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 1.58 2008/11/09 10:18:42 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> 104 105 #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 struct lwkt_port *la_msgport; /* last packet's msgport */ 134 u_short la_preempt; /* countdown for pre-expiry arps */ 135 u_short la_asked; /* #times we QUERIED following expiration */ 136 }; 137 138 static LIST_HEAD(, llinfo_arp) llinfo_arp_list[MAXCPU]; 139 140 static int arp_maxtries = 5; 141 static int useloopback = 1; /* use loopback interface for local traffic */ 142 static int arp_proxyall = 0; 143 144 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, maxtries, CTLFLAG_RW, 145 &arp_maxtries, 0, ""); 146 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, useloopback, CTLFLAG_RW, 147 &useloopback, 0, ""); 148 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, proxyall, CTLFLAG_RW, 149 &arp_proxyall, 0, ""); 150 151 static int arp_mpsafe = 0; 152 TUNABLE_INT("net.link.ether.inet.arp_mpsafe", &arp_mpsafe); 153 154 static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 155 static void arprequest(struct ifnet *, struct in_addr *, struct in_addr *, 156 const u_char *); 157 static void arprequest_async(struct ifnet *, struct in_addr *, 158 struct in_addr *, const u_char *); 159 static void arpintr(struct netmsg *); 160 static void arptfree(struct llinfo_arp *); 161 static void arptimer(void *); 162 static struct llinfo_arp * 163 arplookup(in_addr_t, boolean_t, boolean_t); 164 #ifdef INET 165 static void in_arpinput(struct mbuf *); 166 #endif 167 168 static struct callout arptimer_ch[MAXCPU]; 169 170 /* 171 * Timeout routine. Age arp_tab entries periodically. 172 */ 173 /* ARGSUSED */ 174 static void 175 arptimer(void *ignored_arg) 176 { 177 struct llinfo_arp *la, *nla; 178 179 crit_enter(); 180 LIST_FOREACH_MUTABLE(la, &llinfo_arp_list[mycpuid], la_le, nla) { 181 if (la->la_rt->rt_expire && la->la_rt->rt_expire <= time_second) 182 arptfree(la); 183 } 184 callout_reset(&arptimer_ch[mycpuid], arpt_prune * hz, arptimer, NULL); 185 crit_exit(); 186 } 187 188 /* 189 * Parallel to llc_rtrequest. 190 */ 191 static void 192 arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 193 { 194 struct sockaddr *gate = rt->rt_gateway; 195 struct llinfo_arp *la = rt->rt_llinfo; 196 197 struct sockaddr_dl null_sdl = { sizeof null_sdl, AF_LINK }; 198 static boolean_t arpinit_done[MAXCPU]; 199 200 if (!arpinit_done[mycpuid]) { 201 arpinit_done[mycpuid] = TRUE; 202 callout_init(&arptimer_ch[mycpuid]); 203 callout_reset(&arptimer_ch[mycpuid], hz, arptimer, NULL); 204 } 205 if (rt->rt_flags & RTF_GATEWAY) 206 return; 207 208 switch (req) { 209 case RTM_ADD: 210 /* 211 * XXX: If this is a manually added route to interface 212 * such as older version of routed or gated might provide, 213 * restore cloning bit. 214 */ 215 if (!(rt->rt_flags & RTF_HOST) && 216 SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) 217 rt->rt_flags |= RTF_CLONING; 218 if (rt->rt_flags & RTF_CLONING) { 219 /* 220 * Case 1: This route should come from a route to iface. 