if_ether.c revision 227785
1184588Sdfr/*- 2184588Sdfr * Copyright (c) 1982, 1986, 1988, 1993 3184588Sdfr * The Regents of the University of California. All rights reserved. 4184588Sdfr * 5184588Sdfr * Redistribution and use in source and binary forms, with or without 6184588Sdfr * modification, are permitted provided that the following conditions 7184588Sdfr * are met: 8184588Sdfr * 1. Redistributions of source code must retain the above copyright 9184588Sdfr * notice, this list of conditions and the following disclaimer. 10184588Sdfr * 2. Redistributions in binary form must reproduce the above copyright 11184588Sdfr * notice, this list of conditions and the following disclaimer in the 12184588Sdfr * documentation and/or other materials provided with the distribution. 13184588Sdfr * 4. Neither the name of the University nor the names of its contributors 14184588Sdfr * may be used to endorse or promote products derived from this software 15184588Sdfr * without specific prior written permission. 16184588Sdfr * 17184588Sdfr * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18184588Sdfr * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19184588Sdfr * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20184588Sdfr * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21184588Sdfr * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22184588Sdfr * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23184588Sdfr * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24184588Sdfr * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25184588Sdfr * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26184588Sdfr * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27184588Sdfr * SUCH DAMAGE. 28184588Sdfr * 29184588Sdfr * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 30184588Sdfr */ 31184588Sdfr 32184588Sdfr/* 33184588Sdfr * Ethernet address resolution protocol. 34184588Sdfr * TODO: 35184588Sdfr * add "inuse/lock" bit (or ref. count) along with valid bit 36184588Sdfr */ 37184588Sdfr 38184588Sdfr#include <sys/cdefs.h> 39184588Sdfr__FBSDID("$FreeBSD: head/sys/netinet/if_ether.c 227785 2011-11-21 12:07:18Z glebius $"); 40184588Sdfr 41184588Sdfr#include "opt_inet.h" 42184588Sdfr 43184588Sdfr#include <sys/param.h> 44184588Sdfr#include <sys/kernel.h> 45184588Sdfr#include <sys/queue.h> 46184588Sdfr#include <sys/sysctl.h> 47184588Sdfr#include <sys/systm.h> 48184588Sdfr#include <sys/mbuf.h> 49184588Sdfr#include <sys/malloc.h> 50184588Sdfr#include <sys/proc.h> 51184588Sdfr#include <sys/socket.h> 52184588Sdfr#include <sys/syslog.h> 53184588Sdfr 54184588Sdfr#include <net/if.h> 55184588Sdfr#include <net/if_dl.h> 56184588Sdfr#include <net/if_types.h> 57184588Sdfr#include <net/netisr.h> 58184588Sdfr#include <net/if_llc.h> 59184588Sdfr#include <net/ethernet.h> 60184588Sdfr#include <net/route.h> 61184588Sdfr#include <net/vnet.h> 62184588Sdfr 63184588Sdfr#include <netinet/in.h> 64184588Sdfr#include <netinet/in_var.h> 65184588Sdfr#include <net/if_llatbl.h> 66184588Sdfr#include <netinet/if_ether.h> 67184588Sdfr#if defined(INET) || defined(INET6) 68184588Sdfr#include <netinet/ip_carp.h> 69184588Sdfr#endif 70184588Sdfr 71184588Sdfr#include <net/if_arc.h> 72184588Sdfr#include <net/iso88025.h> 73184588Sdfr 74184588Sdfr#include <security/mac/mac_framework.h> 75184588Sdfr 76184588Sdfr#define SIN(s) ((struct sockaddr_in *)s) 77184588Sdfr#define SDL(s) ((struct sockaddr_dl *)s) 78184588Sdfr 79184588SdfrSYSCTL_DECL(_net_link_ether); 80184588Sdfrstatic SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 81184588Sdfrstatic SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, ""); 82184588Sdfr 83184588Sdfr/* timer values */ 84184588Sdfrstatic VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20 85184588Sdfr * minutes */ 86184588Sdfrstatic VNET_DEFINE(int, arp_maxtries) = 5; 87184588SdfrVNET_DEFINE(int, useloopback) = 1; /* use loopback interface for 88184588Sdfr * local traffic */ 89184588Sdfrstatic VNET_DEFINE(int, arp_proxyall) = 0; 90184588Sdfrstatic VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for 91184588Sdfr * 20 seconds */ 92184588SdfrVNET_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ 93184588Sdfr 94184588Sdfrstatic VNET_DEFINE(int, arp_maxhold) = 1; 95184588Sdfr 96184588Sdfr#define V_arpt_keep VNET(arpt_keep) 97184588Sdfr#define V_arpt_down VNET(arpt_down) 98184588Sdfr#define V_arp_maxtries VNET(arp_maxtries) 99184588Sdfr#define V_arp_proxyall VNET(arp_proxyall) 100184588Sdfr#define V_arpstat VNET(arpstat) 101184588Sdfr#define V_arp_maxhold VNET(arp_maxhold) 102184588Sdfr 103184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 104184588Sdfr &VNET_NAME(arpt_keep), 0, 105184588Sdfr "ARP entry lifetime in seconds"); 106184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 107184588Sdfr &VNET_NAME(arp_maxtries), 0, 108184588Sdfr "ARP resolution attempts before returning error"); 109184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 110184588Sdfr &VNET_NAME(useloopback), 0, 111184588Sdfr "Use the loopback interface for local traffic"); 112184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 113184588Sdfr &VNET_NAME(arp_proxyall), 0, 114184588Sdfr "Enable proxy ARP for all suitable requests"); 115184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_RW, 116184588Sdfr &VNET_NAME(arpt_down), 0, 117184588Sdfr "Incomplete ARP entry lifetime in seconds"); 118184588SdfrSYSCTL_VNET_STRUCT(_net_link_ether_arp, OID_AUTO, stats, CTLFLAG_RW, 119184588Sdfr &VNET_NAME(arpstat), arpstat, 120184588Sdfr "ARP statistics (struct arpstat, net/if_arp.h)"); 121184588SdfrSYSCTL_VNET_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_RW, 122184588Sdfr &VNET_NAME(arp_maxhold), 0, 123184588Sdfr "Number of packets to hold per ARP entry"); 124184588Sdfr 125184588Sdfrstatic void arp_init(void); 126184588Sdfrvoid arprequest(struct ifnet *, 127184588Sdfr struct in_addr *, struct in_addr *, u_char *); 128184588Sdfrstatic void arpintr(struct mbuf *); 129184588Sdfrstatic void arptimer(void *); 130184588Sdfr#ifdef INET 131184588Sdfrstatic void in_arpinput(struct mbuf *); 132184588Sdfr#endif 133184588Sdfr 134184588Sdfrstatic const struct netisr_handler arp_nh = { 135184588Sdfr .nh_name = "arp", 136184588Sdfr .nh_handler = arpintr, 137184588Sdfr .nh_proto = NETISR_ARP, 138184588Sdfr .nh_policy = NETISR_POLICY_SOURCE, 139184588Sdfr}; 140184588Sdfr 141184588Sdfr#ifdef AF_INET 142184588Sdfrvoid arp_ifscrub(struct ifnet *ifp, uint32_t addr); 143184588Sdfr 144184588Sdfr/* 145184588Sdfr * called by in_ifscrub to remove entry from the table when 146184588Sdfr * the interface goes away 147184588Sdfr */ 148184588Sdfrvoid 149184588Sdfrarp_ifscrub(struct ifnet *ifp, uint32_t addr) 150184588Sdfr{ 151184588Sdfr struct sockaddr_in addr4; 152184588Sdfr 153184588Sdfr bzero((void *)&addr4, sizeof(addr4)); 154184588Sdfr addr4.sin_len = sizeof(addr4); 155184588Sdfr addr4.sin_family = AF_INET; 156184588Sdfr addr4.sin_addr.s_addr = addr; 157184588Sdfr IF_AFDATA_LOCK(ifp); 158184588Sdfr lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR), 159184588Sdfr (struct sockaddr *)&addr4); 160184588Sdfr IF_AFDATA_UNLOCK(ifp); 161184588Sdfr} 162184588Sdfr#endif 163184588Sdfr 164184588Sdfr/* 165184588Sdfr * Timeout routine. Age arp_tab entries periodically. 