1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/sockio.h> 45 #include <sys/time.h> 46 #include <sys/kernel.h> 47 #include <sys/protosw.h> 48 #include <sys/errno.h> 49 #include <sys/syslog.h> 50 #include <sys/lock.h> 51 #include <sys/rwlock.h> 52 #include <sys/queue.h> 53 #include <sys/sdt.h> 54 #include <sys/sysctl.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/if_arc.h> 59 #include <net/if_dl.h> 60 #include <net/if_types.h> 61 #include <net/iso88025.h> 62 #include <net/fddi.h> 63 #include <net/route.h> 64 #include <net/vnet.h> 65 66 #include <netinet/in.h> 67 #include <netinet/in_kdtrace.h> 68 #include <net/if_llatbl.h> 69 #include <netinet/if_ether.h> 70 #include <netinet6/in6_var.h> 71 #include <netinet/ip6.h> 72 #include <netinet6/ip6_var.h> 73 #include <netinet6/scope6_var.h> 74 #include <netinet6/nd6.h> 75 #include <netinet6/in6_ifattach.h> 76 #include <netinet/icmp6.h> 77 #include <netinet6/send.h> 78 79 #include <sys/limits.h> 80 81 #include <security/mac/mac_framework.h> 82 83 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 84 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 85 86 #define SIN6(s) ((const struct sockaddr_in6 *)(s)) 87 88 /* timer values */ 89 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 90 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 91 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 92 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 93 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 94 * local traffic */ 95 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 96 * collection timer */ 97 98 /* preventing too many loops in ND option parsing */ 99 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 100 101 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 102 * layer hints */ 103 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 104 * ND entries */ 105 #define V_nd6_maxndopt VNET(nd6_maxndopt) 106 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 107 108 #ifdef ND6_DEBUG 109 VNET_DEFINE(int, nd6_debug) = 1; 110 #else 111 VNET_DEFINE(int, nd6_debug) = 0; 112 #endif 113 114 static eventhandler_tag lle_event_eh; 115 116 /* for debugging? */ 117 #if 0 118 static int nd6_inuse, nd6_allocated; 119 #endif 120 121 VNET_DEFINE(struct nd_drhead, nd_defrouter); 122 VNET_DEFINE(struct nd_prhead, nd_prefix); 123 124 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 125 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 126 127 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 128 129 static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *, 130 struct ifnet *); 131 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 132 static void nd6_slowtimo(void *); 133 static int regen_tmpaddr(struct in6_ifaddr *); 134 static struct llentry *nd6_free(struct llentry *, int); 135 static void nd6_llinfo_timer(void *); 136 static void clear_llinfo_pqueue(struct llentry *); 137 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 138 static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *, 139 struct sockaddr_in6 *); 140 static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *, 141 struct sockaddr_in6 *); 142 143 static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 144 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 145 146 VNET_DEFINE(struct callout, nd6_timer_ch); 147 148 static void 149 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 150 { 151 struct rt_addrinfo rtinfo; 152 struct sockaddr_in6 dst; 153 struct sockaddr_dl gw; 154 struct ifnet *ifp; 155 int type; 156 157 LLE_WLOCK_ASSERT(lle); 158 159 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 160 return; 161 162 switch (evt) { 163 case LLENTRY_RESOLVED: 164 type = RTM_ADD; 165 KASSERT(lle->la_flags & LLE_VALID, 166 ("%s: %p resolved but not valid?", __func__, lle)); 167 break; 168 case LLENTRY_EXPIRED: 169 type = RTM_DELETE; 170 break; 171 default: 172 return; 173 } 174 175 ifp = lltable_get_ifp(lle->lle_tbl); 176 177 bzero(&dst, sizeof(dst)); 178 bzero(&gw, sizeof(gw)); 179 bzero(&rtinfo, sizeof(rtinfo)); 180 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 181 dst.sin6_scope_id = in6_getscopezone(ifp, 182 in6_addrscope(&dst.sin6_addr)); 183 gw.sdl_len = sizeof(struct sockaddr_dl); 184 gw.sdl_family = AF_LINK; 185 gw.sdl_alen = ifp->if_addrlen; 186 gw.sdl_index = ifp->if_index; 187 gw.sdl_type = ifp->if_type; 188 if (evt == LLENTRY_RESOLVED) 189 bcopy(&lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 190 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 191 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 192 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 193 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 194 type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB); 195 } 196 197 void 198 nd6_init(void) 199 { 200 201 LIST_INIT(&V_nd_prefix); 202 203 /* initialization of the default router list */ 204 TAILQ_INIT(&V_nd_defrouter); 205 206 /* start timer */ 207 callout_init(&V_nd6_slowtimo_ch, 0); 208 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 209 nd6_slowtimo, curvnet); 210 211 nd6_dad_init(); 212 if (IS_DEFAULT_VNET(curvnet)) 213 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 214 NULL, EVENTHANDLER_PRI_ANY); 215 } 216 217 #ifdef VIMAGE 218 void 219 nd6_destroy() 220 { 221 222 callout_drain(&V_nd6_slowtimo_ch); 223 callout_drain(&V_nd6_timer_ch); 224 if (IS_DEFAULT_VNET(curvnet)) 225 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 226 } 227 #endif 228 229 struct nd_ifinfo * 230 nd6_ifattach(struct ifnet *ifp) 231 { 232 struct nd_ifinfo *nd; 233 234 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 235 nd->initialized = 1; 236 237 nd->chlim = IPV6_DEFHLIM; 238 nd->basereachable = REACHABLE_TIME; 239 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 240 nd->retrans = RETRANS_TIMER; 241 242 nd->flags = ND6_IFF_PERFORMNUD; 243 244 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 245 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 246 * default regardless of the V_ip6_auto_linklocal configuration to 247 * give a reasonable default behavior. 248 */ 249 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 250 (ifp->if_flags & IFF_LOOPBACK)) 251 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 252 /* 253 * A loopback interface does not need to accept RTADV. 254 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 255 * default regardless of the V_ip6_accept_rtadv configuration to 256 * prevent the interface from accepting RA messages arrived 257 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 258 */ 259 if (V_ip6_accept_rtadv && 260 !(ifp->if_flags & IFF_LOOPBACK) && 261 (ifp->if_type != IFT_BRIDGE)) 262 nd->flags |= ND6_IFF_ACCEPT_RTADV; 263 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 264 nd->flags |= ND6_IFF_NO_RADR; 265 266 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 267 nd6_setmtu0(ifp, nd); 268 269 return nd; 270 } 271 272 void 273 nd6_ifdetach(struct nd_ifinfo *nd) 274 { 275 276 free(nd, M_IP6NDP); 277 } 278 279 /* 280 * Reset ND level link MTU. This function is called when the physical MTU 281 * changes, which means we might have to adjust the ND level MTU. 282 */ 283 void 284 nd6_setmtu(struct ifnet *ifp) 285 { 286 287 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 288 } 289 290 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 291 void 292 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 293 { 294 u_int32_t omaxmtu; 295 296 omaxmtu = ndi->maxmtu; 297 298 switch (ifp->if_type) { 299 case IFT_ARCNET: 300 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 301 break; 302 case IFT_FDDI: 303 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 304 break; 305 case IFT_ISO88025: 306 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 307 break; 308 default: 309 ndi->maxmtu = ifp->if_mtu; 310 break; 311 } 312 313 /* 314 * Decreasing the interface MTU under IPV6 minimum MTU may cause 315 * undesirable situation. We thus notify the operator of the change 316 * explicitly. The check for omaxmtu is necessary to restrict the 317 * log to the case of changing the MTU, not initializing it. 318 */ 319 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 320 log(LOG_NOTICE, "nd6_setmtu0: " 321 "new link MTU on %s (%lu) is too small for IPv6\n", 322 if_name(ifp), (unsigned long)ndi->maxmtu); 323 } 324 325 if (ndi->maxmtu > V_in6_maxmtu) 326 in6_setmaxmtu(); /* check all interfaces just in case */ 327 328 } 329 330 void 331 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 332 { 333 334 bzero(ndopts, sizeof(*ndopts)); 335 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 336 ndopts->nd_opts_last 337 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 338 339 if (icmp6len == 0) { 340 ndopts->nd_opts_done = 1; 341 ndopts->nd_opts_search = NULL; 342 } 343 } 344 345 /* 346 * Take one ND option. 