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 #include "opt_route.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/eventhandler.h> 44 #include <sys/callout.h> 45 #include <sys/lock.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/mutex.h> 49 #include <sys/socket.h> 50 #include <sys/sockio.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/protosw.h> 54 #include <sys/errno.h> 55 #include <sys/syslog.h> 56 #include <sys/rwlock.h> 57 #include <sys/queue.h> 58 #include <sys/sdt.h> 59 #include <sys/sysctl.h> 60 61 #include <net/if.h> 62 #include <net/if_var.h> 63 #include <net/if_dl.h> 64 #include <net/if_types.h> 65 #include <net/route.h> 66 #include <net/route/route_ctl.h> 67 #include <net/route/nhop.h> 68 #include <net/vnet.h> 69 70 #include <netinet/in.h> 71 #include <netinet/in_kdtrace.h> 72 #include <net/if_llatbl.h> 73 #include <netinet/if_ether.h> 74 #include <netinet6/in6_var.h> 75 #include <netinet/ip6.h> 76 #include <netinet6/ip6_var.h> 77 #include <netinet6/scope6_var.h> 78 #include <netinet6/nd6.h> 79 #include <netinet6/in6_ifattach.h> 80 #include <netinet/icmp6.h> 81 #include <netinet6/send.h> 82 83 #include <sys/limits.h> 84 85 #include <security/mac/mac_framework.h> 86 87 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 88 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 89 90 #define SIN6(s) ((const struct sockaddr_in6 *)(s)) 91 92 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery"); 93 94 /* timer values */ 95 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 96 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 97 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 98 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 99 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 100 * local traffic */ 101 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 102 * collection timer */ 103 104 /* preventing too many loops in ND option parsing */ 105 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 106 107 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 108 * layer hints */ 109 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 16; /* max pkts cached in unresolved 110 * ND entries */ 111 #define V_nd6_maxndopt VNET(nd6_maxndopt) 112 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 113 114 #ifdef ND6_DEBUG 115 VNET_DEFINE(int, nd6_debug) = 1; 116 #else 117 VNET_DEFINE(int, nd6_debug) = 0; 118 #endif 119 120 static eventhandler_tag lle_event_eh, iflladdr_event_eh, ifnet_link_event_eh; 121 122 VNET_DEFINE(struct nd_prhead, nd_prefix); 123 VNET_DEFINE(struct rwlock, nd6_lock); 124 VNET_DEFINE(uint64_t, nd6_list_genid); 125 VNET_DEFINE(struct mtx, nd6_onlink_mtx); 126 127 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 128 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 129 130 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 131 132 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *, 133 struct ifnet *); 134 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 135 static void nd6_slowtimo(void *); 136 static int regen_tmpaddr(struct in6_ifaddr *); 137 static void nd6_free(struct llentry **, int); 138 static void nd6_free_redirect(const struct llentry *); 139 static void nd6_llinfo_timer(void *); 140 static void nd6_llinfo_settimer_locked(struct llentry *, long); 141 static void clear_llinfo_pqueue(struct llentry *); 142 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *, 143 const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **); 144 static int nd6_need_cache(struct ifnet *); 145 146 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch); 147 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 148 149 VNET_DEFINE_STATIC(struct callout, nd6_timer_ch); 150 #define V_nd6_timer_ch VNET(nd6_timer_ch) 151 152 SYSCTL_DECL(_net_inet6_icmp6); 153 154 static void 155 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 156 { 157 struct rt_addrinfo rtinfo; 158 struct sockaddr_in6 dst; 159 struct sockaddr_dl gw; 160 struct ifnet *ifp; 161 int type; 162 int fibnum; 163 164 LLE_WLOCK_ASSERT(lle); 165 166 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 167 return; 168 169 switch (evt) { 170 case LLENTRY_RESOLVED: 171 type = RTM_ADD; 172 KASSERT(lle->la_flags & LLE_VALID, 173 ("%s: %p resolved but not valid?", __func__, lle)); 174 break; 175 case LLENTRY_EXPIRED: 176 type = RTM_DELETE; 177 break; 178 default: 179 return; 180 } 181 182 ifp = lltable_get_ifp(lle->lle_tbl); 183 184 bzero(&dst, sizeof(dst)); 185 bzero(&gw, sizeof(gw)); 186 bzero(&rtinfo, sizeof(rtinfo)); 187 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 188 dst.sin6_scope_id = in6_getscopezone(ifp, 189 in6_addrscope(&dst.sin6_addr)); 190 gw.sdl_len = sizeof(struct sockaddr_dl); 191 gw.sdl_family = AF_LINK; 192 gw.sdl_alen = ifp->if_addrlen; 193 gw.sdl_index = ifp->if_index; 194 gw.sdl_type = ifp->if_type; 195 if (evt == LLENTRY_RESOLVED) 196 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 197 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 198 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 199 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 200 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib; 201 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 202 type == RTM_ADD ? RTF_UP: 0), 0, fibnum); 203 } 204 205 /* 206 * A handler for interface link layer address change event. 207 */ 208 static void 209 nd6_iflladdr(void *arg __unused, struct ifnet *ifp) 210 { 211 if (ifp->if_afdata[AF_INET6] == NULL) 212 return; 213 214 lltable_update_ifaddr(LLTABLE6(ifp)); 215 } 216 217 void 218 nd6_init(void) 219 { 220 221 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF); 222 rw_init(&V_nd6_lock, "nd6 list"); 223 224 LIST_INIT(&V_nd_prefix); 225 nd6_defrouter_init(); 226 227 /* Start timers. */ 228 callout_init(&V_nd6_slowtimo_ch, 0); 229 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 230 nd6_slowtimo, curvnet); 231 232 callout_init(&V_nd6_timer_ch, 0); 233 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 234 235 nd6_dad_init(); 236 if (IS_DEFAULT_VNET(curvnet)) { 237 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 238 NULL, EVENTHANDLER_PRI_ANY); 239 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event, 240 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 241 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event, 242 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY); 243 } 244 } 245 246 #ifdef VIMAGE 247 void 248 nd6_destroy() 249 { 250 251 callout_drain(&V_nd6_slowtimo_ch); 252 callout_drain(&V_nd6_timer_ch); 253 if (IS_DEFAULT_VNET(curvnet)) { 254 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh); 255 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 256 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh); 257 } 258 rw_destroy(&V_nd6_lock); 259 mtx_destroy(&V_nd6_onlink_mtx); 260 } 261 #endif 262 263 struct nd_ifinfo * 264 nd6_ifattach(struct ifnet *ifp) 265 { 266 struct nd_ifinfo *nd; 267 268 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 269 nd->initialized = 1; 270 271 nd->chlim = IPV6_DEFHLIM; 272 nd->basereachable = REACHABLE_TIME; 273 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 274 nd->retrans = RETRANS_TIMER; 275 276 nd->flags = ND6_IFF_PERFORMNUD; 277 278 /* Set IPv6 disabled on all interfaces but loopback by default. */ 279 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 280 nd->flags |= ND6_IFF_IFDISABLED; 281 282 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 283 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 284 * default regardless of the V_ip6_auto_linklocal configuration to 285 * give a reasonable default behavior. 286 */ 287 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 288 (ifp->if_flags & IFF_LOOPBACK)) 289 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 290 /* 291 * A loopback interface does not need to accept RTADV. 292 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 293 * default regardless of the V_ip6_accept_rtadv configuration to 294 * prevent the interface from accepting RA messages arrived 295 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 296 */ 297 if (V_ip6_accept_rtadv && 298 !