166184588Sdfr */ 167184588Sdfrstatic void 168184588Sdfrarptimer(void *arg) 169184588Sdfr{ 170184588Sdfr struct ifnet *ifp; 171184588Sdfr struct llentry *lle; 172184588Sdfr int pkts_dropped; 173184588Sdfr 174184588Sdfr KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 175184588Sdfr lle = (struct llentry *)arg; 176184588Sdfr ifp = lle->lle_tbl->llt_ifp; 177184588Sdfr CURVNET_SET(ifp->if_vnet); 178184588Sdfr IF_AFDATA_LOCK(ifp); 179184588Sdfr LLE_WLOCK(lle); 180184588Sdfr if (lle->la_flags & LLE_STATIC) 181184588Sdfr LLE_WUNLOCK(lle); 182184588Sdfr else { 183184588Sdfr if (!callout_pending(&lle->la_timer) && 184184588Sdfr callout_active(&lle->la_timer)) { 185184588Sdfr callout_stop(&lle->la_timer); 186184588Sdfr LLE_REMREF(lle); 187184588Sdfr pkts_dropped = llentry_free(lle); 188184588Sdfr ARPSTAT_ADD(dropped, pkts_dropped); 189184588Sdfr ARPSTAT_INC(timeouts); 190184588Sdfr } else { 191184588Sdfr#ifdef DIAGNOSTIC 192184588Sdfr struct sockaddr *l3addr = L3_ADDR(lle); 193184588Sdfr log(LOG_INFO, 194184588Sdfr "arptimer issue: %p, IPv4 address: \"%s\"\n", lle, 195184588Sdfr inet_ntoa( 196184588Sdfr ((const struct sockaddr_in *)l3addr)->sin_addr)); 197184588Sdfr#endif 198184588Sdfr LLE_WUNLOCK(lle); 199184588Sdfr } 200184588Sdfr } 201184588Sdfr IF_AFDATA_UNLOCK(ifp); 202184588Sdfr CURVNET_RESTORE(); 203184588Sdfr} 204184588Sdfr 205184588Sdfr/* 206184588Sdfr * Broadcast an ARP request. Caller specifies: 207184588Sdfr * - arp header source ip address 208184588Sdfr * - arp header target ip address 209184588Sdfr * - arp header source ethernet address 210184588Sdfr */ 211184588Sdfrvoid 212184588Sdfrarprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 213184588Sdfr u_char *enaddr) 214184588Sdfr{ 215184588Sdfr struct mbuf *m; 216184588Sdfr struct arphdr *ah; 217184588Sdfr struct sockaddr sa; 218184588Sdfr 219184588Sdfr if (sip == NULL) { 220184588Sdfr /* XXX don't believe this can happen (or explain why) */ 221 /* 222 * The caller did not supply a source address, try to find 223 * a compatible one among those assigned to this interface. 224 */ 225 struct ifaddr *ifa; 226 227 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 228 if (!ifa->ifa_addr || 229 ifa->ifa_addr->sa_family != AF_INET) 230 continue; 231 sip = &SIN(ifa->ifa_addr)->sin_addr; 232 if (0 == ((sip->s_addr ^ tip->s_addr) & 233 SIN(ifa->ifa_netmask)->sin_addr.s_addr) ) 234 break; /* found it. */ 235 } 236 if (sip == NULL) { 237 printf("%s: cannot find matching address\n", __func__); 238 return; 239 } 240 } 241 242 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 243 return; 244 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 245 2*ifp->if_data.ifi_addrlen; 246 m->m_pkthdr.len = m->m_len; 247 MH_ALIGN(m, m->m_len); 248 ah = mtod(m, struct arphdr *); 249 bzero((caddr_t)ah, m->m_len); 250#ifdef MAC 251 mac_netinet_arp_send(ifp, m); 252#endif 253 ah->ar_pro = htons(ETHERTYPE_IP); 254 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 255 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 256 ah->ar_op = htons(ARPOP_REQUEST); 257 bcopy((caddr_t)enaddr, (caddr_t)ar_sha(ah), ah->ar_hln); 258 bcopy((caddr_t)sip, (caddr_t)ar_spa(ah), ah->ar_pln); 259 bcopy((caddr_t)tip, (caddr_t)ar_tpa(ah), ah->ar_pln); 260 sa.sa_family = AF_ARP; 261 sa.sa_len = 2; 262 m->m_flags |= M_BCAST; 263 (*ifp->if_output)(ifp, m, &sa, NULL); 264 ARPSTAT_INC(txrequests); 265} 266 267/* 268 * Resolve an IP address into an ethernet address. 269 * On input: 270 * ifp is the interface we use 271 * rt0 is the route to the final destination (possibly useless) 272 * m is the mbuf. May be NULL if we don't have a packet. 273 * dst is the next hop, 274 * desten is where we want the address. 