if_ether.c revision 148883
1114407Snyan/*- 2114407Snyan * Copyright (c) 1982, 1986, 1988, 1993 3114407Snyan * The Regents of the University of California. All rights reserved. 4114407Snyan * 5114407Snyan * Redistribution and use in source and binary forms, with or without 6114407Snyan * modification, are permitted provided that the following conditions 7114407Snyan * are met: 8114407Snyan * 1. Redistributions of source code must retain the above copyright 9114407Snyan * notice, this list of conditions and the following disclaimer. 10114407Snyan * 2. Redistributions in binary form must reproduce the above copyright 11114407Snyan * notice, this list of conditions and the following disclaimer in the 12114407Snyan * documentation and/or other materials provided with the distribution. 13114407Snyan * 4. Neither the name of the University nor the names of its contributors 14114407Snyan * may be used to endorse or promote products derived from this software 15114407Snyan * without specific prior written permission. 16114407Snyan * 17114407Snyan * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18114407Snyan * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19114407Snyan * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20114407Snyan * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21114407Snyan * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22114407Snyan * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23114407Snyan * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24114407Snyan * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25114407Snyan * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26114407Snyan * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27119880Sobrien * SUCH DAMAGE. 28119880Sobrien * 29119880Sobrien * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 30114407Snyan * $FreeBSD: head/sys/netinet/if_ether.c 148883 2005-08-09 08:39:56Z glebius $ 31114407Snyan */ 32114407Snyan 33114407Snyan/* 34114407Snyan * Ethernet address resolution protocol. 35114407Snyan * TODO: 36114407Snyan * add "inuse/lock" bit (or ref. count) along with valid bit 37114407Snyan */ 38114407Snyan 39114407Snyan#include "opt_inet.h" 40114407Snyan#include "opt_bdg.h" 41114407Snyan#include "opt_mac.h" 42114407Snyan#include "opt_carp.h" 43114407Snyan 44114407Snyan#include <sys/param.h> 45114407Snyan#include <sys/kernel.h> 46114407Snyan#include <sys/queue.h> 47114407Snyan#include <sys/sysctl.h> 48114407Snyan#include <sys/systm.h> 49114407Snyan#include <sys/mac.h> 50114407Snyan#include <sys/mbuf.h> 51114407Snyan#include <sys/malloc.h> 52114407Snyan#include <sys/socket.h> 53114407Snyan#include <sys/syslog.h> 54114407Snyan 55114407Snyan#include <net/if.h> 56114407Snyan#include <net/if_dl.h> 57114407Snyan#include <net/if_types.h> 58114407Snyan#include <net/route.h> 59114407Snyan#include <net/netisr.h> 60114407Snyan#include <net/if_llc.h> 61114407Snyan#include <net/ethernet.h> 62114407Snyan#include <net/bridge.h> 63114407Snyan 64114407Snyan#include <netinet/in.h> 65114407Snyan#include <netinet/in_var.h> 66114407Snyan#include <netinet/if_ether.h> 67114407Snyan 68114407Snyan#include <net/if_arc.h> 69114407Snyan#include <net/iso88025.h> 70114407Snyan 71114407Snyan#ifdef DEV_CARP 72114407Snyan#include <netinet/ip_carp.h> 73114407Snyan#endif 74114407Snyan 75114407Snyan#define SIN(s) ((struct sockaddr_in *)s) 76114407Snyan#define SDL(s) ((struct sockaddr_dl *)s) 77114407Snyan 78114407SnyanSYSCTL_DECL(_net_link_ether); 79114407SnyanSYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 80114407Snyan 81114407Snyan/* timer values */ 82114407Snyanstatic int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 