if_ether.c revision 278801
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: stable/10/sys/netinet/if_ether.c 278801 2015-02-15 13:57:44Z rrs $"); 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 79SYSCTL_DECL(_net_link_ether); 80static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 81static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, ""); 82 83/* timer values */ 84static VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20 85 * minutes */ 86static VNET_DEFINE(int, arp_maxtries) = 5; 87VNET_DEFINE(int, useloopback) = 1; /* use loopback interface for 88 * local traffic */ 89static VNET_DEFINE(int, arp_proxyall) = 0; 90static VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for 91 * 20 seconds */ 92VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ 93VNET_PCPUSTAT_SYSINIT(arpstat); 94 95#ifdef VIMAGE 96VNET_PCPUSTAT_SYSUNINIT(arpstat); 97#endif /* VIMAGE */ 98 99static 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 107SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 108 &VNET_NAME(arpt_keep), 0, 109 "ARP entry lifetime in seconds"); 110SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 111 &VNET_NAME(arp_maxtries), 0, 112 "ARP resolution attempts before returning error"); 113SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 114 &VNET_NAME(useloopback), 0, 115 "Use the loopback interface for local traffic"); 116SYSCTL_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"); 119SYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW, 120 &VNET_NAME(arpt_down), 0, 121 "Incomplete ARP entry lifetime in seconds"); 122SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat, 123 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)"); 124SYSCTL_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 128static void arp_init(void); 129static void arpintr(struct mbuf *); 130static void arptimer(void *); 131#ifdef INET 132static void in_arpinput(struct mbuf *); 133#endif 134 135static const struct netisr_handler arp_nh = { 136 .nh_name = "arp", 137 .nh_handler = arpintr, 138 .nh_proto = NETISR_ARP, 139 .nh_policy = NETISR_POLICY_SOURCE, 140}; 141 142#ifdef AF_INET 143/* 144 * called by in_ifscrub to remove entry from the table when 145 * the interface goes away 146 */ 147void 148arp_ifscrub(struct ifnet *ifp, uint32_t addr) 149{ 150 struct sockaddr_in addr4; 151 152 bzero((void *)&addr4, sizeof(addr4)); 153 addr4.sin_len = sizeof(addr4); 154 addr4.sin_family = AF_INET; 155 addr4.sin_addr.s_addr = addr; 156 IF_AFDATA_RLOCK(ifp); 157 lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR), 158 (struct sockaddr *)&addr4); 159 IF_AFDATA_RUNLOCK(ifp); 160} 161#endif 162 163/* 164 * Timeout routine. Age arp_tab entries periodically. 165 */ 166static void 167arptimer(void *arg) 168{ 169 struct llentry *lle = (struct llentry *)arg; 170 struct ifnet *ifp; 171 172 if (lle->la_flags & LLE_STATIC) { 173 return; 174 } 175 LLE_WLOCK(lle); 176 if (callout_pending(&lle->la_timer)) { 177 /* 178 * Here we are a bit odd here in the treatment of 179 * active/pending. If the pending bit is set, it got 180 * rescheduled before I ran. The active 181 * bit we ignore, since if it was stopped 182 * in ll_tablefree() and was currently running 183 * it would have return 0 so the code would 184 * not have deleted it since the callout could 185 * not be stopped so we want to go through 186 * with the delete here now. If the callout 187 * was restarted, the pending bit will be back on and 188 * we just want to bail since the callout_reset would 189 * return 1 and our reference would have been removed 190 * by arpresolve() below. 