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 1.59 2008/11/22 11:03:35 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 #include <sys/lock.h> 89 90 #include <net/if.h> 91 #include <net/if_dl.h> 92 #include <net/if_types.h> 93 #include <net/route.h> 94 #include <net/netisr.h> 95 #include <net/if_llc.h> 96 97 #include <netinet/in.h> 98 #include <netinet/in_var.h> 99 #include <netinet/if_ether.h> 100 101 #include <sys/thread2.h> 102 #include <sys/msgport2.h> 103 #include <net/netmsg2.h> 104 #include <sys/mplock2.h> 105 106 #ifdef CARP 107 #include <netinet/ip_carp.h> 108 #endif 109 110 #define SIN(s) ((struct sockaddr_in *)s) 111 #define SDL(s) ((struct sockaddr_dl *)s) 112 113 SYSCTL_DECL(_net_link_ether); 114 SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); 115 116 /* timer values */ 117 static int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 118 static int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 119 static int arpt_down = 20; /* once declared down, don't send for 20 sec */ 120 121 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, prune_intvl, CTLFLAG_RW, 122 &arpt_prune, 0, ""); 123 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, 124 &arpt_keep, 0, ""); 125 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, host_down_time, CTLFLAG_RW, 126 &arpt_down, 0, ""); 127 128 #define rt_expire rt_rmx.rmx_expire 129 130 struct llinfo_arp { 131 LIST_ENTRY(llinfo_arp) la_le; 132 struct rtentry *la_rt; 133 struct mbuf *la_hold; /* last packet until resolved/timeout */ 134 struct lwkt_port *la_msgport; /* last packet's msgport */ 135 u_short la_preempt; /* countdown for pre-expiry arps */ 136 u_short la_asked; /* #times we QUERIED following expiration */ 137 }; 138 139 static LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU]; 140 141 static int arp_maxtries = 5; 142 static int useloopback = 1; /* use loopback interface for local traffic */ 143 static int arp_proxyall = 0; 144 static int arp_refresh = 60; /* refresh arp cache ~60 (not impl yet) */ 145 static int arp_restricted_match = 0; 146 147 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 148 &arp_maxtries, 0, "ARP resolution attempts before returning error"); 149 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 150 &useloopback, 0, "Use the loopback interface for local traffic"); 151 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 152 &arp_proxyall, 0, "Enable proxy ARP for all suitable requests"); 153 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, restricted_match, CTLFLAG_RW, 154 &arp_restricted_match, 0, "Only match against the sender"); 155 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, refresh, CTLFLAG_RW, 156 &arp_refresh, 0, "Preemptively refresh the ARP"); 157 158 static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 159 static void arprequest(struct ifnet *, const struct in_addr *, 160 const struct in_addr *, const u_char *); 161 static void arprequest_async(struct ifnet *, const struct in_addr *, 162 const struct in_addr *, const u_char *); 163 static void arpintr(netmsg_t msg); 164 static void arptfree(struct llinfo_arp *); 165 static void arptimer(void *); 166 static struct llinfo_arp * 167 arplookup(in_addr_t, boolean_t, boolean_t, boolean_t); 168 #ifdef INET 169 static void in_arpinput(struct mbuf *); 170 #endif 171 172 static struct callout arptimer_ch[MAXCPU]; 173 174 /* 175 * Timeout routine. Age arp_tab entries periodically. 176 */ 177 /* ARGSUSED */ 178 static void 179 arptimer(void *ignored_arg) 180 { 181 struct llinfo_arp *la, *nla; 182 183 crit_enter(); 184 LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) { 185 if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second) 186 arptfree(la); 187 } 188 callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL); 189 crit_exit(); 190 } 191 192 /* 193 * Parallel to llc_rtrequest. 