1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 30 */ 31 32 /* 33 * Ethernet address resolution protocol. 34 * TODO: 35 * add "inuse/lock" bit (or ref. count) along with valid bit 36 */ 37 38 #include <sys/cdefs.h> 39 __FBSDID("$FreeBSD$"); 40 41 #include "opt_inet.h" 42 43 #include <sys/param.h> 44 #include <sys/kernel.h> 45 #include <sys/lock.h> 46 #include <sys/queue.h> 47 #include <sys/sysctl.h> 48 #include <sys/systm.h> 49 #include <sys/mbuf.h> 50 #include <sys/malloc.h> 51 #include <sys/proc.h> 52 #include <sys/rmlock.h> 53 #include <sys/socket.h> 54 #include <sys/syslog.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/if_dl.h> 59 #include <net/if_types.h> 60 #include <net/netisr.h> 61 #include <net/ethernet.h> 62 #include <net/route.h> 63 #include <net/vnet.h> 64 65 #include <netinet/in.h> 66 #include <netinet/in_fib.h> 67 #include <netinet/in_var.h> 68 #include <net/if_llatbl.h> 69 #include <netinet/if_ether.h> 70 #ifdef INET 71 #include <netinet/ip_carp.h> 72 #endif 73 74 #include <security/mac/mac_framework.h> 75 76 #define SIN(s) ((const struct sockaddr_in *)(s)) 77 78 static struct timeval arp_lastlog; 79 static int arp_curpps; 80 static int arp_maxpps = 1; 81 82 /* Simple ARP state machine */ 83 enum arp_llinfo_state { 84 ARP_LLINFO_INCOMPLETE = 0, /* No LLE data */ 85 ARP_LLINFO_REACHABLE, /* LLE is valid */ 86 ARP_LLINFO_VERIFY, /* LLE is valid, need refresh */ 87 ARP_LLINFO_DELETED, /* LLE is deleted */ 88 }; 89 90 SYSCTL_DECL(_net_link_ether); 91 static SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 92 static SYSCTL_NODE(_net_link_ether, PF_ARP, arp, CTLFLAG_RW, 0, ""); 93 94 /* timer values */ 95 static VNET_DEFINE(int, arpt_keep) = (20*60); /* once resolved, good for 20 96 * minutes */ 97 static VNET_DEFINE(int, arp_maxtries) = 5; 98 static VNET_DEFINE(int, arp_proxyall) = 0; 99 static VNET_DEFINE(int, arpt_down) = 20; /* keep incomplete entries for 100 * 20 seconds */ 101 static VNET_DEFINE(int, arpt_rexmit) = 1; /* retransmit arp entries, sec*/ 102 VNET_PCPUSTAT_DEFINE(struct arpstat, arpstat); /* ARP statistics, see if_arp.h */ 103 VNET_PCPUSTAT_SYSINIT(arpstat); 104 105 #ifdef VIMAGE 106 VNET_PCPUSTAT_SYSUNINIT(arpstat); 107 #endif /* VIMAGE */ 108 109 static VNET_DEFINE(int, arp_maxhold) = 1; 110 111 #define V_arpt_keep VNET(arpt_keep) 112 #define V_arpt_down VNET(arpt_down) 113 #define V_arpt_rexmit VNET(arpt_rexmit) 114 #define V_arp_maxtries VNET(arp_maxtries) 115 #define V_arp_proxyall VNET(arp_proxyall) 116 #define V_arp_maxhold VNET(arp_maxhold) 117 118 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW, 119 &VNET_NAME(arpt_keep), 0, 120 "ARP entry lifetime in seconds"); 121 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_VNET | CTLFLAG_RW, 122 &VNET_NAME(arp_maxtries), 0, 123 "ARP resolution attempts before returning error"); 124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_VNET | CTLFLAG_RW, 125 &VNET_NAME(arp_proxyall), 0, 126 "Enable proxy ARP for all suitable requests"); 127 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, wait, CTLFLAG_VNET | CTLFLAG_RW, 128 &VNET_NAME(arpt_down), 0, 129 "Incomplete ARP entry lifetime in seconds"); 130 SYSCTL_VNET_PCPUSTAT(_net_link_ether_arp, OID_AUTO, stats, struct arpstat, 131 arpstat, "ARP statistics (struct arpstat, net/if_arp.h)"); 132 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxhold, CTLFLAG_VNET | CTLFLAG_RW, 133 &VNET_NAME(arp_maxhold), 0, 134 "Number of packets to hold per ARP entry"); 135 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, 136 CTLFLAG_RW, &arp_maxpps, 0, 137 "Maximum number of remotely triggered ARP messages that can be " 138 "logged per second"); 139 140 #define ARP_LOG(pri, ...) do { \ 141 if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ 142 log((pri), "arp: " __VA_ARGS__); \ 143 } while (0) 144 145 146 static void arpintr(struct mbuf *); 147 static void arptimer(void *); 148 #ifdef INET 149 static void in_arpinput(struct mbuf *); 150 #endif 151 152 static void arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, 153 struct ifnet *ifp, int bridged, struct llentry *la); 154 static void arp_mark_lle_reachable(struct llentry *la); 155 static void arp_iflladdr(void *arg __unused, struct ifnet *ifp); 156 157 static eventhandler_tag iflladdr_tag; 158 159 static const struct netisr_handler arp_nh = { 160 .nh_name = "arp", 161 .nh_handler = arpintr, 162 .nh_proto = NETISR_ARP, 163 .nh_policy = NETISR_POLICY_SOURCE, 164 }; 165 166 /* 167 * Timeout routine. Age arp_tab entries periodically. 