if_ether.c revision 174699
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 174699 2007-12-17 04:19:25Z kmacy $"); 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/socket.h> 53#include <sys/syslog.h> 54 55#include <net/if.h> 56#include <net/if_dl.h> 57#include <net/if_types.h> 58#include <net/route.h> 59#include <net/netisr.h> 60#include <net/if_llc.h> 61#include <net/ethernet.h> 62 63#include <netinet/in.h> 64#include <netinet/in_var.h> 65#include <netinet/if_ether.h> 66 67#include <net/if_arc.h> 68#include <net/iso88025.h> 69 70#ifdef DEV_CARP 71#include <netinet/ip_carp.h> 72#endif 73 74#include <security/mac/mac_framework.h> 75 76#define SIN(s) ((struct sockaddr_in *)s) 77#define SDL(s) ((struct sockaddr_dl *)s) 78 79SYSCTL_DECL(_net_link_ether); 80SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 81 82/* timer values */ 83static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 84 85SYSCTL_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 90struct llinfo_arp { 91 struct callout la_timer; 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 98static struct ifqueue arpintrq; 99static int arp_allocated; 100 101static int arp_maxtries = 5; 102static int useloopback = 1; /* use loopback interface for local traffic */ 103static int arp_proxyall = 0; 104 105SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 106 &arp_maxtries, 0, "ARP resolution attempts before returning error"); 107SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 108 &useloopback, 0, "Use the loopback interface for local traffic"); 109SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 110 &arp_proxyall, 0, "Enable proxy ARP for all suitable requests"); 111 112static void arp_init(void); 113static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 114static void arprequest(struct ifnet *, 115 struct in_addr *, struct in_addr *, u_char *); 116static void arpintr(struct mbuf *); 117static void arptimer(void *); 118static struct rtentry 119 *arplookup(u_long, int, int); 120#ifdef INET 121static void in_arpinput(struct mbuf *); 122#endif 123 124/* 125 * Timeout routine. 126 */ 127static void 128arptimer(void *arg) 129{ 130 struct rtentry *rt = (struct rtentry *)arg; 131 132 RT_LOCK_ASSERT(rt); 133 /* 134 * The lock is needed to close a theoretical race 135 * between spontaneous expiry and intentional removal. 136 * We still got an extra reference on rtentry, so can 137 * safely pass pointers to its contents. 138 */ 139 RT_UNLOCK(rt); 140 141 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 142} 143 144/* 145 * Parallel to llc_rtrequest. 146 */ 147static void 148arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 149{ 150 struct sockaddr *gate; 151 struct llinfo_arp *la; 152 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 153 struct in_ifaddr *ia; 154 struct ifaddr *ifa; 155 156 RT_LOCK_ASSERT(rt); 157 158 if (rt->rt_flags & RTF_GATEWAY) 159 return; 160 gate = rt->rt_gateway; 161 la = (struct llinfo_arp *)rt->rt_llinfo; 162 switch (req) { 163 164 case RTM_ADD: 165 /* 166 * XXX: If this is a manually added route to interface 167 * such as older version of routed or gated might provide, 168 * restore cloning bit. 169 */ 170 if ((rt->rt_flags & RTF_HOST) == 0 && 171 rt_mask(rt) != NULL && 172 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 173 rt->rt_flags |= RTF_CLONING; 174 if (rt->rt_flags & RTF_CLONING) { 175 /* 176 * Case 1: This route should come from a route to iface. 177 */ 178 rt_setgate(rt, rt_key(rt), 179 (struct sockaddr *)&null_sdl); 180 gate = rt->rt_gateway; 181 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 182 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 183 rt->rt_expire = time_uptime; 184 break; 185 } 186 /* Announce a new entry if requested. */ 187 if (rt->rt_flags & RTF_ANNOUNCE) 188 arprequest(rt->rt_ifp, 189 &SIN(rt_key(rt))->sin_addr, 190 &SIN(rt_key(rt))->sin_addr, 191 (u_char *)LLADDR(SDL(gate))); 192 /*FALLTHROUGH*/ 193 case RTM_RESOLVE: 194 if (gate->sa_family != AF_LINK || 195 gate->sa_len < sizeof(null_sdl)) { 196 log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, 197 inet_ntoa(SIN(rt_key(rt))->sin_addr), 198 (gate->sa_family != AF_LINK) ? 