1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the project nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_inet.h" 38 #include "opt_inet6.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/callout.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/mutex.h> 47 #include <sys/socket.h> 48 #include <sys/sockio.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/protosw.h> 52 #include <sys/errno.h> 53 #include <sys/syslog.h> 54 #include <sys/rwlock.h> 55 #include <sys/queue.h> 56 #include <sys/sdt.h> 57 #include <sys/sysctl.h> 58 59 #include <net/if.h> 60 #include <net/if_var.h> 61 #include <net/if_dl.h> 62 #include <net/if_types.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 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery"); 89 90 /* timer values */ 91 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 92 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 93 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 94 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 95 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 96 * local traffic */ 97 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 98 * collection timer */ 99 100 /* preventing too many loops in ND option parsing */ 101 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 102 103 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 104 * layer hints */ 105 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 106 * ND entries */ 107 #define V_nd6_maxndopt VNET(nd6_maxndopt) 108 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 109 110 #ifdef ND6_DEBUG 111 VNET_DEFINE(int, nd6_debug) = 1; 112 #else 113 VNET_DEFINE(int, nd6_debug) = 0; 114 #endif 115 116 static eventhandler_tag lle_event_eh, iflladdr_event_eh, ifnet_link_event_eh; 117 118 VNET_DEFINE(struct nd_drhead, nd_defrouter); 119 VNET_DEFINE(struct nd_prhead, nd_prefix); 120 VNET_DEFINE(struct rwlock, nd6_lock); 121 VNET_DEFINE(uint64_t, nd6_list_genid); 122 VNET_DEFINE(struct mtx, nd6_onlink_mtx); 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(const 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 void nd6_free(struct llentry **, int); 135 static void nd6_free_redirect(const struct llentry *); 136 static void nd6_llinfo_timer(void *); 137 static void nd6_llinfo_settimer_locked(struct llentry *, long); 138 static void clear_llinfo_pqueue(struct llentry *); 139 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 140 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *, 141 const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **); 142 static int nd6_need_cache(struct ifnet *); 143 144 145 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch); 146 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 147 148 VNET_DEFINE(struct callout, nd6_timer_ch); 149 #define V_nd6_timer_ch VNET(nd6_timer_ch) 150 151 static void 152 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 153 { 154 struct rt_addrinfo rtinfo; 155 struct sockaddr_in6 dst; 156 struct sockaddr_dl gw; 157 struct ifnet *ifp; 158 int type; 159 int fibnum; 160 161 LLE_WLOCK_ASSERT(lle); 162 163 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 164 return; 165 166 switch (evt) { 167 case LLENTRY_RESOLVED: 168 type = RTM_ADD; 169 KASSERT(lle->la_flags & LLE_VALID, 170 ("%s: %p resolved but not valid?", __func__, lle)); 171 break; 172 case LLENTRY_EXPIRED: 173 type = RTM_DELETE; 174 break; 175 default: 176 return; 177 } 178 179 ifp = lltable_get_ifp(lle->lle_tbl); 180 181 bzero(&dst, sizeof(dst)); 182 bzero(&gw, sizeof(gw)); 183 bzero(&rtinfo, sizeof(rtinfo)); 184 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 185 dst.sin6_scope_id = in6_getscopezone(ifp, 186 in6_addrscope(&dst.sin6_addr)); 187 gw.sdl_len = sizeof(struct sockaddr_dl); 188 gw.sdl_family = AF_LINK; 189 gw.sdl_alen = ifp->if_addrlen; 190 gw.sdl_index = ifp->if_index; 191 gw.sdl_type = ifp->if_type; 192 if (evt == LLENTRY_RESOLVED) 193 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 194 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 195 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 196 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 197 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib; 198 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 199 type == RTM_ADD ? RTF_UP: 0), 0, fibnum); 200 } 201 202 /* 203 * A handler for interface link layer address change event. 204 */ 205 static void 206 nd6_iflladdr(void *arg __unused, struct ifnet *ifp) 207 { 208 209 lltable_update_ifaddr(LLTABLE6(ifp)); 210 } 211 212 void 213 nd6_init(void) 214 { 215 216 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF); 217 rw_init(&V_nd6_lock, "nd6 list"); 218 219 LIST_INIT(&V_nd_prefix); 220 TAILQ_INIT(&V_nd_defrouter); 221 222 /* Start timers. */ 223 callout_init(&V_nd6_slowtimo_ch, 0); 224 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 225 nd6_slowtimo, curvnet); 226 227 callout_init(&V_nd6_timer_ch, 0); 228 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 229 230 nd6_dad_init(); 231 if (IS_DEFAULT_VNET(curvnet)) { 232 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 233 NULL, EVENTHANDLER_PRI_ANY); 234 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event, 235 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 236 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event, 237 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY); 238 } 239 } 240 241 #ifdef VIMAGE 242 void 243 nd6_destroy() 244 { 245 246 callout_drain(&V_nd6_slowtimo_ch); 247 callout_drain(&V_nd6_timer_ch); 248 if (IS_DEFAULT_VNET(curvnet)) { 249 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh); 250 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 251 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh); 252 } 253 rw_destroy(&V_nd6_lock); 254 mtx_destroy(&V_nd6_onlink_mtx); 255 } 256 #endif 257 258 struct nd_ifinfo * 259 nd6_ifattach(struct ifnet *ifp) 260 { 261 struct nd_ifinfo *nd; 262 263 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 264 nd->initialized = 1; 265 266 nd->chlim = IPV6_DEFHLIM; 267 nd->basereachable = REACHABLE_TIME; 268 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 269 nd->retrans = RETRANS_TIMER; 270 271 nd->flags = ND6_IFF_PERFORMNUD; 272 273 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 274 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 275 * default regardless of the V_ip6_auto_linklocal configuration to 276 * give a reasonable default behavior. 277 */ 278 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 279 (ifp->if_flags & IFF_LOOPBACK)) 280 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 281 /* 282 * A loopback interface does not need to accept RTADV. 283 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 284 * default regardless of the V_ip6_accept_rtadv configuration to 285 * prevent the interface from accepting RA messages arrived 286 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 287 */ 288 if (V_ip6_accept_rtadv && 289 !(ifp->if_flags & IFF_LOOPBACK) && 290 (ifp->if_type != IFT_BRIDGE)) 291 nd->flags |= ND6_IFF_ACCEPT_RTADV; 292 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 293 nd->flags |= ND6_IFF_NO_RADR; 294 295 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 296 nd6_setmtu0(ifp, nd); 297 298 return nd; 299 } 300 301 void 302 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd) 303 { 304 struct epoch_tracker et; 305 struct ifaddr *ifa, *next; 306 307 NET_EPOCH_ENTER(et); 308 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 309 if (ifa->ifa_addr->sa_family != AF_INET6) 310 continue; 311 312 /* stop DAD processing */ 313 nd6_dad_stop(ifa); 314 } 315 NET_EPOCH_EXIT(et); 316 317 free(nd, M_IP6NDP); 318 } 319 320 /* 321 * Reset ND level link MTU. This function is called when the physical MTU 322 * changes, which means we might have to adjust the ND level MTU. 323 */ 324 void 325 nd6_setmtu(struct ifnet *ifp) 326 { 327 if (ifp->if_afdata[AF_INET6] == NULL) 328 return; 329 330 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 331 } 332 333 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 334 void 335 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 336 { 337 u_int32_t omaxmtu; 338 339 omaxmtu = ndi->maxmtu; 340 ndi->maxmtu = ifp->if_mtu; 341 342 /* 343 * Decreasing the interface MTU under IPV6 minimum MTU may cause 344 * undesirable situation. We thus notify the operator of the change 345 * explicitly. The check for omaxmtu is necessary to restrict the 346 * log to the case of changing the MTU, not initializing it. 347 */ 348 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 349 log(LOG_NOTICE, "nd6_setmtu0: " 350 "new link MTU on %s (%lu) is too small for IPv6\n", 351 if_name(ifp), (unsigned long)ndi->maxmtu); 352 } 353 354 if (ndi->maxmtu > V_in6_maxmtu) 355 in6_setmaxmtu(); /* check all interfaces just in case */ 356 357 } 358 359 void 360 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 361 { 362 363 bzero(ndopts, sizeof(*ndopts)); 364 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 365 ndopts->nd_opts_last 366 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 367 368 if (icmp6len == 0) { 369 ndopts->nd_opts_done = 1; 370 ndopts->nd_opts_search = NULL; 371 } 372 } 373 374 /* 375 * Take one ND option. 