83114407Snyanstatic int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 84114407Snyanstatic int arpt_down = 20; /* once declared down, don't send for 20 sec */ 85114407Snyan 86114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW, 87114407Snyan &arpt_prune, 0, ""); 88114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 89114407Snyan &arpt_keep, 0, ""); 90114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW, 91114407Snyan &arpt_down, 0, ""); 92114407Snyan 93114407Snyan#define rt_expire rt_rmx.rmx_expire 94114407Snyan 95114407Snyanstruct llinfo_arp { 96114407Snyan LIST_ENTRY(llinfo_arp) la_le; 97114407Snyan struct rtentry *la_rt; 98114407Snyan struct mbuf *la_hold; /* last packet until resolved/timeout */ 99114407Snyan u_short la_preempt; /* countdown for pre-expiry arps */ 100114407Snyan u_short la_asked; /* #times we QUERIED following expiration */ 101114407Snyan#define la_timer la_rt->rt_rmx.rmx_expire /* deletion time in seconds */ 102114407Snyan}; 103114407Snyan 104114407Snyanstatic LIST_HEAD(, llinfo_arp) llinfo_arp; 105114407Snyan 106114407Snyanstatic struct ifqueue arpintrq; 107114407Snyanstatic int arp_allocated; 108114407Snyan 109114407Snyanstatic int arp_maxtries = 5; 110114407Snyanstatic int useloopback = 1; /* use loopback interface for local traffic */ 111114407Snyanstatic int arp_proxyall = 0; 112114407Snyanstatic struct callout arp_callout; 113114407Snyan 114114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 115114407Snyan &arp_maxtries, 0, ""); 116114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 117114407Snyan &useloopback, 0, ""); 118114407SnyanSYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 119 &arp_proxyall, 0, ""); 120 121static void arp_init(void); 122static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 123static void arprequest(struct ifnet *, 124 struct in_addr *, struct in_addr *, u_char *); 125static void arpintr(struct mbuf *); 126static void arptfree(struct llinfo_arp *); 127static void arptimer(void *); 128static struct llinfo_arp 129 *arplookup(u_long, int, int); 130#ifdef INET 131static void in_arpinput(struct mbuf *); 132#endif 133 134/* 135 * Timeout routine. Age arp_tab entries periodically. 136 */ 137/* ARGSUSED */ 138static void 139arptimer(ignored_arg) 140 void *ignored_arg; 141{ 142 struct llinfo_arp *la, *ola; 143 144 RADIX_NODE_HEAD_LOCK(rt_tables[AF_INET]); 145 la = LIST_FIRST(&llinfo_arp); 146 while (la != NULL) { 147 struct rtentry *rt = la->la_rt; 148 ola = la; 149 la = LIST_NEXT(la, la_le); 150 if (rt->rt_expire && rt->rt_expire <= time_second) 151 arptfree(ola); /* timer has expired, clear */ 152 } 153 RADIX_NODE_HEAD_UNLOCK(rt_tables[AF_INET]); 154 155 callout_reset(&arp_callout, arpt_prune * hz, arptimer, NULL); 156} 157 158/* 159 * Parallel to llc_rtrequest. 160 */ 161static void 162arp_rtrequest(req, rt, info) 163 int req; 164 struct rtentry *rt; 165 struct rt_addrinfo *info; 166{ 167 struct sockaddr *gate; 168 struct llinfo_arp *la; 169 static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; 170 struct in_ifaddr *ia; 171 struct ifaddr *ifa; 172 173 RT_LOCK_ASSERT(rt); 174 175 if (rt->rt_flags & RTF_GATEWAY) 176 return; 177 gate = rt->rt_gateway; 178 la = (struct llinfo_arp *)rt->rt_llinfo; 179 switch (req) { 180 181 case RTM_ADD: 182 /* 183 * XXX: If this is a manually added route to interface 184 * such as older version of routed or gated might provide, 185 * restore cloning bit. 186 */ 187 if ((rt->rt_flags & RTF_HOST) == 0 && 188 rt_mask(rt) != NULL && 189 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 190 rt->rt_flags |= RTF_CLONING; 191 if (rt->rt_flags & RTF_CLONING) { 192 /* 193 * Case 1: This route should come from a route to iface. 