199 " (!AF_LINK)": ""); 200 break; 201 } 202 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 203 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 204 if (la != 0) 205 break; /* This happens on a route change */ 206 /* 207 * Case 2: This route may come from cloning, or a manual route 208 * add with a LL address. 209 */ 210 R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); 211 rt->rt_llinfo = (caddr_t)la; 212 if (la == 0) { 213 log(LOG_DEBUG, "%s: malloc failed\n", __func__); 214 break; 215 } 216 arp_allocated++; 217 /* 218 * We are storing a route entry outside of radix tree. So, 219 * it can be found and accessed by other means than radix 220 * lookup. The routing code assumes that any rtentry detached 221 * from radix can be destroyed safely. To prevent this, we 222 * add an additional reference. 223 */ 224 RT_ADDREF(rt); 225 la->la_rt = rt; 226 rt->rt_flags |= RTF_LLINFO; 227 callout_init_mtx(&la->la_timer, &rt->rt_mtx, 228 CALLOUT_RETURNUNLOCKED); 229 230#ifdef INET 231 /* 232 * This keeps the multicast addresses from showing up 233 * in `arp -a' listings as unresolved. It's not actually 234 * functional. Then the same for broadcast. 235 */ 236 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && 237 rt->rt_ifp->if_type != IFT_ARCNET) { 238 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 239 LLADDR(SDL(gate))); 240 SDL(gate)->sdl_alen = 6; 241 rt->rt_expire = 0; 242 } 243 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 244 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 245 rt->rt_ifp->if_addrlen); 246 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 247 rt->rt_expire = 0; 248 } 249#endif 250 251 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 252 if (ia->ia_ifp == rt->rt_ifp && 253 SIN(rt_key(rt))->sin_addr.s_addr == 254 (IA_SIN(ia))->sin_addr.s_addr) 255 break; 256 } 257 if (ia) { 258 /* 259 * This test used to be 260 * if (loif.if_flags & IFF_UP) 261 * It allowed local traffic to be forced 262 * through the hardware by configuring the loopback down. 263 * However, it causes problems during network configuration 264 * for boards that can't receive packets they send. 265 * It is now necessary to clear "useloopback" and remove 266 * the route to force traffic out to the hardware. 267 */ 268 rt->rt_expire = 0; 269 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 270 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 271 if (useloopback) { 272 rt->rt_ifp = loif; 273 rt->rt_rmx.rmx_mtu = loif->if_mtu; 274 } 275 276 /* 277 * make sure to set rt->rt_ifa to the interface 278 * address we are using, otherwise we will have trouble 279 * with source address selection. 280 */ 281 ifa = &ia->ia_ifa; 282 if (ifa != rt->rt_ifa) { 283 IFAFREE(rt->rt_ifa); 284 IFAREF(ifa); 285 rt->rt_ifa = ifa; 286 } 287 } 288 break; 289 290 case RTM_DELETE: 291 if (la == NULL) /* XXX: at least CARP does this. */ 292 break; 293 callout_stop(&la->la_timer); 294 rt->rt_llinfo = NULL; 295 rt->rt_flags &= ~RTF_LLINFO; 296 RT_REMREF(rt); 297 if (la->la_hold) 298 m_freem(la->la_hold); 299 Free((caddr_t)la); 300 } 301} 302 303/* 304 * Broadcast an ARP request. Caller specifies: 305 * - arp header source ip address 306 * - arp header target ip address 307 * - arp header source ethernet address 308 */ 309static void 310arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 311 u_char *enaddr) 312{ 313 struct mbuf *m; 314 struct arphdr *ah; 315 struct sockaddr sa; 316 317 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 318 return; 319 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 