if_ether.c revision 183954
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: head/sys/netinet/if_ether.c 183954 2008-10-16 12:31:03Z zec $"); 40 41#include "opt_inet.h" 42#include "opt_mac.h" 43#include "opt_carp.h" 44 45#include <sys/param.h> 46#include <sys/kernel.h> 47#include <sys/queue.h> 48#include <sys/sysctl.h> 49#include <sys/systm.h> 50#include <sys/mbuf.h> 51#include <sys/malloc.h> 52#include <sys/proc.h> 53#include <sys/socket.h> 54#include <sys/syslog.h> 55#include <sys/vimage.h> 56 57#include <net/if.h> 58#include <net/if_dl.h> 59#include <net/if_types.h> 60#include <net/route.h> 61#include <net/netisr.h> 62#include <net/if_llc.h> 63#include <net/ethernet.h> 64 65#include <netinet/in.h> 66#include <netinet/in_var.h> 67#include <netinet/if_ether.h> 68 69#include <net/if_arc.h> 70#include <net/iso88025.h> 71 72#ifdef DEV_CARP 73#include <netinet/ip_carp.h> 74#endif 75 76#include <security/mac/mac_framework.h> 77 78#define SIN(s) ((struct sockaddr_in *)s) 79#define SDL(s) ((struct sockaddr_dl *)s) 80 81SYSCTL_DECL(_net_link_ether); 82SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 83 84/* timer values */ 85static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 86 87SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 88 &arpt_keep, 0, "ARP entry lifetime in seconds"); 89 90#define rt_expire rt_rmx.rmx_expire 91 92struct llinfo_arp { 93 struct callout la_timer; 94 struct rtentry *la_rt; 95 struct mbuf *la_hold; /* last packet until resolved/timeout */ 96 u_short la_preempt; /* countdown for pre-expiry arps */ 97 u_short la_asked; /* # requests sent */ 98}; 99 100static struct ifqueue arpintrq; 101 102static int arp_maxtries = 5; 103static int useloopback = 1; /* use loopback interface for local traffic */ 104static int arp_proxyall = 0; 105 106SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries, 107 CTLFLAG_RW, arp_maxtries, 0, 108 "ARP resolution attempts before returning error"); 109SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, useloopback, 110 CTLFLAG_RW, useloopback, 0, 111 "Use the loopback interface for local traffic"); 112SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall, 113 CTLFLAG_RW, arp_proxyall, 0, 114 "Enable proxy ARP for all suitable requests"); 115 116static void arp_init(void); 117static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 118static void arprequest(struct ifnet *, 119 struct in_addr *, struct in_addr *, u_char *); 120static void arpintr(struct mbuf *); 121static void arptimer(void *); 122static struct rtentry 123 *arplookup(u_long, int, int, int); 124#ifdef INET 125static void in_arpinput(struct mbuf *); 126#endif 127 128/* 129 * Timeout routine. 130 */ 131static void 132arptimer(void *arg) 133{ 134 struct rtentry *rt = (struct rtentry *)arg; 135 136 RT_LOCK_ASSERT(rt); 137 /* 138 * The lock is needed to close a theoretical race 139 * between spontaneous expiry and intentional removal. 140 * We still got an extra reference on rtentry, so can 141 * safely pass pointers to its contents. 142 */ 143 RT_UNLOCK(rt); 144 145 in_rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL, 146 rt->rt_fibnum); 147} 148 149/* 150 * Parallel to llc_rtrequest. 151 */ 152static void 153arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 154{ 155 INIT_VNET_NET(curvnet); 156 INIT_VNET_INET(curvnet); 157 struct sockaddr *gate; 158 struct llinfo_arp *la; 159 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 160 struct in_ifaddr *ia; 161 struct ifaddr *ifa; 162 163 RT_LOCK_ASSERT(rt); 164 165 if (rt->rt_flags & RTF_GATEWAY) 166 return; 167 gate = rt->rt_gateway; 168 la = (struct llinfo_arp *)rt->rt_llinfo; 169 switch (req) { 170 171 case RTM_ADD: 172 /* 173 * XXX: If this is a manually added route to interface 174 * such as older version of routed or gated might provide, 175 * restore cloning bit. 176 */ 177 if ((rt->rt_flags & RTF_HOST) == 0 && 178 rt_mask(rt) != NULL && 179 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 180 rt->rt_flags |= RTF_CLONING; 181 if (rt->rt_flags & RTF_CLONING) { 182 /* 183 * Case 1: This route should come from a route to iface. 