194 */ 195 rt_setgate(rt, rt_key(rt), 196 (struct sockaddr *)&null_sdl); 197 gate = rt->rt_gateway; 198 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 199 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 200 rt->rt_expire = time_second; 201 break; 202 } 203 /* Announce a new entry if requested. */ 204 if (rt->rt_flags & RTF_ANNOUNCE) 205 arprequest(rt->rt_ifp, 206 &SIN(rt_key(rt))->sin_addr, 207 &SIN(rt_key(rt))->sin_addr, 208 (u_char *)LLADDR(SDL(gate))); 209 /*FALLTHROUGH*/ 210 case RTM_RESOLVE: 211 if (gate->sa_family != AF_LINK || 212 gate->sa_len < sizeof(null_sdl)) { 213 log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, 214 inet_ntoa(SIN(rt_key(rt))->sin_addr), 215 (gate->sa_family != AF_LINK) ? 216 " (!AF_LINK)": ""); 217 break; 218 } 219 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 220 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 221 if (la != 0) 222 break; /* This happens on a route change */ 223 /* 224 * Case 2: This route may come from cloning, or a manual route 225 * add with a LL address. 226 */ 227 R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); 228 rt->rt_llinfo = (caddr_t)la; 229 if (la == 0) { 230 log(LOG_DEBUG, "%s: malloc failed\n", __func__); 231 break; 232 } 233 arp_allocated++; 234 la->la_rt = rt; 235 rt->rt_flags |= RTF_LLINFO; 236 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 237 LIST_INSERT_HEAD(&llinfo_arp, la, la_le); 238 239#ifdef INET 240 /* 241 * This keeps the multicast addresses from showing up 242 * in `arp -a' listings as unresolved. It's not actually 243 * functional. Then the same for broadcast. 244 */ 245 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && 246 rt->rt_ifp->if_type != IFT_ARCNET) { 247 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 248 LLADDR(SDL(gate))); 249 SDL(gate)->sdl_alen = 6; 250 rt->rt_expire = 0; 251 } 252 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 253 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 254 rt->rt_ifp->if_addrlen); 255 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 256 rt->rt_expire = 0; 257 } 258#endif 259 260 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 261 if (ia->ia_ifp == rt->rt_ifp && 262 SIN(rt_key(rt))->sin_addr.s_addr == 263 (IA_SIN(ia))->sin_addr.s_addr) 264 break; 265 } 266 if (ia) { 267 /* 268 * This test used to be 269 * if (loif.if_flags & IFF_UP) 270 * It allowed local traffic to be forced 271 * through the hardware by configuring the loopback down. 272 * However, it causes problems during network configuration 273 * for boards that can't receive packets they send. 274 * It is now necessary to clear "useloopback" and remove 275 * the route to force traffic out to the hardware. 276 */ 277 rt->rt_expire = 0; 278 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 279 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 280 if (useloopback) 281 rt->rt_ifp = loif; 282 283 /* 284 * make sure to set rt->rt_ifa to the interface 285 * address we are using, otherwise we will have trouble 286 * with source address selection. 287 */ 288 ifa = &ia->ia_ifa; 289 if (ifa != rt->rt_ifa) { 290 IFAFREE(rt->rt_ifa); 291 IFAREF(ifa); 292 rt->rt_ifa = ifa; 293 } 294 } 295 break; 296 297 case RTM_DELETE: 298 if (la == 0) 299 break; 300 RADIX_NODE_HEAD_LOCK_ASSERT(rt_tables[AF_INET]); 301 LIST_REMOVE(la, la_le); 302 rt->rt_llinfo = 0; 303 rt->rt_flags &= ~RTF_LLINFO; 304 if (la->la_hold) 305 m_freem(la->la_hold); 306 Free((caddr_t)la); 307 } 308} 309 310/* 311 * Broadcast an ARP request. Caller specifies: 312 * - arp header source ip address 313 * - arp header target ip address 314 * - arp header source ethernet address 315 */ 