(ifp->if_flags & IFF_LOOPBACK) && 299 (ifp->if_type != IFT_BRIDGE)) { 300 nd->flags |= ND6_IFF_ACCEPT_RTADV; 301 /* If we globally accept rtadv, assume IPv6 on. */ 302 nd->flags &= ~ND6_IFF_IFDISABLED; 303 } 304 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 305 nd->flags |= ND6_IFF_NO_RADR; 306 307 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 308 nd6_setmtu0(ifp, nd); 309 310 return nd; 311 } 312 313 void 314 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd) 315 { 316 struct epoch_tracker et; 317 struct ifaddr *ifa, *next; 318 319 NET_EPOCH_ENTER(et); 320 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 321 if (ifa->ifa_addr->sa_family != AF_INET6) 322 continue; 323 324 /* stop DAD processing */ 325 nd6_dad_stop(ifa); 326 } 327 NET_EPOCH_EXIT(et); 328 329 free(nd, M_IP6NDP); 330 } 331 332 /* 333 * Reset ND level link MTU. This function is called when the physical MTU 334 * changes, which means we might have to adjust the ND level MTU. 335 */ 336 void 337 nd6_setmtu(struct ifnet *ifp) 338 { 339 if (ifp->if_afdata[AF_INET6] == NULL) 340 return; 341 342 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 343 } 344 345 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 346 void 347 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 348 { 349 u_int32_t omaxmtu; 350 351 omaxmtu = ndi->maxmtu; 352 ndi->maxmtu = ifp->if_mtu; 353 354 /* 355 * Decreasing the interface MTU under IPV6 minimum MTU may cause 356 * undesirable situation. We thus notify the operator of the change 357 * explicitly. The check for omaxmtu is necessary to restrict the 358 * log to the case of changing the MTU, not initializing it. 359 */ 360 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 361 log(LOG_NOTICE, "nd6_setmtu0: " 362 "new link MTU on %s (%lu) is too small for IPv6\n", 363 if_name(ifp), (unsigned long)ndi->maxmtu); 364 } 365 366 if (ndi->maxmtu > V_in6_maxmtu) 367 in6_setmaxmtu(); /* check all interfaces just in case */ 368 369 } 370 371 void 372 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 373 { 374 375 bzero(ndopts, sizeof(*ndopts)); 376 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 377 ndopts->nd_opts_last 378 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 379 380 if (icmp6len == 0) { 381 ndopts->nd_opts_done = 1; 382 ndopts->nd_opts_search = NULL; 383 } 384 } 385 386 /* 387 * Take one ND option. 388 */ 389 struct nd_opt_hdr * 390 nd6_option(union nd_opts *ndopts) 391 { 392 struct nd_opt_hdr *nd_opt; 393 int olen; 394 395 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 396 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 397 __func__)); 398 if (ndopts->nd_opts_search == NULL) 399 return NULL; 400 if (ndopts->nd_opts_done) 401 return NULL; 402 403 nd_opt = ndopts->nd_opts_search; 404 405 /* make sure nd_opt_len is inside the buffer */ 406 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 407 bzero(ndopts, sizeof(*ndopts)); 408 return NULL; 409 } 410 411 olen = nd_opt->nd_opt_len << 3; 412 if (olen == 0) { 413 /* 414 * Message validation requires that all included 415 * options have a length that is greater than zero. 416 */ 417 bzero(ndopts, sizeof(*ndopts)); 418 return NULL; 419 } 420 421 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 422 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 423 /* option overruns the end of buffer, invalid */ 424 bzero(ndopts, sizeof(*ndopts)); 425 return NULL; 426 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 427 /* reached the end of options chain */ 428 ndopts->nd_opts_done = 1; 429 ndopts->nd_opts_search = NULL; 430 } 431 return nd_opt; 432 } 433 434 /* 435 * Parse multiple ND options. 436 * This function is much easier to use, for ND routines that do not need 437 * multiple options of the same type. 438 */ 439 int 440 nd6_options(union nd_opts *ndopts) 441 { 442 struct nd_opt_hdr *nd_opt; 443 int i = 0; 444 445 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 446 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 447 __func__)); 448 if (ndopts->nd_opts_search == NULL) 449 return 0; 450 451 while (1) { 452 nd_opt = nd6_option(ndopts); 453 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 454 /* 455 * Message validation requires that all included 456 * options have a length that is greater than zero. 457 */ 458 ICMP6STAT_INC(icp6s_nd_badopt); 459 bzero(ndopts, sizeof(*ndopts)); 460 return -1; 461 } 462 463 if (nd_opt == NULL) 464 goto skip1; 465 466 switch (nd_opt->nd_opt_type) { 467 case ND_OPT_SOURCE_LINKADDR: 468 case ND_OPT_TARGET_LINKADDR: 469 case ND_OPT_MTU: 470 case ND_OPT_REDIRECTED_HEADER: 471 case ND_OPT_NONCE: 472 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 473 nd6log((LOG_INFO, 474 "duplicated ND6 option found (type=%d)\n", 475 nd_opt->nd_opt_type)); 476 /* XXX bark? */ 477 } else { 478 ndopts->nd_opt_array[nd_opt->nd_opt_type] 479 = nd_opt; 480 } 481 break; 482 case ND_OPT_PREFIX_INFORMATION: 483 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 484 ndopts->nd_opt_array[nd_opt->nd_opt_type] 485 = nd_opt; 486 } 487 ndopts->nd_opts_pi_end = 488 (struct nd_opt_prefix_info *)nd_opt; 489 break; 490 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 491 case ND_OPT_RDNSS: /* RFC 6106 */ 492 case ND_OPT_DNSSL: /* RFC 6106 */ 493 /* 494 * Silently ignore options we know and do not care about 495 * in the kernel. 496 */ 497 break; 498 default: 499 /* 500 * Unknown options must be silently ignored, 501 * to accommodate future extension to the protocol. 502 */ 503 nd6log((LOG_DEBUG, 504 "nd6_options: unsupported option %d - " 505 "option ignored\n", nd_opt->nd_opt_type)); 506 } 507 508 skip1: 509 i++; 510 if (i > V_nd6_maxndopt) { 511 ICMP6STAT_INC(icp6s_nd_toomanyopt); 512 nd6log((LOG_INFO, "too many loop in nd opt\n")); 513 break; 514 } 515 516 if (ndopts->nd_opts_done) 517 break; 518 } 519 520 return 0; 521 } 522 523 /* 524 * ND6 timer routine to handle ND6 entries 525 */ 526 static void 527 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 528 { 529 int canceled; 530 531 LLE_WLOCK_ASSERT(ln); 532 533 if (tick < 0) { 534 ln->la_expire = 0; 535 ln->ln_ntick = 0; 536 canceled = callout_stop(&ln->lle_timer); 537 } else { 538 ln->la_expire = time_uptime + tick / hz; 539 LLE_ADDREF(ln); 540 if (tick > INT_MAX) { 541 ln->ln_ntick = tick - INT_MAX; 542 canceled = callout_reset(&ln->lle_timer, INT_MAX, 543 nd6_llinfo_timer, ln); 544 } else { 545 ln->ln_ntick = 0; 546 canceled = callout_reset(&ln->lle_timer, tick, 547 nd6_llinfo_timer, ln); 548 } 549 } 550 if (canceled > 0) 551 LLE_REMREF(ln); 552 } 553 554 /* 555 * Gets source address of the first packet in hold queue 556 * and stores it in @src. 557 * Returns pointer to @src (if hold queue is not empty) or NULL. 558 * 559 * Set noinline to be dtrace-friendly 560 */ 561 static __noinline struct in6_addr * 562 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 563 { 564 struct ip6_hdr hdr; 565 struct mbuf *m; 566 567 if (ln->la_hold == NULL) 568 return (NULL); 569 570 /* 571 * assume every packet in la_hold has the same IP header 572 */ 573 m = ln->la_hold; 574 if (sizeof(hdr) > m->m_len) 575 return (NULL); 576 577 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 578 *src = hdr.ip6_src; 579 580 return (src); 581 } 582 583 /* 584 * Checks if we need to switch from STALE state. 585 * 586 * RFC 4861 requires switching from STALE to DELAY state 587 * on first packet matching entry, waiting V_nd6_delay and 588 * transition to PROBE state (if upper layer confirmation was 589 * not received). 590 * 591 * This code performs a bit differently: 592 * On packet hit we don't change state (but desired state 593 * can be guessed by control plane). However, after V_nd6_delay 594 * seconds code will transition to PROBE state (so DELAY state 595 * is kinda skipped in most situations). 596 * 597 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so 598 * we perform the following upon entering STALE state: 599 * 600 * 1) Arm timer to run each V_nd6_delay seconds to make sure that 601 * if packet was transmitted at the start of given interval, we 602 * would be able to switch to PROBE state in V_nd6_delay seconds 603 * as user expects. 604 * 605 * 2) Reschedule timer until original V_nd6_gctimer expires keeping 606 * lle in STALE state (remaining timer value stored in lle_remtime). 607 * 608 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay 609 * seconds ago. 610 * 611 * Returns non-zero value if the entry is still STALE (storing 612 * the next timer interval in @pdelay). 613 * 614 * Returns zero value if original timer expired or we need to switch to 615 * PROBE (store that in @do_switch variable). 