191 */ 192 LLE_WUNLOCK(lle); 193 return; 194 } 195 ifp = lle->lle_tbl->llt_ifp; 196 CURVNET_SET(ifp->if_vnet); 197 198 if ((lle->la_flags & LLE_DELETED) == 0) { 199 int evt; 200 201 if (lle->la_flags & LLE_VALID) 202 evt = LLENTRY_EXPIRED; 203 else 204 evt = LLENTRY_TIMEDOUT; 205 EVENTHANDLER_INVOKE(lle_event, lle, evt); 206 } 207 208 callout_stop(&lle->la_timer); 209 210 /* XXX: LOR avoidance. We still have ref on lle. */ 211 LLE_WUNLOCK(lle); 212 IF_AFDATA_LOCK(ifp); 213 LLE_WLOCK(lle); 214 215 /* Guard against race with other llentry_free(). */ 216 if (lle->la_flags & LLE_LINKED) { 217 size_t pkts_dropped; 218 219 LLE_REMREF(lle); 220 pkts_dropped = llentry_free(lle); 221 ARPSTAT_ADD(dropped, pkts_dropped); 222 } else 223 LLE_FREE_LOCKED(lle); 224 225 IF_AFDATA_UNLOCK(ifp); 226 227 ARPSTAT_INC(timeouts); 228 229 CURVNET_RESTORE(); 230} 231 232/* 233 * Broadcast an ARP request. Caller specifies: 234 * - arp header source ip address 235 * - arp header target ip address 236 * - arp header source ethernet address 237 */ 238void 239arprequest(struct ifnet *ifp, const struct in_addr *sip, 240 const struct in_addr *tip, u_char *enaddr) 241{ 242 struct mbuf *m; 243 struct arphdr *ah; 244 struct sockaddr sa; 245 u_char *carpaddr = NULL; 246 247 if (sip == NULL) { 248 /* 249 * The caller did not supply a source address, try to find 250 * a compatible one among those assigned to this interface. 251 */ 252 struct ifaddr *ifa; 253 254 IF_ADDR_RLOCK(ifp); 255 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 256 if (ifa->ifa_addr->sa_family != AF_INET) 257 continue; 258 259 if (ifa->ifa_carp) { 260 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0) 261 continue; 262 sip = &IA_SIN(ifa)->sin_addr; 263 } else { 264 carpaddr = NULL; 265 sip = &IA_SIN(ifa)->sin_addr; 266 } 267 268 if (0 == ((sip->s_addr ^ tip->s_addr) & 269 IA_MASKSIN(ifa)->sin_addr.s_addr)) 270 break; /* found it. */ 271 } 272 IF_ADDR_RUNLOCK(ifp); 273 if (sip == NULL) { 274 printf("%s: cannot find matching address\n", __func__); 275 return; 276 } 277 } 278 if (enaddr == NULL) 279 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp); 280 281 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) 282 return; 283 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 284 2*ifp->if_data.ifi_addrlen; 285 m->m_pkthdr.len = m->m_len; 286 MH_ALIGN(m, m->m_len); 287 ah = mtod(m, struct arphdr *); 288 bzero((caddr_t)ah, m->m_len); 289#ifdef MAC 290 mac_netinet_arp_send(ifp, m); 291#endif 292 ah->ar_pro = htons(ETHERTYPE_IP); 293 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 294 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 295 ah->ar_op = htons(ARPOP_REQUEST); 296 bcopy(enaddr, ar_sha(ah), ah->ar_hln); 297 bcopy(sip, ar_spa(ah), ah->ar_pln); 298 bcopy(tip, ar_tpa(ah), ah->ar_pln); 299 sa.sa_family = AF_ARP; 300 sa.sa_len = 2; 301 m->m_flags |= M_BCAST; 302 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 303 (*ifp->if_output)(ifp, m, &sa, NULL); 304 ARPSTAT_INC(txrequests); 305} 306 307/* 308 * Resolve an IP address into an ethernet address. 309 * On input: 310 * ifp is the interface we use 311 * rt0 is the route to the final destination (possibly useless) 312 * m is the mbuf. May be NULL if we don't have a packet. 313 * dst is the next hop, 314 * desten is where we want the address. 315 * 316 * On success, desten is filled in and the function returns 0; 317 * If the packet must be held pending resolution, we return EWOULDBLOCK 318 * On other errors, we return the corresponding error code. 319 * Note that m_freem() handles NULL. 320 */ 321int 322arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 323 const struct sockaddr *dst, u_char *desten, struct llentry **lle) 324{ 325 struct llentry *la = 0; 326 u_int flags = 0; 327 struct mbuf *curr = NULL; 328 struct mbuf *next = NULL; 329 int error, renew; 330 331 *lle = NULL; 332 if (m != NULL) { 333 if (m->m_flags & M_BCAST) { 334 /* broadcast */ 335 (void)memcpy(desten, 336 ifp->if_broadcastaddr, ifp->if_addrlen); 337 return (0); 338 } 339 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) { 340 /* multicast */ 341 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 342 return (0); 343 } 344 } 345retry: 346 IF_AFDATA_RLOCK(ifp); 347 la = lla_lookup(LLTABLE(ifp), flags, dst); 348 IF_AFDATA_RUNLOCK(ifp); 349 if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0) 350 && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) { 351 flags |= (LLE_CREATE | LLE_EXCLUSIVE); 352 IF_AFDATA_WLOCK(ifp); 353 la = lla_lookup(LLTABLE(ifp), flags, dst); 354 IF_AFDATA_WUNLOCK(ifp); 355 } 356 if (la == NULL) { 357 if (flags & LLE_CREATE) 358 log(LOG_DEBUG, 359 "arpresolve: can't allocate llinfo for %s on %s\n", 360 inet_ntoa(SIN(dst)->sin_addr), ifp->if_xname); 361 m_freem(m); 362 return (EINVAL); 363 } 364 365 if ((la->la_flags & LLE_VALID) && 366 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { 367 bcopy(&la->ll_addr, desten, ifp->if_addrlen); 368 /* 369 * If entry has an expiry time and it is approaching, 370 * see if we need to send an ARP request within this 371 * arpt_down interval. 372 */ 373 if (!(la->la_flags & LLE_STATIC) && 374 time_uptime + la->la_preempt > la->la_expire) { 375 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); 376 la->la_preempt--; 377 } 378 379 *lle = la; 380 error = 0; 381 goto done; 382 } 383 384 if (la->la_flags & LLE_STATIC) { /* should not happen! */ 385 log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n", 386 inet_ntoa(SIN(dst)->sin_addr)); 387 m_freem(m); 388 error = EINVAL; 389 goto done; 390 } 391 392 renew = (la->la_asked == 0 || la->la_expire != time_uptime); 393 if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) { 394 flags |= LLE_EXCLUSIVE; 395 LLE_RUNLOCK(la); 396 goto retry; 397 } 398 /* 399 * There is an arptab entry, but no ethernet address 400 * response yet. Add the mbuf to the list, dropping 401 * the oldest packet if we have exceeded the system 402 * setting. 403 */ 404 if (m != NULL) { 405 if (la->la_numheld >= V_arp_maxhold) { 406 if (la->la_hold != NULL) { 407 next = la->la_hold->m_nextpkt; 408 m_freem(la->la_hold); 409 la->la_hold = next; 410 la->la_numheld--; 411 ARPSTAT_INC(dropped); 412 } 413 } 414 if (la->la_hold != NULL) { 415 curr = la->la_hold; 416 while (curr->m_nextpkt != NULL) 417 curr = curr->m_nextpkt; 418 curr->m_nextpkt = m; 419 } else 420 la->la_hold = m; 421 la->la_numheld++; 422 if (renew == 0 && (flags & LLE_EXCLUSIVE)) { 423 flags &= ~LLE_EXCLUSIVE; 424 LLE_DOWNGRADE(la); 425 } 426 427 } 428 /* 429 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 430 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 431 * if we have already sent arp_maxtries ARP requests. Retransmit the 432 * ARP request, but not faster than one request per second. 433 */ 434 if (la->la_asked < V_arp_maxtries) 435 error = EWOULDBLOCK; /* First request. */ 436 else 437 error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ? 438 EHOSTUNREACH : EHOSTDOWN; 439 440 if (renew) { 441 int canceled; 442 443 LLE_ADDREF(la); 444 la->la_expire = time_uptime; 445 canceled = callout_reset(&la->la_timer, hz * V_arpt_down, 446 arptimer, la); 447 if (canceled) 448 LLE_REMREF(la); 449 la->la_asked++; 450 LLE_WUNLOCK(la); 451 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); 452 return (error); 453 } 454done: 455 if (flags & LLE_EXCLUSIVE) 456 LLE_WUNLOCK(la); 457 else 458 LLE_RUNLOCK(la); 459 return (error); 460} 461 462/* 463 * Common length and type checks are done here, 464 * then the protocol-specific routine is called. 