347 */ 348 struct nd_opt_hdr * 349 nd6_option(union nd_opts *ndopts) 350 { 351 struct nd_opt_hdr *nd_opt; 352 int olen; 353 354 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 355 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 356 __func__)); 357 if (ndopts->nd_opts_search == NULL) 358 return NULL; 359 if (ndopts->nd_opts_done) 360 return NULL; 361 362 nd_opt = ndopts->nd_opts_search; 363 364 /* make sure nd_opt_len is inside the buffer */ 365 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 366 bzero(ndopts, sizeof(*ndopts)); 367 return NULL; 368 } 369 370 olen = nd_opt->nd_opt_len << 3; 371 if (olen == 0) { 372 /* 373 * Message validation requires that all included 374 * options have a length that is greater than zero. 375 */ 376 bzero(ndopts, sizeof(*ndopts)); 377 return NULL; 378 } 379 380 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 381 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 382 /* option overruns the end of buffer, invalid */ 383 bzero(ndopts, sizeof(*ndopts)); 384 return NULL; 385 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 386 /* reached the end of options chain */ 387 ndopts->nd_opts_done = 1; 388 ndopts->nd_opts_search = NULL; 389 } 390 return nd_opt; 391 } 392 393 /* 394 * Parse multiple ND options. 395 * This function is much easier to use, for ND routines that do not need 396 * multiple options of the same type. 397 */ 398 int 399 nd6_options(union nd_opts *ndopts) 400 { 401 struct nd_opt_hdr *nd_opt; 402 int i = 0; 403 404 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 405 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 406 __func__)); 407 if (ndopts->nd_opts_search == NULL) 408 return 0; 409 410 while (1) { 411 nd_opt = nd6_option(ndopts); 412 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 413 /* 414 * Message validation requires that all included 415 * options have a length that is greater than zero. 416 */ 417 ICMP6STAT_INC(icp6s_nd_badopt); 418 bzero(ndopts, sizeof(*ndopts)); 419 return -1; 420 } 421 422 if (nd_opt == NULL) 423 goto skip1; 424 425 switch (nd_opt->nd_opt_type) { 426 case ND_OPT_SOURCE_LINKADDR: 427 case ND_OPT_TARGET_LINKADDR: 428 case ND_OPT_MTU: 429 case ND_OPT_REDIRECTED_HEADER: 430 case ND_OPT_NONCE: 431 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 432 nd6log((LOG_INFO, 433 "duplicated ND6 option found (type=%d)\n", 434 nd_opt->nd_opt_type)); 435 /* XXX bark? */ 436 } else { 437 ndopts->nd_opt_array[nd_opt->nd_opt_type] 438 = nd_opt; 439 } 440 break; 441 case ND_OPT_PREFIX_INFORMATION: 442 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 443 ndopts->nd_opt_array[nd_opt->nd_opt_type] 444 = nd_opt; 445 } 446 ndopts->nd_opts_pi_end = 447 (struct nd_opt_prefix_info *)nd_opt; 448 break; 449 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 450 case ND_OPT_RDNSS: /* RFC 6106 */ 451 case ND_OPT_DNSSL: /* RFC 6106 */ 452 /* 453 * Silently ignore options we know and do not care about 454 * in the kernel. 455 */ 456 break; 457 default: 458 /* 459 * Unknown options must be silently ignored, 460 * to accomodate future extension to the protocol. 461 */ 462 nd6log((LOG_DEBUG, 463 "nd6_options: unsupported option %d - " 464 "option ignored\n", nd_opt->nd_opt_type)); 465 } 466 467 skip1: 468 i++; 469 if (i > V_nd6_maxndopt) { 470 ICMP6STAT_INC(icp6s_nd_toomanyopt); 471 nd6log((LOG_INFO, "too many loop in nd opt\n")); 472 break; 473 } 474 475 if (ndopts->nd_opts_done) 476 break; 477 } 478 479 return 0; 480 } 481 482 /* 483 * ND6 timer routine to handle ND6 entries 484 */ 485 void 486 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 487 { 488 int canceled; 489 490 LLE_WLOCK_ASSERT(ln); 491 492 if (tick < 0) { 493 ln->la_expire = 0; 494 ln->ln_ntick = 0; 495 canceled = callout_stop(&ln->lle_timer); 496 } else { 497 ln->la_expire = time_uptime + tick / hz; 498 LLE_ADDREF(ln); 499 if (tick > INT_MAX) { 500 ln->ln_ntick = tick - INT_MAX; 501 canceled = callout_reset(&ln->lle_timer, INT_MAX, 502 nd6_llinfo_timer, ln); 503 } else { 504 ln->ln_ntick = 0; 505 canceled = callout_reset(&ln->lle_timer, tick, 506 nd6_llinfo_timer, ln); 507 } 508 } 509 if (canceled) 510 LLE_REMREF(ln); 511 } 512 513 /* 514 * Gets source address of the first packet in hold queue 515 * and stores it in @src. 516 * Returns pointer to @src (if hold queue is not empty) or NULL. 517 * 518 */ 519 static struct in6_addr * 520 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 521 { 522 struct ip6_hdr hdr; 523 struct mbuf *m; 524 525 if (ln->la_hold == NULL) 526 return (NULL); 527 528 /* 529 * assume every packet in la_hold has the same IP header 530 */ 531 m = ln->la_hold; 532 if (sizeof(hdr) < m->m_len) 533 return (NULL); 534 535 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 536 *src = hdr.ip6_src; 537 538 return (src); 539 } 540 541 void 542 nd6_llinfo_settimer(struct llentry *ln, long tick) 543 { 544 545 LLE_WLOCK(ln); 546 nd6_llinfo_settimer_locked(ln, tick); 547 LLE_WUNLOCK(ln); 548 } 549 550 static void 551 nd6_llinfo_timer(void *arg) 552 { 553 struct llentry *ln; 554 struct in6_addr *dst, *pdst, *psrc, src; 555 struct ifnet *ifp; 556 struct nd_ifinfo *ndi = NULL; 557 int send_ns; 558 559 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 560 ln = (struct llentry *)arg; 561 LLE_WLOCK(ln); 562 if (callout_pending(&ln->lle_timer)) { 563 /* 564 * Here we are a bit odd here in the treatment of 565 * active/pending. If the pending bit is set, it got 566 * rescheduled before I ran. The active 567 * bit we ignore, since if it was stopped 568 * in ll_tablefree() and was currently running 569 * it would have return 0 so the code would 570 * not have deleted it since the callout could 571 * not be stopped so we want to go through 572 * with the delete here now. If the callout 573 * was restarted, the pending bit will be back on and 574 * we just want to bail since the callout_reset would 575 * return 1 and our reference would have been removed 576 * by nd6_llinfo_settimer_locked above since canceled 577 * would have been 1. 578 */ 579 LLE_WUNLOCK(ln); 580 return; 581 } 582 ifp = ln->lle_tbl->llt_ifp; 583 CURVNET_SET(ifp->if_vnet); 584 ndi = ND_IFINFO(ifp); 585 send_ns = 0; 586 dst = &ln->r_l3addr.addr6; 587 pdst = dst; 588 589 if (ln->ln_ntick > 0) { 590 if (ln->ln_ntick > INT_MAX) { 591 ln->ln_ntick -= INT_MAX; 592 nd6_llinfo_settimer_locked(ln, INT_MAX); 593 } else { 594 ln->ln_ntick = 0; 595 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 596 } 597 goto done; 598 } 599 600 if (ln->la_flags & LLE_STATIC) { 601 goto done; 602 } 603 604 if (ln->la_flags & LLE_DELETED) { 605 (void)nd6_free(ln, 0); 606 ln = NULL; 607 goto done; 608 } 609 610 switch (ln->ln_state) { 611 case ND6_LLINFO_INCOMPLETE: 612 if (ln->la_asked < V_nd6_mmaxtries) { 613 ln->la_asked++; 614 send_ns = 1; 615 /* Send NS to multicast address */ 616 pdst = NULL; 617 } else { 618 struct mbuf *m = ln->la_hold; 619 if (m) { 620 struct mbuf *m0; 621 622 /* 623 * assuming every packet in la_hold has the 624 * same IP header. Send error after unlock. 625 */ 626 m0 = m->m_nextpkt; 627 m->m_nextpkt = NULL; 628 ln->la_hold = m0; 629 clear_llinfo_pqueue(ln); 630 } 631 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT); 632 (void)nd6_free(ln, 0); 633 ln = NULL; 634 if (m != NULL) 635 icmp6_error2(m, ICMP6_DST_UNREACH, 636 ICMP6_DST_UNREACH_ADDR, 0, ifp); 637 } 638 break; 639 case ND6_LLINFO_REACHABLE: 640 if (!ND6_LLINFO_PERMANENT(ln)) { 641 ln->ln_state = ND6_LLINFO_STALE; 642 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 643 } 644 break; 645 646 case ND6_LLINFO_STALE: 647 /* Garbage Collection(RFC 2461 5.3) */ 648 if (!ND6_LLINFO_PERMANENT(ln)) { 649 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 650 (void)nd6_free(ln, 1); 651 ln = NULL; 652 } 653 break; 654 655 case ND6_LLINFO_DELAY: 656 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 657 /* We need NUD */ 658 ln->la_asked = 1; 659 ln->ln_state = ND6_LLINFO_PROBE; 660 send_ns = 1; 661 } else { 662 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 663 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 664 } 665 break; 666 case ND6_LLINFO_PROBE: 667 if (ln->la_asked < V_nd6_umaxtries) { 668 ln->la_asked++; 669 send_ns = 1; 670 } else { 671 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 672 (void)nd6_free(ln, 0); 673 ln = NULL; 674 } 675 break; 676 default: 677 panic("%s: paths in a dark night can be confusing: %d", 678 __func__, ln->ln_state); 679 } 680 done: 681 if (send_ns != 0) { 682 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 683 psrc = nd6_llinfo_get_holdsrc(ln, &src); 684 LLE_FREE_LOCKED(ln); 685 ln = NULL; 686 nd6_ns_output(ifp, psrc, pdst, dst, NULL); 687 } 688 689 if (ln != NULL) 690 LLE_FREE_LOCKED(ln); 691 CURVNET_RESTORE(); 692 } 693 694 695 /* 696 * ND6 timer routine to expire default route list and prefix list 697 */ 698 void 699 nd6_timer(void *arg) 700 { 701 CURVNET_SET((struct vnet *) arg); 702 struct nd_defrouter *dr, *ndr; 703 struct nd_prefix *pr, *npr; 704 struct in6_ifaddr *ia6, *nia6; 705 706 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 707 nd6_timer, curvnet); 708 709 /* expire default router list */ 710 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 711 if (dr->expire && dr->expire < time_uptime) 712 defrtrlist_del(dr); 713 } 714 715 /* 716 * expire interface addresses. 