1 /* 2 * Copyright (c) 2004, 2005 The DragonFly Project. All rights reserved. 3 * 4 * This code is derived from software contributed to The DragonFly Project 5 * by Jeffrey M. Hsu. 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 DragonFly Project nor the names of its 16 * contributors may be used to endorse or promote products derived 17 * from this software without specific, prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 21 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 22 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 23 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 24 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 25 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 26 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 27 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 28 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 29 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * 37 * Redistribution and use in source and binary forms, with or without 38 * modification, are permitted provided that the following conditions 39 * are met: 40 * 1. Redistributions of source code must retain the above copyright 41 * notice, this list of conditions and the following disclaimer. 42 * 2. Redistributions in binary form must reproduce the above copyright 43 * notice, this list of conditions and the following disclaimer in the 44 * documentation and/or other materials provided with the distribution. 45 * 3. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgement: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * 4. Neither the name of the University nor the names of its contributors 50 * may be used to endorse or promote products derived from this software 51 * without specific prior written permission. 52 * 53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 56 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 * 65 * @(#)if_ether.c 8.1 (Berkeley) 6/10/93 66 * $FreeBSD: src/sys/netinet/if_ether.c,v 1.64.2.23 2003/04/11 07:23:15 fjoe Exp $ 67 * $DragonFly: src/sys/netinet/if_ether.c,v 1.45 2008/03/07 11:34:20 sephe Exp $ 68 */ 69 70 /* 71 * Ethernet address resolution protocol. 72 * TODO: 73 * add "inuse/lock" bit (or ref. count) along with valid bit 74 */ 75 76 #include "opt_inet.h" 77 #include "opt_carp.h" 78 79 #include <sys/param.h> 80 #include <sys/kernel.h> 81 #include <sys/queue.h> 82 #include <sys/sysctl.h> 83 #include <sys/systm.h> 84 #include <sys/mbuf.h> 85 #include <sys/malloc.h> 86 #include <sys/socket.h> 87 #include <sys/syslog.h> 88 89 #include <net/if.h> 90 #include <net/if_dl.h> 91 #include <net/if_types.h> 92 #include <net/route.h> 93 #include <net/netisr.h> 94 #include <net/if_llc.h> 95 96 #include <netinet/in.h> 97 #include <netinet/in_var.h> 98 #include <netinet/if_ether.h> 99 100 #include <sys/thread2.h> 101 #include <sys/msgport2.h> 102 #include <net/netmsg2.h> 103 104 105 #ifdef CARP 106 #include <netinet/ip_carp.h> 107 #endif 108 109 #define SIN(s) ((struct sockaddr_in *)s) 110 #define SDL(s) ((struct sockaddr_dl *)s) 111 112 SYSCTL_DECL(_net_link_ether); 113 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 114 115 /* timer values */ 116 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 117 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 118 static int arpt_down = 20; /* once declared down, don't send for 20 sec */ 119 120 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW, 121 &arpt_prune, 0, ""); 122 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 123 &arpt_keep, 0, ""); 124 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW, 125 &arpt_down, 0, ""); 126 127 #define rt_expire rt_rmx.rmx_expire 128 129 struct llinfo_arp { 130 LIST_ENTRY(llinfo_arp) la_le; 131 struct rtentry *la_rt; 132 struct mbuf *la_hold; /* last packet until resolved/timeout */ 133 u_short la_preempt; /* countdown for pre-expiry arps */ 134 u_short la_asked; /* #times we QUERIED following expiration */ 135 }; 136 137 static LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU]; 138 139 static int arp_maxtries = 5; 140 static