616 */ 617 static int 618 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) 619 { 620 int nd_delay, nd_gctimer; 621 time_t lle_hittime; 622 long delay; 623 624 *do_switch = 0; 625 nd_gctimer = V_nd6_gctimer; 626 nd_delay = V_nd6_delay; 627 628 lle_hittime = llentry_get_hittime(lle); 629 630 if (lle_hittime == 0) { 631 /* 632 * Datapath feedback has been requested upon entering 633 * STALE state. No packets has been passed using this lle. 634 * Ask for the timer reschedule and keep STALE state. 635 */ 636 delay = (long)(MIN(nd_gctimer, nd_delay)); 637 delay *= hz; 638 if (lle->lle_remtime > delay) 639 lle->lle_remtime -= delay; 640 else { 641 delay = lle->lle_remtime; 642 lle->lle_remtime = 0; 643 } 644 645 if (delay == 0) { 646 /* 647 * The original ng6_gctime timeout ended, 648 * no more rescheduling. 649 */ 650 return (0); 651 } 652 653 *pdelay = delay; 654 return (1); 655 } 656 657 /* 658 * Packet received. Verify timestamp 659 */ 660 delay = (long)(time_uptime - lle_hittime); 661 if (delay < nd_delay) { 662 /* 663 * V_nd6_delay still not passed since the first 664 * hit in STALE state. 665 * Reshedule timer and return. 666 */ 667 *pdelay = (long)(nd_delay - delay) * hz; 668 return (1); 669 } 670 671 /* Request switching to probe */ 672 *do_switch = 1; 673 return (0); 674 } 675 676 /* 677 * Switch @lle state to new state optionally arming timers. 678 * 679 * Set noinline to be dtrace-friendly 680 */ 681 __noinline void 682 nd6_llinfo_setstate(struct llentry *lle, int newstate) 683 { 684 struct ifnet *ifp; 685 int nd_gctimer, nd_delay; 686 long delay, remtime; 687 688 delay = 0; 689 remtime = 0; 690 691 switch (newstate) { 692 case ND6_LLINFO_INCOMPLETE: 693 ifp = lle->lle_tbl->llt_ifp; 694 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; 695 break; 696 case ND6_LLINFO_REACHABLE: 697 if (!ND6_LLINFO_PERMANENT(lle)) { 698 ifp = lle->lle_tbl->llt_ifp; 699 delay = (long)ND_IFINFO(ifp)->reachable * hz; 700 } 701 break; 702 case ND6_LLINFO_STALE: 703 704 llentry_request_feedback(lle); 705 nd_delay = V_nd6_delay; 706 nd_gctimer = V_nd6_gctimer; 707 708 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz; 709 remtime = (long)nd_gctimer * hz - delay; 710 break; 711 case ND6_LLINFO_DELAY: 712 lle->la_asked = 0; 713 delay = (long)V_nd6_delay * hz; 714 break; 715 } 716 717 if (delay > 0) 718 nd6_llinfo_settimer_locked(lle, delay); 719 720 lle->lle_remtime = remtime; 721 lle->ln_state = newstate; 722 } 723 724 /* 725 * Timer-dependent part of nd state machine. 726 * 727 * Set noinline to be dtrace-friendly 728 */ 729 static __noinline void 730 nd6_llinfo_timer(void *arg) 731 { 732 struct epoch_tracker et; 733 struct llentry *ln; 734 struct in6_addr *dst, *pdst, *psrc, src; 735 struct ifnet *ifp; 736 struct nd_ifinfo *ndi; 737 int do_switch, send_ns; 738 long delay; 739 740 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 741 ln = (struct llentry *)arg; 742 ifp = lltable_get_ifp(ln->lle_tbl); 743 CURVNET_SET(ifp->if_vnet); 744 745 ND6_RLOCK(); 746 LLE_WLOCK(ln); 747 if (callout_pending(&ln->lle_timer)) { 748 /* 749 * Here we are a bit odd here in the treatment of 750 * active/pending. If the pending bit is set, it got 751 * rescheduled before I ran. The active 752 * bit we ignore, since if it was stopped 753 * in ll_tablefree() and was currently running 754 * it would have return 0 so the code would 755 * not have deleted it since the callout could 756 * not be stopped so we want to go through 757 * with the delete here now. If the callout 758 * was restarted, the pending bit will be back on and 759 * we just want to bail since the callout_reset would 760 * return 1 and our reference would have been removed 761 * by nd6_llinfo_settimer_locked above since canceled 762 * would have been 1. 763 */ 764 LLE_WUNLOCK(ln); 765 ND6_RUNLOCK(); 766 CURVNET_RESTORE(); 767 return; 768 } 769 NET_EPOCH_ENTER(et); 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 struct mbuf *n = m; 819 820 /* 821 * if there are any ummapped mbufs, we 822 * must free them, rather than using 823 * them for an ICMP, as they cannot be 824 * checksummed. 825 */ 826 while ((n = n->m_next) != NULL) { 827 if (n->m_flags & M_EXTPG) 828 break; 829 } 830 if (n != NULL) { 831 m_freem(m); 832 m = NULL; 833 } else { 834 icmp6_error2(m, ICMP6_DST_UNREACH, 835 ICMP6_DST_UNREACH_ADDR, 0, ifp); 836 } 837 } 838 } 839 break; 840 case ND6_LLINFO_REACHABLE: 841 if (!ND6_LLINFO_PERMANENT(ln)) 842 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 843 break; 844 845 case ND6_LLINFO_STALE: 846 if (nd6_is_stale(ln, &delay, &do_switch) != 0) { 847 /* 848 * No packet has used this entry and GC timeout 849 * has not been passed. Reshedule timer and 850 * return. 851 */ 852 nd6_llinfo_settimer_locked(ln, delay); 853 break; 854 } 855 856 if (do_switch == 0) { 857 /* 858 * GC timer has ended and entry hasn't been used. 859 * Run Garbage collector (RFC 4861, 5.3) 860 */ 861 if (!ND6_LLINFO_PERMANENT(ln)) 862 nd6_free(&ln, 1); 863 break; 864 } 865 866 /* Entry has been used AND delay timer has ended. */ 867 868 /* FALLTHROUGH */ 869 870 case ND6_LLINFO_DELAY: 871 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 872 /* We need NUD */ 873 ln->la_asked = 1; 874 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE); 875 send_ns = 1; 876 } else 877 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */ 878 break; 879 case ND6_LLINFO_PROBE: 880 if (ln->la_asked < V_nd6_umaxtries) { 881 ln->la_asked++; 882 send_ns = 1; 883 } else { 884 nd6_free(&ln, 0); 885 } 886 break; 887 default: 888 panic("%s: paths in a dark night can be confusing: %d", 889 __func__, ln->ln_state); 890 } 891 done: 892 if (ln != NULL) 893 ND6_RUNLOCK(); 894 if (send_ns != 0) { 895 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 896 psrc = nd6_llinfo_get_holdsrc(ln, &src); 897 LLE_FREE_LOCKED(ln); 898 ln = NULL; 899 nd6_ns_output(ifp, psrc, pdst, dst, NULL); 900 } 901 902 if (ln != NULL) 903 LLE_FREE_LOCKED(ln); 904 NET_EPOCH_EXIT(et); 905 CURVNET_RESTORE(); 906 } 907 908 /* 909 * ND6 timer routine to expire default route list and prefix list 910 */ 911 void 912 nd6_timer(void *arg) 913 { 914 CURVNET_SET((struct vnet *) arg); 915 struct epoch_tracker et; 916 struct nd_prhead prl; 917 struct nd_prefix *pr, *npr; 918 struct ifnet *ifp; 919 struct in6_ifaddr *ia6, *nia6; 920 uint64_t genid; 921 922 LIST_INIT(&prl); 923 924 NET_EPOCH_ENTER(et); 925 nd6_defrouter_timer(); 926 927 /* 928 * expire interface addresses. 929 * in the past the loop was inside prefix expiry processing. 930 * However, from a stricter speci-confrmance standpoint, we should 931 * rather separate address lifetimes and prefix lifetimes. 932 * 933 * XXXRW: in6_ifaddrhead locking. 934 */ 935 addrloop: 936 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 937 /* check address lifetime */ 938 if (IFA6_IS_INVALID(ia6)) { 939 int regen = 0; 940 941 /* 942 * If the expiring address is temporary, try 943 * regenerating a new one. This would be useful when 944 * we suspended a laptop PC, then turned it on after a 945 * period that could invalidate all temporary 946 * addresses. Although we may have to restart the 947 * loop (see below), it must be after purging the 948 * address. Otherwise, we'd see an infinite loop of 949 * regeneration. 950 */ 951 if (V_ip6_use_tempaddr && 952 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 953 if (regen_tmpaddr(ia6) == 0) 954 regen = 1; 955 } 956 957 in6_purgeaddr(&ia6->ia_ifa); 958 959 if (regen) 960 goto addrloop; /* XXX: see below */ 961 } else if (IFA6_IS_DEPRECATED(ia6)) { 962 int oldflags = ia6->ia6_flags; 963 964 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 965 966 /* 967 * If a temporary address has just become deprecated, 968 * regenerate a new one if possible. 969 */ 970 if (V_ip6_use_tempaddr && 971 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 972 (oldflags & IN6_IFF_DEPRECATED) == 0) { 973 if (regen_tmpaddr(ia6) == 0) { 974 /* 975 * A new temporary address is 976 * generated. 977 * XXX: this means the address chain 978 * has changed while we are still in 979 * the loop. Although the change 980 * would not cause disaster (because 981 * it's not a deletion, but an 982 * addition,) we'd rather restart the 983 * loop just for safety. Or does this 984 * significantly reduce performance?? 985 */ 986 goto addrloop; 987 } 988 } 989 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 990 /* 991 * Schedule DAD for a tentative address. This happens 992 * if the interface was down or not running 993 * when the address was configured. 