275 * 276 * On success, desten is filled in and the function returns 0; 277 * If the packet must be held pending resolution, we return EWOULDBLOCK 278 * On other errors, we return the corresponding error code. 279 * Note that m_freem() handles NULL. 280 */ 281int 282arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 283 struct sockaddr *dst, u_char *desten, struct llentry **lle) 284{ 285 struct llentry *la = 0; 286 u_int flags = 0; 287 struct mbuf *curr = NULL; 288 struct mbuf *next = NULL; 289 int error, renew; 290 291 *lle = NULL; 292 if (m != NULL) { 293 if (m->m_flags & M_BCAST) { 294 /* broadcast */ 295 (void)memcpy(desten, 296 ifp->if_broadcastaddr, ifp->if_addrlen); 297 return (0); 298 } 299 if (m->m_flags & M_MCAST && ifp->if_type != IFT_ARCNET) { 300 /* multicast */ 301 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 302 return (0); 303 } 304 } 305retry: 306 IF_AFDATA_RLOCK(ifp); 307 la = lla_lookup(LLTABLE(ifp), flags, dst); 308 IF_AFDATA_RUNLOCK(ifp); 309 if ((la == NULL) && ((flags & LLE_EXCLUSIVE) == 0) 310 && ((ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0)) { 311 flags |= (LLE_CREATE | LLE_EXCLUSIVE); 312 IF_AFDATA_WLOCK(ifp); 313 la = lla_lookup(LLTABLE(ifp), flags, dst); 314 IF_AFDATA_WUNLOCK(ifp); 315 } 316 if (la == NULL) { 317 if (flags & LLE_CREATE) 318 log(LOG_DEBUG, 319 "arpresolve: can't allocate llinfo for %s\n", 320 inet_ntoa(SIN(dst)->sin_addr)); 321 m_freem(m); 322 return (EINVAL); 323 } 324 325 if ((la->la_flags & LLE_VALID) && 326 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { 327 bcopy(&la->ll_addr, desten, ifp->if_addrlen); 328 /* 329 * If entry has an expiry time and it is approaching, 330 * see if we need to send an ARP request within this 331 * arpt_down interval. 332 */ 333 if (!(la->la_flags & LLE_STATIC) && 334 time_uptime + la->la_preempt > la->la_expire) { 335 arprequest(ifp, NULL, 336 &SIN(dst)->sin_addr, IF_LLADDR(ifp)); 337 338 la->la_preempt--; 339 } 340 341 *lle = la; 342 error = 0; 343 goto done; 344 } 345 346 if (la->la_flags & LLE_STATIC) { /* should not happen! */ 347 log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n", 348 inet_ntoa(SIN(dst)->sin_addr)); 349 m_freem(m); 350 error = EINVAL; 351 goto done; 352 } 353 354 renew = (la->la_asked == 0 || la->la_expire != time_uptime); 355 if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) { 356 flags |= LLE_EXCLUSIVE; 357 LLE_RUNLOCK(la); 358 goto retry; 359 } 360 /* 361 * There is an arptab entry, but no ethernet address 362 * response yet. Add the mbuf to the list, dropping 363 * the oldest packet if we have exceeded the system 364 * setting. 365 */ 366 if (m != NULL) { 367 if (la->la_numheld >= V_arp_maxhold) { 368 if (la->la_hold != NULL) { 369 next = la->la_hold->m_nextpkt; 370 m_freem(la->la_hold); 371 la->la_hold = next; 372 la->la_numheld--; 373 ARPSTAT_INC(dropped); 374 } 375 } 376 if (la->la_hold != NULL) { 377 curr = la->la_hold; 378 while (curr->m_nextpkt != NULL) 379 curr = curr->m_nextpkt; 380 curr->m_nextpkt = m; 381 } else 382 la->la_hold = m; 383 la->la_numheld++; 384 if (renew == 0 && (flags & LLE_EXCLUSIVE)) { 385 flags &= ~LLE_EXCLUSIVE; 386 LLE_DOWNGRADE(la); 387 } 388 389 } 390 /* 391 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 392 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 393 * if we have already sent arp_maxtries ARP requests. Retransmit the 394 * ARP request, but not faster than one request per second. 395 */ 396 if (la->la_asked < V_arp_maxtries) 397 error = EWOULDBLOCK; /* First request. */ 398 else 399 error = rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) ? 