320 2*ifp->if_data.ifi_addrlen; 321 m->m_pkthdr.len = m->m_len; 322 MH_ALIGN(m, m->m_len); 323 ah = mtod(m, struct arphdr *); 324 bzero((caddr_t)ah, m->m_len); 325#ifdef MAC 326 mac_netinet_arp_send(ifp, m); 327#endif 328 ah->ar_pro = htons(ETHERTYPE_IP); 329 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 330 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 331 ah->ar_op = htons(ARPOP_REQUEST); 332 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 333 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 334 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 335 sa.sa_family = AF_ARP; 336 sa.sa_len = 2; 337 m->m_flags |= M_BCAST; 338 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 339 340 return; 341} 342 343/* 344 * Resolve an IP address into an ethernet address. 345 * On input: 346 * ifp is the interface we use 347 * dst is the next hop, 348 * rt0 is the route to the final destination (possibly useless) 349 * m is the mbuf 350 * desten is where we want the address. 351 * 352 * On success, desten is filled in and the function returns 0; 353 * If the packet must be held pending resolution, we return EWOULDBLOCK 354 * On other errors, we return the corresponding error code. 355 */ 356int 357arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 358 struct sockaddr *dst, u_char *desten) 359{ 360 struct llinfo_arp *la = NULL; 361 struct rtentry *rt = NULL; 362 struct sockaddr_dl *sdl; 363 int error; 364 365 if (m != NULL) { 366 367 if (m->m_flags & M_BCAST) { /* broadcast */ 368 (void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 369 return (0); 370 } 371 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */ 372 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 373 return (0); 374 } 375 } 376 377 if (rt0 != NULL) { 378 error = rt_check(&rt, &rt0, dst); 379 if (error) { 380 m_freem(m); 381 return error; 382 } 383 la = (struct llinfo_arp *)rt->rt_llinfo; 384 if (la == NULL) 385 RT_UNLOCK(rt); 386 } 387 if (la == NULL) { 388 /* 389 * We enter this block in case if rt0 was NULL, 390 * or if rt found by rt_check() didn't have llinfo. 391 */ 392 rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0); 393 if (rt == NULL) { 394 log(LOG_DEBUG, 395 "arpresolve: can't allocate route for %s\n", 396 inet_ntoa(SIN(dst)->sin_addr)); 397 m_freem(m); 398 return (EINVAL); /* XXX */ 399 } 400 la = (struct llinfo_arp *)rt->rt_llinfo; 401 if (la == NULL) { 402 RT_UNLOCK(rt); 403 log(LOG_DEBUG, 404 "arpresolve: can't allocate llinfo for %s\n", 405 inet_ntoa(SIN(dst)->sin_addr)); 406 m_freem(m); 407 return (EINVAL); /* XXX */ 408 } 409 } 410 sdl = SDL(rt->rt_gateway); 411 /* 412 * Check the address family and length is valid, the address 413 * is resolved; otherwise, try to resolve. 414 */ 415 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) && 416 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 417 418 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 419 420 /* 421 * If entry has an expiry time and it is approaching, 422 * send an ARP request. 423 */ 424 if ((rt->rt_expire != 0) && 425 (time_uptime + la->la_preempt > rt->rt_expire)) { 426 struct in_addr sin = 427 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 428 429 la->la_preempt--; 430 RT_UNLOCK(rt); 431 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 432 IF_LLADDR(ifp)); 433 return (0); 434 } 435 436 RT_UNLOCK(rt); 437 return (0); 438 } 439 /* 440 * If ARP is disabled or static on this interface, stop. 441 * XXX 442 * Probably should not allocate empty llinfo struct if we are 443 * not going to be sending out an arp request. 