the next hop,
336 * desten is where we want the address.
337 *
338 * On success, desten is filled in and the function returns 0;
339 * If the packet must be held pending resolution, we return EWOULDBLOCK
340 * On other errors, we return the corresponding error code.
341 * Note that m_freem() handles NULL.
342 */
343 int
344 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m,
345 const struct sockaddr *dst, u_char *desten, struct llentry **lle)
346 {
347 struct llentry *la = 0;
348 u_int flags = 0;
349 struct mbuf *curr = NULL;
350 struct mbuf *next = NULL;
351 int error, renew;
352
353 *lle = NULL;
354 if (m != NULL) {
355 if (m->m_flags & M_BCAST) {
356 /* broadcast */
357 (void)memcpy(desten,
358 ifp->if_broadcastaddr, ifp->if_addrlen);
359 return (0);
360 }
361 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {
362 /* multicast */
363 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten);
364 return (0);
365 }
366 }
367 retry:
368 IF_AFDATA_RLOCK(ifp);
369 la = lla_lookup(LLTABLE(ifp), flags, dst);
370 IF_AFDATA_RUNLOCK(ifp);
371 if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0)
372 && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) {
373 flags |= (LLE_CREATE | LLE_EXCLUSIVE);
374 IF_AFDATA_WLOCK(ifp);
375 la = lla_lookup(LLTABLE(ifp), flags, dst);
376 IF_AFDATA_WUNLOCK(ifp);
377 }
378 if (la == NULL) {
379 if (flags & LLE_CREATE)
380 log(LOG_DEBUG,
381 "arpresolve: can't allocate llinfo for %s on %s\n",
382 inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname);
383 m_freem(m);
384 return (EINVAL);
385 }
386
387 if ((la->la_flags & LLE_VALID) &&
388 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) {
389 bcopy(&la->ll_addr, desten, ifp->if_addrlen);
390 renew = 0;
391 /*
392 * If entry has an expiry time and it is approaching,
393 * see if we need to send an ARP request within this
394 * arpt_down interval.
395 */
396 if (!(la->la_flags & LLE_STATIC) &&
397 time_uptime + la->la_preempt > la->la_expire) {
398 renew = 1;
399 la->la_preempt--;
400 }
401
402 *lle = la;
403
404 if (flags & LLE_EXCLUSIVE)
405 LLE_WUNLOCK(la);
406 else
407 LLE_RUNLOCK(la);
408
409 if (renew == 1)
410 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
411
412 return (0);
413 }
414
415 if (la->la_flags & LLE_STATIC) { /* should not happen! */
416 log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n",
417 inet_ntoa(SIN(dst)->sin_addr));
418 m_freem(m);
419 error = EINVAL;
420 goto done;
421 }
422
423 renew = (la->la_asked == 0 || la->la_expire != time_uptime);
424 if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) {
425 flags |= LLE_EXCLUSIVE;
426 LLE_RUNLOCK(la);
427 goto retry;
428 }
429 /*
430 * There is an arptab entry, but no ethernet address
431 * response yet. Add the mbuf to the list, dropping
432 * the oldest packet if we have exceeded the system
433 * setting.
434 */
435 if (m != NULL) {
436 if (la->la_numheld >= V_arp_maxhold) {
437 if (la->la_hold != NULL) {
438 next = la->la_hold->m_nextpkt;
439 m_freem(la->la_hold);
440 la->la_hold = next;
441 la->la_numheld--;
442 ARPSTAT_INC(dropped);
443 }
444 }
445 if (la->la_hold != NULL) {
446 curr = la->la_hold;
447 while (curr->m_nextpkt != NULL)
448 curr = curr->m_nextpkt;
449 curr->m_nextpkt = m;
450 } else
451 la->la_hold = m;
452 la->la_numheld++;
453 if (renew == 0 && (flags & LLE_EXCLUSIVE)) {
454 flags &= ~LLE_EXCLUSIVE;
455 LLE_DOWNGRADE(la);
456 }
457
458 }
459 /*
460 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It
461 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH
462 * if we have already sent arp_maxtries ARP requests. Retransmit the
463 * ARP request, but not faster than one request per second.
464 */
465 if (la->la_asked < V_arp_maxtries)
466 error = EWOULDBLOCK; /* First request. */
467 else
468 error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ?
