if_ether.c revision 284776
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 284776 2015-06-24 19:06:54Z eri $"); 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_WLOCK(ifp); 157 lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR), 158 (struct sockaddr *)&addr4); 159 IF_AFDATA_WUNLOCK(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 renew = 0; 369 /* 370 * If entry has an expiry time and it is approaching, 371 * see if we need to send an ARP request within this 372 * arpt_down interval. 373 */ 374 if (!(la->la_flags & LLE_STATIC) && 375 time_uptime + la->la_preempt > la->la_expire) { 376 renew = 1; 377 la->la_preempt--; 378 } 379 380 *lle = la; 381 382 if (flags & LLE_EXCLUSIVE) 383 LLE_WUNLOCK(la); 384 else 385 LLE_RUNLOCK(la); 386 387 if (renew == 1) 388 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); 389 390 return (0); 391 } 392 393 if (la->la_flags & LLE_STATIC) { /* should not happen! */ 394 log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n", 395 inet_ntoa(SIN(dst)->sin_addr)); 396 m_freem(m); 397 error = EINVAL; 398 goto done; 399 } 400 401 renew = (la->la_asked == 0 || la->la_expire != time_uptime); 402 if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) { 403 flags |= LLE_EXCLUSIVE; 404 LLE_RUNLOCK(la); 405 goto retry; 406 } 407 /* 408 * There is an arptab entry, but no ethernet address 409 * response yet. Add the mbuf to the list, dropping 410 * the oldest packet if we have exceeded the system 411 * setting. 412 */ 413 if (m != NULL) { 414 if (la->la_numheld >= V_arp_maxhold) { 415 if (la->la_hold != NULL) { 416 next = la->la_hold->m_nextpkt; 417 m_freem(la->la_hold); 418 la->la_hold = next; 419 la->la_numheld--; 420 ARPSTAT_INC(dropped); 421 } 422 } 423 if (la->la_hold != NULL) { 424 curr = la->la_hold; 425 while (curr->m_nextpkt != NULL) 426 curr = curr->m_nextpkt; 427 curr->m_nextpkt = m; 428 } else 429 la->la_hold = m; 430 la->la_numheld++; 431 if (renew == 0 && (flags & LLE_EXCLUSIVE)) { 432 flags &= ~LLE_EXCLUSIVE; 433 LLE_DOWNGRADE(la); 434 } 435 436 } 437 /* 438 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 439 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 440 * if we have already sent arp_maxtries ARP requests. Retransmit the 441 * ARP request, but not faster than one request per second. 442 */ 443 if (la->la_asked < V_arp_maxtries) 444 error = EWOULDBLOCK; /* First request. */ 445 else 446 error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ? 447 EHOSTUNREACH : EHOSTDOWN; 448 449 if (renew) { 450 int canceled; 451 452 LLE_ADDREF(la); 453 la->la_expire = time_uptime; 454 canceled = callout_reset(&la->la_timer, hz * V_arpt_down, 455 arptimer, la); 456 if (canceled) 457 LLE_REMREF(la); 458 la->la_asked++; 459 LLE_WUNLOCK(la); 460 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); 461 return (error); 462 } 463done: 464 if (flags & LLE_EXCLUSIVE) 465 LLE_WUNLOCK(la); 466 else 467 LLE_RUNLOCK(la); 468 return (error); 469} 470 471/* 472 * Common length and type checks are done here, 473 * then the protocol-specific routine is called. 474 */ 475static void 476arpintr(struct mbuf *m) 477{ 478 struct arphdr *ar; 479 480 if (m->m_len < sizeof(struct arphdr) && 481 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 482 log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n"); 483 return; 484 } 485 ar = mtod(m, struct arphdr *); 486 487 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 488 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 489 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 490 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 && 491 ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) { 492 log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)" 493 " (from %*D to %*D)\n", (unsigned char *)&ar->ar_hrd, "", 494 ETHER_ADDR_LEN, (u_char *)ar_sha(ar), ":", 495 ETHER_ADDR_LEN, (u_char *)ar_tha(ar), ":"); 496 m_freem(m); 497 return; 498 } 499 500 if (m->m_len < arphdr_len(ar)) { 501 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 502 log(LOG_NOTICE, "arp: runt packet\n"); 503 m_freem(m); 504 return; 505 } 506 ar = mtod(m, struct arphdr *); 507 } 508 509 ARPSTAT_INC(received); 510 switch (ntohs(ar->ar_pro)) { 511#ifdef INET 512 case ETHERTYPE_IP: 513 in_arpinput(m); 514 return; 515#endif 516 } 517 m_freem(m); 518} 519 520#ifdef INET 521/* 522 * ARP for Internet protocols on 10 Mb/s Ethernet. 523 * Algorithm is that given in RFC 826. 524 * In addition, a sanity check is performed on the sender 525 * protocol address, to catch impersonators. 