221 */ 222 rt_setgate(rt, rt_key(rt), 223 (struct sockaddr *)&null_sdl); 224 gate = rt->rt_gateway; 225 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 226 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 227 rt->rt_expire = time_second; 228 break; 229 } 230 /* Announce a new entry if requested. */ 231 if (rt->rt_flags & RTF_ANNOUNCE) { 232 arprequest_async(rt->rt_ifp, 233 &SIN(rt_key(rt))->sin_addr, 234 &SIN(rt_key(rt))->sin_addr, 235 LLADDR(SDL(gate))); 236 } 237 /*FALLTHROUGH*/ 238 case RTM_RESOLVE: 239 if (gate->sa_family != AF_LINK || 240 gate->sa_len < sizeof(struct sockaddr_dl)) { 241 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n"); 242 break; 243 } 244 SDL(gate)->sdl_type = rt->rt_ifp->if_type; 245 SDL(gate)->sdl_index = rt->rt_ifp->if_index; 246 if (la != NULL) 247 break; /* This happens on a route change */ 248 /* 249 * Case 2: This route may come from cloning, or a manual route 250 * add with a LL address. 251 */ 252 R_Malloc(la, struct llinfo_arp *, sizeof *la); 253 rt->rt_llinfo = la; 254 if (la == NULL) { 255 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n"); 256 break; 257 } 258 bzero(la, sizeof *la); 259 la->la_rt = rt; 260 rt->rt_flags |= RTF_LLINFO; 261 LIST_INSERT_HEAD(&llinfo_arp_list[mycpuid], la, la_le); 262 263 #ifdef INET 264 /* 265 * This keeps the multicast addresses from showing up 266 * in `arp -a' listings as unresolved. It's not actually 267 * functional. Then the same for broadcast. 268 */ 269 if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr))) { 270 ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, 271 LLADDR(SDL(gate))); 272 SDL(gate)->sdl_alen = 6; 273 rt->rt_expire = 0; 274 } 275 if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { 276 memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, 277 rt->rt_ifp->if_addrlen); 278 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; 279 rt->rt_expire = 0; 280 } 281 #endif 282 283 if (SIN(rt_key(rt))->sin_addr.s_addr == 284 (IA_SIN(rt->rt_ifa))->sin_addr.s_addr) { 285 /* 286 * This test used to be 287 * if (loif.if_flags & IFF_UP) 288 * It allowed local traffic to be forced 289 * through the hardware by configuring the 290 * loopback down. However, it causes problems 291 * during network configuration for boards 292 * that can't receive packets they send. It 293 * is now necessary to clear "useloopback" and 294 * remove the route to force traffic out to 295 * the hardware. 296 */ 297 rt->rt_expire = 0; 298 bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), 299 SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); 300 if (useloopback) 301 rt->rt_ifp = loif; 302 } 303 break; 304 305 case RTM_DELETE: 306 if (la == NULL) 307 break; 308 LIST_REMOVE(la, la_le); 309 rt->rt_llinfo = NULL; 310 rt->rt_flags &= ~RTF_LLINFO; 311 if (la->la_hold != NULL) 312 m_freem(la->la_hold); 313 Free(la); 314 break; 315 } 316 } 317 318 static struct mbuf * 319 arpreq_alloc(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 320 const u_char *enaddr) 321 { 322 struct mbuf *m; 323 struct arphdr *ah; 324 u_short ar_hrd; 325 326 if ((m = m_gethdr(MB_DONTWAIT, MT_DATA)) == NULL) 327 return NULL; 328 m->m_pkthdr.rcvif = NULL; 329 330 switch (ifp->if_type) { 331 case IFT_ETHER: 332 /* 333 * This may not be correct for types not explicitly 334 * listed, but this is our best guess 335 */ 336 default: 337 ar_hrd = htons(ARPHRD_ETHER); 338 339 m->m_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 340 m->m_pkthdr.len = m->m_len; 341 MH_ALIGN(m, m->m_len); 342 343 ah = mtod(m, struct arphdr *); 344 break; 345 } 346 347 ah->ar_hrd = ar_hrd; 348 ah->ar_pro = htons(ETHERTYPE_IP); 349 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 350 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 351 ah->ar_op = htons(ARPOP_REQUEST); 352 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 353 memset(ar_tha(ah), 0, ah->ar_hln); 354 memcpy(ar_spa(ah), sip, ah->ar_pln); 355 memcpy(ar_tpa(ah), tip, ah->ar_pln); 356 357 return m; 358 } 359 360 static void 361 arpreq_send(struct ifnet *ifp, struct mbuf *m) 362 { 363 struct sockaddr sa; 364 struct ether_header *eh; 365 366 switch (ifp->if_type) { 367 case IFT_ETHER: 368 /* 369 * This may not be correct for types not explicitly 370 * listed, but this is our best guess 371 */ 372 default: 373 eh = (struct ether_header *)sa.sa_data; 374 /* if_output() will not swap */ 375 eh->ether_type = htons(ETHERTYPE_ARP); 376 memcpy(eh->ether_dhost, ifp->if_broadcastaddr, ifp->if_addrlen); 377 break; 378 } 379 380 sa.sa_family = AF_UNSPEC; 381 sa.sa_len = sizeof(sa); 382 ifp->if_output(ifp, m, &sa, NULL); 383 } 384 385 static void 386 arpreq_send_handler(struct netmsg *nmsg) 387 { 388 struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet; 389 struct ifnet *ifp = nmsg->nm_lmsg.u.ms_resultp; 390 391 arpreq_send(ifp, m); 392 /* nmsg was embedded in the mbuf, do not reply! */ 393 } 394 395 /* 396 * Broadcast an ARP request. Caller specifies: 397 * - arp header source ip address 398 * - arp header target ip address 399 * - arp header source ethernet address 400 * 401 * NOTE: Caller MUST NOT hold ifp's serializer 402 */ 403 static void 404 arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 405 const u_char *enaddr) 406 { 407 struct mbuf *m; 408 409 m = arpreq_alloc(ifp, sip, tip, enaddr); 410 if (m == NULL) 411 return; 412 arpreq_send(ifp, m); 413 } 414 415 /* 416 * Same as arprequest(), except: 417 * - Caller is allowed to hold ifp's serializer 418 * - Network output is done in TDF_NETWORK kernel thread 419 */ 420 static void 421 arprequest_async(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, 422 const u_char *enaddr) 423 { 424 struct mbuf *m; 425 struct netmsg_packet *pmsg; 426 427 m = arpreq_alloc(ifp, sip, tip, enaddr); 428 if (m == NULL) 429 return; 430 431 pmsg = &m->m_hdr.mh_netmsg; 432 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0, 433 arpreq_send_handler); 434 pmsg->nm_packet = m; 435 pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = ifp; 436 437 lwkt_sendmsg(cpu_portfn(mycpuid), &pmsg->nm_netmsg.nm_lmsg); 438 } 439 440 /* 441 * Resolve an IP address into an ethernet address. If success, 442 * desten is filled in. If there is no entry in arptab, 443 * set one up and broadcast a request for the IP address. 444 * Hold onto this mbuf and resend it once the address 445 * is finally resolved. A return value of 1 indicates 446 * that desten has been filled in and the packet should be sent 447 * normally; a 0 return indicates that the packet has been 448 * taken over here, either now or for later transmission. 