1/* $NetBSD: if_arp.c,v 1.154 2012/01/02 22:17:11 liamjfoy 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 1.154 2012/01/02 22:17:11 liamjfoy 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 */ 138int arpt_prune = (5*60*1); /* walk list every 5 minutes */ 139int arpt_keep = (20*60); /* once resolved, good for 20 more minutes */ 140int arpt_down = 20; /* once declared down, don't send for 20 secs */ 141int 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 145static struct sockaddr *arp_setgate(struct rtentry *, struct sockaddr *, 146 const struct sockaddr *); 147static void arptfree(struct llinfo_arp *); 148static void arptimer(void *); 149static struct llinfo_arp *arplookup1(struct mbuf *, const struct in_addr *, 150 int, int, struct rtentry *); 151static struct llinfo_arp *arplookup(struct mbuf *, const struct in_addr *, 152 int, int); 153static void in_arpinput(struct mbuf *); 154static void arp_drainstub(void); 155 156LIST_HEAD(, llinfo_arp) llinfo_arp; 157struct ifqueue arpintrq = { 158 .ifq_head = NULL, 159 .ifq_tail = NULL, 160 .ifq_len = 0, 161 .ifq_maxlen = 50, 162 .ifq_drops = 0, 163}; 164int arp_inuse, arp_allocated; 165int arp_maxtries = 5; 166int useloopback = 1; /* use loopback interface for local traffic */ 167int arpinit_done = 0; 168 169static percpu_t *arpstat_percpu; 170 171#define ARP_STAT_GETREF() _NET_STAT_GETREF(arpstat_percpu) 172#define ARP_STAT_PUTREF() _NET_STAT_PUTREF(arpstat_percpu) 173 174#define ARP_STATINC(x) _NET_STATINC(arpstat_percpu, x) 175#define ARP_STATADD(x, v) _NET_STATADD(arpstat_percpu, x, v) 176 177struct callout arptimer_ch; 178 179/* revarp state */ 180struct in_addr myip, srv_ip; 181int myip_initialized = 0; 182int revarp_in_progress = 0; 183struct ifnet *myip_ifp = NULL; 184 185#ifdef DDB 186static void db_print_sa(const struct sockaddr *); 187static void db_print_ifa(struct ifaddr *); 188static void db_print_llinfo(void *); 189static int db_show_rtentry(struct rtentry *, void *); 190#endif 191 192static int arp_drainwanted; 193 194static int log_movements = 1; 195static int log_permanent_modify = 1; 196static int log_wrong_iface = 1; 197 198/* 199 * this should be elsewhere. 200 */ 201 202static char * 203lla_snprintf(u_int8_t *, int); 204 205static char * 206lla_snprintf(u_int8_t *adrp, int len) 207{ 208#define NUMBUFS 3 209 static char buf[NUMBUFS][16*3]; 210 static int bnum = 0; 211 212 int i; 213 char *p; 214 215 p = buf[bnum]; 216 217 *p++ = hexdigits[(*adrp)>>4]; 218 *p++ = hexdigits[(*adrp++)&0xf]; 219 220 for (i=1; i<len && i<16; i++) { 221 *p++ = ':'; 222 *p++ = hexdigits[(*adrp)>>4]; 223 *p++ = hexdigits[(*adrp++)&0xf]; 224 } 225 226 *p = 0; 227 p = buf[bnum]; 228 bnum = (bnum + 1) % NUMBUFS; 229 return p; 230} 231 232DOMAIN_DEFINE(arpdomain); /* forward declare and add to link set */ 233 234static void 235arp_fasttimo(void) 236{ 237 if (arp_drainwanted) { 238 arp_drain(); 239 arp_drainwanted = 0; 240 } 241} 242 243const struct protosw arpsw[] = { 244 { .pr_type = 0, 245 .pr_domain = &arpdomain, 246 .pr_protocol = 0, 247 .pr_flags = 0, 248 .pr_input = 0, 249 .pr_output = 0, 250 .pr_ctlinput = 0, 251 .pr_ctloutput = 0, 252 .pr_usrreq = 0, 253 .pr_init = arp_init, 254 .pr_fasttimo = arp_fasttimo, 255 .pr_slowtimo = 0, 256 .pr_drain = arp_drainstub, 257 } 258}; 259 260 261struct domain arpdomain = { 262 .dom_family = PF_ARP, 263 .dom_name = "arp", 264 .dom_protosw = arpsw, 265 .dom_protoswNPROTOSW = &arpsw[__arraycount(arpsw)], 266}; 267 268/* 269 * ARP table locking. 270 * 271 * to prevent lossage vs. the arp_drain routine (which may be called at 272 * any time, including in a device driver context), we do two things: 273 * 274 * 1) manipulation of la->la_hold is done at splnet() (for all of 275 * about two instructions). 276 * 277 * 2) manipulation of the arp table's linked list is done under the 278 * protection of the ARP_LOCK; if arp_drain() or arptimer is called 279 * while the arp table is locked, we punt and try again later. 280 */ 281 282static int arp_locked; 283static inline int arp_lock_try(int); 284static inline void arp_unlock(void); 285 286static inline int 287arp_lock_try(int recurse) 288{ 289 int s; 290 291 /* 292 * Use splvm() -- we're blocking things that would cause 293 * mbuf allocation. 294 */ 295 s = splvm(); 296 if (!recurse && arp_locked) { 297 splx(s); 298 return 0; 299 } 300 arp_locked++; 301 splx(s); 302 return 1; 303} 304 305static inline void 306arp_unlock(void) 307{ 308 int s; 309 310 s = splvm(); 311 arp_locked--; 312 splx(s); 313} 314 315#ifdef DIAGNOSTIC 316#define ARP_LOCK(recurse) \ 317do { \ 318 if (arp_lock_try(recurse) == 0) { \ 319 printf("%s:%d: arp already locked\n", __FILE__, __LINE__); \ 320 panic("arp_lock"); \ 321 } \ 322} while (/*CONSTCOND*/ 0) 323#define ARP_LOCK_CHECK() \ 324do { \ 325 if (arp_locked == 0) { \ 326 printf("%s:%d: arp lock not held\n", __FILE__, __LINE__); \ 327 panic("arp lock check"); \ 328 } \ 329} while (/*CONSTCOND*/ 0) 330#else 331#define ARP_LOCK(x) (void) arp_lock_try(x) 332#define ARP_LOCK_CHECK() /* nothing */ 333#endif 334 335#define ARP_UNLOCK() arp_unlock() 336 337static void sysctl_net_inet_arp_setup(struct sysctllog **); 338 339void 340arp_init(void) 341{ 342 343 sysctl_net_inet_arp_setup(NULL); 344 arpstat_percpu = percpu_alloc(sizeof(uint64_t) * ARP_NSTATS); 345} 346 347static void 348arp_drainstub(void) 349{ 350 arp_drainwanted = 1; 351} 352 353/* 354 * ARP protocol drain routine. Called when memory is in short supply. 