194 */ 195 static void 196 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 197 { 198 struct sockaddr *gate = rt->rt_gateway; 199 struct llinfo_arp *la = rt->rt_llinfo; 200 201 struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK }; 202 static boolean_t arpinit_done[MAXCPU]; 203 204 if (!arpinit_done[mycpuid]) { 205 arpinit_done[mycpuid] = TRUE; 206 callout_init(&arptimer_ch[mycpuid]); 207 callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL); 208 } 209 if (rt->rt_flags & RTF_GATEWAY) 210 return; 211 212 switch (req) { 213 case RTM_ADD: 214 /* 215 * XXX: If this is a manually added route to interface 216 * such as older version of routed or gated might provide, 217 * restore cloning bit. 218 */ 219 if (!(rt->rt_flags & RTF_HOST) && 220 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 221 rt->rt_flags |= RTF_CLONING; 222 if (rt->rt_flags & RTF_CLONING) { 223 /* 224 * Case 1: This route should come from a route to iface. 225 */ 226 rt_setgate(rt, rt_key(rt), 227 (struct sockaddr *)&null_sdl, 228 RTL_DONTREPORT); 229 gate = rt->rt_gateway; 230 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 231 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 232 rt->rt_expire = time_second; 233 break; 234 } 235 /* Announce a new entry if requested. */ 236 if (rt->rt_flags & RTF_ANNOUNCE) { 237 arprequest_async(rt->rt_ifp, 238 &SIN(rt_key(rt))->sin_addr, 239 &SIN(rt_key(rt))->sin_addr, 240 LLADDR(SDL(gate))); 241 } 242 /*FALLTHROUGH*/ 243 case RTM_RESOLVE: 244 if (gate->sa_family != AF_LINK || 245 gate->sa_len < sizeof(struct sockaddr_dl)) { 246 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n"); 247 break; 248 } 249 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 250 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 251 if (la != NULL) 252 break; /* This happens on a route change */ 253 /* 254 * Case 2: This route may come from cloning, or a manual route 255 * add with a LL address. 256 */ 257 R_Malloc(la, struct llinfo_arp *, sizeof *la); 258 rt->rt_llinfo = la; 259 if (la == NULL) { 260 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n"); 261 break; 262 } 263 bzero(la, sizeof *la); 264 la->la_rt = rt; 265 rt->rt_flags |= RTF_LLINFO; 266 LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le); 267 268 #ifdef INET 269 /* 270 * This keeps the multicast addresses from showing up 271 * in `arp -a' listings as unresolved. It's not actually 272 * functional. Then the same for broadcast. 273 */ 274 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) { 275 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 276 LLADDR(SDL(gate))); 277 SDL(gate)->sdl_alen = 6; 278 rt->rt_expire = 0; 279 } 280 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 281 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 282 rt->rt_ifp->if_addrlen); 283 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 284 rt->rt_expire = 0; 285 } 286 #endif 287 288 if (SIN(rt_key(rt))->sin_addr.s_addr == 289 (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) { 290 /* 291 * This test used to be 292 * if (loif.if_flags & IFF_UP) 293 * It allowed local traffic to be forced 294 * through the hardware by configuring the 295 * loopback down. However, it causes problems 296 * during network configuration for boards 297 * that can't receive packets they send. It 298 * is now necessary to clear "useloopback" and 299 * remove the route to force traffic out to 300 * the hardware. 301 */ 302 rt->rt_expire = 0; 303 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 304 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 305 if (useloopback) 306 rt->rt_ifp = loif; 307 } 308 break; 309 310 case RTM_DELETE: 311 if (la == NULL) 312 break; 313 LIST_REMOVE(la, la_le); 314 rt->rt_llinfo = NULL; 315 rt->rt_flags &= ~RTF_LLINFO; 316 if (la->la_hold != NULL) 317 m_freem(la->la_hold); 318 Free(la); 319 break; 320 } 321 } 322 323 static struct mbuf * 324 arpreq_alloc(struct ifnet *ifp, const struct in_addr *sip, 325 const struct in_addr *tip, const u_char *enaddr) 326 { 327 struct mbuf *m; 328 struct arphdr *ah; 329 u_short ar_hrd; 330 331 if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL) 332 return NULL; 333 m->m_pkthdr.rcvif = NULL; 334 335 switch (ifp->if_type) { 336 case IFT_ETHER: 337 /* 338 * This may not be correct for types not explicitly 339 * listed, but this