449 */ 450 int 451 arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, 452 struct sockaddr *dst, u_char *desten) 453 { 454 struct rtentry *rt; 455 struct llinfo_arp *la = NULL; 456 struct sockaddr_dl *sdl; 457 458 if (m->m_flags & M_BCAST) { /* broadcast */ 459 memcpy(desten, ifp->if_broadcastaddr, ifp->if_addrlen); 460 return (1); 461 } 462 if (m->m_flags & M_MCAST) {/* multicast */ 463 ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); 464 return (1); 465 } 466 if (rt0 != NULL) { 467 if (rt_llroute(dst, rt0, &rt) != 0) { 468 m_freem(m); 469 return 0; 470 } 471 la = rt->rt_llinfo; 472 } 473 if (la == NULL) { 474 la = arplookup(SIN(dst)->sin_addr.s_addr, TRUE, 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 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) { 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 if (curthread->td_flags & TDF_NETWORK) 528 la->la_msgport = &curthread->td_msgport; 529 else 530 la->la_msgport = cpu_portfn(mycpuid); 531 if (rt->rt_expire || ((rt->rt_flags & RTF_STATIC) && !sdl->sdl_alen)) { 532 rt->rt_flags &= ~RTF_REJECT; 533 if (la->la_asked == 0 || rt->rt_expire != time_second) { 534 rt->rt_expire = time_second; 535 if (la->la_asked++ < arp_maxtries) { 536 arprequest(ifp, 537 &SIN(rt->rt_ifa->ifa_addr)->sin_addr, 538 &SIN(dst)->sin_addr, 539 IF_LLADDR(ifp)); 540 } else { 541 rt->rt_flags |= RTF_REJECT; 542 rt->rt_expire += arpt_down; 543 la->la_asked = 0; 544 la->la_preempt = arp_maxtries; 545 } 546 } 547 } 548 return (0); 549 } 550 551 /* 552 * Common length and type checks are done here, 553 * then the protocol-specific routine is called. 554 */ 555 static void 556 arpintr(struct netmsg *msg) 557 { 558 struct mbuf *m = ((struct netmsg_packet *)msg)->nm_packet; 559 struct arphdr *ar; 560 u_short ar_hrd; 561 562 if (m->m_len < sizeof(struct arphdr) && 563 (m = m_pullup(m, sizeof(struct arphdr))) == NULL) { 564 log(LOG_ERR, "arp: runt packet -- m_pullup failed\n"); 565 return; 566 } 567 ar = mtod(m, struct arphdr *); 568 569 ar_hrd = ntohs(ar->ar_hrd); 570 if (ar_hrd != ARPHRD_ETHER && ar_hrd != ARPHRD_IEEE802) { 571 log(LOG_ERR, "arp: unknown hardware address format (0x%2D)\n", 572 (unsigned char *)&ar->ar_hrd, ""); 573 m_freem(m); 574 return; 575 } 576 577 if (m->m_pkthdr.len < arphdr_len(ar)) { 578 if ((m = m_pullup(m, arphdr_len(ar))) == NULL) { 579 log(LOG_ERR, "arp: runt packet\n"); 580 return; 581 } 582 ar = mtod(m, struct arphdr *); 583 } 584 585 switch (ntohs(ar->ar_pro)) { 586 #ifdef INET 587 case ETHERTYPE_IP: 588 in_arpinput(m); 589 return; 590 #endif 591 } 592 m_freem(m); 593 /* msg was embedded in the mbuf, do not reply! */ 594 } 595 596 #ifdef INET 597 /* 598 * ARP for Internet protocols on 10 Mb/s Ethernet. 599 * Algorithm is that given in RFC 826. 600 * In addition, a sanity check is performed on the sender 601 * protocol address, to catch impersonators. 602 * We no longer handle negotiations for use of trailer protocol: 603 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 604 * along with IP replies if we wanted trailers sent to us, 605 * and also sent them in response to IP replies. 606 * This allowed either end to announce the desire to receive 607 * trailer packets. 608 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 609 * but formerly didn't normally send requests. 