355 * Called at splvm(); don't acquire softnet_lock as can be called from 356 * hardware interrupt handlers. 357 */ 358void 359arp_drain(void) 360{ 361 struct llinfo_arp *la, *nla; 362 int count = 0; 363 struct mbuf *mold; 364 365 KERNEL_LOCK(1, NULL); 366 367 if (arp_lock_try(0) == 0) { 368 KERNEL_UNLOCK_ONE(NULL); 369 return; 370 } 371 372 for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) { 373 nla = LIST_NEXT(la, la_list); 374 375 mold = la->la_hold; 376 la->la_hold = 0; 377 378 if (mold) { 379 m_freem(mold); 380 count++; 381 } 382 } 383 ARP_UNLOCK(); 384 ARP_STATADD(ARP_STAT_DFRDROPPED, count); 385 KERNEL_UNLOCK_ONE(NULL); 386} 387 388 389/* 390 * Timeout routine. Age arp_tab entries periodically. 391 */ 392/* ARGSUSED */ 393static void 394arptimer(void *arg) 395{ 396 struct llinfo_arp *la, *nla; 397 398 mutex_enter(softnet_lock); 399 KERNEL_LOCK(1, NULL); 400 401 if (arp_lock_try(0) == 0) { 402 /* get it later.. */ 403 KERNEL_UNLOCK_ONE(NULL); 404 mutex_exit(softnet_lock); 405 return; 406 } 407 408 callout_reset(&arptimer_ch, arpt_prune * hz, arptimer, NULL); 409 for (la = LIST_FIRST(&llinfo_arp); la != NULL; la = nla) { 410 struct rtentry *rt = la->la_rt; 411 412 nla = LIST_NEXT(la, la_list); 413 if (rt->rt_expire == 0) 414 continue; 415 if ((rt->rt_expire - time_second) < arpt_refresh && 416 rt->rt_pksent > (time_second - arpt_keep)) { 417 /* 418 * If the entry has been used during since last 419 * refresh, try to renew it before deleting. 420 */ 421 arprequest(rt->rt_ifp, 422 &satocsin(rt->rt_ifa->ifa_addr)->sin_addr, 423 &satocsin(rt_getkey(rt))->sin_addr, 424 CLLADDR(rt->rt_ifp->if_sadl)); 425 } else if (rt->rt_expire <= time_second) 426 arptfree(la); /* timer has expired; clear */ 427 } 428 429 ARP_UNLOCK(); 430 431 KERNEL_UNLOCK_ONE(NULL); 432 mutex_exit(softnet_lock); 433} 434 435/* 436 * We set the gateway for RTF_CLONING routes to a "prototype" 437 * link-layer sockaddr whose interface type (if_type) and interface 438 * index (if_index) fields are prepared. 439 */ 440static struct sockaddr * 441arp_setgate(struct rtentry *rt, struct sockaddr *gate, 442 const struct sockaddr *netmask) 443{ 444 const struct ifnet *ifp = rt->rt_ifp; 445 uint8_t namelen = strlen(ifp->if_xname); 446 uint8_t addrlen = ifp->if_addrlen; 447 448 /* 449 * XXX: If this is a manually added route to interface 450 * such as older version of routed or gated might provide, 451 * restore cloning bit. 452 */ 453 if ((rt->rt_flags & RTF_HOST) == 0 && netmask != NULL && 454 satocsin(netmask)->sin_addr.s_addr != 0xffffffff) 455 rt->rt_flags |= RTF_CLONING; 456 if (rt->rt_flags & RTF_CLONING) { 457 union { 458 struct sockaddr sa; 459 struct sockaddr_storage ss; 460 struct sockaddr_dl sdl; 461 } u; 462 /* 463 * Case 1: This route should come from a route to iface. 464 */ 465 sockaddr_dl_init(&u.sdl, sizeof(u.ss), 466 ifp->if_index, ifp->if_type, NULL, namelen, NULL, addrlen); 467 rt_setgate(rt, &u.sa); 468 gate = rt->rt_gateway; 469 } 470 return gate; 471} 472 473/* 474 * Parallel to llc_rtrequest. 475 */ 476void 477arp_rtrequest(int req, struct rtentry *rt, const struct rt_addrinfo *info) 478{ 479 struct sockaddr *gate = rt->rt_gateway; 480 struct llinfo_arp *la = (struct llinfo_arp *)rt->rt_llinfo; 481 size_t allocsize; 482 struct mbuf *mold; 483 int s; 484 struct in_ifaddr *ia; 485 struct ifaddr *ifa; 486 struct ifnet *ifp = rt->rt_ifp; 487 488 if (!arpinit_done) { 489 arpinit_done = 1; 490 /* 491 * We generate expiration times from time_second 492 * so avoid accidentally creating permanent routes. 493 */ 494 if (time_second == 0) { 495 struct timespec ts; 496 ts.tv_sec = 1; 497 ts.tv_nsec = 0; 498 tc_setclock(&ts); 499 } 500 callout_init(&arptimer_ch, CALLOUT_MPSAFE); 501 callout_reset(&arptimer_ch, hz, arptimer, NULL); 502 } 503 504 if (req == RTM_LLINFO_UPD) { 505 struct in_addr *in; 506 507 if ((ifa = info->rti_ifa) == NULL) 508 return; 509 510 in = &ifatoia(ifa)->ia_addr.sin_addr; 511 512 arprequest(ifa->ifa_ifp, in, in, 513 CLLADDR(ifa->ifa_ifp->if_sadl)); 514 return; 515 } 516 517 if ((rt->rt_flags & RTF_GATEWAY) != 0) { 518 if (req != RTM_ADD) 519 return; 520 521 /* 522 * linklayers with particular link MTU limitation. 523 */ 524 switch(ifp->if_type) { 525#if NFDDI > 0 526 case IFT_FDDI: 527 if (ifp->if_mtu > FDDIIPMTU) 528 rt->rt_rmx.rmx_mtu = FDDIIPMTU; 529 break; 530#endif 531#if NARC > 0 532 case IFT_ARCNET: 533 { 534 int arcipifmtu; 535 536 if (ifp->if_flags & IFF_LINK0) 537 arcipifmtu = arc_ipmtu; 538 else 539 arcipifmtu = ARCMTU; 540 if (ifp->if_mtu > arcipifmtu) 541 rt->rt_rmx.rmx_mtu = arcipifmtu; 542 break; 543 } 544#endif 545 } 546 return; 547 } 548 549 ARP_LOCK(1); /* we may already be locked here. */ 550 551 switch (req) { 552 case RTM_SETGATE: 553 gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]); 554 break; 555 case RTM_ADD: 556 gate = arp_setgate(rt, gate, info->rti_info[RTAX_NETMASK]); 557 if (rt->rt_flags & RTF_CLONING) { 558 /* 559 * Give this route an expiration time, even though 560 * it's a "permanent" route, so that routes cloned 561 * from it do not need their expiration time set. 562 */ 563 rt->rt_expire = time_second; 564 /* 565 * linklayers with particular link MTU limitation. 