1 /* $NetBSD: if_arp.c,v 1.145 2009/01/11 02:45:54 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1998, 2000, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Public Access Networks Corporation ("Panix"). It was developed under 9 * contract to Panix by Eric Haszlakiewicz and Thor Lancelot Simon. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF 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. Neither the name of the University nor the names of its contributors 46 * may be used to endorse or promote products derived from this software 47 * without specific prior written permission. 48 * 49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 59 * SUCH DAMAGE. 60 * 61 * @(#)if_ether.c 8.2 (Berkeley) 9/26/94 62 */ 63 64 /* 65 * Ethernet address resolution protocol. 66 * TODO: 67 * add "inuse/lock" bit (or ref. count) along with valid bit 68 */ 69 70 #include <sys/cdefs.h> 71 __KERNEL_RCSID(0, "$NetBSD: if_arp.c,v 1.145 2009/01/11 02:45:54 christos Exp $"); 72 73 #include "opt_ddb.h" 74 #include "opt_inet.h" 75 76 #ifdef INET 77 78 #include "bridge.h" 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/callout.h> 83 #include <sys/malloc.h> 84 #include <sys/mbuf.h> 85 #include <sys/socket.h> 86 #include <sys/time.h> 87 #include <sys/timetc.h> 88 #include <sys/kernel.h> 89 #include <sys/errno.h> 90 #include <sys/ioctl.h> 91 #include <sys/syslog.h> 92 #include <sys/proc.h> 93 #include <sys/protosw.h> 94 #include <sys/domain.h> 95 #include <sys/sysctl.h> 96 #include <sys/socketvar.h> 97 #include <sys/percpu.h> 98 99 #include <net/ethertypes.h> 100 #include <net/if.h> 101 #include <net/if_dl.h> 102 #include <net/if_token.h> 103 #include <net/if_types.h> 104 #include <net/if_ether.h> 105 #include <net/route.h> 106 #include <net/net_stats.h> 107 108 #include <netinet/in.h> 109 #include <netinet/in_systm.h> 110 #include <netinet/in_var.h> 111 #include <netinet/ip.h> 112 #include <netinet/if_inarp.h> 113 114 #include "arcnet.h" 115 #if NARCNET > 0 116 #include <net/if_arc.h> 117 #endif 118 #include "fddi.h" 119 #if NFDDI > 0 120 #include <net/if_fddi.h> 121 #endif 122 #include "token.h" 123 #include "carp.h" 124 #if NCARP > 0 125 #include <netinet/ip_carp.h> 126 #endif 127 128 #define SIN(s) ((struct sockaddr_in *)s) 129 #define SRP(s) ((struct sockaddr_inarp *)s) 130 131 /* 132 * ARP trailer negotiation. Trailer protocol is not IP specific, 133 * but ARP request/response use IP addresses. 134 */ 135 #define ETHERTYPE_IPTRAILERS ETHERTYPE_TRAIL 136 137 /* timer values */ 138 int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 139 int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 140 int arpt_down = 20; /* once declared down, don't send for 20 secs */ 141 int arpt_refresh = (5*60); /* time left before refreshing */ 142 #define rt_expire rt_rmx.rmx_expire 143 #define rt_pksent rt_rmx.rmx_pksent 144 145 static struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *, 146 const struct sockaddr *); 147 static void arptfree(struct llinfo_arp *); 148 static void arptimer(void *); 149 static struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *, 150 int, int, struct rtentry *); 151 static struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *, 152 int, int); 153 static void in_arpinput(struct mbuf *); 154 155 LIST_HEAD(, llinfo_arp) llinfo_arp; 156 struct ifqueue arpintrq = { 157 .ifq_head = NULL, 158 .ifq_tail = NULL, 159 .ifq_len = 0, 160 .ifq_maxlen = 50, 161 .ifq_drops = 0, 162 }; 163 int arp_inuse, arp_allocated, arp_intimer; 164 int arp_maxtries = 5; 165 int useloopback = 1; /* use loopback interface for local traffic */ 166 int arpinit_done = 0; 167 168 static percpu_t *arpstat_percpu; 169 170 #define ARP_STAT_GETREF() _NET_STAT_GETREF(arpstat_percpu) 171 #define ARP_STAT_PUTREF() _NET_STAT_PUTREF(arpstat_percpu) 172 173 #define ARP_STATINC(x) _NET_STATINC(arpstat_percpu, x) 174 #define ARP_STATADD(x, v) _NET_STATADD(arpstat_percpu, x, v) 175 176 struct callout arptimer_ch; 177 178 /* revarp state */ 179 struct in_addr myip, srv_ip; 180 int myip_initialized = 0; 181 int revarp_in_progress = 0; 182 struct ifnet *myip_ifp = NULL; 183 184 #ifdef DDB 185 static void db_print_sa(const struct sockaddr *); 186 static void db_print_ifa(struct ifaddr *); 187 static void db_print_llinfo(void *); 188 static int db_show_rtentry(struct rtentry *, void *); 189 #endif 190 191 /* 192 * this should be elsewhere. 193 */ 194 195 static char * 196 lla_snprintf(u_int8_t *, int); 197 198 static char * 199 lla_snprintf(u_int8_t *adrp, int len) 200 { 201 #define NUMBUFS 3 202 static char buf[NUMBUFS][16*3]; 203 static int bnum = 0; 204 205 int i; 206 char *p; 207 208 p = buf[bnum]; 209 210 *p++ = hexdigits[(*adrp)>>4]; 211 *p++ = hexdigits[(*adrp++)&0xf]; 212 213 for (i=1; i<len && i<16; i++) { 214 *p++ = ':'; 215 *p++ = hexdigits[(*adrp)>>4]; 216 *p++ = hexdigits[(*adrp++)&0xf]; 217 } 218 219 *p = 0; 220 p = buf[bnum]; 221 bnum = (bnum + 1) % NUMBUFS; 222 return p; 223 } 224 225 DOMAIN_DEFINE(arpdomain); /* forward declare and add to link set */ 226 227 const struct protosw arpsw[] = { 228 { .pr_type = 0, 229 .pr_domain = &arpdomain, 230 .pr_protocol = 0, 231 .pr_flags = 0, 232 .pr_input = 0, 233 .pr_output = 0, 234 .pr_ctlinput = 0, 235 .pr_ctloutput = 0, 236 .pr_usrreq = 0, 237 .pr_init = arp_init, 238 .pr_fasttimo = 0, 239 .pr_slowtimo = 0, 240 .pr_drain = arp_drain, 241 } 242 }; 243 244 245 struct domain arpdomain = { 246 .dom_family = PF_ARP, 247 .dom_name = "arp", 248 .dom_protosw = arpsw, 249 .dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)], 250 }; 251 252 /* 253 * ARP table locking. 