316static void 317arprequest(ifp, sip, tip, enaddr) 318 struct ifnet *ifp; 319 struct in_addr *sip, *tip; 320 u_char *enaddr; 321{ 322 struct mbuf *m; 323 struct arphdr *ah; 324 struct sockaddr sa; 325 326 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 327 return; 328 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 329 2*ifp->if_data.ifi_addrlen; 330 m->m_pkthdr.len = m->m_len; 331 MH_ALIGN(m, m->m_len); 332 ah = mtod(m, struct arphdr *); 333 bzero((caddr_t)ah, m->m_len); 334#ifdef MAC 335 mac_create_mbuf_linklayer(ifp, m); 336#endif 337 ah->ar_pro = htons(ETHERTYPE_IP); 338 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 339 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 340 ah->ar_op = htons(ARPOP_REQUEST); 341 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 342 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 343 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 344 sa.sa_family = AF_ARP; 345 sa.sa_len = 2; 346 m->m_flags |= M_BCAST; 347 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 348 349 return; 350} 351 352/* 353 * Resolve an IP address into an ethernet address. 354 * On input: 355 * ifp is the interface we use 356 * dst is the next hop, 357 * rt0 is the route to the final destination (possibly useless) 358 * m is the mbuf 359 * desten is where we want the address. 360 * 361 * On success, desten is filled in and the function returns 0; 362 * If the packet must be held pending resolution, we return EWOULDBLOCK 363 * On other errors, we return the corresponding error code. 364 */ 365int 366arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 367 struct sockaddr *dst, u_char *desten) 368{ 369 struct llinfo_arp *la = 0; 370 struct sockaddr_dl *sdl; 371 int error; 372 struct rtentry *rt; 373 374 error = rt_check(&rt, &rt0, dst); 375 if (error) { 376 m_freem(m); 377 return error; 378 } 379 RT_UNLOCK(rt); 380 381 if (m->m_flags & M_BCAST) { /* broadcast */ 382 (void)memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 383 return (0); 384 } 385 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) {/* multicast */ 386 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 387 return (0); 388 } 389 if (rt) 390 la = (struct llinfo_arp *)rt->rt_llinfo; 391 if (la == 0) { 392 la = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0); 393 if (la) 394 rt = la->la_rt; 395 } 396 if (la == 0 || rt == 0) { 397 log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n", 398 inet_ntoa(SIN(dst)->sin_addr), la ? "la" : "", 399 rt ? "rt" : ""); 400 m_freem(m); 401 return (EINVAL); /* XXX */ 402 } 403 sdl = SDL(rt->rt_gateway); 404 /* 405 * Check the address family and length is valid, the address 406 * is resolved; otherwise, try to resolve. 407 */ 408 if ((rt->rt_expire == 0 || rt->rt_expire > time_second) && 409 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 410 /* 411 * If entry has an expiry time and it is approaching, 412 * see if we need to send an ARP request within this 413 * arpt_down interval. 414 */ 415 if ((rt->rt_expire != 0) && 416 (time_second + la->la_preempt > rt->rt_expire)) { 417 arprequest(ifp, 418 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 419 &SIN(dst)->sin_addr, 420 IF_LLADDR(ifp)); 421 la->la_preempt--; 422 } 423 424 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 425 return (0); 426 } 427 /* 428 * If ARP is disabled or static on this interface, stop. 429 * XXX 430 * Probably should not allocate empty llinfo struct if we are 431 * not going to be sending out an arp request. 