717 * in the past the loop was inside prefix expiry processing. 718 * However, from a stricter speci-confrmance standpoint, we should 719 * rather separate address lifetimes and prefix lifetimes. 720 * 721 * XXXRW: in6_ifaddrhead locking. 722 */ 723 addrloop: 724 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 725 /* check address lifetime */ 726 if (IFA6_IS_INVALID(ia6)) { 727 int regen = 0; 728 729 /* 730 * If the expiring address is temporary, try 731 * regenerating a new one. This would be useful when 732 * we suspended a laptop PC, then turned it on after a 733 * period that could invalidate all temporary 734 * addresses. Although we may have to restart the 735 * loop (see below), it must be after purging the 736 * address. Otherwise, we'd see an infinite loop of 737 * regeneration. 738 */ 739 if (V_ip6_use_tempaddr && 740 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 741 if (regen_tmpaddr(ia6) == 0) 742 regen = 1; 743 } 744 745 in6_purgeaddr(&ia6->ia_ifa); 746 747 if (regen) 748 goto addrloop; /* XXX: see below */ 749 } else if (IFA6_IS_DEPRECATED(ia6)) { 750 int oldflags = ia6->ia6_flags; 751 752 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 753 754 /* 755 * If a temporary address has just become deprecated, 756 * regenerate a new one if possible. 757 */ 758 if (V_ip6_use_tempaddr && 759 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 760 (oldflags & IN6_IFF_DEPRECATED) == 0) { 761 762 if (regen_tmpaddr(ia6) == 0) { 763 /* 764 * A new temporary address is 765 * generated. 766 * XXX: this means the address chain 767 * has changed while we are still in 768 * the loop. Although the change 769 * would not cause disaster (because 770 * it's not a deletion, but an 771 * addition,) we'd rather restart the 772 * loop just for safety. Or does this 773 * significantly reduce performance?? 774 */ 775 goto addrloop; 776 } 777 } 778 } else { 779 /* 780 * A new RA might have made a deprecated address 781 * preferred. 782 */ 783 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 784 } 785 } 786 787 /* expire prefix list */ 788 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 789 /* 790 * check prefix lifetime. 791 * since pltime is just for autoconf, pltime processing for 792 * prefix is not necessary. 793 */ 794 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 795 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { 796 797 /* 798 * address expiration and prefix expiration are 799 * separate. NEVER perform in6_purgeaddr here. 800 */ 801 prelist_remove(pr); 802 } 803 } 804 CURVNET_RESTORE(); 805 } 806 807 /* 808 * ia6 - deprecated/invalidated temporary address 809 */ 810 static int 811 regen_tmpaddr(struct in6_ifaddr *ia6) 812 { 813 struct ifaddr *ifa; 814 struct ifnet *ifp; 815 struct in6_ifaddr *public_ifa6 = NULL; 816 817 ifp = ia6->ia_ifa.ifa_ifp; 818 IF_ADDR_RLOCK(ifp); 819 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 820 struct in6_ifaddr *it6; 821 822 if (ifa->ifa_addr->sa_family != AF_INET6) 823 continue; 824 825 it6 = (struct in6_ifaddr *)ifa; 826 827 /* ignore no autoconf addresses. */ 828 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 829 continue; 830 831 /* ignore autoconf addresses with different prefixes. */ 832 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 833 continue; 834 835 /* 836 * Now we are looking at an autoconf address with the same 837 * prefix as ours. If the address is temporary and is still 838 * preferred, do not create another one. It would be rare, but 839 * could happen, for example, when we resume a laptop PC after 840 * a long period. 841 */ 842 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 843 !IFA6_IS_DEPRECATED(it6)) { 844 public_ifa6 = NULL; 845 break; 846 } 847 848 /* 849 * This is a public autoconf address that has the same prefix 850 * as ours. If it is preferred, keep it. We can't break the 851 * loop here, because there may be a still-preferred temporary 852 * address with the prefix. 853 */ 854 if (!IFA6_IS_DEPRECATED(it6)) 855 public_ifa6 = it6; 856 } 857 if (public_ifa6 != NULL) 858 ifa_ref(&public_ifa6->ia_ifa); 859 IF_ADDR_RUNLOCK(ifp); 860 861 if (public_ifa6 != NULL) { 862 int e; 863 864 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 865 ifa_free(&public_ifa6->ia_ifa); 866 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 867 " tmp addr,errno=%d\n", e); 868 return (-1); 869 } 870 ifa_free(&public_ifa6->ia_ifa); 871 return (0); 872 } 873 874 return (-1); 875 } 876 877 /* 878 * Nuke neighbor cache/prefix/default router management table, right before 879 * ifp goes away. 880 */ 881 void 882 nd6_purge(struct ifnet *ifp) 883 { 884 struct nd_defrouter *dr, *ndr; 885 struct nd_prefix *pr, *npr; 886 887 /* 888 * Nuke default router list entries toward ifp. 889 * We defer removal of default router list entries that is installed 890 * in the routing table, in order to keep additional side effects as 891 * small as possible. 892 */ 893 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 894 if (dr->installed) 895 continue; 896 897 if (dr->ifp == ifp) 898 defrtrlist_del(dr); 899 } 900 901 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 902 if (!dr->installed) 903 continue; 904 905 if (dr->ifp == ifp) 906 defrtrlist_del(dr); 907 } 908 909 /* Nuke prefix list entries toward ifp */ 910 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 911 if (pr->ndpr_ifp == ifp) { 912 /* 913 * Because if_detach() does *not* release prefixes 914 * while purging addresses the reference count will 915 * still be above zero. We therefore reset it to 916 * make sure that the prefix really gets purged. 917 */ 918 pr->ndpr_refcnt = 0; 919 920 /* 921 * Previously, pr->ndpr_addr is removed as well, 922 * but I strongly believe we don't have to do it. 923 * nd6_purge() is only called from in6_ifdetach(), 924 * which removes all the associated interface addresses 925 * by itself. 926 * (jinmei@kame.net 20010129) 927 */ 928 prelist_remove(pr); 929 } 930 } 931 932 /* cancel default outgoing interface setting */ 933 if (V_nd6_defifindex == ifp->if_index) 934 nd6_setdefaultiface(0); 935 936 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 937 /* Refresh default router list. */ 938 defrouter_select(); 939 } 940 941 /* XXXXX 942 * We do not nuke the neighbor cache entries here any more 943 * because the neighbor cache is kept in if_afdata[AF_INET6]. 944 * nd6_purge() is invoked by in6_ifdetach() which is called 945 * from if_detach() where everything gets purged. So let 946 * in6_domifdetach() do the actual L2 table purging work. 947 */ 948 } 949 950 /* 951 * the caller acquires and releases the lock on the lltbls 952 * Returns the llentry locked 953 */ 954 struct llentry * 955 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 956 { 957 struct sockaddr_in6 sin6; 958 struct llentry *ln; 959 int llflags; 960 961 bzero(&sin6, sizeof(sin6)); 962 sin6.sin6_len = sizeof(struct sockaddr_in6); 963 sin6.sin6_family = AF_INET6; 964 sin6.sin6_addr = *addr6; 965 966 IF_AFDATA_LOCK_ASSERT(ifp); 967 968 llflags = (flags & ND6_EXCLUSIVE) ? LLE_EXCLUSIVE : 0; 969 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 970 971 return (ln); 972 } 973 974 struct llentry * 975 nd6_alloc(struct in6_addr *addr6, int flags, struct ifnet *ifp) 976 { 977 struct sockaddr_in6 sin6; 978 struct llentry *ln; 979 980 bzero(&sin6, sizeof(sin6)); 981 sin6.sin6_len = sizeof(struct sockaddr_in6); 982 sin6.sin6_family = AF_INET6; 983 sin6.sin6_addr = *addr6; 984 985 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 986 if (ln != NULL) 987 ln->ln_state = ND6_LLINFO_NOSTATE; 988 989 return (ln); 990 } 991 992 /* 993 * Test whether a given IPv6 address is a neighbor or not, ignoring 994 * the actual neighbor cache. The neighbor cache is ignored in order 995 * to not reenter the routing code from within itself. 996 */ 997 static int 998 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 999 { 1000 struct nd_prefix *pr; 1001 struct ifaddr *dstaddr; 1002 1003 /* 1004 * A link-local address is always a neighbor. 