168 */ 169 static void 170 arptimer(void *arg) 171 { 172 struct llentry *lle = (struct llentry *)arg; 173 struct ifnet *ifp; 174 int r_skip_req; 175 176 if (lle->la_flags & LLE_STATIC) { 177 return; 178 } 179 LLE_WLOCK(lle); 180 if (callout_pending(&lle->lle_timer)) { 181 /* 182 * Here we are a bit odd here in the treatment of 183 * active/pending. If the pending bit is set, it got 184 * rescheduled before I ran. The active 185 * bit we ignore, since if it was stopped 186 * in ll_tablefree() and was currently running 187 * it would have return 0 so the code would 188 * not have deleted it since the callout could 189 * not be stopped so we want to go through 190 * with the delete here now. If the callout 191 * was restarted, the pending bit will be back on and 192 * we just want to bail since the callout_reset would 193 * return 1 and our reference would have been removed 194 * by arpresolve() below. 195 */ 196 LLE_WUNLOCK(lle); 197 return; 198 } 199 ifp = lle->lle_tbl->llt_ifp; 200 CURVNET_SET(ifp->if_vnet); 201 202 switch (lle->ln_state) { 203 case ARP_LLINFO_REACHABLE: 204 205 /* 206 * Expiration time is approaching. 207 * Let's try to refresh entry if it is still 208 * in use. 209 * 210 * Set r_skip_req to get feedback from 211 * fast path. Change state and re-schedule 212 * ourselves. 213 */ 214 LLE_REQ_LOCK(lle); 215 lle->r_skip_req = 1; 216 LLE_REQ_UNLOCK(lle); 217 lle->ln_state = ARP_LLINFO_VERIFY; 218 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 219 LLE_WUNLOCK(lle); 220 CURVNET_RESTORE(); 221 return; 222 case ARP_LLINFO_VERIFY: 223 LLE_REQ_LOCK(lle); 224 r_skip_req = lle->r_skip_req; 225 LLE_REQ_UNLOCK(lle); 226 227 if (r_skip_req == 0 && lle->la_preempt > 0) { 228 /* Entry was used, issue refresh request */ 229 struct in_addr dst; 230 dst = lle->r_l3addr.addr4; 231 lle->la_preempt--; 232 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 233 LLE_WUNLOCK(lle); 234 arprequest(ifp, NULL, &dst, NULL); 235 CURVNET_RESTORE(); 236 return; 237 } 238 /* Nothing happened. Reschedule if not too late */ 239 if (lle->la_expire > time_uptime) { 240 callout_schedule(&lle->lle_timer, hz * V_arpt_rexmit); 241 LLE_WUNLOCK(lle); 242 CURVNET_RESTORE(); 243 return; 244 } 245 break; 246 case ARP_LLINFO_INCOMPLETE: 247 case ARP_LLINFO_DELETED: 248 break; 249 } 250 251 if ((lle->la_flags & LLE_DELETED) == 0) { 252 int evt; 253 254 if (lle->la_flags & LLE_VALID) 255 evt = LLENTRY_EXPIRED; 256 else 257 evt = LLENTRY_TIMEDOUT; 258 EVENTHANDLER_INVOKE(lle_event, lle, evt); 259 } 260 261 callout_stop(&lle->lle_timer); 262 263 /* XXX: LOR avoidance. We still have ref on lle. */ 264 LLE_WUNLOCK(lle); 265 IF_AFDATA_LOCK(ifp); 266 LLE_WLOCK(lle); 267 268 /* Guard against race with other llentry_free(). */ 269 if (lle->la_flags & LLE_LINKED) { 270 LLE_REMREF(lle); 271 lltable_unlink_entry(lle->lle_tbl, lle); 272 } 273 IF_AFDATA_UNLOCK(ifp); 274 275 size_t pkts_dropped = llentry_free(lle); 276 277 ARPSTAT_ADD(dropped, pkts_dropped); 278 ARPSTAT_INC(timeouts); 279 280 CURVNET_RESTORE(); 281 } 282 283 /* 284 * Stores link-layer header for @ifp in format suitable for if_output() 285 * into buffer @buf. Resulting header length is stored in @bufsize. 286 * 287 * Returns 0 on success. 288 */ 289 static int 290 arp_fillheader(struct ifnet *ifp, struct arphdr *ah, int bcast, u_char *buf, 291 size_t *bufsize) 292 { 293 struct if_encap_req ereq; 294 int error; 295 296 bzero(buf, *bufsize); 297 bzero(&ereq, sizeof(ereq)); 298 ereq.buf = buf; 299 ereq.bufsize = *bufsize; 300 ereq.rtype = IFENCAP_LL; 301 ereq.family = AF_ARP; 302 ereq.lladdr = ar_tha(ah); 303 ereq.hdata = (u_char *)ah; 304 if (bcast) 305 ereq.flags = IFENCAP_FLAG_BROADCAST; 306 error = ifp->if_requestencap(ifp, &ereq); 307 if (error == 0) 308 *bufsize = ereq.bufsize; 309 310 return (error); 311 } 312 313 314 /* 315 * Broadcast an ARP request. Caller specifies: 316 * - arp header source ip address 317 * - arp header target ip address 318 * - arp header source ethernet address 319 */ 320 void 321 arprequest(struct ifnet *ifp, const struct in_addr *sip, 322 const struct in_addr *tip, u_char *enaddr) 323 { 324 struct mbuf *m; 325 struct arphdr *ah; 326 struct sockaddr sa; 327 u_char *carpaddr = NULL; 328 uint8_t linkhdr[LLE_MAX_LINKHDR]; 329 size_t linkhdrsize; 330 struct route ro; 331 int error; 332 333 if (sip == NULL) { 334 /* 335 * The caller did not supply a source address, try to find 336 * a compatible one among those assigned to this interface. 337 */ 338 struct ifaddr *ifa; 339 340 IF_ADDR_RLOCK(ifp); 341 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 342 if (ifa->ifa_addr->sa_family != AF_INET) 343 continue; 344 345 if (ifa->ifa_carp) { 346 if ((*carp_iamatch_p)(ifa, &carpaddr) == 0) 347 continue; 348 sip = &IA_SIN(ifa)->sin_addr; 349 } else { 350 carpaddr = NULL; 351 sip = &IA_SIN(ifa)->sin_addr; 352 } 353 354 if (0 == ((sip->s_addr ^ tip->s_addr) & 355 IA_MASKSIN(ifa)->sin_addr.s_addr)) 356 break; /* found it. */ 357 } 358 IF_ADDR_RUNLOCK(ifp); 359 if (sip == NULL) { 360 printf("%s: cannot find matching address\n", __func__); 361 return; 362 } 363 } 364 if (enaddr == NULL) 365 enaddr = carpaddr ? carpaddr : (u_char *)IF_LLADDR(ifp); 366 367 if ((m = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) 368 return; 369 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 