1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 32 */ 33 34 /* 35 * Ethernet address resolution protocol. 36 * TODO: 37 * add "inuse/lock" bit (or ref. count) along with valid bit 38 */ 39 40 #include <sys/cdefs.h> 41 __FBSDID("$FreeBSD$"); 42 43 #include "opt_inet.h" 44 45 #include <sys/param.h> 46 #include <sys/eventhandler.h> 47 #include <sys/kernel.h> 48 #include <sys/lock.h> 49 #include <sys/queue.h> 50 #include <sys/sysctl.h> 51 #include <sys/systm.h> 52 #include <sys/mbuf.h> 53 #include <sys/malloc.h> 54 #include <sys/proc.h> 55 #include <sys/rmlock.h> 56 #include <sys/socket.h> 57 #include <sys/syslog.h> 58 59 #include <net/if.h> 60 #include <net/if_var.h> 61 #include <net/if_dl.h> 62 #include <net/if_types.h> 63 #include <net/netisr.h> 64 #include <net/ethernet.h> 65 #include <net/route.h> 66 #include <net/route/nhop.h> 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_fib.h> 71 #include <netinet/in_var.h> 72 #include <net/if_llatbl.h> 73 #include <netinet/if_ether.h> 74 #ifdef INET 75 #include <netinet/ip_carp.h> 76 #endif 77 78 #include <security/mac/mac_framework.h> 79 80 #define SIN(s) ((const struct sockaddr_in *)(s)) 81 82 static struct timeval arp_lastlog; 83 static int arp_curpps; 84 static int arp_maxpps = 1; 85 86 /* Simple ARP state machine */ 87 enum arp_llinfo_state { 88 ARP_LLINFO_INCOMPLETE = 0, /* No LLE data */ 89 ARP_LLINFO_REACHABLE, /* LLE is valid */ 90 ARP_LLINFO_VERIFY, /* LLE is valid, need refresh */ 91 ARP_LLINFO_DELETED, /* LLE is deleted */ 92 }; 93 94 SYSCTL_DECL(_net_link_ether); 95 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, 96 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 97 ""); 98 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, 99 CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 100 ""); 101 102 /* timer values */ 103 VNET_DEFINE_STATIC(int, arpt_keep) = (20*60); /* once resolved, good for 20 104 * minutes */ 105 VNET_DEFINE_STATIC(int, arp_maxtries) = 5; 106 VNET_DEFINE_STATIC(int, arp_proxyall) = 0; 107 VNET_DEFINE_STATIC(int, arpt_down) = 20; /* keep incomplete entries for 108 * 20 seconds */ 109 VNET_DEFINE_STATIC(int, arpt_rexmit) = 1; /* retransmit arp entries, sec*/ 110 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ 111 VNET_PCPUSTAT_SYSINIT(arpstat); 112 113 #ifdef VIMAGE 114 VNET_PCPUSTAT_SYSUNINIT(arpstat); 115 #endif /* VIMAGE */ 116 117 VNET_DEFINE_STATIC(int, arp_maxhold) = 1; 118 119 #define V_arpt_keep VNET(arpt_keep) 120 #define V_arpt_down VNET(arpt_down) 121 #define V_arpt_rexmit VNET(arpt_rexmit) 122 #define V_arp_maxtries VNET(arp_maxtries) 123 #define V_arp_proxyall VNET(arp_proxyall) 124 #define V_arp_maxhold VNET(arp_maxhold) 125 126 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW, 127 &VNET_NAME(arpt_keep), 0, 128 "ARP entry lifetime in seconds"); 129 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW, 130 &VNET_NAME(arp_maxtries), 0, 131 "ARP resolution attempts before returning error"); 132 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW, 133 &VNET_NAME(arp_proxyall), 0, 134 "Enable proxy ARP for all suitable requests"); 135 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW, 136 &VNET_NAME(arpt_down), 0, 137 "Incomplete ARP entry lifetime in seconds"); 138 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat, 139 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)"); 140 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW, 141 &VNET_NAME(arp_maxhold), 0, 142 "Number of packets to hold per ARP entry"); 143 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, 144 CTLFLAG_RW, &arp_maxpps, 0, 145 "Maximum number of remotely triggered ARP messages that can be " 146 "logged per second"); 147 148 /* 149 * Due to the exponential backoff algorithm used for the interval between GARP 150 * retransmissions, the maximum number of retransmissions is limited for 151 * sanity. This limit corresponds to a maximum interval between retransmissions 152 * of 2^16 seconds ~= 18 hours. 153 * 154 * Making this limit more dynamic is more complicated than worthwhile, 155 * especially since sending out GARPs spaced days apart would be of little 156 * use. A maximum dynamic limit would look something like: 157 * 158 * const int max = fls(INT_MAX / hz) - 1; 159 */ 160 #define MAX_GARP_RETRANSMITS 16 161 static int sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS); 162 static int garp_rexmit_count = 0; /* GARP retransmission setting. */ 163 164 SYSCTL_PROC(_net_link_ether_inet, OID_AUTO, garp_rexmit_count, 165 CTLTYPE_INT|CTLFLAG_RW|CTLFLAG_MPSAFE, 166 &garp_rexmit_count, 0, sysctl_garp_rexmit, "I", 167 "Number of times to retransmit GARP packets;" 168 " 0 to disable, maximum of 16"); 169 170 VNET_DEFINE_STATIC(int, arp_log_level) = LOG_INFO; /* Min. log(9) level. */ 171 #define V_arp_log_level VNET(arp_log_level) 172 SYSCTL_INT(_net_link_ether_arp, OID_AUTO, log_level, CTLFLAG_VNET | CTLFLAG_RW, 173 &VNET_NAME(arp_log_level), 0, 174 "Minimum log(9) level for recording rate limited arp log messages. " 175 "The higher will be log more (emerg=0, info=6 (default), debug=7)."); 176 #define ARP_LOG(pri, ...) do { \ 177 if ((pri) <= V_arp_log_level && \ 178 ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ 179 log((pri), "arp: " __VA_ARGS__); \ 180 } while (0) 181 182 183 static void arpintr(struct mbuf *); 184 static void arptimer(void *); 185 #ifdef INET 186 static void in_arpinput(struct mbuf *); 187 #endif 188 189 static void arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, 190 struct ifnet *ifp, int bridged, struct llentry *la); 191 static void arp_mark_lle_reachable(struct llentry *la); 192 static void arp_iflladdr(void *arg __unused, struct ifnet *ifp); 193 194 static eventhandler_tag iflladdr_tag; 195 196 static const struct netisr_handler arp_nh = { 197 .nh_name = "arp", 198 .nh_handler = arpintr, 199 .nh_proto = NETISR_ARP, 200 .nh_policy = NETISR_POLICY_SOURCE, 201 }; 202 203 /* 204 * Timeout routine. Age arp_tab entries periodically. 