566 */ 567 switch (ifp->if_type) { 568#if NFDDI > 0 569 case IFT_FDDI: 570 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 && 571 (rt->rt_rmx.rmx_mtu > FDDIIPMTU || 572 (rt->rt_rmx.rmx_mtu == 0 && 573 ifp->if_mtu > FDDIIPMTU))) 574 rt->rt_rmx.rmx_mtu = FDDIIPMTU; 575 break; 576#endif 577#if NARC > 0 578 case IFT_ARCNET: 579 { 580 int arcipifmtu; 581 if (ifp->if_flags & IFF_LINK0) 582 arcipifmtu = arc_ipmtu; 583 else 584 arcipifmtu = ARCMTU; 585 586 if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 && 587 (rt->rt_rmx.rmx_mtu > arcipifmtu || 588 (rt->rt_rmx.rmx_mtu == 0 && 589 ifp->if_mtu > arcipifmtu))) 590 rt->rt_rmx.rmx_mtu = arcipifmtu; 591 break; 592 } 593#endif 594 } 595 break; 596 } 597 /* Announce a new entry if requested. */ 598 if (rt->rt_flags & RTF_ANNOUNCE) { 599 arprequest(ifp, 600 &satocsin(rt_getkey(rt))->sin_addr, 601 &satocsin(rt_getkey(rt))->sin_addr, 602 CLLADDR(satocsdl(gate))); 603 } 604 /*FALLTHROUGH*/ 605 case RTM_RESOLVE: 606 if (gate->sa_family != AF_LINK || 607 gate->sa_len < sockaddr_dl_measure(0, ifp->if_addrlen)) { 608 log(LOG_DEBUG, "arp_rtrequest: bad gateway value\n"); 609 break; 610 } 611 satosdl(gate)->sdl_type = ifp->if_type; 612 satosdl(gate)->sdl_index = ifp->if_index; 613 if (la != NULL) 614 break; /* This happens on a route change */ 615 /* 616 * Case 2: This route may come from cloning, or a manual route 617 * add with a LL address. 618 */ 619 switch (ifp->if_type) { 620#if NTOKEN > 0 621 case IFT_ISO88025: 622 allocsize = sizeof(*la) + sizeof(struct token_rif); 623 break; 624#endif /* NTOKEN > 0 */ 625 default: 626 allocsize = sizeof(*la); 627 } 628 R_Malloc(la, struct llinfo_arp *, allocsize); 629 rt->rt_llinfo = (void *)la; 630 if (la == NULL) { 631 log(LOG_DEBUG, "arp_rtrequest: malloc failed\n"); 632 break; 633 } 634 arp_inuse++, arp_allocated++; 635 memset(la, 0, allocsize); 636 la->la_rt = rt; 637 rt->rt_flags |= RTF_LLINFO; 638 LIST_INSERT_HEAD(&llinfo_arp, la, la_list); 639 640 INADDR_TO_IA(satocsin(rt_getkey(rt))->sin_addr, ia); 641 while (ia && ia->ia_ifp != ifp) 642 NEXT_IA_WITH_SAME_ADDR(ia); 643 if (ia) { 644 /* 645 * This test used to be 646 * if (lo0ifp->if_flags & IFF_UP) 647 * It allowed local traffic to be forced through 648 * the hardware by configuring the loopback down. 649 * However, it causes problems during network 650 * configuration for boards that can't receive 651 * packets they send. It is now necessary to clear 652 * "useloopback" and remove the route to force 653 * traffic out to the hardware. 654 * 655 * In 4.4BSD, the above "if" statement checked 656 * rt->rt_ifa against rt_getkey(rt). It was changed 657 * to the current form so that we can provide a 658 * better support for multiple IPv4 addresses on a 659 * interface. 660 */ 661 rt->rt_expire = 0; 662 if (sockaddr_dl_init(satosdl(gate), gate->sa_len, 663 ifp->if_index, ifp->if_type, NULL, 0, 664 CLLADDR(ifp->if_sadl), ifp->if_addrlen) == NULL) { 665 panic("%s(%s): sockaddr_dl_init cannot fail", 666 __func__, ifp->if_xname); 667 } 668 if (useloopback) 669 ifp = rt->rt_ifp = lo0ifp; 670 /* 671 * make sure to set rt->rt_ifa to the interface 672 * address we are using, otherwise we will have trouble 673 * with source address selection. 674 */ 675 ifa = &ia->ia_ifa; 676 if (ifa != rt->rt_ifa) 677 rt_replace_ifa(rt, ifa); 678 } 679 break; 680 681 case RTM_DELETE: 682 if (la == NULL) 683 break; 684 arp_inuse--; 685 LIST_REMOVE(la, la_list); 686 rt->rt_llinfo = NULL; 687 rt->rt_flags &= ~RTF_LLINFO; 688 689 s = splnet(); 690 mold = la->la_hold; 691 la->la_hold = 0; 692 splx(s); 693 694 if (mold) 695 m_freem(mold); 696 697 Free((void *)la); 698 } 699 ARP_UNLOCK(); 700} 701 702/* 703 * Broadcast an ARP request. Caller specifies: 704 * - arp header source ip address 705 * - arp header target ip address 706 * - arp header source ethernet address 707 */ 708void 709arprequest(struct ifnet *ifp, 710 const struct in_addr *sip, const struct in_addr *tip, 711 const u_int8_t *enaddr) 712{ 713 struct mbuf *m; 714 struct arphdr *ah; 715 struct sockaddr sa; 716 uint64_t *arps; 717 718 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 719 return; 720 MCLAIM(m, &arpdomain.dom_mowner); 721 switch (ifp->if_type) { 722 case IFT_IEEE1394: 723 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 724 ifp->if_addrlen; 725 break; 726 default: 727 m->m_len = sizeof(*ah) + 2 * sizeof(struct in_addr) + 728 2 * ifp->if_addrlen; 729 break; 730 } 731 m->m_pkthdr.len = m->m_len; 732 MH_ALIGN(m, m->m_len); 733 ah = mtod(m, struct arphdr *); 734 memset(ah, 0, m->m_len); 735 switch (ifp->if_type) { 736 case IFT_IEEE1394: /* RFC2734 */ 737 /* fill it now for ar_tpa computation */ 738 ah->ar_hrd = htons(ARPHRD_IEEE1394); 739 break; 740 default: 741 /* ifp->if_output will fill ar_hrd */ 742 break; 743 } 744 ah->ar_pro = htons(ETHERTYPE_IP); 745 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 746 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 747 ah->ar_op = htons(ARPOP_REQUEST); 748 memcpy(ar_sha(ah), enaddr, ah->ar_hln); 749 memcpy(ar_spa(ah), sip, ah->ar_pln); 750 memcpy(ar_tpa(ah), tip, ah->ar_pln); 751 sa.sa_family = AF_ARP; 752 sa.sa_len = 2; 753 m->m_flags |= M_BCAST; 754 arps = ARP_STAT_GETREF(); 755 arps[ARP_STAT_SNDTOTAL]++; 756 arps[ARP_STAT_SENDREQUEST]++; 757 ARP_STAT_PUTREF(); 758 (*ifp->if_output)(ifp, m, &sa, NULL); 759} 760 761/* 762 * Resolve an IP address into an ethernet address. If success, 763 * desten is filled in. If there is no entry in arptab, 764 * set one up and broadcast a request for the IP address. 