610 */ 611 static int log_arp_wrong_iface = 1; 612 SYSCTL_INT(_net_link_ether_inet, OID_AUTO, log_arp_wrong_iface, CTLFLAG_RW, 613 &log_arp_wrong_iface, 0, 614 "log arp packets arriving on the wrong interface"); 615 616 static void 617 arp_hold_output(struct netmsg *nmsg) 618 { 619 struct mbuf *m = ((struct netmsg_packet *)nmsg)->nm_packet; 620 struct rtentry *rt; 621 struct ifnet *ifp; 622 623 rt = nmsg->nm_lmsg.u.ms_resultp; 624 ifp = m->m_pkthdr.rcvif; 625 m->m_pkthdr.rcvif = NULL; 626 627 ifp->if_output(ifp, m, rt_key(rt), rt); 628 629 /* Drop the reference count bumped by the sender */ 630 RTFREE(rt); 631 632 /* nmsg was embedded in the mbuf, do not reply! */ 633 } 634 635 static void 636 arp_update_oncpu(struct mbuf *m, in_addr_t saddr, boolean_t create, 637 boolean_t dologging) 638 { 639 struct arphdr *ah = mtod(m, struct arphdr *); 640 struct ifnet *ifp = m->m_pkthdr.rcvif; 641 struct llinfo_arp *la; 642 struct sockaddr_dl *sdl; 643 struct rtentry *rt; 644 645 la = arplookup(saddr, create, FALSE); 646 if (la && (rt = la->la_rt) && (sdl = SDL(rt->rt_gateway))) { 647 struct in_addr isaddr = { saddr }; 648 649 /* the following is not an error when doing bridging */ 650 if (rt->rt_ifp != ifp) { 651 if (dologging && log_arp_wrong_iface) { 652 log(LOG_ERR, 653 "arp: %s is on %s " 654 "but got reply from %*D on %s\n", 655 inet_ntoa(isaddr), 656 rt->rt_ifp->if_xname, 657 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 658 ifp->if_xname); 659 } 660 return; 661 } 662 if (sdl->sdl_alen && 663 bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { 664 if (rt->rt_expire != 0) { 665 if (dologging) { 666 log(LOG_INFO, 667 "arp: %s moved from %*D to %*D on %s\n", 668 inet_ntoa(isaddr), 669 ifp->if_addrlen, (u_char *)LLADDR(sdl), 670 ":", ifp->if_addrlen, 671 (u_char *)ar_sha(ah), ":", 672 ifp->if_xname); 673 } 674 } else { 675 if (dologging) { 676 log(LOG_ERR, 677 "arp: %*D attempts to modify " 678 "permanent entry for %s on %s\n", 679 ifp->if_addrlen, (u_char *)ar_sha(ah), 680 ":", inet_ntoa(isaddr), ifp->if_xname); 681 } 682 return; 683 } 684 } 685 /* 686 * sanity check for the address length. 687 * XXX this does not work for protocols with variable address 688 * length. -is 689 */ 690 if (dologging && sdl->sdl_alen && sdl->sdl_alen != ah->ar_hln) { 691 log(LOG_WARNING, 692 "arp from %*D: new addr len %d, was %d", 693 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 694 ah->ar_hln, sdl->sdl_alen); 695 } 696 if (ifp->if_addrlen != ah->ar_hln) { 697 if (dologging) { 698 log(LOG_WARNING, 699 "arp from %*D: addr len: new %d, i/f %d " 700 "(ignored)", 701 ifp->if_addrlen, (u_char *) ar_sha(ah), ":", 702 ah->ar_hln, ifp->if_addrlen); 703 } 704 return; 705 } 706 memcpy(LLADDR(sdl), ar_sha(ah), sdl->sdl_alen = ah->ar_hln); 707 if (rt->rt_expire != 0) 708 rt->rt_expire = time_second + arpt_keep; 709 rt->rt_flags &= ~RTF_REJECT; 710 la->la_asked = 0; 711 la->la_preempt = arp_maxtries; 712 713 /* 714 * This particular cpu might have been holding an mbuf 715 * pending ARP resolution. If so, transmit the mbuf now. 716 */ 717 if (la->la_hold != NULL) { 718 struct mbuf *m = la->la_hold; 719 struct lwkt_port *port = la->la_msgport; 720 struct netmsg_packet *pmsg; 721 722 la->la_hold = NULL; 723 la->la_msgport = NULL; 724 725 m_adj(m, sizeof(struct ether_header)); 726 727 /* 728 * Make sure that this rtentry will not be freed 729 * before the packet is processed on the target 730 * msgport. The reference count will be dropped 731 * in the handler associated with this packet. 