370 2 * ifp->if_addrlen; 371 m->m_pkthdr.len = m->m_len; 372 M_ALIGN(m, m->m_len); 373 ah = mtod(m, struct arphdr *); 374 bzero((caddr_t)ah, m->m_len); 375 #ifdef MAC 376 mac_netinet_arp_send(ifp, m); 377 #endif 378 ah->ar_pro = htons(ETHERTYPE_IP); 379 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 380 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 381 ah->ar_op = htons(ARPOP_REQUEST); 382 bcopy(enaddr, ar_sha(ah), ah->ar_hln); 383 bcopy(sip, ar_spa(ah), ah->ar_pln); 384 bcopy(tip, ar_tpa(ah), ah->ar_pln); 385 sa.sa_family = AF_ARP; 386 sa.sa_len = 2; 387 388 /* Calculate link header for sending frame */ 389 bzero(&ro, sizeof(ro)); 390 linkhdrsize = sizeof(linkhdr); 391 error = arp_fillheader(ifp, ah, 1, linkhdr, &linkhdrsize); 392 if (error != 0 && error != EAFNOSUPPORT) { 393 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n", 394 if_name(ifp), error); 395 return; 396 } 397 398 ro.ro_prepend = linkhdr; 399 ro.ro_plen = linkhdrsize; 400 ro.ro_flags = 0; 401 402 m->m_flags |= M_BCAST; 403 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 404 (*ifp->if_output)(ifp, m, &sa, &ro); 405 ARPSTAT_INC(txrequests); 406 } 407 408 409 /* 410 * Resolve an IP address into an ethernet address - heavy version. 411 * Used internally by arpresolve(). 412 * We have already checked than we can't use existing lle without 413 * modification so we have to acquire LLE_EXCLUSIVE lle lock. 414 * 415 * On success, desten and flags are filled in and the function returns 0; 416 * If the packet must be held pending resolution, we return EWOULDBLOCK 417 * On other errors, we return the corresponding error code. 418 * Note that m_freem() handles NULL. 419 */ 420 static int 421 arpresolve_full(struct ifnet *ifp, int is_gw, int flags, struct mbuf *m, 422 const struct sockaddr *dst, u_char *desten, uint32_t *pflags, 423 struct llentry **plle) 424 { 425 struct llentry *la = NULL, *la_tmp; 426 struct mbuf *curr = NULL; 427 struct mbuf *next = NULL; 428 int error, renew; 429 char *lladdr; 430 int ll_len; 431 432 if (pflags != NULL) 433 *pflags = 0; 434 if (plle != NULL) 435 *plle = NULL; 436 437 if ((flags & LLE_CREATE) == 0) { 438 IF_AFDATA_RLOCK(ifp); 439 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 440 IF_AFDATA_RUNLOCK(ifp); 441 } 442 if (la == NULL && (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) == 0) { 443 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst); 444 if (la == NULL) { 445 log(LOG_DEBUG, 446 "arpresolve: can't allocate llinfo for %s on %s\n", 447 inet_ntoa(SIN(dst)->sin_addr), if_name(ifp)); 448 m_freem(m); 449 return (EINVAL); 450 } 451 452 IF_AFDATA_WLOCK(ifp); 453 LLE_WLOCK(la); 454 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 455 /* Prefer ANY existing lle over newly-created one */ 456 if (la_tmp == NULL) 457 lltable_link_entry(LLTABLE(ifp), la); 458 IF_AFDATA_WUNLOCK(ifp); 459 if (la_tmp != NULL) { 460 lltable_free_entry(LLTABLE(ifp), la); 461 la = la_tmp; 462 } 463 } 464 if (la == NULL) { 465 m_freem(m); 466 return (EINVAL); 467 } 468 469 if ((la->la_flags & LLE_VALID) && 470 ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { 471 if (flags & LLE_ADDRONLY) { 472 lladdr = la->ll_addr; 473 ll_len = ifp->if_addrlen; 474 } else { 475 lladdr = la->r_linkdata; 476 ll_len = la->r_hdrlen; 477 } 478 bcopy(lladdr, desten, ll_len); 479 480 /* Check if we have feedback request from arptimer() */ 481 if (la->r_skip_req != 0) { 482 LLE_REQ_LOCK(la); 483 la->r_skip_req = 0; /* Notify that entry was used */ 484 LLE_REQ_UNLOCK(la); 485 } 486 if (pflags != NULL) 487 *pflags = la->la_flags & (LLE_VALID|LLE_IFADDR); 488 if (plle) { 489 LLE_ADDREF(la); 490 *plle = la; 491 } 492 LLE_WUNLOCK(la); 493 return (0); 494 } 495 496 renew = (la->la_asked == 0 || la->la_expire != time_uptime); 497 /* 498 * There is an arptab entry, but no ethernet address 499 * response yet. Add the mbuf to the list, dropping 500 * the oldest packet if we have exceeded the system 501 * setting. 502 */ 503 if (m != NULL) { 504 if (la->la_numheld >= V_arp_maxhold) { 505 if (la->la_hold != NULL) { 506 next = la->la_hold->m_nextpkt; 507 m_freem(la->la_hold); 508 la->la_hold = next; 509 la->la_numheld--; 510 ARPSTAT_INC(dropped); 511 } 512 } 513 if (la->la_hold != NULL) { 514 curr = la->la_hold; 515 while (curr->m_nextpkt != NULL) 516 curr = curr->m_nextpkt; 517 curr->m_nextpkt = m; 518 } else 519 la->la_hold = m; 520 la->la_numheld++; 521 } 522 /* 523 * Return EWOULDBLOCK if we have tried less than arp_maxtries. It 524 * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH 525 * if we have already sent arp_maxtries ARP requests. Retransmit the 526 * ARP request, but not faster than one request per second. 