184 */ 185 rt_setgate(rt, rt_key(rt), 186 (struct sockaddr *)&null_sdl); 187 gate = rt->rt_gateway; 188 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 189 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 190 rt->rt_expire = time_uptime; 191 break; 192 } 193 /* Announce a new entry if requested. */ 194 if (rt->rt_flags & RTF_ANNOUNCE) 195 arprequest(rt->rt_ifp, 196 &SIN(rt_key(rt))->sin_addr, 197 &SIN(rt_key(rt))->sin_addr, 198 (u_char *)LLADDR(SDL(gate))); 199 /*FALLTHROUGH*/ 200 case RTM_RESOLVE: 201 if (gate->sa_family != AF_LINK || 202 gate->sa_len < sizeof(null_sdl)) { 203 log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, 204 inet_ntoa(SIN(rt_key(rt))->sin_addr), 205 (gate->sa_family != AF_LINK) ? 206 " (!AF_LINK)": ""); 207 break; 208 } 209 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 210 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 211 if (la != 0) 212 break; /* This happens on a route change */ 213 /* 214 * Case 2: This route may come from cloning, or a manual route 215 * add with a LL address. 216 */ 217 R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); 218 rt->rt_llinfo = (caddr_t)la; 219 if (la == 0) { 220 log(LOG_DEBUG, "%s: malloc failed\n", __func__); 221 break; 222 } 223 /* 224 * We are storing a route entry outside of radix tree. So, 225 * it can be found and accessed by other means than radix 226 * lookup. The routing code assumes that any rtentry detached 227 * from radix can be destroyed safely. To prevent this, we 228 * add an additional reference. 229 */ 230 RT_ADDREF(rt); 231 la->la_rt = rt; 232 rt->rt_flags |= RTF_LLINFO; 233 callout_init_mtx(&la->la_timer, &rt->rt_mtx, 234 CALLOUT_RETURNUNLOCKED); 235 236#ifdef INET 237 /* 238 * This keeps the multicast addresses from showing up 239 * in `arp -a' listings as unresolved. It's not actually 240 * functional. Then the same for broadcast. 241 */ 242 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && 243 rt->rt_ifp->if_type != IFT_ARCNET) { 244 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 245 LLADDR(SDL(gate))); 246 SDL(gate)->sdl_alen = 6; 247 rt->rt_expire = 0; 248 } 249 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 250 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 251 rt->rt_ifp->if_addrlen); 252 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 253 rt->rt_expire = 0; 254 } 255#endif 256 257 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 258 if (ia->ia_ifp == rt->rt_ifp && 259 SIN(rt_key(rt))->sin_addr.s_addr == 260 (IA_SIN(ia))->sin_addr.s_addr) 261 break; 262 } 263 if (ia) { 264 /* 265 * This test used to be 266 * if (loif.if_flags & IFF_UP) 267 * It allowed local traffic to be forced 268 * through the hardware by configuring the loopback down. 269 * However, it causes problems during network configuration 270 * for boards that can't receive packets they send. 271 * It is now necessary to clear "useloopback" and remove 272 * the route to force traffic out to the hardware. 273 */ 274 rt->rt_expire = 0; 275 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 276 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 277 if (V_useloopback) { 278 rt->rt_ifp = V_loif; 279 rt->rt_rmx.rmx_mtu = V_loif->if_mtu; 280 } 281 282 /* 283 * make sure to set rt->rt_ifa to the interface 284 * address we are using, otherwise we will have trouble 285 * with source address selection. 