if_ether.c revision 166010
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: head/sys/netinet/if_ether.c 166010 2007-01-14 18:44:17Z maxim $ 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/mbuf.h> 49#include <sys/malloc.h> 50#include <sys/socket.h> 51#include <sys/syslog.h> 52 53#include <net/if.h> 54#include <net/if_dl.h> 55#include <net/if_types.h> 56#include <net/route.h> 57#include <net/netisr.h> 58#include <net/if_llc.h> 59#include <net/ethernet.h> 60 61#include <netinet/in.h> 62#include <netinet/in_var.h> 63#include <netinet/if_ether.h> 64 65#include <net/if_arc.h> 66#include <net/iso88025.h> 67 68#ifdef DEV_CARP 69#include <netinet/ip_carp.h> 70#endif 71 72#include <security/mac/mac_framework.h> 73 74#define SIN(s) ((struct sockaddr_in *)s) 75#define SDL(s) ((struct sockaddr_dl *)s) 76 77SYSCTL_DECL(_net_link_ether); 78SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 79 80/* timer values */ 81static int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 82static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 83 84SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW, 85 &arpt_prune, 0, "ARP table prune interval in seconds"); 86SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 87 &arpt_keep, 0, "ARP entry lifetime in seconds"); 88 89#define rt_expire rt_rmx.rmx_expire 90 91struct llinfo_arp { 92 LIST_ENTRY(llinfo_arp) la_le; 93 struct rtentry *la_rt; 94 struct mbuf *la_hold; /* last packet until resolved/timeout */ 95 u_short la_preempt; /* countdown for pre-expiry arps */ 96 u_short la_asked; /* # requests sent */ 97}; 98 99static LIST_HEAD(, llinfo_arp) llinfo_arp; 100 101static struct ifqueue arpintrq; 102static int arp_allocated; 103 104static int arp_maxtries = 5; 105static int useloopback = 1; /* use loopback interface for local traffic */ 106static int arp_proxyall = 0; 107static struct callout arp_callout; 108 109SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 110 &arp_maxtries, 0, "ARP resolution attempts before returning error"); 111SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 112 &useloopback, 0, "Use the loopback interface for local traffic"); 113SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 114 &arp_proxyall, 0, "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); 124#ifdef INET 125static void in_arpinput(struct mbuf *); 126#endif 127 128/* 129 * Timeout routine. Age arp_tab entries periodically. 130 */ 131/* ARGSUSED */ 132static void 133arptimer(void * __unused unused) 134{ 135 struct llinfo_arp *la, *ola; 136 137 RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]); 138 LIST_FOREACH_SAFE(la, &llinfo_arp, la_le, ola) { 139 struct rtentry *rt = la->la_rt; 140 141 RT_LOCK(rt); 142 if (rt->rt_expire && rt->rt_expire <= time_uptime) { 143 struct sockaddr_dl *sdl = SDL(rt->rt_gateway); 144 145 KASSERT(sdl->sdl_family == AF_LINK, ("sdl_family %d", 146 sdl->sdl_family)); 147 if (rt->rt_refcnt > 1) { 148 sdl->sdl_alen = 0; 149 la->la_preempt = la->la_asked = 0; 150 RT_UNLOCK(rt); 151 continue; 152 } 153 RT_UNLOCK(rt); 154 /* 155 * XXX: LIST_REMOVE() is deep inside rtrequest(). 156 */ 157 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, 158 NULL); 159 continue; 160 } 161 RT_UNLOCK(rt); 162 } 163 RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]); 164 165 callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL); 166} 167 168/* 169 * Parallel to llc_rtrequest. 170 */ 171static void 172arp_rtrequest(req, rt, info) 173 int req; 174 struct rtentry *rt; 175 struct rt_addrinfo *info; 176{ 177 struct sockaddr *gate; 178 struct llinfo_arp *la; 179 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 180 struct in_ifaddr *ia; 181 struct ifaddr *ifa; 182 183 RT_LOCK_ASSERT(rt); 184 185 if (rt->rt_flags & RTF_GATEWAY) 186 return; 187 gate = rt->rt_gateway; 188 la = (struct llinfo_arp *)rt->rt_llinfo; 189 switch (req) { 190 191 case RTM_ADD: 192 /* 193 * XXX: If this is a manually added route to interface 194 * such as older version of routed or gated might provide, 195 * restore cloning bit. 196 */ 197 if ((rt->rt_flags & RTF_HOST) == 0 && 198 rt_mask(rt) != NULL && 199 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 200 rt->rt_flags |= RTF_CLONING; 201 if (rt->rt_flags & RTF_CLONING) { 202 /* 203 * Case 1: This route should come from a route to iface. 