254 * 255 * to prevent lossage vs. the arp_drain routine (which may be called at 256 * any time, including in a device driver context), we do two things: 257 * 258 * 1) manipulation of la->la_hold is done at splnet() (for all of 259 * about two instructions). 260 * 261 * 2) manipulation of the arp table's linked list is done under the 262 * protection of the ARP_LOCK; if arp_drain() or arptimer is called 263 * while the arp table is locked, we punt and try again later. 264 */ 265 266 static int arp_locked; 267 static inline int arp_lock_try(int); 268 static inline void arp_unlock(void); 269 270 static inline int 271 arp_lock_try(int recurse) 272 { 273 int s; 274 275 /* 276 * Use splvm() -- we're blocking things that would cause 277 * mbuf allocation. 278 */ 279 s = splvm(); 280 if (!recurse && arp_locked) { 281 splx(s); 282 return 0; 283 } 284 arp_locked++; 285 splx(s); 286 return 1; 287 } 288 289 static inline void 290 arp_unlock(void) 291 { 292 int s; 293 294 s = splvm(); 295 arp_locked--; 296 splx(s); 297 } 298 299 #ifdef DIAGNOSTIC 300 #define ARP_LOCK(recurse) \ 301 do { \ 302 if (arp_lock_try(recurse) == 0) { \ 303 printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \ 304 panic("arp_lock"); \ 305 } \ 306 } while (/*CONSTCOND*/ 0) 307 #define ARP_LOCK_CHECK() \ 308 do { \ 309 if (arp_locked == 0) { \ 310 printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \ 311 panic("arp lock check"); \ 312 } \ 313 } while (/*CONSTCOND*/ 0) 314 #else 315 #define ARP_LOCK(x) (void) arp_lock_try(x) 316 #define ARP_LOCK_CHECK() /* nothing */ 317 #endif 318 319 #define ARP_UNLOCK() arp_unlock() 320 321 void 322 arp_init(void) 323 { 324 325 arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS); 326 } 327 328 /* 329 * ARP protocol drain routine. Called when memory is in short supply. 330 * Called at splvm(); don't acquire softnet_lock as can be called from 331 * hardware interrupt handlers. 332 */ 333 void 334 arp_drain(void) 335 { 336 struct llinfo_arp *la, *nla; 337 int count = 0; 338 struct mbuf *mold; 339 340 KERNEL_LOCK(1, NULL); 341 342 if (arp_lock_try(0) == 0) { 343 KERNEL_UNLOCK_ONE(NULL); 344 return; 345 } 346 347 for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) { 348 nla = LIST_NEXT(la, la_list); 349 350 mold = la->la_hold; 351 la->la_hold = 0; 352 353 if (mold) { 354 m_freem(mold); 355 count++; 356 } 357 } 358 ARP_UNLOCK(); 359 ARP_STATADD(ARP_STAT_DFRDROPPED, count); 360 KERNEL_UNLOCK_ONE(NULL); 361 } 362 363 364 /* 365 * Timeout routine. Age arp_tab entries periodically. 366 */ 367 /* ARGSUSED */ 368 static void 369 arptimer(void *arg) 370 { 371 struct llinfo_arp *la, *nla; 372 373 mutex_enter(softnet_lock); 374 KERNEL_LOCK(1, NULL); 375 376 if (arp_lock_try(0) == 0) { 377 /* get it later.. */ 378 KERNEL_UNLOCK_ONE(NULL); 379 mutex_exit(softnet_lock); 380 return; 381 } 382 383 callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL); 384 for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) { 385 struct rtentry *rt = la->la_rt; 386 387 nla = LIST_NEXT(la, la_list); 388 if (rt->rt_expire == 0) 389 continue; 390 if ((rt->rt_expire - time_second) < arpt_refresh && 391 rt->rt_pksent > (time_second - arpt_keep)) { 392 /* 393 * If the entry has been used during since last 394 * refresh, try to renew it before deleting. 395 */ 396 arprequest(rt->rt_ifp, 397 &satocsin(rt->rt_ifa->ifa_addr)->sin_addr, 398 &satocsin(rt_getkey(rt))->sin_addr, 399 CLLADDR(rt->rt_ifp->if_sadl)); 400 } else if (rt->rt_expire <= time_second) 401 arptfree(la); /* timer has expired; clear */ 402 } 403 404 ARP_UNLOCK(); 405 406 KERNEL_UNLOCK_ONE(NULL); 407 mutex_exit(softnet_lock); 408 } 409 410 /* 411 * We set the gateway for RTF_CLONING routes to a "prototype" 412 * link-layer sockaddr whose interface type (if_type) and interface 413 * index (if_index) fields are prepared. 414 */ 415 static struct sockaddr * 416 arp_setgate(struct rtentry *rt, struct sockaddr *gate, 417 const struct sockaddr *netmask) 418 { 419 const struct ifnet *ifp = rt->rt_ifp; 420 uint8_t namelen = strlen(ifp->if_xname); 421 uint8_t addrlen = ifp->if_addrlen; 422 423 /* 424 * XXX: If this is a manually added route to interface 425 * such as older version of routed or gated might provide, 426 * restore cloning bit. 427 */ 428 if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL && 429 satocsin(netmask)->sin_addr.s_addr != 0xffffffff) 430 rt->rt_flags |= RTF_CLONING; 431 if (rt->rt_flags & RTF_CLONING) { 432 union { 433 struct sockaddr sa; 434 struct sockaddr_storage ss; 435 struct sockaddr_dl sdl; 436 } u; 437 /* 438 * Case 1: This route should come from a route to iface. 439 */ 440 sockaddr_dl_init(&u.sdl, sizeof(u.ss), 441 ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen); 442 rt_setgate(rt, &u.sa); 443 gate = rt->rt_gateway; 444 } 445 return gate; 446 } 447 448 /* 449 * Parallel to llc_rtrequest. 450 */ 451 void 452 arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 453 { 454 struct sockaddr *gate = rt->rt_gateway; 455 struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo; 456 size_t allocsize; 457 struct mbuf *mold; 458 int s; 459 struct in_ifaddr *ia; 460 struct ifaddr *ifa; 461 struct ifnet *ifp = rt->rt_ifp; 462 uint8_t namelen = strlen(ifp->if_xname); 463 uint8_t addrlen = ifp->if_addrlen; 464 465 if (!arpinit_done) { 466 arpinit_done = 1; 467 /* 468 * We generate expiration times from time_second 469 * so avoid accidentally creating permanent routes. 