432 */ 433 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 434 m_freem(m); 435 return (EINVAL); 436 } 437 /* 438 * There is an arptab entry, but no ethernet address 439 * response yet. Replace the held mbuf with this 440 * latest one. 441 */ 442 if (la->la_hold) 443 m_freem(la->la_hold); 444 la->la_hold = m; 445 if (rt->rt_expire) { 446 RT_LOCK(rt); 447 rt->rt_flags &= ~RTF_REJECT; 448 if (la->la_asked == 0 || rt->rt_expire != time_second) { 449 rt->rt_expire = time_second; 450 if (la->la_asked++ < arp_maxtries) { 451 arprequest(ifp, 452 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 453 &SIN(dst)->sin_addr, 454 IF_LLADDR(ifp)); 455 } else { 456 rt->rt_flags |= RTF_REJECT; 457 rt->rt_expire += arpt_down; 458 la->la_asked = 0; 459 la->la_preempt = arp_maxtries; 460 } 461 462 } 463 RT_UNLOCK(rt); 464 } 465 return (EWOULDBLOCK); 466} 467 468/* 469 * Common length and type checks are done here, 470 * then the protocol-specific routine is called. 471 */ 472static void 473arpintr(struct mbuf *m) 474{ 475 struct arphdr *ar; 476 477 if (m->m_len < sizeof(struct arphdr) && 478 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 479 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 480 return; 481 } 482 ar = mtod(m, struct arphdr *); 483 484 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 485 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 486 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 487 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394) { 488 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 489 (unsigned char *)&ar->ar_hrd, ""); 490 m_freem(m); 491 return; 492 } 493 494 if (m->m_len < arphdr_len(ar)) { 495 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 496 log(LOG_ERR, "arp: runt packet\n"); 497 m_freem(m); 498 return; 499 } 500 ar = mtod(m, struct arphdr *); 501 } 502 503 switch (ntohs(ar->ar_pro)) { 504#ifdef INET 505 case ETHERTYPE_IP: 506 in_arpinput(m); 507 return; 508#endif 509 } 510 m_freem(m); 511} 512 513#ifdef INET 514/* 515 * ARP for Internet protocols on 10 Mb/s Ethernet. 516 * Algorithm is that given in RFC 826. 517 * In addition, a sanity check is performed on the sender 518 * protocol address, to catch impersonators. 519 * We no longer handle negotiations for use of trailer protocol: 520 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 521 * along with IP replies if we wanted trailers sent to us, 522 * and also sent them in response to IP replies. 523 * This allowed either end to announce the desire to receive 524 * trailer packets. 525 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 526 * but formerly didn't normally send requests. 527 */ 528static int log_arp_wrong_iface = 1; 529static int log_arp_movements = 1; 530 531SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 532 &log_arp_wrong_iface, 0, 533 "log arp packets arriving on the wrong interface"); 534SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 535 &log_arp_movements, 0, 536 "log arp replies from MACs different than the one in the cache"); 537 538 539static void 540in_arpinput(m) 541 struct mbuf *m; 542{ 543 struct arphdr *ah; 544 struct ifnet *ifp = m->m_pkthdr.rcvif; 545 struct iso88025_header *th = (struct iso88025_header *)0; 546 struct iso88025_sockaddr_dl_data *trld; 547 struct llinfo_arp *la = 0; 548 struct rtentry *rt; 549 struct ifaddr *ifa; 550 struct in_ifaddr *ia; 551 struct sockaddr_dl *sdl; 552 struct sockaddr sa; 553 struct in_addr isaddr, itaddr, myaddr; 554 u_int8_t *enaddr = NULL; 555 int op, rif_len; 556 int req_len; 557 int bridged = 0; 558#ifdef DEV_CARP 559 int carp_match = 0; 560#endif 561 562 if (do_bridge || ifp->if_bridge) 563 bridged = 1; 564 565 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 566 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 567 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 568 return; 569 } 570 571 ah = mtod(m, struct arphdr *); 572 op = ntohs(ah->ar_op); 573 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 574 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 575 576 /* 577 * For a bridge, we want to check the address irrespective 578 * of the receive interface. (This will change slightly 579 * when we have clusters of interfaces). 