469 EHOSTUNREACH : EHOSTDOWN;
470
471 if (renew) {
472 int canceled;
473
474 LLE_ADDREF(la);
475 la->la_expire = time_uptime;
476 canceled = callout_reset(&la->la_timer, hz * V_arpt_down,
477 arptimer, la);
478 if (canceled)
479 LLE_REMREF(la);
480 la->la_asked++;
481 LLE_WUNLOCK(la);
482 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL);
483 return (error);
484 }
485 done:
486 if (flags & LLE_EXCLUSIVE)
487 LLE_WUNLOCK(la);
488 else
489 LLE_RUNLOCK(la);
490 return (error);
491 }
492
493 /*
494 * Common length and type checks are done here,
495 * then the protocol-specific routine is called.
496 */
497 static void
498 arpintr(struct mbuf *m)
499 {
500 struct arphdr *ar;
501
502 if (m->m_len < sizeof(struct arphdr) &&
503 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) {
504 log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n");
505 return;
506 }
507 ar = mtod(m, struct arphdr *);
508
509 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER &&
510 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 &&
511 ntohs(ar->ar_hrd) != ARPHRD_ARCNET &&
512 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 &&
513 ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) {
514 log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)"
515 " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "",
516 ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":",
517 ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":");
518 m_freem(m);
519 return;
520 }
521
522 if (m->m_len < arphdr_len(ar)) {
523 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) {
524 log(LOG_NOTICE, "arp: runt packet\n");
525 m_freem(m);
526 return;
527 }
528 ar = mtod(m, struct arphdr *);
529 }
530
531 ARPSTAT_INC(received);
532 switch (ntohs(ar->ar_pro)) {
533 #ifdef INET
534 case ETHERTYPE_IP:
535 in_arpinput(m);
536 return;
537 #endif
538 }
539 m_freem(m);
540 }
541
542 #ifdef INET
543 /*
544 * ARP for Internet protocols on 10 Mb/s Ethernet.
545 * Algorithm is that given in RFC 826.
546 * In addition, a sanity check is performed on the sender
547 * protocol address, to catch impersonators.
548 * We no longer handle negotiations for use of trailer protocol:
549 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent
550 * along with IP replies if we wanted trailers sent to us,
551 * and also sent them in response to IP replies.
552 * This allowed either end to announce the desire to receive
553 * trailer packets.
554 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either,
555 * but formerly didn't normally send requests.
556 */
557 static int log_arp_wrong_iface = 1;
558 static int log_arp_movements = 1;
559 static int log_arp_permanent_modify = 1;
560 static int allow_multicast = 0;
561 static struct timeval arp_lastlog;
562 static int arp_curpps;
563 static int arp_maxpps = 1;
564
565 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW,
566 &log_arp_wrong_iface, 0,
567 "log arp packets arriving on the wrong interface");
568 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW,
569 &log_arp_movements, 0,
570 "log arp replies from MACs different than the one in the cache");
571 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW,
572 &log_arp_permanent_modify, 0,
573 "log arp replies from MACs different than the one in the permanent arp entry");
574 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW,
575 &allow_multicast, 0, "accept multicast addresses");
576 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second,
577 CTLFLAG_RW, &arp_maxpps, 0,
578 "Maximum number of remotely triggered ARP messages that can be "
579 "logged per second");
580
581 #define ARP_LOG(pri, ...) do { \
582 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \
583 log((pri), "arp: " __VA_ARGS__); \
584 } while (0)
585
586 static void
587 in_arpinput(struct mbuf *m)
588 {
589 struct arphdr *ah;
590 struct ifnet *ifp = m->m_pkthdr.rcvif;
591 struct llentry *la = NULL;
592 struct rtentry *rt;
593 struct ifaddr *ifa;
594 struct in_ifaddr *ia;
595 struct sockaddr sa;
596 struct in_addr isaddr, itaddr, myaddr;
597 u_int8_t *enaddr = NULL;
598 int op, flags;
599 int req_len;
600 int bridged = 0, is_bridge = 0;
601 int carped;
602 struct sockaddr_in sin;
603 sin.sin_len = sizeof(struct sockaddr_in);
604 sin.sin_family = AF_INET;
605 sin.sin_addr.s_addr = 0;
606
607 if (ifp->if_bridge)
608 bridged = 1;
609 if (ifp->if_type == IFT_BRIDGE)
610 is_bridge = 1;
611
612 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr));
613 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) {
614 ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n");
615 return;
616 }
617
618 ah = mtod(m, struct arphdr *);
619 /*
620 * ARP is only for IPv4 so we can reject packets with
621 * a protocol length not equal to an IPv4 address.
