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 * $FreeBSD$ 31 */ 32 33 /* 34 * Ethernet address resolution protocol. 35 * TODO: 36 * add "inuse/lock" bit (or ref. count) along with valid bit 37 */ 38 39 #include "opt_inet.h" 40 #include "opt_mac.h" 41 #include "opt_carp.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/mac.h> 49 #include <sys/mbuf.h> 50 #include <sys/malloc.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/route.h> 58 #include <net/netisr.h> 59 #include <net/if_llc.h> 60 #include <net/ethernet.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_var.h> 64 #include <netinet/if_ether.h> 65 66 #include <net/if_arc.h> 67 #include <net/iso88025.h> 68 69 #ifdef DEV_CARP 70 #include <netinet/ip_carp.h> 71 #endif 72 73 #define SIN(s) ((struct sockaddr_in *)s) 74 #define SDL(s) ((struct sockaddr_dl *)s) 75 76 SYSCTL_DECL(_net_link_ether); 77 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 78 79 /* timer values */ 80 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 81 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 82 83 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW, 84 &arpt_prune, 0, "ARP table prune interval in seconds"); 85 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 86 &arpt_keep, 0, "ARP entry lifetime in seconds"); 87 88 #define rt_expire rt_rmx.rmx_expire 89 90 struct llinfo_arp { 91 LIST_ENTRY(llinfo_arp) la_le; 92 struct rtentry *la_rt; 93 struct mbuf *la_hold; /* last packet until resolved/timeout */ 94 u_short la_preempt; /* countdown for pre-expiry arps */ 95 u_short la_asked; /* # requests sent */ 96 }; 97 98 static LIST_HEAD(, llinfo_arp) llinfo_arp; 99 100 static struct ifqueue arpintrq; 101 static int arp_allocated; 102 103 static int arp_maxtries = 5; 104 static int useloopback = 1; /* use loopback interface for local traffic */ 105 static int arp_proxyall = 0; 106 static struct callout arp_callout; 107 108 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 109 &arp_maxtries, 0, "ARP resolution attempts before returning error"); 110 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 111 &useloopback, 0, "Use the loopback interface for local traffic"); 112 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 113 &arp_proxyall, 0, "Enable proxy ARP for all suitable requests"); 114 115 static void arp_init(void); 116 static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 117 static void arprequest(struct ifnet *, 118 struct in_addr *, struct in_addr *, u_char *); 119 static void arpintr(struct mbuf *); 120 static void arptimer(void *); 121 static struct rtentry 122 *arplookup(u_long, int, int); 123 #ifdef INET 124 static void in_arpinput(struct mbuf *); 125 #endif 126 127 /* 128 * Timeout routine. Age arp_tab entries periodically. 129 */ 130 /* ARGSUSED */ 131 static void 132 arptimer(void * __unused unused) 133 { 134 struct llinfo_arp *la, *ola; 135 136 RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]); 137 LIST_FOREACH_SAFE(la, &llinfo_arp, la_le, ola) { 138 struct rtentry *rt = la->la_rt; 139 140 RT_LOCK(rt); 141 if (rt->rt_expire && rt->rt_expire <= time_uptime) { 142 struct sockaddr_dl *sdl = SDL(rt->rt_gateway); 143 144 KASSERT(sdl->sdl_family == AF_LINK, ("sdl_family %d", 145 sdl->sdl_family)); 146 if (rt->rt_refcnt > 1) { 147 sdl->sdl_alen = 0; 148 la->la_preempt = la->la_asked = 0; 149 RT_UNLOCK(rt); 150 continue; 151 } 152 RT_UNLOCK(rt); 153 /* 154 * XXX: LIST_REMOVE() is deep inside rtrequest(). 155 */ 156 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, 157 NULL); 158 continue; 159 } 160 RT_UNLOCK(rt); 161 } 162 RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]); 163 164 callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL); 165 } 166 167 /* 168 * Parallel to llc_rtrequest. 