nd6.c revision 194907
10Sduke/*- 211719Sgoetz * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 30Sduke * All rights reserved. 40Sduke * 50Sduke * Redistribution and use in source and binary forms, with or without 60Sduke * modification, are permitted provided that the following conditions 70Sduke * are met: 80Sduke * 1. Redistributions of source code must retain the above copyright 90Sduke * notice, this list of conditions and the following disclaimer. 100Sduke * 2. Redistributions in binary form must reproduce the above copyright 110Sduke * notice, this list of conditions and the following disclaimer in the 120Sduke * documentation and/or other materials provided with the distribution. 130Sduke * 3. Neither the name of the project nor the names of its contributors 140Sduke * may be used to endorse or promote products derived from this software 150Sduke * without specific prior written permission. 160Sduke * 170Sduke * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 180Sduke * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 191472Strims * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 201472Strims * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 211472Strims * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 220Sduke * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 230Sduke * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 240Sduke * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 251879Sstefank * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 261879Sstefank * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 271879Sstefank * SUCH DAMAGE. 281879Sstefank * 291879Sstefank * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 300Sduke */ 310Sduke 320Sduke#include <sys/cdefs.h> 330Sduke__FBSDID("$FreeBSD: head/sys/netinet6/nd6.c 194907 2009-06-24 21:00:25Z rwatson $"); 340Sduke 350Sduke#include "opt_inet.h" 360Sduke#include "opt_inet6.h" 370Sduke 380Sduke#include <sys/param.h> 390Sduke#include <sys/systm.h> 400Sduke#include <sys/callout.h> 410Sduke#include <sys/malloc.h> 420Sduke#include <sys/mbuf.h> 430Sduke#include <sys/socket.h> 440Sduke#include <sys/sockio.h> 450Sduke#include <sys/time.h> 460Sduke#include <sys/kernel.h> 470Sduke#include <sys/protosw.h> 488379Skvn#include <sys/errno.h> 498379Skvn#include <sys/syslog.h> 500Sduke#include <sys/lock.h> 510Sduke#include <sys/rwlock.h> 528379Skvn#include <sys/queue.h> 538379Skvn#include <sys/sysctl.h> 540Sduke#include <sys/vimage.h> 550Sduke 560Sduke#include <net/if.h> 570Sduke#include <net/if_arc.h> 580Sduke#include <net/if_dl.h> 590Sduke#include <net/if_types.h> 600Sduke#include <net/iso88025.h> 610Sduke#include <net/fddi.h> 620Sduke#include <net/route.h> 636760Sgoetz#include <net/vnet.h> 640Sduke 650Sduke#include <netinet/in.h> 660Sduke#include <net/if_llatbl.h> 670Sduke#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) 680Sduke#include <netinet/if_ether.h> 690Sduke#include <netinet6/in6_var.h> 700Sduke#include <netinet/ip6.h> 710Sduke#include <netinet6/ip6_var.h> 720Sduke#include <netinet6/scope6_var.h> 730Sduke#include <netinet6/nd6.h> 740Sduke#include <netinet/icmp6.h> 750Sduke#include <netinet6/vinet6.h> 760Sduke 770Sduke#include <sys/limits.h> 780Sduke 790Sduke#include <security/mac/mac_framework.h> 800Sduke 810Sduke#define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 820Sduke#define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 830Sduke 840Sduke#define SIN6(s) ((struct sockaddr_in6 *)s) 850Sduke 860Sduke#ifdef VIMAGE_GLOBALS 870Sdukeint nd6_prune; 880Sdukeint nd6_delay; 890Sdukeint nd6_umaxtries; 900Sdukeint nd6_mmaxtries; 910Sdukeint nd6_useloopback; 920Sdukeint nd6_gctimer; 930Sduke 940Sduke/* preventing too many loops in ND option parsing */ 950Sdukeint nd6_maxndopt; 960Sduke 970Sdukeint nd6_maxnudhint; 980Sdukeint nd6_maxqueuelen; 990Sduke 1000Sdukeint nd6_debug; 1010Sduke 1020Sduke/* for debugging? */ 1030Sduke#if 0 1040Sdukestatic int nd6_inuse, nd6_allocated; 1050Sduke#endif 1060Sduke 1070Sdukestruct nd_drhead nd_defrouter; 1080Sdukestruct nd_prhead nd_prefix; 1090Sduke 1100Sdukeint nd6_recalc_reachtm_interval; 1110Sduke#endif /* VIMAGE_GLOBALS */ 1120Sduke 1130Sdukestatic struct sockaddr_in6 all1_sa; 1146760Sgoetz 1150Sdukestatic int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, 1160Sduke struct ifnet *)); 1176760Sgoetzstatic void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 1180Sdukestatic void nd6_slowtimo(void *); 1190Sdukestatic int regen_tmpaddr(struct in6_ifaddr *); 1200Sdukestatic struct llentry *nd6_free(struct llentry *, int); 1210Sdukestatic void nd6_llinfo_timer(void *); 1220Sdukestatic void clear_llinfo_pqueue(struct llentry *); 1230Sduke 1240Sduke#ifdef VIMAGE_GLOBALS 1250Sdukestruct callout nd6_slowtimo_ch; 1260Sdukestruct callout nd6_timer_ch; 12711719Sgoetzextern struct callout in6_tmpaddrtimer_ch; 12811719Sgoetzextern int dad_ignore_ns; 1290Sdukeextern int dad_maxtry; 1300Sduke#endif 1310Sduke 1320Sdukevoid 1330Sdukend6_init(void) 1340Sduke{ 1350Sduke INIT_VNET_INET6(curvnet); 1360Sduke int i; 1370Sduke 1380Sduke V_nd6_prune = 1; /* walk list every 1 seconds */ 1390Sduke V_nd6_delay = 5; /* delay first probe time 5 second */ 1400Sduke V_nd6_umaxtries = 3; /* maximum unicast query */ 1410Sduke V_nd6_mmaxtries = 3; /* maximum multicast query */ 1420Sduke V_nd6_useloopback = 1; /* use loopback interface for local traffic */ 1430Sduke V_nd6_gctimer = (60 * 60 * 24); /* 1 day: garbage collection timer */ 1440Sduke 1450Sduke /* preventing too many loops in ND option parsing */ 1460Sduke V_nd6_maxndopt = 10; /* max # of ND options allowed */ 1470Sduke 1480Sduke V_nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */ 1490Sduke V_nd6_maxqueuelen = 1; /* max pkts cached in unresolved ND entries */ 1508379Skvn 1518379Skvn#ifdef ND6_DEBUG 1520Sduke V_nd6_debug = 1; 1538379Skvn#else 1548379Skvn V_nd6_debug = 0; 1550Sduke#endif 1560Sduke 1570Sduke V_nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL; 1580Sduke 1590Sduke V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ 1600Sduke V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ 1616760Sgoetz 1620Sduke /* 1630Sduke * XXX just to get this to compile KMM 1640Sduke */ 1650Sduke#ifdef notyet 1660Sduke V_llinfo_nd6.ln_next = &V_llinfo_nd6; 1679056Sdavid V_llinfo_nd6.ln_prev = &V_llinfo_nd6; 1680Sduke#endif 1690Sduke LIST_INIT(&V_nd_prefix); 1709111Stwisti 1710Sduke V_ip6_use_tempaddr = 0; 1720Sduke V_ip6_temp_preferred_lifetime = DEF_TEMP_PREFERRED_LIFETIME; 1730Sduke V_ip6_temp_valid_lifetime = DEF_TEMP_VALID_LIFETIME; 1740Sduke V_ip6_temp_regen_advance = TEMPADDR_REGEN_ADVANCE; 1750Sduke 1760Sduke V_ip6_desync_factor = 0; 1770Sduke 1780Sduke all1_sa.sin6_family = AF_INET6; 1790Sduke all1_sa.sin6_len = sizeof(struct sockaddr_in6); 1800Sduke for (i = 0; i < sizeof(all1_sa.sin6_addr); i++) 1810Sduke all1_sa.sin6_addr.s6_addr[i] = 0xff; 1820Sduke 1830Sduke /* initialization of the default router list */ 1840Sduke TAILQ_INIT(&V_nd_defrouter); 1850Sduke /* start timer */ 1860Sduke callout_init(&V_nd6_slowtimo_ch, 0); 1870Sduke callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1880Sduke nd6_slowtimo, curvnet); 1890Sduke} 1900Sduke 1910Sduke 1920Sduke#ifdef VIMAGE 1938379Skvnvoid 1948379Skvnnd6_destroy() 1958379Skvn{ 1968379Skvn INIT_VNET_INET6(curvnet); 1978379Skvn 1988379Skvn callout_drain(&V_nd6_slowtimo_ch); 1998379Skvn callout_drain(&V_nd6_timer_ch); 2008379Skvn} 2018379Skvn#endif 2028379Skvn 2038379Skvnstruct nd_ifinfo * 2048379Skvnnd6_ifattach(struct ifnet *ifp) 2058379Skvn{ 2068379Skvn struct nd_ifinfo *nd; 2078379Skvn 2088379Skvn nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK); 2090Sduke bzero(nd, sizeof(*nd)); 2100Sduke 2110Sduke nd->initialized = 1; 2120Sduke 2130Sduke nd->chlim = IPV6_DEFHLIM; 2140Sduke nd->basereachable = REACHABLE_TIME; 2150Sduke nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 2160Sduke nd->retrans = RETRANS_TIMER; 2170Sduke /* 2180Sduke * Note that the default value of ip6_accept_rtadv is 0, which means 2190Sduke * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV 2200Sduke * here. 2210Sduke */ 2220Sduke nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV); 2230Sduke 2240Sduke /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 2250Sduke nd6_setmtu0(ifp, nd); 2268379Skvn 2278379Skvn return nd; 2288379Skvn} 2298379Skvn 2308379Skvnvoid 2318379Skvnnd6_ifdetach(struct nd_ifinfo *nd) 2328379Skvn{ 2338379Skvn 2348379Skvn free(nd, M_IP6NDP); 2358379Skvn} 2368379Skvn 2378379Skvn/* 2388379Skvn * Reset ND level link MTU. This function is called when the physical MTU 2398379Skvn * changes, which means we might have to adjust the ND level MTU. 2408379Skvn */ 2418379Skvnvoid 2428379Skvnnd6_setmtu(struct ifnet *ifp) 2438379Skvn{ 2448379Skvn 2458379Skvn nd6_setmtu0(ifp, ND_IFINFO(ifp)); 2468379Skvn} 2478379Skvn 2488379Skvn/* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 2498379Skvnvoid 2508379Skvnnd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 2519056Sdavid{ 2528379Skvn INIT_VNET_INET6(ifp->if_vnet); 2538379Skvn u_int32_t omaxmtu; 2548379Skvn 2558379Skvn omaxmtu = ndi->maxmtu; 2568379Skvn 2578379Skvn switch (ifp->if_type) { 2588379Skvn case IFT_ARCNET: 2598379Skvn ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 2608379Skvn break; 2618379Skvn case IFT_FDDI: 2628379Skvn ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 2638379Skvn break; 2648379Skvn case IFT_ISO88025: 2658379Skvn ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 2660Sduke break; 2670Sduke default: 2680Sduke ndi->maxmtu = ifp->if_mtu; 2690Sduke break; 2700Sduke } 2710Sduke 2720Sduke /* 2730Sduke * Decreasing the interface MTU under IPV6 minimum MTU may cause 2740Sduke * undesirable situation. We thus notify the operator of the change 2750Sduke * explicitly. The check for omaxmtu is necessary to restrict the 2760Sduke * log to the case of changing the MTU, not initializing it. 2778379Skvn */ 2780Sduke if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 2798379Skvn log(LOG_NOTICE, "nd6_setmtu0: " 2800Sduke "new link MTU on %s (%lu) is too small for IPv6\n", 2818379Skvn if_name(ifp), (unsigned long)ndi->maxmtu); 2828379Skvn } 2838379Skvn 2848379Skvn if (ndi->maxmtu > V_in6_maxmtu) 2850Sduke in6_setmaxmtu(); /* check all interfaces just in case */ 2860Sduke 2870Sduke} 2880Sduke 2890Sdukevoid 2908379Skvnnd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 2910Sduke{ 2920Sduke 2931879Sstefank bzero(ndopts, sizeof(*ndopts)); 2941879Sstefank ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 295 ndopts->nd_opts_last 296 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 297 298 if (icmp6len == 0) { 299 ndopts->nd_opts_done = 1; 300 ndopts->nd_opts_search = NULL; 301 } 302} 303 304/* 305 * Take one ND option. 306 */ 307struct nd_opt_hdr * 308nd6_option(union nd_opts *ndopts) 309{ 310 struct nd_opt_hdr *nd_opt; 311 int olen; 312 313 if (ndopts == NULL) 314 panic("ndopts == NULL in nd6_option"); 315 if (ndopts->nd_opts_last == NULL) 316 panic("uninitialized ndopts in nd6_option"); 317 if (ndopts->nd_opts_search == NULL) 318 return NULL; 319 if (ndopts->nd_opts_done) 320 return NULL; 321 322 nd_opt = ndopts->nd_opts_search; 323 324 /* make sure nd_opt_len is inside the buffer */ 325 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 326 bzero(ndopts, sizeof(*ndopts)); 327 return NULL; 328 } 329 330 olen = nd_opt->nd_opt_len << 3; 331 if (olen == 0) { 332 /* 333 * Message validation requires that all included 334 * options have a length that is greater than zero. 335 */ 336 bzero(ndopts, sizeof(*ndopts)); 337 return NULL; 338 } 339 340 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 341 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 342 /* option overruns the end of buffer, invalid */ 343 bzero(ndopts, sizeof(*ndopts)); 344 return NULL; 345 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 346 /* reached the end of options chain */ 347 ndopts->nd_opts_done = 1; 348 ndopts->nd_opts_search = NULL; 349 } 350 return nd_opt; 351} 352 353/* 354 * Parse multiple ND options. 355 * This function is much easier to use, for ND routines that do not need 356 * multiple options of the same type. 357 */ 358int 359nd6_options(union nd_opts *ndopts) 360{ 361 INIT_VNET_INET6(curvnet); 362 struct nd_opt_hdr *nd_opt; 363 int i = 0; 364 365 if (ndopts == NULL) 366 panic("ndopts == NULL in nd6_options"); 367 if (ndopts->nd_opts_last == NULL) 368 panic("uninitialized ndopts in nd6_options"); 369 if (ndopts->nd_opts_search == NULL) 370 return 0; 371 372 while (1) { 373 nd_opt = nd6_option(ndopts); 374 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 375 /* 376 * Message validation requires that all included 377 * options have a length that is greater than zero. 378 */ 379 ICMP6STAT_INC(icp6s_nd_badopt); 380 bzero(ndopts, sizeof(*ndopts)); 381 return -1; 382 } 383 384 if (nd_opt == NULL) 385 goto skip1; 386 387 switch (nd_opt->nd_opt_type) { 388 case ND_OPT_SOURCE_LINKADDR: 389 case ND_OPT_TARGET_LINKADDR: 390 case ND_OPT_MTU: 391 case ND_OPT_REDIRECTED_HEADER: 392 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 393 nd6log((LOG_INFO, 394 "duplicated ND6 option found (type=%d)\n", 395 nd_opt->nd_opt_type)); 396 /* XXX bark? */ 397 } else { 398 ndopts->nd_opt_array[nd_opt->nd_opt_type] 399 = nd_opt; 400 } 401 break; 402 case ND_OPT_PREFIX_INFORMATION: 403 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 404 ndopts->nd_opt_array[nd_opt->nd_opt_type] 405 = nd_opt; 406 } 407 ndopts->nd_opts_pi_end = 408 (struct nd_opt_prefix_info *)nd_opt; 409 break; 410 default: 411 /* 412 * Unknown options must be silently ignored, 413 * to accomodate future extension to the protocol. 414 */ 415 nd6log((LOG_DEBUG, 416 "nd6_options: unsupported option %d - " 417 "option ignored\n", nd_opt->nd_opt_type)); 418 } 419 420skip1: 421 i++; 422 if (i > V_nd6_maxndopt) { 423 ICMP6STAT_INC(icp6s_nd_toomanyopt); 424 nd6log((LOG_INFO, "too many loop in nd opt\n")); 425 break; 426 } 427 428 if (ndopts->nd_opts_done) 429 break; 430 } 431 432 return 0; 433} 434 435/* 436 * ND6 timer routine to handle ND6 entries 437 */ 438void 439nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 440{ 441 if (tick < 0) { 442 ln->la_expire = 0; 443 ln->ln_ntick = 0; 444 callout_stop(&ln->ln_timer_ch); 445 /* 446 * XXX - do we know that there is 447 * callout installed? i.e. are we 448 * guaranteed that we're not dropping 449 * a reference that we did not add? 450 * KMM 451 */ 452 LLE_REMREF(ln); 453 } else { 454 ln->la_expire = time_second + tick / hz; 455 LLE_ADDREF(ln); 456 if (tick > INT_MAX) { 457 ln->ln_ntick = tick - INT_MAX; 458 callout_reset(&ln->ln_timer_ch, INT_MAX, 459 nd6_llinfo_timer, ln); 460 } else { 461 ln->ln_ntick = 0; 462 callout_reset(&ln->ln_timer_ch, tick, 463 nd6_llinfo_timer, ln); 464 } 465 } 466} 467 468void 469nd6_llinfo_settimer(struct llentry *ln, long tick) 470{ 471 472 LLE_WLOCK(ln); 473 nd6_llinfo_settimer_locked(ln, tick); 474 LLE_WUNLOCK(ln); 475} 476 477static void 478nd6_llinfo_timer(void *arg) 479{ 480 struct llentry *ln; 481 struct in6_addr *dst; 482 struct ifnet *ifp; 483 struct nd_ifinfo *ndi = NULL; 484 485 ln = (struct llentry *)arg; 486 if (ln == NULL) { 487 panic("%s: NULL entry!