205 */ 206 static void 207 arptimer(void *arg) 208 { 209 struct llentry *lle = (struct llentry *)arg; 210 struct ifnet *ifp; 211 int r_skip_req; 212 213 if (lle->la_flags & LLE_STATIC) { 214 return; 215 } 216 LLE_WLOCK(lle); 217 if (callout_pending(&lle->lle_timer)) { 218 /* 219 * Here we are a bit odd here in the treatment of 220 * active/pending. If the pending bit is set, it got 221 * rescheduled before I ran. The active 222 * bit we ignore, since if it was stopped 223 * in ll_tablefree() and was currently running 224 * it would have return 0 so the code would 225 * not have deleted it since the callout could 226 * not be stopped so we want to go through 227 * with the delete here now. If the callout 228 * was restarted, the pending bit will be back on and 229 * we just want to bail since the callout_reset would 230 * return 1 and our reference would have been removed 231 * by arpresolve() below. 232 */ 233 LLE_WUNLOCK(lle); 234 return; 235 } 236 ifp = lle->lle_tbl->llt_ifp; 237 CURVNET_SET(ifp->if_vnet); 238 239 switch (lle->ln_state) { 240 case ARP_LLINFO_REACHABLE: 241 242 /* 243 * Expiration time is approaching. 244 * Let's try to refresh entry if it is still 245 * in use. 246 * 247 * Set r_skip_req to get feedback from 248 * fast path. Change state and re-schedule 249 * ourselves. 250 */ 251 LLE_REQ_LOCK(lle); 252 lle->r_skip_req = 1; 253 LLE_REQ_UNLOCK(lle); 254 lle->ln_state = ARP_LLINFO_VERIFY; 255 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 256 LLE_WUNLOCK(lle); 257 CURVNET_RESTORE(); 258 return; 259 case ARP_LLINFO_VERIFY: 260 LLE_REQ_LOCK(lle); 261 r_skip_req = lle->r_skip_req; 262 LLE_REQ_UNLOCK(lle); 263 264 if (r_skip_req == 0 && lle->la_preempt > 0) { 265 /* Entry was used, issue refresh request */ 266 struct epoch_tracker et; 267 struct in_addr dst; 268 269 dst = lle->r_l3addr.addr4; 270 lle->la_preempt--; 271 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 272 LLE_WUNLOCK(lle); 273 NET_EPOCH_ENTER(et); 274 arprequest(ifp, NULL, &dst, NULL); 275 NET_EPOCH_EXIT(et); 276 CURVNET_RESTORE(); 277 return; 278 } 279 /* Nothing happened. Reschedule if not too late */ 280 if (lle->la_expire > time_uptime) { 281 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 282 LLE_WUNLOCK(lle); 283 CURVNET_RESTORE(); 284 return; 285 } 286 break; 287 case ARP_LLINFO_INCOMPLETE: 288 case ARP_LLINFO_DELETED: 289 break; 290 } 291 292 if ((lle->la_flags & LLE_DELETED) == 0) { 293 int evt; 294 295 if (lle->la_flags & LLE_VALID) 296 evt = LLENTRY_EXPIRED; 297 else 298 evt = LLENTRY_TIMEDOUT; 299 EVENTHANDLER_INVOKE(lle_event, lle, evt); 300 } 301 302 callout_stop(&lle->lle_timer); 303 304 /* XXX: LOR avoidance. We still have ref on lle. */ 305 LLE_WUNLOCK(lle); 306 IF_AFDATA_LOCK(ifp); 307 LLE_WLOCK(lle); 308 309 /* Guard against race with other llentry_free(). */ 310 if (lle->la_flags & LLE_LINKED) { 311 LLE_REMREF(lle); 312 lltable_unlink_entry(lle->lle_tbl, lle); 313 } 314 IF_AFDATA_UNLOCK(ifp); 315 316 size_t pkts_dropped = llentry_free(lle); 317 318 ARPSTAT_ADD(dropped, pkts_dropped); 319 ARPSTAT_INC(timeouts); 320 321 CURVNET_RESTORE(); 322 } 323 324 /* 325 * Stores link-layer header for @ifp in format suitable for if_output() 326 * into buffer @buf. Resulting header length is stored in @bufsize. 327 * 328 * Returns 0 on success. 329 */ 330 static int 331 arp_fillheader(struct ifnet *ifp, struct arphdr *ah, int bcast, u_char *buf, 332 size_t *bufsize) 333 { 334 struct if_encap_req ereq; 335 int error; 336 337 bzero(buf, *bufsize); 338 bzero(&ereq, sizeof(ereq)); 339 ereq.buf = buf; 340 ereq.bufsize = *bufsize; 341 ereq.rtype = IFENCAP_LL; 342 ereq.family = AF_ARP; 343 ereq.lladdr = ar_tha(ah); 344 ereq.hdata = (u_char *)ah; 345 if (bcast) 346 ereq.flags = IFENCAP_FLAG_BROADCAST; 347 error = ifp->if_requestencap(ifp, &ereq); 348 if (error == 0) 349 *bufsize = ereq.bufsize; 350 351 return (error); 352 } 353 354 355 /* 356 * Broadcast an ARP request. Caller specifies: 357 * - arp header source ip address 358 * - arp header target ip address 359 * - arp header source ethernet address 360 */ 361 static int 362 arprequest_internal(struct ifnet *ifp, const struct in_addr *sip, 363 const struct in_addr *tip, u_char *enaddr) 364 { 365 struct mbuf *m; 366 struct arphdr *ah; 367 struct sockaddr sa; 368 u_char *carpaddr = NULL; 369 uint8_t linkhdr[LLE_MAX_LINKHDR]; 370 size_t linkhdrsize; 371 struct route ro; 372 int error; 373 374 NET_EPOCH_ASSERT(); 375 376 if (sip == NULL) { 377 /* 378 * The caller did not supply a source address, try to find 379 * a compatible one among those assigned to this interface. 