622 */
623 if (ah->ar_pln != sizeof(struct in_addr)) {
624 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n",
625 sizeof(struct in_addr));
626 goto drop;
627 }
628
629 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) {
630 ARP_LOG(LOG_NOTICE, "%*D is multicast\n",
631 ifp->if_addrlen, (u_char *)ar_sha(ah), ":");
632 goto drop;
633 }
634
635 op = ntohs(ah->ar_op);
636 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr));
637 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr));
638
639 if (op == ARPOP_REPLY)
640 ARPSTAT_INC(rxreplies);
641
642 /*
643 * For a bridge, we want to check the address irrespective
644 * of the receive interface. (This will change slightly
645 * when we have clusters of interfaces).
646 */
647 IN_IFADDR_RLOCK();
648 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
649 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
650 ia->ia_ifp == ifp) &&
651 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr &&
652 (ia->ia_ifa.ifa_carp == NULL ||
653 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) {
654 ifa_ref(&ia->ia_ifa);
655 IN_IFADDR_RUNLOCK();
656 goto match;
657 }
658 }
659 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash)
660 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) ||
661 ia->ia_ifp == ifp) &&
662 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) {
663 ifa_ref(&ia->ia_ifa);
664 IN_IFADDR_RUNLOCK();
665 goto match;
666 }
667
668 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \
669 (ia->ia_ifp->if_bridge == ifp->if_softc && \
670 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \
671 addr == ia->ia_addr.sin_addr.s_addr)
672 /*
673 * Check the case when bridge shares its MAC address with
674 * some of its children, so packets are claimed by bridge
675 * itself (bridge_input() does it first), but they are really
676 * meant to be destined to the bridge member.
677 */
678 if (is_bridge) {
679 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) {
680 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) {
681 ifa_ref(&ia->ia_ifa);
682 ifp = ia->ia_ifp;
683 IN_IFADDR_RUNLOCK();
684 goto match;
685 }
686 }
687 }
688 #undef BDG_MEMBER_MATCHES_ARP
689 IN_IFADDR_RUNLOCK();
690
691 /*
692 * No match, use the first inet address on the receive interface
693 * as a dummy address for the rest of the function.
694 */
695 IF_ADDR_RLOCK(ifp);
696 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
697 if (ifa->ifa_addr->sa_family == AF_INET &&
698 (ifa->ifa_carp == NULL ||
699 (*carp_iamatch_p)(ifa, &enaddr))) {
700 ia = ifatoia(ifa);
701 ifa_ref(ifa);
702 IF_ADDR_RUNLOCK(ifp);
703 goto match;
704 }
705 IF_ADDR_RUNLOCK(ifp);
706
707 /*
708 * If bridging, fall back to using any inet address.
709 */
710 IN_IFADDR_RLOCK();
711 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) {
712 IN_IFADDR_RUNLOCK();
713 goto drop;
714 }
715 ifa_ref(&ia->ia_ifa);
716 IN_IFADDR_RUNLOCK();
717 match:
718 if (!enaddr)
719 enaddr = (u_int8_t *)IF_LLADDR(ifp);
720 carped = (ia->ia_ifa.ifa_carp != NULL);
721 myaddr = ia->ia_addr.sin_addr;
722 ifa_free(&ia->ia_ifa);
723 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen))
724 goto drop; /* it's from me, ignore it. */
725 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) {
726 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address "
727 "%s!\n", inet_ntoa(isaddr));
728 goto drop;
729 }
730 /*
731 * Warn if another host is using the same IP address, but only if the
732 * IP address isn't 0.0.0.0, which is used for DHCP only, in which
733 * case we suppress the warning to avoid false positive complaints of
734 * potential misconfiguration.