470 */ 471 if (time_second == 0) { 472 struct timespec ts; 473 ts.tv_sec = 1; 474 ts.tv_nsec = 0; 475 tc_setclock(&ts); 476 } 477 callout_init(&arptimer_ch, CALLOUT_MPSAFE); 478 callout_reset(&arptimer_ch, hz, arptimer, NULL); 479 } 480 481 if (req == RTM_LLINFO_UPD) { 482 struct in_addr *in; 483 484 if ((ifa = info->rti_ifa) == NULL) 485 return; 486 487 in = &ifatoia(ifa)->ia_addr.sin_addr; 488 489 arprequest(ifa->ifa_ifp, in, in, 490 CLLADDR(ifa->ifa_ifp->if_sadl)); 491 return; 492 } 493 494 if ((rt->rt_flags & RTF_GATEWAY) != 0) { 495 if (req != RTM_ADD) 496 return; 497 498 /* 499 * linklayers with particular link MTU limitation. 500 */ 501 switch(ifp->if_type) { 502 #if NFDDI > 0 503 case IFT_FDDI: 504 if (ifp->if_mtu > FDDIIPMTU) 505 rt->rt_rmx.rmx_mtu = FDDIIPMTU; 506 break; 507 #endif 508 #if NARC > 0 509 case IFT_ARCNET: 510 { 511 int arcipifmtu; 512 513 if (ifp->if_flags & IFF_LINK0) 514 arcipifmtu = arc_ipmtu; 515 else 516 arcipifmtu = ARCMTU; 517 if (ifp->if_mtu > arcipifmtu) 518 rt->rt_rmx.rmx_mtu = arcipifmtu; 519 break; 520 } 521 #endif 522 } 523 return; 524 } 525 526 ARP_LOCK(1); /* we may already be locked here. */ 527 528 switch (req) { 529 case RTM_SETGATE: 530 gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]); 531 break; 532 case RTM_ADD: 533 gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]); 534 if (rt->rt_flags & RTF_CLONING) { 535 /* 536 * Give this route an expiration time, even though 537 * it's a "permanent" route, so that routes cloned 538 * from it do not need their expiration time set. 539 */ 540 rt->rt_expire = time_second; 541 /* 542 * linklayers with particular link MTU limitation. 543 */ 544 switch (ifp->if_type) { 545 #if NFDDI > 0 546 case IFT_FDDI: 547 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 && 548 (rt->rt_rmx.rmx_mtu > FDDIIPMTU || 549 (rt->rt_rmx.rmx_mtu == 0 && 550 ifp->if_mtu > FDDIIPMTU))) 551 rt->rt_rmx.rmx_mtu = FDDIIPMTU; 552 break; 553 #endif 554 #if NARC > 0 555 case IFT_ARCNET: 556 { 557 int arcipifmtu; 558 if (ifp->if_flags & IFF_LINK0) 559 arcipifmtu = arc_ipmtu; 560 else 561 arcipifmtu = ARCMTU; 562 563 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 && 564 (rt->rt_rmx.rmx_mtu > arcipifmtu || 565 (rt->rt_rmx.rmx_mtu == 0 && 566 ifp->if_mtu > arcipifmtu))) 567 rt->rt_rmx.rmx_mtu = arcipifmtu; 568 break; 569 } 570 #endif 571 } 572 break; 573 } 574 /* Announce a new entry if requested. */ 575 if (rt->rt_flags & RTF_ANNOUNCE) { 576 arprequest(ifp, 577 &satocsin(rt_getkey(rt))->sin_addr, 578 &satocsin(rt_getkey(rt))->sin_addr, 579 CLLADDR(satocsdl(gate))); 580 } 581 /*FALLTHROUGH*/ 582 case RTM_RESOLVE: 583 if (gate->sa_family != AF_LINK || 584 gate->sa_len < sockaddr_dl_measure(namelen, addrlen)) { 585 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n"); 586 break; 587 } 588 satosdl(gate)->sdl_type = ifp->if_type; 589 satosdl(gate)->sdl_index = ifp->if_index; 590 if (la != NULL) 591 break; /* This happens on a route change */ 592 /* 593 * Case 2: This route may come from cloning, or a manual route 594 * add with a LL address. 595 */ 596 switch (satocsdl(gate)->sdl_type) { 597 #if NTOKEN > 0 598 case IFT_ISO88025: 599 allocsize = sizeof(*la) + sizeof(struct token_rif); 600 break; 601 #endif /* NTOKEN > 0 */ 602 default: 603 allocsize = sizeof(*la); 604 } 605 R_Malloc(la, struct llinfo_arp *, allocsize); 606 rt->rt_llinfo = (void *)la; 607 if (la == NULL) { 608 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n"); 609 break; 610 } 611 arp_inuse++, arp_allocated++; 612 memset(la, 0, allocsize); 613 la->la_rt = rt; 614 rt->rt_flags |= RTF_LLINFO; 615 LIST_INSERT_HEAD(&llinfo_arp, la, la_list); 616 617 INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia); 618 while (ia && ia->ia_ifp != ifp) 619 NEXT_IA_WITH_SAME_ADDR(ia); 620 if (ia) { 621 /* 622 * This test used to be 623 * if (lo0ifp->if_flags & IFF_UP) 624 * It allowed local traffic to be forced through 625 * the hardware by configuring the loopback down. 626 * However, it causes problems during network 627 * configuration for boards that can't receive 628 * packets they send. It is now necessary to clear 629 * "useloopback" and remove the route to force 630 * traffic out to the hardware. 631 * 632 * In 4.4BSD, the above "if" statement checked 633 * rt->rt_ifa against rt_getkey(rt). It was changed 634 * to the current form so that we can provide a 635 * better support for multiple IPv4 addresses on a 636 * interface. 637 */ 638 rt->rt_expire = 0; 639 (void)sockaddr_dl_setaddr(satosdl(gate), gate->sa_len, 640 CLLADDR(ifp->if_sadl), ifp->if_addrlen); 641 if (useloopback) 642 ifp = rt->rt_ifp = lo0ifp; 643 /* 644 * make sure to set rt->rt_ifa to the interface 645 * address we are using, otherwise we will have trouble 646 * with source address selection. 