204 */ 205 rt_setgate(rt, rt_key(rt), 206 (struct sockaddr *)&null_sdl); 207 gate = rt->rt_gateway; 208 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 209 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 210 rt->rt_expire = time_uptime; 211 break; 212 } 213 /* Announce a new entry if requested. */ 214 if (rt->rt_flags & RTF_ANNOUNCE) 215 arprequest(rt->rt_ifp, 216 &SIN(rt_key(rt))->sin_addr, 217 &SIN(rt_key(rt))->sin_addr, 218 (u_char *)LLADDR(SDL(gate))); 219 /*FALLTHROUGH*/ 220 case RTM_RESOLVE: 221 if (gate->sa_family != AF_LINK || 222 gate->sa_len < sizeof(null_sdl)) { 223 log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, 224 inet_ntoa(SIN(rt_key(rt))->sin_addr), 225 (gate->sa_family != AF_LINK) ? 226 " (!AF_LINK)": ""); 227 break; 228 } 229 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 230 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 231 if (la != 0) 232 break; /* This happens on a route change */ 233 /* 234 * Case 2: This route may come from cloning, or a manual route 235 * add with a LL address. 236 */ 237 R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); 238 rt->rt_llinfo = (caddr_t)la; 239 if (la == 0) { 240 log(LOG_DEBUG, "%s: malloc failed\n", __func__); 241 break; 242 } 243 arp_allocated++; 244 /* 245 * We are storing a route entry outside of radix tree. So, 246 * it can be found and accessed by other means than radix 247 * lookup. The routing code assumes that any rtentry detached 248 * from radix can be destroyed safely. To prevent this, we 249 * add an additional reference. 250 */ 251 RT_ADDREF(rt); 252 la->la_rt = rt; 253 rt->rt_flags |= RTF_LLINFO; 254 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 255 LIST_INSERT_HEAD(&llinfo_arp, la, la_le); 256 257#ifdef INET 258 /* 259 * This keeps the multicast addresses from showing up 260 * in `arp -a' listings as unresolved. It's not actually 261 * functional. Then the same for broadcast. 262 */ 263 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && 264 rt->rt_ifp->if_type != IFT_ARCNET) { 265 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 266 LLADDR(SDL(gate))); 267 SDL(gate)->sdl_alen = 6; 268 rt->rt_expire = 0; 269 } 270 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 271 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 272 rt->rt_ifp->if_addrlen); 273 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 274 rt->rt_expire = 0; 275 } 276#endif 277 278 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 279 if (ia->ia_ifp == rt->rt_ifp && 280 SIN(rt_key(rt))->sin_addr.s_addr == 281 (IA_SIN(ia))->sin_addr.s_addr) 282 break; 283 } 284 if (ia) { 285 /* 286 * This test used to be 287 * if (loif.if_flags & IFF_UP) 288 * It allowed local traffic to be forced 289 * through the hardware by configuring the loopback down. 290 * However, it causes problems during network configuration 291 * for boards that can't receive packets they send. 292 * It is now necessary to clear "useloopback" and remove 293 * the route to force traffic out to the hardware. 294 */ 295 rt->rt_expire = 0; 296 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 297 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 298 if (useloopback) { 299 rt->rt_ifp = loif; 300 rt->rt_rmx.rmx_mtu = loif->if_mtu; 301 } 302 303 /* 304 * make sure to set rt->rt_ifa to the interface 305 * address we are using, otherwise we will have trouble 306 * with source address selection. 