580 * If the interface does not match, but the recieving interface 581 * is part of carp, we call carp_iamatch to see if this is a 582 * request for the virtual host ip. 583 * XXX: This is really ugly! 584 */ 585 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 586 if ((bridged || (ia->ia_ifp == ifp)) && 587 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 588 goto match; 589#ifdef DEV_CARP 590 if (ifp->if_carp != NULL && 591 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 592 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 593 carp_match = 1; 594 goto match; 595 } 596#endif 597 } 598 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 599 if ((bridged || (ia->ia_ifp == ifp)) && 600 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 601 goto match; 602 /* 603 * No match, use the first inet address on the receive interface 604 * as a dummy address for the rest of the function. 605 */ 606 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 607 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 608 ia = ifatoia(ifa); 609 goto match; 610 } 611 /* 612 * If bridging, fall back to using any inet address. 613 */ 614 if (!bridged || (ia = TAILQ_FIRST(&in_ifaddrhead)) == NULL) 615 goto drop; 616match: 617 if (!enaddr) 618 enaddr = (u_int8_t *)IF_LLADDR(ifp); 619 myaddr = ia->ia_addr.sin_addr; 620 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 621 goto drop; /* it's from me, ignore it. */ 622 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 623 log(LOG_ERR, 624 "arp: link address is broadcast for IP address %s!\n", 625 inet_ntoa(isaddr)); 626 goto drop; 627 } 628 /* 629 * Warn if another host is using the same IP address, but only if the 630 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 631 * case we suppress the warning to avoid false positive complaints of 632 * potential misconfiguration. 633 */ 634 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 635 log(LOG_ERR, 636 "arp: %*D is using my IP address %s!\n", 637 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 638 inet_ntoa(isaddr)); 639 itaddr = myaddr; 640 goto reply; 641 } 642 if (ifp->if_flags & IFF_STATICARP) 643 goto reply; 644 la = arplookup(isaddr.s_addr, itaddr.s_addr == myaddr.s_addr, 0); 645 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) { 646 /* the following is not an error when doing bridging */ 647 if (!bridged && rt->rt_ifp != ifp 648#ifdef DEV_CARP 649 && (ifp->if_type != IFT_CARP || !carp_match) 650#endif 651 ) { 652 if (log_arp_wrong_iface) 653 log(LOG_ERR, "arp: %s is on %s but got reply from %*D on %s\n", 654 inet_ntoa(isaddr), 655 rt->rt_ifp->if_xname, 656 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 657 ifp->if_xname); 658 goto reply; 659 } 660 if (sdl->sdl_alen && 661 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 662 if (rt->rt_expire) { 663 if (log_arp_movements) 664 log(LOG_INFO, "arp: %s moved from %*D to %*D on %s\n", 665 inet_ntoa(isaddr), 666 ifp->if_addrlen, (u_char *)LLADDR(sdl), ":", 667 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 668 ifp->if_xname); 669 } else { 670 log(LOG_ERR, 671 "arp: %*D attempts to modify permanent entry for %s on %s\n", 672 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 673 inet_ntoa(isaddr), ifp->if_xname); 674 goto reply; 675 } 676 } 677 /* 678 * sanity check for the address length. 