int useloopback = 1; /* use loopback interface for local traffic */ 141 static int arp_proxyall = 0; 142 143 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 144 &arp_maxtries, 0, ""); 145 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 146 &useloopback, 0, ""); 147 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 148 &arp_proxyall, 0, ""); 149 150 void arprequest_acces(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, u_char *enaddr); 151 static void arp_rtrequest (int, struct rtentry *, struct rt_addrinfo *); 152 static void arprequest (struct ifnet *, 153 struct in_addr *, struct in_addr *, u_char *); 154 static void arpintr(struct netmsg *); 155 static void arptfree (struct llinfo_arp *); 156 static void arptimer (void *); 157 static struct llinfo_arp 158 *arplookup (in_addr_t addr, boolean_t create, boolean_t proxy); 159 #ifdef INET 160 static void in_arpinput (struct mbuf *); 161 #endif 162 163 static struct callout arptimer_ch[MAXCPU]; 164 165 /* 166 * Timeout routine. Age arp_tab entries periodically. 167 */ 168 /* ARGSUSED */ 169 static void 170 arptimer(void *ignored_arg) 171 { 172 struct llinfo_arp *la, *nla; 173 174 crit_enter(); 175 LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) { 176 if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second) 177 arptfree(la); 178 } 179 callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL); 180 crit_exit(); 181 } 182 183 /* 184 * Parallel to llc_rtrequest. 185 */ 186 static void 187 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 188 { 189 struct sockaddr *gate = rt->rt_gateway; 190 struct llinfo_arp *la = rt->rt_llinfo; 191 192 struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK }; 193 static boolean_t arpinit_done[MAXCPU]; 194 195 if (!arpinit_done[mycpuid]) { 196 arpinit_done[mycpuid] = TRUE; 197 callout_init(&arptimer_ch[mycpuid]); 198 callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL); 199 } 200 if (rt->rt_flags & RTF_GATEWAY) 201 return; 202 203 switch (req) { 204 case RTM_ADD: 205 /* 206 * XXX: If this is a manually added route to interface 207 * such as older version of routed or gated might provide, 208 * restore cloning bit. 209 */ 210 if (!(rt->rt_flags & RTF_HOST) && 211 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 212 rt->rt_flags |= RTF_CLONING; 213 if (rt->rt_flags & RTF_CLONING) { 214 /* 215 * Case 1: This route should come from a route to iface. 216 */ 217 rt_setgate(rt, rt_key(rt), 218 (struct sockaddr *)&null_sdl); 219 gate = rt->rt_gateway; 220 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 221 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 222 rt->rt_expire = time_second; 223 break; 224 } 225 /* Announce a new entry if requested. */ 226 if (rt->rt_flags & RTF_ANNOUNCE) { 227 lwkt_serialize_enter(rt->rt_ifp->if_serializer); 228 arprequest(rt->rt_ifp, 229 &SIN(rt_key(rt))->sin_addr, 230 &SIN(rt_key(rt))->sin_addr, 231 LLADDR(SDL(gate))); 232 lwkt_serialize_exit(rt->rt_ifp->if_serializer); 233 } 234 /*FALLTHROUGH*/ 235 case RTM_RESOLVE: 236 if (gate->sa_family != AF_LINK || 237 gate->sa_len < sizeof(struct sockaddr_dl)) { 238 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n"); 239 break; 240 } 241 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 242 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 243 if (la != NULL) 244 break; /* This happens on a route change */ 245 /* 246 * Case 2: This route may come from cloning, or a manual route 247 * add with a LL address. 248 */ 249 R_Malloc(la, struct llinfo_arp *, sizeof *la); 250 rt->rt_llinfo = la; 251 if (la == NULL) { 252 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n"); 253 break; 254 } 255 bzero(la, sizeof *la); 256 la->la_rt = rt; 257 rt->rt_flags |= RTF_LLINFO; 258 LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le); 259 260 #ifdef INET 261 /* 262 * This keeps the multicast addresses from showing up 263 * in `arp -a' listings as unresolved. It's not actually 264 * functional. Then the same for broadcast. 