1005 * XXX: a link does not necessarily specify a single interface. 1006 */ 1007 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 1008 struct sockaddr_in6 sin6_copy; 1009 u_int32_t zone; 1010 1011 /* 1012 * We need sin6_copy since sa6_recoverscope() may modify the 1013 * content (XXX). 1014 */ 1015 sin6_copy = *addr; 1016 if (sa6_recoverscope(&sin6_copy)) 1017 return (0); /* XXX: should be impossible */ 1018 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 1019 return (0); 1020 if (sin6_copy.sin6_scope_id == zone) 1021 return (1); 1022 else 1023 return (0); 1024 } 1025 1026 /* 1027 * If the address matches one of our addresses, 1028 * it should be a neighbor. 1029 * If the address matches one of our on-link prefixes, it should be a 1030 * neighbor. 1031 */ 1032 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1033 if (pr->ndpr_ifp != ifp) 1034 continue; 1035 1036 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 1037 struct rtentry *rt; 1038 1039 /* Always use the default FIB here. */ 1040 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 1041 0, 0, RT_DEFAULT_FIB); 1042 if (rt == NULL) 1043 continue; 1044 /* 1045 * This is the case where multiple interfaces 1046 * have the same prefix, but only one is installed 1047 * into the routing table and that prefix entry 1048 * is not the one being examined here. In the case 1049 * where RADIX_MPATH is enabled, multiple route 1050 * entries (of the same rt_key value) will be 1051 * installed because the interface addresses all 1052 * differ. 1053 */ 1054 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1055 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 1056 RTFREE_LOCKED(rt); 1057 continue; 1058 } 1059 RTFREE_LOCKED(rt); 1060 } 1061 1062 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1063 &addr->sin6_addr, &pr->ndpr_mask)) 1064 return (1); 1065 } 1066 1067 /* 1068 * If the address is assigned on the node of the other side of 1069 * a p2p interface, the address should be a neighbor. 1070 */ 1071 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr, RT_ALL_FIBS); 1072 if (dstaddr != NULL) { 1073 if (dstaddr->ifa_ifp == ifp) { 1074 ifa_free(dstaddr); 1075 return (1); 1076 } 1077 ifa_free(dstaddr); 1078 } 1079 1080 /* 1081 * If the default router list is empty, all addresses are regarded 1082 * as on-link, and thus, as a neighbor. 1083 */ 1084 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1085 TAILQ_EMPTY(&V_nd_defrouter) && 1086 V_nd6_defifindex == ifp->if_index) { 1087 return (1); 1088 } 1089 1090 return (0); 1091 } 1092 1093 1094 /* 1095 * Detect if a given IPv6 address identifies a neighbor on a given link. 1096 * XXX: should take care of the destination of a p2p link? 1097 */ 1098 int 1099 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1100 { 1101 struct llentry *lle; 1102 int rc = 0; 1103 1104 IF_AFDATA_UNLOCK_ASSERT(ifp); 1105 if (nd6_is_new_addr_neighbor(addr, ifp)) 1106 return (1); 1107 1108 /* 1109 * Even if the address matches none of our addresses, it might be 1110 * in the neighbor cache. 1111 */ 1112 IF_AFDATA_RLOCK(ifp); 1113 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1114 LLE_RUNLOCK(lle); 1115 rc = 1; 1116 } 1117 IF_AFDATA_RUNLOCK(ifp); 1118 return (rc); 1119 } 1120 1121 /* 1122 * Free an nd6 llinfo entry. 1123 * Since the function would cause significant changes in the kernel, DO NOT 1124 * make it global, unless you have a strong reason for the change, and are sure 1125 * that the change is safe. 1126 */ 1127 static struct llentry * 1128 nd6_free(struct llentry *ln, int gc) 1129 { 1130 struct llentry *next; 1131 struct nd_defrouter *dr; 1132 struct ifnet *ifp; 1133 1134 LLE_WLOCK_ASSERT(ln); 1135 1136 /* 1137 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1138 * even though it is not harmful, it was not really necessary. 1139 */ 1140 1141 /* cancel timer */ 1142 nd6_llinfo_settimer_locked(ln, -1); 1143 1144 ifp = ln->lle_tbl->llt_ifp; 1145 1146 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1147 dr = defrouter_lookup(&ln->r_l3addr.addr6, ifp); 1148 1149 if (dr != NULL && dr->expire && 1150 ln->ln_state == ND6_LLINFO_STALE && gc) { 1151 /* 1152 * If the reason for the deletion is just garbage 1153 * collection, and the neighbor is an active default 1154 * router, do not delete it. Instead, reset the GC 1155 * timer using the router's lifetime. 1156 * Simply deleting the entry would affect default 1157 * router selection, which is not necessarily a good 1158 * thing, especially when we're using router preference 1159 * values. 1160 * XXX: the check for ln_state would be redundant, 1161 * but we intentionally keep it just in case. 1162 */ 1163 if (dr->expire > time_uptime) 1164 nd6_llinfo_settimer_locked(ln, 1165 (dr->expire - time_uptime) * hz); 1166 else 1167 nd6_llinfo_settimer_locked(ln, 1168 (long)V_nd6_gctimer * hz); 1169 1170 next = LIST_NEXT(ln, lle_next); 1171 LLE_REMREF(ln); 1172 LLE_WUNLOCK(ln); 1173 return (next); 1174 } 1175 1176 if (dr) { 1177 /* 1178 * Unreachablity of a router might affect the default 1179 * router selection and on-link detection of advertised 1180 * prefixes. 1181 */ 1182 1183 /* 1184 * Temporarily fake the state to choose a new default 1185 * router and to perform on-link determination of 1186 * prefixes correctly. 1187 * Below the state will be set correctly, 1188 * or the entry itself will be deleted. 1189 */ 1190 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1191 } 1192 1193 if (ln->ln_router || dr) { 1194 1195 /* 1196 * We need to unlock to avoid a LOR with rt6_flush() with the 1197 * rnh and for the calls to pfxlist_onlink_check() and 1198 * defrouter_select() in the block further down for calls 1199 * into nd6_lookup(). We still hold a ref. 1200 */ 1201 LLE_WUNLOCK(ln); 1202 1203 /* 1204 * rt6_flush must be called whether or not the neighbor 1205 * is in the Default Router List. 1206 * See a corresponding comment in nd6_na_input(). 1207 */ 1208 rt6_flush(&ln->r_l3addr.addr6, ifp); 1209 } 1210 1211 if (dr) { 1212 /* 1213 * Since defrouter_select() does not affect the 1214 * on-link determination and MIP6 needs the check 1215 * before the default router selection, we perform 1216 * the check now. 1217 */ 1218 pfxlist_onlink_check(); 1219 1220 /* 1221 * Refresh default router list. 1222 */ 1223 defrouter_select(); 1224 } 1225 1226 if (ln->ln_router || dr) 1227 LLE_WLOCK(ln); 1228 } 1229 1230 /* 1231 * Before deleting the entry, remember the next entry as the 1232 * return value. We need this because pfxlist_onlink_check() above 1233 * might have freed other entries (particularly the old next entry) as 1234 * a side effect (XXX). 1235 */ 1236 next = LIST_NEXT(ln, lle_next); 1237 1238 /* 1239 * Save to unlock. We still hold an extra reference and will not 1240 * free(9) in llentry_free() if someone else holds one as well. 1241 */ 1242 LLE_WUNLOCK(ln); 1243 IF_AFDATA_LOCK(ifp); 1244 LLE_WLOCK(ln); 1245 1246 /* Guard against race with other llentry_free(). */ 1247 if (ln->la_flags & LLE_LINKED) { 1248 LLE_REMREF(ln); 1249 llentry_free(ln); 1250 } else 1251 LLE_FREE_LOCKED(ln); 1252 1253 IF_AFDATA_UNLOCK(ifp); 1254 1255 return (next); 1256 } 1257 1258 /* 1259 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1260 * 1261 * XXX cost-effective methods? 1262 */ 1263 void 1264 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1265 { 1266 struct llentry *ln; 1267 struct ifnet *ifp; 1268 1269 if ((dst6 == NULL) || (rt == NULL)) 1270 return; 1271 1272 ifp = rt->rt_ifp; 1273 IF_AFDATA_RLOCK(ifp); 1274 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1275 IF_AFDATA_RUNLOCK(ifp); 1276 if (ln == NULL) 1277 return; 1278 1279 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1280 goto done; 1281 1282 /* 1283 * if we get upper-layer reachability confirmation many times, 1284 * it is possible we have false information. 1285 */ 1286 if (!force) { 1287 ln->ln_byhint++; 1288 if (ln->ln_byhint > V_nd6_maxnudhint) { 1289 goto done; 1290 } 1291 } 1292 1293 ln->ln_state = ND6_LLINFO_REACHABLE; 1294 if (!ND6_LLINFO_PERMANENT(ln)) { 1295 nd6_llinfo_settimer_locked(ln, 1296 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1297 } 1298 done: 1299 LLE_WUNLOCK(ln); 1300 } 1301 1302 1303 /* 1304 * Rejuvenate this function for routing operations related 1305 * processing. 1306 */ 1307 void 1308 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1309 { 1310 struct sockaddr_in6 *gateway; 1311 struct nd_defrouter *dr; 1312 struct ifnet *ifp; 1313 1314 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1315 ifp = rt->rt_ifp; 1316 1317 switch (req) { 1318 case RTM_ADD: 1319 break; 1320 1321 case RTM_DELETE: 1322 if (!ifp) 1323 return; 1324 /* 1325 * Only indirect routes are interesting. 