732 */ 733 rt->rt_refcnt++; 734 735 pmsg = &m->m_hdr.mh_netmsg; 736 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 737 MSGF_PRIORITY | MSGF_MPSAFE, 738 arp_hold_output); 739 pmsg->nm_packet = m; 740 741 /* Record necessary information */ 742 m->m_pkthdr.rcvif = ifp; 743 pmsg->nm_netmsg.nm_lmsg.u.ms_resultp = rt; 744 745 lwkt_sendmsg(port, &pmsg->nm_netmsg.nm_lmsg); 746 } 747 } 748 } 749 750 #ifdef SMP 751 752 struct netmsg_arp_update { 753 struct netmsg netmsg; 754 struct mbuf *m; 755 in_addr_t saddr; 756 boolean_t create; 757 }; 758 759 static void arp_update_msghandler(struct netmsg *); 760 761 #endif 762 763 /* 764 * Called from arpintr() - this routine is run from a single cpu. 765 */ 766 static void 767 in_arpinput(struct mbuf *m) 768 { 769 struct arphdr *ah; 770 struct ifnet *ifp = m->m_pkthdr.rcvif; 771 struct ether_header *eh; 772 struct rtentry *rt; 773 struct ifaddr_container *ifac; 774 struct in_ifaddr_container *iac; 775 struct in_ifaddr *ia; 776 struct sockaddr sa; 777 struct in_addr isaddr, itaddr, myaddr; 778 #ifdef SMP 779 struct netmsg_arp_update msg; 780 #endif 781 u_int8_t *enaddr = NULL; 782 int op; 783 int req_len; 784 785 req_len = arphdr_len2(ifp->if_addrlen, sizeof(struct in_addr)); 786 if (m->m_len < req_len && (m = m_pullup(m, req_len)) == NULL) { 787 log(LOG_ERR, "in_arp: runt packet -- m_pullup failed\n"); 788 return; 789 } 790 791 ah = mtod(m, struct arphdr *); 792 op = ntohs(ah->ar_op); 793 memcpy(&isaddr, ar_spa(ah), sizeof isaddr); 794 memcpy(&itaddr, ar_tpa(ah), sizeof itaddr); 795 /* 796 * Check both target and sender IP addresses: 797 * 798 * If we receive the packet on the interface owning the address, 799 * then accept the address. 800 * 801 * For a bridge, we accept the address if the receive interface and 802 * the interface owning the address are on the same bridge. 803 * (This will change slightly when we have clusters of interfaces). 804 */ 805 LIST_FOREACH(iac, INADDR_HASH(itaddr.s_addr), ia_hash) { 806 ia = iac->ia; 807 808 /* Skip all ia's which don't match */ 809 if (itaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 810 continue; 811 812 if (ia->ia_ifp == ifp) 813 goto match; 814 815 if (ifp->if_bridge && ia->ia_ifp && 816 ifp->if_bridge == ia->ia_ifp->if_bridge) 817 goto match; 818 819 #ifdef CARP 820 /* 821 * If the interface does not match, but the recieving interface 822 * is part of carp, we call carp_iamatch to see if this is a 823 * request for the virtual host ip. 824 * XXX: This is really ugly! 825 */ 826 if (ifp->if_carp != NULL) { 827 get_mplock(); 828 if (ifp->if_carp != NULL && 829 carp_iamatch(ifp->if_carp, ia, &isaddr, &enaddr)) { 830 rel_mplock(); 831 goto match; 832 } 833 rel_mplock(); 834 } 835 #endif 836 } 837 LIST_FOREACH(iac, INADDR_HASH(isaddr.s_addr), ia_hash) { 838 ia = iac->ia; 839 840 /* Skip all ia's which don't match */ 841 if (isaddr.s_addr != ia->ia_addr.sin_addr.s_addr) 842 continue; 843 844 if (ia->ia_ifp == ifp) 845 goto match; 846 847 if (ifp->if_bridge && ia->ia_ifp && 848 ifp->if_bridge == ia->ia_ifp->if_bridge) 849 goto match; 850 } 851 /* 852 * No match, use the first inet address on the receive interface 853 * as a dummy address for the rest of the function. 854 */ 855 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) { 856 struct ifaddr *ifa = ifac->ifa; 857 858 if (ifa->ifa_addr && ifa->ifa_addr->sa_family == AF_INET) { 859 ia = ifatoia(ifa); 860 goto match; 861 } 862 } 863 /* 864 * If we got here, we didn't find any suitable interface, 865 * so drop the packet. 