647 */ 648 ifa = &ia->ia_ifa; 649 if (ifa != rt->rt_ifa) 650 rt_replace_ifa(rt, ifa); 651 } 652 break; 653 654 case RTM_DELETE: 655 if (la == NULL) 656 break; 657 arp_inuse--; 658 LIST_REMOVE(la, la_list); 659 rt->rt_llinfo = NULL; 660 rt->rt_flags &= ~RTF_LLINFO; 661 662 s = splnet(); 663 mold = la->la_hold; 664 la->la_hold = 0; 665 splx(s); 666 667 if (mold) 668 m_freem(mold); 669 670 Free((void *)la); 671 } 672 ARP_UNLOCK(); 673 } 674 675 /* 676 * Broadcast an ARP request. Caller specifies: 677 * - arp header source ip address 678 * - arp header target ip address 679 * - arp header source ethernet address 680 */ 681 void 682 arprequest(struct ifnet *ifp, 683 const struct in_addr *sip, const struct in_addr *tip, 684 const u_int8_t *enaddr) 685 { 686 struct mbuf *m; 687 struct arphdr *ah; 688 struct sockaddr sa; 689 uint64_t *arps; 690 691 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 692 return; 693 MCLAIM(m, &arpdomain.dom_mowner); 694 switch (ifp->if_type) { 695 case IFT_IEEE1394: 696 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 697 ifp->if_addrlen; 698 break; 699 default: 700 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 701 2 * ifp->if_addrlen; 702 break; 703 } 704 m->m_pkthdr.len = m->m_len; 705 MH_ALIGN(m, m->m_len); 706 ah = mtod(m, struct arphdr *); 707 memset(ah, 0, m->m_len); 708 switch (ifp->if_type) { 709 case IFT_IEEE1394: /* RFC2734 */ 710 /* fill it now for ar_tpa computation */ 711 ah->ar_hrd = htons(ARPHRD_IEEE1394); 712 break; 713 default: 714 /* ifp->if_output will fill ar_hrd */ 715 break; 716 } 717 ah->ar_pro = htons(ETHERTYPE_IP); 718 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 719 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 720 ah->ar_op = htons(ARPOP_REQUEST); 721 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 722 memcpy(ar_spa(ah), sip, ah->ar_pln); 723 memcpy(ar_tpa(ah), tip, ah->ar_pln); 724 sa.sa_family = AF_ARP; 725 sa.sa_len = 2; 726 m->m_flags |= M_BCAST; 727 arps = ARP_STAT_GETREF(); 728 arps[ARP_STAT_SNDTOTAL]++; 729 arps[ARP_STAT_SENDREQUEST]++; 730 ARP_STAT_PUTREF(); 731 (*ifp->if_output)(ifp, m, &sa, NULL); 732 } 733 734 /* 735 * Resolve an IP address into an ethernet address. If success, 736 * desten is filled in. If there is no entry in arptab, 737 * set one up and broadcast a request for the IP address. 738 * Hold onto this mbuf and resend it once the address 739 * is finally resolved. A return value of 1 indicates 740 * that desten has been filled in and the packet should be sent 741 * normally; a 0 return indicates that the packet has been 742 * taken over here, either now or for later transmission. 743 */ 744 int 745 arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m, 746 const struct sockaddr *dst, u_char *desten) 747 { 748 struct llinfo_arp *la; 749 const struct sockaddr_dl *sdl; 750 struct mbuf *mold; 751 int s; 752 753 if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL) 754 rt = la->la_rt; 755 756 if (la == NULL || rt == NULL) { 757 ARP_STATINC(ARP_STAT_ALLOCFAIL); 758 log(LOG_DEBUG, 759 "arpresolve: can't allocate llinfo on %s for %s\n", 760 ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr)); 761 m_freem(m); 762 return 0; 763 } 764 sdl = satocsdl(rt->rt_gateway); 765 /* 766 * Check the address family and length is valid, the address 767 * is resolved; otherwise, try to resolve. 768 */ 769 if ((rt->rt_expire == 0 || rt->rt_expire > time_second) && 770 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 771 memcpy(desten, CLLADDR(sdl), 772 min(sdl->sdl_alen, ifp->if_addrlen)); 773 rt->rt_pksent = time_second; /* Time for last pkt sent */ 774 return 1; 775 } 776 /* 777 * There is an arptab entry, but no ethernet address 778 * response yet. Replace the held mbuf with this 779 * latest one. 780 */ 781 782 ARP_STATINC(ARP_STAT_DFRTOTAL); 783 s = splnet(); 784 mold = la->la_hold; 785 la->la_hold = m; 786 splx(s); 787 788 if (mold) { 789 ARP_STATINC(ARP_STAT_DFRDROPPED); 790 m_freem(mold); 791 } 792 793 /* 794 * Re-send the ARP request when appropriate. 795 */ 796 #ifdef DIAGNOSTIC 797 if (rt->rt_expire == 0) { 798 /* This should never happen. (Should it? -gwr) */ 799 printf("arpresolve: unresolved and rt_expire == 0\n"); 800 /* Set expiration time to now (expired). */ 801 rt->rt_expire = time_second; 802 } 803 #endif 804 if (rt->rt_expire) { 805 rt->rt_flags &= ~RTF_REJECT; 806 if (la->la_asked == 0 || rt->rt_expire != time_second) { 807 rt->rt_expire = time_second; 808 if (la->la_asked++ < arp_maxtries) { 809 arprequest(ifp, 810 &satocsin(rt->rt_ifa->ifa_addr)->sin_addr, 811 &satocsin(dst)->sin_addr, 812 #if NCARP > 0 813 (rt->rt_ifp->if_type == IFT_CARP) ? 814 CLLADDR(rt->rt_ifp->if_sadl): 815 #endif 816 CLLADDR(ifp->if_sadl)); 817 } else { 818 rt->rt_flags |= RTF_REJECT; 819 rt->rt_expire += arpt_down; 820 la->la_asked = 0; 821 } 822 } 823 } 824 return 0; 825 } 826 827 /* 828 * Common length and type checks are done here, 829 * then the protocol-specific routine is called. 830 */ 831 void 832 arpintr(void) 833 { 834 struct mbuf *m; 835 struct arphdr *ar; 836 int s; 837 int arplen; 838 839 mutex_enter(softnet_lock); 840 KERNEL_LOCK(1, NULL); 841 while (arpintrq.ifq_head) { 842 s = splnet(); 843 IF_DEQUEUE(&arpintrq, m); 844 splx(s); 845 if (m == 0 || (m->m_flags & M_PKTHDR) == 0) 846 panic("arpintr"); 847 848 MCLAIM(m, &arpdomain.dom_mowner); 849 ARP_STATINC(ARP_STAT_RCVTOTAL); 850 851 /* 852 * First, make sure we have at least struct arphdr. 