400 EHOSTUNREACH : EHOSTDOWN; 401 402 if (renew) { 403 int canceled; 404 405 LLE_ADDREF(la); 406 la->la_expire = time_uptime; 407 canceled = callout_reset(&la->la_timer, hz * V_arpt_down, 408 arptimer, la); 409 if (canceled) 410 LLE_REMREF(la); 411 la->la_asked++; 412 LLE_WUNLOCK(la); 413 arprequest(ifp, NULL, &SIN(dst)->sin_addr, 414 IF_LLADDR(ifp)); 415 return (error); 416 } 417done: 418 if (flags & LLE_EXCLUSIVE) 419 LLE_WUNLOCK(la); 420 else 421 LLE_RUNLOCK(la); 422 return (error); 423} 424 425/* 426 * Common length and type checks are done here, 427 * then the protocol-specific routine is called. 428 */ 429static void 430arpintr(struct mbuf *m) 431{ 432 struct arphdr *ar; 433 434 if (m->m_len < sizeof(struct arphdr) && 435 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 436 log(LOG_NOTICE, "arp: runt packet -- m_pullup failed\n"); 437 return; 438 } 439 ar = mtod(m, struct arphdr *); 440 441 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER && 442 ntohs(ar->ar_hrd) != ARPHRD_IEEE802 && 443 ntohs(ar->ar_hrd) != ARPHRD_ARCNET && 444 ntohs(ar->ar_hrd) != ARPHRD_IEEE1394 && 445 ntohs(ar->ar_hrd) != ARPHRD_INFINIBAND) { 446 log(LOG_NOTICE, "arp: unknown hardware address format (0x%2D)\n", 447 (unsigned char *)&ar->ar_hrd, ""); 448 m_freem(m); 449 return; 450 } 451 452 if (m->m_len < arphdr_len(ar)) { 453 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 454 log(LOG_NOTICE, "arp: runt packet\n"); 455 m_freem(m); 456 return; 457 } 458 ar = mtod(m, struct arphdr *); 459 } 460 461 ARPSTAT_INC(received); 462 switch (ntohs(ar->ar_pro)) { 463#ifdef INET 464 case ETHERTYPE_IP: 465 in_arpinput(m); 466 return; 467#endif 468 } 469 m_freem(m); 470} 471 472#ifdef INET 473/* 474 * ARP for Internet protocols on 10 Mb/s Ethernet. 475 * Algorithm is that given in RFC 826. 476 * In addition, a sanity check is performed on the sender 477 * protocol address, to catch impersonators. 478 * We no longer handle negotiations for use of trailer protocol: 479 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 480 * along with IP replies if we wanted trailers sent to us, 481 * and also sent them in response to IP replies. 482 * This allowed either end to announce the desire to receive 483 * trailer packets. 484 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 485 * but formerly didn't normally send requests. 486 */ 487static int log_arp_wrong_iface = 1; 488static int log_arp_movements = 1; 489static int log_arp_permanent_modify = 1; 490 491SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 492 &log_arp_wrong_iface, 0, 493 "log arp packets arriving on the wrong interface"); 494SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 495 &log_arp_movements, 0, 496 "log arp replies from MACs different than the one in the cache"); 497SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 498 &log_arp_permanent_modify, 0, 499 "log arp replies from MACs different than the one in the permanent arp entry"); 500 501 502static void 503in_arpinput(struct mbuf *m) 504{ 505 struct arphdr *ah; 506 struct ifnet *ifp = m->m_pkthdr.rcvif; 507 struct llentry *la = NULL; 508 struct rtentry *rt; 509 struct ifaddr *ifa; 510 struct in_ifaddr *ia; 511 struct sockaddr sa; 512 struct in_addr isaddr, itaddr, myaddr; 513 u_int8_t *enaddr = NULL; 514 int op, flags; 515 int req_len; 516 int bridged = 0, is_bridge = 0; 517 int carp_match = 0; 518 struct sockaddr_in sin; 519 sin.sin_len = sizeof(struct sockaddr_in); 520 sin.sin_family = AF_INET; 521 sin.sin_addr.s_addr = 0; 522 523 if (ifp->if_bridge) 524 bridged = 1; 525 if (ifp->if_type == IFT_BRIDGE) 526 is_bridge = 1; 527 528 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 529 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 530 log(LOG_NOTICE, "in_arp: runt packet -- m_pullup failed\n"); 531 return; 532 } 533 534 ah = mtod(m, struct arphdr *); 535 /* 536 * ARP is only for IPv4 so we can reject packets with 537 * a protocol length not equal to an IPv4 address. 