265 */ 266 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) { 267 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 268 LLADDR(SDL(gate))); 269 SDL(gate)->sdl_alen = 6; 270 rt->rt_expire = 0; 271 } 272 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 273 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 274 rt->rt_ifp->if_addrlen); 275 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 276 rt->rt_expire = 0; 277 } 278 #endif 279 280 if (SIN(rt_key(rt))->sin_addr.s_addr == 281 (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) { 282 /* 283 * This test used to be 284 * if (loif.if_flags & IFF_UP) 285 * It allowed local traffic to be forced 286 * through the hardware by configuring the 287 * loopback down. However, it causes problems 288 * during network configuration for boards 289 * that can't receive packets they send. It 290 * is now necessary to clear "useloopback" and 291 * remove the route to force traffic out to 292 * the hardware. 293 */ 294 rt->rt_expire = 0; 295 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 296 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 297 if (useloopback) 298 rt->rt_ifp = loif; 299 } 300 break; 301 302 case RTM_DELETE: 303 if (la == NULL) 304 break; 305 LIST_REMOVE(la, la_le); 306 rt->rt_llinfo = NULL; 307 rt->rt_flags &= ~RTF_LLINFO; 308 if (la->la_hold != NULL) 309 m_freem(la->la_hold); 310 Free(la); 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 static void 321 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 322 u_char *enaddr) 323 { 324 struct mbuf *m; 325 struct ether_header *eh; 326 struct arphdr *ah; 327 struct sockaddr sa; 328 u_short ar_hrd; 329 330 ASSERT_SERIALIZED(ifp->if_serializer); 331 332 if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL) 333 return; 334 m->m_pkthdr.rcvif = (struct ifnet *)NULL; 335 336 switch (ifp->if_type) { 337 case IFT_ETHER: 338 /* 339 * This may not be correct for types not explicitly 340 * listed, but this is our best guess 341 */ 342 default: 343 ar_hrd = htons(ARPHRD_ETHER); 344 345 m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 346 m->m_pkthdr.len = m->m_len; 347 MH_ALIGN(m, m->m_len); 348 349 eh = (struct ether_header *)sa.sa_data; 350 /* if_output() will not swap */ 351 eh->ether_type = htons(ETHERTYPE_ARP); 352 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen); 353 354 ah = mtod(m, struct arphdr *); 355 break; 356 } 357 358 ah->ar_hrd = ar_hrd; 359 ah->ar_pro = htons(ETHERTYPE_IP); 360 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 361 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 362 ah->ar_op = htons(ARPOP_REQUEST); 363 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 364 memset(ar_tha(ah), 0, ah->ar_hln); 365 memcpy(ar_spa(ah), sip, ah->ar_pln); 366 memcpy(ar_tpa(ah), tip, ah->ar_pln); 367 368 sa.sa_family = AF_UNSPEC; 369 sa.sa_len = sizeof sa; 370 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)NULL); 371 } 372 373 /* 374 * Resolve an IP address into an ethernet address. If success, 375 * desten is filled in. If there is no entry in arptab, 376 * set one up and broadcast a request for the IP address. 377 * Hold onto this mbuf and resend it once the address 378 * is finally resolved. A return value of 1 indicates 379 * that desten has been filled in and the packet should be sent 380 * normally; a 0 return indicates that the packet has been 381 * taken over here, either now or for later transmission. 382 */ 383 int 384 arpresolve( 385 struct ifnet *ifp, 386 struct rtentry *rt0, 387 struct mbuf *m, 388 struct sockaddr *dst, 389 u_char *desten) 390 { 391 struct rtentry *rt; 392 struct llinfo_arp *la = NULL; 393 struct sockaddr_dl *sdl; 394 395 if (m->m_flags & M_BCAST) { /* broadcast */ 396 memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 397 return (1); 398 } 399 if (m->m_flags & M_MCAST) {/* multicast */ 400 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 401 return (1); 402 } 403 if (rt0 != NULL) { 404 if (rt_llroute(dst, rt0, &rt) != 0) { 405 m_freem(m); 406 return 0; 407 } 408 la = rt->rt_llinfo; 409 } 410 if (la == NULL) { 411 la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, FALSE); 412 if (la != NULL) 413 rt = la->la_rt; 414 } 415 if (la == NULL || rt == NULL) { 416 log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n", 417 inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ", 418 rt ? "rt" : ""); 419 m_freem(m); 420 return (0); 421 } 422 sdl = SDL(rt->rt_gateway); 423 /* 424 * Check the address family and length is valid, the address 425 * is resolved; otherwise, try to resolve. 