853 */ 854 if (m->m_len < sizeof(struct arphdr) || 855 (ar = mtod(m, struct arphdr *)) == NULL) 856 goto badlen; 857 858 switch (m->m_pkthdr.rcvif->if_type) { 859 case IFT_IEEE1394: 860 arplen = sizeof(struct arphdr) + 861 ar->ar_hln + 2 * ar->ar_pln; 862 break; 863 default: 864 arplen = sizeof(struct arphdr) + 865 2 * ar->ar_hln + 2 * ar->ar_pln; 866 break; 867 } 868 869 if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */ 870 m->m_len >= arplen) 871 switch (ntohs(ar->ar_pro)) { 872 case ETHERTYPE_IP: 873 case ETHERTYPE_IPTRAILERS: 874 in_arpinput(m); 875 continue; 876 default: 877 ARP_STATINC(ARP_STAT_RCVBADPROTO); 878 } 879 else { 880 badlen: 881 ARP_STATINC(ARP_STAT_RCVBADLEN); 882 } 883 m_freem(m); 884 } 885 KERNEL_UNLOCK_ONE(NULL); 886 mutex_exit(softnet_lock); 887 } 888 889 /* 890 * ARP for Internet protocols on 10 Mb/s Ethernet. 891 * Algorithm is that given in RFC 826. 892 * In addition, a sanity check is performed on the sender 893 * protocol address, to catch impersonators. 894 * We no longer handle negotiations for use of trailer protocol: 895 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 896 * along with IP replies if we wanted trailers sent to us, 897 * and also sent them in response to IP replies. 898 * This allowed either end to announce the desire to receive 899 * trailer packets. 900 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 901 * but formerly didn't normally send requests. 902 */ 903 static void 904 in_arpinput(struct mbuf *m) 905 { 906 struct arphdr *ah; 907 struct ifnet *ifp = m->m_pkthdr.rcvif; 908 struct llinfo_arp *la = NULL; 909 struct rtentry *rt; 910 struct in_ifaddr *ia; 911 #if NBRIDGE > 0 912 struct in_ifaddr *bridge_ia = NULL; 913 #endif 914 #if NCARP > 0 915 u_int32_t count = 0, index = 0; 916 #endif 917 struct sockaddr_dl *sdl; 918 struct sockaddr sa; 919 struct in_addr isaddr, itaddr, myaddr; 920 int op; 921 struct mbuf *mold; 922 void *tha; 923 int s; 924 uint64_t *arps; 925 926 if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT))) 927 goto out; 928 ah = mtod(m, struct arphdr *); 929 op = ntohs(ah->ar_op); 930 931 /* 932 * Fix up ah->ar_hrd if necessary, before using ar_tha() or 933 * ar_tpa(). 934 */ 935 switch (ifp->if_type) { 936 case IFT_IEEE1394: 937 if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394) 938 ; 939 else { 940 /* XXX this is to make sure we compute ar_tha right */ 941 /* XXX check ar_hrd more strictly? */ 942 ah->ar_hrd = htons(ARPHRD_IEEE1394); 943 } 944 break; 945 default: 946 /* XXX check ar_hrd? */ 947 break; 948 } 949 950 memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 951 memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 952 953 if (m->m_flags & (M_BCAST|M_MCAST)) 954 ARP_STATINC(ARP_STAT_RCVMCAST); 955 956 /* 957 * If the target IP address is zero, ignore the packet. 958 * This prevents the code below from tring to answer 959 * when we are using IP address zero (booting). 960 */ 961 if (in_nullhost(itaddr)) { 962 ARP_STATINC(ARP_STAT_RCVZEROTPA); 963 goto out; 964 } 965 966 /* 967 * If the source IP address is zero, this is most likely a 968 * confused host trying to use IP address zero. (Windoze?) 969 * XXX: Should we bother trying to reply to these? 970 */ 971 if (in_nullhost(isaddr)) { 972 ARP_STATINC(ARP_STAT_RCVZEROSPA); 973 goto out; 974 } 975 976 /* 977 * Search for a matching interface address 978 * or any address on the interface to use 979 * as a dummy address in the rest of this function 980 */ 981 982 INADDR_TO_IA(itaddr, ia); 983 while (ia != NULL) { 984 #if NCARP > 0 985 if (ia->ia_ifp->if_type == IFT_CARP && 986 ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) == 987 (IFF_UP|IFF_RUNNING))) { 988 index++; 989 if (ia->ia_ifp == m->m_pkthdr.rcvif && 990 carp_iamatch(ia, ar_sha(ah), 991 &count, index)) { 992 break; 993 } 994 } else 995 #endif 996 if (ia->ia_ifp == m->m_pkthdr.rcvif) 997 break; 998 #if NBRIDGE > 0 999 /* 1000 * If the interface we received the packet on 1001 * is part of a bridge, check to see if we need 1002 * to "bridge" the packet to ourselves at this 1003 * layer. Note we still prefer a perfect match, 1004 * but allow this weaker match if necessary. 1005 */ 1006 if (m->m_pkthdr.rcvif->if_bridge != NULL && 1007 m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge) 1008 bridge_ia = ia; 1009 #endif /* NBRIDGE > 0 */ 1010 1011 NEXT_IA_WITH_SAME_ADDR(ia); 1012 } 1013 1014 #if NBRIDGE > 0 1015 if (ia == NULL && bridge_ia != NULL) { 1016 ia = bridge_ia; 1017 ifp = bridge_ia->ia_ifp; 1018 } 1019 #endif 1020 1021 if (ia == NULL) { 1022 INADDR_TO_IA(isaddr, ia); 1023 while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif) 1024 NEXT_IA_WITH_SAME_ADDR(ia); 1025 1026 if (ia == NULL) { 1027 IFP_TO_IA(ifp, ia); 1028 if (ia == NULL) { 1029 ARP_STATINC(ARP_STAT_RCVNOINT); 1030 goto out; 1031 } 1032 } 1033 } 1034 1035 myaddr = ia->ia_addr.sin_addr; 1036 1037 /* XXX checks for bridge case? */ 1038 if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) { 1039 ARP_STATINC(ARP_STAT_RCVLOCALSHA); 1040 goto out; /* it's from me, ignore it. */ 1041 } 1042 1043 /* XXX checks for bridge case? */ 1044 if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 1045 ARP_STATINC(ARP_STAT_RCVBCASTSHA); 1046 log(LOG_ERR, 1047 "%s: arp: link address is broadcast for IP address %s!