444 */ 445 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 446 RT_UNLOCK(rt); 447 m_freem(m); 448 return (EINVAL); 449 } 450 /* 451 * There is an arptab entry, but no ethernet address 452 * response yet. Replace the held mbuf with this 453 * latest one. 454 */ 455 if (m != NULL) { 456 if (la->la_hold) 457 m_freem(la->la_hold); 458 la->la_hold = m; 459 } 460 KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry")); 461 462 /* 463 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 464 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 465 * if we have already sent arp_maxtries ARP requests. Retransmit the 466 * ARP request, but not faster than one request per second. 467 */ 468 if (la->la_asked < arp_maxtries) 469 error = EWOULDBLOCK; /* First request. */ 470 else 471 error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH; 472 473 if (la->la_asked == 0 || rt->rt_expire != time_uptime) { 474 struct in_addr sin = 475 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 476 477 rt->rt_expire = time_uptime; 478 callout_reset(&la->la_timer, hz, arptimer, rt); 479 la->la_asked++; 480 RT_UNLOCK(rt); 481 482 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 483 IF_LLADDR(ifp)); 484 } else 485 RT_UNLOCK(rt); 486 487 return (error); 488} 489 490/* 491 * Common length and type checks are done here, 492 * then the protocol-specific routine is called. 493 */ 494static void 495arpintr(struct mbuf *m) 496{ 497 struct arphdr *ar; 498 499 if (m->m_len < sizeof(struct arphdr) && 500 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 501 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 502 return; 503 } 504 ar = mtod(m, struct arphdr *); 505 506 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 507 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 508 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 509 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) { 510 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 511 (unsigned char *)&ar->ar_hrd, ""); 512 m_freem(m); 513 return; 514 } 515 516 if (m->m_len < arphdr_len(ar)) { 517 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 518 log(LOG_ERR, "arp: runt packet\n"); 519 m_freem(m); 520 return; 521 } 522 ar = mtod(m, struct arphdr *); 523 } 524 525 switch (ntohs(ar->ar_pro)) { 526#ifdef INET 527 case ETHERTYPE_IP: 528 in_arpinput(m); 529 return; 530#endif 531 } 532 m_freem(m); 533} 534 535#ifdef INET 536/* 537 * ARP for Internet protocols on 10 Mb/s Ethernet. 538 * Algorithm is that given in RFC 826. 539 * In addition, a sanity check is performed on the sender 540 * protocol address, to catch impersonators. 541 * We no longer handle negotiations for use of trailer protocol: 542 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 543 * along with IP replies if we wanted trailers sent to us, 544 * and also sent them in response to IP replies. 545 * This allowed either end to announce the desire to receive 546 * trailer packets. 547 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 548 * but formerly didn't normally send requests. 549 */ 550static int log_arp_wrong_iface = 1; 551static int log_arp_movements = 1; 552static int log_arp_permanent_modify = 1; 553 554SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 555 &log_arp_wrong_iface, 0, 556 "log arp packets arriving on the wrong interface"); 557SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 558 &log_arp_movements, 0, 559 "log arp replies from MACs different than the one in the cache"); 560SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 561 &log_arp_permanent_modify, 0, 562 "log arp replies from MACs different than the one in the permanent arp entry"); 563 564 565static void 566in_arpinput(struct mbuf *m) 567{ 568 struct arphdr *ah; 569 struct ifnet *ifp = m->m_pkthdr.rcvif; 570 struct llinfo_arp *la; 571 struct rtentry *rt; 572 struct ifaddr *ifa; 573 struct in_ifaddr *ia; 574 struct sockaddr_dl *sdl; 575 struct sockaddr sa; 576 struct in_addr isaddr, itaddr, myaddr; 577 struct mbuf *hold; 578 u_int8_t *enaddr = NULL; 579 int op, rif_len; 580 int req_len; 581 int bridged = 0; 582#ifdef DEV_CARP 583 int carp_match = 