169 */ 170 static void 171 arp_rtrequest(req, rt, info) 172 int req; 173 struct rtentry *rt; 174 struct rt_addrinfo *info; 175 { 176 struct sockaddr *gate; 177 struct llinfo_arp *la; 178 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 179 struct in_ifaddr *ia; 180 struct ifaddr *ifa; 181 182 RT_LOCK_ASSERT(rt); 183 184 if (rt->rt_flags & RTF_GATEWAY) 185 return; 186 gate = rt->rt_gateway; 187 la = (struct llinfo_arp *)rt->rt_llinfo; 188 switch (req) { 189 190 case RTM_ADD: 191 /* 192 * XXX: If this is a manually added route to interface 193 * such as older version of routed or gated might provide, 194 * restore cloning bit. 195 */ 196 if ((rt->rt_flags & RTF_HOST) == 0 && 197 rt_mask(rt) != NULL && 198 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 199 rt->rt_flags |= RTF_CLONING; 200 if (rt->rt_flags & RTF_CLONING) { 201 /* 202 * Case 1: This route should come from a route to iface. 203 */ 204 rt_setgate(rt, rt_key(rt), 205 (struct sockaddr *)&null_sdl); 206 gate = rt->rt_gateway; 207 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 208 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 209 rt->rt_expire = time_uptime; 210 break; 211 } 212 /* Announce a new entry if requested. */ 213 if (rt->rt_flags & RTF_ANNOUNCE) 214 arprequest(rt->rt_ifp, 215 &SIN(rt_key(rt))->sin_addr, 216 &SIN(rt_key(rt))->sin_addr, 217 (u_char *)LLADDR(SDL(gate))); 218 /*FALLTHROUGH*/ 219 case RTM_RESOLVE: 220 if (gate->sa_family != AF_LINK || 221 gate->sa_len < sizeof(null_sdl)) { 222 log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, 223 inet_ntoa(SIN(rt_key(rt))->sin_addr), 224 (gate->sa_family != AF_LINK) ? 225 " (!AF_LINK)": ""); 226 break; 227 } 228 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 229 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 230 if (la != 0) 231 break; /* This happens on a route change */ 232 /* 233 * Case 2: This route may come from cloning, or a manual route 234 * add with a LL address. 235 */ 236 R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); 237 rt->rt_llinfo = (caddr_t)la; 238 if (la == 0) { 239 log(LOG_DEBUG, "%s: malloc failed\n", __func__); 240 break; 241 } 242 arp_allocated++; 243 /* 244 * We are storing a route entry outside of radix tree. So, 245 * it can be found and accessed by other means than radix 246 * lookup. The routing code assumes that any rtentry detached 247 * from radix can be destroyed safely. To prevent this, we 248 * add an additional reference. 249 */ 250 RT_ADDREF(rt); 251 la->la_rt = rt; 252 rt->rt_flags |= RTF_LLINFO; 253 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 254 LIST_INSERT_HEAD(&llinfo_arp, la, la_le); 255 256 #ifdef INET 257 /* 258 * This keeps the multicast addresses from showing up 259 * in `arp -a' listings as unresolved. It's not actually 260 * functional. Then the same for broadcast. 261 */ 262 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && 263 rt->rt_ifp->if_type != IFT_ARCNET) { 264 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 265 LLADDR(SDL(gate))); 266 SDL(gate)->sdl_alen = 6; 267 rt->rt_expire = 0; 268 } 269 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 270 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 271 rt->rt_ifp->if_addrlen); 272 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 273 rt->rt_expire = 0; 274 } 275 #endif 276 277 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 278 if (ia->ia_ifp == rt->rt_ifp && 279 SIN(rt_key(rt))->sin_addr.s_addr == 280 (IA_SIN(ia))->sin_addr.s_addr) 281 break; 282 } 283 if (ia) { 284 /* 285 * This test used to be 286 * if (loif.if_flags & IFF_UP) 287 * It allowed local traffic to be forced 288 * through the hardware by configuring the loopback down. 289 * However, it causes problems during network configuration 290 * for boards that can't receive packets they send. 291 * It is now necessary to clear "useloopback" and remove 292 * the route to force traffic out to the hardware. 293 */ 294 rt->rt_expire = 0; 295 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 296 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 297 if (useloopback) { 298 rt->rt_ifp = loif; 299 rt->rt_rmx.rmx_mtu = loif->if_mtu; 300 } 301 302 /* 303 * make sure to set rt->rt_ifa to the interface 304 * address we are using, otherwise we will have trouble 305 * with source address selection. 