\n", 1048 ifp->if_xname, in_fmtaddr(isaddr)); 1049 goto out; 1050 } 1051 1052 if (in_hosteq(isaddr, myaddr)) { 1053 ARP_STATINC(ARP_STAT_RCVLOCALSPA); 1054 log(LOG_ERR, 1055 "duplicate IP address %s sent from link address %s\n", 1056 in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln)); 1057 itaddr = myaddr; 1058 goto reply; 1059 } 1060 la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0); 1061 if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) { 1062 if (sdl->sdl_alen && 1063 memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) { 1064 if (rt->rt_flags & RTF_STATIC) { 1065 ARP_STATINC(ARP_STAT_RCVOVERPERM); 1066 log(LOG_INFO, 1067 "%s tried to overwrite permanent arp info" 1068 " for %s\n", 1069 lla_snprintf(ar_sha(ah), ah->ar_hln), 1070 in_fmtaddr(isaddr)); 1071 goto out; 1072 } else if (rt->rt_ifp != ifp) { 1073 ARP_STATINC(ARP_STAT_RCVOVERINT); 1074 log(LOG_INFO, 1075 "%s on %s tried to overwrite " 1076 "arp info for %s on %s\n", 1077 lla_snprintf(ar_sha(ah), ah->ar_hln), 1078 ifp->if_xname, in_fmtaddr(isaddr), 1079 rt->rt_ifp->if_xname); 1080 goto out; 1081 } else { 1082 ARP_STATINC(ARP_STAT_RCVOVER); 1083 log(LOG_INFO, 1084 "arp info overwritten for %s by %s\n", 1085 in_fmtaddr(isaddr), 1086 lla_snprintf(ar_sha(ah), ah->ar_hln)); 1087 } 1088 } 1089 /* 1090 * sanity check for the address length. 1091 * XXX this does not work for protocols with variable address 1092 * length. -is 1093 */ 1094 if (sdl->sdl_alen && 1095 sdl->sdl_alen != ah->ar_hln) { 1096 ARP_STATINC(ARP_STAT_RCVLENCHG); 1097 log(LOG_WARNING, 1098 "arp from %s: new addr len %d, was %d\n", 1099 in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen); 1100 } 1101 if (ifp->if_addrlen != ah->ar_hln) { 1102 ARP_STATINC(ARP_STAT_RCVBADLEN); 1103 log(LOG_WARNING, 1104 "arp from %s: addr len: new %d, i/f %d (ignored)\n", 1105 in_fmtaddr(isaddr), ah->ar_hln, 1106 ifp->if_addrlen); 1107 goto reply; 1108 } 1109 #if NTOKEN > 0 1110 /* 1111 * XXX uses m_data and assumes the complete answer including 1112 * XXX token-ring headers is in the same buf 1113 */ 1114 if (ifp->if_type == IFT_ISO88025) { 1115 struct token_header *trh; 1116 1117 trh = (struct token_header *)M_TRHSTART(m); 1118 if (trh->token_shost[0] & TOKEN_RI_PRESENT) { 1119 struct token_rif *rif; 1120 size_t riflen; 1121 1122 rif = TOKEN_RIF(trh); 1123 riflen = (ntohs(rif->tr_rcf) & 1124 TOKEN_RCF_LEN_MASK) >> 8; 1125 1126 if (riflen > 2 && 1127 riflen < sizeof(struct token_rif) && 1128 (riflen & 1) == 0) { 1129 rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION); 1130 rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK); 1131 memcpy(TOKEN_RIF(la), rif, riflen); 1132 } 1133 } 1134 } 1135 #endif /* NTOKEN > 0 */ 1136 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah), 1137 ah->ar_hln); 1138 if (rt->rt_expire) 1139 rt->rt_expire = time_second + arpt_keep; 1140 rt->rt_flags &= ~RTF_REJECT; 1141 la->la_asked = 0; 1142 1143 s = splnet(); 1144 mold = la->la_hold; 1145 la->la_hold = 0; 1146 splx(s); 1147 1148 if (mold) { 1149 ARP_STATINC(ARP_STAT_DFRSENT); 1150 (*ifp->if_output)(ifp, mold, rt_getkey(rt), rt); 1151 } 1152 } 1153 reply: 1154 if (op != ARPOP_REQUEST) { 1155 if (op == ARPOP_REPLY) 1156 ARP_STATINC(ARP_STAT_RCVREPLY); 1157 out: 1158 m_freem(m); 1159 return; 1160 } 1161 ARP_STATINC(ARP_STAT_RCVREQUEST); 1162 if (in_hosteq(itaddr, myaddr)) { 1163 /* I am the target */ 1164 tha = ar_tha(ah); 1165 if (tha) 1166 memcpy(tha, ar_sha(ah), ah->ar_hln); 1167 memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln); 1168 } else { 1169 la = arplookup(m, &itaddr, 0, SIN_PROXY); 1170 if (la == NULL) 1171 goto out; 1172 rt = la->la_rt; 1173 if (rt->rt_ifp->if_type == IFT_CARP && 1174 m->m_pkthdr.rcvif->if_type != IFT_CARP) 1175 goto out; 1176 tha = ar_tha(ah); 1177 if (tha) 1178 memcpy(tha, ar_sha(ah), ah->ar_hln); 1179 sdl = satosdl(rt->rt_gateway); 1180 memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln); 1181 } 1182 1183 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 1184 memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 1185 ah->ar_op = htons(ARPOP_REPLY); 1186 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 1187 switch (ifp->if_type) { 1188 case IFT_IEEE1394: 1189 /* 1190 * ieee1394 arp reply is broadcast 1191 */ 1192 m->m_flags &= ~M_MCAST; 1193 m->m_flags |= M_BCAST; 1194 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln; 1195 break; 1196 1197 default: 1198 m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */ 1199 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 1200 break; 1201 } 1202 m->m_pkthdr.len = m->m_len; 1203 sa.sa_family = AF_ARP; 1204 sa.sa_len = 2; 1205 arps = ARP_STAT_GETREF(); 1206 arps[ARP_STAT_SNDTOTAL]++; 1207 arps[ARP_STAT_SNDREPLY]++; 1208 ARP_STAT_PUTREF(); 1209 (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); 1210 return; 1211 } 1212 1213 /* 1214 * Free an arp entry. 1215 */ 1216 static void arptfree(struct llinfo_arp *la) 1217 { 1218 struct rtentry *rt = la->la_rt; 1219 struct sockaddr_dl *sdl; 1220 1221 ARP_LOCK_CHECK(); 1222 1223 if (rt == NULL) 1224 panic("arptfree"); 1225 if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) && 1226 sdl->sdl_family == AF_LINK) { 1227 sdl->sdl_alen = 0; 1228 la->la_asked = 0; 1229 rt->rt_flags &= ~RTF_REJECT; 1230 return; 1231 } 1232 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1233 } 1234 1235 static struct llinfo_arp * 1236 arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy) 1237 { 1238 return arplookup1(m, addr, create, proxy, NULL); 1239 } 1240 1241 /* 1242 * Lookup or enter a new address in arptab. 