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