286 */ 287 ifa = &ia->ia_ifa; 288 if (ifa != rt->rt_ifa) { 289 IFAFREE(rt->rt_ifa); 290 IFAREF(ifa); 291 rt->rt_ifa = ifa; 292 } 293 } 294 break; 295 296 case RTM_DELETE: 297 if (la == NULL) /* XXX: at least CARP does this. */ 298 break; 299 callout_stop(&la->la_timer); 300 rt->rt_llinfo = NULL; 301 rt->rt_flags &= ~RTF_LLINFO; 302 RT_REMREF(rt); 303 if (la->la_hold) 304 m_freem(la->la_hold); 305 Free((caddr_t)la); 306 } 307} 308 309/* 310 * Broadcast an ARP request. Caller specifies: 311 * - arp header source ip address 312 * - arp header target ip address 313 * - arp header source ethernet address 314 */ 315static void 316arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 317 u_char *enaddr) 318{ 319 struct mbuf *m; 320 struct arphdr *ah; 321 struct sockaddr sa; 322 323 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 324 return; 325 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 326 2*ifp->if_data.ifi_addrlen; 327 m->m_pkthdr.len = m->m_len; 328 MH_ALIGN(m, m->m_len); 329 ah = mtod(m, struct arphdr *); 330 bzero((caddr_t)ah, m->m_len); 331#ifdef MAC 332 mac_netinet_arp_send(ifp, m); 333#endif 334 ah->ar_pro = htons(ETHERTYPE_IP); 335 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 336 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 337 ah->ar_op = htons(ARPOP_REQUEST); 338 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 339 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 340 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 341 sa.sa_family = AF_ARP; 342 sa.sa_len = 2; 343 m->m_flags |= M_BCAST; 344 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 345 346 return; 347} 348 349/* 350 * Resolve an IP address into an ethernet address. 351 * On input: 352 * ifp is the interface we use 353 * rt0 is the route to the final destination (possibly useless) 354 * m is the mbuf. May be NULL if we don't have a packet. 355 * dst is the next hop, 356 * desten is where we want the address. 357 * 358 * On success, desten is filled in and the function returns 0; 359 * If the packet must be held pending resolution, we return EWOULDBLOCK 360 * On other errors, we return the corresponding error code. 361 * Note that m_freem() handles NULL. 362 */ 363int 364arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 365 struct sockaddr *dst, u_char *desten) 366{ 367 INIT_VNET_INET(ifp->if_vnet); 368 struct llinfo_arp *la = NULL; 369 struct rtentry *rt = NULL; 370 struct sockaddr_dl *sdl; 371 int error; 372 int fibnum = -1; 373 374 if (m) { 375 if (m->m_flags & M_BCAST) { 376 /* broadcast */ 377 (void)memcpy(desten, 378 ifp->if_broadcastaddr, ifp->if_addrlen); 379 return (0); 380 } 381 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) { 382 /* multicast */ 383 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 384 return (0); 385 } 386 fibnum = M_GETFIB(m); 387 } 388 389 if (rt0 != NULL) { 390 /* Look for a cached arp (ll) entry. */ 391 if (m == NULL) 392 fibnum = rt0->rt_fibnum; 393 error = rt_check(&rt, &rt0, dst); 394 if (error) { 395 m_freem(m); 396 return error; 397 } 398 la = (struct llinfo_arp *)rt->rt_llinfo; 399 if (la == NULL) 400 RT_UNLOCK(rt); 401 } 402 403 /* 404 * If we had no mbuf and no route, then hope the caller 405 * has a fib in mind because we are running out of ideas. 406 * I think this should not happen in current code. 407 * (kmacy would know). 408 */ 409 if (fibnum == -1) 410 fibnum = curthread->td_proc->p_fibnum; /* last gasp */ 411 412 if (la == NULL) { 413 /* 414 * We enter this block if rt0 was NULL, 415 * or if rt found by rt_check() didn't have llinfo. 416 * we should get a cloned route, which since it should 417 * come from the local interface should have a ll entry. 418 * It may be incomplete but that's ok. 419 */ 420 rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0, fibnum); 421 if (rt == NULL) { 422 log(LOG_DEBUG, 423 "arpresolve: can't allocate route for %s\n", 424 inet_ntoa(SIN(dst)->sin_addr)); 425 m_freem(m); 426 return (EINVAL); /* XXX */ 427 } 428 la = (struct llinfo_arp *)rt->rt_llinfo; 429 if (la == NULL) { 430 RT_UNLOCK(rt); 431 log(LOG_DEBUG, 432 "arpresolve: can't allocate llinfo for %s\n", 433 inet_ntoa(SIN(dst)->sin_addr)); 434 m_freem(m); 435 return (EINVAL); /* XXX */ 436 } 437 } 438 sdl = SDL(rt->rt_gateway); 439 /* 440 * Check the address family and length is valid, the address 441 * is resolved; otherwise, try to resolve. 