527 */ 528 if (la->la_asked < V_arp_maxtries) 529 error = EWOULDBLOCK; /* First request. */ 530 else 531 error = is_gw != 0 ? EHOSTUNREACH : EHOSTDOWN; 532 533 if (renew) { 534 int canceled; 535 536 LLE_ADDREF(la); 537 la->la_expire = time_uptime; 538 canceled = callout_reset(&la->lle_timer, hz * V_arpt_down, 539 arptimer, la); 540 if (canceled) 541 LLE_REMREF(la); 542 la->la_asked++; 543 LLE_WUNLOCK(la); 544 arprequest(ifp, NULL, &SIN(dst)->sin_addr, NULL); 545 return (error); 546 } 547 548 LLE_WUNLOCK(la); 549 return (error); 550 } 551 552 /* 553 * Resolve an IP address into an ethernet address. 554 */ 555 int 556 arpresolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 557 char *desten, uint32_t *pflags, struct llentry **plle) 558 { 559 int error; 560 561 flags |= LLE_ADDRONLY; 562 error = arpresolve_full(ifp, 0, flags, NULL, dst, desten, pflags, plle); 563 return (error); 564 } 565 566 567 /* 568 * Lookups link header based on an IP address. 569 * On input: 570 * ifp is the interface we use 571 * is_gw != 0 if @dst represents gateway to some destination 572 * m is the mbuf. May be NULL if we don't have a packet. 573 * dst is the next hop, 574 * desten is the storage to put LL header. 575 * flags returns subset of lle flags: LLE_VALID | LLE_IFADDR 576 * 577 * On success, full/partial link header and flags are filled in and 578 * the function returns 0. 579 * If the packet must be held pending resolution, we return EWOULDBLOCK 580 * On other errors, we return the corresponding error code. 581 * Note that m_freem() handles NULL. 582 */ 583 int 584 arpresolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 585 const struct sockaddr *dst, u_char *desten, uint32_t *pflags, 586 struct llentry **plle) 587 { 588 struct llentry *la = NULL; 589 590 if (pflags != NULL) 591 *pflags = 0; 592 if (plle != NULL) 593 *plle = NULL; 594 595 if (m != NULL) { 596 if (m->m_flags & M_BCAST) { 597 /* broadcast */ 598 (void)memcpy(desten, 599 ifp->if_broadcastaddr, ifp->if_addrlen); 600 return (0); 601 } 602 if (m->m_flags & M_MCAST) { 603 /* multicast */ 604 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 605 return (0); 606 } 607 } 608 609 IF_AFDATA_RLOCK(ifp); 610 la = lla_lookup(LLTABLE(ifp), LLE_UNLOCKED, dst); 611 if (la != NULL && (la->r_flags & RLLE_VALID) != 0) { 612 /* Entry found, let's copy lle info */ 613 bcopy(la->r_linkdata, desten, la->r_hdrlen); 614 if (pflags != NULL) 615 *pflags = LLE_VALID | (la->r_flags & RLLE_IFADDR); 616 /* Check if we have feedback request from arptimer() */ 617 if (la->r_skip_req != 0) { 618 LLE_REQ_LOCK(la); 619 la->r_skip_req = 0; /* Notify that entry was used */ 620 LLE_REQ_UNLOCK(la); 621 } 622 IF_AFDATA_RUNLOCK(ifp); 623 return (0); 624 } 625 IF_AFDATA_RUNLOCK(ifp); 626 627 return (arpresolve_full(ifp, is_gw, la == NULL ? LLE_CREATE : 0, m, dst, 628 desten, pflags, plle)); 629 } 630 631 /* 632 * Common length and type checks are done here, 633 * then the protocol-specific routine is called. 634 */ 635 static void 636 arpintr(struct mbuf *m) 637 { 638 struct arphdr *ar; 639 struct ifnet *ifp; 640 char *layer; 641 int hlen; 642 643 ifp = m->m_pkthdr.rcvif; 644 645 if (m->m_len < sizeof(struct arphdr) && 646 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 647 ARP_LOG(LOG_NOTICE, "packet with short header received on %s\n", 648 if_name(ifp)); 649 return; 650 } 651 ar = mtod(m, struct arphdr *); 652 653 /* Check if length is sufficient */ 654 if (m->m_len < arphdr_len(ar)) { 655 m = m_pullup(m, arphdr_len(ar)); 656 if (m == NULL) { 657 ARP_LOG(LOG_NOTICE, "short packet received on %s\n", 658 if_name(ifp)); 659 return; 660 } 661 ar = mtod(m, struct arphdr *); 662 } 663 664 hlen = 0; 665 layer = ""; 666 switch (ntohs(ar->ar_hrd)) { 667 case ARPHRD_ETHER: 668 hlen = ETHER_ADDR_LEN; /* RFC 826 */ 669 layer = "ethernet"; 670 break; 671 case ARPHRD_IEEE802: 672 hlen = 6; /* RFC 1390, FDDI_ADDR_LEN */ 673 layer = "fddi"; 674 break; 675 case ARPHRD_ARCNET: 676 hlen = 1; /* RFC 1201, ARC_ADDR_LEN */ 677 layer = "arcnet"; 678 break; 679 case ARPHRD_INFINIBAND: 680 hlen = 20; /* RFC 4391, INFINIBAND_ALEN */ 681 layer = "infiniband"; 682 break; 683 case ARPHRD_IEEE1394: 684 hlen = 0; /* SHALL be 16 */ /* RFC 2734 */ 685 layer = "firewire"; 686 687 /* 688 * Restrict too long hardware addresses. 689 * Currently we are capable of handling 20-byte 690 * addresses ( sizeof(lle->ll_addr) ) 691 */ 692 if (ar->ar_hln >= 20) 693 hlen = 16; 694 break; 695 default: 696 ARP_LOG(LOG_NOTICE, 697 "packet with unknown hardware format 0x%02d received on " 698 "%s\n", ntohs(ar->ar_hrd), if_name(ifp)); 699 m_freem(m); 700 return; 701 } 702 703 if (hlen != 0 && hlen != ar->ar_hln) { 704 ARP_LOG(LOG_NOTICE, 705 "packet with invalid %s address length %d received on %s\n", 706 layer, ar->ar_hln, if_name(ifp)); 707 m_freem(m); 708 return; 709 } 710 711 ARPSTAT_INC(received); 712 switch (ntohs(ar->ar_pro)) { 713 #ifdef INET 714 case ETHERTYPE_IP: 715 in_arpinput(m); 716 return; 717 #endif 718 } 719 m_freem(m); 720 } 721 722 #ifdef INET 723 /* 724 * ARP for Internet protocols on 10 Mb/s Ethernet. 725 * Algorithm is that given in RFC 826. 