765 * Hold onto this mbuf and resend it once the address 766 * is finally resolved. A return value of 1 indicates 767 * that desten has been filled in and the packet should be sent 768 * normally; a 0 return indicates that the packet has been 769 * taken over here, either now or for later transmission. 770 */ 771int 772arpresolve(struct ifnet *ifp, struct rtentry *rt, struct mbuf *m, 773 const struct sockaddr *dst, u_char *desten) 774{ 775 struct llinfo_arp *la; 776 const struct sockaddr_dl *sdl; 777 struct mbuf *mold; 778 int s; 779 780 if ((la = arplookup1(m, &satocsin(dst)->sin_addr, 1, 0, rt)) != NULL) 781 rt = la->la_rt; 782 783 if (la == NULL || rt == NULL) { 784 ARP_STATINC(ARP_STAT_ALLOCFAIL); 785 log(LOG_DEBUG, 786 "arpresolve: can't allocate llinfo on %s for %s\n", 787 ifp->if_xname, in_fmtaddr(satocsin(dst)->sin_addr)); 788 m_freem(m); 789 return 0; 790 } 791 sdl = satocsdl(rt->rt_gateway); 792 /* 793 * Check the address family and length is valid, the address 794 * is resolved; otherwise, try to resolve. 795 */ 796 if ((rt->rt_expire == 0 || rt->rt_expire > time_second) && 797 sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { 798 memcpy(desten, CLLADDR(sdl), 799 min(sdl->sdl_alen, ifp->if_addrlen)); 800 rt->rt_pksent = time_second; /* Time for last pkt sent */ 801 return 1; 802 } 803 /* 804 * There is an arptab entry, but no ethernet address 805 * response yet. Replace the held mbuf with this 806 * latest one. 807 */ 808 809 ARP_STATINC(ARP_STAT_DFRTOTAL); 810 s = splnet(); 811 mold = la->la_hold; 812 la->la_hold = m; 813 splx(s); 814 815 if (mold) { 816 ARP_STATINC(ARP_STAT_DFRDROPPED); 817 m_freem(mold); 818 } 819 820 /* 821 * Re-send the ARP request when appropriate. 822 */ 823#ifdef DIAGNOSTIC 824 if (rt->rt_expire == 0) { 825 /* This should never happen. (Should it? -gwr) */ 826 printf("arpresolve: unresolved and rt_expire == 0\n"); 827 /* Set expiration time to now (expired). */ 828 rt->rt_expire = time_second; 829 } 830#endif 831 if (rt->rt_expire) { 832 rt->rt_flags &= ~RTF_REJECT; 833 if (la->la_asked == 0 || rt->rt_expire != time_second) { 834 rt->rt_expire = time_second; 835 if (la->la_asked++ < arp_maxtries) { 836 arprequest(ifp, 837 &satocsin(rt->rt_ifa->ifa_addr)->sin_addr, 838 &satocsin(dst)->sin_addr, 839#if NCARP > 0 840 (rt->rt_ifp->if_type == IFT_CARP) ? 841 CLLADDR(rt->rt_ifp->if_sadl): 842#endif 843 CLLADDR(ifp->if_sadl)); 844 } else { 845 rt->rt_flags |= RTF_REJECT; 846 rt->rt_expire += arpt_down; 847 la->la_asked = 0; 848 } 849 } 850 } 851 return 0; 852} 853 854/* 855 * Common length and type checks are done here, 856 * then the protocol-specific routine is called. 857 */ 858void 859arpintr(void) 860{ 861 struct mbuf *m; 862 struct arphdr *ar; 863 int s; 864 int arplen; 865 866 mutex_enter(softnet_lock); 867 KERNEL_LOCK(1, NULL); 868 while (arpintrq.ifq_head) { 869 s = splnet(); 870 IF_DEQUEUE(&arpintrq, m); 871 splx(s); 872 if (m == 0 || (m->m_flags & M_PKTHDR) == 0) 873 panic("arpintr"); 874 875 MCLAIM(m, &arpdomain.dom_mowner); 876 ARP_STATINC(ARP_STAT_RCVTOTAL); 877 878 /* 879 * First, make sure we have at least struct arphdr. 880 */ 881 if (m->m_len < sizeof(struct arphdr) || 882 (ar = mtod(m, struct arphdr *)) == NULL) 883 goto badlen; 884 885 switch (m->m_pkthdr.rcvif->if_type) { 886 case IFT_IEEE1394: 887 arplen = sizeof(struct arphdr) + 888 ar->ar_hln + 2 * ar->ar_pln; 889 break; 890 default: 891 arplen = sizeof(struct arphdr) + 892 2 * ar->ar_hln + 2 * ar->ar_pln; 893 break; 894 } 895 896 if (/* XXX ntohs(ar->ar_hrd) == ARPHRD_ETHER && */ 897 m->m_len >= arplen) 898 switch (ntohs(ar->ar_pro)) { 899 case ETHERTYPE_IP: 900 case ETHERTYPE_IPTRAILERS: 901 in_arpinput(m); 902 continue; 903 default: 904 ARP_STATINC(ARP_STAT_RCVBADPROTO); 905 } 906 else { 907badlen: 908 ARP_STATINC(ARP_STAT_RCVBADLEN); 909 } 910 m_freem(m); 911 } 912 KERNEL_UNLOCK_ONE(NULL); 913 mutex_exit(softnet_lock); 914} 915 916/* 917 * ARP for Internet protocols on 10 Mb/s Ethernet. 918 * Algorithm is that given in RFC 826. 919 * In addition, a sanity check is performed on the sender 920 * protocol address, to catch impersonators. 921 * We no longer handle negotiations for use of trailer protocol: 922 * Formerly, ARP replied for protocol type ETHERTYPE_TRAIL sent 923 * along with IP replies if we wanted trailers sent to us, 924 * and also sent them in response to IP replies. 925 * This allowed either end to announce the desire to receive 926 * trailer packets. 927 * We no longer reply to requests for ETHERTYPE_TRAIL protocol either, 928 * but formerly didn't normally send requests. 929 */ 930static void 931in_arpinput(struct mbuf *m) 932{ 933 struct arphdr *ah; 934 struct ifnet *ifp = m->m_pkthdr.rcvif; 935 struct llinfo_arp *la = NULL; 936 struct rtentry *rt; 937 struct in_ifaddr *ia; 938#if NBRIDGE > 0 939 struct in_ifaddr *bridge_ia = NULL; 940#endif 941#if NCARP > 0 942 u_int32_t count = 0, index = 0; 943#endif 944 struct sockaddr_dl *sdl; 945 struct sockaddr sa; 946 struct in_addr isaddr, itaddr, myaddr; 947 int op; 948 struct mbuf *mold; 949 void *tha; 950 int s; 951 uint64_t *arps; 952 953 if (__predict_false(m_makewritable(&m, 0, m->m_pkthdr.len, M_DONTWAIT))) 954 goto out; 955 ah = mtod(m, struct arphdr *); 956 op = ntohs(ah->ar_op); 957 958 /* 959 * Fix up ah->ar_hrd if necessary, before using ar_tha() or 960 * ar_tpa(). 