\n", __func__); 488 return; 489 } 490 491 if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) 492 panic("ln ifp == NULL"); 493 494 CURVNET_SET(ifp->if_vnet); 495 INIT_VNET_INET6(curvnet); 496 497 if (ln->ln_ntick > 0) { 498 if (ln->ln_ntick > INT_MAX) { 499 ln->ln_ntick -= INT_MAX; 500 nd6_llinfo_settimer(ln, INT_MAX); 501 } else { 502 ln->ln_ntick = 0; 503 nd6_llinfo_settimer(ln, ln->ln_ntick); 504 } 505 goto done; 506 } 507 508 ndi = ND_IFINFO(ifp); 509 dst = &L3_ADDR_SIN6(ln)->sin6_addr; 510 if ((ln->la_flags & LLE_STATIC) || (ln->la_expire > time_second)) { 511 goto done; 512 } 513 514 if (ln->la_flags & LLE_DELETED) { 515 (void)nd6_free(ln, 0); 516 goto done; 517 } 518 519 switch (ln->ln_state) { 520 case ND6_LLINFO_INCOMPLETE: 521 if (ln->la_asked < V_nd6_mmaxtries) { 522 ln->la_asked++; 523 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 524 nd6_ns_output(ifp, NULL, dst, ln, 0); 525 } else { 526 struct mbuf *m = ln->la_hold; 527 if (m) { 528 struct mbuf *m0; 529 530 /* 531 * assuming every packet in la_hold has the 532 * same IP header 533 */ 534 m0 = m->m_nextpkt; 535 m->m_nextpkt = NULL; 536 icmp6_error2(m, ICMP6_DST_UNREACH, 537 ICMP6_DST_UNREACH_ADDR, 0, ifp); 538 539 ln->la_hold = m0; 540 clear_llinfo_pqueue(ln); 541 } 542 (void)nd6_free(ln, 0); 543 ln = NULL; 544 } 545 break; 546 case ND6_LLINFO_REACHABLE: 547 if (!ND6_LLINFO_PERMANENT(ln)) { 548 ln->ln_state = ND6_LLINFO_STALE; 549 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 550 } 551 break; 552 553 case ND6_LLINFO_STALE: 554 /* Garbage Collection(RFC 2461 5.3) */ 555 if (!ND6_LLINFO_PERMANENT(ln)) { 556 (void)nd6_free(ln, 1); 557 ln = NULL; 558 } 559 break; 560 561 case ND6_LLINFO_DELAY: 562 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 563 /* We need NUD */ 564 ln->la_asked = 1; 565 ln->ln_state = ND6_LLINFO_PROBE; 566 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 567 nd6_ns_output(ifp, dst, dst, ln, 0); 568 } else { 569 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 570 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 571 } 572 break; 573 case ND6_LLINFO_PROBE: 574 if (ln->la_asked < V_nd6_umaxtries) { 575 ln->la_asked++; 576 nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); 577 nd6_ns_output(ifp, dst, dst, ln, 0); 578 } else { 579 (void)nd6_free(ln, 0); 580 ln = NULL; 581 } 582 break; 583 } 584 CURVNET_RESTORE(); 585done: 586 if (ln != NULL) 587 LLE_FREE(ln); 588} 589 590 591/* 592 * ND6 timer routine to expire default route list and prefix list 593 */ 594void 595nd6_timer(void *arg) 596{ 597 CURVNET_SET((struct vnet *) arg); 598 INIT_VNET_INET6(curvnet); 599 int s; 600 struct nd_defrouter *dr; 601 struct nd_prefix *pr; 602 struct in6_ifaddr *ia6, *nia6; 603 struct in6_addrlifetime *lt6; 604 605 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 606 nd6_timer, curvnet); 607 608 /* expire default router list */ 609 s = splnet(); 610 dr = TAILQ_FIRST(&V_nd_defrouter); 611 while (dr) { 612 if (dr->expire && dr->expire < time_second) { 613 struct nd_defrouter *t; 614 t = TAILQ_NEXT(dr, dr_entry); 615 defrtrlist_del(dr); 616 dr = t; 617 } else { 618 dr = TAILQ_NEXT(dr, dr_entry); 619 } 620 } 621 622 /* 623 * expire interface addresses. 624 * in the past the loop was inside prefix expiry processing. 625 * However, from a stricter speci-confrmance standpoint, we should 626 * rather separate address lifetimes and prefix lifetimes. 627 */ 628 addrloop: 629 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 630 /* check address lifetime */ 631 lt6 = &ia6->ia6_lifetime; 632 if (IFA6_IS_INVALID(ia6)) { 633 int regen = 0; 634 635 /* 636 * If the expiring address is temporary, try 637 * regenerating a new one. This would be useful when 638 * we suspended a laptop PC, then turned it on after a 639 * period that could invalidate all temporary 640 * addresses. Although we may have to restart the 641 * loop (see below), it must be after purging the 642 * address. Otherwise, we'd see an infinite loop of 643 * regeneration. 644 */ 645 if (V_ip6_use_tempaddr && 646 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 647 if (regen_tmpaddr(ia6) == 0) 648 regen = 1; 649 } 650 651 in6_purgeaddr(&ia6->ia_ifa); 652 653 if (regen) 654 goto addrloop; /* XXX: see below */ 655 } else if (IFA6_IS_DEPRECATED(ia6)) { 656 int oldflags = ia6->ia6_flags; 657 658 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 659 660 /* 661 * If a temporary address has just become deprecated, 662 * regenerate a new one if possible. 663 */ 664 if (V_ip6_use_tempaddr && 665 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 666 (oldflags & IN6_IFF_DEPRECATED) == 0) { 667 668 if (regen_tmpaddr(ia6) == 0) { 669 /* 670 * A new temporary address is 671 * generated. 672 * XXX: this means the address chain 673 * has changed while we are still in 674 * the loop. Although the change 675 * would not cause disaster (because 676 * it's not a deletion, but an 677 * addition,) we'd rather restart the 678 * loop just for safety. Or does this 679 * significantly reduce performance?? 680 */ 681 goto addrloop; 682 } 683 } 684 } else { 685 /* 686 * A new RA might have made a deprecated address 687 * preferred. 688 */ 689 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 690 } 691 } 692 693 /* expire prefix list */ 694 pr = V_nd_prefix.lh_first; 695 while (pr) { 696 /* 697 * check prefix lifetime. 698 * since pltime is just for autoconf, pltime processing for 699 * prefix is not necessary. 700 */ 701 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 702 time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) { 703 struct nd_prefix *t; 704 t = pr->ndpr_next; 705 706 /* 707 * address expiration and prefix expiration are 708 * separate. NEVER perform in6_purgeaddr here. 709 */ 710 711 prelist_remove(pr); 712 pr = t; 713 } else 714 pr = pr->ndpr_next; 715 } 716 splx(s); 717 CURVNET_RESTORE(); 718} 719 720/* 721 * ia6 - deprecated/invalidated temporary address 722 */ 723static int 724regen_tmpaddr(struct in6_ifaddr *ia6) 725{ 726 struct ifaddr *ifa; 727 struct ifnet *ifp; 728 struct in6_ifaddr *public_ifa6 = NULL; 729 730 ifp = ia6->ia_ifa.ifa_ifp; 731 IF_ADDR_LOCK(ifp); 732 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 733 struct in6_ifaddr *it6; 734 735 if (ifa->ifa_addr->sa_family != AF_INET6) 736 continue; 737 738 it6 = (struct in6_ifaddr *)ifa; 739 740 /* ignore no autoconf addresses. */ 741 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 742 continue; 743 744 /* ignore autoconf addresses with different prefixes. */ 745 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 746 continue; 747 748 /* 749 * Now we are looking at an autoconf address with the same 750 * prefix as ours. If the address is temporary and is still 751 * preferred, do not create another one. It would be rare, but 752 * could happen, for example, when we resume a laptop PC after 753 * a long period. 754 */ 755 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 756 !IFA6_IS_DEPRECATED(it6)) { 757 public_ifa6 = NULL; 758 break; 759 } 760 761 /* 762 * This is a public autoconf address that has the same prefix 763 * as ours. If it is preferred, keep it. We can't break the 764 * loop here, because there may be a still-preferred temporary 765 * address with the prefix. 766 */ 767 if (!IFA6_IS_DEPRECATED(it6)) 768 public_ifa6 = it6; 769 } 770 771 if (public_ifa6 != NULL) { 772 int e; 773 774 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 775 IF_ADDR_UNLOCK(ifp); 776 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 777 " tmp addr,errno=%d\n", e); 778 return (-1); 779 } 780 IF_ADDR_UNLOCK(ifp); 781 return (0); 782 } 783 784 IF_ADDR_UNLOCK(ifp); 785 return (-1); 786} 787 788/* 789 * Nuke neighbor cache/prefix/default router management table, right before 790 * ifp goes away. 791 */ 792void 793nd6_purge(struct ifnet *ifp) 794{ 795 INIT_VNET_INET6(ifp->if_vnet); 796 struct nd_defrouter *dr, *ndr; 797 struct nd_prefix *pr, *npr; 798 799 /* 800 * Nuke default router list entries toward ifp. 801 * We defer removal of default router list entries that is installed 802 * in the routing table, in order to keep additional side effects as 803 * small as possible. 