726 * In addition, a sanity check is performed on the sender 727 * protocol address, to catch impersonators. 728 * We no longer handle negotiations for use of trailer protocol: 729 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 730 * along with IP replies if we wanted trailers sent to us, 731 * and also sent them in response to IP replies. 732 * This allowed either end to announce the desire to receive 733 * trailer packets. 734 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 735 * but formerly didn't normally send requests. 736 */ 737 static int log_arp_wrong_iface = 1; 738 static int log_arp_movements = 1; 739 static int log_arp_permanent_modify = 1; 740 static int allow_multicast = 0; 741 742 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 743 &log_arp_wrong_iface, 0, 744 "log arp packets arriving on the wrong interface"); 745 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 746 &log_arp_movements, 0, 747 "log arp replies from MACs different than the one in the cache"); 748 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 749 &log_arp_permanent_modify, 0, 750 "log arp replies from MACs different than the one in the permanent arp entry"); 751 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, allow_multicast, CTLFLAG_RW, 752 &allow_multicast, 0, "accept multicast addresses"); 753 754 static void 755 in_arpinput(struct mbuf *m) 756 { 757 struct rm_priotracker in_ifa_tracker; 758 struct arphdr *ah; 759 struct ifnet *ifp = m->m_pkthdr.rcvif; 760 struct llentry *la = NULL, *la_tmp; 761 struct ifaddr *ifa; 762 struct in_ifaddr *ia; 763 struct sockaddr sa; 764 struct in_addr isaddr, itaddr, myaddr; 765 u_int8_t *enaddr = NULL; 766 int op; 767 int bridged = 0, is_bridge = 0; 768 int carped; 769 struct sockaddr_in sin; 770 struct sockaddr *dst; 771 struct nhop4_basic nh4; 772 uint8_t linkhdr[LLE_MAX_LINKHDR]; 773 struct route ro; 774 size_t linkhdrsize; 775 int lladdr_off; 776 int error; 777 778 sin.sin_len = sizeof(struct sockaddr_in); 779 sin.sin_family = AF_INET; 780 sin.sin_addr.s_addr = 0; 781 782 if (ifp->if_bridge) 783 bridged = 1; 784 if (ifp->if_type == IFT_BRIDGE) 785 is_bridge = 1; 786 787 /* 788 * We already have checked that mbuf contains enough contiguous data 789 * to hold entire arp message according to the arp header. 790 */ 791 ah = mtod(m, struct arphdr *); 792 793 /* 794 * ARP is only for IPv4 so we can reject packets with 795 * a protocol length not equal to an IPv4 address. 796 */ 797 if (ah->ar_pln != sizeof(struct in_addr)) { 798 ARP_LOG(LOG_NOTICE, "requested protocol length != %zu\n", 799 sizeof(struct in_addr)); 800 goto drop; 801 } 802 803 if (allow_multicast == 0 && ETHER_IS_MULTICAST(ar_sha(ah))) { 804 ARP_LOG(LOG_NOTICE, "%*D is multicast\n", 805 ifp->if_addrlen, (u_char *)ar_sha(ah), ":"); 806 goto drop; 807 } 808 809 op = ntohs(ah->ar_op); 810 (void)memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 811 (void)memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 812 813 if (op == ARPOP_REPLY) 814 ARPSTAT_INC(rxreplies); 815 816 /* 817 * For a bridge, we want to check the address irrespective 818 * of the receive interface. (This will change slightly 819 * when we have clusters of interfaces). 820 */ 821 IN_IFADDR_RLOCK(&in_ifa_tracker); 822 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 823 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 824 ia->ia_ifp == ifp) && 825 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr && 826 (ia->ia_ifa.ifa_carp == NULL || 827 (*carp_iamatch_p)(&ia->ia_ifa, &enaddr))) { 828 ifa_ref(&ia->ia_ifa); 829 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 830 goto match; 831 } 832 } 833 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) 834 if (((bridged && ia->ia_ifp->if_bridge == ifp->if_bridge) || 835 ia->ia_ifp == ifp) && 836 isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) { 837 ifa_ref(&ia->ia_ifa); 838 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 839 goto match; 840 } 841 842 #define BDG_MEMBER_MATCHES_ARP(addr, ifp, ia) \ 843 (ia->ia_ifp->if_bridge == ifp->if_softc && \ 844 !bcmp(IF_LLADDR(ia->ia_ifp), IF_LLADDR(ifp), ifp->if_addrlen) && \ 845 addr == ia->ia_addr.sin_addr.s_addr) 846 /* 847 * Check the case when bridge shares its MAC address with 848 * some of its children, so packets are claimed by bridge 849 * itself (bridge_input() does it first), but they are really 850 * meant to be destined to the bridge member. 851 */ 852 if (is_bridge) { 853 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 854 if (BDG_MEMBER_MATCHES_ARP(itaddr.s_addr, ifp, ia)) { 855 ifa_ref(&ia->ia_ifa); 856 ifp = ia->ia_ifp; 857 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 858 goto match; 859 } 860 } 861 } 862 #undef BDG_MEMBER_MATCHES_ARP 863 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 864 865 /* 866 * No match, use the first inet address on the receive interface 867 * as a dummy address for the rest of the function. 