961 */ 962 switch (ifp->if_type) { 963 case IFT_IEEE1394: 964 if (ntohs(ah->ar_hrd) == ARPHRD_IEEE1394) 965 ; 966 else { 967 /* XXX this is to make sure we compute ar_tha right */ 968 /* XXX check ar_hrd more strictly? */ 969 ah->ar_hrd = htons(ARPHRD_IEEE1394); 970 } 971 break; 972 default: 973 /* XXX check ar_hrd? */ 974 break; 975 } 976 977 memcpy(&isaddr, ar_spa(ah), sizeof (isaddr)); 978 memcpy(&itaddr, ar_tpa(ah), sizeof (itaddr)); 979 980 if (m->m_flags & (M_BCAST|M_MCAST)) 981 ARP_STATINC(ARP_STAT_RCVMCAST); 982 983 /* 984 * If the target IP address is zero, ignore the packet. 985 * This prevents the code below from tring to answer 986 * when we are using IP address zero (booting). 987 */ 988 if (in_nullhost(itaddr)) { 989 ARP_STATINC(ARP_STAT_RCVZEROTPA); 990 goto out; 991 } 992 993 994 /* 995 * Search for a matching interface address 996 * or any address on the interface to use 997 * as a dummy address in the rest of this function 998 */ 999 1000 INADDR_TO_IA(itaddr, ia); 1001 while (ia != NULL) { 1002#if NCARP > 0 1003 if (ia->ia_ifp->if_type == IFT_CARP && 1004 ((ia->ia_ifp->if_flags & (IFF_UP|IFF_RUNNING)) == 1005 (IFF_UP|IFF_RUNNING))) { 1006 index++; 1007 if (ia->ia_ifp == m->m_pkthdr.rcvif && 1008 carp_iamatch(ia, ar_sha(ah), 1009 &count, index)) { 1010 break; 1011 } 1012 } else 1013#endif 1014 if (ia->ia_ifp == m->m_pkthdr.rcvif) 1015 break; 1016#if NBRIDGE > 0 1017 /* 1018 * If the interface we received the packet on 1019 * is part of a bridge, check to see if we need 1020 * to "bridge" the packet to ourselves at this 1021 * layer. Note we still prefer a perfect match, 1022 * but allow this weaker match if necessary. 1023 */ 1024 if (m->m_pkthdr.rcvif->if_bridge != NULL && 1025 m->m_pkthdr.rcvif->if_bridge == ia->ia_ifp->if_bridge) 1026 bridge_ia = ia; 1027#endif /* NBRIDGE > 0 */ 1028 1029 NEXT_IA_WITH_SAME_ADDR(ia); 1030 } 1031 1032#if NBRIDGE > 0 1033 if (ia == NULL && bridge_ia != NULL) { 1034 ia = bridge_ia; 1035 ifp = bridge_ia->ia_ifp; 1036 } 1037#endif 1038 1039 if (ia == NULL) { 1040 INADDR_TO_IA(isaddr, ia); 1041 while ((ia != NULL) && ia->ia_ifp != m->m_pkthdr.rcvif) 1042 NEXT_IA_WITH_SAME_ADDR(ia); 1043 1044 if (ia == NULL) { 1045 IFP_TO_IA(ifp, ia); 1046 if (ia == NULL) { 1047 ARP_STATINC(ARP_STAT_RCVNOINT); 1048 goto out; 1049 } 1050 } 1051 } 1052 1053 myaddr = ia->ia_addr.sin_addr; 1054 1055 /* XXX checks for bridge case? */ 1056 if (!memcmp(ar_sha(ah), CLLADDR(ifp->if_sadl), ifp->if_addrlen)) { 1057 ARP_STATINC(ARP_STAT_RCVLOCALSHA); 1058 goto out; /* it's from me, ignore it. */ 1059 } 1060 1061 /* XXX checks for bridge case? */ 1062 if (!memcmp(ar_sha(ah), ifp->if_broadcastaddr, ifp->if_addrlen)) { 1063 ARP_STATINC(ARP_STAT_RCVBCASTSHA); 1064 log(LOG_ERR, 1065 "%s: arp: link address is broadcast for IP address %s!\n", 1066 ifp->if_xname, in_fmtaddr(isaddr)); 1067 goto out; 1068 } 1069 1070 /* 1071 * If the source IP address is zero, this is an RFC 5227 ARP probe 1072 */ 1073 if (in_nullhost(isaddr)) { 1074 ARP_STATINC(ARP_STAT_RCVZEROSPA); 1075 goto reply; 1076 } 1077 1078 if (in_hosteq(isaddr, myaddr)) { 1079 ARP_STATINC(ARP_STAT_RCVLOCALSPA); 1080 log(LOG_ERR, 1081 "duplicate IP address %s sent from link address %s\n", 1082 in_fmtaddr(isaddr), lla_snprintf(ar_sha(ah), ah->ar_hln)); 1083 itaddr = myaddr; 1084 goto reply; 1085 } 1086 la = arplookup(m, &isaddr, in_hosteq(itaddr, myaddr), 0); 1087 if (la != NULL && (rt = la->la_rt) && (sdl = satosdl(rt->rt_gateway))) { 1088 if (sdl->sdl_alen && 1089 memcmp(ar_sha(ah), CLLADDR(sdl), sdl->sdl_alen)) { 1090 if (rt->rt_flags & RTF_STATIC) { 1091 ARP_STATINC(ARP_STAT_RCVOVERPERM); 1092 if (!log_permanent_modify) 1093 goto out; 1094 log(LOG_INFO, 1095 "%s tried to overwrite permanent arp info" 1096 " for %s\n", 1097 lla_snprintf(ar_sha(ah), ah->ar_hln), 1098 in_fmtaddr(isaddr)); 1099 goto out; 1100 } else if (rt->rt_ifp != ifp) { 1101 ARP_STATINC(ARP_STAT_RCVOVERINT); 1102 if (!log_wrong_iface) 1103 goto out; 1104 log(LOG_INFO, 1105 "%s on %s tried to overwrite " 1106 "arp info for %s on %s\n", 1107 lla_snprintf(ar_sha(ah), ah->ar_hln), 1108 ifp->if_xname, in_fmtaddr(isaddr), 1109 rt->rt_ifp->if_xname); 1110 goto out; 1111 } else { 1112 ARP_STATINC(ARP_STAT_RCVOVER); 1113 if (log_movements) 1114 log(LOG_INFO, "arp info overwritten " 1115 "for %s by %s\n", 1116 in_fmtaddr(isaddr), 1117 lla_snprintf(ar_sha(ah), 1118 ah->ar_hln)); 1119 } 1120 } 1121 /* 1122 * sanity check for the address length. 