679 * XXX this does not work for protocols with variable address 680 * length. -is 681 */ 682 if (sdl->sdl_alen && 683 sdl->sdl_alen != ah->ar_hln) { 684 log(LOG_WARNING, 685 "arp from %*D: new addr len %d, was %d", 686 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 687 ah->ar_hln, sdl->sdl_alen); 688 } 689 if (ifp->if_addrlen != ah->ar_hln) { 690 log(LOG_WARNING, 691 "arp from %*D: addr len: new %d, i/f %d (ignored)", 692 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 693 ah->ar_hln, ifp->if_addrlen); 694 goto reply; 695 } 696 (void)memcpy(LLADDR(sdl), ar_sha(ah), 697 sdl->sdl_alen = ah->ar_hln); 698 /* 699 * If we receive an arp from a token-ring station over 700 * a token-ring nic then try to save the source 701 * routing info. 702 */ 703 if (ifp->if_type == IFT_ISO88025) { 704 th = (struct iso88025_header *)m->m_pkthdr.header; 705 trld = SDL_ISO88025(sdl); 706 rif_len = TR_RCF_RIFLEN(th->rcf); 707 if ((th->iso88025_shost[0] & TR_RII) && 708 (rif_len > 2)) { 709 trld->trld_rcf = th->rcf; 710 trld->trld_rcf ^= htons(TR_RCF_DIR); 711 memcpy(trld->trld_route, th->rd, rif_len - 2); 712 trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); 713 /* 714 * Set up source routing information for 715 * reply packet (XXX) 716 */ 717 m->m_data -= rif_len; 718 m->m_len += rif_len; 719 m->m_pkthdr.len += rif_len; 720 } else { 721 th->iso88025_shost[0] &= ~TR_RII; 722 trld->trld_rcf = 0; 723 } 724 m->m_data -= 8; 725 m->m_len += 8; 726 m->m_pkthdr.len += 8; 727 th->rcf = trld->trld_rcf; 728 } 729 RT_LOCK(rt); 730 if (rt->rt_expire) 731 rt->rt_expire = time_second + arpt_keep; 732 rt->rt_flags &= ~RTF_REJECT; 733 RT_UNLOCK(rt); 734 la->la_asked = 0; 735 la->la_preempt = arp_maxtries; 736 if (la->la_hold) { 737 (*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt); 738 la->la_hold = 0; 739 } 740 } 741reply: 742 if (op != ARPOP_REQUEST) 743 goto drop; 744 if (itaddr.s_addr == myaddr.s_addr) { 745 /* I am the target */ 746 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 747 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 748 } else { 749 la = arplookup(itaddr.s_addr, 0, SIN_PROXY); 750 if (la == NULL) { 751 struct sockaddr_in sin; 752 753 if (!arp_proxyall) 754 goto drop; 755 756 bzero(&sin, sizeof sin); 757 sin.sin_family = AF_INET; 758 sin.sin_len = sizeof sin; 759 sin.sin_addr = itaddr; 760 761 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 762 if (!rt) 763 goto drop; 764 /* 765 * Don't send proxies for nodes on the same interface 766 * as this one came out of, or we'll get into a fight 767 * over who claims what Ether address. 768 */ 769 if (rt->rt_ifp == ifp) { 770 rtfree(rt); 771 goto drop; 772 } 773 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 774 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 775 rtfree(rt); 776 777 /* 778 * Also check that the node which sent the ARP packet 779 * is on the the interface we expect it to be on. This 780 * avoids ARP chaos if an interface is connected to the 781 * wrong network. 782 */ 783 sin.sin_addr = isaddr; 784 785 rt = rtalloc1((struct sockaddr *)&sin, 0, 0UL); 786 if (!rt) 787 goto drop; 788 if (rt->rt_ifp != ifp) { 789 log(LOG_INFO, "arp_proxy: ignoring request" 790 " from %s via %s, expecting %s\n", 791 inet_ntoa(isaddr), ifp->if_xname, 792 rt->rt_ifp->if_xname); 793 rtfree(rt); 794 goto drop; 795 } 796 rtfree(rt); 797 798#ifdef DEBUG_PROXY 799 printf("arp: proxying for %s\n", 800 inet_ntoa(itaddr)); 801#endif 802 } else { 803 rt = la->la_rt; 804 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 805 sdl = SDL(rt->rt_gateway); 806 (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 