is our best guess 340 */ 341 default: 342 ar_hrd = htons(ARPHRD_ETHER); 343 344 m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 345 m->m_pkthdr.len = m->m_len; 346 MH_ALIGN(m, m->m_len); 347 348 ah = mtod(m, struct arphdr *); 349 break; 350 } 351 352 ah->ar_hrd = ar_hrd; 353 ah->ar_pro = htons(ETHERTYPE_IP); 354 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 355 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 356 ah->ar_op = htons(ARPOP_REQUEST); 357 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 358 memset(ar_tha(ah), 0, ah->ar_hln); 359 memcpy(ar_spa(ah), sip, ah->ar_pln); 360 memcpy(ar_tpa(ah), tip, ah->ar_pln); 361 362 return m; 363 } 364 365 static void 366 arpreq_send(struct ifnet *ifp, struct mbuf *m) 367 { 368 struct sockaddr sa; 369 struct ether_header *eh; 370 371 switch (ifp->if_type) { 372 case IFT_ETHER: 373 /* 374 * This may not be correct for types not explicitly 375 * listed, but this is our best guess 376 */ 377 default: 378 eh = (struct ether_header *)sa.sa_data; 379 /* if_output() will not swap */ 380 eh->ether_type = htons(ETHERTYPE_ARP); 381 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen); 382 break; 383 } 384 385 sa.sa_family = AF_UNSPEC; 386 sa.sa_len = sizeof(sa); 387 ifp->if_output(ifp, m, &sa, NULL); 388 } 389 390 static void 391 arpreq_send_handler(netmsg_t msg) 392 { 393 struct mbuf *m = msg->packet.nm_packet; 394 struct ifnet *ifp = msg->lmsg.u.ms_resultp; 395 396 arpreq_send(ifp, m); 397 /* nmsg was embedded in the mbuf, do not reply! */ 398 } 399 400 /* 401 * Broadcast an ARP request. Caller specifies: 402 * - arp header source ip address 403 * - arp header target ip address 404 * - arp header source ethernet address 405 * 406 * NOTE: Caller MUST NOT hold ifp's serializer 407 */ 408 static void 409 arprequest(struct ifnet *ifp, const struct in_addr *sip, 410 const struct in_addr *tip, const u_char *enaddr) 411 { 412 struct mbuf *m; 413 414 if (enaddr == NULL) { 415 if (ifp->if_bridge) { 416 enaddr = IF_LLADDR(ether_bridge_interface(ifp)); 417 } else { 418 enaddr = IF_LLADDR(ifp); 419 } 420 } 421 422 m = arpreq_alloc(ifp, sip, tip, enaddr); 423 if (m == NULL) 424 return; 425 arpreq_send(ifp, m); 426 } 427 428 /* 429 * Same as arprequest(), except: 430 * - Caller is allowed to hold ifp's serializer 431 * - Network output is done in protocol thead 432 */ 433 static void 434 arprequest_async(struct ifnet *ifp, const struct in_addr *sip, 435 const struct in_addr *tip, const u_char *enaddr) 436 { 437 struct mbuf *m; 438 struct netmsg_packet *pmsg; 439 440 if (enaddr == NULL) { 441 if (ifp->if_bridge) { 442 enaddr = IF_LLADDR(ether_bridge_interface(ifp)); 443 } else { 444 enaddr = IF_LLADDR(ifp); 445 } 446 } 447 m = arpreq_alloc(ifp, sip, tip, enaddr); 448 if (m == NULL) 449 return; 450 451 pmsg = &m->m_hdr.mh_netmsg; 452 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport, 453 0, arpreq_send_handler); 454 pmsg->nm_packet = m; 455 pmsg->base.lmsg.u.ms_resultp = ifp; 456 457 lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->base.lmsg); 458 } 459 460 /* 461 * Resolve an IP address into an ethernet address. If success, 462 * desten is filled in. If there is no entry in arptab, 463 * set one up and broadcast a request for the IP address. 464 * Hold onto this mbuf and resend it once the address 465 * is finally resolved. A return value of 1 indicates 466 * that desten has been filled in and the packet should be sent 467 * normally; a 0 return indicates that the packet has been 468 * taken over here, either now or for later transmission. 