306 */ 307 ifa = &ia->ia_ifa; 308 if (ifa != rt->rt_ifa) { 309 IFAFREE(rt->rt_ifa); 310 IFAREF(ifa); 311 rt->rt_ifa = ifa; 312 } 313 } 314 break; 315 316 case RTM_DELETE: 317 if (la == 0) 318 break; 319 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 320 LIST_REMOVE(la, la_le); 321 RT_REMREF(rt); 322 rt->rt_llinfo = 0; 323 rt->rt_flags &= ~RTF_LLINFO; 324 if (la->la_hold) 325 m_freem(la->la_hold); 326 Free((caddr_t)la); 327 } 328 } 329 330 /* 331 * Broadcast an ARP request. Caller specifies: 332 * - arp header source ip address 333 * - arp header target ip address 334 * - arp header source ethernet address 335 */ 336 static void 337 arprequest(ifp, sip, tip, enaddr) 338 struct ifnet *ifp; 339 struct in_addr *sip, *tip; 340 u_char *enaddr; 341 { 342 struct mbuf *m; 343 struct arphdr *ah; 344 struct sockaddr sa; 345 346 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 347 return; 348 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 349 2*ifp->if_data.ifi_addrlen; 350 m->m_pkthdr.len = m->m_len; 351 MH_ALIGN(m, m->m_len); 352 ah = mtod(m, struct arphdr *); 353 bzero((caddr_t)ah, m->m_len); 354 #ifdef MAC 355 mac_create_mbuf_linklayer(ifp, m); 356 #endif 357 ah->ar_pro = htons(ETHERTYPE_IP); 358 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 359 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 360 ah->ar_op = htons(ARPOP_REQUEST); 361 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 362 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 363 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 364 sa.sa_family = AF_ARP; 365 sa.sa_len = 2; 366 m->m_flags |= M_BCAST; 367 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 368 369 return; 370 } 371 372 /* 373 * Resolve an IP address into an ethernet address. 374 * On input: 375 * ifp is the interface we use 376 * dst is the next hop, 377 * rt0 is the route to the final destination (possibly useless) 378 * m is the mbuf 379 * desten is where we want the address. 380 * 381 * On success, desten is filled in and the function returns 0; 382 * If the packet must be held pending resolution, we return EWOULDBLOCK 383 * On other errors, we return the corresponding error code. 384 */ 385 int 386 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 387 struct sockaddr *dst, u_char *desten) 388 { 389 struct llinfo_arp *la = NULL; 390 struct rtentry *rt = NULL; 391 struct sockaddr_dl *sdl; 392 int error; 393 394 if (m->m_flags & M_BCAST) { /* broadcast */ 395 (void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 396 return (0); 397 } 398 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */ 399 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 400 return (0); 401 } 402 403 if (rt0 != NULL) { 404 error = rt_check(&rt, &rt0, dst); 405 if (error) { 406 m_freem(m); 407 return error; 408 } 409 la = (struct llinfo_arp *)rt->rt_llinfo; 410 if (la == NULL) 411 RT_UNLOCK(rt); 412 } 413 if (la == NULL) { 414 /* 415 * We enter this block in case if rt0 was NULL, 416 * or if rt found by rt_check() didn't have llinfo. 417 */ 418 rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0); 419 if (rt == NULL) { 420 log(LOG_DEBUG, 421 "arpresolve: can't allocate route for %s\n", 422 inet_ntoa(SIN(dst)->sin_addr)); 423 m_freem(m); 424 return (EINVAL); /* XXX */ 425 } 426 la = (struct llinfo_arp *)rt->rt_llinfo; 427 if (la == NULL) { 428 RT_UNLOCK(rt); 429 log(LOG_DEBUG, 430 "arpresolve: can't allocate llinfo for %s\n", 431 inet_ntoa(SIN(dst)->sin_addr)); 432 m_freem(m); 433 return (EINVAL); /* XXX */ 434 } 435 } 436 sdl = SDL(rt->rt_gateway); 437 /* 438 * Check the address family and length is valid, the address 439 * is resolved; otherwise, try to resolve. 440 */ 441 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) && 442 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 443 444 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 445 446 /* 447 * If entry has an expiry time and it is approaching, 448 * send an ARP request. 