1326 */ 1327 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1328 return; 1329 /* 1330 * check for default route 1331 */ 1332 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1333 &SIN6(rt_key(rt))->sin6_addr)) { 1334 1335 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1336 if (dr != NULL) 1337 dr->installed = 0; 1338 } 1339 break; 1340 } 1341 } 1342 1343 1344 int 1345 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1346 { 1347 struct in6_drlist *drl = (struct in6_drlist *)data; 1348 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1349 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1350 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1351 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1352 struct nd_defrouter *dr; 1353 struct nd_prefix *pr; 1354 int i = 0, error = 0; 1355 1356 if (ifp->if_afdata[AF_INET6] == NULL) 1357 return (EPFNOSUPPORT); 1358 switch (cmd) { 1359 case SIOCGDRLST_IN6: 1360 /* 1361 * obsolete API, use sysctl under net.inet6.icmp6 1362 */ 1363 bzero(drl, sizeof(*drl)); 1364 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1365 if (i >= DRLSTSIZ) 1366 break; 1367 drl->defrouter[i].rtaddr = dr->rtaddr; 1368 in6_clearscope(&drl->defrouter[i].rtaddr); 1369 1370 drl->defrouter[i].flags = dr->flags; 1371 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1372 drl->defrouter[i].expire = dr->expire + 1373 (time_second - time_uptime); 1374 drl->defrouter[i].if_index = dr->ifp->if_index; 1375 i++; 1376 } 1377 break; 1378 case SIOCGPRLST_IN6: 1379 /* 1380 * obsolete API, use sysctl under net.inet6.icmp6 1381 * 1382 * XXX the structure in6_prlist was changed in backward- 1383 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1384 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1385 */ 1386 /* 1387 * XXX meaning of fields, especialy "raflags", is very 1388 * differnet between RA prefix list and RR/static prefix list. 1389 * how about separating ioctls into two? 1390 */ 1391 bzero(oprl, sizeof(*oprl)); 1392 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1393 struct nd_pfxrouter *pfr; 1394 int j; 1395 1396 if (i >= PRLSTSIZ) 1397 break; 1398 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1399 oprl->prefix[i].raflags = pr->ndpr_raf; 1400 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1401 oprl->prefix[i].vltime = pr->ndpr_vltime; 1402 oprl->prefix[i].pltime = pr->ndpr_pltime; 1403 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1404 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1405 oprl->prefix[i].expire = 0; 1406 else { 1407 time_t maxexpire; 1408 1409 /* XXX: we assume time_t is signed. */ 1410 maxexpire = (-1) & 1411 ~((time_t)1 << 1412 ((sizeof(maxexpire) * 8) - 1)); 1413 if (pr->ndpr_vltime < 1414 maxexpire - pr->ndpr_lastupdate) { 1415 oprl->prefix[i].expire = 1416 pr->ndpr_lastupdate + 1417 pr->ndpr_vltime + 1418 (time_second - time_uptime); 1419 } else 1420 oprl->prefix[i].expire = maxexpire; 1421 } 1422 1423 j = 0; 1424 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1425 if (j < DRLSTSIZ) { 1426 #define RTRADDR oprl->prefix[i].advrtr[j] 1427 RTRADDR = pfr->router->rtaddr; 1428 in6_clearscope(&RTRADDR); 1429 #undef RTRADDR 1430 } 1431 j++; 1432 } 1433 oprl->prefix[i].advrtrs = j; 1434 oprl->prefix[i].origin = PR_ORIG_RA; 1435 1436 i++; 1437 } 1438 1439 break; 1440 case OSIOCGIFINFO_IN6: 1441 #define ND ndi->ndi 1442 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1443 bzero(&ND, sizeof(ND)); 1444 ND.linkmtu = IN6_LINKMTU(ifp); 1445 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1446 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1447 ND.reachable = ND_IFINFO(ifp)->reachable; 1448 ND.retrans = ND_IFINFO(ifp)->retrans; 1449 ND.flags = ND_IFINFO(ifp)->flags; 1450 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1451 ND.chlim = ND_IFINFO(ifp)->chlim; 1452 break; 1453 case SIOCGIFINFO_IN6: 1454 ND = *ND_IFINFO(ifp); 1455 break; 1456 case SIOCSIFINFO_IN6: 1457 /* 1458 * used to change host variables from userland. 1459 * intented for a use on router to reflect RA configurations. 1460 */ 1461 /* 0 means 'unspecified' */ 1462 if (ND.linkmtu != 0) { 1463 if (ND.linkmtu < IPV6_MMTU || 1464 ND.linkmtu > IN6_LINKMTU(ifp)) { 1465 error = EINVAL; 1466 break; 1467 } 1468 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1469 } 1470 1471 if (ND.basereachable != 0) { 1472 int obasereachable = ND_IFINFO(ifp)->basereachable; 1473 1474 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1475 if (ND.basereachable != obasereachable) 1476 ND_IFINFO(ifp)->reachable = 1477 ND_COMPUTE_RTIME(ND.basereachable); 1478 } 1479 if (ND.retrans != 0) 1480 ND_IFINFO(ifp)->retrans = ND.retrans; 1481 if (ND.chlim != 0) 1482 ND_IFINFO(ifp)->chlim = ND.chlim; 1483 /* FALLTHROUGH */ 1484 case SIOCSIFINFO_FLAGS: 1485 { 1486 struct ifaddr *ifa; 1487 struct in6_ifaddr *ia; 1488 1489 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1490 !(ND.flags & ND6_IFF_IFDISABLED)) { 1491 /* ifdisabled 1->0 transision */ 1492 1493 /* 1494 * If the interface is marked as ND6_IFF_IFDISABLED and 1495 * has an link-local address with IN6_IFF_DUPLICATED, 1496 * do not clear ND6_IFF_IFDISABLED. 1497 * See RFC 4862, Section 5.4.5. 1498 */ 1499 int duplicated_linklocal = 0; 1500 1501 IF_ADDR_RLOCK(ifp); 1502 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1503 if (ifa->ifa_addr->sa_family != AF_INET6) 1504 continue; 1505 ia = (struct in6_ifaddr *)ifa; 1506 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1507 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1508 duplicated_linklocal = 1; 1509 break; 1510 } 1511 } 1512 IF_ADDR_RUNLOCK(ifp); 1513 1514 if (duplicated_linklocal) { 1515 ND.flags |= ND6_IFF_IFDISABLED; 1516 log(LOG_ERR, "Cannot enable an interface" 1517 " with a link-local address marked" 1518 " duplicate.\n"); 1519 } else { 1520 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1521 if (ifp->if_flags & IFF_UP) 1522 in6_if_up(ifp); 1523 } 1524 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1525 (ND.flags & ND6_IFF_IFDISABLED)) { 1526 /* ifdisabled 0->1 transision */ 1527 /* Mark all IPv6 address as tentative. */ 1528 1529 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1530 IF_ADDR_RLOCK(ifp); 1531 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1532 if (ifa->ifa_addr->sa_family != AF_INET6) 1533 continue; 1534 ia = (struct in6_ifaddr *)ifa; 1535 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1536 } 1537 IF_ADDR_RUNLOCK(ifp); 1538 } 1539 1540 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1541 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1542 /* auto_linklocal 0->1 transision */ 1543 1544 /* If no link-local address on ifp, configure */ 1545 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1546 in6_ifattach(ifp, NULL); 1547 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1548 ifp->if_flags & IFF_UP) { 1549 /* 1550 * When the IF already has 1551 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1552 * address is assigned, and IFF_UP, try to 1553 * assign one. 1554 */ 1555 int haslinklocal = 0; 1556 1557 IF_ADDR_RLOCK(ifp); 1558 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1559 if (ifa->ifa_addr->sa_family != AF_INET6) 1560 continue; 1561 ia = (struct in6_ifaddr *)ifa; 1562 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1563 haslinklocal = 1; 1564 break; 1565 } 1566 } 1567 IF_ADDR_RUNLOCK(ifp); 1568 if (!haslinklocal) 1569 in6_ifattach(ifp, NULL); 1570 } 1571 } 1572 } 1573 ND_IFINFO(ifp)->flags = ND.flags; 1574 break; 1575 #undef ND 1576 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1577 /* sync kernel routing table with the default router list */ 1578 defrouter_reset(); 1579 defrouter_select(); 1580 break; 1581 case SIOCSPFXFLUSH_IN6: 1582 { 1583 /* flush all the prefix advertised by routers */ 1584 struct nd_prefix *pr, *next; 1585 1586 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1587 struct in6_ifaddr *ia, *ia_next; 1588 1589 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1590 continue; /* XXX */ 1591 1592 /* do we really have to remove addresses as well? */ 1593 /* XXXRW: in6_ifaddrhead locking. */ 1594 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1595 ia_next) { 1596 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1597 continue; 1598 1599 if (ia->ia6_ndpr == pr) 1600 in6_purgeaddr(&ia->ia_ifa); 1601 } 1602 prelist_remove(pr); 1603 } 1604 break; 1605 } 1606 case SIOCSRTRFLUSH_IN6: 1607 { 1608 /* flush all the default routers */ 1609 struct nd_defrouter *dr, *next; 1610 1611 defrouter_reset(); 1612 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1613 defrtrlist_del(dr); 1614 } 1615 defrouter_select(); 1616 break; 1617 } 1618 case SIOCGNBRINFO_IN6: 1619 { 1620 struct llentry *ln; 1621 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1622 1623 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1624 return (error); 1625 1626 IF_AFDATA_RLOCK(ifp); 1627 ln = nd6_lookup(&nb_addr, 0, ifp); 1628 IF_AFDATA_RUNLOCK(ifp); 1629 1630 if (ln == NULL) { 1631 error = EINVAL; 1632 break; 1633 } 1634 nbi->state = ln->ln_state; 1635 nbi->asked = ln->la_asked; 1636 nbi->isrouter = ln->ln_router; 1637 if (ln->la_expire == 0) 1638 nbi->expire = 0; 1639 else 1640 nbi->expire = ln->la_expire + 1641 (time_second - time_uptime); 1642 LLE_RUNLOCK(ln); 1643 break; 1644 } 1645 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1646 ndif->ifindex = V_nd6_defifindex; 1647 break; 1648 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1649 return (nd6_setdefaultiface(ndif->ifindex)); 1650 } 1651 return (error); 1652 } 1653 1654 /* 1655 * Create neighbor cache entry and cache link-layer address, 1656 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1657 * 1658 * type - ICMP6 type 1659 * code - type dependent information 1660 * 1661 * XXXXX 1662 * The caller of this function already acquired the ndp 1663 * cache table lock because the cache entry is returned. 