376 */ 377 struct nd_opt_hdr * 378 nd6_option(union nd_opts *ndopts) 379 { 380 struct nd_opt_hdr *nd_opt; 381 int olen; 382 383 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 384 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 385 __func__)); 386 if (ndopts->nd_opts_search == NULL) 387 return NULL; 388 if (ndopts->nd_opts_done) 389 return NULL; 390 391 nd_opt = ndopts->nd_opts_search; 392 393 /* make sure nd_opt_len is inside the buffer */ 394 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 395 bzero(ndopts, sizeof(*ndopts)); 396 return NULL; 397 } 398 399 olen = nd_opt->nd_opt_len << 3; 400 if (olen == 0) { 401 /* 402 * Message validation requires that all included 403 * options have a length that is greater than zero. 404 */ 405 bzero(ndopts, sizeof(*ndopts)); 406 return NULL; 407 } 408 409 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 410 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 411 /* option overruns the end of buffer, invalid */ 412 bzero(ndopts, sizeof(*ndopts)); 413 return NULL; 414 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 415 /* reached the end of options chain */ 416 ndopts->nd_opts_done = 1; 417 ndopts->nd_opts_search = NULL; 418 } 419 return nd_opt; 420 } 421 422 /* 423 * Parse multiple ND options. 424 * This function is much easier to use, for ND routines that do not need 425 * multiple options of the same type. 426 */ 427 int 428 nd6_options(union nd_opts *ndopts) 429 { 430 struct nd_opt_hdr *nd_opt; 431 int i = 0; 432 433 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 434 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 435 __func__)); 436 if (ndopts->nd_opts_search == NULL) 437 return 0; 438 439 while (1) { 440 nd_opt = nd6_option(ndopts); 441 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 442 /* 443 * Message validation requires that all included 444 * options have a length that is greater than zero. 445 */ 446 ICMP6STAT_INC(icp6s_nd_badopt); 447 bzero(ndopts, sizeof(*ndopts)); 448 return -1; 449 } 450 451 if (nd_opt == NULL) 452 goto skip1; 453 454 switch (nd_opt->nd_opt_type) { 455 case ND_OPT_SOURCE_LINKADDR: 456 case ND_OPT_TARGET_LINKADDR: 457 case ND_OPT_MTU: 458 case ND_OPT_REDIRECTED_HEADER: 459 case ND_OPT_NONCE: 460 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 461 nd6log((LOG_INFO, 462 "duplicated ND6 option found (type=%d)\n", 463 nd_opt->nd_opt_type)); 464 /* XXX bark? */ 465 } else { 466 ndopts->nd_opt_array[nd_opt->nd_opt_type] 467 = nd_opt; 468 } 469 break; 470 case ND_OPT_PREFIX_INFORMATION: 471 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 472 ndopts->nd_opt_array[nd_opt->nd_opt_type] 473 = nd_opt; 474 } 475 ndopts->nd_opts_pi_end = 476 (struct nd_opt_prefix_info *)nd_opt; 477 break; 478 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 479 case ND_OPT_RDNSS: /* RFC 6106 */ 480 case ND_OPT_DNSSL: /* RFC 6106 */ 481 /* 482 * Silently ignore options we know and do not care about 483 * in the kernel. 484 */ 485 break; 486 default: 487 /* 488 * Unknown options must be silently ignored, 489 * to accommodate future extension to the protocol. 490 */ 491 nd6log((LOG_DEBUG, 492 "nd6_options: unsupported option %d - " 493 "option ignored\n", nd_opt->nd_opt_type)); 494 } 495 496 skip1: 497 i++; 498 if (i > V_nd6_maxndopt) { 499 ICMP6STAT_INC(icp6s_nd_toomanyopt); 500 nd6log((LOG_INFO, "too many loop in nd opt\n")); 501 break; 502 } 503 504 if (ndopts->nd_opts_done) 505 break; 506 } 507 508 return 0; 509 } 510 511 /* 512 * ND6 timer routine to handle ND6 entries 513 */ 514 static void 515 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 516 { 517 int canceled; 518 519 LLE_WLOCK_ASSERT(ln); 520 521 if (tick < 0) { 522 ln->la_expire = 0; 523 ln->ln_ntick = 0; 524 canceled = callout_stop(&ln->lle_timer); 525 } else { 526 ln->la_expire = time_uptime + tick / hz; 527 LLE_ADDREF(ln); 528 if (tick > INT_MAX) { 529 ln->ln_ntick = tick - INT_MAX; 530 canceled = callout_reset(&ln->lle_timer, INT_MAX, 531 nd6_llinfo_timer, ln); 532 } else { 533 ln->ln_ntick = 0; 534 canceled = callout_reset(&ln->lle_timer, tick, 535 nd6_llinfo_timer, ln); 536 } 537 } 538 if (canceled > 0) 539 LLE_REMREF(ln); 540 } 541 542 /* 543 * Gets source address of the first packet in hold queue 544 * and stores it in @src. 545 * Returns pointer to @src (if hold queue is not empty) or NULL. 546 * 547 * Set noinline to be dtrace-friendly 548 */ 549 static __noinline struct in6_addr * 550 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 551 { 552 struct ip6_hdr hdr; 553 struct mbuf *m; 554 555 if (ln->la_hold == NULL) 556 return (NULL); 557 558 /* 559 * assume every packet in la_hold has the same IP header 560 */ 561 m = ln->la_hold; 562 if (sizeof(hdr) > m->m_len) 563 return (NULL); 564 565 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 566 *src = hdr.ip6_src; 567 568 return (src); 569 } 570 571 /* 572 * Checks if we need to switch from STALE state. 573 * 574 * RFC 4861 requires switching from STALE to DELAY state 575 * on first packet matching entry, waiting V_nd6_delay and 576 * transition to PROBE state (if upper layer confirmation was 577 * not received). 578 * 579 * This code performs a bit differently: 580 * On packet hit we don't change state (but desired state 581 * can be guessed by control plane). However, after V_nd6_delay 582 * seconds code will transition to PROBE state (so DELAY state 583 * is kinda skipped in most situations). 584 * 585 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so 586 * we perform the following upon entering STALE state: 587 * 588 * 1) Arm timer to run each V_nd6_delay seconds to make sure that 589 * if packet was transmitted at the start of given interval, we 590 * would be able to switch to PROBE state in V_nd6_delay seconds 591 * as user expects. 592 * 593 * 2) Reschedule timer until original V_nd6_gctimer expires keeping 594 * lle in STALE state (remaining timer value stored in lle_remtime). 595 * 596 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay 597 * seconds ago. 598 * 599 * Returns non-zero value if the entry is still STALE (storing 600 * the next timer interval in @pdelay). 601 * 602 * Returns zero value if original timer expired or we need to switch to 603 * PROBE (store that in @do_switch variable). 604 */ 605 static int 606 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) 607 { 608 int nd_delay, nd_gctimer, r_skip_req; 609 time_t lle_hittime; 610 long delay; 611 612 *do_switch = 0; 613 nd_gctimer = V_nd6_gctimer; 614 nd_delay = V_nd6_delay; 615 616 LLE_REQ_LOCK(lle); 617 r_skip_req = lle->r_skip_req; 618 lle_hittime = lle->lle_hittime; 619 LLE_REQ_UNLOCK(lle); 620 621 if (r_skip_req > 0) { 622 623 /* 624 * Nonzero r_skip_req value was set upon entering 625 * STALE state. Since value was not changed, no 626 * packets were passed using this lle. Ask for 627 * timer reschedule and keep STALE state. 628 */ 629 delay = (long)(MIN(nd_gctimer, nd_delay)); 630 delay *= hz; 631 if (lle->lle_remtime > delay) 632 lle->lle_remtime -= delay; 633 else { 634 delay = lle->lle_remtime; 635 lle->lle_remtime = 0; 636 } 637 638 if (delay == 0) { 639 640 /* 641 * The original ng6_gctime timeout ended, 642 * no more rescheduling. 643 */ 644 return (0); 645 } 646 647 *pdelay = delay; 648 return (1); 649 } 650 651 /* 652 * Packet received. Verify timestamp 653 */ 654 delay = (long)(time_uptime - lle_hittime); 655 if (delay < nd_delay) { 656 657 /* 658 * V_nd6_delay still not passed since the first 659 * hit in STALE state. 660 * Reshedule timer and return. 661 */ 662 *pdelay = (long)(nd_delay - delay) * hz; 663 return (1); 664 } 665 666 /* Request switching to probe */ 667 *do_switch = 1; 668 return (0); 669 } 670 671 672 /* 673 * Switch @lle state to new state optionally arming timers. 674 * 675 * Set noinline to be dtrace-friendly 676 */ 677 __noinline void 678 nd6_llinfo_setstate(struct llentry *lle, int newstate) 679 { 680 struct ifnet *ifp; 681 int nd_gctimer, nd_delay; 682 long delay, remtime; 683 684 delay = 0; 685 remtime = 0; 686 687 switch (newstate) { 688 case ND6_LLINFO_INCOMPLETE: 689 ifp = lle->lle_tbl->llt_ifp; 690 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; 691 break; 692 case ND6_LLINFO_REACHABLE: 693 if (!ND6_LLINFO_PERMANENT(lle)) { 694 ifp = lle->lle_tbl->llt_ifp; 695 delay = (long)ND_IFINFO(ifp)->reachable * hz; 696 } 697 break; 698 case ND6_LLINFO_STALE: 699 700 /* 701 * Notify fast path that we want to know if any packet 702 * is transmitted by setting r_skip_req. 703 */ 704 LLE_REQ_LOCK(lle); 705 lle->r_skip_req = 1; 706 LLE_REQ_UNLOCK(lle); 707 nd_delay = V_nd6_delay; 708 nd_gctimer = V_nd6_gctimer; 709 710 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz; 711 remtime = (long)nd_gctimer * hz - delay; 712 break; 713 case ND6_LLINFO_DELAY: 714 lle->la_asked = 0; 715 delay = (long)V_nd6_delay * hz; 716 break; 717 } 718 719 if (delay > 0) 720 nd6_llinfo_settimer_locked(lle, delay); 721 722 lle->lle_remtime = remtime; 723 lle->ln_state = newstate; 724 } 725 726 /* 727 * Timer-dependent part of nd state machine. 