465 */ 466static void 467arpintr(struct mbuf *m) 468{ 469 struct arphdr *ar; 470 471 if (m->m_len < sizeof(struct arphdr) && 472 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 473 log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n"); 474 return; 475 } 476 ar = mtod(m, struct arphdr *); 477 478 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 479 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 480 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 481 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 && 482 ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) { 483 log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)" 484 " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "", 485 ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":", 486 ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":"); 487 m_freem(m); 488 return; 489 } 490 491 if (m->m_len < arphdr_len(ar)) { 492 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 493 log(LOG_NOTICE, "arp: runt packet\n"); 494 m_freem(m); 495 return; 496 } 497 ar = mtod(m, struct arphdr *); 498 } 499 500 ARPSTAT_INC(received); 501 switch (ntohs(ar->ar_pro)) { 502#ifdef INET 503 case ETHERTYPE_IP: 504 in_arpinput(m); 505 return; 506#endif 507 } 508 m_freem(m); 509} 510 511#ifdef INET 512/* 513 * ARP for Internet protocols on 10 Mb/s Ethernet. 514 * Algorithm is that given in RFC 826. 515 * In addition, a sanity check is performed on the sender 516 * protocol address, to catch impersonators. 517 * We no longer handle negotiations for use of trailer protocol: 518 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 519 * along with IP replies if we wanted trailers sent to us, 520 * and also sent them in response to IP replies. 521 * This allowed either end to announce the desire to receive 522 * trailer packets. 523 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 524 * but formerly didn't normally send requests. 525 */ 526static int log_arp_wrong_iface = 1; 527static int log_arp_movements = 1; 528static int log_arp_permanent_modify = 1; 529static int allow_multicast = 0; 530static struct timeval arp_lastlog; 531static int arp_curpps; 532static int arp_maxpps = 1; 533 534SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 535 &log_arp_wrong_iface, 0, 536 "log arp packets arriving on the wrong interface"); 537SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 538 &log_arp_movements, 0, 539 "log arp replies from MACs different than the one in the cache"); 540SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 541 &log_arp_permanent_modify, 0, 542 "log arp replies from MACs different than the one in the permanent arp entry"); 543SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW, 544 &allow_multicast, 0, "accept multicast addresses"); 545SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, 546 CTLFLAG_RW, &arp_maxpps, 0, 547 "Maximum number of remotely triggered ARP messages that can be " 548 "logged per second"); 549 550#define ARP_LOG(pri, ...) do { \ 551 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ 552 log((pri), "arp: " __VA_ARGS__); \ 553} while (0) 554 555static void 556in_arpinput(struct mbuf *m) 557{ 558 struct arphdr *ah; 559 struct ifnet *ifp = m->m_pkthdr.rcvif; 560 struct llentry *la = NULL; 561 struct rtentry *rt; 562 struct ifaddr *ifa; 563 struct in_ifaddr *ia; 564 struct sockaddr sa; 565 struct in_addr isaddr, itaddr, myaddr; 566 u_int8_t *enaddr = NULL; 567 int op, flags; 568 int req_len; 569 int bridged = 0, is_bridge = 0; 570 int carped; 571 struct sockaddr_in sin; 572 sin.sin_len = sizeof(struct sockaddr_in); 573 sin.sin_family = AF_INET; 574 sin.sin_addr.s_addr = 0; 575 576 if (ifp->if_bridge) 577 bridged = 1; 578 if (ifp->if_type == IFT_BRIDGE) 579 is_bridge = 1; 580 581 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 582 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 583 ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n"); 584 return; 585 } 586 587 ah = mtod(m, struct arphdr *); 588 /* 589 * ARP is only for IPv4 so we can reject packets with 590 * a protocol length not equal to an IPv4 address. 