469 */ 470 int 471 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 472 struct sockaddr *dst, u_char *desten) 473 { 474 struct rtentry *rt; 475 struct llinfo_arp *la = NULL; 476 struct sockaddr_dl *sdl; 477 478 if (m->m_flags & M_BCAST) { /* broadcast */ 479 memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 480 return (1); 481 } 482 if (m->m_flags & M_MCAST) {/* multicast */ 483 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 484 return (1); 485 } 486 if (rt0 != NULL) { 487 if (rt_llroute(dst, rt0, &rt) != 0) { 488 m_freem(m); 489 return 0; 490 } 491 la = rt->rt_llinfo; 492 } 493 if (la == NULL) { 494 la = arplookup(SIN(dst)->sin_addr.s_addr, 495 TRUE, RTL_REPORTMSG, FALSE); 496 if (la != NULL) 497 rt = la->la_rt; 498 } 499 if (la == NULL || rt == NULL) { 500 log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s%s%s\n", 501 inet_ntoa(SIN(dst)->sin_addr), la ? "la" : " ", 502 rt ? "rt" : ""); 503 m_freem(m); 504 return (0); 505 } 506 sdl = SDL(rt->rt_gateway); 507 /* 508 * Check the address family and length is valid, the address 509 * is resolved; otherwise, try to resolve. 510 */ 511 if ((rt->rt_expire == 0 || rt->rt_expire > time_second) && 512 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 513 /* 514 * If entry has an expiry time and it is approaching, 515 * see if we need to send an ARP request within this 516 * arpt_down interval. 517 */ 518 if ((rt->rt_expire != 0) && 519 (time_second + la->la_preempt > rt->rt_expire)) { 520 arprequest(ifp, 521 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 522 &SIN(dst)->sin_addr, 523 NULL); 524 la->la_preempt--; 525 } 526 527 bcopy(LLADDR(sdl), desten, sdl->sdl_alen); 528 return 1; 529 } 530 /* 531 * If ARP is disabled or static on this interface, stop. 532 * XXX 533 * Probably should not allocate empty llinfo struct if we are 534 * not going to be sending out an arp request. 535 */ 536 if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { 537 m_freem(m); 538 return (0); 539 } 540 /* 541 * There is an arptab entry, but no ethernet address 542 * response yet. Replace the held mbuf with this 543 * latest one. 544 */ 545 if (la->la_hold != NULL) 546 m_freem(la->la_hold); 547 la->la_hold = m; 548 la->la_msgport = cur_netport(); 549 if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) { 550 rt->rt_flags &= ~RTF_REJECT; 551 if (la->la_asked == 0 || rt->rt_expire != time_second) { 552 rt->rt_expire = time_second; 553 if (la->la_asked++ < arp_maxtries) { 554 arprequest(ifp, 555 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 556 &SIN(dst)->sin_addr, 557 NULL); 558 } else { 559 rt->rt_flags |= RTF_REJECT; 560 rt->rt_expire += arpt_down; 561 la->la_asked = 0; 562 la->la_preempt = arp_maxtries; 563 } 564 } 565 } 566 return (0); 567 } 568 569 /* 570 * Common length and type checks are done here, 571 * then the protocol-specific routine is called. 572 */ 573 static void 574 arpintr(netmsg_t msg) 575 { 576 struct mbuf *m = msg->packet.nm_packet; 577 struct arphdr *ar; 578 u_short ar_hrd; 579 580 if (m->m_len < sizeof(struct arphdr) && 581 (m = m_pullup(m, sizeof(struct arphdr))) == NULL) { 582 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 583 return; 584 } 585 ar = mtod(m, struct arphdr *); 586 587 ar_hrd = ntohs(ar->ar_hrd); 588 if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) { 589 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 590 (unsigned char *)&ar->ar_hrd, ""); 591 m_freem(m); 592 return; 593 } 594 595 if (m->m_pkthdr.len < arphdr_len(ar)) { 596 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 597 log(LOG_ERR, "arp: runt packet\n"); 598 return; 599 } 600 ar = mtod(m, struct arphdr *); 601 } 602 603 switch (ntohs(ar->ar_pro)) { 604 #ifdef INET 605 case ETHERTYPE_IP: 606 in_arpinput(m); 607 return; 608 #endif 609 } 610 m_freem(m); 611 /* msg was embedded in the mbuf, do not reply! */ 612 } 613 614 #ifdef INET 615 /* 616 * ARP for Internet protocols on 10 Mb/s Ethernet. 617 * Algorithm is that given in RFC 826. 618 * In addition, a sanity check is performed on the sender 619 * protocol address, to catch impersonators. 620 * We no longer handle negotiations for use of trailer protocol: 621 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 622 * along with IP replies if we wanted trailers sent to us, 623 * and also sent them in response to IP replies. 624 * This allowed either end to announce the desire to receive 625 * trailer packets. 626 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 627 * but formerly didn't normally send requests. 