426 */ 427 if ((rt->rt_expire == 0 || rt->rt_expire > time_second) && 428 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 429 /* 430 * If entry has an expiry time and it is approaching, 431 * see if we need to send an ARP request within this 432 * arpt_down interval. 433 */ 434 if ((rt->rt_expire != 0) && 435 (time_second + la->la_preempt > rt->rt_expire)) { 436 arprequest(ifp, 437 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 438 &SIN(dst)->sin_addr, 439 IF_LLADDR(ifp)); 440 la->la_preempt--; 441 } 442 443 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 444 return 1; 445 } 446 /* 447 * If ARP is disabled on this interface, stop. 448 * XXX 449 * Probably should not allocate empty llinfo struct if we are 450 * not going to be sending out an arp request. 451 */ 452 if (ifp->if_flags & IFF_NOARP) { 453 m_freem(m); 454 return (0); 455 } 456 /* 457 * There is an arptab entry, but no ethernet address 458 * response yet. Replace the held mbuf with this 459 * latest one. 460 */ 461 if (la->la_hold != NULL) 462 m_freem(la->la_hold); 463 la->la_hold = m; 464 if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) { 465 rt->rt_flags &= ~RTF_REJECT; 466 if (la->la_asked == 0 || rt->rt_expire != time_second) { 467 rt->rt_expire = time_second; 468 if (la->la_asked++ < arp_maxtries) { 469 arprequest(ifp, 470 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 471 &SIN(dst)->sin_addr, 472 IF_LLADDR(ifp)); 473 } else { 474 rt->rt_flags |= RTF_REJECT; 475 rt->rt_expire += arpt_down; 476 la->la_asked = 0; 477 la->la_preempt = arp_maxtries; 478 } 479 480 } 481 } 482 return (0); 483 } 484 485 /* 486 * Common length and type checks are done here, 487 * then the protocol-specific routine is called. 488 */ 489 static void 490 arpintr(struct netmsg *msg) 491 { 492 struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet; 493 struct arphdr *ar; 494 u_short ar_hrd; 495 496 if (m->m_len < sizeof(struct arphdr) && 497 ((m = m_pullup(m, sizeof(struct arphdr))) == NULL)) { 498 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 499 goto out2; 500 } 501 ar = mtod(m, struct arphdr *); 502 503 ar_hrd = ntohs(ar->ar_hrd); 504 if (ar_hrd != ARPHRD_ETHER && 505 ar_hrd != ARPHRD_IEEE802) { 506 log(LOG_ERR, 507 "arp: unknown hardware address format (0x%2D)\n", 508 (unsigned char *)&ar->ar_hrd, ""); 509 goto out1; 510 } 511 512 if (m->m_pkthdr.len < arphdr_len(ar) && 513 (m = m_pullup(m, arphdr_len(ar))) == NULL) { 514 log(LOG_ERR, "arp: runt packet\n"); 515 goto out1; 516 } 517 518 switch (ntohs(ar->ar_pro)) { 519 #ifdef INET 520 case ETHERTYPE_IP: 521 in_arpinput(m); 522 goto out2; 523 #endif 524 } 525 out1: 526 m_freem(m); 527 out2: 528 ; 529 /* msg was embedded in the mbuf, do not reply! */ 530 } 531 532 #ifdef INET 533 /* 534 * ARP for Internet protocols on 10 Mb/s Ethernet. 535 * Algorithm is that given in RFC 826. 536 * In addition, a sanity check is performed on the sender 537 * protocol address, to catch impersonators. 538 * We no longer handle negotiations for use of trailer protocol: 539 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 540 * along with IP replies if we wanted trailers sent to us, 541 * and also sent them in response to IP replies. 542 * This allowed either end to announce the desire to receive 543 * trailer packets. 544 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 545 * but formerly didn't normally send requests. 