538 */ 539 if (ah->ar_pln != sizeof(struct in_addr)) { 540 log(LOG_NOTICE, "in_arp: requested protocol length != %zu\n", 541 sizeof(struct in_addr)); 542 return; 543 } 544 545 if (ETHER_IS_MULTICAST(ar_sha(ah))) { 546 log(LOG_NOTICE, "in_arp: %*D is multicast\n", 547 ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); 548 return; 549 } 550 551 op = ntohs(ah->ar_op); 552 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 553 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 554 555 if (op == ARPOP_REPLY) 556 ARPSTAT_INC(rxreplies); 557 558 /* 559 * For a bridge, we want to check the address irrespective 560 * of the receive interface. (This will change slightly 561 * when we have clusters of interfaces). 562 * If the interface does not match, but the recieving interface 563 * is part of carp, we call carp_iamatch to see if this is a 564 * request for the virtual host ip. 565 * XXX: This is really ugly! 566 */ 567 IN_IFADDR_RLOCK(); 568 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 569 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 570 ia->ia_ifp == ifp) && 571 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 572 ifa_ref(&ia->ia_ifa); 573 IN_IFADDR_RUNLOCK(); 574 goto match; 575 } 576 if (ifp->if_carp != NULL && 577 (*carp_iamatch_p)(ifp, ia, &isaddr, &enaddr) && 578 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 579 carp_match = 1; 580 ifa_ref(&ia->ia_ifa); 581 IN_IFADDR_RUNLOCK(); 582 goto match; 583 } 584 } 585 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 586 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 587 ia->ia_ifp == ifp) && 588 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 589 ifa_ref(&ia->ia_ifa); 590 IN_IFADDR_RUNLOCK(); 591 goto match; 592 } 593 594#define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 595 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 596 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 597 addr == ia->ia_addr.sin_addr.s_addr) 598 /* 599 * Check the case when bridge shares its MAC address with 600 * some of its children, so packets are claimed by bridge 601 * itself (bridge_input() does it first), but they are really 602 * meant to be destined to the bridge member. 603 */ 604 if (is_bridge) { 605 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 606 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 607 ifa_ref(&ia->ia_ifa); 608 ifp = ia->ia_ifp; 609 IN_IFADDR_RUNLOCK(); 610 goto match; 611 } 612 } 613 } 614#undef BDG_MEMBER_MATCHES_ARP 615 IN_IFADDR_RUNLOCK(); 616 617 /* 618 * No match, use the first inet address on the receive interface 619 * as a dummy address for the rest of the function. 620 */ 621 IF_ADDR_LOCK(ifp); 622 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 623 if (ifa->ifa_addr->sa_family == AF_INET) { 624 ia = ifatoia(ifa); 625 ifa_ref(ifa); 626 IF_ADDR_UNLOCK(ifp); 627 goto match; 628 } 629 IF_ADDR_UNLOCK(ifp); 630 631 /* 632 * If bridging, fall back to using any inet address. 633 */ 634 IN_IFADDR_RLOCK(); 635 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { 636 IN_IFADDR_RUNLOCK(); 637 goto drop; 638 } 639 ifa_ref(&ia->ia_ifa); 640 IN_IFADDR_RUNLOCK(); 641match: 642 if (!enaddr) 643 enaddr = (u_int8_t *)IF_LLADDR(ifp); 644 myaddr = ia->ia_addr.sin_addr; 645 ifa_free(&ia->ia_ifa); 646 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 647 goto drop; /* it's from me, ignore it. */ 648 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 649 log(LOG_NOTICE, 650 "arp: link address is broadcast for IP address %s!\n", 651 inet_ntoa(isaddr)); 652 goto drop; 653 } 654 /* 655 * Warn if another host is using the same IP address, but only if the 656 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 657 * case we suppress the warning to avoid false positive complaints of 658 * potential misconfiguration. 