0; 584#endif 585 struct sockaddr_in sin; 586 sin.sin_len = sizeof(struct sockaddr_in); 587 sin.sin_family = AF_INET; 588 589 if (ifp->if_bridge) 590 bridged = 1; 591 592 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 593 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 594 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 595 return; 596 } 597 598 ah = mtod(m, struct arphdr *); 599 op = ntohs(ah->ar_op); 600 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 601 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 602 603 /* 604 * For a bridge, we want to check the address irrespective 605 * of the receive interface. (This will change slightly 606 * when we have clusters of interfaces). 607 * If the interface does not match, but the recieving interface 608 * is part of carp, we call carp_iamatch to see if this is a 609 * request for the virtual host ip. 610 * XXX: This is really ugly! 611 */ 612 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 613 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 614 (ia->ia_ifp == ifp)) && 615 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 616 goto match; 617#ifdef DEV_CARP 618 if (ifp->if_carp != NULL && 619 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 620 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 621 carp_match = 1; 622 goto match; 623 } 624#endif 625 } 626 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 627 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 628 (ia->ia_ifp == ifp)) && 629 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 630 goto match; 631 /* 632 * No match, use the first inet address on the receive interface 633 * as a dummy address for the rest of the function. 634 */ 635 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 636 if (ifa->ifa_addr->sa_family == AF_INET) { 637 ia = ifatoia(ifa); 638 goto match; 639 } 640 /* 641 * If bridging, fall back to using any inet address. 642 */ 643 if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) 644 goto drop; 645match: 646 if (!enaddr) 647 enaddr = (u_int8_t *)IF_LLADDR(ifp); 648 myaddr = ia->ia_addr.sin_addr; 649 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 650 goto drop; /* it's from me, ignore it. */ 651 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 652 log(LOG_ERR, 653 "arp: link address is broadcast for IP address %s!\n", 654 inet_ntoa(isaddr)); 655 goto drop; 656 } 657 /* 658 * Warn if another host is using the same IP address, but only if the 659 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 660 * case we suppress the warning to avoid false positive complaints of 661 * potential misconfiguration. 662 */ 663 if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 664 log(LOG_ERR, 665 "arp: %*D is using my IP address %s on %s!\n", 666 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 667 inet_ntoa(isaddr), ifp->if_xname); 668 itaddr = myaddr; 669 goto reply; 670 } 671 if (ifp->if_flags & IFF_STATICARP) 672 goto reply; 673 rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0); 674 if (rt != NULL) { 675 sin.sin_addr.s_addr = isaddr.s_addr; 676 EVENTHANDLER_INVOKE(route_event, RTEVENT_ARP_UPDATE, rt, NULL, 677 (struct sockaddr *)&sin); 678 679 la = (struct llinfo_arp *)rt->rt_llinfo; 680 if (la == NULL) { 681 RT_UNLOCK(rt); 682 goto reply; 683 } 684 } else 685 goto reply; 686 687 /* The following is not an error when doing bridging. */ 688 if (!bridged && rt->rt_ifp != ifp 689#ifdef DEV_CARP 690 && (ifp->if_type != IFT_CARP || !carp_match) 691#endif 692 ) { 693 if (log_arp_wrong_iface) 694 log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n", 695 inet_ntoa(isaddr), 696 rt->rt_ifp->if_xname, 697 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 698 ifp->if_xname); 699 RT_UNLOCK(rt); 700 goto reply; 701 } 702 sdl = SDL(rt->rt_gateway); 703 if (sdl->sdl_alen && 704 