728 * 729 * Set noinline to be dtrace-friendly 730 */ 731 static __noinline void 732 nd6_llinfo_timer(void *arg) 733 { 734 struct llentry *ln; 735 struct in6_addr *dst, *pdst, *psrc, src; 736 struct ifnet *ifp; 737 struct nd_ifinfo *ndi; 738 int do_switch, send_ns; 739 long delay; 740 741 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 742 ln = (struct llentry *)arg; 743 ifp = lltable_get_ifp(ln->lle_tbl); 744 CURVNET_SET(ifp->if_vnet); 745 746 ND6_RLOCK(); 747 LLE_WLOCK(ln); 748 if (callout_pending(&ln->lle_timer)) { 749 /* 750 * Here we are a bit odd here in the treatment of 751 * active/pending. If the pending bit is set, it got 752 * rescheduled before I ran. The active 753 * bit we ignore, since if it was stopped 754 * in ll_tablefree() and was currently running 755 * it would have return 0 so the code would 756 * not have deleted it since the callout could 757 * not be stopped so we want to go through 758 * with the delete here now. If the callout 759 * was restarted, the pending bit will be back on and 760 * we just want to bail since the callout_reset would 761 * return 1 and our reference would have been removed 762 * by nd6_llinfo_settimer_locked above since canceled 763 * would have been 1. 764 */ 765 LLE_WUNLOCK(ln); 766 ND6_RUNLOCK(); 767 CURVNET_RESTORE(); 768 return; 769 } 770 ndi = ND_IFINFO(ifp); 771 send_ns = 0; 772 dst = &ln->r_l3addr.addr6; 773 pdst = dst; 774 775 if (ln->ln_ntick > 0) { 776 if (ln->ln_ntick > INT_MAX) { 777 ln->ln_ntick -= INT_MAX; 778 nd6_llinfo_settimer_locked(ln, INT_MAX); 779 } else { 780 ln->ln_ntick = 0; 781 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 782 } 783 goto done; 784 } 785 786 if (ln->la_flags & LLE_STATIC) { 787 goto done; 788 } 789 790 if (ln->la_flags & LLE_DELETED) { 791 nd6_free(&ln, 0); 792 goto done; 793 } 794 795 switch (ln->ln_state) { 796 case ND6_LLINFO_INCOMPLETE: 797 if (ln->la_asked < V_nd6_mmaxtries) { 798 ln->la_asked++; 799 send_ns = 1; 800 /* Send NS to multicast address */ 801 pdst = NULL; 802 } else { 803 struct mbuf *m = ln->la_hold; 804 if (m) { 805 struct mbuf *m0; 806 807 /* 808 * assuming every packet in la_hold has the 809 * same IP header. Send error after unlock. 810 */ 811 m0 = m->m_nextpkt; 812 m->m_nextpkt = NULL; 813 ln->la_hold = m0; 814 clear_llinfo_pqueue(ln); 815 } 816 nd6_free(&ln, 0); 817 if (m != NULL) 818 icmp6_error2(m, ICMP6_DST_UNREACH, 819 ICMP6_DST_UNREACH_ADDR, 0, ifp); 820 } 821 break; 822 case ND6_LLINFO_REACHABLE: 823 if (!ND6_LLINFO_PERMANENT(ln)) 824 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 825 break; 826 827 case ND6_LLINFO_STALE: 828 if (nd6_is_stale(ln, &delay, &do_switch) != 0) { 829 830 /* 831 * No packet has used this entry and GC timeout 832 * has not been passed. Reshedule timer and 833 * return. 834 */ 835 nd6_llinfo_settimer_locked(ln, delay); 836 break; 837 } 838 839 if (do_switch == 0) { 840 841 /* 842 * GC timer has ended and entry hasn't been used. 843 * Run Garbage collector (RFC 4861, 5.3) 844 */ 845 if (!ND6_LLINFO_PERMANENT(ln)) 846 nd6_free(&ln, 1); 847 break; 848 } 849 850 /* Entry has been used AND delay timer has ended. */ 851 852 /* FALLTHROUGH */ 853 854 case ND6_LLINFO_DELAY: 855 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 856 /* We need NUD */ 857 ln->la_asked = 1; 858 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE); 859 send_ns = 1; 860 } else 861 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */ 862 break; 863 case ND6_LLINFO_PROBE: 864 if (ln->la_asked < V_nd6_umaxtries) { 865 ln->la_asked++; 866 send_ns = 1; 867 } else { 868 nd6_free(&ln, 0); 869 } 870 break; 871 default: 872 panic("%s: paths in a dark night can be confusing: %d", 873 __func__, ln->ln_state); 874 } 875 done: 876 if (ln != NULL) 877 ND6_RUNLOCK(); 878 if (send_ns != 0) { 879 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 880 psrc = nd6_llinfo_get_holdsrc(ln, &src); 881 LLE_FREE_LOCKED(ln); 882 ln = NULL; 883 nd6_ns_output(ifp, psrc, pdst, dst, NULL); 884 } 885 886 if (ln != NULL) 887 LLE_FREE_LOCKED(ln); 888 CURVNET_RESTORE(); 889 } 890 891 892 /* 893 * ND6 timer routine to expire default route list and prefix list 894 */ 895 void 896 nd6_timer(void *arg) 897 { 898 CURVNET_SET((struct vnet *) arg); 899 struct nd_drhead drq; 900 struct nd_prhead prl; 901 struct nd_defrouter *dr, *ndr; 902 struct nd_prefix *pr, *npr; 903 struct in6_ifaddr *ia6, *nia6; 904 uint64_t genid; 905 906 TAILQ_INIT(&drq); 907 LIST_INIT(&prl); 908 909 ND6_WLOCK(); 910 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) 911 if (dr->expire && dr->expire < time_uptime) 912 defrouter_unlink(dr, &drq); 913 ND6_WUNLOCK(); 914 915 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 916 TAILQ_REMOVE(&drq, dr, dr_entry); 917 defrouter_del(dr); 918 } 919 920 /* 921 * expire interface addresses. 922 * in the past the loop was inside prefix expiry processing. 923 * However, from a stricter speci-confrmance standpoint, we should 924 * rather separate address lifetimes and prefix lifetimes. 925 * 926 * XXXRW: in6_ifaddrhead locking. 927 */ 928 addrloop: 929 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 930 /* check address lifetime */ 931 if (IFA6_IS_INVALID(ia6)) { 932 int regen = 0; 933 934 /* 935 * If the expiring address is temporary, try 936 * regenerating a new one. This would be useful when 937 * we suspended a laptop PC, then turned it on after a 938 * period that could invalidate all temporary 939 * addresses. Although we may have to restart the 940 * loop (see below), it must be after purging the 941 * address. Otherwise, we'd see an infinite loop of 942 * regeneration. 943 */ 944 if (V_ip6_use_tempaddr && 945 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 946 if (regen_tmpaddr(ia6) == 0) 947 regen = 1; 948 } 949 950 in6_purgeaddr(&ia6->ia_ifa); 951 952 if (regen) 953 goto addrloop; /* XXX: see below */ 954 } else if (IFA6_IS_DEPRECATED(ia6)) { 955 int oldflags = ia6->ia6_flags; 956 957 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 958 959 /* 960 * If a temporary address has just become deprecated, 961 * regenerate a new one if possible. 962 */ 963 if (V_ip6_use_tempaddr && 964 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 965 (oldflags & IN6_IFF_DEPRECATED) == 0) { 966 967 if (regen_tmpaddr(ia6) == 0) { 968 /* 969 * A new temporary address is 970 * generated. 971 * XXX: this means the address chain 972 * has changed while we are still in 973 * the loop. Although the change 974 * would not cause disaster (because 975 * it's not a deletion, but an 976 * addition,) we'd rather restart the 977 * loop just for safety. Or does this 978 * significantly reduce performance?? 979 */ 980 goto addrloop; 981 } 982 } 983 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 984 /* 985 * Schedule DAD for a tentative address. This happens 986 * if the interface was down or not running 987 * when the address was configured. 988 */ 989 int delay; 990 991 delay = arc4random() % 992 (MAX_RTR_SOLICITATION_DELAY * hz); 993 nd6_dad_start((struct ifaddr *)ia6, delay); 994 } else { 995 /* 996 * Check status of the interface. If it is down, 997 * mark the address as tentative for future DAD. 998 */ 999 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 || 1000 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING) 1001 == 0 || 1002 (ND_IFINFO(ia6->ia_ifp)->flags & 1003 ND6_IFF_IFDISABLED) != 0) { 1004 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 1005 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 1006 } 1007 /* 1008 * A new RA might have made a deprecated address 1009 * preferred. 1010 */ 1011 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 1012 } 1013 } 1014 1015 ND6_WLOCK(); 1016 restart: 1017 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1018 /* 1019 * Expire prefixes. Since the pltime is only used for 1020 * autoconfigured addresses, pltime processing for prefixes is 1021 * not necessary. 1022 * 1023 * Only unlink after all derived addresses have expired. This 1024 * may not occur until two hours after the prefix has expired 1025 * per RFC 4862. If the prefix expires before its derived 1026 * addresses, mark it off-link. This will be done automatically 1027 * after unlinking if no address references remain. 1028 */ 1029 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME || 1030 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime) 1031 continue; 1032 1033 if (pr->ndpr_addrcnt == 0) { 1034 nd6_prefix_unlink(pr, &prl); 1035 continue; 1036 } 1037 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1038 genid = V_nd6_list_genid; 1039 nd6_prefix_ref(pr); 1040 ND6_WUNLOCK(); 1041 ND6_ONLINK_LOCK(); 1042 (void)nd6_prefix_offlink(pr); 1043 ND6_ONLINK_UNLOCK(); 1044 ND6_WLOCK(); 1045 nd6_prefix_rele(pr); 1046 if (genid != V_nd6_list_genid) 1047 goto restart; 1048 } 1049 } 1050 ND6_WUNLOCK(); 1051 1052 while ((pr = LIST_FIRST(&prl)) != NULL) { 1053 LIST_REMOVE(pr, ndpr_entry); 1054 nd6_prefix_del(pr); 1055 } 1056 1057 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 1058 nd6_timer, curvnet); 1059 1060 CURVNET_RESTORE(); 1061 } 1062 1063 /* 1064 * ia6 - deprecated/invalidated temporary address 1065 */ 1066 static int 1067 regen_tmpaddr(struct in6_ifaddr *ia6) 1068 { 1069 struct epoch_tracker et; 1070 struct ifaddr *ifa; 1071 struct ifnet *ifp; 1072 struct in6_ifaddr *public_ifa6 = NULL; 1073 1074 ifp = ia6->ia_ifa.ifa_ifp; 1075 NET_EPOCH_ENTER(et); 1076 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1077 struct in6_ifaddr *it6; 1078 1079 if (ifa->ifa_addr->sa_family != AF_INET6) 1080 continue; 1081 1082 it6 = (struct in6_ifaddr *)ifa; 1083 1084 /* ignore no autoconf addresses. */ 1085 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1086 continue; 1087 1088 /* ignore autoconf addresses with different prefixes. */ 1089 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 1090 continue; 1091 1092 /* 1093 * Now we are looking at an autoconf address with the same 1094 * prefix as ours. If the address is temporary and is still 1095 * preferred, do not create another one. It would be rare, but 1096 * could happen, for example, when we resume a laptop PC after 1097 * a long period. 