307 */ 308 ifa = &ia->ia_ifa; 309 if (ifa != rt->rt_ifa) { 310 IFAFREE(rt->rt_ifa); 311 IFAREF(ifa); 312 rt->rt_ifa = ifa; 313 } 314 } 315 break; 316 317 case RTM_DELETE: 318 if (la == 0) 319 break; 320 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 321 LIST_REMOVE(la, la_le); 322 RT_REMREF(rt); 323 rt->rt_llinfo = 0; 324 rt->rt_flags &= ~RTF_LLINFO; 325 if (la->la_hold) 326 m_freem(la->la_hold); 327 Free((caddr_t)la); 328 } 329} 330 331/* 332 * Broadcast an ARP request. Caller specifies: 333 * - arp header source ip address 334 * - arp header target ip address 335 * - arp header source ethernet address 336 */ 337static void 338arprequest(ifp, sip, tip, enaddr) 339 struct ifnet *ifp; 340 struct in_addr *sip, *tip; 341 u_char *enaddr; 342{ 343 struct mbuf *m; 344 struct arphdr *ah; 345 struct sockaddr sa; 346 347 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 348 return; 349 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 350 2*ifp->if_data.ifi_addrlen; 351 m->m_pkthdr.len = m->m_len; 352 MH_ALIGN(m, m->m_len); 353 ah = mtod(m, struct arphdr *); 354 bzero((caddr_t)ah, m->m_len); 355#ifdef MAC 356 mac_create_mbuf_linklayer(ifp, m); 357#endif 358 ah->ar_pro = htons(ETHERTYPE_IP); 359 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 360 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 361 ah->ar_op = htons(ARPOP_REQUEST); 362 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 363 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 364 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 365 sa.sa_family = AF_ARP; 366 sa.sa_len = 2; 367 m->m_flags |= M_BCAST; 368 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 369 370 return; 371} 372 373/* 374 * Resolve an IP address into an ethernet address. 375 * On input: 376 * ifp is the interface we use 377 * dst is the next hop, 378 * rt0 is the route to the final destination (possibly useless) 379 * m is the mbuf 380 * desten is where we want the address. 381 * 382 * On success, desten is filled in and the function returns 0; 383 * If the packet must be held pending resolution, we return EWOULDBLOCK 384 * On other errors, we return the corresponding error code. 385 */ 386int 387arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 388 struct sockaddr *dst, u_char *desten) 389{ 390 struct llinfo_arp *la = NULL; 391 struct rtentry *rt = NULL; 392 struct sockaddr_dl *sdl; 393 int error; 394 395 if (m->m_flags & M_BCAST) { /* broadcast */ 396 (void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 397 return (0); 398 } 399 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */ 400 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 401 return (0); 402 } 403 404 if (rt0 != NULL) { 405 error = rt_check(&rt, &rt0, dst); 406 if (error) { 407 m_freem(m); 408 return error; 409 } 410 la = (struct llinfo_arp *)rt->rt_llinfo; 411 if (la == NULL) 412 RT_UNLOCK(rt); 413 } 414 if (la == NULL) { 415 /* 416 * We enter this block in case if rt0 was NULL, 417 * or if rt found by rt_check() didn't have llinfo. 418 */ 419 rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0); 420 if (rt == NULL) { 421 log(LOG_DEBUG, 422 "arpresolve: can't allocate route for %s\n", 423 inet_ntoa(SIN(dst)->sin_addr)); 424 m_freem(m); 425 return (EINVAL); /* XXX */ 426 } 427 la = (struct llinfo_arp *)rt->rt_llinfo; 428 if (la == NULL) { 429 RT_UNLOCK(rt); 430 log(LOG_DEBUG, 431 "arpresolve: can't allocate llinfo for %s\n", 432 inet_ntoa(SIN(dst)->sin_addr)); 433 m_freem(m); 434 return (EINVAL); /* XXX */ 435 } 436 } 437 sdl = SDL(rt->rt_gateway); 438 /* 439 * Check the address family and length is valid, the address 440 * is resolved; otherwise, try to resolve. 441 */ 442 if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) && 443 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 444 445 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 446 447 /* 448 * If entry has an expiry time and it is approaching, 449 * send an ARP request. 