449 */ 450 if ((rt->rt_expire != 0) && 451 (time_uptime + la->la_preempt > rt->rt_expire)) { 452 struct in_addr sin = 453 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 454 455 la->la_preempt--; 456 RT_UNLOCK(rt); 457 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 458 IF_LLADDR(ifp)); 459 return (0); 460 } 461 462 RT_UNLOCK(rt); 463 return (0); 464 } 465 /* 466 * If ARP is disabled or static on this interface, stop. 467 * XXX 468 * Probably should not allocate empty llinfo struct if we are 469 * not going to be sending out an arp request. 470 */ 471 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 472 RT_UNLOCK(rt); 473 m_freem(m); 474 return (EINVAL); 475 } 476 /* 477 * There is an arptab entry, but no ethernet address 478 * response yet. Replace the held mbuf with this 479 * latest one. 480 */ 481 if (la->la_hold) 482 m_freem(la->la_hold); 483 la->la_hold = m; 484 485 KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry")); 486 487 /* 488 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 489 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 490 * if we have already sent arp_maxtries ARP requests. Retransmit the 491 * ARP request, but not faster than one request per second. 492 */ 493 if (la->la_asked < arp_maxtries) 494 error = EWOULDBLOCK; /* First request. */ 495 else 496 error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH; 497 498 if (la->la_asked++ == 0 || rt->rt_expire != time_uptime) { 499 struct in_addr sin = 500 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 501 502 rt->rt_expire = time_uptime; 503 RT_UNLOCK(rt); 504 505 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 506 IF_LLADDR(ifp)); 507 } else 508 RT_UNLOCK(rt); 509 510 return (error); 511 } 512 513 /* 514 * Common length and type checks are done here, 515 * then the protocol-specific routine is called. 516 */ 517 static void 518 arpintr(struct mbuf *m) 519 { 520 struct arphdr *ar; 521 522 if (m->m_len < sizeof(struct arphdr) && 523 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 524 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 525 return; 526 } 527 ar = mtod(m, struct arphdr *); 528 529 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 530 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 531 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 532 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) { 533 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 534 (unsigned char *)&ar->ar_hrd, ""); 535 m_freem(m); 536 return; 537 } 538 539 if (m->m_len < arphdr_len(ar)) { 540 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 541 log(LOG_ERR, "arp: runt packet\n"); 542 m_freem(m); 543 return; 544 } 545 ar = mtod(m, struct arphdr *); 546 } 547 548 switch (ntohs(ar->ar_pro)) { 549 #ifdef INET 550 case ETHERTYPE_IP: 551 in_arpinput(m); 552 return; 553 #endif 554 } 555 m_freem(m); 556 } 557 558 #ifdef INET 559 /* 560 * ARP for Internet protocols on 10 Mb/s Ethernet. 561 * Algorithm is that given in RFC 826. 562 * In addition, a sanity check is performed on the sender 563 * protocol address, to catch impersonators. 564 * We no longer handle negotiations for use of trailer protocol: 565 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 566 * along with IP replies if we wanted trailers sent to us, 567 * and also sent them in response to IP replies. 568 * This allowed either end to announce the desire to receive 569 * trailer packets. 570 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 571 * but formerly didn't normally send requests. 