868 */ 869 IF_ADDR_RLOCK(ifp); 870 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) 871 if (ifa->ifa_addr->sa_family == AF_INET && 872 (ifa->ifa_carp == NULL || 873 (*carp_iamatch_p)(ifa, &enaddr))) { 874 ia = ifatoia(ifa); 875 ifa_ref(ifa); 876 IF_ADDR_RUNLOCK(ifp); 877 goto match; 878 } 879 IF_ADDR_RUNLOCK(ifp); 880 881 /* 882 * If bridging, fall back to using any inet address. 883 */ 884 IN_IFADDR_RLOCK(&in_ifa_tracker); 885 if (!bridged || (ia = TAILQ_FIRST(&V_in_ifaddrhead)) == NULL) { 886 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 887 goto drop; 888 } 889 ifa_ref(&ia->ia_ifa); 890 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 891 match: 892 if (!enaddr) 893 enaddr = (u_int8_t *)IF_LLADDR(ifp); 894 carped = (ia->ia_ifa.ifa_carp != NULL); 895 myaddr = ia->ia_addr.sin_addr; 896 ifa_free(&ia->ia_ifa); 897 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) 898 goto drop; /* it's from me, ignore it. */ 899 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 900 ARP_LOG(LOG_NOTICE, "link address is broadcast for IP address " 901 "%s!\n", inet_ntoa(isaddr)); 902 goto drop; 903 } 904 905 if (ifp->if_addrlen != ah->ar_hln) { 906 ARP_LOG(LOG_WARNING, "from %*D: addr len: new %d, " 907 "i/f %d (ignored)\n", ifp->if_addrlen, 908 (u_char *) ar_sha(ah), ":", ah->ar_hln, 909 ifp->if_addrlen); 910 goto drop; 911 } 912 913 /* 914 * Warn if another host is using the same IP address, but only if the 915 * IP address isn't 0.0.0.0, which is used for DHCP only, in which 916 * case we suppress the warning to avoid false positive complaints of 917 * potential misconfiguration. 918 */ 919 if (!bridged && !carped && isaddr.s_addr == myaddr.s_addr && 920 myaddr.s_addr != 0) { 921 ARP_LOG(LOG_ERR, "%*D is using my IP address %s on %s!\n", 922 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 923 inet_ntoa(isaddr), ifp->if_xname); 924 itaddr = myaddr; 925 ARPSTAT_INC(dupips); 926 goto reply; 927 } 928 if (ifp->if_flags & IFF_STATICARP) 929 goto reply; 930 931 bzero(&sin, sizeof(sin)); 932 sin.sin_len = sizeof(struct sockaddr_in); 933 sin.sin_family = AF_INET; 934 sin.sin_addr = isaddr; 935 dst = (struct sockaddr *)&sin; 936 IF_AFDATA_RLOCK(ifp); 937 la = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 938 IF_AFDATA_RUNLOCK(ifp); 939 if (la != NULL) 940 arp_check_update_lle(ah, isaddr, ifp, bridged, la); 941 else if (itaddr.s_addr == myaddr.s_addr) { 942 /* 943 * Request/reply to our address, but no lle exists yet. 944 * Calculate full link prepend to use in lle. 945 */ 946 linkhdrsize = sizeof(linkhdr); 947 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr, 948 &linkhdrsize, &lladdr_off) != 0) 949 goto reply; 950 951 /* Allocate new entry */ 952 la = lltable_alloc_entry(LLTABLE(ifp), 0, dst); 953 if (la == NULL) { 954 955 /* 956 * lle creation may fail if source address belongs 957 * to non-directly connected subnet. However, we 958 * will try to answer the request instead of dropping 959 * frame. 960 */ 961 goto reply; 962 } 963 lltable_set_entry_addr(ifp, la, linkhdr, linkhdrsize, 964 lladdr_off); 965 966 IF_AFDATA_WLOCK(ifp); 967 LLE_WLOCK(la); 968 la_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 969 970 /* 971 * Check if lle still does not exists. 972 * If it does, that means that we either 973 * 1) have configured it explicitly, via 974 * 1a) 'arp -s' static entry or 975 * 1b) interface address static record 976 * or 977 * 2) it was the result of sending first packet to-host 978 * or 979 * 3) it was another arp reply packet we handled in 980 * different thread. 981 * 982 * In all cases except 3) we definitely need to prefer 983 * existing lle. For the sake of simplicity, prefer any 984 * existing lle over newly-create one. 985 */ 986 if (la_tmp == NULL) 987 lltable_link_entry(LLTABLE(ifp), la); 988 IF_AFDATA_WUNLOCK(ifp); 989 990 if (la_tmp == NULL) { 991 arp_mark_lle_reachable(la); 992 LLE_WUNLOCK(la); 993 } else { 994 /* Free newly-create entry and handle packet */ 995 lltable_free_entry(LLTABLE(ifp), la); 996 la = la_tmp; 997 la_tmp = NULL; 998 arp_check_update_lle(ah, isaddr, ifp, bridged, la); 999 /* arp_check_update_lle() returns @la unlocked */ 1000 } 1001 la = NULL; 1002 } 1003 reply: 1004 if (op != ARPOP_REQUEST) 1005 goto drop; 1006 ARPSTAT_INC(rxrequests); 1007 1008 if (itaddr.s_addr == myaddr.s_addr) { 1009 /* Shortcut.. the receiving interface is the target. */ 1010 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1011 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1012 } else { 1013 struct llentry *lle = NULL; 1014 1015 sin.sin_addr = itaddr; 1016 IF_AFDATA_RLOCK(ifp); 1017 lle = lla_lookup(LLTABLE(ifp), 0, (struct sockaddr *)&sin); 1018 IF_AFDATA_RUNLOCK(ifp); 1019 1020 if ((lle != NULL) && (lle->la_flags & LLE_PUB)) { 1021 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1022 (void)memcpy(ar_sha(ah), lle->ll_addr, ah->ar_hln); 1023 LLE_RUNLOCK(lle); 1024 } else { 1025 1026 if (lle != NULL) 1027 LLE_RUNLOCK(lle); 1028 1029 if (!V_arp_proxyall) 1030 goto drop; 1031 1032 /* XXX MRT use table 0 for arp reply */ 1033 if (fib4_lookup_nh_basic(0, itaddr, 0, 0, &nh4) != 0) 1034 goto drop; 1035 1036 /* 1037 * Don't send proxies for nodes on the same interface 1038 * as this one came out of, or we'll get into a fight 1039 * over who claims what Ether address. 