735 */
736 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr &&
737 myaddr.s_addr != 0) {
738 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n",
739 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
740 inet_ntoa(isaddr), ifp->if_xname);
741 itaddr = myaddr;
742 ARPSTAT_INC(dupips);
743 goto reply;
744 }
745 if (ifp->if_flags & IFF_STATICARP)
746 goto reply;
747
748 bzero(&sin, sizeof(sin));
749 sin.sin_len = sizeof(struct sockaddr_in);
750 sin.sin_family = AF_INET;
751 sin.sin_addr = isaddr;
752 flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0;
753 flags |= LLE_EXCLUSIVE;
754 IF_AFDATA_LOCK(ifp);
755 la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin);
756 IF_AFDATA_UNLOCK(ifp);
757 if (la != NULL) {
758 /* the following is not an error when doing bridging */
759 if (!bridged && la->lle_tbl->llt_ifp != ifp) {
760 if (log_arp_wrong_iface)
761 ARP_LOG(LOG_WARNING, "%s is on %s "
762 "but got reply from %*D on %s\n",
763 inet_ntoa(isaddr),
764 la->lle_tbl->llt_ifp->if_xname,
765 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
766 ifp->if_xname);
767 LLE_WUNLOCK(la);
768 goto reply;
769 }
770 if ((la->la_flags & LLE_VALID) &&
771 bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) {
772 if (la->la_flags & LLE_STATIC) {
773 LLE_WUNLOCK(la);
774 if (log_arp_permanent_modify)
775 ARP_LOG(LOG_ERR,
776 "%*D attempts to modify "
777 "permanent entry for %s on %s\n",
778 ifp->if_addrlen,
779 (u_char *)ar_sha(ah), ":",
780 inet_ntoa(isaddr), ifp->if_xname);
781 goto reply;
782 }
783 if (log_arp_movements) {
784 ARP_LOG(LOG_INFO, "%s moved from %*D "
785 "to %*D on %s\n",
786 inet_ntoa(isaddr),
787 ifp->if_addrlen,
788 (u_char *)&la->ll_addr, ":",
789 ifp->if_addrlen, (u_char *)ar_sha(ah), ":",
790 ifp->if_xname);
791 }
792 }
793
794 if (ifp->if_addrlen != ah->ar_hln) {
795 LLE_WUNLOCK(la);
796 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, "
797 "i/f %d (ignored)\n", ifp->if_addrlen,
798 (u_char *) ar_sha(ah), ":", ah->ar_hln,
799 ifp->if_addrlen);
800 goto drop;
801 }
802 (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen);
803 la->la_flags |= LLE_VALID;
804
805 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED);
806
807 if (!(la->la_flags & LLE_STATIC)) {
808 int canceled;
809
810 LLE_ADDREF(la);
811 la->la_expire = time_uptime + V_arpt_keep;
812 canceled = callout_reset(&la->la_timer,
813 hz * V_arpt_keep, arptimer, la);
814 if (canceled)
815 LLE_REMREF(la);
816 }
817 la->la_asked = 0;
818 la->la_preempt = V_arp_maxtries;
819 /*
820 * The packets are all freed within the call to the output
821 * routine.
822 *
823 * NB: The lock MUST be released before the call to the
824 * output routine.
825 */
826 if (la->la_hold != NULL) {
827 struct mbuf *m_hold, *m_hold_next;
828
829 m_hold = la->la_hold;
830 la->la_hold = NULL;
831 la->la_numheld = 0;
832 memcpy(&sa, L3_ADDR(la), sizeof(sa));
833 LLE_WUNLOCK(la);
834 for (; m_hold != NULL; m_hold = m_hold_next) {
835 m_hold_next = m_hold->m_nextpkt;
836 m_hold->m_nextpkt = NULL;
837 /* Avoid confusing lower layers. */
838 m_clrprotoflags(m_hold);
839 (*ifp->if_output)(ifp, m_hold, &sa, NULL);
840 }
841 } else
842 LLE_WUNLOCK(la);
843 }
844 reply:
845 if (op != ARPOP_REQUEST)
846 goto drop;
847 ARPSTAT_INC(rxrequests);
848
849 if (itaddr.s_addr == myaddr.s_addr) {
850 /* Shortcut.. the receiving interface is the target. */
851 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
852 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
853 } else {
854 struct llentry *lle = NULL;
855
856 sin.sin_addr = itaddr;
857 IF_AFDATA_RLOCK(ifp);
858 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin);
859 IF_AFDATA_RUNLOCK(ifp);
860
861 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) {
862 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
863 (void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln);
864 LLE_RUNLOCK(lle);
865 } else {
866
867 if (lle != NULL)
868 LLE_RUNLOCK(lle);
869
870 if (!V_arp_proxyall)
871 goto drop;
872
873 sin.sin_addr = itaddr;
874 /* XXX MRT use table 0 for arp reply */
875 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
876 if (!rt)
877 goto drop;
878
879 /*
880 * Don't send proxies for nodes on the same interface
881 * as this one came out of, or we'll get into a fight
882 * over who claims what Ether address.