1098 */ 1099 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1100 !IFA6_IS_DEPRECATED(it6)) { 1101 public_ifa6 = NULL; 1102 break; 1103 } 1104 1105 /* 1106 * This is a public autoconf address that has the same prefix 1107 * as ours. If it is preferred, keep it. We can't break the 1108 * loop here, because there may be a still-preferred temporary 1109 * address with the prefix. 1110 */ 1111 if (!IFA6_IS_DEPRECATED(it6)) 1112 public_ifa6 = it6; 1113 } 1114 if (public_ifa6 != NULL) 1115 ifa_ref(&public_ifa6->ia_ifa); 1116 NET_EPOCH_EXIT(et); 1117 1118 if (public_ifa6 != NULL) { 1119 int e; 1120 1121 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 1122 ifa_free(&public_ifa6->ia_ifa); 1123 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 1124 " tmp addr,errno=%d\n", e); 1125 return (-1); 1126 } 1127 ifa_free(&public_ifa6->ia_ifa); 1128 return (0); 1129 } 1130 1131 return (-1); 1132 } 1133 1134 /* 1135 * Remove prefix and default router list entries corresponding to ifp. Neighbor 1136 * cache entries are freed in in6_domifdetach(). 1137 */ 1138 void 1139 nd6_purge(struct ifnet *ifp) 1140 { 1141 struct nd_drhead drq; 1142 struct nd_prhead prl; 1143 struct nd_defrouter *dr, *ndr; 1144 struct nd_prefix *pr, *npr; 1145 1146 TAILQ_INIT(&drq); 1147 LIST_INIT(&prl); 1148 1149 /* 1150 * Nuke default router list entries toward ifp. 1151 * We defer removal of default router list entries that is installed 1152 * in the routing table, in order to keep additional side effects as 1153 * small as possible. 1154 */ 1155 ND6_WLOCK(); 1156 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 1157 if (dr->installed) 1158 continue; 1159 if (dr->ifp == ifp) 1160 defrouter_unlink(dr, &drq); 1161 } 1162 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 1163 if (!dr->installed) 1164 continue; 1165 if (dr->ifp == ifp) 1166 defrouter_unlink(dr, &drq); 1167 } 1168 1169 /* 1170 * Remove prefixes on ifp. We should have already removed addresses on 1171 * this interface, so no addresses should be referencing these prefixes. 1172 */ 1173 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1174 if (pr->ndpr_ifp == ifp) 1175 nd6_prefix_unlink(pr, &prl); 1176 } 1177 ND6_WUNLOCK(); 1178 1179 /* Delete the unlinked router and prefix objects. */ 1180 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 1181 TAILQ_REMOVE(&drq, dr, dr_entry); 1182 defrouter_del(dr); 1183 } 1184 while ((pr = LIST_FIRST(&prl)) != NULL) { 1185 LIST_REMOVE(pr, ndpr_entry); 1186 nd6_prefix_del(pr); 1187 } 1188 1189 /* cancel default outgoing interface setting */ 1190 if (V_nd6_defifindex == ifp->if_index) 1191 nd6_setdefaultiface(0); 1192 1193 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1194 /* Refresh default router list. */ 1195 defrouter_select_fib(ifp->if_fib); 1196 } 1197 } 1198 1199 /* 1200 * the caller acquires and releases the lock on the lltbls 1201 * Returns the llentry locked 1202 */ 1203 struct llentry * 1204 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1205 { 1206 struct sockaddr_in6 sin6; 1207 struct llentry *ln; 1208 1209 bzero(&sin6, sizeof(sin6)); 1210 sin6.sin6_len = sizeof(struct sockaddr_in6); 1211 sin6.sin6_family = AF_INET6; 1212 sin6.sin6_addr = *addr6; 1213 1214 IF_AFDATA_LOCK_ASSERT(ifp); 1215 1216 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6); 1217 1218 return (ln); 1219 } 1220 1221 struct llentry * 1222 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1223 { 1224 struct sockaddr_in6 sin6; 1225 struct llentry *ln; 1226 1227 bzero(&sin6, sizeof(sin6)); 1228 sin6.sin6_len = sizeof(struct sockaddr_in6); 1229 sin6.sin6_family = AF_INET6; 1230 sin6.sin6_addr = *addr6; 1231 1232 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 1233 if (ln != NULL) 1234 ln->ln_state = ND6_LLINFO_NOSTATE; 1235 1236 return (ln); 1237 } 1238 1239 /* 1240 * Test whether a given IPv6 address is a neighbor or not, ignoring 1241 * the actual neighbor cache. The neighbor cache is ignored in order 1242 * to not reenter the routing code from within itself. 1243 */ 1244 static int 1245 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1246 { 1247 struct nd_prefix *pr; 1248 struct ifaddr *ifa; 1249 struct rt_addrinfo info; 1250 struct sockaddr_in6 rt_key; 1251 const struct sockaddr *dst6; 1252 uint64_t genid; 1253 int error, fibnum; 1254 1255 /* 1256 * A link-local address is always a neighbor. 1257 * XXX: a link does not necessarily specify a single interface. 1258 */ 1259 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 1260 struct sockaddr_in6 sin6_copy; 1261 u_int32_t zone; 1262 1263 /* 1264 * We need sin6_copy since sa6_recoverscope() may modify the 1265 * content (XXX). 1266 */ 1267 sin6_copy = *addr; 1268 if (sa6_recoverscope(&sin6_copy)) 1269 return (0); /* XXX: should be impossible */ 1270 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 1271 return (0); 1272 if (sin6_copy.sin6_scope_id == zone) 1273 return (1); 1274 else 1275 return (0); 1276 } 1277 1278 bzero(&rt_key, sizeof(rt_key)); 1279 bzero(&info, sizeof(info)); 1280 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1281 1282 /* 1283 * If the address matches one of our addresses, 1284 * it should be a neighbor. 1285 * If the address matches one of our on-link prefixes, it should be a 1286 * neighbor. 1287 */ 1288 ND6_RLOCK(); 1289 restart: 1290 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1291 if (pr->ndpr_ifp != ifp) 1292 continue; 1293 1294 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1295 dst6 = (const struct sockaddr *)&pr->ndpr_prefix; 1296 1297 /* 1298 * We only need to check all FIBs if add_addr_allfibs 1299 * is unset. If set, checking any FIB will suffice. 1300 */ 1301 fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0; 1302 for (; fibnum < rt_numfibs; fibnum++) { 1303 genid = V_nd6_list_genid; 1304 ND6_RUNLOCK(); 1305 1306 /* 1307 * Restore length field before 1308 * retrying lookup 1309 */ 1310 rt_key.sin6_len = sizeof(rt_key); 1311 error = rib_lookup_info(fibnum, dst6, 0, 0, 1312 &info); 1313 1314 ND6_RLOCK(); 1315 if (genid != V_nd6_list_genid) 1316 goto restart; 1317 if (error == 0) 1318 break; 1319 } 1320 if (error != 0) 1321 continue; 1322 1323 /* 1324 * This is the case where multiple interfaces 1325 * have the same prefix, but only one is installed 1326 * into the routing table and that prefix entry 1327 * is not the one being examined here. In the case 1328 * where RADIX_MPATH is enabled, multiple route 1329 * entries (of the same rt_key value) will be 1330 * installed because the interface addresses all 1331 * differ. 1332 */ 1333 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1334 &rt_key.sin6_addr)) 1335 continue; 1336 } 1337 1338 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1339 &addr->sin6_addr, &pr->ndpr_mask)) { 1340 ND6_RUNLOCK(); 1341 return (1); 1342 } 1343 } 1344 ND6_RUNLOCK(); 1345 1346 /* 1347 * If the address is assigned on the node of the other side of 1348 * a p2p interface, the address should be a neighbor. 1349 */ 1350 if (ifp->if_flags & IFF_POINTOPOINT) { 1351 struct epoch_tracker et; 1352 1353 NET_EPOCH_ENTER(et); 1354 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1355 if (ifa->ifa_addr->sa_family != addr->sin6_family) 1356 continue; 1357 if (ifa->ifa_dstaddr != NULL && 1358 sa_equal(addr, ifa->ifa_dstaddr)) { 1359 NET_EPOCH_EXIT(et); 1360 return 1; 1361 } 1362 } 1363 NET_EPOCH_EXIT(et); 1364 } 1365 1366 /* 1367 * If the default router list is empty, all addresses are regarded 1368 * as on-link, and thus, as a neighbor. 1369 */ 1370 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1371 TAILQ_EMPTY(&V_nd_defrouter) && 1372 V_nd6_defifindex == ifp->if_index) { 1373 return (1); 1374 } 1375 1376 return (0); 1377 } 1378 1379 1380 /* 1381 * Detect if a given IPv6 address identifies a neighbor on a given link. 1382 * XXX: should take care of the destination of a p2p link? 1383 */ 1384 int 1385 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1386 { 1387 struct epoch_tracker et; 1388 struct llentry *lle; 1389 int rc = 0; 1390 1391 IF_AFDATA_UNLOCK_ASSERT(ifp); 1392 if (nd6_is_new_addr_neighbor(addr, ifp)) 1393 return (1); 1394 1395 /* 1396 * Even if the address matches none of our addresses, it might be 1397 * in the neighbor cache. 1398 */ 1399 NET_EPOCH_ENTER(et); 1400 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1401 LLE_RUNLOCK(lle); 1402 rc = 1; 1403 } 1404 NET_EPOCH_EXIT(et); 1405 return (rc); 1406 } 1407 1408 /* 1409 * Free an nd6 llinfo entry. 1410 * Since the function would cause significant changes in the kernel, DO NOT 1411 * make it global, unless you have a strong reason for the change, and are sure 1412 * that the change is safe. 1413 * 1414 * Set noinline to be dtrace-friendly 1415 */ 1416 static __noinline void 1417 nd6_free(struct llentry **lnp, int gc) 1418 { 1419 struct ifnet *ifp; 1420 struct llentry *ln; 1421 struct nd_defrouter *dr; 1422 1423 ln = *lnp; 1424 *lnp = NULL; 1425 1426 LLE_WLOCK_ASSERT(ln); 1427 ND6_RLOCK_ASSERT(); 1428 1429 ifp = lltable_get_ifp(ln->lle_tbl); 1430 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0) 1431 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp); 1432 else 1433 dr = NULL; 1434 ND6_RUNLOCK(); 1435 1436 if ((ln->la_flags & LLE_DELETED) == 0) 1437 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 1438 1439 /* 1440 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1441 * even though it is not harmful, it was not really necessary. 