591 */ 592 if (ah->ar_pln != sizeof(struct in_addr)) { 593 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n", 594 sizeof(struct in_addr)); 595 goto drop; 596 } 597 598 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) { 599 ARP_LOG(LOG_NOTICE, "%*D is multicast\n", 600 ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); 601 goto drop; 602 } 603 604 op = ntohs(ah->ar_op); 605 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 606 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 607 608 if (op == ARPOP_REPLY) 609 ARPSTAT_INC(rxreplies); 610 611 /* 612 * For a bridge, we want to check the address irrespective 613 * of the receive interface. (This will change slightly 614 * when we have clusters of interfaces). 615 */ 616 IN_IFADDR_RLOCK(); 617 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 618 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 619 ia->ia_ifp == ifp) && 620 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr && 621 (ia->ia_ifa.ifa_carp == NULL || 622 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) { 623 ifa_ref(&ia->ia_ifa); 624 IN_IFADDR_RUNLOCK(); 625 goto match; 626 } 627 } 628 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 629 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 630 ia->ia_ifp == ifp) && 631 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 632 ifa_ref(&ia->ia_ifa); 633 IN_IFADDR_RUNLOCK(); 634 goto match; 635 } 636 637#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 638 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 639 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 640 addr == ia->ia_addr.sin_addr.s_addr) 641 /* 642 * Check the case when bridge shares its MAC address with 643 * some of its children, so packets are claimed by bridge 644 * itself (bridge_input() does it first), but they are really 645 * meant to be destined to the bridge member. 646 */ 647 if (is_bridge) { 648 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 649 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 650 ifa_ref(&ia->ia_ifa); 651 ifp = ia->ia_ifp; 652 IN_IFADDR_RUNLOCK(); 653 goto match; 654 } 655 } 656 } 657#undef BDG_MEMBER_MATCHES_ARP 658 IN_IFADDR_RUNLOCK(); 659 660 /* 661 * No match, use the first inet address on the receive interface 662 * as a dummy address for the rest of the function. 663 */ 664 IF_ADDR_RLOCK(ifp); 665 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 666 if (ifa->ifa_addr->sa_family == AF_INET && 667 (ifa->ifa_carp == NULL || 668 (*carp_iamatch_p)(ifa, &enaddr))) { 669 ia = ifatoia(ifa); 670 ifa_ref(ifa); 671 IF_ADDR_RUNLOCK(ifp); 672 goto match; 673 } 674 IF_ADDR_RUNLOCK(ifp); 675 676 /* 677 * If bridging, fall back to using any inet address. 678 */ 679 IN_IFADDR_RLOCK(); 680 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { 681 IN_IFADDR_RUNLOCK(); 682 goto drop; 683 } 684 ifa_ref(&ia->ia_ifa); 685 IN_IFADDR_RUNLOCK(); 686match: 687 if (!enaddr) 688 enaddr = (u_int8_t *)IF_LLADDR(ifp); 689 carped = (ia->ia_ifa.ifa_carp != NULL); 690 myaddr = ia->ia_addr.sin_addr; 691 ifa_free(&ia->ia_ifa); 692 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 693 goto drop; /* it's from me, ignore it. */ 694 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 695 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address " 696 "%s!\n", inet_ntoa(isaddr)); 697 goto drop; 698 } 699 /* 700 * Warn if another host is using the same IP address, but only if the 701 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 702 * case we suppress the warning to avoid false positive complaints of 703 * potential misconfiguration. 