1243 */ 1244 static struct llinfo_arp * 1245 arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy, 1246 struct rtentry *rt0) 1247 { 1248 struct arphdr *ah; 1249 struct ifnet *ifp = m->m_pkthdr.rcvif; 1250 struct rtentry *rt; 1251 struct sockaddr_inarp sin; 1252 const char *why = NULL; 1253 1254 ah = mtod(m, struct arphdr *); 1255 if (rt0 == NULL) { 1256 memset(&sin, 0, sizeof(sin)); 1257 sin.sin_len = sizeof(sin); 1258 sin.sin_family = AF_INET; 1259 sin.sin_addr = *addr; 1260 sin.sin_other = proxy ? SIN_PROXY : 0; 1261 rt = rtalloc1(sintosa(&sin), create); 1262 if (rt == NULL) 1263 return NULL; 1264 rt->rt_refcnt--; 1265 } else 1266 rt = rt0; 1267 1268 #define IS_LLINFO(__rt) \ 1269 (((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \ 1270 (__rt)->rt_gateway->sa_family == AF_LINK) 1271 1272 1273 if (IS_LLINFO(rt)) 1274 return (struct llinfo_arp *)rt->rt_llinfo; 1275 1276 if (create) { 1277 if (rt->rt_flags & RTF_GATEWAY) 1278 why = "host is not on local network"; 1279 else if ((rt->rt_flags & RTF_LLINFO) == 0) { 1280 ARP_STATINC(ARP_STAT_ALLOCFAIL); 1281 why = "could not allocate llinfo"; 1282 } else 1283 why = "gateway route is not ours"; 1284 log(LOG_DEBUG, "arplookup: unable to enter address" 1285 " for %s@%s on %s (%s)\n", 1286 in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln), 1287 (ifp) ? ifp->if_xname : "null", why); 1288 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) { 1289 rtrequest(RTM_DELETE, rt_getkey(rt), 1290 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1291 } 1292 } 1293 return NULL; 1294 } 1295 1296 int 1297 arpioctl(u_long cmd, void *data) 1298 { 1299 1300 return EOPNOTSUPP; 1301 } 1302 1303 void 1304 arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1305 { 1306 struct in_addr *ip; 1307 1308 /* 1309 * Warn the user if another station has this IP address, 1310 * but only if the interface IP address is not zero. 1311 */ 1312 ip = &IA_SIN(ifa)->sin_addr; 1313 if (!in_nullhost(*ip)) 1314 arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl)); 1315 1316 ifa->ifa_rtrequest = arp_rtrequest; 1317 ifa->ifa_flags |= RTF_CLONING; 1318 } 1319 1320 /* 1321 * Called from 10 Mb/s Ethernet interrupt handlers 1322 * when ether packet type ETHERTYPE_REVARP 1323 * is received. Common length and type checks are done here, 1324 * then the protocol-specific routine is called. 1325 */ 1326 void 1327 revarpinput(struct mbuf *m) 1328 { 1329 struct arphdr *ar; 1330 1331 if (m->m_len < sizeof(struct arphdr)) 1332 goto out; 1333 ar = mtod(m, struct arphdr *); 1334 #if 0 /* XXX I don't think we need this... and it will prevent other LL */ 1335 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) 1336 goto out; 1337 #endif 1338 if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln)) 1339 goto out; 1340 switch (ntohs(ar->ar_pro)) { 1341 case ETHERTYPE_IP: 1342 case ETHERTYPE_IPTRAILERS: 1343 in_revarpinput(m); 1344 return; 1345 1346 default: 1347 break; 1348 } 1349 out: 1350 m_freem(m); 1351 } 1352 1353 /* 1354 * RARP for Internet protocols on 10 Mb/s Ethernet. 1355 * Algorithm is that given in RFC 903. 1356 * We are only using for bootstrap purposes to get an ip address for one of 1357 * our interfaces. Thus we support no user-interface. 1358 * 1359 * Since the contents of the RARP reply are specific to the interface that 1360 * sent the request, this code must ensure that they are properly associated. 1361 * 1362 * Note: also supports ARP via RARP packets, per the RFC. 1363 */ 1364 void 1365 in_revarpinput(struct mbuf *m) 1366 { 1367 struct ifnet *ifp; 1368 struct arphdr *ah; 1369 void *tha; 1370 int op; 1371 1372 ah = mtod(m, struct arphdr *); 1373 op = ntohs(ah->ar_op); 1374 1375 switch (m->m_pkthdr.rcvif->if_type) { 1376 case IFT_IEEE1394: 1377 /* ARP without target hardware address is not supported */ 1378 goto out; 1379 default: 1380 break; 1381 } 1382 1383 switch (op) { 1384 case ARPOP_REQUEST: 1385 case ARPOP_REPLY: /* per RFC */ 1386 in_arpinput(m); 1387 return; 1388 case ARPOP_REVREPLY: 1389 break; 1390 case ARPOP_REVREQUEST: /* handled by rarpd(8) */ 1391 default: 1392 goto out; 1393 } 1394 if (!revarp_in_progress) 1395 goto out; 1396 ifp = m->m_pkthdr.rcvif; 1397 if (ifp != myip_ifp) /* !same interface */ 1398 goto out; 1399 if (myip_initialized) 1400 goto wake; 1401 tha = ar_tha(ah); 1402 KASSERT(tha); 1403 if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen)) 1404 goto out; 1405 memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip)); 1406 memcpy(&myip, ar_tpa(ah), sizeof(myip)); 1407 myip_initialized = 1; 1408 wake: /* Do wakeup every time in case it was missed. */ 1409 wakeup((void *)&myip); 1410 1411 out: 1412 m_freem(m); 1413 } 1414 1415 /* 1416 * Send a RARP request for the ip address of the specified interface. 1417 * The request should be RFC 903-compliant. 1418 */ 1419 void 1420 revarprequest(struct ifnet *ifp) 1421 { 1422 struct sockaddr sa; 1423 struct mbuf *m; 1424 struct arphdr *ah; 1425 void *tha; 1426 1427 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 1428 return; 1429 MCLAIM(m, &arpdomain.dom_mowner); 1430 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 1431 2*ifp->if_addrlen; 1432 m->m_pkthdr.len = m->m_len; 1433 MH_ALIGN(m, m->m_len); 1434 ah = mtod(m, struct arphdr *); 1435 memset(ah, 0, m->m_len); 1436 ah->ar_pro = htons(ETHERTYPE_IP); 1437 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 1438 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 1439 ah->ar_op = htons(ARPOP_REVREQUEST); 1440 1441 memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln); 1442 tha = ar_tha(ah); 1443 KASSERT(tha); 1444 memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln); 1445 1446 sa.sa_family = AF_ARP; 1447 sa.sa_len = 2; 1448 m->m_flags |= M_BCAST; 1449 (*ifp->if_output)(ifp, m, &sa, NULL); 1450 1451 } 1452 1453 /* 1454 * RARP for the ip address of the specified interface, but also 1455 * save the ip address of the server that sent the answer. 