442 */ 443 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) && 444 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 445 446 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 447 448 /* 449 * If entry has an expiry time and it is approaching, 450 * send an ARP request. 451 */ 452 if ((rt->rt_expire != 0) && 453 (time_uptime + la->la_preempt > rt->rt_expire)) { 454 struct in_addr sin = 455 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 456 457 la->la_preempt--; 458 RT_UNLOCK(rt); 459 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 460 IF_LLADDR(ifp)); 461 return (0); 462 } 463 464 RT_UNLOCK(rt); 465 return (0); 466 } 467 /* 468 * If ARP is disabled or static on this interface, stop. 469 * XXX 470 * Probably should not allocate empty llinfo struct if we are 471 * not going to be sending out an arp request. 472 */ 473 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 474 RT_UNLOCK(rt); 475 m_freem(m); 476 return (EINVAL); 477 } 478 /* 479 * There is an arptab entry, but no ethernet address 480 * response yet. Replace the held mbuf with this 481 * latest one. 482 */ 483 if (m) { 484 if (la->la_hold) 485 m_freem(la->la_hold); 486 la->la_hold = m; 487 } 488 KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry")); 489 490 /* 491 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 492 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 493 * if we have already sent arp_maxtries ARP requests. Retransmit the 494 * ARP request, but not faster than one request per second. 495 */ 496 if (la->la_asked < V_arp_maxtries) 497 error = EWOULDBLOCK; /* First request. */ 498 else 499 error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH; 500 501 if (la->la_asked == 0 || rt->rt_expire != time_uptime) { 502 struct in_addr sin = 503 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 504 505 rt->rt_expire = time_uptime; 506 callout_reset(&la->la_timer, hz, arptimer, rt); 507 la->la_asked++; 508 RT_UNLOCK(rt); 509 510 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 511 IF_LLADDR(ifp)); 512 } else 513 RT_UNLOCK(rt); 514 515 return (error); 516} 517 518/* 519 * Common length and type checks are done here, 520 * then the protocol-specific routine is called. 521 */ 522static void 523arpintr(struct mbuf *m) 524{ 525 struct arphdr *ar; 526 527 if (m->m_len < sizeof(struct arphdr) && 528 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 529 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 530 return; 531 } 532 ar = mtod(m, struct arphdr *); 533 534 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 535 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 536 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 537 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) { 538 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 539 (unsigned char *)&ar->ar_hrd, ""); 540 m_freem(m); 541 return; 542 } 543 544 if (m->m_len < arphdr_len(ar)) { 545 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 546 log(LOG_ERR, "arp: runt packet\n"); 547 m_freem(m); 548 return; 549 } 550 ar = mtod(m, struct arphdr *); 551 } 552 553 switch (ntohs(ar->ar_pro)) { 554#ifdef INET 555 case ETHERTYPE_IP: 556 in_arpinput(m); 557 return; 558#endif 559 } 560 m_freem(m); 561} 562 563#ifdef INET 564/* 565 * ARP for Internet protocols on 10 Mb/s Ethernet. 566 * Algorithm is that given in RFC 826. 567 * In addition, a sanity check is performed on the sender 568 * protocol address, to catch impersonators. 