994 */ 995 int delay; 996 997 delay = arc4random() % 998 (MAX_RTR_SOLICITATION_DELAY * hz); 999 nd6_dad_start((struct ifaddr *)ia6, delay); 1000 } else { 1001 /* 1002 * Check status of the interface. If it is down, 1003 * mark the address as tentative for future DAD. 1004 */ 1005 ifp = ia6->ia_ifp; 1006 if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 && 1007 ((ifp->if_flags & IFF_UP) == 0 || 1008 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1009 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){ 1010 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 1011 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 1012 } 1013 1014 /* 1015 * A new RA might have made a deprecated address 1016 * preferred. 1017 */ 1018 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 1019 } 1020 } 1021 NET_EPOCH_EXIT(et); 1022 1023 ND6_WLOCK(); 1024 restart: 1025 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1026 /* 1027 * Expire prefixes. Since the pltime is only used for 1028 * autoconfigured addresses, pltime processing for prefixes is 1029 * not necessary. 1030 * 1031 * Only unlink after all derived addresses have expired. This 1032 * may not occur until two hours after the prefix has expired 1033 * per RFC 4862. If the prefix expires before its derived 1034 * addresses, mark it off-link. This will be done automatically 1035 * after unlinking if no address references remain. 1036 */ 1037 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME || 1038 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime) 1039 continue; 1040 1041 if (pr->ndpr_addrcnt == 0) { 1042 nd6_prefix_unlink(pr, &prl); 1043 continue; 1044 } 1045 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1046 genid = V_nd6_list_genid; 1047 nd6_prefix_ref(pr); 1048 ND6_WUNLOCK(); 1049 ND6_ONLINK_LOCK(); 1050 (void)nd6_prefix_offlink(pr); 1051 ND6_ONLINK_UNLOCK(); 1052 ND6_WLOCK(); 1053 nd6_prefix_rele(pr); 1054 if (genid != V_nd6_list_genid) 1055 goto restart; 1056 } 1057 } 1058 ND6_WUNLOCK(); 1059 1060 while ((pr = LIST_FIRST(&prl)) != NULL) { 1061 LIST_REMOVE(pr, ndpr_entry); 1062 nd6_prefix_del(pr); 1063 } 1064 1065 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 1066 nd6_timer, curvnet); 1067 1068 CURVNET_RESTORE(); 1069 } 1070 1071 /* 1072 * ia6 - deprecated/invalidated temporary address 1073 */ 1074 static int 1075 regen_tmpaddr(struct in6_ifaddr *ia6) 1076 { 1077 struct ifaddr *ifa; 1078 struct ifnet *ifp; 1079 struct in6_ifaddr *public_ifa6 = NULL; 1080 1081 NET_EPOCH_ASSERT(); 1082 1083 ifp = ia6->ia_ifa.ifa_ifp; 1084 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1085 struct in6_ifaddr *it6; 1086 1087 if (ifa->ifa_addr->sa_family != AF_INET6) 1088 continue; 1089 1090 it6 = (struct in6_ifaddr *)ifa; 1091 1092 /* ignore no autoconf addresses. */ 1093 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1094 continue; 1095 1096 /* ignore autoconf addresses with different prefixes. */ 1097 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 1098 continue; 1099 1100 /* 1101 * Now we are looking at an autoconf address with the same 1102 * prefix as ours. If the address is temporary and is still 1103 * preferred, do not create another one. It would be rare, but 1104 * could happen, for example, when we resume a laptop PC after 1105 * a long period. 1106 */ 1107 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1108 !IFA6_IS_DEPRECATED(it6)) { 1109 public_ifa6 = NULL; 1110 break; 1111 } 1112 1113 /* 1114 * This is a public autoconf address that has the same prefix 1115 * as ours. If it is preferred, keep it. We can't break the 1116 * loop here, because there may be a still-preferred temporary 1117 * address with the prefix. 1118 */ 1119 if (!IFA6_IS_DEPRECATED(it6)) 1120 public_ifa6 = it6; 1121 } 1122 if (public_ifa6 != NULL) 1123 ifa_ref(&public_ifa6->ia_ifa); 1124 1125 if (public_ifa6 != NULL) { 1126 int e; 1127 1128 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 1129 ifa_free(&public_ifa6->ia_ifa); 1130 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 1131 " tmp addr,errno=%d\n", e); 1132 return (-1); 1133 } 1134 ifa_free(&public_ifa6->ia_ifa); 1135 return (0); 1136 } 1137 1138 return (-1); 1139 } 1140 1141 /* 1142 * Remove prefix and default router list entries corresponding to ifp. Neighbor 1143 * cache entries are freed in in6_domifdetach(). 1144 */ 1145 void 1146 nd6_purge(struct ifnet *ifp) 1147 { 1148 struct nd_prhead prl; 1149 struct nd_prefix *pr, *npr; 1150 1151 LIST_INIT(&prl); 1152 1153 /* Purge default router list entries toward ifp. */ 1154 nd6_defrouter_purge(ifp); 1155 1156 ND6_WLOCK(); 1157 /* 1158 * Remove prefixes on ifp. We should have already removed addresses on 1159 * this interface, so no addresses should be referencing these prefixes. 1160 */ 1161 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1162 if (pr->ndpr_ifp == ifp) 1163 nd6_prefix_unlink(pr, &prl); 1164 } 1165 ND6_WUNLOCK(); 1166 1167 /* Delete the unlinked prefix objects. */ 1168 while ((pr = LIST_FIRST(&prl)) != NULL) { 1169 LIST_REMOVE(pr, ndpr_entry); 1170 nd6_prefix_del(pr); 1171 } 1172 1173 /* cancel default outgoing interface setting */ 1174 if (V_nd6_defifindex == ifp->if_index) 1175 nd6_setdefaultiface(0); 1176 1177 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1178 /* Refresh default router list. */ 1179 defrouter_select_fib(ifp->if_fib); 1180 } 1181 } 1182 1183 /* 1184 * the caller acquires and releases the lock on the lltbls 1185 * Returns the llentry locked 1186 */ 1187 struct llentry * 1188 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1189 { 1190 struct sockaddr_in6 sin6; 1191 struct llentry *ln; 1192 1193 bzero(&sin6, sizeof(sin6)); 1194 sin6.sin6_len = sizeof(struct sockaddr_in6); 1195 sin6.sin6_family = AF_INET6; 1196 sin6.sin6_addr = *addr6; 1197 1198 IF_AFDATA_LOCK_ASSERT(ifp); 1199 1200 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6); 1201 1202 return (ln); 1203 } 1204 1205 static struct llentry * 1206 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1207 { 1208 struct sockaddr_in6 sin6; 1209 struct llentry *ln; 1210 1211 bzero(&sin6, sizeof(sin6)); 1212 sin6.sin6_len = sizeof(struct sockaddr_in6); 1213 sin6.sin6_family = AF_INET6; 1214 sin6.sin6_addr = *addr6; 1215 1216 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 1217 if (ln != NULL) 1218 ln->ln_state = ND6_LLINFO_NOSTATE; 1219 1220 return (ln); 1221 } 1222 1223 /* 1224 * Test whether a given IPv6 address is a neighbor or not, ignoring 1225 * the actual neighbor cache. The neighbor cache is ignored in order 1226 * to not reenter the routing code from within itself. 1227 */ 1228 static int 1229 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1230 { 1231 struct nd_prefix *pr; 1232 struct ifaddr *ifa; 1233 struct rt_addrinfo info; 1234 struct sockaddr_in6 rt_key; 1235 const struct sockaddr *dst6; 1236 uint64_t genid; 1237 int error, fibnum; 1238 1239 /* 1240 * A link-local address is always a neighbor. 1241 * XXX: a link does not necessarily specify a single interface. 1242 */ 1243 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 1244 struct sockaddr_in6 sin6_copy; 1245 u_int32_t zone; 1246 1247 /* 1248 * We need sin6_copy since sa6_recoverscope() may modify the 1249 * content (XXX). 1250 */ 1251 sin6_copy = *addr; 1252 if (sa6_recoverscope(&sin6_copy)) 1253 return (0); /* XXX: should be impossible */ 1254 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 1255 return (0); 1256 if (sin6_copy.sin6_scope_id == zone) 1257 return (1); 1258 else 1259 return (0); 1260 } 1261 1262 bzero(&rt_key, sizeof(rt_key)); 1263 bzero(&info, sizeof(info)); 1264 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1265 1266 /* 1267 * If the address matches one of our addresses, 1268 * it should be a neighbor. 1269 * If the address matches one of our on-link prefixes, it should be a 1270 * neighbor. 1271 */ 1272 ND6_RLOCK(); 1273 restart: 1274 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1275 if (pr->ndpr_ifp != ifp) 1276 continue; 1277 1278 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1279 dst6 = (const struct sockaddr *)&pr->ndpr_prefix; 1280 1281 /* 1282 * We only need to check all FIBs if add_addr_allfibs 1283 * is unset. If set, checking any FIB will suffice. 