628 */ 629 630 static int log_arp_wrong_iface = 1; 631 static int log_arp_movements = 1; 632 static int log_arp_permanent_modify = 1; 633 634 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 635 &log_arp_wrong_iface, 0, 636 "Log arp packets arriving on the wrong interface"); 637 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_movements, CTLFLAG_RW, 638 &log_arp_movements, 0, 639 "Log arp replies from MACs different than the one in the cache"); 640 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_permanent_modify, CTLFLAG_RW, 641 &log_arp_permanent_modify, 0, 642 "Log arp replies from MACs different than the one " 643 "in the permanent arp entry"); 644 645 646 static void 647 arp_hold_output(netmsg_t msg) 648 { 649 struct mbuf *m = msg->packet.nm_packet; 650 struct rtentry *rt; 651 struct ifnet *ifp; 652 653 rt = msg->lmsg.u.ms_resultp; 654 ifp = m->m_pkthdr.rcvif; 655 m->m_pkthdr.rcvif = NULL; 656 657 ifp->if_output(ifp, m, rt_key(rt), rt); 658 659 /* Drop the reference count bumped by the sender */ 660 RTFREE(rt); 661 662 /* nmsg was embedded in the mbuf, do not reply! */ 663 } 664 665 static void 666 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create, 667 boolean_t generate_report, boolean_t dologging) 668 { 669 struct arphdr *ah = mtod(m, struct arphdr *); 670 struct ifnet *ifp = m->m_pkthdr.rcvif; 671 struct llinfo_arp *la; 672 struct sockaddr_dl *sdl; 673 struct rtentry *rt; 674 675 la = arplookup(saddr, create, generate_report, FALSE); 676 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) { 677 struct in_addr isaddr = { saddr }; 678 679 /* 680 * Normally arps coming in on the wrong interface are ignored, 681 * but if we are bridging and the two interfaces belong to 682 * the same bridge, or one is a member of the bridge which 683 * is the other, then it isn't an error. 684 */ 685 if (rt->rt_ifp != ifp) { 686 /* 687 * (1) ifp and rt_ifp both members of same bridge 688 * (2) rt_ifp member of bridge ifp 689 * (3) ifp member of bridge rt_ifp 690 * 691 * Always replace rt_ifp with the bridge ifc. 692 */ 693 struct ifnet *nifp; 694 695 if (ifp->if_bridge && 696 rt->rt_ifp->if_bridge == ifp->if_bridge) { 697 nifp = ether_bridge_interface(ifp); 698 } else if (rt->rt_ifp->if_bridge && 699 ether_bridge_interface(rt->rt_ifp) == ifp) { 700 nifp = ifp; 701 } else if (ifp->if_bridge && 702 ether_bridge_interface(ifp) == rt->rt_ifp) { 703 nifp = rt->rt_ifp; 704 } else { 705 nifp = NULL; 706 } 707 708 if ((log_arp_wrong_iface == 1 && nifp == NULL) || 709 log_arp_wrong_iface == 2) { 710 log(LOG_ERR, 711 "arp: %s is on %s " 712 "but got reply from %*D on %s\n", 713 inet_ntoa(isaddr), 714 rt->rt_ifp->if_xname, 715 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 716 ifp->if_xname); 717 } 718 if (nifp == NULL) 719 return; 720 721 /* 722 * nifp is our man! Replace rt_ifp and adjust 723 * the sdl. 724 */ 725 ifp = rt->rt_ifp = nifp; 726 sdl->sdl_type = ifp->if_type; 727 sdl->sdl_index = ifp->if_index; 728 } 729 if (sdl->sdl_alen && 730 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 731 if (rt->rt_expire != 0) { 732 if (dologging && log_arp_movements) { 733 log(LOG_INFO, 734 "arp: %s moved from %*D to %*D on %s\n", 735 inet_ntoa(isaddr), 736 ifp->if_addrlen, (u_char *)LLADDR(sdl), 737 ":", ifp->if_addrlen, 738 (u_char *)ar_sha(ah), ":", 739 ifp->if_xname); 740 } 741 } else { 742 if (dologging && log_arp_permanent_modify) { 743 log(LOG_ERR, 744 "arp: %*D attempts to modify " 745 "permanent entry for %s on %s\n", 746 ifp->if_addrlen, (u_char *)ar_sha(ah), 747 ":", inet_ntoa(isaddr), ifp->if_xname); 748 } 749 return; 750 } 751 } 752 /* 753 * sanity check for the address length. 