1040 */ 1041 if (nh4.nh_ifp == ifp) 1042 goto drop; 1043 1044 (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1045 (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1046 1047 /* 1048 * Also check that the node which sent the ARP packet 1049 * is on the interface we expect it to be on. This 1050 * avoids ARP chaos if an interface is connected to the 1051 * wrong network. 1052 */ 1053 1054 /* XXX MRT use table 0 for arp checks */ 1055 if (fib4_lookup_nh_basic(0, isaddr, 0, 0, &nh4) != 0) 1056 goto drop; 1057 if (nh4.nh_ifp != ifp) { 1058 ARP_LOG(LOG_INFO, "proxy: ignoring request" 1059 " from %s via %s\n", 1060 inet_ntoa(isaddr), ifp->if_xname); 1061 goto drop; 1062 } 1063 1064 #ifdef DEBUG_PROXY 1065 printf("arp: proxying for %s\n", inet_ntoa(itaddr)); 1066 #endif 1067 } 1068 } 1069 1070 if (itaddr.s_addr == myaddr.s_addr && 1071 IN_LINKLOCAL(ntohl(itaddr.s_addr))) { 1072 /* RFC 3927 link-local IPv4; always reply by broadcast. */ 1073 #ifdef DEBUG_LINKLOCAL 1074 printf("arp: sending reply for link-local addr %s\n", 1075 inet_ntoa(itaddr)); 1076 #endif 1077 m->m_flags |= M_BCAST; 1078 m->m_flags &= ~M_MCAST; 1079 } else { 1080 /* default behaviour; never reply by broadcast. */ 1081 m->m_flags &= ~(M_BCAST|M_MCAST); 1082 } 1083 (void)memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 1084 (void)memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 1085 ah->ar_op = htons(ARPOP_REPLY); 1086 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 1087 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 1088 m->m_pkthdr.len = m->m_len; 1089 m->m_pkthdr.rcvif = NULL; 1090 sa.sa_family = AF_ARP; 1091 sa.sa_len = 2; 1092 1093 /* Calculate link header for sending frame */ 1094 bzero(&ro, sizeof(ro)); 1095 linkhdrsize = sizeof(linkhdr); 1096 error = arp_fillheader(ifp, ah, 0, linkhdr, &linkhdrsize); 1097 1098 /* 1099 * arp_fillheader() may fail due to lack of support inside encap request 1100 * routing. This is not necessary an error, AF_ARP can/should be handled 1101 * by if_output(). 1102 */ 1103 if (error != 0 && error != EAFNOSUPPORT) { 1104 ARP_LOG(LOG_ERR, "Failed to calculate ARP header on %s: %d\n", 1105 if_name(ifp), error); 1106 return; 1107 } 1108 1109 ro.ro_prepend = linkhdr; 1110 ro.ro_plen = linkhdrsize; 1111 ro.ro_flags = 0; 1112 1113 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1114 (*ifp->if_output)(ifp, m, &sa, &ro); 1115 ARPSTAT_INC(txreplies); 1116 return; 1117 1118 drop: 1119 m_freem(m); 1120 } 1121 #endif 1122 1123 /* 1124 * Checks received arp data against existing @la. 1125 * Updates lle state/performs notification if necessary. 1126 */ 1127 static void 1128 arp_check_update_lle(struct arphdr *ah, struct in_addr isaddr, struct ifnet *ifp, 1129 int bridged, struct llentry *la) 1130 { 1131 struct sockaddr sa; 1132 struct mbuf *m_hold, *m_hold_next; 1133 uint8_t linkhdr[LLE_MAX_LINKHDR]; 1134 size_t linkhdrsize; 1135 int lladdr_off; 1136 1137 LLE_WLOCK_ASSERT(la); 1138 1139 /* the following is not an error when doing bridging */ 1140 if (!bridged && la->lle_tbl->llt_ifp != ifp) { 1141 if (log_arp_wrong_iface) 1142 ARP_LOG(LOG_WARNING, "%s is on %s " 1143 "but got reply from %*D on %s\n", 1144 inet_ntoa(isaddr), 1145 la->lle_tbl->llt_ifp->if_xname, 1146 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 1147 ifp->if_xname); 1148 LLE_WUNLOCK(la); 1149 return; 1150 } 1151 if ((la->la_flags & LLE_VALID) && 1152 bcmp(ar_sha(ah), la->ll_addr, ifp->if_addrlen)) { 1153 if (la->la_flags & LLE_STATIC) { 1154 LLE_WUNLOCK(la); 1155 if (log_arp_permanent_modify) 1156 ARP_LOG(LOG_ERR, 1157 "%*D attempts to modify " 1158 "permanent entry for %s on %s\n", 1159 ifp->if_addrlen, 1160 (u_char *)ar_sha(ah), ":", 1161 inet_ntoa(isaddr), ifp->if_xname); 1162 return; 1163 } 1164 if (log_arp_movements) { 1165 ARP_LOG(LOG_INFO, "%s moved from %*D " 1166 "to %*D on %s\n", 1167 inet_ntoa(isaddr), 1168 ifp->if_addrlen, 1169 (u_char *)&la->ll_addr, ":", 1170 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 1171 ifp->if_xname); 1172 } 1173 } 1174 1175 /* Calculate full link prepend to use in lle */ 1176 linkhdrsize = sizeof(linkhdr); 1177 if (lltable_calc_llheader(ifp, AF_INET, ar_sha(ah), linkhdr, 1178 &linkhdrsize, &lladdr_off) != 0) 1179 return; 1180 1181 /* Check if something has changed */ 1182 if (memcmp(la->r_linkdata, linkhdr, linkhdrsize) != 0 || 1183 (la->la_flags & LLE_VALID) == 0) { 1184 /* Try to perform LLE update */ 1185 if (lltable_try_set_entry_addr(ifp, la, linkhdr, linkhdrsize, 1186 lladdr_off) == 0) 1187 return; 1188 1189 /* Clear fast path feedback request if set */ 1190 la->r_skip_req = 0; 1191 } 1192 1193 arp_mark_lle_reachable(la); 1194 1195 /* 1196 * The packets are all freed within the call to the output 1197 * routine. 