1664 */ 1665 struct llentry * 1666 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1667 int lladdrlen, int type, int code) 1668 { 1669 struct llentry *ln = NULL, *ln_tmp; 1670 int is_newentry; 1671 int do_update; 1672 int olladdr; 1673 int llchange; 1674 int flags; 1675 int newstate = 0; 1676 uint16_t router = 0; 1677 struct sockaddr_in6 sin6; 1678 struct mbuf *chain = NULL; 1679 int static_route = 0; 1680 1681 IF_AFDATA_UNLOCK_ASSERT(ifp); 1682 1683 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1684 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1685 1686 /* nothing must be updated for unspecified address */ 1687 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1688 return NULL; 1689 1690 /* 1691 * Validation about ifp->if_addrlen and lladdrlen must be done in 1692 * the caller. 1693 * 1694 * XXX If the link does not have link-layer adderss, what should 1695 * we do? (ifp->if_addrlen == 0) 1696 * Spec says nothing in sections for RA, RS and NA. There's small 1697 * description on it in NS section (RFC 2461 7.2.3). 1698 */ 1699 flags = lladdr ? ND6_EXCLUSIVE : 0; 1700 IF_AFDATA_RLOCK(ifp); 1701 ln = nd6_lookup(from, flags, ifp); 1702 IF_AFDATA_RUNLOCK(ifp); 1703 is_newentry = 0; 1704 if (ln == NULL) { 1705 flags |= ND6_EXCLUSIVE; 1706 ln = nd6_alloc(from, 0, ifp); 1707 if (ln == NULL) 1708 return (NULL); 1709 IF_AFDATA_WLOCK(ifp); 1710 LLE_WLOCK(ln); 1711 /* Prefer any existing lle over newly-created one */ 1712 ln_tmp = nd6_lookup(from, ND6_EXCLUSIVE, ifp); 1713 if (ln_tmp == NULL) 1714 lltable_link_entry(LLTABLE6(ifp), ln); 1715 IF_AFDATA_WUNLOCK(ifp); 1716 if (ln_tmp == NULL) 1717 /* No existing lle, mark as new entry */ 1718 is_newentry = 1; 1719 else { 1720 lltable_free_entry(LLTABLE6(ifp), ln); 1721 ln = ln_tmp; 1722 ln_tmp = NULL; 1723 } 1724 } 1725 /* do nothing if static ndp is set */ 1726 if ((ln->la_flags & LLE_STATIC)) { 1727 static_route = 1; 1728 goto done; 1729 } 1730 1731 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1732 if (olladdr && lladdr) { 1733 llchange = bcmp(lladdr, &ln->ll_addr, 1734 ifp->if_addrlen); 1735 } else 1736 llchange = 0; 1737 1738 /* 1739 * newentry olladdr lladdr llchange (*=record) 1740 * 0 n n -- (1) 1741 * 0 y n -- (2) 1742 * 0 n y -- (3) * STALE 1743 * 0 y y n (4) * 1744 * 0 y y y (5) * STALE 1745 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1746 * 1 -- y -- (7) * STALE 1747 */ 1748 1749 if (lladdr) { /* (3-5) and (7) */ 1750 /* 1751 * Record source link-layer address 1752 * XXX is it dependent to ifp->if_type? 1753 */ 1754 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1755 ln->la_flags |= LLE_VALID; 1756 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 1757 } 1758 1759 if (!is_newentry) { 1760 if ((!olladdr && lladdr != NULL) || /* (3) */ 1761 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1762 do_update = 1; 1763 newstate = ND6_LLINFO_STALE; 1764 } else /* (1-2,4) */ 1765 do_update = 0; 1766 } else { 1767 do_update = 1; 1768 if (lladdr == NULL) /* (6) */ 1769 newstate = ND6_LLINFO_NOSTATE; 1770 else /* (7) */ 1771 newstate = ND6_LLINFO_STALE; 1772 } 1773 1774 if (do_update) { 1775 /* 1776 * Update the state of the neighbor cache. 1777 */ 1778 ln->ln_state = newstate; 1779 1780 if (ln->ln_state == ND6_LLINFO_STALE) { 1781 if (ln->la_hold != NULL) 1782 nd6_grab_holdchain(ln, &chain, &sin6); 1783 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1784 /* probe right away */ 1785 nd6_llinfo_settimer_locked((void *)ln, 0); 1786 } 1787 } 1788 1789 /* 1790 * ICMP6 type dependent behavior. 1791 * 1792 * NS: clear IsRouter if new entry 1793 * RS: clear IsRouter 1794 * RA: set IsRouter if there's lladdr 1795 * redir: clear IsRouter if new entry 1796 * 1797 * RA case, (1): 1798 * The spec says that we must set IsRouter in the following cases: 1799 * - If lladdr exist, set IsRouter. This means (1-5). 1800 * - If it is old entry (!newentry), set IsRouter. This means (7). 1801 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1802 * A quetion arises for (1) case. (1) case has no lladdr in the 1803 * neighbor cache, this is similar to (6). 1804 * This case is rare but we figured that we MUST NOT set IsRouter. 1805 * 1806 * newentry olladdr lladdr llchange NS RS RA redir 1807 * D R 1808 * 0 n n -- (1) c ? s 1809 * 0 y n -- (2) c s s 1810 * 0 n y -- (3) c s s 1811 * 0 y y n (4) c s s 1812 * 0 y y y (5) c s s 1813 * 1 -- n -- (6) c c c s 1814 * 1 -- y -- (7) c c s c s 1815 * 1816 * (c=clear s=set) 1817 */ 1818 switch (type & 0xff) { 1819 case ND_NEIGHBOR_SOLICIT: 1820 /* 1821 * New entry must have is_router flag cleared. 1822 */ 1823 if (is_newentry) /* (6-7) */ 1824 ln->ln_router = 0; 1825 break; 1826 case ND_REDIRECT: 1827 /* 1828 * If the icmp is a redirect to a better router, always set the 1829 * is_router flag. Otherwise, if the entry is newly created, 1830 * clear the flag. [RFC 2461, sec 8.3] 1831 */ 1832 if (code == ND_REDIRECT_ROUTER) 1833 ln->ln_router = 1; 1834 else if (is_newentry) /* (6-7) */ 1835 ln->ln_router = 0; 1836 break; 1837 case ND_ROUTER_SOLICIT: 1838 /* 1839 * is_router flag must always be cleared. 1840 */ 1841 ln->ln_router = 0; 1842 break; 1843 case ND_ROUTER_ADVERT: 1844 /* 1845 * Mark an entry with lladdr as a router. 1846 */ 1847 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1848 (is_newentry && lladdr)) { /* (7) */ 1849 ln->ln_router = 1; 1850 } 1851 break; 1852 } 1853 1854 if (ln != NULL) { 1855 static_route = (ln->la_flags & LLE_STATIC); 1856 router = ln->ln_router; 1857 1858 if (flags & ND6_EXCLUSIVE) 1859 LLE_WUNLOCK(ln); 1860 else 1861 LLE_RUNLOCK(ln); 1862 if (static_route) 1863 ln = NULL; 1864 } 1865 if (chain != NULL) 1866 nd6_flush_holdchain(ifp, ifp, chain, &sin6); 1867 1868 /* 1869 * When the link-layer address of a router changes, select the 1870 * best router again. In particular, when the neighbor entry is newly 1871 * created, it might affect the selection policy. 1872 * Question: can we restrict the first condition to the "is_newentry" 1873 * case? 1874 * XXX: when we hear an RA from a new router with the link-layer 1875 * address option, defrouter_select() is called twice, since 1876 * defrtrlist_update called the function as well. However, I believe 1877 * we can compromise the overhead, since it only happens the first 1878 * time. 1879 * XXX: although defrouter_select() should not have a bad effect 1880 * for those are not autoconfigured hosts, we explicitly avoid such 1881 * cases for safety. 1882 */ 1883 if (do_update && router && 1884 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1885 /* 1886 * guaranteed recursion 1887 */ 1888 defrouter_select(); 1889 } 1890 1891 return (ln); 1892 done: 1893 if (ln != NULL) { 1894 if (flags & ND6_EXCLUSIVE) 1895 LLE_WUNLOCK(ln); 1896 else 1897 LLE_RUNLOCK(ln); 1898 if (static_route) 1899 ln = NULL; 1900 } 1901 return (ln); 1902 } 1903 1904 static void 1905 nd6_slowtimo(void *arg) 1906 { 1907 CURVNET_SET((struct vnet *) arg); 1908 struct nd_ifinfo *nd6if; 1909 struct ifnet *ifp; 1910 1911 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1912 nd6_slowtimo, curvnet); 1913 IFNET_RLOCK_NOSLEEP(); 1914 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1915 if (ifp->if_afdata[AF_INET6] == NULL) 1916 continue; 1917 nd6if = ND_IFINFO(ifp); 1918 if (nd6if->basereachable && /* already initialized */ 1919 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1920 /* 1921 * Since reachable time rarely changes by router 1922 * advertisements, we SHOULD insure that a new random 1923 * value gets recomputed at least once every few hours. 