450 */ 451 if ((rt->rt_expire != 0) && 452 (time_uptime + la->la_preempt > rt->rt_expire)) { 453 struct in_addr sin = 454 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 455 456 la->la_preempt--; 457 RT_UNLOCK(rt); 458 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 459 IF_LLADDR(ifp)); 460 return (0); 461 } 462 463 RT_UNLOCK(rt); 464 return (0); 465 } 466 /* 467 * If ARP is disabled or static on this interface, stop. 468 * XXX 469 * Probably should not allocate empty llinfo struct if we are 470 * not going to be sending out an arp request. 471 */ 472 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 473 RT_UNLOCK(rt); 474 m_freem(m); 475 return (EINVAL); 476 } 477 /* 478 * There is an arptab entry, but no ethernet address 479 * response yet. Replace the held mbuf with this 480 * latest one. 481 */ 482 if (la->la_hold) 483 m_freem(la->la_hold); 484 la->la_hold = m; 485 486 KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry")); 487 488 /* 489 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 490 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 491 * if we have already sent arp_maxtries ARP requests. Retransmit the 492 * ARP request, but not faster than one request per second. 493 */ 494 if (la->la_asked < arp_maxtries) 495 error = EWOULDBLOCK; /* First request. */ 496 else 497 error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH; 498 499 if (la->la_asked == 0 || rt->rt_expire != time_uptime) { 500 struct in_addr sin = 501 SIN(rt->rt_ifa->ifa_addr)->sin_addr; 502 503 rt->rt_expire = time_uptime; 504 la->la_asked++; 505 RT_UNLOCK(rt); 506 507 arprequest(ifp, &sin, &SIN(dst)->sin_addr, 508 IF_LLADDR(ifp)); 509 } else 510 RT_UNLOCK(rt); 511 512 return (error); 513} 514 515/* 516 * Common length and type checks are done here, 517 * then the protocol-specific routine is called. 518 */ 519static void 520arpintr(struct mbuf *m) 521{ 522 struct arphdr *ar; 523 524 if (m->m_len < sizeof(struct arphdr) && 525 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 526 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 527 return; 528 } 529 ar = mtod(m, struct arphdr *); 530 531 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 532 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 533 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 534 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) { 535 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 536 (unsigned char *)&ar->ar_hrd, ""); 537 m_freem(m); 538 return; 539 } 540 541 if (m->m_len < arphdr_len(ar)) { 542 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 543 log(LOG_ERR, "arp: runt packet\n"); 544 m_freem(m); 545 return; 546 } 547 ar = mtod(m, struct arphdr *); 548 } 549 550 switch (ntohs(ar->ar_pro)) { 551#ifdef INET 552 case ETHERTYPE_IP: 553 in_arpinput(m); 554 return; 555#endif 556 } 557 m_freem(m); 558} 559 560#ifdef INET 561/* 562 * ARP for Internet protocols on 10 Mb/s Ethernet. 563 * Algorithm is that given in RFC 826. 564 * In addition, a sanity check is performed on the sender 565 * protocol address, to catch impersonators. 566 * We no longer handle negotiations for use of trailer protocol: 567 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 568 * along with IP replies if we wanted trailers sent to us, 569 * and also sent them in response to IP replies. 570 * This allowed either end to announce the desire to receive 571 * trailer packets. 572 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 573 * but formerly didn't normally send requests. 