569 * We no longer handle negotiations for use of trailer protocol: 570 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 571 * along with IP replies if we wanted trailers sent to us, 572 * and also sent them in response to IP replies. 573 * This allowed either end to announce the desire to receive 574 * trailer packets. 575 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 576 * but formerly didn't normally send requests. 577 */ 578static int log_arp_wrong_iface = 1; 579static int log_arp_movements = 1; 580static int log_arp_permanent_modify = 1; 581 582SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 583 &log_arp_wrong_iface, 0, 584 "log arp packets arriving on the wrong interface"); 585SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 586 &log_arp_movements, 0, 587 "log arp replies from MACs different than the one in the cache"); 588SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 589 &log_arp_permanent_modify, 0, 590 "log arp replies from MACs different than the one in the permanent arp entry"); 591 592 593static void 594in_arpinput(struct mbuf *m) 595{ 596 struct arphdr *ah; 597 struct ifnet *ifp = m->m_pkthdr.rcvif; 598 struct llinfo_arp *la; 599 struct rtentry *rt; 600 struct ifaddr *ifa; 601 struct in_ifaddr *ia; 602 struct sockaddr_dl *sdl; 603 struct sockaddr sa; 604 struct in_addr isaddr, itaddr, myaddr; 605 struct mbuf *hold; 606 u_int8_t *enaddr = NULL; 607 int op, rif_len; 608 int req_len; 609 int bridged = 0, is_bridge = 0; 610 u_int fibnum; 611 u_int goodfib = 0; 612 int firstpass = 1; 613#ifdef DEV_CARP 614 int carp_match = 0; 615#endif 616 struct sockaddr_in sin; 617 sin.sin_len = sizeof(struct sockaddr_in); 618 sin.sin_family = AF_INET; 619 sin.sin_addr.s_addr = 0; 620 INIT_VNET_INET(ifp->if_vnet); 621 622 if (ifp->if_bridge) 623 bridged = 1; 624 if (ifp->if_type == IFT_BRIDGE) 625 is_bridge = 1; 626 627 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 628 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 629 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 630 return; 631 } 632 633 ah = mtod(m, struct arphdr *); 634 op = ntohs(ah->ar_op); 635 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 636 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 637 638 /* 639 * For a bridge, we want to check the address irrespective 640 * of the receive interface. (This will change slightly 641 * when we have clusters of interfaces). 642 * If the interface does not match, but the recieving interface 643 * is part of carp, we call carp_iamatch to see if this is a 644 * request for the virtual host ip. 645 * XXX: This is really ugly! 646 */ 647 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 648 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 649 (ia->ia_ifp == ifp)) && 650 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 651 goto match; 652#ifdef DEV_CARP 653 if (ifp->if_carp != NULL && 654 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 655 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 656 carp_match = 1; 657 goto match; 658 } 659#endif 660 } 661 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 662 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 663 (ia->ia_ifp == ifp)) && 664 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 665 goto match; 666 667#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 668 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 669 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 670 addr == ia->ia_addr.sin_addr.s_addr) 671 /* 672 * Check the case when bridge shares its MAC address with 673 * some of its children, so packets are claimed by bridge 674 * itself (bridge_input() does it first), but they are really 675 * meant to be destined to the bridge member. 676 */ 677 if (is_bridge) { 678 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 679 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 680 ifp = ia->ia_ifp; 681 goto match; 682 } 683 } 684 } 685#undef BDG_MEMBER_MATCHES_ARP 686 687 /* 688 * No match, use the first inet address on the receive interface 689 * as a dummy address for the rest of the function. 