883 */
884 if (!rt->rt_ifp || rt->rt_ifp == ifp) {
885 RTFREE_LOCKED(rt);
886 goto drop;
887 }
888 RTFREE_LOCKED(rt);
889
890 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln);
891 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln);
892
893 /*
894 * Also check that the node which sent the ARP packet
895 * is on the interface we expect it to be on. This
896 * avoids ARP chaos if an interface is connected to the
897 * wrong network.
898 */
899 sin.sin_addr = isaddr;
900
901 /* XXX MRT use table 0 for arp checks */
902 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0);
903 if (!rt)
904 goto drop;
905 if (rt->rt_ifp != ifp) {
906 ARP_LOG(LOG_INFO, "proxy: ignoring request"
907 " from %s via %s, expecting %s\n",
908 inet_ntoa(isaddr), ifp->if_xname,
909 rt->rt_ifp->if_xname);
910 RTFREE_LOCKED(rt);
911 goto drop;
912 }
913 RTFREE_LOCKED(rt);
914
915 #ifdef DEBUG_PROXY
916 printf("arp: proxying for %s\n", inet_ntoa(itaddr));
917 #endif
918 }
919 }
920
921 if (itaddr.s_addr == myaddr.s_addr &&
922 IN_LINKLOCAL(ntohl(itaddr.s_addr))) {
923 /* RFC 3927 link-local IPv4; always reply by broadcast. */
924 #ifdef DEBUG_LINKLOCAL
925 printf("arp: sending reply for link-local addr %s\n",
926 inet_ntoa(itaddr));
927 #endif
928 m->m_flags |= M_BCAST;
929 m->m_flags &= ~M_MCAST;
930 } else {
931 /* default behaviour; never reply by broadcast. */
932 m->m_flags &= ~(M_BCAST|M_MCAST);
933 }
934 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln);
935 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln);
936 ah->ar_op = htons(ARPOP_REPLY);
937 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */
938 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln);
939 m->m_pkthdr.len = m->m_len;
940 m->m_pkthdr.rcvif = NULL;
941 sa.sa_family = AF_ARP;
942 sa.sa_len = 2;
943 m_clrprotoflags(m); /* Avoid confusing lower layers. */
944 (*ifp->if_output)(ifp, m, &sa, NULL);
945 ARPSTAT_INC(txreplies);
946 return;
947
948 drop:
949 m_freem(m);
950 }
951 #endif
952
953 /*
954 * Handle the garp_rexmit_count. Like sysctl_handle_int(), but limits the range
955 * of valid values.
956 */
957 static int
958 sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS)
959 {
960 int error;
961 int rexmit_count = *(int *)arg1;
962
963 error = sysctl_handle_int(oidp, &rexmit_count, 0, req);
964
965 /* Enforce limits on any new value that may have been set. */
966 if (!error && req->newptr) {
967 /* A new value was set. */
968 if (rexmit_count < 0) {
969 rexmit_count = 0;
970 } else if (rexmit_count > MAX_GARP_RETRANSMITS) {
971 rexmit_count = MAX_GARP_RETRANSMITS;
972 }
973 *(int *)arg1 = rexmit_count;
974 }
975
976 return (error);
977 }
978
979 /*
980 * Retransmit a Gratuitous ARP (GARP) and, if necessary, schedule a callout to
981 * retransmit it again. A pending callout owns a reference to the ifa.
982 */
983 static void
984 garp_rexmit(void *arg)
985 {
986 struct in_ifaddr *ia = arg;
987
988 if (callout_pending(&ia->ia_garp_timer) ||
989 !callout_active(&ia->ia_garp_timer)) {
990 IFA_UNLOCK(&ia->ia_ifa);
991 ifa_free(&ia->ia_ifa);
992 return;
993 }
994
995 /*
996 * Drop ifa lock while the ARP request is generated.