1442 */ 1443 1444 /* cancel timer */ 1445 nd6_llinfo_settimer_locked(ln, -1); 1446 1447 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1448 if (dr != NULL && dr->expire && 1449 ln->ln_state == ND6_LLINFO_STALE && gc) { 1450 /* 1451 * If the reason for the deletion is just garbage 1452 * collection, and the neighbor is an active default 1453 * router, do not delete it. Instead, reset the GC 1454 * timer using the router's lifetime. 1455 * Simply deleting the entry would affect default 1456 * router selection, which is not necessarily a good 1457 * thing, especially when we're using router preference 1458 * values. 1459 * XXX: the check for ln_state would be redundant, 1460 * but we intentionally keep it just in case. 1461 */ 1462 if (dr->expire > time_uptime) 1463 nd6_llinfo_settimer_locked(ln, 1464 (dr->expire - time_uptime) * hz); 1465 else 1466 nd6_llinfo_settimer_locked(ln, 1467 (long)V_nd6_gctimer * hz); 1468 1469 LLE_REMREF(ln); 1470 LLE_WUNLOCK(ln); 1471 defrouter_rele(dr); 1472 return; 1473 } 1474 1475 if (dr) { 1476 /* 1477 * Unreachablity of a router might affect the default 1478 * router selection and on-link detection of advertised 1479 * prefixes. 1480 */ 1481 1482 /* 1483 * Temporarily fake the state to choose a new default 1484 * router and to perform on-link determination of 1485 * prefixes correctly. 1486 * Below the state will be set correctly, 1487 * or the entry itself will be deleted. 1488 */ 1489 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1490 } 1491 1492 if (ln->ln_router || dr) { 1493 1494 /* 1495 * We need to unlock to avoid a LOR with rt6_flush() with the 1496 * rnh and for the calls to pfxlist_onlink_check() and 1497 * defrouter_select_fib() in the block further down for calls 1498 * into nd6_lookup(). We still hold a ref. 1499 */ 1500 LLE_WUNLOCK(ln); 1501 1502 /* 1503 * rt6_flush must be called whether or not the neighbor 1504 * is in the Default Router List. 1505 * See a corresponding comment in nd6_na_input(). 1506 */ 1507 rt6_flush(&ln->r_l3addr.addr6, ifp); 1508 } 1509 1510 if (dr) { 1511 /* 1512 * Since defrouter_select_fib() does not affect the 1513 * on-link determination and MIP6 needs the check 1514 * before the default router selection, we perform 1515 * the check now. 1516 */ 1517 pfxlist_onlink_check(); 1518 1519 /* 1520 * Refresh default router list. 1521 */ 1522 defrouter_select_fib(dr->ifp->if_fib); 1523 } 1524 1525 /* 1526 * If this entry was added by an on-link redirect, remove the 1527 * corresponding host route. 1528 */ 1529 if (ln->la_flags & LLE_REDIRECT) 1530 nd6_free_redirect(ln); 1531 1532 if (ln->ln_router || dr) 1533 LLE_WLOCK(ln); 1534 } 1535 1536 /* 1537 * Save to unlock. We still hold an extra reference and will not 1538 * free(9) in llentry_free() if someone else holds one as well. 1539 */ 1540 LLE_WUNLOCK(ln); 1541 IF_AFDATA_LOCK(ifp); 1542 LLE_WLOCK(ln); 1543 /* Guard against race with other llentry_free(). */ 1544 if (ln->la_flags & LLE_LINKED) { 1545 /* Remove callout reference */ 1546 LLE_REMREF(ln); 1547 lltable_unlink_entry(ln->lle_tbl, ln); 1548 } 1549 IF_AFDATA_UNLOCK(ifp); 1550 1551 llentry_free(ln); 1552 if (dr != NULL) 1553 defrouter_rele(dr); 1554 } 1555 1556 static int 1557 nd6_isdynrte(const struct rtentry *rt, void *xap) 1558 { 1559 1560 if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC)) 1561 return (1); 1562 1563 return (0); 1564 } 1565 /* 1566 * Remove the rtentry for the given llentry, 1567 * both of which were installed by a redirect. 1568 */ 1569 static void 1570 nd6_free_redirect(const struct llentry *ln) 1571 { 1572 int fibnum; 1573 struct sockaddr_in6 sin6; 1574 struct rt_addrinfo info; 1575 1576 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); 1577 memset(&info, 0, sizeof(info)); 1578 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; 1579 info.rti_filter = nd6_isdynrte; 1580 1581 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 1582 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum); 1583 } 1584 1585 /* 1586 * Rejuvenate this function for routing operations related 1587 * processing. 1588 */ 1589 void 1590 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1591 { 1592 struct sockaddr_in6 *gateway; 1593 struct nd_defrouter *dr; 1594 struct ifnet *ifp; 1595 1596 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1597 ifp = rt->rt_ifp; 1598 1599 switch (req) { 1600 case RTM_ADD: 1601 break; 1602 1603 case RTM_DELETE: 1604 if (!ifp) 1605 return; 1606 /* 1607 * Only indirect routes are interesting. 1608 */ 1609 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1610 return; 1611 /* 1612 * check for default route 1613 */ 1614 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1615 &SIN6(rt_key(rt))->sin6_addr)) { 1616 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1617 if (dr != NULL) { 1618 dr->installed = 0; 1619 defrouter_rele(dr); 1620 } 1621 } 1622 break; 1623 } 1624 } 1625 1626 1627 int 1628 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1629 { 1630 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1631 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1632 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1633 struct epoch_tracker et; 1634 int error = 0; 1635 1636 if (ifp->if_afdata[AF_INET6] == NULL) 1637 return (EPFNOSUPPORT); 1638 switch (cmd) { 1639 case OSIOCGIFINFO_IN6: 1640 #define ND ndi->ndi 1641 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1642 bzero(&ND, sizeof(ND)); 1643 ND.linkmtu = IN6_LINKMTU(ifp); 1644 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1645 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1646 ND.reachable = ND_IFINFO(ifp)->reachable; 1647 ND.retrans = ND_IFINFO(ifp)->retrans; 1648 ND.flags = ND_IFINFO(ifp)->flags; 1649 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1650 ND.chlim = ND_IFINFO(ifp)->chlim; 1651 break; 1652 case SIOCGIFINFO_IN6: 1653 ND = *ND_IFINFO(ifp); 1654 break; 1655 case SIOCSIFINFO_IN6: 1656 /* 1657 * used to change host variables from userland. 1658 * intended for a use on router to reflect RA configurations. 1659 */ 1660 /* 0 means 'unspecified' */ 1661 if (ND.linkmtu != 0) { 1662 if (ND.linkmtu < IPV6_MMTU || 1663 ND.linkmtu > IN6_LINKMTU(ifp)) { 1664 error = EINVAL; 1665 break; 1666 } 1667 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1668 } 1669 1670 if (ND.basereachable != 0) { 1671 int obasereachable = ND_IFINFO(ifp)->basereachable; 1672 1673 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1674 if (ND.basereachable != obasereachable) 1675 ND_IFINFO(ifp)->reachable = 1676 ND_COMPUTE_RTIME(ND.basereachable); 1677 } 1678 if (ND.retrans != 0) 1679 ND_IFINFO(ifp)->retrans = ND.retrans; 1680 if (ND.chlim != 0) 1681 ND_IFINFO(ifp)->chlim = ND.chlim; 1682 /* FALLTHROUGH */ 1683 case SIOCSIFINFO_FLAGS: 1684 { 1685 struct ifaddr *ifa; 1686 struct in6_ifaddr *ia; 1687 1688 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1689 !(ND.flags & ND6_IFF_IFDISABLED)) { 1690 /* ifdisabled 1->0 transision */ 1691 1692 /* 1693 * If the interface is marked as ND6_IFF_IFDISABLED and 1694 * has an link-local address with IN6_IFF_DUPLICATED, 1695 * do not clear ND6_IFF_IFDISABLED. 1696 * See RFC 4862, Section 5.4.5. 1697 */ 1698 NET_EPOCH_ENTER(et); 1699 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1700 if (ifa->ifa_addr->sa_family != AF_INET6) 1701 continue; 1702 ia = (struct in6_ifaddr *)ifa; 1703 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1704 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1705 break; 1706 } 1707 NET_EPOCH_EXIT(et); 1708 1709 if (ifa != NULL) { 1710 /* LLA is duplicated. */ 1711 ND.flags |= ND6_IFF_IFDISABLED; 1712 log(LOG_ERR, "Cannot enable an interface" 1713 " with a link-local address marked" 1714 " duplicate.\n"); 1715 } else { 1716 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1717 if (ifp->if_flags & IFF_UP) 1718 in6_if_up(ifp); 1719 } 1720 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1721 (ND.flags & ND6_IFF_IFDISABLED)) { 1722 /* ifdisabled 0->1 transision */ 1723 /* Mark all IPv6 address as tentative. */ 1724 1725 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1726 if (V_ip6_dad_count > 0 && 1727 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1728 NET_EPOCH_ENTER(et); 1729 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1730 ifa_link) { 1731 if (ifa->ifa_addr->sa_family != 1732 AF_INET6) 1733 continue; 1734 ia = (struct in6_ifaddr *)ifa; 1735 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1736 } 1737 NET_EPOCH_EXIT(et); 1738 } 1739 } 1740 1741 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1742 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1743 /* auto_linklocal 0->1 transision */ 1744 1745 /* If no link-local address on ifp, configure */ 1746 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1747 in6_ifattach(ifp, NULL); 1748 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1749 ifp->if_flags & IFF_UP) { 1750 /* 1751 * When the IF already has 1752 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1753 * address is assigned, and IFF_UP, try to 1754 * assign one. 1755 */ 1756 NET_EPOCH_ENTER(et); 1757 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1758 ifa_link) { 1759 if (ifa->ifa_addr->sa_family != 1760 AF_INET6) 1761 continue; 1762 ia = (struct in6_ifaddr *)ifa; 1763 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1764 break; 1765 } 1766 NET_EPOCH_EXIT(et); 1767 if (ifa != NULL) 1768 /* No LLA is configured. */ 1769 in6_ifattach(ifp, NULL); 1770 } 1771 } 1772 } 1773 ND_IFINFO(ifp)->flags = ND.flags; 1774 break; 1775 #undef ND 1776 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1777 /* sync kernel routing table with the default router list */ 1778 defrouter_reset(); 1779 defrouter_select(); 1780 break; 1781 case SIOCSPFXFLUSH_IN6: 1782 { 1783 /* flush all the prefix advertised by routers */ 1784 struct in6_ifaddr *ia, *ia_next; 1785 struct nd_prefix *pr, *next; 1786 struct nd_prhead prl; 1787 1788 LIST_INIT(&prl); 1789 1790 ND6_WLOCK(); 1791 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1792 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1793 continue; /* XXX */ 1794 nd6_prefix_unlink(pr, &prl); 1795 } 1796 ND6_WUNLOCK(); 1797 1798 while ((pr = LIST_FIRST(&prl)) != NULL) { 1799 LIST_REMOVE(pr, ndpr_entry); 1800 /* XXXRW: in6_ifaddrhead locking. */ 1801 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1802 ia_next) { 1803 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1804 continue; 1805 1806 if (ia->ia6_ndpr == pr) 1807 in6_purgeaddr(&ia->ia_ifa); 1808 } 1809 nd6_prefix_del(pr); 1810 } 1811 break; 1812 } 1813 case SIOCSRTRFLUSH_IN6: 1814 { 1815 /* flush all the default routers */ 1816 struct nd_drhead drq; 1817 struct nd_defrouter *dr; 1818 1819 TAILQ_INIT(&drq); 1820 1821 defrouter_reset(); 1822 1823 ND6_WLOCK(); 1824 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL) 1825 defrouter_unlink(dr, &drq); 1826 ND6_WUNLOCK(); 1827 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 1828 TAILQ_REMOVE(&drq, dr, dr_entry); 1829 defrouter_del(dr); 1830 } 1831 1832 defrouter_select(); 1833 break; 1834 } 1835 case SIOCGNBRINFO_IN6: 1836 { 1837 struct llentry *ln; 1838 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1839 1840 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1841 return (error); 1842 1843 NET_EPOCH_ENTER(et); 1844 ln = nd6_lookup(&nb_addr, 0, ifp); 1845 NET_EPOCH_EXIT(et); 1846 1847 if (ln == NULL) { 1848 error = EINVAL; 1849 break; 1850 } 1851 nbi->state = ln->ln_state; 1852 nbi->asked = ln->la_asked; 1853 nbi->isrouter = ln->ln_router; 1854 if (ln->la_expire == 0) 1855 nbi->expire = 0; 1856 else 1857 nbi->expire = ln->la_expire + ln->lle_remtime / hz + 1858 (time_second - time_uptime); 1859 LLE_RUNLOCK(ln); 1860 break; 1861 } 1862 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1863 ndif->ifindex = V_nd6_defifindex; 1864 break; 1865 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1866 return (nd6_setdefaultiface(ndif->ifindex)); 1867 } 1868 return (error); 1869 } 1870 1871 /* 1872 * Calculates new isRouter value based on provided parameters and 1873 * returns it. 1874 */ 1875 static int 1876 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, 1877 int ln_router) 1878 { 1879 1880 /* 1881 * ICMP6 type dependent behavior. 1882 * 1883 * NS: clear IsRouter if new entry 1884 * RS: clear IsRouter 1885 * RA: set IsRouter if there's lladdr 1886 * redir: clear IsRouter if new entry 1887 * 1888 * RA case, (1): 1889 * The spec says that we must set IsRouter in the following cases: 1890 * - If lladdr exist, set IsRouter. This means (1-5). 1891 * - If it is old entry (!newentry), set IsRouter. This means (7). 1892 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1893 * A quetion arises for (1) case. (1) case has no lladdr in the 1894 * neighbor cache, this is similar to (6). 1895 * This case is rare but we figured that we MUST NOT set IsRouter. 1896 * 1897 * is_new old_addr new_addr NS RS RA redir 1898 * D R 1899 * 0 n n (1) c ? s 1900 * 0 y n (2) c s s 1901 * 0 n y (3) c s s 1902 * 0 y y (4) c s s 1903 * 0 y y (5) c s s 1904 * 1 -- n (6) c c c s 1905 * 1 -- y (7) c c s c s 1906 * 1907 * (c=clear s=set) 1908 */ 1909 switch (type & 0xff) { 1910 case ND_NEIGHBOR_SOLICIT: 1911 /* 1912 * New entry must have is_router flag cleared. 1913 */ 1914 if (is_new) /* (6-7) */ 1915 ln_router = 0; 1916 break; 1917 case ND_REDIRECT: 1918 /* 1919 * If the icmp is a redirect to a better router, always set the 1920 * is_router flag. Otherwise, if the entry is newly created, 1921 * clear the flag. [RFC 2461, sec 8.3] 1922 */ 1923 if (code == ND_REDIRECT_ROUTER) 1924 ln_router = 1; 1925 else { 1926 if (is_new) /* (6-7) */ 1927 ln_router = 0; 1928 } 1929 break; 1930 case ND_ROUTER_SOLICIT: 1931 /* 1932 * is_router flag must always be cleared. 1933 */ 1934 ln_router = 0; 1935 break; 1936 case ND_ROUTER_ADVERT: 1937 /* 1938 * Mark an entry with lladdr as a router. 1939 */ 1940 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ 1941 (is_new && new_addr)) { /* (7) */ 1942 ln_router = 1; 1943 } 1944 break; 1945 } 1946 1947 return (ln_router); 1948 } 1949 1950 /* 1951 * Create neighbor cache entry and cache link-layer address, 1952 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1953 * 1954 * type - ICMP6 type 1955 * code - type dependent information 1956 * 1957 */ 1958 void 1959 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1960 int lladdrlen, int type, int code) 1961 { 1962 struct llentry *ln = NULL, *ln_tmp; 1963 int is_newentry; 1964 int do_update; 1965 int olladdr; 1966 int llchange; 1967 int flags; 1968 uint16_t router = 0; 1969 struct sockaddr_in6 sin6; 1970 struct epoch_tracker et; 1971 struct mbuf *chain = NULL; 1972 u_char linkhdr[LLE_MAX_LINKHDR]; 1973 size_t linkhdrsize; 1974 int lladdr_off; 1975 1976 IF_AFDATA_UNLOCK_ASSERT(ifp); 1977 1978 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1979 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1980 1981 /* nothing must be updated for unspecified address */ 1982 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1983 return; 1984 1985 /* 1986 * Validation about ifp->if_addrlen and lladdrlen must be done in 1987 * the caller. 1988 * 1989 * XXX If the link does not have link-layer adderss, what should 1990 * we do? (ifp->if_addrlen == 0) 1991 * Spec says nothing in sections for RA, RS and NA. There's small 1992 * description on it in NS section (RFC 2461 7.2.3). 1993 */ 1994 flags = lladdr ? LLE_EXCLUSIVE : 0; 1995 NET_EPOCH_ENTER(et); 1996 ln = nd6_lookup(from, flags, ifp); 1997 NET_EPOCH_EXIT(et); 1998 is_newentry = 0; 1999 if (ln == NULL) { 2000 flags |= LLE_EXCLUSIVE; 2001 ln = nd6_alloc(from, 0, ifp); 2002 if (ln == NULL) 2003 return; 2004 2005 /* 2006 * Since we already know all the data for the new entry, 2007 * fill it before insertion. 2008 */ 2009 if (lladdr != NULL) { 2010 linkhdrsize = sizeof(linkhdr); 2011 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 2012 linkhdr, &linkhdrsize, &lladdr_off) != 0) 2013 return; 2014 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 2015 lladdr_off); 2016 } 2017 2018 IF_AFDATA_WLOCK(ifp); 2019 LLE_WLOCK(ln); 2020 /* Prefer any existing lle over newly-created one */ 2021 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); 2022 if (ln_tmp == NULL) 2023 lltable_link_entry(LLTABLE6(ifp), ln); 2024 IF_AFDATA_WUNLOCK(ifp); 2025 if (ln_tmp == NULL) { 2026 /* No existing lle, mark as new entry (6,7) */ 2027 is_newentry = 1; 2028 if (lladdr != NULL) { /* (7) */ 2029 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2030 EVENTHANDLER_INVOKE(lle_event, ln, 2031 LLENTRY_RESOLVED); 2032 } 2033 } else { 2034 lltable_free_entry(LLTABLE6(ifp), ln); 2035 ln = ln_tmp; 2036 ln_tmp = NULL; 2037 } 2038 } 2039 /* do nothing if static ndp is set */ 2040 if ((ln->la_flags & LLE_STATIC)) { 2041 if (flags & LLE_EXCLUSIVE) 2042 LLE_WUNLOCK(ln); 2043 else 2044 LLE_RUNLOCK(ln); 2045 return; 2046 } 2047 2048 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 2049 if (olladdr && lladdr) { 2050 llchange = bcmp(lladdr, ln->ll_addr, 2051 ifp->if_addrlen); 2052 } else if (!olladdr && lladdr) 2053 llchange = 1; 2054 else 2055 llchange = 0; 2056 2057 /* 2058 * newentry olladdr lladdr llchange (*=record) 2059 * 0 n n -- (1) 2060 * 0 y n -- (2) 2061 * 0 n y y (3) * STALE 2062 * 0 y y n (4) * 2063 * 0 y y y (5) * STALE 2064 * 1 -- n -- (6) NOSTATE(= PASSIVE) 2065 * 1 -- y -- (7) * STALE 2066 */ 2067 2068 do_update = 0; 2069 if (is_newentry == 0 && llchange != 0) { 2070 do_update = 1; /* (3,5) */ 2071 2072 /* 2073 * Record source link-layer address 2074 * XXX is it dependent to ifp->if_type? 2075 */ 2076 linkhdrsize = sizeof(linkhdr); 2077 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 2078 linkhdr, &linkhdrsize, &lladdr_off) != 0) 2079 return; 2080 2081 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 2082 lladdr_off) == 0) { 2083 /* Entry was deleted */ 2084 return; 2085 } 2086 2087 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2088 2089 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2090 2091 if (ln->la_hold != NULL) 2092 nd6_grab_holdchain(ln, &chain, &sin6); 2093 } 2094 2095 /* Calculates new router status */ 2096 router = nd6_is_router(type, code, is_newentry, olladdr, 2097 lladdr != NULL ? 1 : 0, ln->ln_router); 2098 2099 ln->ln_router = router; 2100 /* Mark non-router redirects with special flag */ 2101 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) 2102 ln->la_flags |= LLE_REDIRECT; 2103 2104 if (flags & LLE_EXCLUSIVE) 2105 LLE_WUNLOCK(ln); 2106 else 2107 LLE_RUNLOCK(ln); 2108 2109 if (chain != NULL) 2110 nd6_flush_holdchain(ifp, chain, &sin6); 2111 2112 /* 2113 * When the link-layer address of a router changes, select the 2114 * best router again. In particular, when the neighbor entry is newly 2115 * created, it might affect the selection policy. 