690 */ 691 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 692 if (ifa->ifa_addr->sa_family == AF_INET) { 693 ia = ifatoia(ifa); 694 goto match; 695 } 696 /* 697 * If bridging, fall back to using any inet address. 698 */ 699 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) 700 goto drop; 701match: 702 if (!enaddr) 703 enaddr = (u_int8_t *)IF_LLADDR(ifp); 704 myaddr = ia->ia_addr.sin_addr; 705 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 706 goto drop; /* it's from me, ignore it. */ 707 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 708 log(LOG_ERR, 709 "arp: link address is broadcast for IP address %s!\n", 710 inet_ntoa(isaddr)); 711 goto drop; 712 } 713 /* 714 * Warn if another host is using the same IP address, but only if the 715 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 716 * case we suppress the warning to avoid false positive complaints of 717 * potential misconfiguration. 718 */ 719 if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 720 log(LOG_ERR, 721 "arp: %*D is using my IP address %s on %s!\n", 722 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 723 inet_ntoa(isaddr), ifp->if_xname); 724 itaddr = myaddr; 725 goto reply; 726 } 727 if (ifp->if_flags & IFF_STATICARP) 728 goto reply; 729 /* 730 * We look for any FIB that has this address to find 731 * the interface etc. 732 * For sanity checks that are FIB independent we abort the loop. 733 */ 734 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { 735 rt = arplookup(isaddr.s_addr, 736 itaddr.s_addr == myaddr.s_addr, 0, fibnum); 737 if (rt == NULL) 738 continue; 739 740 sdl = SDL(rt->rt_gateway); 741 /* Only call this once */ 742 if (firstpass) { 743 sin.sin_addr.s_addr = isaddr.s_addr; 744 EVENTHANDLER_INVOKE(route_arp_update_event, rt, 745 ar_sha(ah), (struct sockaddr *)&sin); 746 } 747 748 la = (struct llinfo_arp *)rt->rt_llinfo; 749 if (la == NULL) { 750 RT_UNLOCK(rt); 751 continue; 752 } 753 754 if (firstpass) { 755 /* The following is not an error when doing bridging. */ 756 if (!bridged && rt->rt_ifp != ifp 757#ifdef DEV_CARP 758 && (ifp->if_type != IFT_CARP || !carp_match) 759#endif 760 ) { 761 if (log_arp_wrong_iface) 762 log(LOG_ERR, "arp: %s is on %s " 763 "but got reply from %*D " 764 "on %s\n", 765 inet_ntoa(isaddr), 766 rt->rt_ifp->if_xname, 767 ifp->if_addrlen, 768 (u_char *)ar_sha(ah), ":", 769 ifp->if_xname); 770 RT_UNLOCK(rt); 771 break; 772 } 773 if (sdl->sdl_alen && 774 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 775 if (rt->rt_expire) { 776 if (log_arp_movements) 777 log(LOG_INFO, 778 "arp: %s moved from %*D to %*D " 779 "on %s\n", 780 inet_ntoa(isaddr), 781 ifp->if_addrlen, 782 (u_char *)LLADDR(sdl), ":", 783 ifp->if_addrlen, 784 (u_char *)ar_sha(ah), ":", 785 ifp->if_xname); 786 } else { 787 RT_UNLOCK(rt); 788 if (log_arp_permanent_modify) 789 log(LOG_ERR, 790 "arp: %*D attempts to " 791 "modify permanent entry " 792 "for %s on %s\n", 793 ifp->if_addrlen, 794 (u_char *)ar_sha(ah), ":", 795 inet_ntoa(isaddr), 796 ifp->if_xname); 797 break; 798 } 799 } 800 /* 801 * sanity check for the address length. 802 * XXX this does not work for protocols 803 * with variable address length. -is 804 */ 805 if (sdl->sdl_alen && 806 sdl->sdl_alen != ah->ar_hln) { 807 log(LOG_WARNING, 808 "arp from %*D: new addr len %d, was %d", 809 ifp->if_addrlen, (u_char *) ar_sha(ah), 810 ":", ah->ar_hln, sdl->sdl_alen); 811 } 812 if (ifp->if_addrlen != ah->ar_hln) { 813 log(LOG_WARNING, 814 "arp from %*D: addr len: " 815 "new %d, i/f %d (ignored)", 816 ifp->if_addrlen, (u_char *) ar_sha(ah), 817 ":", ah->ar_hln, ifp->if_addrlen); 818 RT_UNLOCK(rt); 819 break; 820 } 821 firstpass = 0; 822 goodfib = fibnum; 823 } 