804 */ 805 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 806 ndr = TAILQ_NEXT(dr, dr_entry); 807 if (dr->installed) 808 continue; 809 810 if (dr->ifp == ifp) 811 defrtrlist_del(dr); 812 } 813 814 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = ndr) { 815 ndr = TAILQ_NEXT(dr, dr_entry); 816 if (!dr->installed) 817 continue; 818 819 if (dr->ifp == ifp) 820 defrtrlist_del(dr); 821 } 822 823 /* Nuke prefix list entries toward ifp */ 824 for (pr = V_nd_prefix.lh_first; pr; pr = npr) { 825 npr = pr->ndpr_next; 826 if (pr->ndpr_ifp == ifp) { 827 /* 828 * Because if_detach() does *not* release prefixes 829 * while purging addresses the reference count will 830 * still be above zero. We therefore reset it to 831 * make sure that the prefix really gets purged. 832 */ 833 pr->ndpr_refcnt = 0; 834 835 /* 836 * Previously, pr->ndpr_addr is removed as well, 837 * but I strongly believe we don't have to do it. 838 * nd6_purge() is only called from in6_ifdetach(), 839 * which removes all the associated interface addresses 840 * by itself. 841 * (jinmei@kame.net 20010129) 842 */ 843 prelist_remove(pr); 844 } 845 } 846 847 /* cancel default outgoing interface setting */ 848 if (V_nd6_defifindex == ifp->if_index) 849 nd6_setdefaultiface(0); 850 851 if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */ 852 /* refresh default router list 853 * 854 * 855 */ 856 defrouter_select(); 857 858 } 859 860 /* XXXXX 861 * We do not nuke the neighbor cache entries here any more 862 * because the neighbor cache is kept in if_afdata[AF_INET6]. 863 * nd6_purge() is invoked by in6_ifdetach() which is called 864 * from if_detach() where everything gets purged. So let 865 * in6_domifdetach() do the actual L2 table purging work. 866 */ 867} 868 869/* 870 * the caller acquires and releases the lock on the lltbls 871 * Returns the llentry locked 872 */ 873struct llentry * 874nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 875{ 876 struct sockaddr_in6 sin6; 877 struct llentry *ln; 878 int llflags = 0; 879 880 bzero(&sin6, sizeof(sin6)); 881 sin6.sin6_len = sizeof(struct sockaddr_in6); 882 sin6.sin6_family = AF_INET6; 883 sin6.sin6_addr = *addr6; 884 885 IF_AFDATA_LOCK_ASSERT(ifp); 886 887 if (flags & ND6_CREATE) 888 llflags |= LLE_CREATE; 889 if (flags & ND6_EXCLUSIVE) 890 llflags |= LLE_EXCLUSIVE; 891 892 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 893 if ((ln != NULL) && (flags & LLE_CREATE)) { 894 ln->ln_state = ND6_LLINFO_NOSTATE; 895 callout_init(&ln->ln_timer_ch, 0); 896 } 897 898 return (ln); 899} 900 901/* 902 * Test whether a given IPv6 address is a neighbor or not, ignoring 903 * the actual neighbor cache. The neighbor cache is ignored in order 904 * to not reenter the routing code from within itself. 905 */ 906static int 907nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 908{ 909 INIT_VNET_INET6(ifp->if_vnet); 910 struct nd_prefix *pr; 911 struct ifaddr *dstaddr; 912 913 /* 914 * A link-local address is always a neighbor. 915 * XXX: a link does not necessarily specify a single interface. 916 */ 917 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 918 struct sockaddr_in6 sin6_copy; 919 u_int32_t zone; 920 921 /* 922 * We need sin6_copy since sa6_recoverscope() may modify the 923 * content (XXX). 924 */ 925 sin6_copy = *addr; 926 if (sa6_recoverscope(&sin6_copy)) 927 return (0); /* XXX: should be impossible */ 928 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 929 return (0); 930 if (sin6_copy.sin6_scope_id == zone) 931 return (1); 932 else 933 return (0); 934 } 935 936 /* 937 * If the address matches one of our addresses, 938 * it should be a neighbor. 939 * If the address matches one of our on-link prefixes, it should be a 940 * neighbor. 941 */ 942 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 943 if (pr->ndpr_ifp != ifp) 944 continue; 945 946 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) 947 continue; 948 949 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 950 &addr->sin6_addr, &pr->ndpr_mask)) 951 return (1); 952 } 953 954 /* 955 * If the address is assigned on the node of the other side of 956 * a p2p interface, the address should be a neighbor. 957 */ 958 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr); 959 if (dstaddr != NULL) { 960 if (dstaddr->ifa_ifp == ifp) { 961 ifa_free(dstaddr); 962 return (1); 963 } 964 ifa_free(dstaddr); 965 } 966 967 /* 968 * If the default router list is empty, all addresses are regarded 969 * as on-link, and thus, as a neighbor. 970 * XXX: we restrict the condition to hosts, because routers usually do 971 * not have the "default router list". 972 */ 973 if (!V_ip6_forwarding && TAILQ_FIRST(&V_nd_defrouter) == NULL && 974 V_nd6_defifindex == ifp->if_index) { 975 return (1); 976 } 977 978 return (0); 979} 980 981 982/* 983 * Detect if a given IPv6 address identifies a neighbor on a given link. 984 * XXX: should take care of the destination of a p2p link? 985 */ 986int 987nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 988{ 989 struct llentry *lle; 990 int rc = 0; 991 992 IF_AFDATA_UNLOCK_ASSERT(ifp); 993 if (nd6_is_new_addr_neighbor(addr, ifp)) 994 return (1); 995 996 /* 997 * Even if the address matches none of our addresses, it might be 998 * in the neighbor cache. 999 */ 1000 IF_AFDATA_LOCK(ifp); 1001 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1002 LLE_RUNLOCK(lle); 1003 rc = 1; 1004 } 1005 IF_AFDATA_UNLOCK(ifp); 1006 return (rc); 1007} 1008 1009/* 1010 * Free an nd6 llinfo entry. 1011 * Since the function would cause significant changes in the kernel, DO NOT 1012 * make it global, unless you have a strong reason for the change, and are sure 1013 * that the change is safe. 1014 */ 1015static struct llentry * 1016nd6_free(struct llentry *ln, int gc) 1017{ 1018 INIT_VNET_INET6(curvnet); 1019 struct llentry *next; 1020 struct nd_defrouter *dr; 1021 struct ifnet *ifp=NULL; 1022 1023 /* 1024 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1025 * even though it is not harmful, it was not really necessary. 1026 */ 1027 1028 /* cancel timer */ 1029 nd6_llinfo_settimer(ln, -1); 1030 1031 if (!V_ip6_forwarding) { 1032 int s; 1033 s = splnet(); 1034 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1035 1036 if (dr != NULL && dr->expire && 1037 ln->ln_state == ND6_LLINFO_STALE && gc) { 1038 /* 1039 * If the reason for the deletion is just garbage 1040 * collection, and the neighbor is an active default 1041 * router, do not delete it. Instead, reset the GC 1042 * timer using the router's lifetime. 1043 * Simply deleting the entry would affect default 1044 * router selection, which is not necessarily a good 1045 * thing, especially when we're using router preference 1046 * values. 1047 * XXX: the check for ln_state would be redundant, 1048 * but we intentionally keep it just in case. 1049 */ 1050 if (dr->expire > time_second) 1051 nd6_llinfo_settimer(ln, 1052 (dr->expire - time_second) * hz); 1053 else 1054 nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); 1055 splx(s); 1056 return (LIST_NEXT(ln, lle_next)); 1057 } 1058 1059 if (ln->ln_router || dr) { 1060 /* 1061 * rt6_flush must be called whether or not the neighbor 1062 * is in the Default Router List. 1063 * See a corresponding comment in nd6_na_input(). 1064 */ 1065 rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); 1066 } 1067 1068 if (dr) { 1069 /* 1070 * Unreachablity of a router might affect the default 1071 * router selection and on-link detection of advertised 1072 * prefixes. 1073 */ 1074 1075 /* 1076 * Temporarily fake the state to choose a new default 1077 * router and to perform on-link determination of 1078 * prefixes correctly. 1079 * Below the state will be set correctly, 1080 * or the entry itself will be deleted. 