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 705 if (rt->rt_expire) { 706 if (log_arp_movements) 707 log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n", 708 inet_ntoa(isaddr), 709 ifp->if_addrlen, (u_char *)LLADDR(sdl), ":", 710 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 711 ifp->if_xname); 712 } else { 713 RT_UNLOCK(rt); 714 if (log_arp_permanent_modify) 715 log(LOG_ERR, "arp: %*D attempts to modify " 716 "permanent entry for %s on %s\n", 717 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 718 inet_ntoa(isaddr), ifp->if_xname); 719 goto reply; 720 } 721 } 722 /* 723 * sanity check for the address length. 724 * XXX this does not work for protocols with variable address 725 * length. -is 726 */ 727 if (sdl->sdl_alen && 728 sdl->sdl_alen != ah->ar_hln) { 729 log(LOG_WARNING, 730 "arp from %*D: new addr len %d, was %d", 731 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 732 ah->ar_hln, sdl->sdl_alen); 733 } 734 if (ifp->if_addrlen != ah->ar_hln) { 735 log(LOG_WARNING, 736 "arp from %*D: addr len: new %d, i/f %d (ignored)", 737 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 738 ah->ar_hln, ifp->if_addrlen); 739 RT_UNLOCK(rt); 740 goto reply; 741 } 742 (void)memcpy(LLADDR(sdl), ar_sha(ah), 743 sdl->sdl_alen = ah->ar_hln); 744 /* 745 * If we receive an arp from a token-ring station over 746 * a token-ring nic then try to save the source 747 * routing info. 748 */ 749 if (ifp->if_type == IFT_ISO88025) { 750 struct iso88025_header *th = NULL; 751 struct iso88025_sockaddr_dl_data *trld; 752 753 th = (struct iso88025_header *)m->m_pkthdr.header; 754 trld = SDL_ISO88025(sdl); 755 rif_len = TR_RCF_RIFLEN(th->rcf); 756 if ((th->iso88025_shost[0] & TR_RII) && 757 (rif_len > 2)) { 758 trld->trld_rcf = th->rcf; 759 trld->trld_rcf ^= htons(TR_RCF_DIR); 760 memcpy(trld->trld_route, th->rd, rif_len - 2); 761 trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); 762 /* 763 * Set up source routing information for 764 * reply packet (XXX) 765 */ 766 m->m_data -= rif_len; 767 m->m_len += rif_len; 768 m->m_pkthdr.len += rif_len; 769 } else { 770 th->iso88025_shost[0] &= ~TR_RII; 771 trld->trld_rcf = 0; 772 } 773 m->m_data -= 8; 774 m->m_len += 8; 775 m->m_pkthdr.len += 8; 776 th->rcf = trld->trld_rcf; 777 } 778 if (rt->rt_expire) { 779 rt->rt_expire = time_uptime + arpt_keep; 780 callout_reset(&la->la_timer, hz * arpt_keep, arptimer, rt); 781 } 782 la->la_asked = 0; 783 la->la_preempt = arp_maxtries; 784 hold = la->la_hold; 785 la->la_hold = NULL; 786 RT_UNLOCK(rt); 787 if (hold != NULL) 788 (*ifp->if_output)(ifp, hold, rt_key(rt), rt); 789 790reply: 791 if (op != ARPOP_REQUEST) 792 goto drop; 793 if (itaddr.s_addr == myaddr.s_addr) { 794 /* I am the target */ 795 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 796 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 797 } else { 798 rt = arplookup(itaddr.s_addr, 0, SIN_PROXY); 799 if (rt == NULL) { 800 struct sockaddr_in sin; 801 802 if (!arp_proxyall) 803 goto drop; 804 805 bzero(&sin, sizeof sin); 806 sin.sin_family = AF_INET; 807 sin.sin_len = sizeof sin; 808 sin.sin_addr = itaddr; 809 810 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 811 if (!rt) 812 goto drop; 813 /* 814 * Don't send proxies for nodes on the same interface 815 * as this one came out of, or we'll get into a fight 816 * over who claims what Ether address. 817 */ 818 if (rt->rt_ifp == ifp) { 819 rtfree(rt); 820 goto drop; 821 } 822 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 823 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 824 rtfree(rt); 825 826 /* 827 * Also check that the node which sent the ARP packet 828 * is on the the interface we expect it to be on. This 829 * avoids ARP chaos if an interface is connected to the 830 * wrong network. 