1924 * (RFC 2461, 6.3.4) 1925 */ 1926 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1927 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1928 } 1929 } 1930 IFNET_RUNLOCK_NOSLEEP(); 1931 CURVNET_RESTORE(); 1932 } 1933 1934 void 1935 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 1936 struct sockaddr_in6 *sin6) 1937 { 1938 1939 LLE_WLOCK_ASSERT(ln); 1940 1941 *chain = ln->la_hold; 1942 ln->la_hold = NULL; 1943 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 1944 1945 if (ln->ln_state == ND6_LLINFO_STALE) { 1946 1947 /* 1948 * The first time we send a packet to a 1949 * neighbor whose entry is STALE, we have 1950 * to change the state to DELAY and a sets 1951 * a timer to expire in DELAY_FIRST_PROBE_TIME 1952 * seconds to ensure do neighbor unreachability 1953 * detection on expiration. 1954 * (RFC 2461 7.3.3) 1955 */ 1956 ln->la_asked = 0; 1957 ln->ln_state = ND6_LLINFO_DELAY; 1958 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1959 } 1960 } 1961 1962 static int 1963 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1964 struct sockaddr_in6 *dst) 1965 { 1966 int error; 1967 int ip6len; 1968 struct ip6_hdr *ip6; 1969 struct m_tag *mtag; 1970 1971 #ifdef MAC 1972 mac_netinet6_nd6_send(ifp, m); 1973 #endif 1974 1975 /* 1976 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 1977 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 1978 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 1979 * to be diverted to user space. When re-injected into the kernel, 1980 * send_output() will directly dispatch them to the outgoing interface. 1981 */ 1982 if (send_sendso_input_hook != NULL) { 1983 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 1984 if (mtag != NULL) { 1985 ip6 = mtod(m, struct ip6_hdr *); 1986 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 1987 /* Use the SEND socket */ 1988 error = send_sendso_input_hook(m, ifp, SND_OUT, 1989 ip6len); 1990 /* -1 == no app on SEND socket */ 1991 if (error == 0 || error != -1) 1992 return (error); 1993 } 1994 } 1995 1996 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1997 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 1998 mtod(m, struct ip6_hdr *)); 1999 2000 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2001 origifp = ifp; 2002 2003 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL); 2004 return (error); 2005 } 2006 2007 /* 2008 * IPv6 packet output - light version. 2009 * Checks if destination LLE exists and is in proper state 2010 * (e.g no modification required). If not true, fall back to 2011 * "heavy" version. 2012 */ 2013 int 2014 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2015 struct sockaddr_in6 *dst, struct rtentry *rt0) 2016 { 2017 struct llentry *ln = NULL; 2018 2019 /* discard the packet if IPv6 operation is disabled on the interface */ 2020 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2021 m_freem(m); 2022 return (ENETDOWN); /* better error? */ 2023 } 2024 2025 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 2026 goto sendpkt; 2027 2028 if (nd6_need_cache(ifp) == 0) 2029 goto sendpkt; 2030 2031 IF_AFDATA_RLOCK(ifp); 2032 ln = nd6_lookup(&dst->sin6_addr, 0, ifp); 2033 IF_AFDATA_RUNLOCK(ifp); 2034 2035 /* 2036 * Perform fast path for the following cases: 2037 * 1) lle state is REACHABLE 2038 * 2) lle state is DELAY (NS message sentNS message sent) 2039 * 2040 * Every other case involves lle modification, so we handle 2041 * them separately. 2042 */ 2043 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE && 2044 ln->ln_state != ND6_LLINFO_DELAY)) { 2045 /* Fall back to slow processing path */ 2046 if (ln != NULL) 2047 LLE_RUNLOCK(ln); 2048 return (nd6_output_lle(ifp, origifp, m, dst)); 2049 } 2050 2051 sendpkt: 2052 if (ln != NULL) 2053 LLE_RUNLOCK(ln); 2054 2055 return (nd6_output_ifp(ifp, origifp, m, dst)); 2056 } 2057 2058 2059 /* 2060 * Output IPv6 packet - heavy version. 2061 * Function assume that either 2062 * 1) destination LLE does not exist, is invalid or stale, so 2063 * ND6_EXCLUSIVE lock needs to be acquired 2064 * 2) destination lle is provided (with ND6_EXCLUSIVE lock), 2065 * in that case packets are queued in &chain. 2066 * 2067 */ 2068 static int 2069 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2070 struct sockaddr_in6 *dst) 2071 { 2072 struct llentry *lle = NULL, *lle_tmp; 2073 2074 KASSERT(m != NULL, ("NULL mbuf, nothing to send")); 2075 /* discard the packet if IPv6 operation is disabled on the interface */ 2076 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2077 m_freem(m); 2078 return (ENETDOWN); /* better error? */ 2079 } 2080 2081 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 2082 goto sendpkt; 2083 2084 if (nd6_need_cache(ifp) == 0) 2085 goto sendpkt; 2086 2087 /* 2088 * Address resolution or Neighbor Unreachability Detection 2089 * for the next hop. 2090 * At this point, the destination of the packet must be a unicast 2091 * or an anycast address(i.e. not a multicast). 2092 */ 2093 if (lle == NULL) { 2094 IF_AFDATA_RLOCK(ifp); 2095 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 2096 IF_AFDATA_RUNLOCK(ifp); 2097 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2098 /* 2099 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2100 * the condition below is not very efficient. But we believe 2101 * it is tolerable, because this should be a rare case. 2102 */ 2103 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2104 if (lle == NULL) { 2105 char ip6buf[INET6_ADDRSTRLEN]; 2106 log(LOG_DEBUG, 2107 "nd6_output: can't allocate llinfo for %s " 2108 "(ln=%p)\n", 2109 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2110 m_freem(m); 2111 return (ENOBUFS); 2112 } 2113 lle->ln_state = ND6_LLINFO_NOSTATE; 2114 2115 IF_AFDATA_WLOCK(ifp); 2116 LLE_WLOCK(lle); 2117 /* Prefer any existing entry over newly-created one */ 2118 lle_tmp = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 2119 if (lle_tmp == NULL) 2120 lltable_link_entry(LLTABLE6(ifp), lle); 2121 IF_AFDATA_WUNLOCK(ifp); 2122 if (lle_tmp != NULL) { 2123 lltable_free_entry(LLTABLE6(ifp), lle); 2124 lle = lle_tmp; 2125 lle_tmp = NULL; 2126 } 2127 } 2128 } 2129 if (lle == NULL) { 2130 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2131 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2132 m_freem(m); 2133 return (ENOBUFS); 2134 } 2135 goto sendpkt; /* send anyway */ 2136 } 2137 2138 LLE_WLOCK_ASSERT(lle); 2139 2140 /* We don't have to do link-layer address resolution on a p2p link. */ 2141 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2142 lle->ln_state < ND6_LLINFO_REACHABLE) { 2143 lle->ln_state = ND6_LLINFO_STALE; 2144 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz); 2145 } 2146 2147 /* 2148 * The first time we send a packet to a neighbor whose entry is 2149 * STALE, we have to change the state to DELAY and a sets a timer to 2150 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2151 * neighbor unreachability detection on expiration. 2152 * (RFC 2461 7.3.3) 2153 */ 2154 if (lle->ln_state == ND6_LLINFO_STALE) { 2155 lle->la_asked = 0; 2156 lle->ln_state = ND6_LLINFO_DELAY; 2157 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz); 2158 } 2159 2160 /* 2161 * If the neighbor cache entry has a state other than INCOMPLETE 2162 * (i.e. its link-layer address is already resolved), just 2163 * send the packet. 2164 */ 2165 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) 2166 goto sendpkt; 2167 2168 /* 2169 * There is a neighbor cache entry, but no ethernet address 2170 * response yet. Append this latest packet to the end of the 2171 * packet queue in the mbuf, unless the number of the packet 2172 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2173 * the oldest packet in the queue will be removed. 2174 */ 2175 if (lle->ln_state == ND6_LLINFO_NOSTATE) 2176 lle->ln_state = ND6_LLINFO_INCOMPLETE; 2177 2178 if (lle->la_hold != NULL) { 2179 struct mbuf *m_hold; 2180 int i; 2181 2182 i = 0; 2183 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2184 i++; 2185 if (m_hold->m_nextpkt == NULL) { 2186 m_hold->m_nextpkt = m; 2187 break; 2188 } 2189 } 2190 while (i >= V_nd6_maxqueuelen) { 2191 m_hold = lle->la_hold; 2192 lle->la_hold = lle->la_hold->m_nextpkt; 2193 m_freem(m_hold); 2194 i--; 2195 } 2196 } else { 2197 lle->la_hold = m; 2198 } 2199 2200 /* 2201 * If there has been no NS for the neighbor after entering the 2202 * INCOMPLETE state, send the first solicitation. 