1198 * 1199 * NB: The lock MUST be released before the call to the 1200 * output routine. 1201 */ 1202 if (la->la_hold != NULL) { 1203 m_hold = la->la_hold; 1204 la->la_hold = NULL; 1205 la->la_numheld = 0; 1206 lltable_fill_sa_entry(la, &sa); 1207 LLE_WUNLOCK(la); 1208 for (; m_hold != NULL; m_hold = m_hold_next) { 1209 m_hold_next = m_hold->m_nextpkt; 1210 m_hold->m_nextpkt = NULL; 1211 /* Avoid confusing lower layers. */ 1212 m_clrprotoflags(m_hold); 1213 (*ifp->if_output)(ifp, m_hold, &sa, NULL); 1214 } 1215 } else 1216 LLE_WUNLOCK(la); 1217 } 1218 1219 static void 1220 arp_mark_lle_reachable(struct llentry *la) 1221 { 1222 int canceled, wtime; 1223 1224 LLE_WLOCK_ASSERT(la); 1225 1226 la->ln_state = ARP_LLINFO_REACHABLE; 1227 EVENTHANDLER_INVOKE(lle_event, la, LLENTRY_RESOLVED); 1228 1229 if (!(la->la_flags & LLE_STATIC)) { 1230 LLE_ADDREF(la); 1231 la->la_expire = time_uptime + V_arpt_keep; 1232 wtime = V_arpt_keep - V_arp_maxtries * V_arpt_rexmit; 1233 if (wtime < 0) 1234 wtime = V_arpt_keep; 1235 canceled = callout_reset(&la->lle_timer, 1236 hz * wtime, arptimer, la); 1237 if (canceled) 1238 LLE_REMREF(la); 1239 } 1240 la->la_asked = 0; 1241 la->la_preempt = V_arp_maxtries; 1242 } 1243 1244 /* 1245 * Add pernament link-layer record for given interface address. 1246 */ 1247 static __noinline void 1248 arp_add_ifa_lle(struct ifnet *ifp, const struct sockaddr *dst) 1249 { 1250 struct llentry *lle, *lle_tmp; 1251 1252 /* 1253 * Interface address LLE record is considered static 1254 * because kernel code relies on LLE_STATIC flag to check 1255 * if these entries can be rewriten by arp updates. 1256 */ 1257 lle = lltable_alloc_entry(LLTABLE(ifp), LLE_IFADDR | LLE_STATIC, dst); 1258 if (lle == NULL) { 1259 log(LOG_INFO, "arp_ifinit: cannot create arp " 1260 "entry for interface address\n"); 1261 return; 1262 } 1263 1264 IF_AFDATA_WLOCK(ifp); 1265 LLE_WLOCK(lle); 1266 /* Unlink any entry if exists */ 1267 lle_tmp = lla_lookup(LLTABLE(ifp), LLE_EXCLUSIVE, dst); 1268 if (lle_tmp != NULL) 1269 lltable_unlink_entry(LLTABLE(ifp), lle_tmp); 1270 1271 lltable_link_entry(LLTABLE(ifp), lle); 1272 IF_AFDATA_WUNLOCK(ifp); 1273 1274 if (lle_tmp != NULL) 1275 EVENTHANDLER_INVOKE(lle_event, lle_tmp, LLENTRY_EXPIRED); 1276 1277 EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_RESOLVED); 1278 LLE_WUNLOCK(lle); 1279 if (lle_tmp != NULL) 1280 lltable_free_entry(LLTABLE(ifp), lle_tmp); 1281 } 1282 1283 void 1284 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1285 { 1286 const struct sockaddr_in *dst_in; 1287 const struct sockaddr *dst; 1288 1289 if (ifa->ifa_carp != NULL) 1290 return; 1291 1292 dst = ifa->ifa_addr; 1293 dst_in = (const struct sockaddr_in *)dst; 1294 1295 if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY) 1296 return; 1297 arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp)); 1298 1299 arp_add_ifa_lle(ifp, dst); 1300 } 1301 1302 void 1303 arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr) 1304 { 1305 1306 if (ntohl(addr.s_addr) != INADDR_ANY) 1307 arprequest(ifp, &addr, &addr, enaddr); 1308 } 1309 1310 /* 1311 * Sends gratuitous ARPs for each ifaddr to notify other 1312 * nodes about the address change. 1313 */ 1314 static __noinline void 1315 arp_handle_ifllchange(struct ifnet *ifp) 1316 { 1317 struct ifaddr *ifa; 1318 1319 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1320 if (ifa->ifa_addr->sa_family == AF_INET) 1321 arp_ifinit(ifp, ifa); 1322 } 1323 } 1324 1325 /* 1326 * A handler for interface link layer address change event. 1327 */ 1328 static void 1329 arp_iflladdr(void *arg __unused, struct ifnet *ifp) 1330 { 1331 1332 lltable_update_ifaddr(LLTABLE(ifp)); 1333 1334 if ((ifp->if_flags & IFF_UP) != 0) 1335 arp_handle_ifllchange(ifp); 1336 } 1337 1338 static void 1339 vnet_arp_init(void) 1340 { 1341 1342 if (IS_DEFAULT_VNET(curvnet)) { 1343 netisr_register(&arp_nh); 1344 iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event, 1345 arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 1346 } 1347 #ifdef VIMAGE 1348 else 1349 netisr_register_vnet(&arp_nh); 1350 #endif 1351 } 1352 VNET_SYSINIT(vnet_arp_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND, 1353 vnet_arp_init, 0); 1354 1355 #ifdef VIMAGE 1356 /* 1357 * We have to unregister ARP along with IP otherwise we risk doing INADDR_HASH 1358 * lookups after destroying the hash. Ideally this would go on SI_ORDER_3.5. 1359 */ 1360 static void 1361 vnet_arp_destroy(__unused void *arg) 1362 { 1363 1364 netisr_unregister_vnet(&arp_nh); 1365 } 1366 VNET_SYSUNINIT(vnet_arp_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, 1367 vnet_arp_destroy, NULL); 1368 #endif 1369