2116 * Question: can we restrict the first condition to the "is_newentry" 2117 * case? 2118 * XXX: when we hear an RA from a new router with the link-layer 2119 * address option, defrouter_select_fib() is called twice, since 2120 * defrtrlist_update called the function as well. However, I believe 2121 * we can compromise the overhead, since it only happens the first 2122 * time. 2123 * XXX: although defrouter_select_fib() should not have a bad effect 2124 * for those are not autoconfigured hosts, we explicitly avoid such 2125 * cases for safety. 2126 */ 2127 if ((do_update || is_newentry) && router && 2128 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 2129 /* 2130 * guaranteed recursion 2131 */ 2132 defrouter_select_fib(ifp->if_fib); 2133 } 2134 } 2135 2136 static void 2137 nd6_slowtimo(void *arg) 2138 { 2139 struct epoch_tracker et; 2140 CURVNET_SET((struct vnet *) arg); 2141 struct nd_ifinfo *nd6if; 2142 struct ifnet *ifp; 2143 2144 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 2145 nd6_slowtimo, curvnet); 2146 NET_EPOCH_ENTER(et); 2147 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2148 if (ifp->if_afdata[AF_INET6] == NULL) 2149 continue; 2150 nd6if = ND_IFINFO(ifp); 2151 if (nd6if->basereachable && /* already initialized */ 2152 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 2153 /* 2154 * Since reachable time rarely changes by router 2155 * advertisements, we SHOULD insure that a new random 2156 * value gets recomputed at least once every few hours. 2157 * (RFC 2461, 6.3.4) 2158 */ 2159 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 2160 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 2161 } 2162 } 2163 NET_EPOCH_EXIT(et); 2164 CURVNET_RESTORE(); 2165 } 2166 2167 void 2168 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 2169 struct sockaddr_in6 *sin6) 2170 { 2171 2172 LLE_WLOCK_ASSERT(ln); 2173 2174 *chain = ln->la_hold; 2175 ln->la_hold = NULL; 2176 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 2177 2178 if (ln->ln_state == ND6_LLINFO_STALE) { 2179 2180 /* 2181 * The first time we send a packet to a 2182 * neighbor whose entry is STALE, we have 2183 * to change the state to DELAY and a sets 2184 * a timer to expire in DELAY_FIRST_PROBE_TIME 2185 * seconds to ensure do neighbor unreachability 2186 * detection on expiration. 2187 * (RFC 2461 7.3.3) 2188 */ 2189 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); 2190 } 2191 } 2192 2193 int 2194 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2195 struct sockaddr_in6 *dst, struct route *ro) 2196 { 2197 int error; 2198 int ip6len; 2199 struct ip6_hdr *ip6; 2200 struct m_tag *mtag; 2201 2202 #ifdef MAC 2203 mac_netinet6_nd6_send(ifp, m); 2204 #endif 2205 2206 /* 2207 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2208 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2209 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2210 * to be diverted to user space. When re-injected into the kernel, 2211 * send_output() will directly dispatch them to the outgoing interface. 2212 */ 2213 if (send_sendso_input_hook != NULL) { 2214 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2215 if (mtag != NULL) { 2216 ip6 = mtod(m, struct ip6_hdr *); 2217 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2218 /* Use the SEND socket */ 2219 error = send_sendso_input_hook(m, ifp, SND_OUT, 2220 ip6len); 2221 /* -1 == no app on SEND socket */ 2222 if (error == 0 || error != -1) 2223 return (error); 2224 } 2225 } 2226 2227 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 2228 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 2229 mtod(m, struct ip6_hdr *)); 2230 2231 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2232 origifp = ifp; 2233 2234 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); 2235 return (error); 2236 } 2237 2238 /* 2239 * Lookup link headerfor @sa_dst address. Stores found 2240 * data in @desten buffer. Copy of lle ln_flags can be also 2241 * saved in @pflags if @pflags is non-NULL. 2242 * 2243 * If destination LLE does not exists or lle state modification 2244 * is required, call "slow" version. 2245 * 2246 * Return values: 2247 * - 0 on success (address copied to buffer). 2248 * - EWOULDBLOCK (no local error, but address is still unresolved) 2249 * - other errors (alloc failure, etc) 2250 */ 2251 int 2252 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 2253 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags, 2254 struct llentry **plle) 2255 { 2256 struct epoch_tracker et; 2257 struct llentry *ln = NULL; 2258 const struct sockaddr_in6 *dst6; 2259 2260 if (pflags != NULL) 2261 *pflags = 0; 2262 2263 dst6 = (const struct sockaddr_in6 *)sa_dst; 2264 2265 /* discard the packet if IPv6 operation is disabled on the interface */ 2266 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2267 m_freem(m); 2268 return (ENETDOWN); /* better error? */ 2269 } 2270 2271 if (m != NULL && m->m_flags & M_MCAST) { 2272 switch (ifp->if_type) { 2273 case IFT_ETHER: 2274 case IFT_L2VLAN: 2275 case IFT_BRIDGE: 2276 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, 2277 desten); 2278 return (0); 2279 default: 2280 m_freem(m); 2281 return (EAFNOSUPPORT); 2282 } 2283 } 2284 2285 NET_EPOCH_ENTER(et); 2286 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, 2287 ifp); 2288 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { 2289 /* Entry found, let's copy lle info */ 2290 bcopy(ln->r_linkdata, desten, ln->r_hdrlen); 2291 if (pflags != NULL) 2292 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); 2293 /* Check if we have feedback request from nd6 timer */ 2294 if (ln->r_skip_req != 0) { 2295 LLE_REQ_LOCK(ln); 2296 ln->r_skip_req = 0; /* Notify that entry was used */ 2297 ln->lle_hittime = time_uptime; 2298 LLE_REQ_UNLOCK(ln); 2299 } 2300 if (plle) { 2301 LLE_ADDREF(ln); 2302 *plle = ln; 2303 LLE_WUNLOCK(ln); 2304 } 2305 NET_EPOCH_EXIT(et); 2306 return (0); 2307 } else if (plle && ln) 2308 LLE_WUNLOCK(ln); 2309 NET_EPOCH_EXIT(et); 2310 2311 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle)); 2312 } 2313 2314 2315 /* 2316 * Do L2 address resolution for @sa_dst address. Stores found 2317 * address in @desten buffer. Copy of lle ln_flags can be also 2318 * saved in @pflags if @pflags is non-NULL. 2319 * 2320 * Heavy version. 2321 * Function assume that destination LLE does not exist, 2322 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. 2323 * 2324 * Set noinline to be dtrace-friendly 2325 */ 2326 static __noinline int 2327 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m, 2328 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags, 2329 struct llentry **plle) 2330 { 2331 struct llentry *lle = NULL, *lle_tmp; 2332 struct in6_addr *psrc, src; 2333 int send_ns, ll_len; 2334 char *lladdr; 2335 2336 /* 2337 * Address resolution or Neighbor Unreachability Detection 2338 * for the next hop. 2339 * At this point, the destination of the packet must be a unicast 2340 * or an anycast address(i.e. not a multicast). 2341 */ 2342 if (lle == NULL) { 2343 struct epoch_tracker et; 2344 2345 NET_EPOCH_ENTER(et); 2346 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2347 NET_EPOCH_EXIT(et); 2348 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2349 /* 2350 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2351 * the condition below is not very efficient. But we believe 2352 * it is tolerable, because this should be a rare case. 2353 */ 2354 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2355 if (lle == NULL) { 2356 char ip6buf[INET6_ADDRSTRLEN]; 2357 log(LOG_DEBUG, 2358 "nd6_output: can't allocate llinfo for %s " 2359 "(ln=%p)\n", 2360 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2361 m_freem(m); 2362 return (ENOBUFS); 2363 } 2364 2365 IF_AFDATA_WLOCK(ifp); 2366 LLE_WLOCK(lle); 2367 /* Prefer any existing entry over newly-created one */ 2368 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2369 if (lle_tmp == NULL) 2370 lltable_link_entry(LLTABLE6(ifp), lle); 2371 IF_AFDATA_WUNLOCK(ifp); 2372 if (lle_tmp != NULL) { 2373 lltable_free_entry(LLTABLE6(ifp), lle); 2374 lle = lle_tmp; 2375 lle_tmp = NULL; 2376 } 2377 } 2378 } 2379 if (lle == NULL) { 2380 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2381 m_freem(m); 2382 return (ENOBUFS); 2383 } 2384 2385 if (m != NULL) 2386 m_freem(m); 2387 return (ENOBUFS); 2388 } 2389 2390 LLE_WLOCK_ASSERT(lle); 2391 2392 /* 2393 * The first time we send a packet to a neighbor whose entry is 2394 * STALE, we have to change the state to DELAY and a sets a timer to 2395 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2396 * neighbor unreachability detection on expiration. 2397 * (RFC 2461 7.3.3) 2398 */ 2399 if (lle->ln_state == ND6_LLINFO_STALE) 2400 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); 2401 2402 /* 2403 * If the neighbor cache entry has a state other than INCOMPLETE 2404 * (i.e. its link-layer address is already resolved), just 2405 * send the packet. 