380 */ 381 struct ifaddr *ifa; 382 383 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 384 if (ifa->ifa_addr->sa_family != AF_INET) 385 continue; 386 387 if (ifa->ifa_carp) { 388 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0) 389 continue; 390 sip = &IA_SIN(ifa)->sin_addr; 391 } else { 392 carpaddr = NULL; 393 sip = &IA_SIN(ifa)->sin_addr; 394 } 395 396 if (0 == ((sip->s_addr ^ tip->s_addr) & 397 IA_MASKSIN(ifa)->sin_addr.s_addr)) 398 break; /* found it. */ 399 } 400 if (sip == NULL) { 401 printf("%s: cannot find matching address\n", __func__); 402 return (EADDRNOTAVAIL); 403 } 404 } 405 if (enaddr == NULL) 406 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp); 407 408 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) 409 return (ENOMEM); 410 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 411 2 * ifp->if_addrlen; 412 m->m_pkthdr.len = m->m_len; 413 M_ALIGN(m, m->m_len); 414 ah = mtod(m, struct arphdr *); 415 bzero((caddr_t)ah, m->m_len); 416 #ifdef MAC 417 mac_netinet_arp_send(ifp, m); 418 #endif 419 ah->ar_pro = htons(ETHERTYPE_IP); 420 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 421 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 422 ah->ar_op = htons(ARPOP_REQUEST); 423 bcopy(enaddr, ar_sha(ah), ah->ar_hln); 424 bcopy(sip, ar_spa(ah), ah->ar_pln); 425 bcopy(tip, ar_tpa(ah), ah->ar_pln); 426 sa.sa_family = AF_ARP; 427 sa.sa_len = 2; 428 429 /* Calculate link header for sending frame */ 430 bzero(&ro, sizeof(ro)); 431 linkhdrsize = sizeof(linkhdr); 432 error = arp_fillheader(ifp, ah, 1, linkhdr, &linkhdrsize); 433 if (error != 0 && error != EAFNOSUPPORT) { 434 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n", 435 if_name(ifp), error); 436 return (error); 437 } 438 439 ro.ro_prepend = linkhdr; 440 ro.ro_plen = linkhdrsize; 441 ro.ro_flags = 0; 442 443 m->m_flags |= M_BCAST; 444 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 445 error = (*ifp->if_output)(ifp, m, &sa, &ro); 446 ARPSTAT_INC(txrequests); 447 if (error) { 448 ARPSTAT_INC(txerrors); 449 ARP_LOG(LOG_DEBUG, "Failed to send ARP packet on %s: %d\n", 450 if_name(ifp), error); 451 } 452 return (error); 453 } 454 455 void 456 arprequest(struct ifnet *ifp, const struct in_addr *sip, 457 const struct in_addr *tip, u_char *enaddr) 458 { 459 460 (void) arprequest_internal(ifp, sip, tip, enaddr); 461 } 462 463 /* 464 * Resolve an IP address into an ethernet address - heavy version. 465 * Used internally by arpresolve(). 466 * We have already checked that we can't use an existing lle without 467 * modification so we have to acquire an LLE_EXCLUSIVE lle lock. 468 * 469 * On success, desten and pflags are filled in and the function returns 0; 470 * If the packet must be held pending resolution, we return EWOULDBLOCK 471 * On other errors, we return the corresponding error code. 472 * Note that m_freem() handles NULL. 473 */ 474 static int 475 arpresolve_full(struct ifnet *ifp, int is_gw, int flags, struct mbuf *m, 476 const struct sockaddr *dst, u_char *desten, uint32_t *pflags, 477 struct llentry **plle) 478 { 479 struct llentry *la = NULL, *la_tmp; 480 struct mbuf *curr = NULL; 481 struct mbuf *next = NULL; 482 int error, renew; 483 char *lladdr; 484 int ll_len; 485 486 NET_EPOCH_ASSERT(); 487 488 if (pflags != NULL) 489 *pflags = 0; 490 if (plle != NULL) 491 *plle = NULL; 492 493 if ((flags & LLE_CREATE) == 0) 494 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 495 if (la == NULL && (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) { 496 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst); 497 if (la == NULL) { 498 char addrbuf[INET_ADDRSTRLEN]; 499 500 log(LOG_DEBUG, 501 "arpresolve: can't allocate llinfo for %s on %s\n", 502 inet_ntoa_r(SIN(dst)->sin_addr, addrbuf), 503 if_name(ifp)); 504 m_freem(m); 505 return (EINVAL); 506 } 507 508 IF_AFDATA_WLOCK(ifp); 509 LLE_WLOCK(la); 510 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 511 /* Prefer ANY existing lle over newly-created one */ 512 if (la_tmp == NULL) 513 lltable_link_entry(LLTABLE(ifp), la); 514 IF_AFDATA_WUNLOCK(ifp); 515 if (la_tmp != NULL) { 516 lltable_free_entry(LLTABLE(ifp), la); 517 la = la_tmp; 518 } 519 } 520 if (la == NULL) { 521 m_freem(m); 522 return (EINVAL); 523 } 524 525 if ((la->la_flags & LLE_VALID) && 526 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { 527 if (flags & LLE_ADDRONLY) { 528 lladdr = la->ll_addr; 529 ll_len = ifp->if_addrlen; 530 } else { 531 lladdr = la->r_linkdata; 532 ll_len = la->r_hdrlen; 533 } 534 bcopy(lladdr, desten, ll_len); 535 536 /* Notify LLE code that the entry was used by datapath */ 537 llentry_mark_used(la); 538 if (pflags != NULL) 539 *pflags = la->la_flags & (LLE_VALID|LLE_IFADDR); 540 if (plle) { 541 LLE_ADDREF(la); 542 *plle = la; 543 } 544 LLE_WUNLOCK(la); 545 return (0); 546 } 547 548 renew = (la->la_asked == 0 || la->la_expire != time_uptime); 549 /* 550 * There is an arptab entry, but no ethernet address 551 * response yet. Add the mbuf to the list, dropping 552 * the oldest packet if we have exceeded the system 553 * setting. 554 */ 555 if (m != NULL) { 556 if (la->la_numheld >= V_arp_maxhold) { 557 if (la->la_hold != NULL) { 558 next = la->la_hold->m_nextpkt; 559 m_freem(la->la_hold); 560 la->la_hold = next; 561 la->la_numheld--; 562 ARPSTAT_INC(dropped); 563 } 564 } 565 if (la->la_hold != NULL) { 566 curr = la->la_hold; 567 while (curr->m_nextpkt != NULL) 568 curr = curr->m_nextpkt; 569 curr->m_nextpkt = m; 570 } else 571 la->la_hold = m; 572 la->la_numheld++; 573 } 574 /* 575 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 576 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 577 * if we have already sent arp_maxtries ARP requests. Retransmit the 578 * ARP request, but not faster than one request per second. 579 */ 580 if (la->la_asked < V_arp_maxtries) 581 error = EWOULDBLOCK; /* First request. */ 582 else 583 error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN; 584 585 if (renew) { 586 int canceled, e; 587 588 LLE_ADDREF(la); 589 la->la_expire = time_uptime; 590 canceled = callout_reset(&la->lle_timer, hz * V_arpt_down, 591 arptimer, la); 592 if (canceled) 593 LLE_REMREF(la); 594 la->la_asked++; 595 LLE_WUNLOCK(la); 596 e = arprequest_internal(ifp, NULL, &SIN(dst)->sin_addr, NULL); 597 /* 598 * Only overwrite 'error' in case of error; in case of success 599 * the proper return value was already set above. 