1081 */ 1082 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1083 1084 /* 1085 * Since defrouter_select() does not affect the 1086 * on-link determination and MIP6 needs the check 1087 * before the default router selection, we perform 1088 * the check now. 1089 */ 1090 pfxlist_onlink_check(); 1091 1092 /* 1093 * refresh default router list 1094 */ 1095 defrouter_select(); 1096 } 1097 splx(s); 1098 } 1099 1100 /* 1101 * Before deleting the entry, remember the next entry as the 1102 * return value. We need this because pfxlist_onlink_check() above 1103 * might have freed other entries (particularly the old next entry) as 1104 * a side effect (XXX). 1105 */ 1106 next = LIST_NEXT(ln, lle_next); 1107 1108 ifp = ln->lle_tbl->llt_ifp; 1109 IF_AFDATA_LOCK(ifp); 1110 LLE_WLOCK(ln); 1111 llentry_free(ln); 1112 IF_AFDATA_UNLOCK(ifp); 1113 1114 return (next); 1115} 1116 1117/* 1118 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1119 * 1120 * XXX cost-effective methods? 1121 */ 1122void 1123nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1124{ 1125 INIT_VNET_INET6(curvnet); 1126 struct llentry *ln; 1127 struct ifnet *ifp; 1128 1129 if ((dst6 == NULL) || (rt == NULL)) 1130 return; 1131 1132 ifp = rt->rt_ifp; 1133 IF_AFDATA_LOCK(ifp); 1134 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1135 IF_AFDATA_UNLOCK(ifp); 1136 if (ln == NULL) 1137 return; 1138 1139 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1140 goto done; 1141 1142 /* 1143 * if we get upper-layer reachability confirmation many times, 1144 * it is possible we have false information. 1145 */ 1146 if (!force) { 1147 ln->ln_byhint++; 1148 if (ln->ln_byhint > V_nd6_maxnudhint) { 1149 goto done; 1150 } 1151 } 1152 1153 ln->ln_state = ND6_LLINFO_REACHABLE; 1154 if (!ND6_LLINFO_PERMANENT(ln)) { 1155 nd6_llinfo_settimer(ln, 1156 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1157 } 1158done: 1159 LLE_WUNLOCK(ln); 1160} 1161 1162 1163int 1164nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1165{ 1166 INIT_VNET_INET6(ifp->if_vnet); 1167 struct in6_drlist *drl = (struct in6_drlist *)data; 1168 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1169 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1170 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1171 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1172 struct nd_defrouter *dr; 1173 struct nd_prefix *pr; 1174 int i = 0, error = 0; 1175 int s; 1176 1177 switch (cmd) { 1178 case SIOCGDRLST_IN6: 1179 /* 1180 * obsolete API, use sysctl under net.inet6.icmp6 1181 */ 1182 bzero(drl, sizeof(*drl)); 1183 s = splnet(); 1184 dr = TAILQ_FIRST(&V_nd_defrouter); 1185 while (dr && i < DRLSTSIZ) { 1186 drl->defrouter[i].rtaddr = dr->rtaddr; 1187 in6_clearscope(&drl->defrouter[i].rtaddr); 1188 1189 drl->defrouter[i].flags = dr->flags; 1190 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1191 drl->defrouter[i].expire = dr->expire; 1192 drl->defrouter[i].if_index = dr->ifp->if_index; 1193 i++; 1194 dr = TAILQ_NEXT(dr, dr_entry); 1195 } 1196 splx(s); 1197 break; 1198 case SIOCGPRLST_IN6: 1199 /* 1200 * obsolete API, use sysctl under net.inet6.icmp6 1201 * 1202 * XXX the structure in6_prlist was changed in backward- 1203 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1204 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1205 */ 1206 /* 1207 * XXX meaning of fields, especialy "raflags", is very 1208 * differnet between RA prefix list and RR/static prefix list. 1209 * how about separating ioctls into two? 1210 */ 1211 bzero(oprl, sizeof(*oprl)); 1212 s = splnet(); 1213 pr = V_nd_prefix.lh_first; 1214 while (pr && i < PRLSTSIZ) { 1215 struct nd_pfxrouter *pfr; 1216 int j; 1217 1218 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1219 oprl->prefix[i].raflags = pr->ndpr_raf; 1220 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1221 oprl->prefix[i].vltime = pr->ndpr_vltime; 1222 oprl->prefix[i].pltime = pr->ndpr_pltime; 1223 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1224 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1225 oprl->prefix[i].expire = 0; 1226 else { 1227 time_t maxexpire; 1228 1229 /* XXX: we assume time_t is signed. */ 1230 maxexpire = (-1) & 1231 ~((time_t)1 << 1232 ((sizeof(maxexpire) * 8) - 1)); 1233 if (pr->ndpr_vltime < 1234 maxexpire - pr->ndpr_lastupdate) { 1235 oprl->prefix[i].expire = 1236 pr->ndpr_lastupdate + 1237 pr->ndpr_vltime; 1238 } else 1239 oprl->prefix[i].expire = maxexpire; 1240 } 1241 1242 pfr = pr->ndpr_advrtrs.lh_first; 1243 j = 0; 1244 while (pfr) { 1245 if (j < DRLSTSIZ) { 1246#define RTRADDR oprl->prefix[i].advrtr[j] 1247 RTRADDR = pfr->router->rtaddr; 1248 in6_clearscope(&RTRADDR); 1249#undef RTRADDR 1250 } 1251 j++; 1252 pfr = pfr->pfr_next; 1253 } 1254 oprl->prefix[i].advrtrs = j; 1255 oprl->prefix[i].origin = PR_ORIG_RA; 1256 1257 i++; 1258 pr = pr->ndpr_next; 1259 } 1260 splx(s); 1261 1262 break; 1263 case OSIOCGIFINFO_IN6: 1264#define ND ndi->ndi 1265 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1266 bzero(&ND, sizeof(ND)); 1267 ND.linkmtu = IN6_LINKMTU(ifp); 1268 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1269 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1270 ND.reachable = ND_IFINFO(ifp)->reachable; 1271 ND.retrans = ND_IFINFO(ifp)->retrans; 1272 ND.flags = ND_IFINFO(ifp)->flags; 1273 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1274 ND.chlim = ND_IFINFO(ifp)->chlim; 1275 break; 1276 case SIOCGIFINFO_IN6: 1277 ND = *ND_IFINFO(ifp); 1278 break; 1279 case SIOCSIFINFO_IN6: 1280 /* 1281 * used to change host variables from userland. 1282 * intented for a use on router to reflect RA configurations. 1283 */ 1284 /* 0 means 'unspecified' */ 1285 if (ND.linkmtu != 0) { 1286 if (ND.linkmtu < IPV6_MMTU || 1287 ND.linkmtu > IN6_LINKMTU(ifp)) { 1288 error = EINVAL; 1289 break; 1290 } 1291 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1292 } 1293 1294 if (ND.basereachable != 0) { 1295 int obasereachable = ND_IFINFO(ifp)->basereachable; 1296 1297 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1298 if (ND.basereachable != obasereachable) 1299 ND_IFINFO(ifp)->reachable = 1300 ND_COMPUTE_RTIME(ND.basereachable); 1301 } 1302 if (ND.retrans != 0) 1303 ND_IFINFO(ifp)->retrans = ND.retrans; 1304 if (ND.chlim != 0) 1305 ND_IFINFO(ifp)->chlim = ND.chlim; 1306 /* FALLTHROUGH */ 1307 case SIOCSIFINFO_FLAGS: 1308 ND_IFINFO(ifp)->flags = ND.flags; 1309 break; 1310#undef ND 1311 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1312 /* sync kernel routing table with the default router list */ 1313 defrouter_reset(); 1314 defrouter_select(); 1315 break; 1316 case SIOCSPFXFLUSH_IN6: 1317 { 1318 /* flush all the prefix advertised by routers */ 1319 struct nd_prefix *pr, *next; 1320 1321 s = splnet(); 1322 for (pr = V_nd_prefix.lh_first; pr; pr = next) { 1323 struct in6_ifaddr *ia, *ia_next; 1324 1325 next = pr->ndpr_next; 1326 1327 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1328 continue; /* XXX */ 1329 1330 /* do we really have to remove addresses as well? */ 1331 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1332 ia_next) { 1333 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1334 continue; 1335 1336 if (ia->ia6_ndpr == pr) 1337 in6_purgeaddr(&ia->ia_ifa); 1338 } 1339 prelist_remove(pr); 1340 } 1341 splx(s); 1342 break; 1343 } 1344 case SIOCSRTRFLUSH_IN6: 1345 { 1346 /* flush all the default routers */ 1347 struct nd_defrouter *dr, *next; 1348 1349 s = splnet(); 1350 defrouter_reset(); 1351 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = next) { 1352 next = TAILQ_NEXT(dr, dr_entry); 1353 defrtrlist_del(dr); 1354 } 1355 defrouter_select(); 1356 splx(s); 1357 break; 1358 } 1359 case SIOCGNBRINFO_IN6: 1360 { 1361 struct llentry *ln; 1362 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1363 1364 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1365 return (error); 1366 1367 IF_AFDATA_LOCK(ifp); 1368 ln = nd6_lookup(&nb_addr, 0, ifp); 1369 IF_AFDATA_UNLOCK(ifp); 1370 1371 if (ln == NULL) { 1372 error = EINVAL; 1373 break; 1374 } 1375 nbi->state = ln->ln_state; 1376 nbi->asked = ln->la_asked; 1377 nbi->isrouter = ln->ln_router; 1378 nbi->expire = ln->la_expire; 1379 LLE_RUNLOCK(ln); 1380 break; 1381 } 1382 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1383 ndif->ifindex = V_nd6_defifindex; 1384 break; 1385 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1386 return (nd6_setdefaultiface(ndif->ifindex)); 1387 } 1388 return (error); 1389} 1390 1391/* 1392 * Create neighbor cache entry and cache link-layer address, 1393 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1394 * 1395 * type - ICMP6 type 1396 * code - type dependent information 1397 * 1398 * XXXXX 1399 * The caller of this function already acquired the ndp 1400 * cache table lock because the cache entry is returned. 