824 825 /* Copy in the information received. */ 826 (void)memcpy(LLADDR(sdl), ar_sha(ah), 827 sdl->sdl_alen = ah->ar_hln); 828 /* 829 * If we receive an arp from a token-ring station over 830 * a token-ring nic then try to save the source routing info. 831 * XXXMRT Only minimal Token Ring support for MRT. 832 * Only do this on the first pass as if modifies the mbuf. 833 */ 834 if (ifp->if_type == IFT_ISO88025) { 835 struct iso88025_header *th = NULL; 836 struct iso88025_sockaddr_dl_data *trld; 837 838 /* force the fib loop to end after this pass */ 839 fibnum = rt_numfibs - 1; 840 841 th = (struct iso88025_header *)m->m_pkthdr.header; 842 trld = SDL_ISO88025(sdl); 843 rif_len = TR_RCF_RIFLEN(th->rcf); 844 if ((th->iso88025_shost[0] & TR_RII) && 845 (rif_len > 2)) { 846 trld->trld_rcf = th->rcf; 847 trld->trld_rcf ^= htons(TR_RCF_DIR); 848 memcpy(trld->trld_route, th->rd, rif_len - 2); 849 trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); 850 /* 851 * Set up source routing information for 852 * reply packet (XXX) 853 */ 854 m->m_data -= rif_len; 855 m->m_len += rif_len; 856 m->m_pkthdr.len += rif_len; 857 } else { 858 th->iso88025_shost[0] &= ~TR_RII; 859 trld->trld_rcf = 0; 860 } 861 m->m_data -= 8; 862 m->m_len += 8; 863 m->m_pkthdr.len += 8; 864 th->rcf = trld->trld_rcf; 865 } 866 867 if (rt->rt_expire) { 868 rt->rt_expire = time_uptime + V_arpt_keep; 869 callout_reset(&la->la_timer, hz * V_arpt_keep, 870 arptimer, rt); 871 } 872 la->la_asked = 0; 873 la->la_preempt = V_arp_maxtries; 874 hold = la->la_hold; 875 la->la_hold = NULL; 876 RT_UNLOCK(rt); 877 if (hold != NULL) 878 (*ifp->if_output)(ifp, hold, rt_key(rt), rt); 879 } /* end of FIB loop */ 880reply: 881 882 /* 883 * Decide if we have to respond to something. 884 */ 885 if (op != ARPOP_REQUEST) 886 goto drop; 887 if (itaddr.s_addr == myaddr.s_addr) { 888 /* Shortcut.. the receiving interface is the target. */ 889 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 890 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 891 } else { 892 /* It's not asking for our address. But it still may 893 * be something we should answer. 894 * 895 * XXX MRT 896 * We assume that link level info is independent of 897 * the table used and so we use whichever we can and don't 898 * have a better option. 899 */ 900 /* Have we been asked to proxy for the target. */ 901 rt = arplookup(itaddr.s_addr, 0, SIN_PROXY, goodfib); 902 if (rt == NULL) { 903 /* Nope, only intersted now if proxying everything. */ 904 struct sockaddr_in sin; 905 906 if (!V_arp_proxyall) 907 goto drop; 908 909 bzero(&sin, sizeof sin); 910 sin.sin_family = AF_INET; 911 sin.sin_len = sizeof sin; 912 sin.sin_addr = itaddr; 913 914 /* XXX MRT use table 0 for arp reply */ 915 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 916 if (!rt) 917 goto drop; 918 /* 919 * Don't send proxies for nodes on the same interface 920 * as this one came out of, or we'll get into a fight 921 * over who claims what Ether address. 922 */ 923 if (rt->rt_ifp == ifp) { 924 rtfree(rt); 925 goto drop; 926 } 927 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 928 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 929 rtfree(rt); 930 931 /* 932 * Also check that the node which sent the ARP packet 933 * is on the the interface we expect it to be on. This 934 * avoids ARP chaos if an interface is connected to the 935 * wrong network. 936 */ 937 sin.sin_addr = isaddr; 938 939 /* XXX MRT use table 0 for arp checks */ 940 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 941 if (!rt) 942 goto drop; 943 if (rt->rt_ifp != ifp) { 944 log(LOG_INFO, "arp_proxy: ignoring request" 945 " from %s via %s, expecting %s\n", 946 inet_ntoa(isaddr), ifp->if_xname, 947 rt->rt_ifp->if_xname); 948 rtfree(rt); 949 goto drop; 950 } 951 rtfree(rt); 952 953#ifdef DEBUG_PROXY 954 printf("arp: proxying for %s\n", 955 inet_ntoa(itaddr)); 956#endif 957 } else { 958 /* 959 * Return proxied ARP replies only on the interface 960 * or bridge cluster where this network resides. 