600 */ 601 if (e != 0) 602 return (e); 603 return (error); 604 } 605 606 LLE_WUNLOCK(la); 607 return (error); 608 } 609 610 /* 611 * Lookups link header based on an IP address. 612 * On input: 613 * ifp is the interface we use 614 * is_gw != 0 if @dst represents gateway to some destination 615 * m is the mbuf. May be NULL if we don't have a packet. 616 * dst is the next hop, 617 * desten is the storage to put LL header. 618 * flags returns subset of lle flags: LLE_VALID | LLE_IFADDR 619 * 620 * On success, full/partial link header and flags are filled in and 621 * the function returns 0. 622 * If the packet must be held pending resolution, we return EWOULDBLOCK 623 * On other errors, we return the corresponding error code. 624 * Note that m_freem() handles NULL. 625 */ 626 int 627 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 628 const struct sockaddr *dst, u_char *desten, uint32_t *pflags, 629 struct llentry **plle) 630 { 631 struct llentry *la = NULL; 632 633 NET_EPOCH_ASSERT(); 634 635 if (pflags != NULL) 636 *pflags = 0; 637 if (plle != NULL) 638 *plle = NULL; 639 640 if (m != NULL) { 641 if (m->m_flags & M_BCAST) { 642 /* broadcast */ 643 (void)memcpy(desten, 644 ifp->if_broadcastaddr, ifp->if_addrlen); 645 return (0); 646 } 647 if (m->m_flags & M_MCAST) { 648 /* multicast */ 649 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 650 return (0); 651 } 652 } 653 654 la = lla_lookup(LLTABLE(ifp), plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, dst); 655 if (la != NULL && (la->r_flags & RLLE_VALID) != 0) { 656 /* Entry found, let's copy lle info */ 657 bcopy(la->r_linkdata, desten, la->r_hdrlen); 658 if (pflags != NULL) 659 *pflags = LLE_VALID | (la->r_flags & RLLE_IFADDR); 660 /* Notify the LLE handling code that the entry was used. */ 661 llentry_mark_used(la); 662 if (plle) { 663 LLE_ADDREF(la); 664 *plle = la; 665 LLE_WUNLOCK(la); 666 } 667 return (0); 668 } 669 if (plle && la) 670 LLE_WUNLOCK(la); 671 672 return (arpresolve_full(ifp, is_gw, la == NULL ? LLE_CREATE : 0, m, dst, 673 desten, pflags, plle)); 674 } 675 676 /* 677 * Common length and type checks are done here, 678 * then the protocol-specific routine is called. 679 */ 680 static void 681 arpintr(struct mbuf *m) 682 { 683 struct arphdr *ar; 684 struct ifnet *ifp; 685 char *layer; 686 int hlen; 687 688 ifp = m->m_pkthdr.rcvif; 689 690 if (m->m_len < sizeof(struct arphdr) && 691 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 692 ARP_LOG(LOG_NOTICE, "packet with short header received on %s\n", 693 if_name(ifp)); 694 return; 695 } 696 ar = mtod(m, struct arphdr *); 697 698 /* Check if length is sufficient */ 699 if (m->m_len < arphdr_len(ar)) { 700 m = m_pullup(m, arphdr_len(ar)); 701 if (m == NULL) { 702 ARP_LOG(LOG_NOTICE, "short packet received on %s\n", 703 if_name(ifp)); 704 return; 705 } 706 ar = mtod(m, struct arphdr *); 707 } 708 709 hlen = 0; 710 layer = ""; 711 switch (ntohs(ar->ar_hrd)) { 712 case ARPHRD_ETHER: 713 hlen = ETHER_ADDR_LEN; /* RFC 826 */ 714 layer = "ethernet"; 715 break; 716 case ARPHRD_INFINIBAND: 717 hlen = 20; /* RFC 4391, INFINIBAND_ALEN */ 718 layer = "infiniband"; 719 break; 720 case ARPHRD_IEEE1394: 721 hlen = 0; /* SHALL be 16 */ /* RFC 2734 */ 722 layer = "firewire"; 723 724 /* 725 * Restrict too long hardware addresses. 726 * Currently we are capable of handling 20-byte 727 * addresses ( sizeof(lle->ll_addr) ) 728 */ 729 if (ar->ar_hln >= 20) 730 hlen = 16; 731 break; 732 default: 733 ARP_LOG(LOG_NOTICE, 734 "packet with unknown hardware format 0x%02d received on " 735 "%s\n", ntohs(ar->ar_hrd), if_name(ifp)); 736 m_freem(m); 737 return; 738 } 739 740 if (hlen != 0 && hlen != ar->ar_hln) { 741 ARP_LOG(LOG_NOTICE, 742 "packet with invalid %s address length %d received on %s\n", 743 layer, ar->ar_hln, if_name(ifp)); 744 m_freem(m); 745 return; 746 } 747 748 ARPSTAT_INC(received); 749 switch (ntohs(ar->ar_pro)) { 750 #ifdef INET 751 case ETHERTYPE_IP: 752 in_arpinput(m); 753 return; 754 #endif 755 } 756 m_freem(m); 757 } 758 759 #ifdef INET 760 /* 761 * ARP for Internet protocols on 10 Mb/s Ethernet. 762 * Algorithm is that given in RFC 826. 763 * In addition, a sanity check is performed on the sender 764 * protocol address, to catch impersonators. 765 * We no longer handle negotiations for use of trailer protocol: 766 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 767 * along with IP replies if we wanted trailers sent to us, 768 * and also sent them in response to IP replies. 769 * This allowed either end to announce the desire to receive 770 * trailer packets. 771 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 772 * but formerly didn't normally send requests. 