1401 */ 1402struct llentry * 1403nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1404 int lladdrlen, int type, int code) 1405{ 1406 INIT_VNET_INET6(curvnet); 1407 struct llentry *ln = NULL; 1408 int is_newentry; 1409 int do_update; 1410 int olladdr; 1411 int llchange; 1412 int flags = 0; 1413 int newstate = 0; 1414 uint16_t router = 0; 1415 struct sockaddr_in6 sin6; 1416 struct mbuf *chain = NULL; 1417 int static_route = 0; 1418 1419 IF_AFDATA_UNLOCK_ASSERT(ifp); 1420 1421 if (ifp == NULL) 1422 panic("ifp == NULL in nd6_cache_lladdr"); 1423 if (from == NULL) 1424 panic("from == NULL in nd6_cache_lladdr"); 1425 1426 /* nothing must be updated for unspecified address */ 1427 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1428 return NULL; 1429 1430 /* 1431 * Validation about ifp->if_addrlen and lladdrlen must be done in 1432 * the caller. 1433 * 1434 * XXX If the link does not have link-layer adderss, what should 1435 * we do? (ifp->if_addrlen == 0) 1436 * Spec says nothing in sections for RA, RS and NA. There's small 1437 * description on it in NS section (RFC 2461 7.2.3). 1438 */ 1439 flags |= lladdr ? ND6_EXCLUSIVE : 0; 1440 IF_AFDATA_LOCK(ifp); 1441 ln = nd6_lookup(from, flags, ifp); 1442 1443 if (ln == NULL) { 1444 flags |= LLE_EXCLUSIVE; 1445 ln = nd6_lookup(from, flags |ND6_CREATE, ifp); 1446 IF_AFDATA_UNLOCK(ifp); 1447 is_newentry = 1; 1448 } else { 1449 IF_AFDATA_UNLOCK(ifp); 1450 /* do nothing if static ndp is set */ 1451 if (ln->la_flags & LLE_STATIC) { 1452 static_route = 1; 1453 goto done; 1454 } 1455 is_newentry = 0; 1456 } 1457 if (ln == NULL) 1458 return (NULL); 1459 1460 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1461 if (olladdr && lladdr) { 1462 llchange = bcmp(lladdr, &ln->ll_addr, 1463 ifp->if_addrlen); 1464 } else 1465 llchange = 0; 1466 1467 /* 1468 * newentry olladdr lladdr llchange (*=record) 1469 * 0 n n -- (1) 1470 * 0 y n -- (2) 1471 * 0 n y -- (3) * STALE 1472 * 0 y y n (4) * 1473 * 0 y y y (5) * STALE 1474 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1475 * 1 -- y -- (7) * STALE 1476 */ 1477 1478 if (lladdr) { /* (3-5) and (7) */ 1479 /* 1480 * Record source link-layer address 1481 * XXX is it dependent to ifp->if_type? 1482 */ 1483 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1484 ln->la_flags |= LLE_VALID; 1485 } 1486 1487 if (!is_newentry) { 1488 if ((!olladdr && lladdr != NULL) || /* (3) */ 1489 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1490 do_update = 1; 1491 newstate = ND6_LLINFO_STALE; 1492 } else /* (1-2,4) */ 1493 do_update = 0; 1494 } else { 1495 do_update = 1; 1496 if (lladdr == NULL) /* (6) */ 1497 newstate = ND6_LLINFO_NOSTATE; 1498 else /* (7) */ 1499 newstate = ND6_LLINFO_STALE; 1500 } 1501 1502 if (do_update) { 1503 /* 1504 * Update the state of the neighbor cache. 1505 */ 1506 ln->ln_state = newstate; 1507 1508 if (ln->ln_state == ND6_LLINFO_STALE) { 1509 /* 1510 * XXX: since nd6_output() below will cause 1511 * state tansition to DELAY and reset the timer, 1512 * we must set the timer now, although it is actually 1513 * meaningless. 1514 */ 1515 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1516 1517 if (ln->la_hold) { 1518 struct mbuf *m_hold, *m_hold_next; 1519 1520 /* 1521 * reset the la_hold in advance, to explicitly 1522 * prevent a la_hold lookup in nd6_output() 1523 * (wouldn't happen, though...) 1524 */ 1525 for (m_hold = ln->la_hold, ln->la_hold = NULL; 1526 m_hold; m_hold = m_hold_next) { 1527 m_hold_next = m_hold->m_nextpkt; 1528 m_hold->m_nextpkt = NULL; 1529 1530 /* 1531 * we assume ifp is not a p2p here, so 1532 * just set the 2nd argument as the 1533 * 1st one. 1534 */ 1535 nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); 1536 } 1537 /* 1538 * If we have mbufs in the chain we need to do 1539 * deferred transmit. Copy the address from the 1540 * llentry before dropping the lock down below. 1541 */ 1542 if (chain != NULL) 1543 memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); 1544 } 1545 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1546 /* probe right away */ 1547 nd6_llinfo_settimer_locked((void *)ln, 0); 1548 } 1549 } 1550 1551 /* 1552 * ICMP6 type dependent behavior. 1553 * 1554 * NS: clear IsRouter if new entry 1555 * RS: clear IsRouter 1556 * RA: set IsRouter if there's lladdr 1557 * redir: clear IsRouter if new entry 1558 * 1559 * RA case, (1): 1560 * The spec says that we must set IsRouter in the following cases: 1561 * - If lladdr exist, set IsRouter. This means (1-5). 1562 * - If it is old entry (!newentry), set IsRouter. This means (7). 1563 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1564 * A quetion arises for (1) case. (1) case has no lladdr in the 1565 * neighbor cache, this is similar to (6). 1566 * This case is rare but we figured that we MUST NOT set IsRouter. 1567 * 1568 * newentry olladdr lladdr llchange NS RS RA redir 1569 * D R 1570 * 0 n n -- (1) c ? s 1571 * 0 y n -- (2) c s s 1572 * 0 n y -- (3) c s s 1573 * 0 y y n (4) c s s 1574 * 0 y y y (5) c s s 1575 * 1 -- n -- (6) c c c s 1576 * 1 -- y -- (7) c c s c s 1577 * 1578 * (c=clear s=set) 1579 */ 1580 switch (type & 0xff) { 1581 case ND_NEIGHBOR_SOLICIT: 1582 /* 1583 * New entry must have is_router flag cleared. 1584 */ 1585 if (is_newentry) /* (6-7) */ 1586 ln->ln_router = 0; 1587 break; 1588 case ND_REDIRECT: 1589 /* 1590 * If the icmp is a redirect to a better router, always set the 1591 * is_router flag. Otherwise, if the entry is newly created, 1592 * clear the flag. [RFC 2461, sec 8.3] 1593 */ 1594 if (code == ND_REDIRECT_ROUTER) 1595 ln->ln_router = 1; 1596 else if (is_newentry) /* (6-7) */ 1597 ln->ln_router = 0; 1598 break; 1599 case ND_ROUTER_SOLICIT: 1600 /* 1601 * is_router flag must always be cleared. 1602 */ 1603 ln->ln_router = 0; 1604 break; 1605 case ND_ROUTER_ADVERT: 1606 /* 1607 * Mark an entry with lladdr as a router. 1608 */ 1609 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1610 (is_newentry && lladdr)) { /* (7) */ 1611 ln->ln_router = 1; 1612 } 1613 break; 1614 } 1615 1616 if (ln != NULL) { 1617 static_route = (ln->la_flags & LLE_STATIC); 1618 router = ln->ln_router; 1619 1620 if (flags & ND6_EXCLUSIVE) 1621 LLE_WUNLOCK(ln); 1622 else 1623 LLE_RUNLOCK(ln); 1624 if (static_route) 1625 ln = NULL; 1626 } 1627 if (chain) 1628 nd6_output_flush(ifp, ifp, chain, &sin6, NULL); 1629 1630 /* 1631 * When the link-layer address of a router changes, select the 1632 * best router again. In particular, when the neighbor entry is newly 1633 * created, it might affect the selection policy. 1634 * Question: can we restrict the first condition to the "is_newentry" 1635 * case? 1636 * XXX: when we hear an RA from a new router with the link-layer 1637 * address option, defrouter_select() is called twice, since 1638 * defrtrlist_update called the function as well. However, I believe 1639 * we can compromise the overhead, since it only happens the first 1640 * time. 1641 * XXX: although defrouter_select() should not have a bad effect 1642 * for those are not autoconfigured hosts, we explicitly avoid such 1643 * cases for safety. 