1123 * XXX this does not work for protocols with variable address 1124 * length. -is 1125 */ 1126 if (sdl->sdl_alen && 1127 sdl->sdl_alen != ah->ar_hln) { 1128 ARP_STATINC(ARP_STAT_RCVLENCHG); 1129 log(LOG_WARNING, 1130 "arp from %s: new addr len %d, was %d\n", 1131 in_fmtaddr(isaddr), ah->ar_hln, sdl->sdl_alen); 1132 } 1133 if (ifp->if_addrlen != ah->ar_hln) { 1134 ARP_STATINC(ARP_STAT_RCVBADLEN); 1135 log(LOG_WARNING, 1136 "arp from %s: addr len: new %d, i/f %d (ignored)\n", 1137 in_fmtaddr(isaddr), ah->ar_hln, 1138 ifp->if_addrlen); 1139 goto reply; 1140 } 1141#if NTOKEN > 0 1142 /* 1143 * XXX uses m_data and assumes the complete answer including 1144 * XXX token-ring headers is in the same buf 1145 */ 1146 if (ifp->if_type == IFT_ISO88025) { 1147 struct token_header *trh; 1148 1149 trh = (struct token_header *)M_TRHSTART(m); 1150 if (trh->token_shost[0] & TOKEN_RI_PRESENT) { 1151 struct token_rif *rif; 1152 size_t riflen; 1153 1154 rif = TOKEN_RIF(trh); 1155 riflen = (ntohs(rif->tr_rcf) & 1156 TOKEN_RCF_LEN_MASK) >> 8; 1157 1158 if (riflen > 2 && 1159 riflen < sizeof(struct token_rif) && 1160 (riflen & 1) == 0) { 1161 rif->tr_rcf ^= htons(TOKEN_RCF_DIRECTION); 1162 rif->tr_rcf &= htons(~TOKEN_RCF_BROADCAST_MASK); 1163 memcpy(TOKEN_RIF(la), rif, riflen); 1164 } 1165 } 1166 } 1167#endif /* NTOKEN > 0 */ 1168 (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, ar_sha(ah), 1169 ah->ar_hln); 1170 if (rt->rt_expire) 1171 rt->rt_expire = time_second + arpt_keep; 1172 rt->rt_flags &= ~RTF_REJECT; 1173 la->la_asked = 0; 1174 1175 s = splnet(); 1176 mold = la->la_hold; 1177 la->la_hold = 0; 1178 splx(s); 1179 1180 if (mold) { 1181 ARP_STATINC(ARP_STAT_DFRSENT); 1182 (*ifp->if_output)(ifp, mold, rt_getkey(rt), rt); 1183 } 1184 } 1185reply: 1186 if (op != ARPOP_REQUEST) { 1187 if (op == ARPOP_REPLY) 1188 ARP_STATINC(ARP_STAT_RCVREPLY); 1189 out: 1190 m_freem(m); 1191 return; 1192 } 1193 ARP_STATINC(ARP_STAT_RCVREQUEST); 1194 if (in_hosteq(itaddr, myaddr)) { 1195 /* I am the target */ 1196 tha = ar_tha(ah); 1197 if (tha) 1198 memcpy(tha, ar_sha(ah), ah->ar_hln); 1199 memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln); 1200 } else { 1201 la = arplookup(m, &itaddr, 0, SIN_PROXY); 1202 if (la == NULL) 1203 goto out; 1204 rt = la->la_rt; 1205 if (rt->rt_ifp->if_type == IFT_CARP && 1206 m->m_pkthdr.rcvif->if_type != IFT_CARP) 1207 goto out; 1208 tha = ar_tha(ah); 1209 if (tha) 1210 memcpy(tha, ar_sha(ah), ah->ar_hln); 1211 sdl = satosdl(rt->rt_gateway); 1212 memcpy(ar_sha(ah), CLLADDR(sdl), ah->ar_hln); 1213 } 1214 1215 memcpy(ar_tpa(ah), ar_spa(ah), ah->ar_pln); 1216 memcpy(ar_spa(ah), &itaddr, ah->ar_pln); 1217 ah->ar_op = htons(ARPOP_REPLY); 1218 ah->ar_pro = htons(ETHERTYPE_IP); /* let's be sure! */ 1219 switch (ifp->if_type) { 1220 case IFT_IEEE1394: 1221 /* 1222 * ieee1394 arp reply is broadcast 1223 */ 1224 m->m_flags &= ~M_MCAST; 1225 m->m_flags |= M_BCAST; 1226 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + ah->ar_hln; 1227 break; 1228 1229 default: 1230 m->m_flags &= ~(M_BCAST|M_MCAST); /* never reply by broadcast */ 1231 m->m_len = sizeof(*ah) + (2 * ah->ar_pln) + (2 * ah->ar_hln); 1232 break; 1233 } 1234 m->m_pkthdr.len = m->m_len; 1235 sa.sa_family = AF_ARP; 1236 sa.sa_len = 2; 1237 arps = ARP_STAT_GETREF(); 1238 arps[ARP_STAT_SNDTOTAL]++; 1239 arps[ARP_STAT_SNDREPLY]++; 1240 ARP_STAT_PUTREF(); 1241 (*ifp->if_output)(ifp, m, &sa, NULL); 1242 return; 1243} 1244 1245/* 1246 * Free an arp entry. 1247 */ 1248static void arptfree(struct llinfo_arp *la) 1249{ 1250 struct rtentry *rt = la->la_rt; 1251 struct sockaddr_dl *sdl; 1252 1253 ARP_LOCK_CHECK(); 1254 1255 if (rt == NULL) 1256 panic("arptfree"); 1257 if (rt->rt_refcnt > 0 && (sdl = satosdl(rt->rt_gateway)) && 1258 sdl->sdl_family == AF_LINK) { 1259 sdl->sdl_alen = 0; 1260 la->la_asked = 0; 1261 rt->rt_flags &= ~RTF_REJECT; 1262 return; 1263 } 1264 rtrequest(RTM_DELETE, rt_getkey(rt), NULL, rt_mask(rt), 0, NULL); 1265} 1266 1267static struct llinfo_arp * 1268arplookup(struct mbuf *m, const struct in_addr *addr, int create, int proxy) 1269{ 1270 return arplookup1(m, addr, create, proxy, NULL); 1271} 1272 1273/* 1274 * Lookup or enter a new address in arptab. 1275 */ 1276static struct llinfo_arp * 1277arplookup1(struct mbuf *m, const struct in_addr *addr, int create, int proxy, 1278 struct rtentry *rt0) 1279{ 1280 struct arphdr *ah; 1281 struct ifnet *ifp = m->m_pkthdr.rcvif; 1282 struct rtentry *rt; 1283 struct sockaddr_inarp sin; 1284 const char *why = NULL; 1285 1286 ah = mtod(m, struct arphdr *); 1287 if (rt0 == NULL) { 1288 memset(&sin, 0, sizeof(sin)); 1289 sin.sin_len = sizeof(sin); 1290 sin.sin_family = AF_INET; 1291 sin.sin_addr = *addr; 1292 sin.sin_other = proxy ? SIN_PROXY : 0; 1293 rt = rtalloc1(sintosa(&sin), create); 1294 if (rt == NULL) 1295 return NULL; 1296 rt->rt_refcnt--; 1297 } else 1298 rt = rt0; 1299 1300#define IS_LLINFO(__rt) \ 1301 (((__rt)->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) == RTF_LLINFO && \ 1302 (__rt)->rt_gateway->sa_family == AF_LINK) 1303 1304 1305 if (IS_LLINFO(rt)) 1306 return (struct llinfo_arp *)rt->rt_llinfo; 1307 1308 if (create) { 1309 if (rt->rt_flags & RTF_GATEWAY) 1310 why = "host is not on local network"; 1311 else if ((rt->rt_flags & RTF_LLINFO) == 0) { 1312 ARP_STATINC(ARP_STAT_ALLOCFAIL); 1313 why = "could not allocate llinfo"; 1314 } else 1315 why = "gateway route is not ours"; 1316 log(LOG_DEBUG, "arplookup: unable to enter address" 1317 " for %s@%s on %s (%s)\n", 1318 in_fmtaddr(*addr), lla_snprintf(ar_sha(ah), ah->ar_hln), 1319 (ifp) ? ifp->if_xname : "null", why); 1320 if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_CLONED) != 0) { 1321 rtrequest(RTM_DELETE, rt_getkey(rt), 1322 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL); 1323 } 1324 } 1325 return NULL; 1326} 1327 1328int 1329arpioctl(u_long cmd, void *data) 1330{ 1331 1332 return EOPNOTSUPP; 1333} 1334 1335void 1336arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) 1337{ 1338 struct in_addr *ip; 1339 1340 /* 1341 * Warn the user if another station has this IP address, 1342 * but only if the interface IP address is not zero. 1343 */ 1344 ip = &IA_SIN(ifa)->sin_addr; 1345 if (!in_nullhost(*ip)) 1346 arprequest(ifp, ip, ip, CLLADDR(ifp->if_sadl)); 1347 1348 ifa->ifa_rtrequest = arp_rtrequest; 1349 ifa->ifa_flags |= RTF_CLONING; 1350} 1351 1352/* 1353 * Called from 10 Mb/s Ethernet interrupt handlers 1354 * when ether packet type ETHERTYPE_REVARP 1355 * is received. Common length and type checks are done here, 1356 * then the protocol-specific routine is called. 