659 */ 660 if (!bridged && isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 661 log(LOG_ERR, 662 "arp: %*D is using my IP address %s on %s!\n", 663 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 664 inet_ntoa(isaddr), ifp->if_xname); 665 itaddr = myaddr; 666 ARPSTAT_INC(dupips); 667 goto reply; 668 } 669 if (ifp->if_flags & IFF_STATICARP) 670 goto reply; 671 672 bzero(&sin, sizeof(sin)); 673 sin.sin_len = sizeof(struct sockaddr_in); 674 sin.sin_family = AF_INET; 675 sin.sin_addr = isaddr; 676 flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0; 677 flags |= LLE_EXCLUSIVE; 678 IF_AFDATA_LOCK(ifp); 679 la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin); 680 IF_AFDATA_UNLOCK(ifp); 681 if (la != NULL) { 682 /* the following is not an error when doing bridging */ 683 if (!bridged && la->lle_tbl->llt_ifp != ifp && !carp_match) { 684 if (log_arp_wrong_iface) 685 log(LOG_WARNING, "arp: %s is on %s " 686 "but got reply from %*D on %s\n", 687 inet_ntoa(isaddr), 688 la->lle_tbl->llt_ifp->if_xname, 689 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 690 ifp->if_xname); 691 LLE_WUNLOCK(la); 692 goto reply; 693 } 694 if ((la->la_flags & LLE_VALID) && 695 bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) { 696 if (la->la_flags & LLE_STATIC) { 697 LLE_WUNLOCK(la); 698 if (log_arp_permanent_modify) 699 log(LOG_ERR, 700 "arp: %*D attempts to modify " 701 "permanent entry for %s on %s\n", 702 ifp->if_addrlen, 703 (u_char *)ar_sha(ah), ":", 704 inet_ntoa(isaddr), ifp->if_xname); 705 goto reply; 706 } 707 if (log_arp_movements) { 708 log(LOG_INFO, "arp: %s moved from %*D " 709 "to %*D on %s\n", 710 inet_ntoa(isaddr), 711 ifp->if_addrlen, 712 (u_char *)&la->ll_addr, ":", 713 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 714 ifp->if_xname); 715 } 716 } 717 718 if (ifp->if_addrlen != ah->ar_hln) { 719 LLE_WUNLOCK(la); 720 log(LOG_WARNING, "arp from %*D: addr len: new %d, " 721 "i/f %d (ignored)\n", ifp->if_addrlen, 722 (u_char *) ar_sha(ah), ":", ah->ar_hln, 723 ifp->if_addrlen); 724 goto drop; 725 } 726 (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen); 727 la->la_flags |= LLE_VALID; 728 729 EVENTHANDLER_INVOKE(arp_update_event, la); 730 731 if (!(la->la_flags & LLE_STATIC)) { 732 int canceled; 733 734 LLE_ADDREF(la); 735 la->la_expire = time_uptime + V_arpt_keep; 736 canceled = callout_reset(&la->la_timer, 737 hz * V_arpt_keep, arptimer, la); 738 if (canceled) 739 LLE_REMREF(la); 740 } 741 la->la_asked = 0; 742 la->la_preempt = V_arp_maxtries; 743 /* 744 * The packets are all freed within the call to the output 745 * routine. 746 * 747 * NB: The lock MUST be released before the call to the 748 * output routine. 749 */ 750 if (la->la_hold != NULL) { 751 struct mbuf *m_hold, *m_hold_next; 752 753 m_hold = la->la_hold; 754 la->la_hold = NULL; 755 la->la_numheld = 0; 756 memcpy(&sa, L3_ADDR(la), sizeof(sa)); 757 LLE_WUNLOCK(la); 758 for (; m_hold != NULL; m_hold = m_hold_next) { 759 m_hold_next = m_hold->m_nextpkt; 760 m_hold->m_nextpkt = NULL; 761 (*ifp->if_output)(ifp, m_hold, &sa, NULL); 762 } 763 } else 764 LLE_WUNLOCK(la); 765 } 766reply: 767 if (op != ARPOP_REQUEST) 768 goto drop; 769 ARPSTAT_INC(rxrequests); 770 771 if (itaddr.s_addr == myaddr.s_addr) { 772 /* Shortcut.. the receiving interface is the target. */ 773 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 774 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 775 } else { 776 struct llentry *lle = NULL; 777 778 sin.sin_addr = itaddr; 779 IF_AFDATA_LOCK(ifp); 780 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); 781 IF_AFDATA_UNLOCK(ifp); 