1644 */ 1645 if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) { 1646 /* 1647 * guaranteed recursion 1648 */ 1649 defrouter_select(); 1650 } 1651 1652 return (ln); 1653done: 1654 if (ln != NULL) { 1655 if (flags & ND6_EXCLUSIVE) 1656 LLE_WUNLOCK(ln); 1657 else 1658 LLE_RUNLOCK(ln); 1659 if (static_route) 1660 ln = NULL; 1661 } 1662 return (ln); 1663} 1664 1665static void 1666nd6_slowtimo(void *arg) 1667{ 1668 CURVNET_SET((struct vnet *) arg); 1669 INIT_VNET_NET((struct vnet *) arg); 1670 INIT_VNET_INET6((struct vnet *) arg); 1671 struct nd_ifinfo *nd6if; 1672 struct ifnet *ifp; 1673 1674 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1675 nd6_slowtimo, curvnet); 1676 IFNET_RLOCK(); 1677 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 1678 ifp = TAILQ_NEXT(ifp, if_list)) { 1679 nd6if = ND_IFINFO(ifp); 1680 if (nd6if->basereachable && /* already initialized */ 1681 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1682 /* 1683 * Since reachable time rarely changes by router 1684 * advertisements, we SHOULD insure that a new random 1685 * value gets recomputed at least once every few hours. 1686 * (RFC 2461, 6.3.4) 1687 */ 1688 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1689 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1690 } 1691 } 1692 IFNET_RUNLOCK(); 1693 CURVNET_RESTORE(); 1694} 1695 1696int 1697nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1698 struct sockaddr_in6 *dst, struct rtentry *rt0) 1699{ 1700 1701 return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); 1702} 1703 1704 1705/* 1706 * Note that I'm not enforcing any global serialization 1707 * lle state or asked changes here as the logic is too 1708 * complicated to avoid having to always acquire an exclusive 1709 * lock 1710 * KMM 1711 * 1712 */ 1713#define senderr(e) { error = (e); goto bad;} 1714 1715int 1716nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, 1717 struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, 1718 struct mbuf **chain) 1719{ 1720 INIT_VNET_INET6(curvnet); 1721 struct mbuf *m = m0; 1722 struct llentry *ln = lle; 1723 int error = 0; 1724 int flags = 0; 1725 1726#ifdef INVARIANTS 1727 if (lle != NULL) { 1728 1729 LLE_WLOCK_ASSERT(lle); 1730 1731 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed")); 1732 } 1733#endif 1734 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1735 goto sendpkt; 1736 1737 if (nd6_need_cache(ifp) == 0) 1738 goto sendpkt; 1739 1740 /* 1741 * next hop determination. This routine is derived from ether_output. 1742 */ 1743 1744 /* 1745 * Address resolution or Neighbor Unreachability Detection 1746 * for the next hop. 1747 * At this point, the destination of the packet must be a unicast 1748 * or an anycast address(i.e. not a multicast). 1749 */ 1750 1751 flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; 1752 if (ln == NULL) { 1753 retry: 1754 IF_AFDATA_LOCK(ifp); 1755 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); 1756 IF_AFDATA_UNLOCK(ifp); 1757 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 1758 /* 1759 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 1760 * the condition below is not very efficient. But we believe 1761 * it is tolerable, because this should be a rare case. 1762 */ 1763 flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); 1764 IF_AFDATA_LOCK(ifp); 1765 ln = nd6_lookup(&dst->sin6_addr, flags, ifp); 1766 IF_AFDATA_UNLOCK(ifp); 1767 } 1768 } 1769 if (ln == NULL) { 1770 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 1771 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 1772 char ip6buf[INET6_ADDRSTRLEN]; 1773 log(LOG_DEBUG, 1774 "nd6_output: can't allocate llinfo for %s " 1775 "(ln=%p)\n", 1776 ip6_sprintf(ip6buf, &dst->sin6_addr), ln); 1777 senderr(EIO); /* XXX: good error? */ 1778 } 1779 goto sendpkt; /* send anyway */ 1780 } 1781 1782 /* We don't have to do link-layer address resolution on a p2p link. */ 1783 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 1784 ln->ln_state < ND6_LLINFO_REACHABLE) { 1785 if ((flags & LLE_EXCLUSIVE) == 0) { 1786 flags |= LLE_EXCLUSIVE; 1787 goto retry; 1788 } 1789 ln->ln_state = ND6_LLINFO_STALE; 1790 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 1791 } 1792 1793 /* 1794 * The first time we send a packet to a neighbor whose entry is 1795 * STALE, we have to change the state to DELAY and a sets a timer to 1796 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 1797 * neighbor unreachability detection on expiration. 1798 * (RFC 2461 7.3.3) 1799 */ 1800 if (ln->ln_state == ND6_LLINFO_STALE) { 1801 if ((flags & LLE_EXCLUSIVE) == 0) { 1802 flags |= LLE_EXCLUSIVE; 1803 LLE_RUNLOCK(ln); 1804 goto retry; 1805 } 1806 ln->la_asked = 0; 1807 ln->ln_state = ND6_LLINFO_DELAY; 1808 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1809 } 1810 1811 /* 1812 * If the neighbor cache entry has a state other than INCOMPLETE 1813 * (i.e. its link-layer address is already resolved), just 1814 * send the packet. 1815 */ 1816 if (ln->ln_state > ND6_LLINFO_INCOMPLETE) 1817 goto sendpkt; 1818 1819 /* 1820 * There is a neighbor cache entry, but no ethernet address 1821 * response yet. Append this latest packet to the end of the 1822 * packet queue in the mbuf, unless the number of the packet 1823 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 1824 * the oldest packet in the queue will be removed. 1825 */ 1826 if (ln->ln_state == ND6_LLINFO_NOSTATE) 1827 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1828 1829 if ((flags & LLE_EXCLUSIVE) == 0) { 1830 flags |= LLE_EXCLUSIVE; 1831 LLE_RUNLOCK(ln); 1832 goto retry; 1833 } 1834 if (ln->la_hold) { 1835 struct mbuf *m_hold; 1836 int i; 1837 1838 i = 0; 1839 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { 1840 i++; 1841 if (m_hold->m_nextpkt == NULL) { 1842 m_hold->m_nextpkt = m; 1843 break; 1844 } 1845 } 1846 while (i >= V_nd6_maxqueuelen) { 1847 m_hold = ln->la_hold; 1848 ln->la_hold = ln->la_hold->m_nextpkt; 1849 m_freem(m_hold); 1850 i--; 1851 } 1852 } else { 1853 ln->la_hold = m; 1854 } 1855 /* 1856 * We did the lookup (no lle arg) so we 1857 * need to do the unlock here 1858 */ 1859 if (lle == NULL) { 1860 if (flags & LLE_EXCLUSIVE) 1861 LLE_WUNLOCK(ln); 1862 else 1863 LLE_RUNLOCK(ln); 1864 } 1865 1866 /* 1867 * If there has been no NS for the neighbor after entering the 1868 * INCOMPLETE state, send the first solicitation. 1869 */ 1870 if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { 1871 ln->la_asked++; 1872 1873 nd6_llinfo_settimer(ln, 1874 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 1875 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); 1876 } 1877 return (0); 1878 1879 sendpkt: 1880 /* discard the packet if IPv6 operation is disabled on the interface */ 1881 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1882 error = ENETDOWN; /* better error? */ 1883 goto bad; 1884 } 1885 /* 1886 * ln is valid and the caller did not pass in 1887 * an llentry 1888 */ 1889 if ((ln != NULL) && (lle == NULL)) { 1890 if (flags & LLE_EXCLUSIVE) 1891 LLE_WUNLOCK(ln); 1892 else 1893 LLE_RUNLOCK(ln); 1894 } 1895 1896#ifdef MAC 1897 mac_netinet6_nd6_send(ifp, m); 1898#endif 1899 /* 1900 * We were passed in a pointer to an lle with the lock held 1901 * this means that we can't call if_output as we will 1902 * recurse on the lle lock - so what we do is we create 1903 * a list of mbufs to send and transmit them in the caller 1904 * after the lock is dropped 1905 */ 1906 if (lle != NULL) { 1907 if (*chain == NULL) 1908 *chain = m; 1909 else { 1910 struct mbuf *m = *chain; 1911 1912 /* 1913 * append mbuf to end of deferred chain 1914 */ 1915 while (m->m_nextpkt != NULL) 1916 m = m->m_nextpkt; 1917 m->m_nextpkt = m; 1918 } 1919 return (error); 1920 } 1921 if ((ifp->if_flags & IFF_LOOPBACK) != 0) { 1922 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, 1923 NULL)); 1924 } 