572 */ 573 static int log_arp_wrong_iface = 1; 574 static int log_arp_movements = 1; 575 static int log_arp_permanent_modify = 1; 576 577 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 578 &log_arp_wrong_iface, 0, 579 "log arp packets arriving on the wrong interface"); 580 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 581 &log_arp_movements, 0, 582 "log arp replies from MACs different than the one in the cache"); 583 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 584 &log_arp_permanent_modify, 0, 585 "log arp replies from MACs different than the one in the permanent arp entry"); 586 587 588 static void 589 in_arpinput(m) 590 struct mbuf *m; 591 { 592 struct arphdr *ah; 593 struct ifnet *ifp = m->m_pkthdr.rcvif; 594 struct llinfo_arp *la; 595 struct rtentry *rt; 596 struct ifaddr *ifa; 597 struct in_ifaddr *ia; 598 struct sockaddr_dl *sdl; 599 struct sockaddr sa; 600 struct in_addr isaddr, itaddr, myaddr; 601 struct mbuf *hold; 602 u_int8_t *enaddr = NULL; 603 int op, rif_len; 604 int req_len; 605 int bridged = 0; 606 #ifdef DEV_CARP 607 int carp_match = 0; 608 #endif 609 610 if (ifp->if_bridge) 611 bridged = 1; 612 613 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 614 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 615 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 616 return; 617 } 618 619 ah = mtod(m, struct arphdr *); 620 op = ntohs(ah->ar_op); 621 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 622 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 623 624 /* 625 * For a bridge, we want to check the address irrespective 626 * of the receive interface. (This will change slightly 627 * when we have clusters of interfaces). 628 * If the interface does not match, but the recieving interface 629 * is part of carp, we call carp_iamatch to see if this is a 630 * request for the virtual host ip. 631 * XXX: This is really ugly! 632 */ 633 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 634 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 635 (ia->ia_ifp == ifp)) && 636 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 637 goto match; 638 #ifdef DEV_CARP 639 if (ifp->if_carp != NULL && 640 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 641 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 642 carp_match = 1; 643 goto match; 644 } 645 #endif 646 } 647 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 648 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 649 (ia->ia_ifp == ifp)) && 650 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 651 goto match; 652 /* 653 * No match, use the first inet address on the receive interface 654 * as a dummy address for the rest of the function. 655 */ 656 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 657 if (ifa->ifa_addr->sa_family == AF_INET) { 658 ia = ifatoia(ifa); 659 goto match; 660 } 661 /* 662 * If bridging, fall back to using any inet address. 663 */ 664 if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) 665 goto drop; 666 match: 667 if (!enaddr) 668 enaddr = (u_int8_t *)IF_LLADDR(ifp); 669 myaddr = ia->ia_addr.sin_addr; 670 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 671 goto drop; /* it's from me, ignore it. */ 672 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 673 log(LOG_ERR, 674 "arp: link address is broadcast for IP address %s!\n", 675 inet_ntoa(isaddr)); 676 goto drop; 677 } 678 /* 679 * Warn if another host is using the same IP address, but only if the 680 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 681 * case we suppress the warning to avoid false positive complaints of 682 * potential misconfiguration. 683 */ 684 if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 685 log(LOG_ERR, 686 "arp: %*D is using my IP address %s!\n", 687 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 688 inet_ntoa(isaddr)); 689 itaddr = myaddr; 690 goto reply; 691 } 692 if (ifp->if_flags & IFF_STATICARP) 693 goto reply; 694 rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0); 695 if (rt != NULL) { 696 la = (struct llinfo_arp *)rt->rt_llinfo; 697 if (la == NULL) { 698 RT_UNLOCK(rt); 699 goto reply; 700 } 701 } else 702 goto reply; 703 704 /* The following is not an error when doing bridging. */ 705 if (!bridged && rt->rt_ifp != ifp 706 #ifdef DEV_CARP 707 && (ifp->if_type != IFT_CARP || !carp_match) 708 #endif 709 ) { 710 if (log_arp_wrong_iface) 711 