546 */ 547 static int log_arp_wrong_iface = 1; 548 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 549 &log_arp_wrong_iface, 0, 550 "log arp packets arriving on the wrong interface"); 551 552 static void 553 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create, 554 boolean_t dologging) 555 { 556 struct arphdr *ah = mtod(m, struct arphdr *); 557 struct ifnet *ifp = m->m_pkthdr.rcvif; 558 struct llinfo_arp *la; 559 struct sockaddr_dl *sdl; 560 struct rtentry *rt; 561 int cpu = mycpuid; 562 563 la = arplookup(saddr, create, FALSE); 564 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) { 565 struct in_addr isaddr = { saddr }; 566 567 /* the following is not an error when doing bridging */ 568 if (rt->rt_ifp != ifp) { 569 if (dologging && log_arp_wrong_iface && cpu == 0) { 570 log(LOG_ERR, 571 "arp: %s is on %s " 572 "but got reply from %*D on %s\n", 573 inet_ntoa(isaddr), 574 rt->rt_ifp->if_xname, 575 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 576 ifp->if_xname); 577 } 578 return; 579 } 580 if (sdl->sdl_alen && 581 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 582 if (rt->rt_expire != 0) { 583 if (dologging && cpu == 0) { 584 log(LOG_INFO, 585 "arp: %s moved from %*D to %*D on %s\n", 586 inet_ntoa(isaddr), 587 ifp->if_addrlen, (u_char *)LLADDR(sdl), 588 ":", ifp->if_addrlen, 589 (u_char *)ar_sha(ah), ":", 590 ifp->if_xname); 591 } 592 } else { 593 if (dologging && cpu == 0) { 594 log(LOG_ERR, 595 "arp: %*D attempts to modify permanent entry for %s on %s\n", 596 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 597 inet_ntoa(isaddr), ifp->if_xname); 598 } 599 return; 600 } 601 } 602 /* 603 * sanity check for the address length. 604 * XXX this does not work for protocols with variable address 605 * length. -is 606 */ 607 if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln && 608 cpu == 0) 609 { 610 log(LOG_WARNING, 611 "arp from %*D: new addr len %d, was %d", 612 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 613 ah->ar_hln, sdl->sdl_alen); 614 } 615 if (ifp->if_addrlen != ah->ar_hln) { 616 if (dologging && cpu == 0) { 617 log(LOG_WARNING, 618 "arp from %*D: addr len: new %d, i/f %d (ignored)", 619 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 620 ah->ar_hln, ifp->if_addrlen); 621 } 622 return; 623 } 624 memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln); 625 if (rt->rt_expire != 0) 626 rt->rt_expire = time_second + arpt_keep; 627 rt->rt_flags &= ~RTF_REJECT; 628 la->la_asked = 0; 629 la->la_preempt = arp_maxtries; 630 631 /* 632 * This particular cpu might have been holding an mbuf 633 * pending ARP resolution. If so, transmit the mbuf now. 634 */ 635 if (la->la_hold != NULL) { 636 m_adj(la->la_hold, sizeof(struct ether_header)); 637 lwkt_serialize_enter(ifp->if_serializer); 638 (*ifp->if_output)(ifp, la->la_hold, rt_key(rt), rt); 639 lwkt_serialize_exit(ifp->if_serializer); 640 la->la_hold = NULL; 641 } 642 } 643 } 644 645 #ifdef SMP 646 647 struct netmsg_arp_update { 648 struct netmsg netmsg; 649 struct mbuf *m; 650 in_addr_t saddr; 651 boolean_t create; 652 }; 653 654 static void arp_update_msghandler(struct netmsg *netmsg); 655 656 #endif 657 658 /* 659 * Called from arpintr() - this routine is run from a single cpu. 660 */ 661 static void 662 in_arpinput(struct mbuf *m) 663 { 664 struct arphdr *ah; 665 struct ifnet *ifp = m->m_pkthdr.rcvif; 666 struct ether_header *eh; 667 struct rtentry *rt; 668 struct ifaddr_container *ifac; 669 struct in_ifaddr *ia; 670 struct sockaddr sa; 671 struct in_addr isaddr, itaddr, myaddr; 672 #ifdef SMP 673 struct netmsg_arp_update msg; 674 #endif 675 u_int8_t *enaddr = NULL; 676 int op; 677 int req_len; 678 679 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 680 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 681 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 682 return; 683 } 684 685 ah = mtod(m, struct arphdr *); 686 op = ntohs(ah->ar_op); 687 memcpy(&isaddr, ar_spa(ah), sizeof isaddr); 688 memcpy(&itaddr, ar_tpa(ah), sizeof itaddr); 689 /* 690 * Check both target and sender IP addresses: 691 * 692 * If we receive the packet on the interface owning the address, 693 * then accept the address. 694 * 695 * For a bridge, we accept the address if the receive interface and 696 * the interface owning the address are on the same bridge. 