961 * Otherwise we may conflict with the host we are 962 * proxying for. 963 */ 964 if (rt->rt_ifp != ifp && 965 (rt->rt_ifp->if_bridge != ifp->if_bridge || 966 ifp->if_bridge == NULL)) { 967 RT_UNLOCK(rt); 968 goto drop; 969 } 970 sdl = SDL(rt->rt_gateway); 971 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 972 (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 973 RT_UNLOCK(rt); 974 } 975 } 976 977 if (itaddr.s_addr == myaddr.s_addr && 978 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 979 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 980#ifdef DEBUG_LINKLOCAL 981 printf("arp: sending reply for link-local addr %s\n", 982 inet_ntoa(itaddr)); 983#endif 984 m->m_flags |= M_BCAST; 985 m->m_flags &= ~M_MCAST; 986 } else { 987 /* default behaviour; never reply by broadcast. */ 988 m->m_flags &= ~(M_BCAST|M_MCAST); 989 } 990 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 991 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 992 ah->ar_op = htons(ARPOP_REPLY); 993 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 994 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 995 m->m_pkthdr.len = m->m_len; 996 sa.sa_family = AF_ARP; 997 sa.sa_len = 2; 998 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 999 return; 1000 1001drop: 1002 m_freem(m); 1003} 1004#endif 1005 1006/* 1007 * Lookup or enter a new address in arptab. 1008 */ 1009static struct rtentry * 1010arplookup(u_long addr, int create, int proxy, int fibnum) 1011{ 1012 struct rtentry *rt; 1013 struct sockaddr_inarp sin; 1014 const char *why = 0; 1015 1016 bzero(&sin, sizeof(sin)); 1017 sin.sin_len = sizeof(sin); 1018 sin.sin_family = AF_INET; 1019 sin.sin_addr.s_addr = addr; 1020 if (proxy) 1021 sin.sin_other = SIN_PROXY; 1022 rt = in_rtalloc1((struct sockaddr *)&sin, create, 0UL, fibnum); 1023 if (rt == 0) 1024 return (0); 1025 1026 if (rt->rt_flags & RTF_GATEWAY) 1027 why = "host is not on local network"; 1028 else if ((rt->rt_flags & RTF_LLINFO) == 0) 1029 why = "could not allocate llinfo"; 1030 else if (rt->rt_gateway->sa_family != AF_LINK) 1031 why = "gateway route is not ours"; 1032 1033 if (why) { 1034#define ISDYNCLONE(_rt) \ 1035 (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) 1036 if (create) 1037 log(LOG_DEBUG, "arplookup %s failed: %s\n", 1038 inet_ntoa(sin.sin_addr), why); 1039 /* 1040 * If there are no references to this Layer 2 route, 1041 * and it is a cloned route, and not static, and 1042 * arplookup() is creating the route, then purge 1043 * it from the routing table as it is probably bogus. 1044 */ 1045 if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) 1046 rtexpunge(rt); 1047 RTFREE_LOCKED(rt); 1048 return (0); 1049#undef ISDYNCLONE 1050 } else { 1051 RT_REMREF(rt); 1052 return (rt); 1053 } 1054} 1055 1056void 1057arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1058{ 1059 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 1060 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 1061 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 1062 ifa->ifa_rtrequest = arp_rtrequest; 1063 ifa->ifa_flags |= RTF_CLONING; 1064} 1065 1066void 1067arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 1068{ 1069 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 1070 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 1071 &IA_SIN(ifa)->sin_addr, enaddr); 1072 ifa->ifa_rtrequest = arp_rtrequest; 1073 ifa->ifa_flags |= RTF_CLONING; 1074} 1075 1076static void 1077arp_init(void) 1078{ 1079 1080 arpintrq.ifq_maxlen = 50; 1081 mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF); 1082 netisr_register(NETISR_ARP, arpintr, &arpintrq, 0); 1083} 1084SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 1085