574 */ 575static int log_arp_wrong_iface = 1; 576static int log_arp_movements = 1; 577static int log_arp_permanent_modify = 1; 578 579SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 580 &log_arp_wrong_iface, 0, 581 "log arp packets arriving on the wrong interface"); 582SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 583 &log_arp_movements, 0, 584 "log arp replies from MACs different than the one in the cache"); 585SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 586 &log_arp_permanent_modify, 0, 587 "log arp replies from MACs different than the one in the permanent arp entry"); 588 589 590static void 591in_arpinput(m) 592 struct mbuf *m; 593{ 594 struct arphdr *ah; 595 struct ifnet *ifp = m->m_pkthdr.rcvif; 596 struct llinfo_arp *la; 597 struct rtentry *rt; 598 struct ifaddr *ifa; 599 struct in_ifaddr *ia; 600 struct sockaddr_dl *sdl; 601 struct sockaddr sa; 602 struct in_addr isaddr, itaddr, myaddr; 603 struct mbuf *hold; 604 u_int8_t *enaddr = NULL; 605 int op, rif_len; 606 int req_len; 607 int bridged = 0; 608#ifdef DEV_CARP 609 int carp_match = 0; 610#endif 611 612 if (ifp->if_bridge) 613 bridged = 1; 614 615 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 616 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 617 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 618 return; 619 } 620 621 ah = mtod(m, struct arphdr *); 622 op = ntohs(ah->ar_op); 623 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 624 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 625 626 /* 627 * For a bridge, we want to check the address irrespective 628 * of the receive interface. (This will change slightly 629 * when we have clusters of interfaces). 630 * If the interface does not match, but the recieving interface 631 * is part of carp, we call carp_iamatch to see if this is a 632 * request for the virtual host ip. 633 * XXX: This is really ugly! 634 */ 635 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 636 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 637 (ia->ia_ifp == ifp)) && 638 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 639 goto match; 640#ifdef DEV_CARP 641 if (ifp->if_carp != NULL && 642 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 643 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 644 carp_match = 1; 645 goto match; 646 } 647#endif 648 } 649 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 650 if (((bridged && ia->ia_ifp->if_bridge != NULL) || 651 (ia->ia_ifp == ifp)) && 652 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 653 goto match; 654 /* 655 * No match, use the first inet address on the receive interface 656 * as a dummy address for the rest of the function. 657 */ 658 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 659 if (ifa->ifa_addr->sa_family == AF_INET) { 660 ia = ifatoia(ifa); 661 goto match; 662 } 663 /* 664 * If bridging, fall back to using any inet address. 665 */ 666 if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) 667 goto drop; 668match: 669 if (!enaddr) 670 enaddr = (u_int8_t *)IF_LLADDR(ifp); 671 myaddr = ia->ia_addr.sin_addr; 672 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 673 goto drop; /* it's from me, ignore it. */ 674 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 675 log(LOG_ERR, 676 "arp: link address is broadcast for IP address %s!\n", 677 inet_ntoa(isaddr)); 678 goto drop; 679 } 680 /* 681 * Warn if another host is using the same IP address, but only if the 682 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 683 * case we suppress the warning to avoid false positive complaints of 684 * potential misconfiguration. 685 */ 686 if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 687 log(LOG_ERR, 688 "arp: %*D is using my IP address %s!\n", 689 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 690 inet_ntoa(isaddr)); 691 itaddr = myaddr; 692 goto reply; 693 } 694 if (ifp->if_flags & IFF_STATICARP) 695 goto reply; 696 rt = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0); 697 if (rt != NULL) { 698 la = (struct llinfo_arp *)rt->rt_llinfo; 699 if (la == NULL) { 700 RT_UNLOCK(rt); 701 goto reply; 702 } 703 } else 704 goto reply; 705 706 /* The following is not an error when doing bridging. */ 707 if (!bridged && rt->rt_ifp != ifp 708#ifdef DEV_CARP 709 && (ifp->if_type != IFT_CARP || !carp_match) 710#endif 711 ) { 712 if (log_arp_wrong_iface) 713 log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n", 714 inet_ntoa(isaddr), 715 rt->rt_ifp->if_xname, 716 