773 */ 774 static int log_arp_wrong_iface = 1; 775 static int log_arp_movements = 1; 776 static int log_arp_permanent_modify = 1; 777 static int allow_multicast = 0; 778 779 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 780 &log_arp_wrong_iface, 0, 781 "log arp packets arriving on the wrong interface"); 782 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 783 &log_arp_movements, 0, 784 "log arp replies from MACs different than the one in the cache"); 785 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 786 &log_arp_permanent_modify, 0, 787 "log arp replies from MACs different than the one in the permanent arp entry"); 788 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW, 789 &allow_multicast, 0, "accept multicast addresses"); 790 791 static void 792 in_arpinput(struct mbuf *m) 793 { 794 struct rm_priotracker in_ifa_tracker; 795 struct arphdr *ah; 796 struct ifnet *ifp = m->m_pkthdr.rcvif; 797 struct llentry *la = NULL, *la_tmp; 798 struct ifaddr *ifa; 799 struct in_ifaddr *ia; 800 struct sockaddr sa; 801 struct in_addr isaddr, itaddr, myaddr; 802 u_int8_t *enaddr = NULL; 803 int op; 804 int bridged = 0, is_bridge = 0; 805 int carped; 806 struct sockaddr_in sin; 807 struct sockaddr *dst; 808 struct nhop_object *nh; 809 uint8_t linkhdr[LLE_MAX_LINKHDR]; 810 struct route ro; 811 size_t linkhdrsize; 812 int lladdr_off; 813 int error; 814 char addrbuf[INET_ADDRSTRLEN]; 815 816 NET_EPOCH_ASSERT(); 817 818 sin.sin_len = sizeof(struct sockaddr_in); 819 sin.sin_family = AF_INET; 820 sin.sin_addr.s_addr = 0; 821 822 if (ifp->if_bridge) 823 bridged = 1; 824 if (ifp->if_type == IFT_BRIDGE) 825 is_bridge = 1; 826 827 /* 828 * We already have checked that mbuf contains enough contiguous data 829 * to hold entire arp message according to the arp header. 830 */ 831 ah = mtod(m, struct arphdr *); 832 833 /* 834 * ARP is only for IPv4 so we can reject packets with 835 * a protocol length not equal to an IPv4 address. 836 */ 837 if (ah->ar_pln != sizeof(struct in_addr)) { 838 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n", 839 sizeof(struct in_addr)); 840 goto drop; 841 } 842 843 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) { 844 ARP_LOG(LOG_NOTICE, "%*D is multicast\n", 845 ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); 846 goto drop; 847 } 848 849 op = ntohs(ah->ar_op); 850 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 851 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 852 853 if (op == ARPOP_REPLY) 854 ARPSTAT_INC(rxreplies); 855 856 /* 857 * For a bridge, we want to check the address irrespective 858 * of the receive interface. (This will change slightly 859 * when we have clusters of interfaces). 860 */ 861 IN_IFADDR_RLOCK(&in_ifa_tracker); 862 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 863 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 864 ia->ia_ifp == ifp) && 865 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr && 866 (ia->ia_ifa.ifa_carp == NULL || 867 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) { 868 ifa_ref(&ia->ia_ifa); 869 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 870 goto match; 871 } 872 } 873 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 874 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 875 ia->ia_ifp == ifp) && 876 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 877 ifa_ref(&ia->ia_ifa); 878 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 879 goto match; 880 } 881 882 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 883 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 884 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 885 addr == ia->ia_addr.sin_addr.s_addr) 886 /* 887 * Check the case when bridge shares its MAC address with 888 * some of its children, so packets are claimed by bridge 889 * itself (bridge_input() does it first), but they are really 890 * meant to be destined to the bridge member. 891 */ 892 if (is_bridge) { 893 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 894 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 895 ifa_ref(&ia->ia_ifa); 896 ifp = ia->ia_ifp; 897 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 898 goto match; 899 } 900 } 901 } 902 #undef BDG_MEMBER_MATCHES_ARP 903 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 904 905 /* 906 * No match, use the first inet address on the receive interface 907 * as a dummy address for the rest of the function. 908 */ 909 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 910 if (ifa->ifa_addr->sa_family == AF_INET && 911 (ifa->ifa_carp == NULL || 912 (*carp_iamatch_p)(ifa, &enaddr))) { 913 ia = ifatoia(ifa); 914 ifa_ref(ifa); 915 goto match; 916 } 917 918 /* 919 * If bridging, fall back to using any inet address. 920 */ 921 IN_IFADDR_RLOCK(&in_ifa_tracker); 922 if (!bridged || (ia = CK_STAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { 923 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 924 goto drop; 925 } 926 ifa_ref(&ia->ia_ifa); 927 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 928 match: 929 if (!enaddr) 930 enaddr = (u_int8_t *)IF_LLADDR(ifp); 931 carped = (ia->ia_ifa.ifa_carp != NULL); 932 myaddr = ia->ia_addr.sin_addr; 933 ifa_free(&ia->ia_ifa); 934 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 935 goto drop; /* it's from me, ignore it. */ 936 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 937 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address " 938 "%s!\n", inet_ntoa_r(isaddr, addrbuf)); 939 goto drop; 940 } 941 942 if (ifp->if_addrlen != ah->ar_hln) { 943 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, " 944 "i/f %d (ignored)\n", ifp->if_addrlen, 945 (u_char *) ar_sha(ah), ":", ah->ar_hln, 946 ifp->if_addrlen); 947 goto drop; 948 } 949 950 /* 951 * Warn if another host is using the same IP address, but only if the 952 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 953 * case we suppress the warning to avoid false positive complaints of 954 * potential misconfiguration. 