697 * (This will change slightly when we have clusters of interfaces). 698 */ 699 LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { 700 /* Skip all ia's which don't match */ 701 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 702 continue; 703 704 if (ia->ia_ifp == ifp) 705 goto match; 706 707 if (ifp->if_bridge && ia->ia_ifp && 708 ifp->if_bridge == ia->ia_ifp->if_bridge) 709 goto match; 710 711 #ifdef CARP 712 /* 713 * If the interface does not match, but the recieving interface 714 * is part of carp, we call carp_iamatch to see if this is a 715 * request for the virtual host ip. 716 * XXX: This is really ugly! 717 */ 718 if (ifp->if_carp != NULL && 719 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr) && 720 itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) 721 goto match; 722 #endif 723 } 724 LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) { 725 /* Skip all ia's which don't match */ 726 if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 727 continue; 728 729 if (ia->ia_ifp == ifp) 730 goto match; 731 732 if (ifp->if_bridge && ia->ia_ifp && 733 ifp->if_bridge == ia->ia_ifp->if_bridge) 734 goto match; 735 } 736 /* 737 * No match, use the first inet address on the receive interface 738 * as a dummy address for the rest of the function. 739 */ 740 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 741 struct ifaddr *ifa = ifac->ifa; 742 743 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 744 ia = ifatoia(ifa); 745 goto match; 746 } 747 } 748 /* 749 * If we got here, we didn't find any suitable interface, 750 * so drop the packet. 751 */ 752 m_freem(m); 753 return; 754 755 match: 756 if (!enaddr) 757 enaddr = (u_int8_t *)IF_LLADDR(ifp); 758 myaddr = ia->ia_addr.sin_addr; 759 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) { 760 m_freem(m); /* it's from me, ignore it. */ 761 return; 762 } 763 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 764 log(LOG_ERR, 765 "arp: link address is broadcast for IP address %s!\n", 766 inet_ntoa(isaddr)); 767 m_freem(m); 768 return; 769 } 770 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 771 log(LOG_ERR, 772 "arp: %*D is using my IP address %s!\n", 773 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 774 inet_ntoa(isaddr)); 775 itaddr = myaddr; 776 goto reply; 777 } 778 #ifdef SMP 779 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0, 780 arp_update_msghandler); 781 msg.m = m; 782 msg.saddr = isaddr.s_addr; 783 msg.create = (itaddr.s_addr == myaddr.s_addr); 784 lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0); 785 #endif 786 arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr), 787 TRUE); 788 reply: 789 if (op != ARPOP_REQUEST) { 790 m_freem(m); 791 return; 792 } 793 if (itaddr.s_addr == myaddr.s_addr) { 794 /* I am the target */ 795 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 796 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 797 } else { 798 struct llinfo_arp *la; 799 800 la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY); 801 if (la == NULL) { 802 struct sockaddr_in sin; 803 804 if (!arp_proxyall) { 805 m_freem(m); 806 return; 807 } 808 809 bzero(&sin, sizeof sin); 810 sin.sin_family = AF_INET; 811 sin.sin_len = sizeof sin; 812 sin.sin_addr = itaddr; 813 814 rt = rtpurelookup((struct sockaddr *)&sin); 815 if (rt == NULL) { 816 m_freem(m); 817 return; 818 } 819 --rt->rt_refcnt; 820 /* 821 * Don't send proxies for nodes on the same interface 822 * as this one came out of, or we'll get into a fight 823 * over who claims what Ether address. 824 */ 825 if (rt->rt_ifp == ifp) { 826 m_freem(m); 827 return; 828 } 829 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 830 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 831 #ifdef DEBUG_PROXY 832 kprintf("arp: proxying for %s\n", inet_ntoa(itaddr)); 833 #endif 834 } else { 835 struct sockaddr_dl *sdl; 836 837 rt = la->la_rt; 838 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 839 sdl = SDL(rt->rt_gateway); 840 memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 841 } 842 } 843 844 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 845 memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 846 ah->ar_op = htons(ARPOP_REPLY); 847 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 848 switch (ifp->if_type) { 849 case IFT_ETHER: 850 /* 851 * May not be correct for types not explictly 852 * listed, but it is our best guess. 