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 717 ifp->if_xname); 718 RT_UNLOCK(rt); 719 goto reply; 720 } 721 sdl = SDL(rt->rt_gateway); 722 if (sdl->sdl_alen && 723 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 724 if (rt->rt_expire) { 725 if (log_arp_movements) 726 log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n", 727 inet_ntoa(isaddr), 728 ifp->if_addrlen, (u_char *)LLADDR(sdl), ":", 729 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 730 ifp->if_xname); 731 } else { 732 RT_UNLOCK(rt); 733 if (log_arp_permanent_modify) 734 log(LOG_ERR, "arp: %*D attempts to modify " 735 "permanent entry for %s on %s\n", 736 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 737 inet_ntoa(isaddr), ifp->if_xname); 738 goto reply; 739 } 740 } 741 /* 742 * sanity check for the address length. 743 * XXX this does not work for protocols with variable address 744 * length. -is 745 */ 746 if (sdl->sdl_alen && 747 sdl->sdl_alen != ah->ar_hln) { 748 log(LOG_WARNING, 749 "arp from %*D: new addr len %d, was %d", 750 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 751 ah->ar_hln, sdl->sdl_alen); 752 } 753 if (ifp->if_addrlen != ah->ar_hln) { 754 log(LOG_WARNING, 755 "arp from %*D: addr len: new %d, i/f %d (ignored)", 756 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 757 ah->ar_hln, ifp->if_addrlen); 758 RT_UNLOCK(rt); 759 goto reply; 760 } 761 (void)memcpy(LLADDR(sdl), ar_sha(ah), 762 sdl->sdl_alen = ah->ar_hln); 763 /* 764 * If we receive an arp from a token-ring station over 765 * a token-ring nic then try to save the source 766 * routing info. 767 */ 768 if (ifp->if_type == IFT_ISO88025) { 769 struct iso88025_header *th = NULL; 770 struct iso88025_sockaddr_dl_data *trld; 771 772 th = (struct iso88025_header *)m->m_pkthdr.header; 773 trld = SDL_ISO88025(sdl); 774 rif_len = TR_RCF_RIFLEN(th->rcf); 775 if ((th->iso88025_shost[0] & TR_RII) && 776 (rif_len > 2)) { 777 trld->trld_rcf = th->rcf; 778 trld->trld_rcf ^= htons(TR_RCF_DIR); 779 memcpy(trld->trld_route, th->rd, rif_len - 2); 780 trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); 781 /* 782 * Set up source routing information for 783 * reply packet (XXX) 784 */ 785 m->m_data -= rif_len; 786 m->m_len += rif_len; 787 m->m_pkthdr.len += rif_len; 788 } else { 789 th->iso88025_shost[0] &= ~TR_RII; 790 trld->trld_rcf = 0; 791 } 792 m->m_data -= 8; 793 m->m_len += 8; 794 m->m_pkthdr.len += 8; 795 th->rcf = trld->trld_rcf; 796 } 797 if (rt->rt_expire) 798 rt->rt_expire = time_uptime + arpt_keep; 799 la->la_asked = 0; 800 la->la_preempt = arp_maxtries; 801 hold = la->la_hold; 802 la->la_hold = NULL; 803 RT_UNLOCK(rt); 804 if (hold != NULL) 805 (*ifp->if_output)(ifp, hold, rt_key(rt), rt); 806 807reply: 808 if (op != ARPOP_REQUEST) 809 goto drop; 810 if (itaddr.s_addr == myaddr.s_addr) { 811 /* I am the target */ 812 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 813 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 814 } else { 815 rt = arplookup(itaddr.s_addr, 0, SIN_PROXY); 816 if (rt == NULL) { 817 struct sockaddr_in sin; 818 819 if (!arp_proxyall) 820 goto drop; 821 822 bzero(&sin, sizeof sin); 823 sin.sin_family = AF_INET; 824 sin.sin_len = sizeof sin; 825 sin.sin_addr = itaddr; 826 827 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 828 if (!rt) 829 goto drop; 830 /* 831 * Don't send proxies for nodes on the same interface 832 * as this one came out of, or we'll get into a fight 833 * over who claims what Ether address. 834 */ 835 if (rt->rt_ifp == ifp) { 836 rtfree(rt); 837 goto drop; 838 } 839 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 840 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 841 rtfree(rt); 842 843 /* 844 * Also check that the node which sent the ARP packet 845 * is on the the interface we expect it to be on. This 846 * avoids ARP chaos if an interface is connected to the 847 * wrong network. 