1284 */ 1285 fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0; 1286 for (; fibnum < rt_numfibs; fibnum++) { 1287 genid = V_nd6_list_genid; 1288 ND6_RUNLOCK(); 1289 1290 /* 1291 * Restore length field before 1292 * retrying lookup 1293 */ 1294 rt_key.sin6_len = sizeof(rt_key); 1295 error = rib_lookup_info(fibnum, dst6, 0, 0, 1296 &info); 1297 1298 ND6_RLOCK(); 1299 if (genid != V_nd6_list_genid) 1300 goto restart; 1301 if (error == 0) 1302 break; 1303 } 1304 if (error != 0) 1305 continue; 1306 1307 /* 1308 * This is the case where multiple interfaces 1309 * have the same prefix, but only one is installed 1310 * into the routing table and that prefix entry 1311 * is not the one being examined here. 1312 */ 1313 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1314 &rt_key.sin6_addr)) 1315 continue; 1316 } 1317 1318 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1319 &addr->sin6_addr, &pr->ndpr_mask)) { 1320 ND6_RUNLOCK(); 1321 return (1); 1322 } 1323 } 1324 ND6_RUNLOCK(); 1325 1326 /* 1327 * If the address is assigned on the node of the other side of 1328 * a p2p interface, the address should be a neighbor. 1329 */ 1330 if (ifp->if_flags & IFF_POINTOPOINT) { 1331 struct epoch_tracker et; 1332 1333 NET_EPOCH_ENTER(et); 1334 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1335 if (ifa->ifa_addr->sa_family != addr->sin6_family) 1336 continue; 1337 if (ifa->ifa_dstaddr != NULL && 1338 sa_equal(addr, ifa->ifa_dstaddr)) { 1339 NET_EPOCH_EXIT(et); 1340 return 1; 1341 } 1342 } 1343 NET_EPOCH_EXIT(et); 1344 } 1345 1346 /* 1347 * If the default router list is empty, all addresses are regarded 1348 * as on-link, and thus, as a neighbor. 1349 */ 1350 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1351 nd6_defrouter_list_empty() && 1352 V_nd6_defifindex == ifp->if_index) { 1353 return (1); 1354 } 1355 1356 return (0); 1357 } 1358 1359 /* 1360 * Detect if a given IPv6 address identifies a neighbor on a given link. 1361 * XXX: should take care of the destination of a p2p link? 1362 */ 1363 int 1364 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1365 { 1366 struct llentry *lle; 1367 int rc = 0; 1368 1369 NET_EPOCH_ASSERT(); 1370 IF_AFDATA_UNLOCK_ASSERT(ifp); 1371 if (nd6_is_new_addr_neighbor(addr, ifp)) 1372 return (1); 1373 1374 /* 1375 * Even if the address matches none of our addresses, it might be 1376 * in the neighbor cache. 1377 */ 1378 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1379 LLE_RUNLOCK(lle); 1380 rc = 1; 1381 } 1382 return (rc); 1383 } 1384 1385 /* 1386 * Tries to update @lle address/prepend data with new @lladdr. 1387 * 1388 * Returns true on success. 1389 * In any case, @lle is returned wlocked. 1390 */ 1391 bool 1392 nd6_try_set_entry_addr(struct ifnet *ifp, struct llentry *lle, char *lladdr) 1393 { 1394 u_char linkhdr[LLE_MAX_LINKHDR]; 1395 size_t linkhdrsize; 1396 int lladdr_off; 1397 1398 LLE_WLOCK_ASSERT(lle); 1399 1400 linkhdrsize = sizeof(linkhdr); 1401 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 1402 linkhdr, &linkhdrsize, &lladdr_off) != 0) { 1403 return (false); 1404 } 1405 1406 if (!lltable_acquire_wlock(ifp, lle)) 1407 return (false); 1408 lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize, lladdr_off); 1409 IF_AFDATA_WUNLOCK(ifp); 1410 1411 return (true); 1412 } 1413 1414 /* 1415 * Free an nd6 llinfo entry. 1416 * Since the function would cause significant changes in the kernel, DO NOT 1417 * make it global, unless you have a strong reason for the change, and are sure 1418 * that the change is safe. 1419 * 1420 * Set noinline to be dtrace-friendly 1421 */ 1422 static __noinline void 1423 nd6_free(struct llentry **lnp, int gc) 1424 { 1425 struct ifnet *ifp; 1426 struct llentry *ln; 1427 struct nd_defrouter *dr; 1428 1429 ln = *lnp; 1430 *lnp = NULL; 1431 1432 LLE_WLOCK_ASSERT(ln); 1433 ND6_RLOCK_ASSERT(); 1434 1435 ifp = lltable_get_ifp(ln->lle_tbl); 1436 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0) 1437 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp); 1438 else 1439 dr = NULL; 1440 ND6_RUNLOCK(); 1441 1442 if ((ln->la_flags & LLE_DELETED) == 0) 1443 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 1444 1445 /* 1446 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1447 * even though it is not harmful, it was not really necessary. 1448 */ 1449 1450 /* cancel timer */ 1451 nd6_llinfo_settimer_locked(ln, -1); 1452 1453 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1454 if (dr != NULL && dr->expire && 1455 ln->ln_state == ND6_LLINFO_STALE && gc) { 1456 /* 1457 * If the reason for the deletion is just garbage 1458 * collection, and the neighbor is an active default 1459 * router, do not delete it. Instead, reset the GC 1460 * timer using the router's lifetime. 1461 * Simply deleting the entry would affect default 1462 * router selection, which is not necessarily a good 1463 * thing, especially when we're using router preference 1464 * values. 1465 * XXX: the check for ln_state would be redundant, 1466 * but we intentionally keep it just in case. 1467 */ 1468 if (dr->expire > time_uptime) 1469 nd6_llinfo_settimer_locked(ln, 1470 (dr->expire - time_uptime) * hz); 1471 else 1472 nd6_llinfo_settimer_locked(ln, 1473 (long)V_nd6_gctimer * hz); 1474 1475 LLE_REMREF(ln); 1476 LLE_WUNLOCK(ln); 1477 defrouter_rele(dr); 1478 return; 1479 } 1480 1481 if (dr) { 1482 /* 1483 * Unreachablity of a router might affect the default 1484 * router selection and on-link detection of advertised 1485 * prefixes. 1486 */ 1487 1488 /* 1489 * Temporarily fake the state to choose a new default 1490 * router and to perform on-link determination of 1491 * prefixes correctly. 1492 * Below the state will be set correctly, 1493 * or the entry itself will be deleted. 1494 */ 1495 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1496 } 1497 1498 if (ln->ln_router || dr) { 1499 /* 1500 * We need to unlock to avoid a LOR with rt6_flush() with the 1501 * rnh and for the calls to pfxlist_onlink_check() and 1502 * defrouter_select_fib() in the block further down for calls 1503 * into nd6_lookup(). We still hold a ref. 1504 */ 1505 LLE_WUNLOCK(ln); 1506 1507 /* 1508 * rt6_flush must be called whether or not the neighbor 1509 * is in the Default Router List. 1510 * See a corresponding comment in nd6_na_input(). 1511 */ 1512 rt6_flush(&ln->r_l3addr.addr6, ifp); 1513 } 1514 1515 if (dr) { 1516 /* 1517 * Since defrouter_select_fib() does not affect the 1518 * on-link determination and MIP6 needs the check 1519 * before the default router selection, we perform 1520 * the check now. 1521 */ 1522 pfxlist_onlink_check(); 1523 1524 /* 1525 * Refresh default router list. 1526 */ 1527 defrouter_select_fib(dr->ifp->if_fib); 1528 } 1529 1530 /* 1531 * If this entry was added by an on-link redirect, remove the 1532 * corresponding host route. 1533 */ 1534 if (ln->la_flags & LLE_REDIRECT) 1535 nd6_free_redirect(ln); 1536 1537 if (ln->ln_router || dr) 1538 LLE_WLOCK(ln); 1539 } 1540 1541 /* 1542 * Save to unlock. We still hold an extra reference and will not 1543 * free(9) in llentry_free() if someone else holds one as well. 1544 */ 1545 LLE_WUNLOCK(ln); 1546 IF_AFDATA_LOCK(ifp); 1547 LLE_WLOCK(ln); 1548 /* Guard against race with other llentry_free(). */ 1549 if (ln->la_flags & LLE_LINKED) { 1550 /* Remove callout reference */ 1551 LLE_REMREF(ln); 1552 lltable_unlink_entry(ln->lle_tbl, ln); 1553 } 1554 IF_AFDATA_UNLOCK(ifp); 1555 1556 llentry_free(ln); 1557 if (dr != NULL) 1558 defrouter_rele(dr); 1559 } 1560 1561 static int 1562 nd6_isdynrte(const struct rtentry *rt, const struct nhop_object *nh, void *xap) 1563 { 1564 1565 if (nh->nh_flags & NHF_REDIRECT) 1566 return (1); 1567 1568 return (0); 1569 } 1570 1571 /* 1572 * Remove the rtentry for the given llentry, 1573 * both of which were installed by a redirect. 1574 */ 1575 static void 1576 nd6_free_redirect(const struct llentry *ln) 1577 { 1578 int fibnum; 1579 struct sockaddr_in6 sin6; 1580 struct rt_addrinfo info; 1581 struct rib_cmd_info rc; 1582 struct epoch_tracker et; 1583 1584 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); 1585 memset(&info, 0, sizeof(info)); 1586 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; 1587 info.rti_filter = nd6_isdynrte; 1588 1589 NET_EPOCH_ENTER(et); 1590 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 1591 rib_action(fibnum, RTM_DELETE, &info, &rc); 1592 NET_EPOCH_EXIT(et); 1593 } 1594 1595 /* 1596 * Updates status of the default router route. 