853 */ 854 default: 855 eh = (struct ether_header *)sa.sa_data; 856 memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost); 857 eh->ether_type = htons(ETHERTYPE_ARP); 858 break; 859 } 860 sa.sa_family = AF_UNSPEC; 861 sa.sa_len = sizeof sa; 862 lwkt_serialize_enter(ifp->if_serializer); 863 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 864 lwkt_serialize_exit(ifp->if_serializer); 865 return; 866 } 867 868 #ifdef SMP 869 870 static 871 void 872 arp_update_msghandler(struct netmsg *netmsg) 873 { 874 struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg; 875 int nextcpu; 876 877 arp_update_oncpu(msg->m, msg->saddr, msg->create, FALSE); 878 879 nextcpu = mycpuid + 1; 880 if (nextcpu < ncpus) { 881 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg); 882 } else { 883 lwkt_replymsg(&msg->netmsg.nm_lmsg, 0); 884 } 885 } 886 887 #endif 888 889 #endif 890 891 /* 892 * Free an arp entry. If the arp entry is actively referenced or represents 893 * a static entry we only clear it back to an unresolved state, otherwise 894 * we destroy the entry entirely. 895 * 896 * Note that static entries are created when route add ... -interface is used 897 * to create an interface route to a (direct) destination. 898 */ 899 static void 900 arptfree(struct llinfo_arp *la) 901 { 902 struct rtentry *rt = la->la_rt; 903 struct sockaddr_dl *sdl; 904 905 if (rt == NULL) 906 panic("arptfree"); 907 sdl = SDL(rt->rt_gateway); 908 if (sdl != NULL && 909 ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) || 910 (rt->rt_flags & RTF_STATIC))) { 911 sdl->sdl_alen = 0; 912 la->la_preempt = la->la_asked = 0; 913 rt->rt_flags &= ~RTF_REJECT; 914 return; 915 } 916 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 917 } 918 919 /* 920 * Lookup or enter a new address in arptab. 921 */ 922 static struct llinfo_arp * 923 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy) 924 { 925 struct rtentry *rt; 926 struct sockaddr_inarp sin = { sizeof sin, AF_INET }; 927 const char *why = NULL; 928 929 sin.sin_addr.s_addr = addr; 930 sin.sin_other = proxy ? SIN_PROXY : 0; 931 if (create) 932 rt = rtlookup((struct sockaddr *)&sin); 933 else 934 rt = rtpurelookup((struct sockaddr *)&sin); 935 if (rt == NULL) 936 return (NULL); 937 rt->rt_refcnt--; 938 939 if (rt->rt_flags & RTF_GATEWAY) 940 why = "host is not on local network"; 941 else if (!(rt->rt_flags & RTF_LLINFO)) 942 why = "could not allocate llinfo"; 943 else if (rt->rt_gateway->sa_family != AF_LINK) 944 why = "gateway route is not ours"; 945 946 if (why) { 947 if (create) { 948 log(LOG_DEBUG, "arplookup %s failed: %s\n", 949 inet_ntoa(sin.sin_addr), why); 950 } 951 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) { 952 /* No references to this route. Purge it. */ 953 rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 954 rt_mask(rt), rt->rt_flags, NULL); 955 } 956 return (NULL); 957 } 958 return (rt->rt_llinfo); 959 } 960 961 void 962 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 963 { 964 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) 965 arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, 966 IF_LLADDR(ifp)); 967 ifa->ifa_rtrequest = arp_rtrequest; 968 ifa->ifa_flags |= RTF_CLONING; 969 } 970 971 void 972 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 973 { 974 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) 975 arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, 976 enaddr); 977 ifa->ifa_rtrequest = arp_rtrequest; 978 ifa->ifa_flags |= RTF_CLONING; 979 } 980 981 static void 982 arp_init(void) 983 { 984 int cpu; 985 986 for (cpu = 0; cpu < ncpus2; cpu++) 987 LIST_INIT(&llinfo_arp_list[cpu]); 988 netisr_register(NETISR_ARP, cpu0_portfn, arpintr); 989 } 990 991 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 992