log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n", 712 inet_ntoa(isaddr), 713 rt->rt_ifp->if_xname, 714 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 715 ifp->if_xname); 716 RT_UNLOCK(rt); 717 goto reply; 718 } 719 sdl = SDL(rt->rt_gateway); 720 if (sdl->sdl_alen && 721 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 722 if (rt->rt_expire) { 723 if (log_arp_movements) 724 log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n", 725 inet_ntoa(isaddr), 726 ifp->if_addrlen, (u_char *)LLADDR(sdl), ":", 727 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 728 ifp->if_xname); 729 } else { 730 RT_UNLOCK(rt); 731 if (log_arp_permanent_modify) 732 log(LOG_ERR, "arp: %*D attempts to modify " 733 "permanent entry for %s on %s\n", 734 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 735 inet_ntoa(isaddr), ifp->if_xname); 736 goto reply; 737 } 738 } 739 /* 740 * sanity check for the address length. 741 * XXX this does not work for protocols with variable address 742 * length. -is 743 */ 744 if (sdl->sdl_alen && 745 sdl->sdl_alen != ah->ar_hln) { 746 log(LOG_WARNING, 747 "arp from %*D: new addr len %d, was %d", 748 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 749 ah->ar_hln, sdl->sdl_alen); 750 } 751 if (ifp->if_addrlen != ah->ar_hln) { 752 log(LOG_WARNING, 753 "arp from %*D: addr len: new %d, i/f %d (ignored)", 754 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 755 ah->ar_hln, ifp->if_addrlen); 756 RT_UNLOCK(rt); 757 goto reply; 758 } 759 (void)memcpy(LLADDR(sdl), ar_sha(ah), 760 sdl->sdl_alen = ah->ar_hln); 761 /* 762 * If we receive an arp from a token-ring station over 763 * a token-ring nic then try to save the source 764 * routing info. 765 */ 766 if (ifp->if_type == IFT_ISO88025) { 767 struct iso88025_header *th = NULL; 768 struct iso88025_sockaddr_dl_data *trld; 769 770 th = (struct iso88025_header *)m->m_pkthdr.header; 771 trld = SDL_ISO88025(sdl); 772 rif_len = TR_RCF_RIFLEN(th->rcf); 773 if ((th->iso88025_shost[0] & TR_RII) && 774 (rif_len > 2)) { 775 trld->trld_rcf = th->rcf; 776 trld->trld_rcf ^= htons(TR_RCF_DIR); 777 memcpy(trld->trld_route, th->rd, rif_len - 2); 778 trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); 779 /* 780 * Set up source routing information for 781 * reply packet (XXX) 782 */ 783 m->m_data -= rif_len; 784 m->m_len += rif_len; 785 m->m_pkthdr.len += rif_len; 786 } else { 787 th->iso88025_shost[0] &= ~TR_RII; 788 trld->trld_rcf = 0; 789 } 790 m->m_data -= 8; 791 m->m_len += 8; 792 m->m_pkthdr.len += 8; 793 th->rcf = trld->trld_rcf; 794 } 795 if (rt->rt_expire) 796 rt->rt_expire = time_uptime + arpt_keep; 797 la->la_asked = 0; 798 la->la_preempt = arp_maxtries; 799 hold = la->la_hold; 800 la->la_hold = NULL; 801 RT_UNLOCK(rt); 802 if (hold != NULL) 803 (*ifp->if_output)(ifp, hold, rt_key(rt), rt); 804 805 reply: 806 if (op != ARPOP_REQUEST) 807 goto drop; 808 if (itaddr.s_addr == myaddr.s_addr) { 809 /* I am the target */ 810 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 811 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 812 } else { 813 rt = arplookup(itaddr.s_addr, 0, SIN_PROXY); 814 if (rt == NULL) { 815 struct sockaddr_in sin; 816 817 if (!arp_proxyall) 818 goto drop; 819 820 bzero(&sin, sizeof sin); 821 sin.sin_family = AF_INET; 822 sin.sin_len = sizeof sin; 823 sin.sin_addr = itaddr; 824 825 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 826 if (!rt) 827 goto drop; 828 /* 829 * Don't send proxies for nodes on the same interface 830 * as this one came out of, or we'll get into a fight 831 * over who claims what Ether address. 832 */ 833 if (rt->rt_ifp == ifp) { 834 rtfree(rt); 835 goto drop; 836 } 837 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 838 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 839 rtfree(rt); 840 841 /* 842 * Also check that the node which sent the ARP packet 843 * is on the the interface we expect it to be on. This 844 * avoids ARP chaos if an interface is connected to the 845 * wrong network. 