2406 */ 2407 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { 2408 if (flags & LLE_ADDRONLY) { 2409 lladdr = lle->ll_addr; 2410 ll_len = ifp->if_addrlen; 2411 } else { 2412 lladdr = lle->r_linkdata; 2413 ll_len = lle->r_hdrlen; 2414 } 2415 bcopy(lladdr, desten, ll_len); 2416 if (pflags != NULL) 2417 *pflags = lle->la_flags; 2418 if (plle) { 2419 LLE_ADDREF(lle); 2420 *plle = lle; 2421 } 2422 LLE_WUNLOCK(lle); 2423 return (0); 2424 } 2425 2426 /* 2427 * There is a neighbor cache entry, but no ethernet address 2428 * response yet. Append this latest packet to the end of the 2429 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen, 2430 * the oldest packet in the queue will be removed. 2431 */ 2432 2433 if (lle->la_hold != NULL) { 2434 struct mbuf *m_hold; 2435 int i; 2436 2437 i = 0; 2438 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2439 i++; 2440 if (m_hold->m_nextpkt == NULL) { 2441 m_hold->m_nextpkt = m; 2442 break; 2443 } 2444 } 2445 while (i >= V_nd6_maxqueuelen) { 2446 m_hold = lle->la_hold; 2447 lle->la_hold = lle->la_hold->m_nextpkt; 2448 m_freem(m_hold); 2449 i--; 2450 } 2451 } else { 2452 lle->la_hold = m; 2453 } 2454 2455 /* 2456 * If there has been no NS for the neighbor after entering the 2457 * INCOMPLETE state, send the first solicitation. 2458 * Note that for newly-created lle la_asked will be 0, 2459 * so we will transition from ND6_LLINFO_NOSTATE to 2460 * ND6_LLINFO_INCOMPLETE state here. 2461 */ 2462 psrc = NULL; 2463 send_ns = 0; 2464 if (lle->la_asked == 0) { 2465 lle->la_asked++; 2466 send_ns = 1; 2467 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2468 2469 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); 2470 } 2471 LLE_WUNLOCK(lle); 2472 if (send_ns != 0) 2473 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2474 2475 return (EWOULDBLOCK); 2476 } 2477 2478 /* 2479 * Do L2 address resolution for @sa_dst address. Stores found 2480 * address in @desten buffer. Copy of lle ln_flags can be also 2481 * saved in @pflags if @pflags is non-NULL. 2482 * 2483 * Return values: 2484 * - 0 on success (address copied to buffer). 2485 * - EWOULDBLOCK (no local error, but address is still unresolved) 2486 * - other errors (alloc failure, etc) 2487 */ 2488 int 2489 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 2490 char *desten, uint32_t *pflags) 2491 { 2492 int error; 2493 2494 flags |= LLE_ADDRONLY; 2495 error = nd6_resolve_slow(ifp, flags, NULL, 2496 (const struct sockaddr_in6 *)dst, desten, pflags, NULL); 2497 return (error); 2498 } 2499 2500 int 2501 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain, 2502 struct sockaddr_in6 *dst) 2503 { 2504 struct mbuf *m, *m_head; 2505 int error = 0; 2506 2507 m_head = chain; 2508 2509 while (m_head) { 2510 m = m_head; 2511 m_head = m_head->m_nextpkt; 2512 error = nd6_output_ifp(ifp, ifp, m, dst, NULL); 2513 } 2514 2515 /* 2516 * XXX 2517 * note that intermediate errors are blindly ignored 2518 */ 2519 return (error); 2520 } 2521 2522 static int 2523 nd6_need_cache(struct ifnet *ifp) 2524 { 2525 /* 2526 * XXX: we currently do not make neighbor cache on any interface 2527 * other than Ethernet and GIF. 2528 * 2529 * RFC2893 says: 2530 * - unidirectional tunnels needs no ND 2531 */ 2532 switch (ifp->if_type) { 2533 case IFT_ETHER: 2534 case IFT_IEEE1394: 2535 case IFT_L2VLAN: 2536 case IFT_INFINIBAND: 2537 case IFT_BRIDGE: 2538 case IFT_PROPVIRTUAL: 2539 return (1); 2540 default: 2541 return (0); 2542 } 2543 } 2544 2545 /* 2546 * Add pernament ND6 link-layer record for given 2547 * interface address. 2548 * 2549 * Very similar to IPv4 arp_ifinit(), but: 2550 * 1) IPv6 DAD is performed in different place 2551 * 2) It is called by IPv6 protocol stack in contrast to 2552 * arp_ifinit() which is typically called in SIOCSIFADDR 2553 * driver ioctl handler. 2554 * 2555 */ 2556 int 2557 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2558 { 2559 struct ifnet *ifp; 2560 struct llentry *ln, *ln_tmp; 2561 struct sockaddr *dst; 2562 2563 ifp = ia->ia_ifa.ifa_ifp; 2564 if (nd6_need_cache(ifp) == 0) 2565 return (0); 2566 2567 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; 2568 dst = (struct sockaddr *)&ia->ia_addr; 2569 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2570 if (ln == NULL) 2571 return (ENOBUFS); 2572 2573 IF_AFDATA_WLOCK(ifp); 2574 LLE_WLOCK(ln); 2575 /* Unlink any entry if exists */ 2576 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2577 if (ln_tmp != NULL) 2578 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2579 lltable_link_entry(LLTABLE6(ifp), ln); 2580 IF_AFDATA_WUNLOCK(ifp); 2581 2582 if (ln_tmp != NULL) 2583 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2584 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2585 2586 LLE_WUNLOCK(ln); 2587 if (ln_tmp != NULL) 2588 llentry_free(ln_tmp); 2589 2590 return (0); 2591 } 2592 2593 /* 2594 * Removes either all lle entries for given @ia, or lle 2595 * corresponding to @ia address. 2596 */ 2597 void 2598 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) 2599 { 2600 struct sockaddr_in6 mask, addr; 2601 struct sockaddr *saddr, *smask; 2602 struct ifnet *ifp; 2603 2604 ifp = ia->ia_ifa.ifa_ifp; 2605 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2606 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2607 saddr = (struct sockaddr *)&addr; 2608 smask = (struct sockaddr *)&mask; 2609 2610 if (all != 0) 2611 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); 2612 else 2613 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); 2614 } 2615 2616 static void 2617 clear_llinfo_pqueue(struct llentry *ln) 2618 { 2619 struct mbuf *m_hold, *m_hold_next; 2620 2621 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2622 m_hold_next = m_hold->m_nextpkt; 2623 m_freem(m_hold); 2624 } 2625 2626 ln->la_hold = NULL; 2627 } 2628 2629 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2630 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2631 2632 SYSCTL_DECL(_net_inet6_icmp6); 2633 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2634 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2635 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter", 2636 "NDP default router list"); 2637 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2638 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2639 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix", 2640 "NDP prefix list"); 2641 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2642 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2643 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2644 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2645 2646 static int 2647 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2648 { 2649 struct in6_defrouter d; 2650 struct nd_defrouter *dr; 2651 int error; 2652 2653 if (req->newptr != NULL) 2654 return (EPERM); 2655 2656 error = sysctl_wire_old_buffer(req, 0); 2657 if (error != 0) 2658 return (error); 2659 2660 bzero(&d, sizeof(d)); 2661 d.rtaddr.sin6_family = AF_INET6; 2662 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2663 2664 ND6_RLOCK(); 2665 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2666 d.rtaddr.sin6_addr = dr->rtaddr; 2667 error = sa6_recoverscope(&d.rtaddr); 2668 if (error != 0) 2669 break; 2670 d.flags = dr->raflags; 2671 d.rtlifetime = dr->rtlifetime; 2672 d.expire = dr->expire + (time_second - time_uptime); 2673 d.if_index = dr->ifp->if_index; 2674 error = SYSCTL_OUT(req, &d, sizeof(d)); 2675 if (error != 0) 2676 break; 2677 } 2678 ND6_RUNLOCK(); 2679 return (error); 2680 } 2681 2682 static int 2683 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2684 { 2685 struct in6_prefix p; 2686 struct sockaddr_in6 s6; 2687 struct nd_prefix *pr; 2688 struct nd_pfxrouter *pfr; 2689 time_t maxexpire; 2690 int error; 2691 char ip6buf[INET6_ADDRSTRLEN]; 2692 2693 if (req->newptr) 2694 return (EPERM); 2695 2696 error = sysctl_wire_old_buffer(req, 0); 2697 if (error != 0) 2698 return (error); 2699 2700 bzero(&p, sizeof(p)); 2701 p.origin = PR_ORIG_RA; 2702 bzero(&s6, sizeof(s6)); 2703 s6.sin6_family = AF_INET6; 2704 s6.sin6_len = sizeof(s6); 2705 2706 ND6_RLOCK(); 2707 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2708 p.prefix = pr->ndpr_prefix; 2709 if (sa6_recoverscope(&p.prefix)) { 2710 log(LOG_ERR, "scope error in prefix list (%s)\n", 2711 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2712 /* XXX: press on... */ 2713 } 2714 p.raflags = pr->ndpr_raf; 2715 p.prefixlen = pr->ndpr_plen; 2716 p.vltime = pr->ndpr_vltime; 2717 p.pltime = pr->ndpr_pltime; 2718 p.if_index = pr->ndpr_ifp->if_index; 2719 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2720 p.expire = 0; 2721 else { 2722 /* XXX: we assume time_t is signed. */ 2723 maxexpire = (-1) & 2724 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2725 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2726 p.expire = pr->ndpr_lastupdate + 2727 pr->ndpr_vltime + 2728 (time_second - time_uptime); 2729 else 2730 p.expire = maxexpire; 2731 } 2732 p.refcnt = pr->ndpr_addrcnt; 2733 p.flags = pr->ndpr_stateflags; 2734 p.advrtrs = 0; 2735 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2736 p.advrtrs++; 2737 error = SYSCTL_OUT(req, &p, sizeof(p)); 2738 if (error != 0) 2739 break; 2740 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2741 s6.sin6_addr = pfr->router->rtaddr; 2742 if (sa6_recoverscope(&s6)) 2743 log(LOG_ERR, 2744 "scope error in prefix list (%s)\n", 2745 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2746 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2747 if (error != 0) 2748 goto out; 2749 } 2750 } 2751 out: 2752 ND6_RUNLOCK(); 2753 return (error); 2754 } 2755