754 * XXX this does not work for protocols with variable address 755 * length. -is 756 */ 757 if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) { 758 log(LOG_WARNING, 759 "arp from %*D: new addr len %d, was %d", 760 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 761 ah->ar_hln, sdl->sdl_alen); 762 } 763 if (ifp->if_addrlen != ah->ar_hln) { 764 if (dologging) { 765 log(LOG_WARNING, 766 "arp from %*D: addr len: new %d, i/f %d " 767 "(ignored)", 768 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 769 ah->ar_hln, ifp->if_addrlen); 770 } 771 return; 772 } 773 memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln); 774 if (rt->rt_expire != 0) { 775 rt->rt_expire = time_second + arpt_keep; 776 } 777 rt->rt_flags &= ~RTF_REJECT; 778 la->la_asked = 0; 779 la->la_preempt = arp_maxtries; 780 781 /* 782 * This particular cpu might have been holding an mbuf 783 * pending ARP resolution. If so, transmit the mbuf now. 784 */ 785 if (la->la_hold != NULL) { 786 struct mbuf *m = la->la_hold; 787 struct lwkt_port *port = la->la_msgport; 788 struct netmsg_packet *pmsg; 789 790 la->la_hold = NULL; 791 la->la_msgport = NULL; 792 793 m_adj(m, sizeof(struct ether_header)); 794 795 /* 796 * Make sure that this rtentry will not be freed 797 * before the packet is processed on the target 798 * msgport. The reference count will be dropped 799 * in the handler associated with this packet. 800 */ 801 rt->rt_refcnt++; 802 803 pmsg = &m->m_hdr.mh_netmsg; 804 netmsg_init(&pmsg->base, NULL, 805 &netisr_apanic_rport, 806 MSGF_PRIORITY, arp_hold_output); 807 pmsg->nm_packet = m; 808 809 /* Record necessary information */ 810 m->m_pkthdr.rcvif = ifp; 811 pmsg->base.lmsg.u.ms_resultp = rt; 812 813 lwkt_sendmsg(port, &pmsg->base.lmsg); 814 } 815 } 816 } 817 818 #ifdef SMP 819 820 struct netmsg_arp_update { 821 struct netmsg_base base; 822 struct mbuf *m; 823 in_addr_t saddr; 824 boolean_t create; 825 }; 826 827 static void arp_update_msghandler(netmsg_t msg); 828 829 #endif 830 831 /* 832 * Called from arpintr() - this routine is run from a single cpu. 833 */ 834 static void 835 in_arpinput(struct mbuf *m) 836 { 837 struct arphdr *ah; 838 struct ifnet *ifp = m->m_pkthdr.rcvif; 839 struct ether_header *eh; 840 struct rtentry *rt; 841 struct ifaddr_container *ifac; 842 struct in_ifaddr_container *iac; 843 struct in_ifaddr *ia = NULL; 844 struct sockaddr sa; 845 struct in_addr isaddr, itaddr, myaddr; 846 #ifdef SMP 847 struct netmsg_arp_update msg; 848 #endif 849 uint8_t *enaddr = NULL; 850 int op; 851 int req_len; 852 853 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 854 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 855 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 856 return; 857 } 858 859 ah = mtod(m, struct arphdr *); 860 op = ntohs(ah->ar_op); 861 memcpy(&isaddr, ar_spa(ah), sizeof isaddr); 862 memcpy(&itaddr, ar_tpa(ah), sizeof itaddr); 863 864 myaddr.s_addr = INADDR_ANY; 865 #ifdef CARP 866 if (ifp->if_carp != NULL) { 867 get_mplock(); 868 if (ifp->if_carp != NULL && 869 carp_iamatch(ifp->if_carp, &itaddr, &isaddr, &enaddr)) { 870 rel_mplock(); 871 myaddr = itaddr; 872 goto match; 873 } 874 rel_mplock(); 875 } 876 #endif 877 878 /* 879 * Check both target and sender IP addresses: 880 * 881 * If we receive the packet on the interface owning the address, 882 * then accept the address. 883 * 884 * For a bridge, we accept the address if the receive interface and 885 * the interface owning the address are on the same bridge, and 886 * use the bridge MAC as the is-at response. The bridge will be 887 * responsible for handling the packet. 888 * 889 * (1) Check target IP against our local IPs 890 */ 891 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) { 892 ia = iac->ia; 893 894 /* Skip all ia's which don't match */ 895 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 896 continue; 897 #ifdef CARP 898 if (ia->ia_ifp->if_type == IFT_CARP) 899 continue; 900 #endif 901 if (ifp->if_bridge && ia->ia_ifp && 902 ifp->if_bridge == ia->ia_ifp->if_bridge) { 903 ifp = ether_bridge_interface(ifp); 904 goto match; 905 } 906 if (ia->ia_ifp && ia->ia_ifp->if_bridge && 907 ether_bridge_interface(ia->ia_ifp) == ifp) { 908 goto match; 909 } 910 if (ifp->if_bridge && ether_bridge_interface(ifp) == 911 ia->ia_ifp) { 912 goto match; 913 } 914 if (ia->ia_ifp == ifp) 915 goto match; 916 917 } 918 919 /* 920 * (2) Check sender IP against our local IPs 921 */ 922 LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) { 923 ia = iac->ia; 924 925 /* Skip all ia's which don't match */ 926 if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 927 continue; 928 #ifdef CARP 929 if (ia->ia_ifp->if_type == IFT_CARP) 930 continue; 931 #endif 932 if (ifp->if_bridge && ia->ia_ifp && 933 ifp->if_bridge == ia->ia_ifp->if_bridge) { 934 ifp = ether_bridge_interface(ifp); 935 goto match; 936 } 937 if (ia->ia_ifp && ia->ia_ifp->if_bridge && 938 ether_bridge_interface(ia->ia_ifp) == ifp) { 939 goto match; 940 } 941 if (ifp->if_bridge && ether_bridge_interface(ifp) == 942 ia->ia_ifp) { 943 goto match; 944 } 945 946 if (ia->ia_ifp == ifp) 947 goto match; 948 } 949 950 /* 951 * No match, use the first inet address on the receive interface 952 * as a dummy address for the rest of the function. 