526 * We no longer handle negotiations for use of trailer protocol: 527 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 528 * along with IP replies if we wanted trailers sent to us, 529 * and also sent them in response to IP replies. 530 * This allowed either end to announce the desire to receive 531 * trailer packets. 532 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 533 * but formerly didn't normally send requests. 534 */ 535static int log_arp_wrong_iface = 1; 536static int log_arp_movements = 1; 537static int log_arp_permanent_modify = 1; 538static int allow_multicast = 0; 539static struct timeval arp_lastlog; 540static int arp_curpps; 541static int arp_maxpps = 1; 542 543SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 544 &log_arp_wrong_iface, 0, 545 "log arp packets arriving on the wrong interface"); 546SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 547 &log_arp_movements, 0, 548 "log arp replies from MACs different than the one in the cache"); 549SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 550 &log_arp_permanent_modify, 0, 551 "log arp replies from MACs different than the one in the permanent arp entry"); 552SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW, 553 &allow_multicast, 0, "accept multicast addresses"); 554SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, 555 CTLFLAG_RW, &arp_maxpps, 0, 556 "Maximum number of remotely triggered ARP messages that can be " 557 "logged per second"); 558 559#define ARP_LOG(pri, ...) do { \ 560 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ 561 log((pri), "arp: " __VA_ARGS__); \ 562} while (0) 563 564static void 565in_arpinput(struct mbuf *m) 566{ 567 struct arphdr *ah; 568 struct ifnet *ifp = m->m_pkthdr.rcvif; 569 struct llentry *la = NULL; 570 struct rtentry *rt; 571 struct ifaddr *ifa; 572 struct in_ifaddr *ia; 573 struct sockaddr sa; 574 struct in_addr isaddr, itaddr, myaddr; 575 u_int8_t *enaddr = NULL; 576 int op, flags; 577 int req_len; 578 int bridged = 0, is_bridge = 0; 579 int carped; 580 struct sockaddr_in sin; 581 sin.sin_len = sizeof(struct sockaddr_in); 582 sin.sin_family = AF_INET; 583 sin.sin_addr.s_addr = 0; 584 585 if (ifp->if_bridge) 586 bridged = 1; 587 if (ifp->if_type == IFT_BRIDGE) 588 is_bridge = 1; 589 590 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 591 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 592 ARP_LOG(LOG_NOTICE, "runt packet -- m_pullup failed\n"); 593 return; 594 } 595 596 ah = mtod(m, struct arphdr *); 597 /* 598 * ARP is only for IPv4 so we can reject packets with 599 * a protocol length not equal to an IPv4 address. 600 */ 601 if (ah->ar_pln != sizeof(struct in_addr)) { 602 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n", 603 sizeof(struct in_addr)); 604 goto drop; 605 } 606 607 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) { 608 ARP_LOG(LOG_NOTICE, "%*D is multicast\n", 609 ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); 610 goto drop; 611 } 612 613 op = ntohs(ah->ar_op); 614 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 615 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 616 617 if (op == ARPOP_REPLY) 618 ARPSTAT_INC(rxreplies); 619 620 /* 621 * For a bridge, we want to check the address irrespective 622 * of the receive interface. (This will change slightly 623 * when we have clusters of interfaces). 624 */ 625 IN_IFADDR_RLOCK(); 626 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 627 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 628 ia->ia_ifp == ifp) && 629 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr && 630 (ia->ia_ifa.ifa_carp == NULL || 631 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) { 632 ifa_ref(&ia->ia_ifa); 633 IN_IFADDR_RUNLOCK(); 634 goto match; 635 } 636 } 637 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 638 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 639 ia->ia_ifp == ifp) && 640 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 641 ifa_ref(&ia->ia_ifa); 642 IN_IFADDR_RUNLOCK(); 643 goto match; 644 } 645 646#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 647 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 648 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 649 addr == ia->ia_addr.sin_addr.s_addr) 650 /* 651 * Check the case when bridge shares its MAC address with 652 * some of its children, so packets are claimed by bridge 653 * itself (bridge_input() does it first), but they are really 654 * meant to be destined to the bridge member. 