848 */ 849 sin.sin_addr = isaddr; 850 851 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 852 if (!rt) 853 goto drop; 854 if (rt->rt_ifp != ifp) { 855 log(LOG_INFO, "arp_proxy: ignoring request" 856 " from %s via %s, expecting %s\n", 857 inet_ntoa(isaddr), ifp->if_xname, 858 rt->rt_ifp->if_xname); 859 rtfree(rt); 860 goto drop; 861 } 862 rtfree(rt); 863 864#ifdef DEBUG_PROXY 865 printf("arp: proxying for %s\n", 866 inet_ntoa(itaddr)); 867#endif 868 } else { 869 /* 870 * Return proxied ARP replies only on the interface 871 * or bridge cluster where this network resides. 872 * Otherwise we may conflict with the host we are 873 * proxying for. 874 */ 875 if (rt->rt_ifp != ifp && 876 (rt->rt_ifp->if_bridge != ifp->if_bridge || 877 ifp->if_bridge == NULL)) { 878 RT_UNLOCK(rt); 879 goto drop; 880 } 881 sdl = SDL(rt->rt_gateway); 882 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 883 (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 884 RT_UNLOCK(rt); 885 } 886 } 887 888 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 889 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 890 ah->ar_op = htons(ARPOP_REPLY); 891 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 892 m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */ 893 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 894 m->m_pkthdr.len = m->m_len; 895 sa.sa_family = AF_ARP; 896 sa.sa_len = 2; 897 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 898 return; 899 900drop: 901 m_freem(m); 902} 903#endif 904 905/* 906 * Lookup or enter a new address in arptab. 907 */ 908static struct rtentry * 909arplookup(addr, create, proxy) 910 u_long addr; 911 int create, proxy; 912{ 913 struct rtentry *rt; 914 struct sockaddr_inarp sin; 915 const char *why = 0; 916 917 bzero(&sin, sizeof(sin)); 918 sin.sin_len = sizeof(sin); 919 sin.sin_family = AF_INET; 920 sin.sin_addr.s_addr = addr; 921 if (proxy) 922 sin.sin_other = SIN_PROXY; 923 rt = rtalloc1((struct sockaddr *)&sin, create, 0UL); 924 if (rt == 0) 925 return (0); 926 927 if (rt->rt_flags & RTF_GATEWAY) 928 why = "host is not on local network"; 929 else if ((rt->rt_flags & RTF_LLINFO) == 0) 930 why = "could not allocate llinfo"; 931 else if (rt->rt_gateway->sa_family != AF_LINK) 932 why = "gateway route is not ours"; 933 934 if (why) { 935#define ISDYNCLONE(_rt) \ 936 (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) 937 if (create) 938 log(LOG_DEBUG, "arplookup %s failed: %s\n", 939 inet_ntoa(sin.sin_addr), why); 940 /* 941 * If there are no references to this Layer 2 route, 942 * and it is a cloned route, and not static, and 943 * arplookup() is creating the route, then purge 944 * it from the routing table as it is probably bogus. 945 */ 946 if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) 947 rtexpunge(rt); 948 RTFREE_LOCKED(rt); 949 return (0); 950#undef ISDYNCLONE 951 } else { 952 RT_REMREF(rt); 953 return (rt); 954 } 955} 956 957void 958arp_ifinit(ifp, ifa) 959 struct ifnet *ifp; 960 struct ifaddr *ifa; 961{ 962 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 963 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 964 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 965 ifa->ifa_rtrequest = arp_rtrequest; 966 ifa->ifa_flags |= RTF_CLONING; 967} 968 969void 970arp_ifinit2(ifp, ifa, enaddr) 971 struct ifnet *ifp; 972 struct ifaddr *ifa; 973 u_char *enaddr; 974{ 975 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 976 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 977 &IA_SIN(ifa)->sin_addr, enaddr); 978 ifa->ifa_rtrequest = arp_rtrequest; 979 ifa->ifa_flags |= RTF_CLONING; 980} 981 982static void 983arp_init(void) 984{ 985 986 arpintrq.ifq_maxlen = 50; 987 mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF); 988 LIST_INIT(&llinfo_arp); 989 callout_init(&arp_callout, CALLOUT_MPSAFE); 990 netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE); 991 callout_reset(&arp_callout, hz, arptimer, NULL); 992} 993SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 994