955 */ 956 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr && 957 myaddr.s_addr != 0) { 958 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n", 959 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 960 inet_ntoa_r(isaddr, addrbuf), ifp->if_xname); 961 itaddr = myaddr; 962 ARPSTAT_INC(dupips); 963 goto reply; 964 } 965 if (ifp->if_flags & IFF_STATICARP) 966 goto reply; 967 968 bzero(&sin, sizeof(sin)); 969 sin.sin_len = sizeof(struct sockaddr_in); 970 sin.sin_family = AF_INET; 971 sin.sin_addr = isaddr; 972 dst = (struct sockaddr *)&sin; 973 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 974 if (la != NULL) 975 arp_check_update_lle(ah, isaddr, ifp, bridged, la); 976 else if (itaddr.s_addr == myaddr.s_addr) { 977 /* 978 * Request/reply to our address, but no lle exists yet. 979 * Calculate full link prepend to use in lle. 980 */ 981 linkhdrsize = sizeof(linkhdr); 982 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr, 983 &linkhdrsize, &lladdr_off) != 0) 984 goto reply; 985 986 /* Allocate new entry */ 987 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst); 988 if (la == NULL) { 989 990 /* 991 * lle creation may fail if source address belongs 992 * to non-directly connected subnet. However, we 993 * will try to answer the request instead of dropping 994 * frame. 995 */ 996 goto reply; 997 } 998 lltable_set_entry_addr(ifp, la, linkhdr, linkhdrsize, 999 lladdr_off); 1000 1001 IF_AFDATA_WLOCK(ifp); 1002 LLE_WLOCK(la); 1003 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 1004 1005 /* 1006 * Check if lle still does not exists. 1007 * If it does, that means that we either 1008 * 1) have configured it explicitly, via 1009 * 1a) 'arp -s' static entry or 1010 * 1b) interface address static record 1011 * or 1012 * 2) it was the result of sending first packet to-host 1013 * or 1014 * 3) it was another arp reply packet we handled in 1015 * different thread. 1016 * 1017 * In all cases except 3) we definitely need to prefer 1018 * existing lle. For the sake of simplicity, prefer any 1019 * existing lle over newly-create one. 1020 */ 1021 if (la_tmp == NULL) 1022 lltable_link_entry(LLTABLE(ifp), la); 1023 IF_AFDATA_WUNLOCK(ifp); 1024 1025 if (la_tmp == NULL) { 1026 arp_mark_lle_reachable(la); 1027 LLE_WUNLOCK(la); 1028 } else { 1029 /* Free newly-create entry and handle packet */ 1030 lltable_free_entry(LLTABLE(ifp), la); 1031 la = la_tmp; 1032 la_tmp = NULL; 1033 arp_check_update_lle(ah, isaddr, ifp, bridged, la); 1034 /* arp_check_update_lle() returns @la unlocked */ 1035 } 1036 la = NULL; 1037 } 1038 reply: 1039 if (op != ARPOP_REQUEST) 1040 goto drop; 1041 ARPSTAT_INC(rxrequests); 1042 1043 if (itaddr.s_addr == myaddr.s_addr) { 1044 /* Shortcut.. the receiving interface is the target. */ 1045 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1046 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1047 } else { 1048 /* 1049 * Destination address is not ours. Check if 1050 * proxyarp entry exists or proxyarp is turned on globally. 1051 */ 1052 struct llentry *lle; 1053 1054 sin.sin_addr = itaddr; 1055 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); 1056 1057 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { 1058 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1059 (void)memcpy(ar_sha(ah), lle->ll_addr, ah->ar_hln); 1060 LLE_RUNLOCK(lle); 1061 } else { 1062 1063 if (lle != NULL) 1064 LLE_RUNLOCK(lle); 1065 1066 if (!V_arp_proxyall) 1067 goto drop; 1068 1069 NET_EPOCH_ASSERT(); 1070 nh = fib4_lookup(ifp->if_fib, itaddr, 0, 0, 0); 1071 if (nh == NULL) 1072 goto drop; 1073 1074 /* 1075 * Don't send proxies for nodes on the same interface 1076 * as this one came out of, or we'll get into a fight 1077 * over who claims what Ether address. 1078 */ 1079 if (nh->nh_ifp == ifp) 1080 goto drop; 1081 1082 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1083 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1084 1085 /* 1086 * Also check that the node which sent the ARP packet 1087 * is on the interface we expect it to be on. This 1088 * avoids ARP chaos if an interface is connected to the 1089 * wrong network. 1090 */ 1091 1092 nh = fib4_lookup(ifp->if_fib, isaddr, 0, 0, 0); 1093 if (nh == NULL) 1094 goto drop; 1095 if (nh->nh_ifp != ifp) { 1096 ARP_LOG(LOG_INFO, "proxy: ignoring request" 1097 " from %s via %s\n", 1098 inet_ntoa_r(isaddr, addrbuf), 1099 ifp->if_xname); 1100 goto drop; 1101 } 1102 1103 #ifdef DEBUG_PROXY 1104 printf("arp: proxying for %s\n", 1105 inet_ntoa_r(itaddr, addrbuf)); 1106 #endif 1107 } 1108 } 1109 1110 if (itaddr.s_addr == myaddr.s_addr && 1111 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 1112 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 1113 #ifdef DEBUG_LINKLOCAL 1114 printf("arp: sending reply for link-local addr %s\n", 1115 inet_ntoa_r(itaddr, addrbuf)); 1116 #endif 1117 m->m_flags |= M_BCAST; 1118 m->m_flags &= ~M_MCAST; 1119 } else { 1120 /* default behaviour; never reply by broadcast. */ 1121 m->m_flags &= ~(M_BCAST|M_MCAST); 1122 } 1123 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 1124 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 1125 ah->ar_op = htons(ARPOP_REPLY); 1126 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 1127 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 1128 m->m_pkthdr.len = m->m_len; 1129 m->m_pkthdr.rcvif = NULL; 1130 sa.sa_family = AF_ARP; 1131 sa.sa_len = 2; 1132 1133 /* Calculate link header for sending frame */ 1134 bzero(&ro, sizeof(ro)); 1135 linkhdrsize = sizeof(linkhdr); 1136 error = arp_fillheader(ifp, ah, 0, linkhdr, &linkhdrsize); 1137 1138 /* 1139 * arp_fillheader() may fail due to lack of support inside encap request 1140 * routing. This is not necessary an error, AF_ARP can/should be handled 1141 * by if_output(). 