1597 */ 1598 static void 1599 check_release_defrouter(struct rib_cmd_info *rc, void *_cbdata) 1600 { 1601 struct nd_defrouter *dr; 1602 struct nhop_object *nh; 1603 1604 nh = rc->rc_nh_old; 1605 1606 if ((nh != NULL) && (nh->nh_flags & NHF_DEFAULT)) { 1607 dr = defrouter_lookup(&nh->gw6_sa.sin6_addr, nh->nh_ifp); 1608 if (dr != NULL) { 1609 dr->installed = 0; 1610 defrouter_rele(dr); 1611 } 1612 } 1613 } 1614 1615 void 1616 nd6_subscription_cb(struct rib_head *rnh, struct rib_cmd_info *rc, void *arg) 1617 { 1618 1619 #ifdef ROUTE_MPATH 1620 rib_decompose_notification(rc, check_release_defrouter, NULL); 1621 #else 1622 check_release_defrouter(rc, NULL); 1623 #endif 1624 } 1625 1626 int 1627 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1628 { 1629 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1630 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1631 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1632 struct epoch_tracker et; 1633 int error = 0; 1634 1635 if (ifp->if_afdata[AF_INET6] == NULL) 1636 return (EPFNOSUPPORT); 1637 switch (cmd) { 1638 case OSIOCGIFINFO_IN6: 1639 #define ND ndi->ndi 1640 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1641 bzero(&ND, sizeof(ND)); 1642 ND.linkmtu = IN6_LINKMTU(ifp); 1643 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1644 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1645 ND.reachable = ND_IFINFO(ifp)->reachable; 1646 ND.retrans = ND_IFINFO(ifp)->retrans; 1647 ND.flags = ND_IFINFO(ifp)->flags; 1648 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1649 ND.chlim = ND_IFINFO(ifp)->chlim; 1650 break; 1651 case SIOCGIFINFO_IN6: 1652 ND = *ND_IFINFO(ifp); 1653 break; 1654 case SIOCSIFINFO_IN6: 1655 /* 1656 * used to change host variables from userland. 1657 * intended for a use on router to reflect RA configurations. 1658 */ 1659 /* 0 means 'unspecified' */ 1660 if (ND.linkmtu != 0) { 1661 if (ND.linkmtu < IPV6_MMTU || 1662 ND.linkmtu > IN6_LINKMTU(ifp)) { 1663 error = EINVAL; 1664 break; 1665 } 1666 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1667 } 1668 1669 if (ND.basereachable != 0) { 1670 int obasereachable = ND_IFINFO(ifp)->basereachable; 1671 1672 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1673 if (ND.basereachable != obasereachable) 1674 ND_IFINFO(ifp)->reachable = 1675 ND_COMPUTE_RTIME(ND.basereachable); 1676 } 1677 if (ND.retrans != 0) 1678 ND_IFINFO(ifp)->retrans = ND.retrans; 1679 if (ND.chlim != 0) 1680 ND_IFINFO(ifp)->chlim = ND.chlim; 1681 /* FALLTHROUGH */ 1682 case SIOCSIFINFO_FLAGS: 1683 { 1684 struct ifaddr *ifa; 1685 struct in6_ifaddr *ia; 1686 1687 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1688 !(ND.flags & ND6_IFF_IFDISABLED)) { 1689 /* ifdisabled 1->0 transision */ 1690 1691 /* 1692 * If the interface is marked as ND6_IFF_IFDISABLED and 1693 * has an link-local address with IN6_IFF_DUPLICATED, 1694 * do not clear ND6_IFF_IFDISABLED. 1695 * See RFC 4862, Section 5.4.5. 1696 */ 1697 NET_EPOCH_ENTER(et); 1698 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1699 if (ifa->ifa_addr->sa_family != AF_INET6) 1700 continue; 1701 ia = (struct in6_ifaddr *)ifa; 1702 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1703 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1704 break; 1705 } 1706 NET_EPOCH_EXIT(et); 1707 1708 if (ifa != NULL) { 1709 /* LLA is duplicated. */ 1710 ND.flags |= ND6_IFF_IFDISABLED; 1711 log(LOG_ERR, "Cannot enable an interface" 1712 " with a link-local address marked" 1713 " duplicate.\n"); 1714 } else { 1715 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1716 if (ifp->if_flags & IFF_UP) 1717 in6_if_up(ifp); 1718 } 1719 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1720 (ND.flags & ND6_IFF_IFDISABLED)) { 1721 /* ifdisabled 0->1 transision */ 1722 /* Mark all IPv6 address as tentative. */ 1723 1724 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1725 if (V_ip6_dad_count > 0 && 1726 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1727 NET_EPOCH_ENTER(et); 1728 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1729 ifa_link) { 1730 if (ifa->ifa_addr->sa_family != 1731 AF_INET6) 1732 continue; 1733 ia = (struct in6_ifaddr *)ifa; 1734 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1735 } 1736 NET_EPOCH_EXIT(et); 1737 } 1738 } 1739 1740 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1741 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1742 /* auto_linklocal 0->1 transision */ 1743 1744 /* If no link-local address on ifp, configure */ 1745 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1746 in6_ifattach(ifp, NULL); 1747 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1748 ifp->if_flags & IFF_UP) { 1749 /* 1750 * When the IF already has 1751 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1752 * address is assigned, and IFF_UP, try to 1753 * assign one. 1754 */ 1755 NET_EPOCH_ENTER(et); 1756 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1757 ifa_link) { 1758 if (ifa->ifa_addr->sa_family != 1759 AF_INET6) 1760 continue; 1761 ia = (struct in6_ifaddr *)ifa; 1762 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1763 break; 1764 } 1765 NET_EPOCH_EXIT(et); 1766 if (ifa != NULL) 1767 /* No LLA is configured. */ 1768 in6_ifattach(ifp, NULL); 1769 } 1770 } 1771 ND_IFINFO(ifp)->flags = ND.flags; 1772 break; 1773 } 1774 #undef ND 1775 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1776 /* sync kernel routing table with the default router list */ 1777 defrouter_reset(); 1778 defrouter_select_fib(RT_ALL_FIBS); 1779 break; 1780 case SIOCSPFXFLUSH_IN6: 1781 { 1782 /* flush all the prefix advertised by routers */ 1783 struct in6_ifaddr *ia, *ia_next; 1784 struct nd_prefix *pr, *next; 1785 struct nd_prhead prl; 1786 1787 LIST_INIT(&prl); 1788 1789 ND6_WLOCK(); 1790 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1791 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1792 continue; /* XXX */ 1793 nd6_prefix_unlink(pr, &prl); 1794 } 1795 ND6_WUNLOCK(); 1796 1797 while ((pr = LIST_FIRST(&prl)) != NULL) { 1798 LIST_REMOVE(pr, ndpr_entry); 1799 /* XXXRW: in6_ifaddrhead locking. */ 1800 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1801 ia_next) { 1802 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1803 continue; 1804 1805 if (ia->ia6_ndpr == pr) 1806 in6_purgeaddr(&ia->ia_ifa); 1807 } 1808 nd6_prefix_del(pr); 1809 } 1810 break; 1811 } 1812 case SIOCSRTRFLUSH_IN6: 1813 { 1814 /* flush all the default routers */ 1815 1816 defrouter_reset(); 1817 nd6_defrouter_flush_all(); 1818 defrouter_select_fib(RT_ALL_FIBS); 1819 break; 1820 } 1821 case SIOCGNBRINFO_IN6: 1822 { 1823 struct llentry *ln; 1824 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1825 1826 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1827 return (error); 1828 1829 NET_EPOCH_ENTER(et); 1830 ln = nd6_lookup(&nb_addr, 0, ifp); 1831 NET_EPOCH_EXIT(et); 1832 1833 if (ln == NULL) { 1834 error = EINVAL; 1835 break; 1836 } 1837 nbi->state = ln->ln_state; 1838 nbi->asked = ln->la_asked; 1839 nbi->isrouter = ln->ln_router; 1840 if (ln->la_expire == 0) 1841 nbi->expire = 0; 1842 else 1843 nbi->expire = ln->la_expire + ln->lle_remtime / hz + 1844 (time_second - time_uptime); 1845 LLE_RUNLOCK(ln); 1846 break; 1847 } 1848 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1849 ndif->ifindex = V_nd6_defifindex; 1850 break; 1851 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1852 return (nd6_setdefaultiface(ndif->ifindex)); 1853 } 1854 return (error); 1855 } 1856 1857 /* 1858 * Calculates new isRouter value based on provided parameters and 1859 * returns it. 1860 */ 1861 static int 1862 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, 1863 int ln_router) 1864 { 1865 1866 /* 1867 * ICMP6 type dependent behavior. 1868 * 1869 * NS: clear IsRouter if new entry 1870 * RS: clear IsRouter 1871 * RA: set IsRouter if there's lladdr 1872 * redir: clear IsRouter if new entry 1873 * 1874 * RA case, (1): 1875 * The spec says that we must set IsRouter in the following cases: 1876 * - If lladdr exist, set IsRouter. This means (1-5). 1877 * - If it is old entry (!newentry), set IsRouter. This means (7). 1878 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1879 * A quetion arises for (1) case. (1) case has no lladdr in the 1880 * neighbor cache, this is similar to (6). 1881 * This case is rare but we figured that we MUST NOT set IsRouter. 1882 * 1883 * is_new old_addr new_addr NS RS RA redir 1884 * D R 1885 * 0 n n (1) c ? s 1886 * 0 y n (2) c s s 1887 * 0 n y (3) c s s 1888 * 0 y y (4) c s s 1889 * 0 y y (5) c s s 1890 * 1 -- n (6) c c c s 1891 * 1 -- y (7) c c s c s 1892 * 1893 * (c=clear s=set) 1894 */ 1895 switch (type & 0xff) { 1896 case ND_NEIGHBOR_SOLICIT: 1897 /* 1898 * New entry must have is_router flag cleared. 1899 */ 1900 if (is_new) /* (6-7) */ 1901 ln_router = 0; 1902 break; 1903 case ND_REDIRECT: 1904 /* 1905 * If the icmp is a redirect to a better router, always set the 1906 * is_router flag. Otherwise, if the entry is newly created, 1907 * clear the flag. [RFC 2461, sec 8.3] 1908 */ 1909 if (code == ND_REDIRECT_ROUTER) 1910 ln_router = 1; 1911 else { 1912 if (is_new) /* (6-7) */ 1913 ln_router = 0; 1914 } 1915 break; 1916 case ND_ROUTER_SOLICIT: 1917 /* 1918 * is_router flag must always be cleared. 