655 */ 656 if (is_bridge) { 657 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 658 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 659 ifa_ref(&ia->ia_ifa); 660 ifp = ia->ia_ifp; 661 IN_IFADDR_RUNLOCK(); 662 goto match; 663 } 664 } 665 } 666#undef BDG_MEMBER_MATCHES_ARP 667 IN_IFADDR_RUNLOCK(); 668 669 /* 670 * No match, use the first inet address on the receive interface 671 * as a dummy address for the rest of the function. 672 */ 673 IF_ADDR_RLOCK(ifp); 674 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 675 if (ifa->ifa_addr->sa_family == AF_INET && 676 (ifa->ifa_carp == NULL || 677 (*carp_iamatch_p)(ifa, &enaddr))) { 678 ia = ifatoia(ifa); 679 ifa_ref(ifa); 680 IF_ADDR_RUNLOCK(ifp); 681 goto match; 682 } 683 IF_ADDR_RUNLOCK(ifp); 684 685 /* 686 * If bridging, fall back to using any inet address. 687 */ 688 IN_IFADDR_RLOCK(); 689 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { 690 IN_IFADDR_RUNLOCK(); 691 goto drop; 692 } 693 ifa_ref(&ia->ia_ifa); 694 IN_IFADDR_RUNLOCK(); 695match: 696 if (!enaddr) 697 enaddr = (u_int8_t *)IF_LLADDR(ifp); 698 carped = (ia->ia_ifa.ifa_carp != NULL); 699 myaddr = ia->ia_addr.sin_addr; 700 ifa_free(&ia->ia_ifa); 701 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 702 goto drop; /* it's from me, ignore it. */ 703 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 704 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address " 705 "%s!\n", inet_ntoa(isaddr)); 706 goto drop; 707 } 708 /* 709 * Warn if another host is using the same IP address, but only if the 710 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 711 * case we suppress the warning to avoid false positive complaints of 712 * potential misconfiguration. 713 */ 714 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr && 715 myaddr.s_addr != 0) { 716 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n", 717 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 718 inet_ntoa(isaddr), ifp->if_xname); 719 itaddr = myaddr; 720 ARPSTAT_INC(dupips); 721 goto reply; 722 } 723 if (ifp->if_flags & IFF_STATICARP) 724 goto reply; 725 726 bzero(&sin, sizeof(sin)); 727 sin.sin_len = sizeof(struct sockaddr_in); 728 sin.sin_family = AF_INET; 729 sin.sin_addr = isaddr; 730 flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0; 731 flags |= LLE_EXCLUSIVE; 732 IF_AFDATA_LOCK(ifp); 733 la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin); 734 IF_AFDATA_UNLOCK(ifp); 735 if (la != NULL) { 736 /* the following is not an error when doing bridging */ 737 if (!bridged && la->lle_tbl->llt_ifp != ifp) { 738 if (log_arp_wrong_iface) 739 ARP_LOG(LOG_WARNING, "%s is on %s " 740 "but got reply from %*D on %s\n", 741 inet_ntoa(isaddr), 742 la->lle_tbl->llt_ifp->if_xname, 743 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 744 ifp->if_xname); 745 LLE_WUNLOCK(la); 746 goto reply; 747 } 748 if ((la->la_flags & LLE_VALID) && 749 bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) { 750 if (la->la_flags & LLE_STATIC) { 751 LLE_WUNLOCK(la); 752 if (log_arp_permanent_modify) 753 ARP_LOG(LOG_ERR, 754 "%*D attempts to modify " 755 "permanent entry for %s on %s\n", 756 ifp->if_addrlen, 757 (u_char *)ar_sha(ah), ":", 758 inet_ntoa(isaddr), ifp->if_xname); 759 goto reply; 760 } 761 if (log_arp_movements) { 762 ARP_LOG(LOG_INFO, "%s moved from %*D " 763 "to %*D on %s\n", 764 inet_ntoa(isaddr), 765 ifp->if_addrlen, 766 (u_char *)&la->ll_addr, ":", 767 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 768 ifp->if_xname); 769 } 770 } 771 772 if (ifp->if_addrlen != ah->ar_hln) { 773 LLE_WUNLOCK(la); 774 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, " 775 "i/f %d (ignored)\n", ifp->if_addrlen, 776 (u_char *) ar_sha(ah), ":", ah->ar_hln, 777 ifp->if_addrlen); 778 goto drop; 779 } 780 (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen); 781 la->la_flags |= LLE_VALID; 782 783 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED); 784 785 if (!(la->la_flags & LLE_STATIC)) { 786 int canceled; 787 788 LLE_ADDREF(la); 789 la->la_expire = time_uptime + V_arpt_keep; 790 canceled = callout_reset(&la->la_timer, 791 hz * V_arpt_keep, arptimer, la); 792 if (canceled) 793 LLE_REMREF(la); 794 } 795 la->la_asked = 0; 796 la->la_preempt = V_arp_maxtries; 797 /* 798 * The packets are all freed within the call to the output 799 * routine. 800 * 801 * NB: The lock MUST be released before the call to the 802 * output routine. 