704 */ 705 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr && 706 myaddr.s_addr != 0) { 707 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n", 708 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 709 inet_ntoa(isaddr), ifp->if_xname); 710 itaddr = myaddr; 711 ARPSTAT_INC(dupips); 712 goto reply; 713 } 714 if (ifp->if_flags & IFF_STATICARP) 715 goto reply; 716 717 bzero(&sin, sizeof(sin)); 718 sin.sin_len = sizeof(struct sockaddr_in); 719 sin.sin_family = AF_INET; 720 sin.sin_addr = isaddr; 721 flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0; 722 flags |= LLE_EXCLUSIVE; 723 IF_AFDATA_LOCK(ifp); 724 la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin); 725 IF_AFDATA_UNLOCK(ifp); 726 if (la != NULL) { 727 /* the following is not an error when doing bridging */ 728 if (!bridged && la->lle_tbl->llt_ifp != ifp) { 729 if (log_arp_wrong_iface) 730 ARP_LOG(LOG_WARNING, "%s is on %s " 731 "but got reply from %*D on %s\n", 732 inet_ntoa(isaddr), 733 la->lle_tbl->llt_ifp->if_xname, 734 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 735 ifp->if_xname); 736 LLE_WUNLOCK(la); 737 goto reply; 738 } 739 if ((la->la_flags & LLE_VALID) && 740 bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) { 741 if (la->la_flags & LLE_STATIC) { 742 LLE_WUNLOCK(la); 743 if (log_arp_permanent_modify) 744 ARP_LOG(LOG_ERR, 745 "%*D attempts to modify " 746 "permanent entry for %s on %s\n", 747 ifp->if_addrlen, 748 (u_char *)ar_sha(ah), ":", 749 inet_ntoa(isaddr), ifp->if_xname); 750 goto reply; 751 } 752 if (log_arp_movements) { 753 ARP_LOG(LOG_INFO, "%s moved from %*D " 754 "to %*D on %s\n", 755 inet_ntoa(isaddr), 756 ifp->if_addrlen, 757 (u_char *)&la->ll_addr, ":", 758 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 759 ifp->if_xname); 760 } 761 } 762 763 if (ifp->if_addrlen != ah->ar_hln) { 764 LLE_WUNLOCK(la); 765 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, " 766 "i/f %d (ignored)\n", ifp->if_addrlen, 767 (u_char *) ar_sha(ah), ":", ah->ar_hln, 768 ifp->if_addrlen); 769 goto drop; 770 } 771 (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen); 772 la->la_flags |= LLE_VALID; 773 774 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED); 775 776 if (!(la->la_flags & LLE_STATIC)) { 777 int canceled; 778 779 LLE_ADDREF(la); 780 la->la_expire = time_uptime + V_arpt_keep; 781 canceled = callout_reset(&la->la_timer, 782 hz * V_arpt_keep, arptimer, la); 783 if (canceled) 784 LLE_REMREF(la); 785 } 786 la->la_asked = 0; 787 la->la_preempt = V_arp_maxtries; 788 /* 789 * The packets are all freed within the call to the output 790 * routine. 791 * 792 * NB: The lock MUST be released before the call to the 793 * output routine. 794 */ 795 if (la->la_hold != NULL) { 796 struct mbuf *m_hold, *m_hold_next; 797 798 m_hold = la->la_hold; 799 la->la_hold = NULL; 800 la->la_numheld = 0; 801 memcpy(&sa, L3_ADDR(la), sizeof(sa)); 802 LLE_WUNLOCK(la); 803 for (; m_hold != NULL; m_hold = m_hold_next) { 804 m_hold_next = m_hold->m_nextpkt; 805 m_hold->m_nextpkt = NULL; 806 /* Avoid confusing lower layers. */ 807 m_clrprotoflags(m_hold); 808 (*ifp->if_output)(ifp, m_hold, &sa, NULL); 809 } 810 } else 811 LLE_WUNLOCK(la); 812 } 813reply: 814 if (op != ARPOP_REQUEST) 815 goto drop; 816 ARPSTAT_INC(rxrequests); 817 818 if (itaddr.s_addr == myaddr.s_addr) { 819 /* Shortcut.. the receiving interface is the target. */ 820 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 821 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 822 } else { 823 struct llentry *lle = NULL; 824 825 sin.sin_addr = itaddr; 826 IF_AFDATA_RLOCK(ifp); 827 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); 828 IF_AFDATA_RUNLOCK(ifp); 829 830 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { 831 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 832 (void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln); 833 LLE_RUNLOCK(lle); 834 } else { 835 836 if (lle != NULL) 837 LLE_RUNLOCK(lle); 838 839 if (!V_arp_proxyall) 840 goto drop; 841 842 sin.sin_addr = itaddr; 843 /* XXX MRT use table 0 for arp reply */ 844 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 845 if (!rt) 846 goto drop; 847 848 /* 849 * Don't send proxies for nodes on the same interface 850 * as this one came out of, or we'll get into a fight 851 * over who claims what Ether address. 852 */ 853 if (!rt->rt_ifp || rt->rt_ifp == ifp) { 854 RTFREE_LOCKED(rt); 855 goto drop; 856 } 857 RTFREE_LOCKED(rt); 858 859 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 860 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 861 862 /* 863 * Also check that the node which sent the ARP packet 864 * is on the interface we expect it to be on. This 865 * avoids ARP chaos if an interface is connected to the 866 * wrong network. 867 */ 868 sin.sin_addr = isaddr; 869 870 /* XXX MRT use table 0 for arp checks */ 871 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 872 if (!rt) 873 goto drop; 874 if (rt->rt_ifp != ifp) { 875 ARP_LOG(LOG_INFO, "proxy: ignoring request" 876 " from %s via %s, expecting %s\n", 877 inet_ntoa(isaddr), ifp->if_xname, 878 rt->rt_ifp->if_xname); 879 RTFREE_LOCKED(rt); 880 goto drop; 881 } 882 RTFREE_LOCKED(rt); 883 884#ifdef DEBUG_PROXY 885 printf("arp: proxying for %s\n", inet_ntoa(itaddr)); 886#endif 887 } 888 } 889 890 if (itaddr.s_addr == myaddr.s_addr && 891 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 892 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 893#ifdef DEBUG_LINKLOCAL 894 printf("arp: sending reply for link-local addr %s\n", 895 inet_ntoa(itaddr)); 896#endif 897 m->m_flags |= M_BCAST; 898 m->m_flags &= ~M_MCAST; 899 } else { 900 /* default behaviour; never reply by broadcast. */ 901 m->m_flags &= ~(M_BCAST|M_MCAST); 902 } 903 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 904 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 905 ah->ar_op = htons(ARPOP_REPLY); 906 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 907 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 908 m->m_pkthdr.len = m->m_len; 909 m->m_pkthdr.rcvif = NULL; 910 sa.sa_family = AF_ARP; 911 sa.sa_len = 2; 912 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 913 (*ifp->if_output)(ifp, m, &sa, NULL); 914 ARPSTAT_INC(txreplies); 915 return; 916 917drop: 918 m_freem(m); 919} 920#endif 921 922void 923arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 924{ 925 struct llentry *lle; 926 927 if (ifa->ifa_carp != NULL) 928 return; 929 930 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) { 931 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 932 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 933 /* 934 * interface address is considered static entry 935 * because the output of the arp utility shows 936 * that L2 entry as permanent 937 */ 938 IF_AFDATA_LOCK(ifp); 939 lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC), 940 (struct sockaddr *)IA_SIN(ifa)); 941 IF_AFDATA_UNLOCK(ifp); 942 if (lle == NULL) 943 log(LOG_INFO, "arp_ifinit: cannot create arp " 944 "entry for interface address\n"); 945 else 946 LLE_RUNLOCK(lle); 947 } 948 ifa->ifa_rtrequest = NULL; 949} 950 951void 952arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 953{ 954 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 955 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 956 &IA_SIN(ifa)->sin_addr, enaddr); 957 ifa->ifa_rtrequest = NULL; 958} 959 960static void 961arp_init(void) 962{ 963 964 netisr_register(&arp_nh); 965} 966SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 967