807 } 808 } 809 810 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 811 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 812 ah->ar_op = htons(ARPOP_REPLY); 813 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 814 m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */ 815 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 816 m->m_pkthdr.len = m->m_len; 817 sa.sa_family = AF_ARP; 818 sa.sa_len = 2; 819 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 820 return; 821 822drop: 823 m_freem(m); 824} 825#endif 826 827/* 828 * Free an arp entry. 829 */ 830static void 831arptfree(la) 832 struct llinfo_arp *la; 833{ 834 struct rtentry *rt = la->la_rt; 835 struct sockaddr_dl *sdl; 836 837 if (rt == 0) 838 panic("arptfree"); 839 if (rt->rt_refcnt > 0 && (sdl = SDL(rt->rt_gateway)) && 840 sdl->sdl_family == AF_LINK) { 841 sdl->sdl_alen = 0; 842 la->la_preempt = la->la_asked = 0; 843 RT_LOCK(rt); /* XXX needed or move higher? */ 844 rt->rt_flags &= ~RTF_REJECT; 845 RT_UNLOCK(rt); 846 return; 847 } 848 rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, rt_mask(rt), 849 0, (struct rtentry **)0); 850} 851/* 852 * Lookup or enter a new address in arptab. 853 */ 854static struct llinfo_arp * 855arplookup(addr, create, proxy) 856 u_long addr; 857 int create, proxy; 858{ 859 struct rtentry *rt; 860 struct sockaddr_inarp sin; 861 const char *why = 0; 862 863 bzero(&sin, sizeof(sin)); 864 sin.sin_len = sizeof(sin); 865 sin.sin_family = AF_INET; 866 sin.sin_addr.s_addr = addr; 867 if (proxy) 868 sin.sin_other = SIN_PROXY; 869 rt = rtalloc1((struct sockaddr *)&sin, create, 0UL); 870 if (rt == 0) 871 return (0); 872 873 if (rt->rt_flags & RTF_GATEWAY) 874 why = "host is not on local network"; 875 else if ((rt->rt_flags & RTF_LLINFO) == 0) 876 why = "could not allocate llinfo"; 877 else if (rt->rt_gateway->sa_family != AF_LINK) 878 why = "gateway route is not ours"; 879 880 if (why) { 881#define ISDYNCLONE(_rt) \ 882 (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) 883 if (create) 884 log(LOG_DEBUG, "arplookup %s failed: %s\n", 885 inet_ntoa(sin.sin_addr), why); 886 /* 887 * If there are no references to this Layer 2 route, 888 * and it is a cloned route, and not static, and 889 * arplookup() is creating the route, then purge 890 * it from the routing table as it is probably bogus. 891 */ 892 if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) 893 rtexpunge(rt); 894 RTFREE_LOCKED(rt); 895 return (0); 896#undef ISDYNCLONE 897 } else { 898 RT_REMREF(rt); 899 RT_UNLOCK(rt); 900 return ((struct llinfo_arp *)rt->rt_llinfo); 901 } 902} 903 904void 905arp_ifinit(ifp, ifa) 906 struct ifnet *ifp; 907 struct ifaddr *ifa; 908{ 909 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 910 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 911 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 912 ifa->ifa_rtrequest = arp_rtrequest; 913 ifa->ifa_flags |= RTF_CLONING; 914} 915 916void 917arp_ifinit2(ifp, ifa, enaddr) 918 struct ifnet *ifp; 919 struct ifaddr *ifa; 920 u_char *enaddr; 921{ 922 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 923 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 924 &IA_SIN(ifa)->sin_addr, enaddr); 925 ifa->ifa_rtrequest = arp_rtrequest; 926 ifa->ifa_flags |= RTF_CLONING; 927} 928 929static void 930arp_init(void) 931{ 932 933 arpintrq.ifq_maxlen = 50; 934 mtx_init(&arpintrq.ifq_mtx, "arp_inq", NULL, MTX_DEF); 935 LIST_INIT(&llinfo_arp); 936 callout_init(&arp_callout, CALLOUT_MPSAFE); 937 netisr_register(NETISR_ARP, arpintr, &arpintrq, NETISR_MPSAFE); 938 callout_reset(&arp_callout, hz, arptimer, NULL); 939} 940SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 941