1357 */ 1358void 1359revarpinput(struct mbuf *m) 1360{ 1361 struct arphdr *ar; 1362 1363 if (m->m_len < sizeof(struct arphdr)) 1364 goto out; 1365 ar = mtod(m, struct arphdr *); 1366#if 0 /* XXX I don't think we need this... and it will prevent other LL */ 1367 if (ntohs(ar->ar_hrd) != ARPHRD_ETHER) 1368 goto out; 1369#endif 1370 if (m->m_len < sizeof(struct arphdr) + 2 * (ar->ar_hln + ar->ar_pln)) 1371 goto out; 1372 switch (ntohs(ar->ar_pro)) { 1373 case ETHERTYPE_IP: 1374 case ETHERTYPE_IPTRAILERS: 1375 in_revarpinput(m); 1376 return; 1377 1378 default: 1379 break; 1380 } 1381out: 1382 m_freem(m); 1383} 1384 1385/* 1386 * RARP for Internet protocols on 10 Mb/s Ethernet. 1387 * Algorithm is that given in RFC 903. 1388 * We are only using for bootstrap purposes to get an ip address for one of 1389 * our interfaces. Thus we support no user-interface. 1390 * 1391 * Since the contents of the RARP reply are specific to the interface that 1392 * sent the request, this code must ensure that they are properly associated. 1393 * 1394 * Note: also supports ARP via RARP packets, per the RFC. 1395 */ 1396void 1397in_revarpinput(struct mbuf *m) 1398{ 1399 struct ifnet *ifp; 1400 struct arphdr *ah; 1401 void *tha; 1402 int op; 1403 1404 ah = mtod(m, struct arphdr *); 1405 op = ntohs(ah->ar_op); 1406 1407 switch (m->m_pkthdr.rcvif->if_type) { 1408 case IFT_IEEE1394: 1409 /* ARP without target hardware address is not supported */ 1410 goto out; 1411 default: 1412 break; 1413 } 1414 1415 switch (op) { 1416 case ARPOP_REQUEST: 1417 case ARPOP_REPLY: /* per RFC */ 1418 in_arpinput(m); 1419 return; 1420 case ARPOP_REVREPLY: 1421 break; 1422 case ARPOP_REVREQUEST: /* handled by rarpd(8) */ 1423 default: 1424 goto out; 1425 } 1426 if (!revarp_in_progress) 1427 goto out; 1428 ifp = m->m_pkthdr.rcvif; 1429 if (ifp != myip_ifp) /* !same interface */ 1430 goto out; 1431 if (myip_initialized) 1432 goto wake; 1433 tha = ar_tha(ah); 1434 if (tha == NULL) 1435 goto out; 1436 if (memcmp(tha, CLLADDR(ifp->if_sadl), ifp->if_sadl->sdl_alen)) 1437 goto out; 1438 memcpy(&srv_ip, ar_spa(ah), sizeof(srv_ip)); 1439 memcpy(&myip, ar_tpa(ah), sizeof(myip)); 1440 myip_initialized = 1; 1441wake: /* Do wakeup every time in case it was missed. */ 1442 wakeup((void *)&myip); 1443 1444out: 1445 m_freem(m); 1446} 1447 1448/* 1449 * Send a RARP request for the ip address of the specified interface. 1450 * The request should be RFC 903-compliant. 1451 */ 1452void 1453revarprequest(struct ifnet *ifp) 1454{ 1455 struct sockaddr sa; 1456 struct mbuf *m; 1457 struct arphdr *ah; 1458 void *tha; 1459 1460 if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) 1461 return; 1462 MCLAIM(m, &arpdomain.dom_mowner); 1463 m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + 1464 2*ifp->if_addrlen; 1465 m->m_pkthdr.len = m->m_len; 1466 MH_ALIGN(m, m->m_len); 1467 ah = mtod(m, struct arphdr *); 1468 memset(ah, 0, m->m_len); 1469 ah->ar_pro = htons(ETHERTYPE_IP); 1470 ah->ar_hln = ifp->if_addrlen; /* hardware address length */ 1471 ah->ar_pln = sizeof(struct in_addr); /* protocol address length */ 1472 ah->ar_op = htons(ARPOP_REVREQUEST); 1473 1474 memcpy(ar_sha(ah), CLLADDR(ifp->if_sadl), ah->ar_hln); 1475 tha = ar_tha(ah); 1476 if (tha == NULL) 1477 return; 1478 memcpy(tha, CLLADDR(ifp->if_sadl), ah->ar_hln); 1479 1480 sa.sa_family = AF_ARP; 1481 sa.sa_len = 2; 1482 m->m_flags |= M_BCAST; 1483 1484 KERNEL_LOCK(1, NULL); 1485 (*ifp->if_output)(ifp, m, &sa, NULL); 1486 KERNEL_UNLOCK_ONE(NULL); 1487} 1488 1489/* 1490 * RARP for the ip address of the specified interface, but also 1491 * save the ip address of the server that sent the answer. 1492 * Timeout if no response is received. 1493 */ 1494int 1495revarpwhoarewe(struct ifnet *ifp, struct in_addr *serv_in, 1496 struct in_addr *clnt_in) 1497{ 1498 int result, count = 20; 1499 1500 myip_initialized = 0; 1501 myip_ifp = ifp; 1502 1503 revarp_in_progress = 1; 1504 while (count--) { 1505 revarprequest(ifp); 1506 result = tsleep((void *)&myip, PSOCK, "revarp", hz/2); 1507 if (result != EWOULDBLOCK) 1508 break; 1509 } 1510 revarp_in_progress = 0; 1511 1512 if (!myip_initialized) 1513 return ENETUNREACH; 1514 1515 memcpy(serv_in, &srv_ip, sizeof(*serv_in)); 1516 memcpy(clnt_in, &myip, sizeof(*clnt_in)); 1517 return 0; 1518} 1519 1520 1521 1522#ifdef DDB 1523 1524#include <machine/db_machdep.h> 1525#include <ddb/db_interface.h> 1526#include <ddb/db_output.h> 1527 1528static void 1529db_print_sa(const struct sockaddr *sa) 1530{ 1531 int len; 