1456 * Timeout if no response is received. 1457 */ 1458 int 1459 revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in, 1460 struct in_addr *clnt_in) 1461 { 1462 int result, count = 20; 1463 1464 myip_initialized = 0; 1465 myip_ifp = ifp; 1466 1467 revarp_in_progress = 1; 1468 while (count--) { 1469 revarprequest(ifp); 1470 result = tsleep((void *)&myip, PSOCK, "revarp", hz/2); 1471 if (result != EWOULDBLOCK) 1472 break; 1473 } 1474 revarp_in_progress = 0; 1475 1476 if (!myip_initialized) 1477 return ENETUNREACH; 1478 1479 memcpy(serv_in, &srv_ip, sizeof(*serv_in)); 1480 memcpy(clnt_in, &myip, sizeof(*clnt_in)); 1481 return 0; 1482 } 1483 1484 1485 1486 #ifdef DDB 1487 1488 #include <machine/db_machdep.h> 1489 #include <ddb/db_interface.h> 1490 #include <ddb/db_output.h> 1491 1492 static void 1493 db_print_sa(const struct sockaddr *sa) 1494 { 1495 int len; 1496 const u_char *p; 1497 1498 if (sa == NULL) { 1499 db_printf("[NULL]"); 1500 return; 1501 } 1502 1503 p = (const u_char *)sa; 1504 len = sa->sa_len; 1505 db_printf("["); 1506 while (len > 0) { 1507 db_printf("%d", *p); 1508 p++; len--; 1509 if (len) db_printf(","); 1510 } 1511 db_printf("]\n"); 1512 } 1513 1514 static void 1515 db_print_ifa(struct ifaddr *ifa) 1516 { 1517 if (ifa == NULL) 1518 return; 1519 db_printf(" ifa_addr="); 1520 db_print_sa(ifa->ifa_addr); 1521 db_printf(" ifa_dsta="); 1522 db_print_sa(ifa->ifa_dstaddr); 1523 db_printf(" ifa_mask="); 1524 db_print_sa(ifa->ifa_netmask); 1525 db_printf(" flags=0x%x,refcnt=%d,metric=%d\n", 1526 ifa->ifa_flags, 1527 ifa->ifa_refcnt, 1528 ifa->ifa_metric); 1529 } 1530 1531 static void 1532 db_print_llinfo(void *li) 1533 { 1534 struct llinfo_arp *la; 1535 1536 if (li == NULL) 1537 return; 1538 la = (struct llinfo_arp *)li; 1539 db_printf(" la_rt=%p la_hold=%p, la_asked=0x%lx\n", 1540 la->la_rt, la->la_hold, la->la_asked); 1541 } 1542 1543 /* 1544 * Function to pass to rt_walktree(). 1545 * Return non-zero error to abort walk. 1546 */ 1547 static int 1548 db_show_rtentry(struct rtentry *rt, void *w) 1549 { 1550 db_printf("rtentry=%p", rt); 1551 1552 db_printf(" flags=0x%x refcnt=%d use=%ld expire=%lld\n", 1553 rt->rt_flags, rt->rt_refcnt, 1554 rt->rt_use, (long long)rt->rt_expire); 1555 1556 db_printf(" key="); db_print_sa(rt_getkey(rt)); 1557 db_printf(" mask="); db_print_sa(rt_mask(rt)); 1558 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1559 1560 db_printf(" ifp=%p ", rt->rt_ifp); 1561 if (rt->rt_ifp) 1562 db_printf("(%s)", rt->rt_ifp->if_xname); 1563 else 1564 db_printf("(NULL)"); 1565 1566 db_printf(" ifa=%p\n", rt->rt_ifa); 1567 db_print_ifa(rt->rt_ifa); 1568 1569 db_printf(" gwroute=%p llinfo=%p\n", 1570 rt->rt_gwroute, rt->rt_llinfo); 1571 db_print_llinfo(rt->rt_llinfo); 1572 1573 return 0; 1574 } 1575 1576 /* 1577 * Function to print all the route trees. 1578 * Use this from ddb: "show arptab" 1579 */ 1580 void 1581 db_show_arptab(db_expr_t addr, bool have_addr, 1582 db_expr_t count, const char *modif) 1583 { 1584 rt_walktree(AF_INET, db_show_rtentry, NULL); 1585 } 1586 #endif 1587 1588 static int 1589 sysctl_net_inet_arp_stats(SYSCTLFN_ARGS) 1590 { 1591 1592 return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS); 1593 } 1594 1595 SYSCTL_SETUP(sysctl_net_inet_arp_setup, "sysctl net.inet.arp subtree setup") 1596 { 1597 const struct sysctlnode *node; 1598 1599 sysctl_createv(clog, 0, NULL, NULL, 1600 CTLFLAG_PERMANENT, 1601 CTLTYPE_NODE, "net", NULL, 1602 NULL, 0, NULL, 0, 1603 CTL_NET, CTL_EOL); 1604 sysctl_createv(clog, 0, NULL, NULL, 1605 CTLFLAG_PERMANENT, 1606 CTLTYPE_NODE, "inet", NULL, 1607 NULL, 0, NULL, 0, 1608 CTL_NET, PF_INET, CTL_EOL); 1609 sysctl_createv(clog, 0, NULL, &node, 1610 CTLFLAG_PERMANENT, 1611 CTLTYPE_NODE, "arp", 1612 SYSCTL_DESCR("Address Resolution Protocol"), 1613 NULL, 0, NULL, 0, 1614 CTL_NET, PF_INET, CTL_CREATE, CTL_EOL); 1615 1616 sysctl_createv(clog, 0, NULL, NULL, 1617 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1618 CTLTYPE_INT, "prune", 1619 SYSCTL_DESCR("ARP cache pruning interval"), 1620 NULL, 0, &arpt_prune, 0, 1621 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1622 1623 sysctl_createv(clog, 0, NULL, NULL, 1624 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1625 CTLTYPE_INT, "keep", 1626 SYSCTL_DESCR("Valid ARP entry lifetime"), 1627 NULL, 0, &arpt_keep, 0, 1628 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1629 1630 sysctl_createv(clog, 0, NULL, NULL, 1631 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1632 CTLTYPE_INT, "down", 1633 SYSCTL_DESCR("Failed ARP entry lifetime"), 1634 NULL, 0, &arpt_down, 0, 1635 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1636 1637 sysctl_createv(clog, 0, NULL, NULL, 1638 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1639 CTLTYPE_INT, "refresh", 1640 SYSCTL_DESCR("ARP entry refresh interval"), 1641 NULL, 0, &arpt_refresh, 0, 1642 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1643 1644 sysctl_createv(clog, 0, NULL, NULL, 1645 CTLFLAG_PERMANENT, 1646 CTLTYPE_STRUCT, "stats", 1647 SYSCTL_DESCR("ARP statistics"), 1648 sysctl_net_inet_arp_stats, 0, NULL, 0, 1649 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1650 } 1651 1652 #endif /* INET */ 1653