866 */ 867 m_freem(m); 868 return; 869 870 match: 871 if (!enaddr) 872 enaddr = (u_int8_t *)IF_LLADDR(ifp); 873 myaddr = ia->ia_addr.sin_addr; 874 if (!bcmp(ar_sha(ah), enaddr, ifp->if_addrlen)) { 875 m_freem(m); /* it's from me, ignore it. */ 876 return; 877 } 878 if (!bcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 879 log(LOG_ERR, 880 "arp: link address is broadcast for IP address %s!\n", 881 inet_ntoa(isaddr)); 882 m_freem(m); 883 return; 884 } 885 if (isaddr.s_addr == myaddr.s_addr && myaddr.s_addr != 0) { 886 log(LOG_ERR, 887 "arp: %*D is using my IP address %s!\n", 888 ifp->if_addrlen, (u_char *)ar_sha(ah), ":", 889 inet_ntoa(isaddr)); 890 itaddr = myaddr; 891 goto reply; 892 } 893 #ifdef SMP 894 netmsg_init(&msg.netmsg, &curthread->td_msgport, 0, 895 arp_update_msghandler); 896 msg.m = m; 897 msg.saddr = isaddr.s_addr; 898 msg.create = (itaddr.s_addr == myaddr.s_addr); 899 lwkt_domsg(rtable_portfn(0), &msg.netmsg.nm_lmsg, 0); 900 #else 901 arp_update_oncpu(m, isaddr.s_addr, (itaddr.s_addr == myaddr.s_addr), 902 TRUE); 903 #endif 904 reply: 905 if (op != ARPOP_REQUEST) { 906 m_freem(m); 907 return; 908 } 909 if (itaddr.s_addr == myaddr.s_addr) { 910 /* I am the target */ 911 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 912 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 913 } else { 914 struct llinfo_arp *la; 915 916 la = arplookup(itaddr.s_addr, FALSE, SIN_PROXY); 917 if (la == NULL) { 918 struct sockaddr_in sin; 919 920 if (!arp_proxyall) { 921 m_freem(m); 922 return; 923 } 924 925 bzero(&sin, sizeof sin); 926 sin.sin_family = AF_INET; 927 sin.sin_len = sizeof sin; 928 sin.sin_addr = itaddr; 929 930 rt = rtpurelookup((struct sockaddr *)&sin); 931 if (rt == NULL) { 932 m_freem(m); 933 return; 934 } 935 --rt->rt_refcnt; 936 /* 937 * Don't send proxies for nodes on the same interface 938 * as this one came out of, or we'll get into a fight 939 * over who claims what Ether address. 940 */ 941 if (rt->rt_ifp == ifp) { 942 m_freem(m); 943 return; 944 } 945 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 946 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 947 #ifdef DEBUG_PROXY 948 kprintf("arp: proxying for %s\n", inet_ntoa(itaddr)); 949 #endif 950 } else { 951 struct sockaddr_dl *sdl; 952 953 rt = la->la_rt; 954 memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); 955 sdl = SDL(rt->rt_gateway); 956 memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); 957 } 958 } 959 960 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 961 memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 962 ah->ar_op = htons(ARPOP_REPLY); 963 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 964 switch (ifp->if_type) { 965 case IFT_ETHER: 966 /* 967 * May not be correct for types not explictly 968 * listed, but it is our best guess. 969 */ 970 default: 971 eh = (struct ether_header *)sa.sa_data; 972 memcpy(eh->ether_dhost, ar_tha(ah), sizeof eh->ether_dhost); 973 eh->ether_type = htons(ETHERTYPE_ARP); 974 break; 975 } 976 sa.sa_family = AF_UNSPEC; 977 sa.sa_len = sizeof sa; 978 ifp->if_output(ifp, m, &sa, NULL); 979 } 980 981 #ifdef SMP 982 983 static void 984 arp_update_msghandler(struct netmsg *netmsg) 985 { 986 struct netmsg_arp_update *msg = (struct netmsg_arp_update *)netmsg; 987 int nextcpu; 988 989 arp_update_oncpu(msg->m, msg->saddr, msg->create, mycpuid == 0); 990 991 nextcpu = mycpuid + 1; 992 if (nextcpu < ncpus) 993 lwkt_forwardmsg(rtable_portfn(nextcpu), &msg->netmsg.nm_lmsg); 994 else 995 lwkt_replymsg(&msg->netmsg.nm_lmsg, 0); 996 } 997 998 #endif /* SMP */ 999 1000 #endif /* INET */ 1001 1002 /* 1003 * Free an arp entry. If the arp entry is actively referenced or represents 1004 * a static entry we only clear it back to an unresolved state, otherwise 1005 * we destroy the entry entirely. 