1919 */ 1920 ln_router = 0; 1921 break; 1922 case ND_ROUTER_ADVERT: 1923 /* 1924 * Mark an entry with lladdr as a router. 1925 */ 1926 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ 1927 (is_new && new_addr)) { /* (7) */ 1928 ln_router = 1; 1929 } 1930 break; 1931 } 1932 1933 return (ln_router); 1934 } 1935 1936 /* 1937 * Create neighbor cache entry and cache link-layer address, 1938 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1939 * 1940 * type - ICMP6 type 1941 * code - type dependent information 1942 * 1943 */ 1944 void 1945 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1946 int lladdrlen, int type, int code) 1947 { 1948 struct llentry *ln = NULL, *ln_tmp; 1949 int is_newentry; 1950 int do_update; 1951 int olladdr; 1952 int llchange; 1953 int flags; 1954 uint16_t router = 0; 1955 struct mbuf *chain = NULL; 1956 u_char linkhdr[LLE_MAX_LINKHDR]; 1957 size_t linkhdrsize; 1958 int lladdr_off; 1959 1960 NET_EPOCH_ASSERT(); 1961 IF_AFDATA_UNLOCK_ASSERT(ifp); 1962 1963 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1964 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1965 1966 /* nothing must be updated for unspecified address */ 1967 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1968 return; 1969 1970 /* 1971 * Validation about ifp->if_addrlen and lladdrlen must be done in 1972 * the caller. 1973 * 1974 * XXX If the link does not have link-layer adderss, what should 1975 * we do? (ifp->if_addrlen == 0) 1976 * Spec says nothing in sections for RA, RS and NA. There's small 1977 * description on it in NS section (RFC 2461 7.2.3). 1978 */ 1979 flags = lladdr ? LLE_EXCLUSIVE : 0; 1980 ln = nd6_lookup(from, flags, ifp); 1981 is_newentry = 0; 1982 if (ln == NULL) { 1983 flags |= LLE_EXCLUSIVE; 1984 ln = nd6_alloc(from, 0, ifp); 1985 if (ln == NULL) 1986 return; 1987 1988 /* 1989 * Since we already know all the data for the new entry, 1990 * fill it before insertion. 1991 */ 1992 if (lladdr != NULL) { 1993 linkhdrsize = sizeof(linkhdr); 1994 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 1995 linkhdr, &linkhdrsize, &lladdr_off) != 0) 1996 return; 1997 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 1998 lladdr_off); 1999 } 2000 2001 IF_AFDATA_WLOCK(ifp); 2002 LLE_WLOCK(ln); 2003 /* Prefer any existing lle over newly-created one */ 2004 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); 2005 if (ln_tmp == NULL) 2006 lltable_link_entry(LLTABLE6(ifp), ln); 2007 IF_AFDATA_WUNLOCK(ifp); 2008 if (ln_tmp == NULL) { 2009 /* No existing lle, mark as new entry (6,7) */ 2010 is_newentry = 1; 2011 if (lladdr != NULL) { /* (7) */ 2012 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2013 EVENTHANDLER_INVOKE(lle_event, ln, 2014 LLENTRY_RESOLVED); 2015 } 2016 } else { 2017 lltable_free_entry(LLTABLE6(ifp), ln); 2018 ln = ln_tmp; 2019 ln_tmp = NULL; 2020 } 2021 } 2022 /* do nothing if static ndp is set */ 2023 if ((ln->la_flags & LLE_STATIC)) { 2024 if (flags & LLE_EXCLUSIVE) 2025 LLE_WUNLOCK(ln); 2026 else 2027 LLE_RUNLOCK(ln); 2028 return; 2029 } 2030 2031 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 2032 if (olladdr && lladdr) { 2033 llchange = bcmp(lladdr, ln->ll_addr, 2034 ifp->if_addrlen); 2035 } else if (!olladdr && lladdr) 2036 llchange = 1; 2037 else 2038 llchange = 0; 2039 2040 /* 2041 * newentry olladdr lladdr llchange (*=record) 2042 * 0 n n -- (1) 2043 * 0 y n -- (2) 2044 * 0 n y y (3) * STALE 2045 * 0 y y n (4) * 2046 * 0 y y y (5) * STALE 2047 * 1 -- n -- (6) NOSTATE(= PASSIVE) 2048 * 1 -- y -- (7) * STALE 2049 */ 2050 2051 do_update = 0; 2052 if (is_newentry == 0 && llchange != 0) { 2053 do_update = 1; /* (3,5) */ 2054 2055 /* 2056 * Record source link-layer address 2057 * XXX is it dependent to ifp->if_type? 2058 */ 2059 if (!nd6_try_set_entry_addr(ifp, ln, lladdr)) { 2060 /* Entry was deleted */ 2061 LLE_WUNLOCK(ln); 2062 return; 2063 } 2064 2065 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2066 2067 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2068 2069 if (ln->la_hold != NULL) 2070 chain = nd6_grab_holdchain(ln); 2071 } 2072 2073 /* Calculates new router status */ 2074 router = nd6_is_router(type, code, is_newentry, olladdr, 2075 lladdr != NULL ? 1 : 0, ln->ln_router); 2076 2077 ln->ln_router = router; 2078 /* Mark non-router redirects with special flag */ 2079 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) 2080 ln->la_flags |= LLE_REDIRECT; 2081 2082 if (flags & LLE_EXCLUSIVE) 2083 LLE_WUNLOCK(ln); 2084 else 2085 LLE_RUNLOCK(ln); 2086 2087 if (chain != NULL) 2088 nd6_flush_holdchain(ifp, ln, chain); 2089 2090 /* 2091 * When the link-layer address of a router changes, select the 2092 * best router again. In particular, when the neighbor entry is newly 2093 * created, it might affect the selection policy. 2094 * Question: can we restrict the first condition to the "is_newentry" 2095 * case? 2096 * XXX: when we hear an RA from a new router with the link-layer 2097 * address option, defrouter_select_fib() is called twice, since 2098 * defrtrlist_update called the function as well. However, I believe 2099 * we can compromise the overhead, since it only happens the first 2100 * time. 2101 * XXX: although defrouter_select_fib() should not have a bad effect 2102 * for those are not autoconfigured hosts, we explicitly avoid such 2103 * cases for safety. 2104 */ 2105 if ((do_update || is_newentry) && router && 2106 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 2107 /* 2108 * guaranteed recursion 2109 */ 2110 defrouter_select_fib(ifp->if_fib); 2111 } 2112 } 2113 2114 static void 2115 nd6_slowtimo(void *arg) 2116 { 2117 struct epoch_tracker et; 2118 CURVNET_SET((struct vnet *) arg); 2119 struct nd_ifinfo *nd6if; 2120 struct ifnet *ifp; 2121 2122 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 2123 nd6_slowtimo, curvnet); 2124 NET_EPOCH_ENTER(et); 2125 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2126 if (ifp->if_afdata[AF_INET6] == NULL) 2127 continue; 2128 nd6if = ND_IFINFO(ifp); 2129 if (nd6if->basereachable && /* already initialized */ 2130 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 2131 /* 2132 * Since reachable time rarely changes by router 2133 * advertisements, we SHOULD insure that a new random 2134 * value gets recomputed at least once every few hours. 2135 * (RFC 2461, 6.3.4) 2136 */ 2137 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 2138 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 2139 } 2140 } 2141 NET_EPOCH_EXIT(et); 2142 CURVNET_RESTORE(); 2143 } 2144 2145 struct mbuf * 2146 nd6_grab_holdchain(struct llentry *ln) 2147 { 2148 struct mbuf *chain; 2149 2150 LLE_WLOCK_ASSERT(ln); 2151 2152 chain = ln->la_hold; 2153 ln->la_hold = NULL; 2154 2155 if (ln->ln_state == ND6_LLINFO_STALE) { 2156 /* 2157 * The first time we send a packet to a 2158 * neighbor whose entry is STALE, we have 2159 * to change the state to DELAY and a sets 2160 * a timer to expire in DELAY_FIRST_PROBE_TIME 2161 * seconds to ensure do neighbor unreachability 2162 * detection on expiration. 2163 * (RFC 2461 7.3.3) 2164 */ 2165 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); 2166 } 2167 2168 return (chain); 2169 } 2170 2171 int 2172 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2173 struct sockaddr_in6 *dst, struct route *ro) 2174 { 2175 int error; 2176 int ip6len; 2177 struct ip6_hdr *ip6; 2178 struct m_tag *mtag; 2179 2180 #ifdef MAC 2181 mac_netinet6_nd6_send(ifp, m); 2182 #endif 2183 2184 /* 2185 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2186 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2187 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2188 * to be diverted to user space. When re-injected into the kernel, 2189 * send_output() will directly dispatch them to the outgoing interface. 2190 */ 2191 if (send_sendso_input_hook != NULL) { 2192 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2193 if (mtag != NULL) { 2194 ip6 = mtod(m, struct ip6_hdr *); 2195 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2196 /* Use the SEND socket */ 2197 error = send_sendso_input_hook(m, ifp, SND_OUT, 2198 ip6len); 2199 /* -1 == no app on SEND socket */ 2200 if (error == 0 || error != -1) 2201 return (error); 2202 } 2203 } 2204 2205 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 2206 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 2207 mtod(m, struct ip6_hdr *)); 2208 2209 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2210 origifp = ifp; 2211 2212 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); 2213 return (error); 2214 } 2215 2216 /* 2217 * Lookup link headerfor @sa_dst address. Stores found 2218 * data in @desten buffer. Copy of lle ln_flags can be also 2219 * saved in @pflags if @pflags is non-NULL. 2220 * 2221 * If destination LLE does not exists or lle state modification 2222 * is required, call "slow" version. 