1532 const u_char *p; 1533 1534 if (sa == NULL) { 1535 db_printf("[NULL]"); 1536 return; 1537 } 1538 1539 p = (const u_char *)sa; 1540 len = sa->sa_len; 1541 db_printf("["); 1542 while (len > 0) { 1543 db_printf("%d", *p); 1544 p++; len--; 1545 if (len) db_printf(","); 1546 } 1547 db_printf("]\n"); 1548} 1549 1550static void 1551db_print_ifa(struct ifaddr *ifa) 1552{ 1553 if (ifa == NULL) 1554 return; 1555 db_printf(" ifa_addr="); 1556 db_print_sa(ifa->ifa_addr); 1557 db_printf(" ifa_dsta="); 1558 db_print_sa(ifa->ifa_dstaddr); 1559 db_printf(" ifa_mask="); 1560 db_print_sa(ifa->ifa_netmask); 1561 db_printf(" flags=0x%x,refcnt=%d,metric=%d\n", 1562 ifa->ifa_flags, 1563 ifa->ifa_refcnt, 1564 ifa->ifa_metric); 1565} 1566 1567static void 1568db_print_llinfo(void *li) 1569{ 1570 struct llinfo_arp *la; 1571 1572 if (li == NULL) 1573 return; 1574 la = (struct llinfo_arp *)li; 1575 db_printf(" la_rt=%p la_hold=%p, la_asked=0x%lx\n", 1576 la->la_rt, la->la_hold, la->la_asked); 1577} 1578 1579/* 1580 * Function to pass to rt_walktree(). 1581 * Return non-zero error to abort walk. 1582 */ 1583static int 1584db_show_rtentry(struct rtentry *rt, void *w) 1585{ 1586 db_printf("rtentry=%p", rt); 1587 1588 db_printf(" flags=0x%x refcnt=%d use=%"PRId64" expire=%"PRId64"\n", 1589 rt->rt_flags, rt->rt_refcnt, 1590 rt->rt_use, (uint64_t)rt->rt_expire); 1591 1592 db_printf(" key="); db_print_sa(rt_getkey(rt)); 1593 db_printf(" mask="); db_print_sa(rt_mask(rt)); 1594 db_printf(" gw="); db_print_sa(rt->rt_gateway); 1595 1596 db_printf(" ifp=%p ", rt->rt_ifp); 1597 if (rt->rt_ifp) 1598 db_printf("(%s)", rt->rt_ifp->if_xname); 1599 else 1600 db_printf("(NULL)"); 1601 1602 db_printf(" ifa=%p\n", rt->rt_ifa); 1603 db_print_ifa(rt->rt_ifa); 1604 1605 db_printf(" gwroute=%p llinfo=%p\n", 1606 rt->rt_gwroute, rt->rt_llinfo); 1607 db_print_llinfo(rt->rt_llinfo); 1608 1609 return 0; 1610} 1611 1612/* 1613 * Function to print all the route trees. 1614 * Use this from ddb: "show arptab" 1615 */ 1616void 1617db_show_arptab(db_expr_t addr, bool have_addr, 1618 db_expr_t count, const char *modif) 1619{ 1620 rt_walktree(AF_INET, db_show_rtentry, NULL); 1621} 1622#endif 1623 1624static int 1625sysctl_net_inet_arp_stats(SYSCTLFN_ARGS) 1626{ 1627 1628 return NETSTAT_SYSCTL(arpstat_percpu, ARP_NSTATS); 1629} 1630 1631static void 1632sysctl_net_inet_arp_setup(struct sysctllog **clog) 1633{ 1634 const struct sysctlnode *node; 1635 1636 sysctl_createv(clog, 0, NULL, NULL, 1637 CTLFLAG_PERMANENT, 1638 CTLTYPE_NODE, "net", NULL, 1639 NULL, 0, NULL, 0, 1640 CTL_NET, CTL_EOL); 1641 sysctl_createv(clog, 0, NULL, NULL, 1642 CTLFLAG_PERMANENT, 1643 CTLTYPE_NODE, "inet", NULL, 1644 NULL, 0, NULL, 0, 1645 CTL_NET, PF_INET, CTL_EOL); 1646 sysctl_createv(clog, 0, NULL, &node, 1647 CTLFLAG_PERMANENT, 1648 CTLTYPE_NODE, "arp", 1649 SYSCTL_DESCR("Address Resolution Protocol"), 1650 NULL, 0, NULL, 0, 1651 CTL_NET, PF_INET, CTL_CREATE, CTL_EOL); 1652 1653 sysctl_createv(clog, 0, NULL, NULL, 1654 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1655 CTLTYPE_INT, "prune", 1656 SYSCTL_DESCR("ARP cache pruning interval in seconds"), 1657 NULL, 0, &arpt_prune, 0, 1658 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1659 1660 sysctl_createv(clog, 0, NULL, NULL, 1661 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1662 CTLTYPE_INT, "keep", 1663 SYSCTL_DESCR("Valid ARP entry lifetime in seconds"), 1664 NULL, 0, &arpt_keep, 0, 1665 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1666 1667 sysctl_createv(clog, 0, NULL, NULL, 1668 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1669 CTLTYPE_INT, "down", 1670 SYSCTL_DESCR("Failed ARP entry lifetime in seconds"), 1671 NULL, 0, &arpt_down, 0, 1672 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1673 1674 sysctl_createv(clog, 0, NULL, NULL, 1675 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1676 CTLTYPE_INT, "refresh", 1677 SYSCTL_DESCR("ARP entry refresh interval"), 1678 NULL, 0, &arpt_refresh, 0, 1679 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1680 1681 sysctl_createv(clog, 0, NULL, NULL, 1682 CTLFLAG_PERMANENT, 1683 CTLTYPE_STRUCT, "stats", 1684 SYSCTL_DESCR("ARP statistics"), 1685 sysctl_net_inet_arp_stats, 0, NULL, 0, 1686 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1687 1688 sysctl_createv(clog, 0, NULL, NULL, 1689 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1690 CTLTYPE_INT, "log_movements", 1691 SYSCTL_DESCR("log ARP replies from MACs different than" 1692 " the one in the cache"), 1693 NULL, 0, &log_movements, 0, 1694 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1695 1696 sysctl_createv(clog, 0, NULL, NULL, 1697 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1698 CTLTYPE_INT, "log_permanent_modify", 1699 SYSCTL_DESCR("log ARP replies from MACs different than" 1700 " the one in the permanent arp entry"), 1701 NULL, 0, &log_permanent_modify, 0, 1702 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1703 1704 sysctl_createv(clog, 0, NULL, NULL, 1705 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 1706 CTLTYPE_INT, "log_wrong_iface", 1707 SYSCTL_DESCR("log ARP packets arriving on the wrong" 1708 " interface"), 1709 NULL, 0, &log_wrong_iface, 0, 1710 CTL_NET,PF_INET, node->sysctl_num, CTL_CREATE, CTL_EOL); 1711} 1712 1713#endif /* INET */ 1714