1006 * 1007 * Note that static entries are created when route add ... -interface is used 1008 * to create an interface route to a (direct) destination. 1009 */ 1010 static void 1011 arptfree(struct llinfo_arp *la) 1012 { 1013 struct rtentry *rt = la->la_rt; 1014 struct sockaddr_dl *sdl; 1015 1016 if (rt == NULL) 1017 panic("arptfree"); 1018 sdl = SDL(rt->rt_gateway); 1019 if (sdl != NULL && 1020 ((rt->rt_refcnt > 0 && sdl->sdl_family == AF_LINK) || 1021 (rt->rt_flags & RTF_STATIC))) { 1022 sdl->sdl_alen = 0; 1023 la->la_preempt = la->la_asked = 0; 1024 rt->rt_flags &= ~RTF_REJECT; 1025 return; 1026 } 1027 rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL); 1028 } 1029 1030 /* 1031 * Lookup or enter a new address in arptab. 1032 */ 1033 static struct llinfo_arp * 1034 arplookup(in_addr_t addr, boolean_t create, boolean_t proxy) 1035 { 1036 struct rtentry *rt; 1037 struct sockaddr_inarp sin = { sizeof sin, AF_INET }; 1038 const char *why = NULL; 1039 1040 sin.sin_addr.s_addr = addr; 1041 sin.sin_other = proxy ? SIN_PROXY : 0; 1042 if (create) 1043 rt = rtlookup((struct sockaddr *)&sin); 1044 else 1045 rt = rtpurelookup((struct sockaddr *)&sin); 1046 if (rt == NULL) 1047 return (NULL); 1048 rt->rt_refcnt--; 1049 1050 if (rt->rt_flags & RTF_GATEWAY) 1051 why = "host is not on local network"; 1052 else if (!(rt->rt_flags & RTF_LLINFO)) 1053 why = "could not allocate llinfo"; 1054 else if (rt->rt_gateway->sa_family != AF_LINK) 1055 why = "gateway route is not ours"; 1056 1057 if (why) { 1058 if (create) { 1059 log(LOG_DEBUG, "arplookup %s failed: %s\n", 1060 inet_ntoa(sin.sin_addr), why); 1061 } 1062 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_WASCLONED)) { 1063 /* No references to this route. Purge it. */ 1064 rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway, 1065 rt_mask(rt), rt->rt_flags, NULL); 1066 } 1067 return (NULL); 1068 } 1069 return (rt->rt_llinfo); 1070 } 1071 1072 void 1073 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1074 { 1075 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) { 1076 arprequest_async(ifp, &IA_SIN(ifa)->sin_addr, 1077 &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); 1078 } 1079 ifa->ifa_rtrequest = arp_rtrequest; 1080 ifa->ifa_flags |= RTF_CLONING; 1081 } 1082 1083 void 1084 arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) 1085 { 1086 if (IA_SIN(ifa)->sin_addr.s_addr != INADDR_ANY) { 1087 arprequest_async(ifp, &IA_SIN(ifa)->sin_addr, 1088 &IA_SIN(ifa)->sin_addr, enaddr); 1089 } 1090 ifa->ifa_rtrequest = arp_rtrequest; 1091 ifa->ifa_flags |= RTF_CLONING; 1092 } 1093 1094 static void 1095 arp_init(void) 1096 { 1097 uint32_t flags; 1098 int cpu; 1099 1100 for (cpu = 0; cpu < ncpus2; cpu++) 1101 LIST_INIT(&llinfo_arp_list[cpu]); 1102 1103 if (arp_mpsafe) { 1104 flags = NETISR_FLAG_MPSAFE; 1105 kprintf("arp: MPSAFE\n"); 1106 } else { 1107 flags = NETISR_FLAG_NOTMPSAFE; 1108 } 1109 netisr_register(NETISR_ARP, cpu0_portfn, arpintr, flags); 1110 } 1111 1112 SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); 1113