2203 */ 2204 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) { 2205 struct in6_addr src, *psrc; 2206 lle->la_asked++; 2207 2208 nd6_llinfo_settimer_locked(lle, 2209 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2210 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2211 LLE_WUNLOCK(lle); 2212 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2213 } else { 2214 /* We did the lookup so we need to do the unlock here. */ 2215 LLE_WUNLOCK(lle); 2216 } 2217 2218 return (0); 2219 2220 sendpkt: 2221 if (lle != NULL) 2222 LLE_WUNLOCK(lle); 2223 2224 return (nd6_output_ifp(ifp, origifp, m, dst)); 2225 } 2226 2227 2228 int 2229 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2230 struct sockaddr_in6 *dst) 2231 { 2232 struct mbuf *m, *m_head; 2233 struct ifnet *outifp; 2234 int error = 0; 2235 2236 m_head = chain; 2237 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2238 outifp = origifp; 2239 else 2240 outifp = ifp; 2241 2242 while (m_head) { 2243 m = m_head; 2244 m_head = m_head->m_nextpkt; 2245 error = nd6_output_ifp(ifp, origifp, m, dst); 2246 } 2247 2248 /* 2249 * XXX 2250 * note that intermediate errors are blindly ignored - but this is 2251 * the same convention as used with nd6_output when called by 2252 * nd6_cache_lladdr 2253 */ 2254 return (error); 2255 } 2256 2257 2258 int 2259 nd6_need_cache(struct ifnet *ifp) 2260 { 2261 /* 2262 * XXX: we currently do not make neighbor cache on any interface 2263 * other than ARCnet, Ethernet, FDDI and GIF. 2264 * 2265 * RFC2893 says: 2266 * - unidirectional tunnels needs no ND 2267 */ 2268 switch (ifp->if_type) { 2269 case IFT_ARCNET: 2270 case IFT_ETHER: 2271 case IFT_FDDI: 2272 case IFT_IEEE1394: 2273 case IFT_L2VLAN: 2274 case IFT_IEEE80211: 2275 case IFT_INFINIBAND: 2276 case IFT_BRIDGE: 2277 case IFT_PROPVIRTUAL: 2278 return (1); 2279 default: 2280 return (0); 2281 } 2282 } 2283 2284 /* 2285 * Add pernament ND6 link-layer record for given 2286 * interface address. 2287 * 2288 * Very similar to IPv4 arp_ifinit(), but: 2289 * 1) IPv6 DAD is performed in different place 2290 * 2) It is called by IPv6 protocol stack in contrast to 2291 * arp_ifinit() which is typically called in SIOCSIFADDR 2292 * driver ioctl handler. 2293 * 2294 */ 2295 int 2296 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2297 { 2298 struct ifnet *ifp; 2299 struct llentry *ln, *ln_tmp; 2300 struct sockaddr *dst; 2301 2302 ifp = ia->ia_ifa.ifa_ifp; 2303 if (nd6_need_cache(ifp) == 0) 2304 return (0); 2305 2306 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; 2307 dst = (struct sockaddr *)&ia->ia_addr; 2308 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2309 if (ln == NULL) 2310 return (ENOBUFS); 2311 2312 IF_AFDATA_WLOCK(ifp); 2313 LLE_WLOCK(ln); 2314 /* Unlink any entry if exists */ 2315 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2316 if (ln_tmp != NULL) 2317 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2318 lltable_link_entry(LLTABLE6(ifp), ln); 2319 IF_AFDATA_WUNLOCK(ifp); 2320 2321 if (ln_tmp != NULL) 2322 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2323 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2324 2325 LLE_WUNLOCK(ln); 2326 if (ln_tmp != NULL) 2327 llentry_free(ln_tmp); 2328 2329 return (0); 2330 } 2331 2332 /* 2333 * Removes ALL lle records for interface address prefix. 2334 * XXXME: That's probably not we really want to do, we need 2335 * to remove address record only and keep other records 2336 * until we determine if given prefix is really going 2337 * to be removed. 2338 */ 2339 void 2340 nd6_rem_ifa_lle(struct in6_ifaddr *ia) 2341 { 2342 struct sockaddr_in6 mask, addr; 2343 struct ifnet *ifp; 2344 2345 ifp = ia->ia_ifa.ifa_ifp; 2346 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2347 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2348 lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr, 2349 (struct sockaddr *)&mask, LLE_STATIC); 2350 } 2351 2352 /* 2353 * the callers of this function need to be re-worked to drop 2354 * the lle lock, drop here for now 2355 */ 2356 int 2357 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2358 const struct sockaddr *dst, u_char *desten, uint32_t *pflags) 2359 { 2360 struct llentry *ln; 2361 2362 if (pflags != NULL) 2363 *pflags = 0; 2364 IF_AFDATA_UNLOCK_ASSERT(ifp); 2365 if (m != NULL && m->m_flags & M_MCAST) { 2366 switch (ifp->if_type) { 2367 case IFT_ETHER: 2368 case IFT_FDDI: 2369 case IFT_L2VLAN: 2370 case IFT_IEEE80211: 2371 case IFT_BRIDGE: 2372 case IFT_ISO88025: 2373 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2374 desten); 2375 return (0); 2376 default: 2377 m_freem(m); 2378 return (EAFNOSUPPORT); 2379 } 2380 } 2381 2382 2383 /* 2384 * the entry should have been created in nd6_store_lladdr 2385 */ 2386 IF_AFDATA_RLOCK(ifp); 2387 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2388 IF_AFDATA_RUNLOCK(ifp); 2389 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2390 if (ln != NULL) 2391 LLE_RUNLOCK(ln); 2392 /* this could happen, if we could not allocate memory */ 2393 m_freem(m); 2394 return (1); 2395 } 2396 2397 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2398 if (pflags != NULL) 2399 *pflags = ln->la_flags; 2400 LLE_RUNLOCK(ln); 2401 /* 2402 * A *small* use after free race exists here 2403 */ 2404 return (0); 2405 } 2406 2407 static void 2408 clear_llinfo_pqueue(struct llentry *ln) 2409 { 2410 struct mbuf *m_hold, *m_hold_next; 2411 2412 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2413 m_hold_next = m_hold->m_nextpkt; 2414 m_freem(m_hold); 2415 } 2416 2417 ln->la_hold = NULL; 2418 return; 2419 } 2420 2421 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2422 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2423 #ifdef SYSCTL_DECL 2424 SYSCTL_DECL(_net_inet6_icmp6); 2425 #endif 2426 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2427 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2428 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2429 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2430 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2431 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2432 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2433 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2434 2435 static int 2436 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2437 { 2438 struct in6_defrouter d; 2439 struct nd_defrouter *dr; 2440 int error; 2441 2442 if (req->newptr) 2443 return (EPERM); 2444 2445 bzero(&d, sizeof(d)); 2446 d.rtaddr.sin6_family = AF_INET6; 2447 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2448 2449 /* 2450 * XXX locking 2451 */ 2452 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2453 d.rtaddr.sin6_addr = dr->rtaddr; 2454 error = sa6_recoverscope(&d.rtaddr); 2455 if (error != 0) 2456 return (error); 2457 d.flags = dr->flags; 2458 d.rtlifetime = dr->rtlifetime; 2459 d.expire = dr->expire + (time_second - time_uptime); 2460 d.if_index = dr->ifp->if_index; 2461 error = SYSCTL_OUT(req, &d, sizeof(d)); 2462 if (error != 0) 2463 return (error); 2464 } 2465 return (0); 2466 } 2467 2468 static int 2469 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2470 { 2471 struct in6_prefix p; 2472 struct sockaddr_in6 s6; 2473 struct nd_prefix *pr; 2474 struct nd_pfxrouter *pfr; 2475 time_t maxexpire; 2476 int error; 2477 char ip6buf[INET6_ADDRSTRLEN]; 2478 2479 if (req->newptr) 2480 return (EPERM); 2481 2482 bzero(&p, sizeof(p)); 2483 p.origin = PR_ORIG_RA; 2484 bzero(&s6, sizeof(s6)); 2485 s6.sin6_family = AF_INET6; 2486 s6.sin6_len = sizeof(s6); 2487 2488 /* 2489 * XXX locking 2490 */ 2491 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2492 p.prefix = pr->ndpr_prefix; 2493 if (sa6_recoverscope(&p.prefix)) { 2494 log(LOG_ERR, "scope error in prefix list (%s)\n", 2495 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2496 /* XXX: press on... */ 2497 } 2498 p.raflags = pr->ndpr_raf; 2499 p.prefixlen = pr->ndpr_plen; 2500 p.vltime = pr->ndpr_vltime; 2501 p.pltime = pr->ndpr_pltime; 2502 p.if_index = pr->ndpr_ifp->if_index; 2503 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2504 p.expire = 0; 2505 else { 2506 /* XXX: we assume time_t is signed. */ 2507 maxexpire = (-1) & 2508 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2509 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2510 p.expire = pr->ndpr_lastupdate + 2511 pr->ndpr_vltime + 2512 (time_second - time_uptime); 2513 else 2514 p.expire = maxexpire; 2515 } 2516 p.refcnt = pr->ndpr_refcnt; 2517 p.flags = pr->ndpr_stateflags; 2518 p.advrtrs = 0; 2519 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2520 p.advrtrs++; 2521 error = SYSCTL_OUT(req, &p, sizeof(p)); 2522 if (error != 0) 2523 return (error); 2524 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2525 s6.sin6_addr = pfr->router->rtaddr; 2526 if (sa6_recoverscope(&s6)) 2527 log(LOG_ERR, 2528 "scope error in prefix list (%s)\n", 2529 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2530 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2531 if (error != 0) 2532 return (error); 2533 } 2534 } 2535 return (0); 2536 } 2537