846 */ 847 sin.sin_addr = isaddr; 848 849 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 850 if (!rt) 851 goto drop; 852 if (rt->rt_ifp != ifp) { 853 log(LOG_INFO, "arp_proxy: ignoring request" 854 " from %s via %s, expecting %s\n", 855 inet_ntoa(isaddr), ifp->if_xname, 856 rt->rt_ifp->if_xname); 857 rtfree(rt); 858 goto drop; 859 } 860 rtfree(rt); 861 862 #ifdef DEBUG_PROXY 863 printf("arp: proxying for %s\n", 864 inet_ntoa(itaddr)); 865 #endif 866 } else { 867 /* 868 * Return proxied ARP replies only on the interface 869 * or bridge cluster where this network resides. 870 * Otherwise we may conflict with the host we are 871 * proxying for. 872 */ 873 if (rt->rt_ifp != ifp && 874 (rt->rt_ifp->if_bridge != ifp->if_bridge || 875 ifp->if_bridge == NULL)) { 876 RT_UNLOCK(rt); 877 goto drop; 878 } 879 sdl = SDL(rt->rt_gateway); 880 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 881 (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 882 RT_UNLOCK(rt); 883 } 884 } 885 886 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 887 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 888 ah->ar_op = htons(ARPOP_REPLY); 889 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 890 m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */ 891 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 892 m->m_pkthdr.len = m->m_len; 893 sa.sa_family = AF_ARP; 894 sa.sa_len = 2; 895 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 896 return; 897 898 drop: 899 m_freem(m); 900 } 901 #endif 902 903 /* 904 * Lookup or enter a new address in arptab. 905 */ 906 static struct rtentry * 907 arplookup(addr, create, proxy) 908 u_long addr; 909 int create, proxy; 910 { 911 struct rtentry *rt; 912 struct sockaddr_inarp sin; 913 const char *why = 0; 914 915 bzero(&sin, sizeof(sin)); 916 sin.sin_len = sizeof(sin); 917 sin.sin_family = AF_INET; 918 sin.sin_addr.s_addr = addr; 919 if (proxy) 920 sin.sin_other = SIN_PROXY; 921 rt = rtalloc1((struct sockaddr *)&sin, create, 0UL); 922 if (rt == 0) 923 return (0); 924 925 if (rt->rt_flags & RTF_GATEWAY) 926 why = "host is not on local network"; 927 else if ((rt->rt_flags & RTF_LLINFO) == 0) 928 why = "could not allocate llinfo"; 929 else if (rt->rt_gateway->sa_family != AF_LINK) 930 why = "gateway route is not ours"; 931 932 if (why) { 933 #define ISDYNCLONE(_rt) \ 934 (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) 935 if (create) 936 log(LOG_DEBUG, "arplookup %s failed: %s\n", 937 inet_ntoa(sin.sin_addr), why); 938 /* 939 * If there are no references to this Layer 2 route, 940 * and it is a cloned route, and not static, and 941 * arplookup() is creating the route, then purge 942 * it from the routing table as it is probably bogus. 943 */ 944 if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) 945 rtexpunge(rt); 946 RTFREE_LOCKED(rt); 947 return (0); 948 #undef ISDYNCLONE 949 } else { 950 RT_REMREF(rt); 951 return (rt); 952 } 953 } 954 955 void 956 arp_ifinit(ifp, ifa) 957 struct ifnet *ifp; 958 struct ifaddr *ifa; 959 { 960 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 961 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 962 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 963 ifa->ifa_rtrequest = arp_rtrequest; 964 ifa->ifa_flags |= RTF_CLONING; 965 } 966 967 void 968 arp_ifinit2(ifp, ifa, enaddr) 969 struct ifnet *ifp; 970 struct ifaddr *ifa; 971 u_char *enaddr; 972 { 973 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 974 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 975 &IA_SIN(ifa)->sin_addr, enaddr); 976 ifa->ifa_rtrequest = arp_rtrequest; 977 ifa->ifa_flags |= RTF_CLONING; 978 } 979 980 static void 981 arp_init(void) 982 { 983 984 arpintrq.ifq_maxlen = 50; 985 mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF); 986 LIST_INIT(&llinfo_arp); 987 callout_init(&arp_callout, CALLOUT_MPSAFE); 988 netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE); 989 callout_reset(&arp_callout, hz, arptimer, NULL); 990 } 991 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 992