1142 */ 1143 if (error != 0 && error != EAFNOSUPPORT) { 1144 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n", 1145 if_name(ifp), error); 1146 return; 1147 } 1148 1149 ro.ro_prepend = linkhdr; 1150 ro.ro_plen = linkhdrsize; 1151 ro.ro_flags = 0; 1152 1153 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1154 (*ifp->if_output)(ifp, m, &sa, &ro); 1155 ARPSTAT_INC(txreplies); 1156 return; 1157 1158 drop: 1159 m_freem(m); 1160 } 1161 #endif 1162 1163 /* 1164 * Checks received arp data against existing @la. 1165 * Updates lle state/performs notification if necessary. 1166 */ 1167 static void 1168 arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, struct ifnet *ifp, 1169 int bridged, struct llentry *la) 1170 { 1171 struct sockaddr sa; 1172 struct mbuf *m_hold, *m_hold_next; 1173 uint8_t linkhdr[LLE_MAX_LINKHDR]; 1174 size_t linkhdrsize; 1175 int lladdr_off; 1176 char addrbuf[INET_ADDRSTRLEN]; 1177 1178 LLE_WLOCK_ASSERT(la); 1179 1180 /* the following is not an error when doing bridging */ 1181 if (!bridged && la->lle_tbl->llt_ifp != ifp) { 1182 if (log_arp_wrong_iface) 1183 ARP_LOG(LOG_WARNING, "%s is on %s " 1184 "but got reply from %*D on %s\n", 1185 inet_ntoa_r(isaddr, addrbuf), 1186 la->lle_tbl->llt_ifp->if_xname, 1187 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 1188 ifp->if_xname); 1189 LLE_WUNLOCK(la); 1190 return; 1191 } 1192 if ((la->la_flags & LLE_VALID) && 1193 bcmp(ar_sha(ah), la->ll_addr, ifp->if_addrlen)) { 1194 if (la->la_flags & LLE_STATIC) { 1195 LLE_WUNLOCK(la); 1196 if (log_arp_permanent_modify) 1197 ARP_LOG(LOG_ERR, 1198 "%*D attempts to modify " 1199 "permanent entry for %s on %s\n", 1200 ifp->if_addrlen, 1201 (u_char *)ar_sha(ah), ":", 1202 inet_ntoa_r(isaddr, addrbuf), 1203 ifp->if_xname); 1204 return; 1205 } 1206 if (log_arp_movements) { 1207 ARP_LOG(LOG_INFO, "%s moved from %*D " 1208 "to %*D on %s\n", 1209 inet_ntoa_r(isaddr, addrbuf), 1210 ifp->if_addrlen, 1211 (u_char *)la->ll_addr, ":", 1212 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 1213 ifp->if_xname); 1214 } 1215 } 1216 1217 /* Calculate full link prepend to use in lle */ 1218 linkhdrsize = sizeof(linkhdr); 1219 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr, 1220 &linkhdrsize, &lladdr_off) != 0) 1221 return; 1222 1223 /* Check if something has changed */ 1224 if (memcmp(la->r_linkdata, linkhdr, linkhdrsize) != 0 || 1225 (la->la_flags & LLE_VALID) == 0) { 1226 /* Try to perform LLE update */ 1227 if (lltable_try_set_entry_addr(ifp, la, linkhdr, linkhdrsize, 1228 lladdr_off) == 0) 1229 return; 1230 1231 /* Clear fast path feedback request if set */ 1232 la->r_skip_req = 0; 1233 } 1234 1235 arp_mark_lle_reachable(la); 1236 1237 /* 1238 * The packets are all freed within the call to the output 1239 * routine. 1240 * 1241 * NB: The lock MUST be released before the call to the 1242 * output routine. 1243 */ 1244 if (la->la_hold != NULL) { 1245 m_hold = la->la_hold; 1246 la->la_hold = NULL; 1247 la->la_numheld = 0; 1248 lltable_fill_sa_entry(la, &sa); 1249 LLE_WUNLOCK(la); 1250 for (; m_hold != NULL; m_hold = m_hold_next) { 1251 m_hold_next = m_hold->m_nextpkt; 1252 m_hold->m_nextpkt = NULL; 1253 /* Avoid confusing lower layers. */ 1254 m_clrprotoflags(m_hold); 1255 (*ifp->if_output)(ifp, m_hold, &sa, NULL); 1256 } 1257 } else 1258 LLE_WUNLOCK(la); 1259 } 1260 1261 static void 1262 arp_mark_lle_reachable(struct llentry *la) 1263 { 1264 int canceled, wtime; 1265 1266 LLE_WLOCK_ASSERT(la); 1267 1268 la->ln_state = ARP_LLINFO_REACHABLE; 1269 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED); 1270 1271 if (!(la->la_flags & LLE_STATIC)) { 1272 LLE_ADDREF(la); 1273 la->la_expire = time_uptime + V_arpt_keep; 1274 wtime = V_arpt_keep - V_arp_maxtries * V_arpt_rexmit; 1275 if (wtime < 0) 1276 wtime = V_arpt_keep; 1277 canceled = callout_reset(&la->lle_timer, 1278 hz * wtime, arptimer, la); 1279 if (canceled) 1280 LLE_REMREF(la); 1281 } 1282 la->la_asked = 0; 1283 la->la_preempt = V_arp_maxtries; 1284 } 1285 1286 /* 1287 * Add permanent link-layer record for given interface address. 1288 */ 1289 static __noinline void 1290 arp_add_ifa_lle(struct ifnet *ifp, const struct sockaddr *dst) 1291 { 1292 struct llentry *lle, *lle_tmp; 1293 1294 /* 1295 * Interface address LLE record is considered static 1296 * because kernel code relies on LLE_STATIC flag to check 1297 * if these entries can be rewriten by arp updates. 1298 */ 1299 lle = lltable_alloc_entry(LLTABLE(ifp), LLE_IFADDR | LLE_STATIC, dst); 1300 if (lle == NULL) { 1301 log(LOG_INFO, "arp_ifinit: cannot create arp " 1302 "entry for interface address\n"); 1303 return; 1304 } 1305 1306 IF_AFDATA_WLOCK(ifp); 1307 LLE_WLOCK(lle); 1308 /* Unlink any entry if exists */ 1309 lle_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 1310 if (lle_tmp != NULL) 1311 lltable_unlink_entry(LLTABLE(ifp), lle_tmp); 1312 1313 lltable_link_entry(LLTABLE(ifp), lle); 1314 IF_AFDATA_WUNLOCK(ifp); 1315 1316 if (lle_tmp != NULL) 1317 EVENTHANDLER_INVOKE(lle_event, lle_tmp, LLENTRY_EXPIRED); 1318 1319 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED); 1320 LLE_WUNLOCK(lle); 1321 if (lle_tmp != NULL) 1322 lltable_free_entry(LLTABLE(ifp), lle_tmp); 1323 } 1324 1325 /* 1326 * Handle the garp_rexmit_count. Like sysctl_handle_int(), but limits the range 1327 * of valid values. 