953 */ 954 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 955 struct ifaddr *ifa = ifac->ifa; 956 957 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 958 ia = ifatoia(ifa); 959 goto match; 960 } 961 } 962 963 /* 964 * If we got here, we didn't find any suitable interface, 965 * so drop the packet. 966 */ 967 m_freem(m); 968 return; 969 970 match: 971 if (!enaddr) 972 enaddr = (uint8_t *)IF_LLADDR(ifp); 973 if (myaddr.s_addr == INADDR_ANY) 974 myaddr = ia->ia_addr.sin_addr; 975 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) { 976 m_freem(m); /* it's from me, ignore it. */ 977 return; 978 } 979 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 980 log(LOG_ERR, 981 "arp: link address is broadcast for IP address %s!\n", 982 inet_ntoa(isaddr)); 983 m_freem(m); 984 return; 985 } 986 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 987 log(LOG_ERR, 988 "arp: %*D is using my IP address %s!\n", 989 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 990 inet_ntoa(isaddr)); 991 itaddr = myaddr; 992 goto reply; 993 } 994 if (ifp->if_flags & IFF_STATICARP) 995 goto reply; 996 997 /* 998 * When arp_restricted_match is true and the ARP response is not 999 * specifically targetted to me, ignore it. Otherwise the entry 1000 * timeout may be updated for an old MAC. 1001 */ 1002 if (arp_restricted_match && itaddr.s_addr != myaddr.s_addr) { 1003 m_freem(m); 1004 return; 1005 } 1006 1007 #ifdef SMP 1008 netmsg_init(&msg.base, NULL, &curthread->td_msgport, 1009 0, arp_update_msghandler); 1010 msg.m = m; 1011 msg.saddr = isaddr.s_addr; 1012 msg.create = (itaddr.s_addr == myaddr.s_addr); 1013 lwkt_domsg(rtable_portfn(0), &msg.base.lmsg, 0); 1014 #else 1015 arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr), 1016 RTL_REPORTMSG, TRUE); 1017 #endif 1018 reply: 1019 if (op != ARPOP_REQUEST) { 1020 m_freem(m); 1021 return; 1022 } 1023 if (itaddr.s_addr == myaddr.s_addr) { 1024 /* I am the target */ 1025 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1026 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1027 } else { 1028 struct llinfo_arp *la; 1029 1030 la = arplookup(itaddr.s_addr, FALSE, RTL_DONTREPORT, SIN_PROXY); 1031 if (la == NULL) { 1032 struct sockaddr_in sin; 1033 1034 if (!arp_proxyall) { 1035 m_freem(m); 1036 return; 1037 } 1038 1039 bzero(&sin, sizeof sin); 1040 sin.sin_family = AF_INET; 1041 sin.sin_len = sizeof sin; 1042 sin.sin_addr = itaddr; 1043 1044 rt = rtpurelookup((struct sockaddr *)&sin); 1045 if (rt == NULL) { 1046 m_freem(m); 1047 return; 1048 } 1049 --rt->rt_refcnt; 1050 /* 1051 * Don't send proxies for nodes on the same interface 1052 * as this one came out of, or we'll get into a fight 1053 * over who claims what Ether address. 1054 */ 1055 if (rt->rt_ifp == ifp) { 1056 m_freem(m); 1057 return; 1058 } 1059 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1060 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 1061 #ifdef DEBUG_PROXY 1062 kprintf("arp: proxying for %s\n", inet_ntoa(itaddr)); 1063 #endif 1064 } else { 1065 struct sockaddr_dl *sdl; 1066 1067 rt = la->la_rt; 1068 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 1069 sdl = SDL(rt->rt_gateway); 1070 memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 1071 } 1072 } 1073 1074 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 1075 memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 1076 ah->ar_op = htons(ARPOP_REPLY); 1077 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 1078 switch (ifp->if_type) { 1079 case IFT_ETHER: 1080 /* 1081 * May not be correct for types not explictly 1082 * listed, but it is our best guess. 