831 */ 832 sin.sin_addr = isaddr; 833 834 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 835 if (!rt) 836 goto drop; 837 if (rt->rt_ifp != ifp) { 838 log(LOG_INFO, "arp_proxy: ignoring request" 839 " from %s via %s, expecting %s\n", 840 inet_ntoa(isaddr), ifp->if_xname, 841 rt->rt_ifp->if_xname); 842 rtfree(rt); 843 goto drop; 844 } 845 rtfree(rt); 846 847#ifdef DEBUG_PROXY 848 printf("arp: proxying for %s\n", 849 inet_ntoa(itaddr)); 850#endif 851 } else { 852 /* 853 * Return proxied ARP replies only on the interface 854 * or bridge cluster where this network resides. 855 * Otherwise we may conflict with the host we are 856 * proxying for. 857 */ 858 if (rt->rt_ifp != ifp && 859 (rt->rt_ifp->if_bridge != ifp->if_bridge || 860 ifp->if_bridge == NULL)) { 861 RT_UNLOCK(rt); 862 goto drop; 863 } 864 sdl = SDL(rt->rt_gateway); 865 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 866 (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 867 RT_UNLOCK(rt); 868 } 869 } 870 871 if (itaddr.s_addr == myaddr.s_addr && 872 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 873 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 874#ifdef DEBUG_LINKLOCAL 875 printf("arp: sending reply for link-local addr %s\n", 876 inet_ntoa(itaddr)); 877#endif 878 m->m_flags |= M_BCAST; 879 m->m_flags &= ~M_MCAST; 880 } else { 881 /* default behaviour; never reply by broadcast. */ 882 m->m_flags &= ~(M_BCAST|M_MCAST); 883 } 884 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 885 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 886 ah->ar_op = htons(ARPOP_REPLY); 887 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 888 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 889 m->m_pkthdr.len = m->m_len; 890 sa.sa_family = AF_ARP; 891 sa.sa_len = 2; 892 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 893 return; 894 895drop: 896 m_freem(m); 897} 898#endif 899 900/* 901 * Lookup or enter a new address in arptab. 902 */ 903static struct rtentry * 904arplookup(u_long addr, int create, int proxy) 905{ 906 struct rtentry *rt; 907 struct sockaddr_inarp sin; 908 const char *why = 0; 909 910 bzero(&sin, sizeof(sin)); 911 sin.sin_len = sizeof(sin); 912 sin.sin_family = AF_INET; 913 sin.sin_addr.s_addr = addr; 914 if (proxy) 915 sin.sin_other = SIN_PROXY; 916 rt = rtalloc1((struct sockaddr *)&sin, create, 0UL); 917 if (rt == 0) 918 return (0); 919 920 if (rt->rt_flags & RTF_GATEWAY) 921 why = "host is not on local network"; 922 else if ((rt->rt_flags & RTF_LLINFO) == 0) 923 why = "could not allocate llinfo"; 924 else if (rt->rt_gateway->sa_family != AF_LINK) 925 why = "gateway route is not ours"; 926 927 if (why) { 928#define ISDYNCLONE(_rt) \ 929 (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) 930 if (create) 931 log(LOG_DEBUG, "arplookup %s failed: %s\n", 932 inet_ntoa(sin.sin_addr), why); 933 /* 934 * If there are no references to this Layer 2 route, 935 * and it is a cloned route, and not static, and 936 * arplookup() is creating the route, then purge 937 * it from the routing table as it is probably bogus. 938 */ 939 if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) 940 rtexpunge(rt); 941 RTFREE_LOCKED(rt); 942 return (0); 943#undef ISDYNCLONE 944 } else { 945 RT_REMREF(rt); 946 return (rt); 947 } 948} 949 950void 951arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 952{ 953 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 954 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 955 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 956 ifa->ifa_rtrequest = arp_rtrequest; 957 ifa->ifa_flags |= RTF_CLONING; 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 = arp_rtrequest; 967 ifa->ifa_flags |= RTF_CLONING; 968} 969 970static void 971arp_init(void) 972{ 973 974 arpintrq.ifq_maxlen = 50; 975 mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF); 976 netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE); 977} 978SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 979