782 783 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { 784 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 785 (void)memcpy(ar_sha(ah), &lle->ll_addr, ah->ar_hln); 786 LLE_RUNLOCK(lle); 787 } else { 788 789 if (lle != NULL) 790 LLE_RUNLOCK(lle); 791 792 if (!V_arp_proxyall) 793 goto drop; 794 795 sin.sin_addr = itaddr; 796 /* XXX MRT use table 0 for arp reply */ 797 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 798 if (!rt) 799 goto drop; 800 801 /* 802 * Don't send proxies for nodes on the same interface 803 * as this one came out of, or we'll get into a fight 804 * over who claims what Ether address. 805 */ 806 if (!rt->rt_ifp || rt->rt_ifp == ifp) { 807 RTFREE_LOCKED(rt); 808 goto drop; 809 } 810 RTFREE_LOCKED(rt); 811 812 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 813 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 814 815 /* 816 * Also check that the node which sent the ARP packet 817 * is on the interface we expect it to be on. This 818 * avoids ARP chaos if an interface is connected to the 819 * wrong network. 820 */ 821 sin.sin_addr = isaddr; 822 823 /* XXX MRT use table 0 for arp checks */ 824 rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); 825 if (!rt) 826 goto drop; 827 if (rt->rt_ifp != ifp) { 828 log(LOG_INFO, "arp_proxy: ignoring request" 829 " from %s via %s, expecting %s\n", 830 inet_ntoa(isaddr), ifp->if_xname, 831 rt->rt_ifp->if_xname); 832 RTFREE_LOCKED(rt); 833 goto drop; 834 } 835 RTFREE_LOCKED(rt); 836 837#ifdef DEBUG_PROXY 838 printf("arp: proxying for %s\n", 839 inet_ntoa(itaddr)); 840#endif 841 } 842 } 843 844 if (itaddr.s_addr == myaddr.s_addr && 845 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 846 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 847#ifdef DEBUG_LINKLOCAL 848 printf("arp: sending reply for link-local addr %s\n", 849 inet_ntoa(itaddr)); 850#endif 851 m->m_flags |= M_BCAST; 852 m->m_flags &= ~M_MCAST; 853 } else { 854 /* default behaviour; never reply by broadcast. */ 855 m->m_flags &= ~(M_BCAST|M_MCAST); 856 } 857 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 858 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 859 ah->ar_op = htons(ARPOP_REPLY); 860 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 861 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 862 m->m_pkthdr.len = m->m_len; 863 m->m_pkthdr.rcvif = NULL; 864 sa.sa_family = AF_ARP; 865 sa.sa_len = 2; 866 (*ifp->if_output)(ifp, m, &sa, NULL); 867 ARPSTAT_INC(txreplies); 868 return; 869 870drop: 871 m_freem(m); 872} 873#endif 874 875void 876arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 877{ 878 struct llentry *lle; 879 880 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) { 881 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 882 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 883 /* 884 * interface address is considered static entry 885 * because the output of the arp utility shows 886 * that L2 entry as permanent 887 */ 888 IF_AFDATA_LOCK(ifp); 889 lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC), 890 (struct sockaddr *)IA_SIN(ifa)); 891 IF_AFDATA_UNLOCK(ifp); 892 if (lle == NULL) 893 log(LOG_INFO, "arp_ifinit: cannot create arp " 894 "entry for interface address\n"); 895 else 896 LLE_RUNLOCK(lle); 897 } 898 ifa->ifa_rtrequest = NULL; 899} 900 901void 902arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 903{ 904 if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) 905 arprequest(ifp, &IA_SIN(ifa)->sin_addr, 906 &IA_SIN(ifa)->sin_addr, enaddr); 907 ifa->ifa_rtrequest = NULL; 908} 909 910static void 911arp_init(void) 912{ 913 914 netisr_register(&arp_nh); 915} 916SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 917