803 */ 804 if (la->la_hold != NULL) { 805 struct mbuf *m_hold, *m_hold_next; 806 807 m_hold = la->la_hold; 808 la->la_hold = NULL; 809 la->la_numheld = 0; 810 memcpy(&sa, L3_ADDR(la), sizeof(sa)); 811 LLE_WUNLOCK(la); 812 for (; m_hold != NULL; m_hold = m_hold_next) { 813 m_hold_next = m_hold->m_nextpkt; 814 m_hold->m_nextpkt = NULL; 815 /* Avoid confusing lower layers. */ 816 m_clrprotoflags(m_hold); 817 (*ifp->if_output)(ifp, m_hold, &sa, NULL); 818 } 819 } else 820 LLE_WUNLOCK(la); 821 } 822reply: 823 if (op != ARPOP_REQUEST) 824 goto drop; 825 ARPSTAT_INC(rxrequests); 826 827 if (itaddr.s_addr == myaddr.s_addr) { 828 /* Shortcut.. the receiving interface is the target. */ 829 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 830 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 831 } else { 832 struct llentry *lle = NULL; 833 834 sin.sin_addr = itaddr; 835 IF_AFDATA_RLOCK(ifp); 836 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); 837 IF_AFDATA_RUNLOCK(ifp); 838 839 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { 840 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 841 (void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln); 842 LLE_RUNLOCK(lle); 843 } else { 844 845 if (lle != NULL) 846 LLE_RUNLOCK(lle); 847 848 if (!V_arp_proxyall) 849 goto drop; 850 851 sin.sin_addr = itaddr; 852 /* XXX MRT use table 0 for arp reply */ 853 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 854 if (!rt) 855 goto drop; 856 857 /* 858 * Don't send proxies for nodes on the same interface 859 * as this one came out of, or we'll get into a fight 860 * over who claims what Ether address. 861 */ 862 if (!rt->rt_ifp || rt->rt_ifp == ifp) { 863 RTFREE_LOCKED(rt); 864 goto drop; 865 } 866 RTFREE_LOCKED(rt); 867 868 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 869 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 870 871 /* 872 * Also check that the node which sent the ARP packet 873 * is on the interface we expect it to be on. This 874 * avoids ARP chaos if an interface is connected to the 875 * wrong network. 876 */ 877 sin.sin_addr = isaddr; 878 879 /* XXX MRT use table 0 for arp checks */ 880 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 881 if (!rt) 882 goto drop; 883 if (rt->rt_ifp != ifp) { 884 ARP_LOG(LOG_INFO, "proxy: ignoring request" 885 " from %s via %s, expecting %s\n", 886 inet_ntoa(isaddr), ifp->if_xname, 887 rt->rt_ifp->if_xname); 888 RTFREE_LOCKED(rt); 889 goto drop; 890 } 891 RTFREE_LOCKED(rt); 892 893#ifdef DEBUG_PROXY 894 printf("arp: proxying for %s\n", inet_ntoa(itaddr)); 895#endif 896 } 897 } 898 899 if (itaddr.s_addr == myaddr.s_addr && 900 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 901 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 902#ifdef DEBUG_LINKLOCAL 903 printf("arp: sending reply for link-local addr %s\n", 904 inet_ntoa(itaddr)); 905#endif 906 m->m_flags |= M_BCAST; 907 m->m_flags &= ~M_MCAST; 908 } else { 909 /* default behaviour; never reply by broadcast. */ 910 m->m_flags &= ~(M_BCAST|M_MCAST); 911 } 912 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 913 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 914 ah->ar_op = htons(ARPOP_REPLY); 915 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 916 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 917 m->m_pkthdr.len = m->m_len; 918 m->m_pkthdr.rcvif = NULL; 919 sa.sa_family = AF_ARP; 920 sa.sa_len = 2; 921 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 922 (*ifp->if_output)(ifp, m, &sa, NULL); 923 ARPSTAT_INC(txreplies); 924 return; 925 926drop: 927 m_freem(m); 928} 929#endif 930 931void 932arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 933{ 934 struct llentry *lle; 935 936 if (ifa->ifa_carp != NULL) 937 return; 938 939 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) { 940 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 941 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 942 /* 943 * interface address is considered static entry 944 * because the output of the arp utility shows 945 * that L2 entry as permanent 946 */ 947 IF_AFDATA_LOCK(ifp); 948 lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC), 949 (struct sockaddr *)IA_SIN(ifa)); 950 IF_AFDATA_UNLOCK(ifp); 951 if (lle == NULL) 952 log(LOG_INFO, "arp_ifinit: cannot create arp " 953 "entry for interface address\n"); 954 else 955 LLE_RUNLOCK(lle); 956 } 957 ifa->ifa_rtrequest = NULL; 958} 959 960void 961arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 962{ 963 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 964 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 965 &IA_SIN(ifa)->sin_addr, enaddr); 966 ifa->ifa_rtrequest = NULL; 967} 968 969static void 970arp_init(void) 971{ 972 973 netisr_register(&arp_nh); 974} 975SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 976