1083 */ 1084 default: 1085 eh = (struct ether_header *)sa.sa_data; 1086 memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost); 1087 eh->ether_type = htons(ETHERTYPE_ARP); 1088 break; 1089 } 1090 sa.sa_family = AF_UNSPEC; 1091 sa.sa_len = sizeof sa; 1092 ifp->if_output(ifp, m, &sa, NULL); 1093 } 1094 1095 #ifdef SMP 1096 1097 static void 1098 arp_update_msghandler(netmsg_t msg) 1099 { 1100 struct netmsg_arp_update *rmsg = (struct netmsg_arp_update *)msg; 1101 int nextcpu; 1102 1103 /* 1104 * This message handler will be called on all of the CPUs, 1105 * however, we only need to generate rtmsg on CPU0. 1106 */ 1107 arp_update_oncpu(rmsg->m, rmsg->saddr, rmsg->create, 1108 mycpuid == 0 ? RTL_REPORTMSG : RTL_DONTREPORT, 1109 mycpuid == 0); 1110 1111 nextcpu = mycpuid + 1; 1112 if (nextcpu < ncpus) 1113 lwkt_forwardmsg(rtable_portfn(nextcpu), &rmsg->base.lmsg); 1114 else 1115 lwkt_replymsg(&rmsg->base.lmsg, 0); 1116 } 1117 1118 #endif /* SMP */ 1119 1120 #endif /* INET */ 1121 1122 /* 1123 * Free an arp entry. If the arp entry is actively referenced or represents 1124 * a static entry we only clear it back to an unresolved state, otherwise 1125 * we destroy the entry entirely. 1126 * 1127 * Note that static entries are created when route add ... -interface is used 1128 * to create an interface route to a (direct) destination. 1129 */ 1130 static void 1131 arptfree(struct llinfo_arp *la) 1132 { 1133 struct rtentry *rt = la->la_rt; 1134 struct sockaddr_dl *sdl; 1135 1136 if (rt == NULL) 1137 panic("arptfree"); 1138 sdl = SDL(rt->rt_gateway); 1139 if (sdl != NULL && 1140 ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) || 1141 (rt->rt_flags & RTF_STATIC))) { 1142 sdl->sdl_alen = 0; 1143 la->la_preempt = la->la_asked = 0; 1144 rt->rt_flags &= ~RTF_REJECT; 1145 return; 1146 } 1147 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 1148 } 1149 1150 /* 1151 * Lookup or enter a new address in arptab. 1152 */ 1153 static struct llinfo_arp * 1154 arplookup(in_addr_t addr, boolean_t create, boolean_t generate_report, 1155 boolean_t proxy) 1156 { 1157 struct rtentry *rt; 1158 struct sockaddr_inarp sin = { sizeof sin, AF_INET }; 1159 const char *why = NULL; 1160 1161 sin.sin_addr.s_addr = addr; 1162 sin.sin_other = proxy ? SIN_PROXY : 0; 1163 if (create) { 1164 rt = _rtlookup((struct sockaddr *)&sin, 1165 generate_report, RTL_DOCLONE); 1166 } else { 1167 rt = rtpurelookup((struct sockaddr *)&sin); 1168 } 1169 if (rt == NULL) 1170 return (NULL); 1171 rt->rt_refcnt--; 1172 1173 if (rt->rt_flags & RTF_GATEWAY) 1174 why = "host is not on local network"; 1175 else if (!(rt->rt_flags & RTF_LLINFO)) 1176 why = "could not allocate llinfo"; 1177 else if (rt->rt_gateway->sa_family != AF_LINK) 1178 why = "gateway route is not ours"; 1179 1180 if (why) { 1181 if (create) { 1182 log(LOG_DEBUG, "arplookup %s failed: %s\n", 1183 inet_ntoa(sin.sin_addr), why); 1184 } 1185 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) { 1186 /* No references to this route. Purge it. */ 1187 rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1188 rt_mask(rt), rt->rt_flags, NULL); 1189 } 1190 return (NULL); 1191 } 1192 return (rt->rt_llinfo); 1193 } 1194 1195 void 1196 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1197 { 1198 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) { 1199 arprequest_async(ifp, &IA_SIN(ifa)->sin_addr, 1200 &IA_SIN(ifa)->sin_addr, NULL); 1201 } 1202 ifa->ifa_rtrequest = arp_rtrequest; 1203 ifa->ifa_flags |= RTF_CLONING; 1204 } 1205 1206 void 1207 arp_iainit(struct ifnet *ifp, const struct in_addr *addr, const u_char *enaddr) 1208 { 1209 if (addr->s_addr != INADDR_ANY) 1210 arprequest_async(ifp, addr, addr, enaddr); 1211 } 1212 1213 static void 1214 arp_init(void) 1215 { 1216 int cpu; 1217 1218 for (cpu = 0; cpu < ncpus2; cpu++) 1219 LIST_INIT(&llinfo_arp_list[cpu]); 1220 1221 netisr_register(NETISR_ARP, arpintr, NULL); 1222 } 1223 1224 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 1225