1328 */ 1329 static int 1330 sysctl_garp_rexmit(SYSCTL_HANDLER_ARGS) 1331 { 1332 int error; 1333 int rexmit_count = *(int *)arg1; 1334 1335 error = sysctl_handle_int(oidp, &rexmit_count, 0, req); 1336 1337 /* Enforce limits on any new value that may have been set. */ 1338 if (!error && req->newptr) { 1339 /* A new value was set. */ 1340 if (rexmit_count < 0) { 1341 rexmit_count = 0; 1342 } else if (rexmit_count > MAX_GARP_RETRANSMITS) { 1343 rexmit_count = MAX_GARP_RETRANSMITS; 1344 } 1345 *(int *)arg1 = rexmit_count; 1346 } 1347 1348 return (error); 1349 } 1350 1351 /* 1352 * Retransmit a Gratuitous ARP (GARP) and, if necessary, schedule a callout to 1353 * retransmit it again. A pending callout owns a reference to the ifa. 1354 */ 1355 static void 1356 garp_rexmit(void *arg) 1357 { 1358 struct in_ifaddr *ia = arg; 1359 1360 if (callout_pending(&ia->ia_garp_timer) || 1361 !callout_active(&ia->ia_garp_timer)) { 1362 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp); 1363 ifa_free(&ia->ia_ifa); 1364 return; 1365 } 1366 1367 CURVNET_SET(ia->ia_ifa.ifa_ifp->if_vnet); 1368 1369 /* 1370 * Drop lock while the ARP request is generated. 1371 */ 1372 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp); 1373 1374 arprequest(ia->ia_ifa.ifa_ifp, &IA_SIN(ia)->sin_addr, 1375 &IA_SIN(ia)->sin_addr, IF_LLADDR(ia->ia_ifa.ifa_ifp)); 1376 1377 /* 1378 * Increment the count of retransmissions. If the count has reached the 1379 * maximum value, stop sending the GARP packets. Otherwise, schedule 1380 * the callout to retransmit another GARP packet. 1381 */ 1382 ++ia->ia_garp_count; 1383 if (ia->ia_garp_count >= garp_rexmit_count) { 1384 ifa_free(&ia->ia_ifa); 1385 } else { 1386 int rescheduled; 1387 IF_ADDR_WLOCK(ia->ia_ifa.ifa_ifp); 1388 rescheduled = callout_reset(&ia->ia_garp_timer, 1389 (1 << ia->ia_garp_count) * hz, 1390 garp_rexmit, ia); 1391 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp); 1392 if (rescheduled) { 1393 ifa_free(&ia->ia_ifa); 1394 } 1395 } 1396 1397 CURVNET_RESTORE(); 1398 } 1399 1400 /* 1401 * Start the GARP retransmit timer. 1402 * 1403 * A single GARP is always transmitted when an IPv4 address is added 1404 * to an interface and that is usually sufficient. However, in some 1405 * circumstances, such as when a shared address is passed between 1406 * cluster nodes, this single GARP may occasionally be dropped or 1407 * lost. This can lead to neighbors on the network link working with a 1408 * stale ARP cache and sending packets destined for that address to 1409 * the node that previously owned the address, which may not respond. 1410 * 1411 * To avoid this situation, GARP retransmits can be enabled by setting 1412 * the net.link.ether.inet.garp_rexmit_count sysctl to a value greater 1413 * than zero. The setting represents the maximum number of 1414 * retransmissions. The interval between retransmissions is calculated 1415 * using an exponential backoff algorithm, doubling each time, so the 1416 * retransmission intervals are: {1, 2, 4, 8, 16, ...} (seconds). 1417 */ 1418 static void 1419 garp_timer_start(struct ifaddr *ifa) 1420 { 1421 struct in_ifaddr *ia = (struct in_ifaddr *) ifa; 1422 1423 IF_ADDR_WLOCK(ia->ia_ifa.ifa_ifp); 1424 ia->ia_garp_count = 0; 1425 if (callout_reset(&ia->ia_garp_timer, (1 << ia->ia_garp_count) * hz, 1426 garp_rexmit, ia) == 0) { 1427 ifa_ref(ifa); 1428 } 1429 IF_ADDR_WUNLOCK(ia->ia_ifa.ifa_ifp); 1430 } 1431 1432 void 1433 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1434 { 1435 struct epoch_tracker et; 1436 const struct sockaddr_in *dst_in; 1437 const struct sockaddr *dst; 1438 1439 if (ifa->ifa_carp != NULL) 1440 return; 1441 1442 dst = ifa->ifa_addr; 1443 dst_in = (const struct sockaddr_in *)dst; 1444 1445 if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY) 1446 return; 1447 NET_EPOCH_ENTER(et); 1448 arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp)); 1449 NET_EPOCH_EXIT(et); 1450 if (garp_rexmit_count > 0) { 1451 garp_timer_start(ifa); 1452 } 1453 1454 arp_add_ifa_lle(ifp, dst); 1455 } 1456 1457 void 1458 arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr) 1459 { 1460 1461 if (ntohl(addr.s_addr) != INADDR_ANY) 1462 arprequest(ifp, &addr, &addr, enaddr); 1463 } 1464 1465 /* 1466 * Sends gratuitous ARPs for each ifaddr to notify other 1467 * nodes about the address change. 1468 */ 1469 static __noinline void 1470 arp_handle_ifllchange(struct ifnet *ifp) 1471 { 1472 struct ifaddr *ifa; 1473 1474 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1475 if (ifa->ifa_addr->sa_family == AF_INET) 1476 arp_ifinit(ifp, ifa); 1477 } 1478 } 1479 1480 /* 1481 * A handler for interface link layer address change event. 1482 */ 1483 static void 1484 arp_iflladdr(void *arg __unused, struct ifnet *ifp) 1485 { 1486 1487 lltable_update_ifaddr(LLTABLE(ifp)); 1488 1489 if ((ifp->if_flags & IFF_UP) != 0) 1490 arp_handle_ifllchange(ifp); 1491 } 1492 1493 static void 1494 vnet_arp_init(void) 1495 { 1496 1497 if (IS_DEFAULT_VNET(curvnet)) { 1498 netisr_register(&arp_nh); 1499 iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event, 1500 arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 1501 } 1502 #ifdef VIMAGE 1503 else 1504 netisr_register_vnet(&arp_nh); 1505 #endif 1506 } 1507 VNET_SYSINIT(vnet_arp_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND, 1508 vnet_arp_init, 0); 1509 1510 #ifdef VIMAGE 1511 /* 1512 * We have to unregister ARP along with IP otherwise we risk doing INADDR_HASH 1513 * lookups after destroying the hash. Ideally this would go on SI_ORDER_3.5. 1514 */ 1515 static void 1516 vnet_arp_destroy(__unused void *arg) 1517 { 1518 1519 netisr_unregister_vnet(&arp_nh); 1520 } 1521 VNET_SYSUNINIT(vnet_arp_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1522 vnet_arp_destroy, NULL); 1523 #endif 1524