997 */
998 IFA_UNLOCK(&ia->ia_ifa);
999
1000 arprequest(ia->ia_ifa.ifa_ifp, &IA_SIN(ia)->sin_addr,
1001 &IA_SIN(ia)->sin_addr, IF_LLADDR(ia->ia_ifa.ifa_ifp));
1002
1003 /*
1004 * Increment the count of retransmissions. If the count has reached the
1005 * maximum value, stop sending the GARP packets. Otherwise, schedule
1006 * the callout to retransmit another GARP packet.
1007 */
1008 ++ia->ia_garp_count;
1009 if (ia->ia_garp_count >= garp_rexmit_count) {
1010 ifa_free(&ia->ia_ifa);
1011 } else {
1012 int rescheduled;
1013 IFA_LOCK(&ia->ia_ifa);
1014 rescheduled = callout_reset(&ia->ia_garp_timer,
1015 (1 << ia->ia_garp_count) * hz,
1016 garp_rexmit, ia);
1017 IFA_UNLOCK(&ia->ia_ifa);
1018 if (rescheduled) {
1019 ifa_free(&ia->ia_ifa);
1020 }
1021 }
1022 }
1023
1024 /*
1025 * Start the GARP retransmit timer.
1026 *
1027 * A single GARP is always transmitted when an IPv4 address is added
1028 * to an interface and that is usually sufficient. However, in some
1029 * circumstances, such as when a shared address is passed between
1030 * cluster nodes, this single GARP may occasionally be dropped or
1031 * lost. This can lead to neighbors on the network link working with a
1032 * stale ARP cache and sending packets destined for that address to
1033 * the node that previously owned the address, which may not respond.
1034 *
1035 * To avoid this situation, GARP retransmits can be enabled by setting
1036 * the net.link.ether.inet.garp_rexmit_count sysctl to a value greater
1037 * than zero. The setting represents the maximum number of
1038 * retransmissions. The interval between retransmissions is calculated
1039 * using an exponential backoff algorithm, doubling each time, so the
1040 * retransmission intervals are: {1, 2, 4, 8, 16, ...} (seconds).
1041 */
1042 static void
1043 garp_timer_start(struct ifaddr *ifa)
1044 {
1045 struct in_ifaddr *ia = (struct in_ifaddr *) ifa;
1046
1047 IFA_LOCK(ifa);
1048 ia->ia_garp_count = 0;
1049 if (callout_reset(&ia->ia_garp_timer, (1 << ia->ia_garp_count) * hz,
1050 garp_rexmit, ia) == 0) {
1051 ifa_ref(ifa);
1052 }
1053 IFA_UNLOCK(ifa);
1054 }
1055
1056 void
1057 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa)
1058 {
1059 struct llentry *lle;
1060
1061 if (ifa->ifa_carp != NULL)
1062 return;
1063
1064 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) {
1065 arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1066 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp));
1067 if (garp_rexmit_count > 0) {
1068 garp_timer_start(ifa);
1069 }
1070
1071 /*
1072 * interface address is considered static entry
1073 * because the output of the arp utility shows
1074 * that L2 entry as permanent
1075 */
1076 IF_AFDATA_LOCK(ifp);
1077 lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC),
1078 (struct sockaddr *)IA_SIN(ifa));
1079 IF_AFDATA_UNLOCK(ifp);
1080 if (lle == NULL)
1081 log(LOG_INFO, "arp_ifinit: cannot create arp "
1082 "entry for interface address\n");
1083 else
1084 LLE_RUNLOCK(lle);
1085 }
1086 ifa->ifa_rtrequest = NULL;
1087 }
1088
1089 void
1090 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr)
1091 {
1092 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY)
1093 arprequest(ifp, &IA_SIN(ifa)->sin_addr,
1094 &IA_SIN(ifa)->sin_addr, enaddr);
1095 ifa->ifa_rtrequest = NULL;
1096 }
1097
1098 static void
1099 arp_init(void)
1100 {
1101
1102 netisr_register(&arp_nh);
1103 }
1104 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0);
Cache object: 33a9e8911e6d819aba8a81cc54b3d8b1
|