1925 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL); 1926 return (error); 1927 1928 bad: 1929 /* 1930 * ln is valid and the caller did not pass in 1931 * an llentry 1932 */ 1933 if ((ln != NULL) && (lle == NULL)) { 1934 if (flags & LLE_EXCLUSIVE) 1935 LLE_WUNLOCK(ln); 1936 else 1937 LLE_RUNLOCK(ln); 1938 } 1939 if (m) 1940 m_freem(m); 1941 return (error); 1942} 1943#undef senderr 1944 1945 1946int 1947nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 1948 struct sockaddr_in6 *dst, struct route *ro) 1949{ 1950 struct mbuf *m, *m_head; 1951 struct ifnet *outifp; 1952 int error = 0; 1953 1954 m_head = chain; 1955 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 1956 outifp = origifp; 1957 else 1958 outifp = ifp; 1959 1960 while (m_head) { 1961 m = m_head; 1962 m_head = m_head->m_nextpkt; 1963 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro); 1964 } 1965 1966 /* 1967 * XXX 1968 * note that intermediate errors are blindly ignored - but this is 1969 * the same convention as used with nd6_output when called by 1970 * nd6_cache_lladdr 1971 */ 1972 return (error); 1973} 1974 1975 1976int 1977nd6_need_cache(struct ifnet *ifp) 1978{ 1979 /* 1980 * XXX: we currently do not make neighbor cache on any interface 1981 * other than ARCnet, Ethernet, FDDI and GIF. 1982 * 1983 * RFC2893 says: 1984 * - unidirectional tunnels needs no ND 1985 */ 1986 switch (ifp->if_type) { 1987 case IFT_ARCNET: 1988 case IFT_ETHER: 1989 case IFT_FDDI: 1990 case IFT_IEEE1394: 1991#ifdef IFT_L2VLAN 1992 case IFT_L2VLAN: 1993#endif 1994#ifdef IFT_IEEE80211 1995 case IFT_IEEE80211: 1996#endif 1997#ifdef IFT_CARP 1998 case IFT_CARP: 1999#endif 2000 case IFT_GIF: /* XXX need more cases? */ 2001 case IFT_PPP: 2002 case IFT_TUNNEL: 2003 case IFT_BRIDGE: 2004 case IFT_PROPVIRTUAL: 2005 return (1); 2006 default: 2007 return (0); 2008 } 2009} 2010 2011/* 2012 * the callers of this function need to be re-worked to drop 2013 * the lle lock, drop here for now 2014 */ 2015int 2016nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2017 struct sockaddr *dst, u_char *desten, struct llentry **lle) 2018{ 2019 struct llentry *ln; 2020 2021 *lle = NULL; 2022 IF_AFDATA_UNLOCK_ASSERT(ifp); 2023 if (m->m_flags & M_MCAST) { 2024 int i; 2025 2026 switch (ifp->if_type) { 2027 case IFT_ETHER: 2028 case IFT_FDDI: 2029#ifdef IFT_L2VLAN 2030 case IFT_L2VLAN: 2031#endif 2032#ifdef IFT_IEEE80211 2033 case IFT_IEEE80211: 2034#endif 2035 case IFT_BRIDGE: 2036 case IFT_ISO88025: 2037 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2038 desten); 2039 return (0); 2040 case IFT_IEEE1394: 2041 /* 2042 * netbsd can use if_broadcastaddr, but we don't do so 2043 * to reduce # of ifdef. 2044 */ 2045 for (i = 0; i < ifp->if_addrlen; i++) 2046 desten[i] = ~0; 2047 return (0); 2048 case IFT_ARCNET: 2049 *desten = 0; 2050 return (0); 2051 default: 2052 m_freem(m); 2053 return (EAFNOSUPPORT); 2054 } 2055 } 2056 2057 2058 /* 2059 * the entry should have been created in nd6_store_lladdr 2060 */ 2061 IF_AFDATA_LOCK(ifp); 2062 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2063 IF_AFDATA_UNLOCK(ifp); 2064 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2065 if (ln != NULL) 2066 LLE_RUNLOCK(ln); 2067 /* this could happen, if we could not allocate memory */ 2068 m_freem(m); 2069 return (1); 2070 } 2071 2072 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2073 *lle = ln; 2074 LLE_RUNLOCK(ln); 2075 /* 2076 * A *small* use after free race exists here 2077 */ 2078 return (0); 2079} 2080 2081static void 2082clear_llinfo_pqueue(struct llentry *ln) 2083{ 2084 struct mbuf *m_hold, *m_hold_next; 2085 2086 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2087 m_hold_next = m_hold->m_nextpkt; 2088 m_hold->m_nextpkt = NULL; 2089 m_freem(m_hold); 2090 } 2091 2092 ln->la_hold = NULL; 2093 return; 2094} 2095 2096static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2097static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2098#ifdef SYSCTL_DECL 2099SYSCTL_DECL(_net_inet6_icmp6); 2100#endif 2101SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2102 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2103SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2104 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2105SYSCTL_V_INT(V_NET, vnet_inet6, _net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, 2106 nd6_maxqueuelen, CTLFLAG_RW, nd6_maxqueuelen, 1, ""); 2107 2108static int 2109nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2110{ 2111 INIT_VNET_INET6(curvnet); 2112 int error; 2113 char buf[1024] __aligned(4); 2114 struct in6_defrouter *d, *de; 2115 struct nd_defrouter *dr; 2116 2117 if (req->newptr) 2118 return EPERM; 2119 error = 0; 2120 2121 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; 2122 dr = TAILQ_NEXT(dr, dr_entry)) { 2123 d = (struct in6_defrouter *)buf; 2124 de = (struct in6_defrouter *)(buf + sizeof(buf)); 2125 2126 if (d + 1 <= de) { 2127 bzero(d, sizeof(*d)); 2128 d->rtaddr.sin6_family = AF_INET6; 2129 d->rtaddr.sin6_len = sizeof(d->rtaddr); 2130 d->rtaddr.sin6_addr = dr->rtaddr; 2131 error = sa6_recoverscope(&d->rtaddr); 2132 if (error != 0) 2133 return (error); 2134 d->flags = dr->flags; 2135 d->rtlifetime = dr->rtlifetime; 2136 d->expire = dr->expire; 2137 d->if_index = dr->ifp->if_index; 2138 } else 2139 panic("buffer too short"); 2140 2141 error = SYSCTL_OUT(req, buf, sizeof(*d)); 2142 if (error) 2143 break; 2144 } 2145 2146 return (error); 2147} 2148 2149static int 2150nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2151{ 2152 INIT_VNET_INET6(curvnet); 2153 int error; 2154 char buf[1024] __aligned(4); 2155 struct in6_prefix *p, *pe; 2156 struct nd_prefix *pr; 2157 char ip6buf[INET6_ADDRSTRLEN]; 2158 2159 if (req->newptr) 2160 return EPERM; 2161 error = 0; 2162 2163 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) { 2164 u_short advrtrs; 2165 size_t advance; 2166 struct sockaddr_in6 *sin6, *s6; 2167 struct nd_pfxrouter *pfr; 2168 2169 p = (struct in6_prefix *)buf; 2170 pe = (struct in6_prefix *)(buf + sizeof(buf)); 2171 2172 if (p + 1 <= pe) { 2173 bzero(p, sizeof(*p)); 2174 sin6 = (struct sockaddr_in6 *)(p + 1); 2175 2176 p->prefix = pr->ndpr_prefix; 2177 if (sa6_recoverscope(&p->prefix)) { 2178 log(LOG_ERR, 2179 "scope error in prefix list (%s)\n", 2180 ip6_sprintf(ip6buf, &p->prefix.sin6_addr)); 2181 /* XXX: press on... */ 2182 } 2183 p->raflags = pr->ndpr_raf; 2184 p->prefixlen = pr->ndpr_plen; 2185 p->vltime = pr->ndpr_vltime; 2186 p->pltime = pr->ndpr_pltime; 2187 p->if_index = pr->ndpr_ifp->if_index; 2188 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2189 p->expire = 0; 2190 else { 2191 time_t maxexpire; 2192 2193 /* XXX: we assume time_t is signed. */ 2194 maxexpire = (-1) & 2195 ~((time_t)1 << 2196 ((sizeof(maxexpire) * 8) - 1)); 2197 if (pr->ndpr_vltime < 2198 maxexpire - pr->ndpr_lastupdate) { 2199 p->expire = pr->ndpr_lastupdate + 2200 pr->ndpr_vltime; 2201 } else 2202 p->expire = maxexpire; 2203 } 2204 p->refcnt = pr->ndpr_refcnt; 2205 p->flags = pr->ndpr_stateflags; 2206 p->origin = PR_ORIG_RA; 2207 advrtrs = 0; 2208 for (pfr = pr->ndpr_advrtrs.lh_first; pfr; 2209 pfr = pfr->pfr_next) { 2210 if ((void *)&sin6[advrtrs + 1] > (void *)pe) { 2211 advrtrs++; 2212 continue; 2213 } 2214 s6 = &sin6[advrtrs]; 2215 bzero(s6, sizeof(*s6)); 2216 s6->sin6_family = AF_INET6; 2217 s6->sin6_len = sizeof(*sin6); 2218 s6->sin6_addr = pfr->router->rtaddr; 2219 if (sa6_recoverscope(s6)) { 2220 log(LOG_ERR, 2221 "scope error in " 2222 "prefix list (%s)\n", 2223 ip6_sprintf(ip6buf, 2224 &pfr->router->rtaddr)); 2225 } 2226 advrtrs++; 2227 } 2228 p->advrtrs = advrtrs; 2229 } else 2230 panic("buffer too short"); 2231 2232 advance = sizeof(*p) + sizeof(*sin6) * advrtrs; 2233 error = SYSCTL_OUT(req, buf, advance); 2234 if (error) 2235 break; 2236 } 2237 2238 return (error); 2239} 2240