2223 * 2224 * Return values: 2225 * - 0 on success (address copied to buffer). 2226 * - EWOULDBLOCK (no local error, but address is still unresolved) 2227 * - other errors (alloc failure, etc) 2228 */ 2229 int 2230 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 2231 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags, 2232 struct llentry **plle) 2233 { 2234 struct llentry *ln = NULL; 2235 const struct sockaddr_in6 *dst6; 2236 2237 NET_EPOCH_ASSERT(); 2238 2239 if (pflags != NULL) 2240 *pflags = 0; 2241 2242 dst6 = (const struct sockaddr_in6 *)sa_dst; 2243 2244 /* discard the packet if IPv6 operation is disabled on the interface */ 2245 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2246 m_freem(m); 2247 return (ENETDOWN); /* better error? */ 2248 } 2249 2250 if (m != NULL && m->m_flags & M_MCAST) { 2251 switch (ifp->if_type) { 2252 case IFT_ETHER: 2253 case IFT_L2VLAN: 2254 case IFT_BRIDGE: 2255 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, 2256 desten); 2257 return (0); 2258 default: 2259 m_freem(m); 2260 return (EAFNOSUPPORT); 2261 } 2262 } 2263 2264 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, 2265 ifp); 2266 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { 2267 /* Entry found, let's copy lle info */ 2268 bcopy(ln->r_linkdata, desten, ln->r_hdrlen); 2269 if (pflags != NULL) 2270 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); 2271 llentry_provide_feedback(ln); 2272 if (plle) { 2273 LLE_ADDREF(ln); 2274 *plle = ln; 2275 LLE_WUNLOCK(ln); 2276 } 2277 return (0); 2278 } else if (plle && ln) 2279 LLE_WUNLOCK(ln); 2280 2281 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle)); 2282 } 2283 2284 /* 2285 * Finds or creates a new llentry for @addr. 2286 * Returns wlocked llentry or NULL. 2287 */ 2288 static __noinline struct llentry * 2289 nd6_get_llentry(struct ifnet *ifp, const struct in6_addr *addr) 2290 { 2291 struct llentry *lle, *lle_tmp; 2292 2293 lle = nd6_alloc(addr, 0, ifp); 2294 if (lle == NULL) { 2295 char ip6buf[INET6_ADDRSTRLEN]; 2296 log(LOG_DEBUG, 2297 "nd6_get_llentry: can't allocate llinfo for %s " 2298 "(ln=%p)\n", 2299 ip6_sprintf(ip6buf, addr), lle); 2300 return (NULL); 2301 } 2302 2303 IF_AFDATA_WLOCK(ifp); 2304 LLE_WLOCK(lle); 2305 /* Prefer any existing entry over newly-created one */ 2306 lle_tmp = nd6_lookup(addr, LLE_EXCLUSIVE, ifp); 2307 if (lle_tmp == NULL) 2308 lltable_link_entry(LLTABLE6(ifp), lle); 2309 IF_AFDATA_WUNLOCK(ifp); 2310 if (lle_tmp != NULL) { 2311 lltable_free_entry(LLTABLE6(ifp), lle); 2312 return (lle_tmp); 2313 } else 2314 return (lle); 2315 } 2316 2317 /* 2318 * Do L2 address resolution for @sa_dst address. Stores found 2319 * address in @desten buffer. Copy of lle ln_flags can be also 2320 * saved in @pflags if @pflags is non-NULL. 2321 * 2322 * Heavy version. 2323 * Function assume that destination LLE does not exist, 2324 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. 2325 * 2326 * Set noinline to be dtrace-friendly 2327 */ 2328 static __noinline int 2329 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m, 2330 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags, 2331 struct llentry **plle) 2332 { 2333 struct llentry *lle = NULL; 2334 struct in6_addr *psrc, src; 2335 int send_ns, ll_len; 2336 char *lladdr; 2337 2338 NET_EPOCH_ASSERT(); 2339 2340 /* 2341 * Address resolution or Neighbor Unreachability Detection 2342 * for the next hop. 2343 * At this point, the destination of the packet must be a unicast 2344 * or an anycast address(i.e. not a multicast). 2345 */ 2346 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2347 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2348 /* 2349 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2350 * the condition below is not very efficient. But we believe 2351 * it is tolerable, because this should be a rare case. 2352 */ 2353 lle = nd6_get_llentry(ifp, &dst->sin6_addr); 2354 } 2355 2356 if (lle == NULL) { 2357 m_freem(m); 2358 return (ENOBUFS); 2359 } 2360 2361 LLE_WLOCK_ASSERT(lle); 2362 2363 /* 2364 * The first time we send a packet to a neighbor whose entry is 2365 * STALE, we have to change the state to DELAY and a sets a timer to 2366 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2367 * neighbor unreachability detection on expiration. 2368 * (RFC 2461 7.3.3) 2369 */ 2370 if (lle->ln_state == ND6_LLINFO_STALE) 2371 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); 2372 2373 /* 2374 * If the neighbor cache entry has a state other than INCOMPLETE 2375 * (i.e. its link-layer address is already resolved), just 2376 * send the packet. 2377 */ 2378 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { 2379 if (flags & LLE_ADDRONLY) { 2380 lladdr = lle->ll_addr; 2381 ll_len = ifp->if_addrlen; 2382 } else { 2383 lladdr = lle->r_linkdata; 2384 ll_len = lle->r_hdrlen; 2385 } 2386 bcopy(lladdr, desten, ll_len); 2387 if (pflags != NULL) 2388 *pflags = lle->la_flags; 2389 if (plle) { 2390 LLE_ADDREF(lle); 2391 *plle = lle; 2392 } 2393 LLE_WUNLOCK(lle); 2394 return (0); 2395 } 2396 2397 /* 2398 * There is a neighbor cache entry, but no ethernet address 2399 * response yet. Append this latest packet to the end of the 2400 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen, 2401 * the oldest packet in the queue will be removed. 2402 */ 2403 2404 if (lle->la_hold != NULL) { 2405 struct mbuf *m_hold; 2406 int i; 2407 2408 i = 0; 2409 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2410 i++; 2411 if (m_hold->m_nextpkt == NULL) { 2412 m_hold->m_nextpkt = m; 2413 break; 2414 } 2415 } 2416 while (i >= V_nd6_maxqueuelen) { 2417 m_hold = lle->la_hold; 2418 lle->la_hold = lle->la_hold->m_nextpkt; 2419 m_freem(m_hold); 2420 i--; 2421 } 2422 } else { 2423 lle->la_hold = m; 2424 } 2425 2426 /* 2427 * If there has been no NS for the neighbor after entering the 2428 * INCOMPLETE state, send the first solicitation. 2429 * Note that for newly-created lle la_asked will be 0, 2430 * so we will transition from ND6_LLINFO_NOSTATE to 2431 * ND6_LLINFO_INCOMPLETE state here. 2432 */ 2433 psrc = NULL; 2434 send_ns = 0; 2435 if (lle->la_asked == 0) { 2436 lle->la_asked++; 2437 send_ns = 1; 2438 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2439 2440 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); 2441 } 2442 LLE_WUNLOCK(lle); 2443 if (send_ns != 0) 2444 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2445 2446 return (EWOULDBLOCK); 2447 } 2448 2449 /* 2450 * Do L2 address resolution for @sa_dst address. Stores found 2451 * address in @desten buffer. Copy of lle ln_flags can be also 2452 * saved in @pflags if @pflags is non-NULL. 2453 * 2454 * Return values: 2455 * - 0 on success (address copied to buffer). 2456 * - EWOULDBLOCK (no local error, but address is still unresolved) 2457 * - other errors (alloc failure, etc) 2458 */ 2459 int 2460 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 2461 char *desten, uint32_t *pflags) 2462 { 2463 int error; 2464 2465 flags |= LLE_ADDRONLY; 2466 error = nd6_resolve_slow(ifp, flags, NULL, 2467 (const struct sockaddr_in6 *)dst, desten, pflags, NULL); 2468 return (error); 2469 } 2470 2471 int 2472 nd6_flush_holdchain(struct ifnet *ifp, struct llentry *lle, struct mbuf *chain) 2473 { 2474 struct mbuf *m, *m_head; 2475 struct sockaddr_in6 dst6; 2476 int error = 0; 2477 2478 NET_EPOCH_ASSERT(); 2479 2480 struct route_in6 ro = { 2481 .ro_prepend = lle->r_linkdata, 2482 .ro_plen = lle->r_hdrlen, 2483 }; 2484 2485 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst6); 2486 m_head = chain; 2487 2488 while (m_head) { 2489 m = m_head; 2490 m_head = m_head->m_nextpkt; 2491 m->m_nextpkt = NULL; 2492 error = nd6_output_ifp(ifp, ifp, m, &dst6, (struct route *)&ro); 2493 } 2494 2495 /* 2496 * XXX 2497 * note that intermediate errors are blindly ignored 2498 */ 2499 return (error); 2500 } 2501 2502 static int 2503 nd6_need_cache(struct ifnet *ifp) 2504 { 2505 /* 2506 * XXX: we currently do not make neighbor cache on any interface 2507 * other than Ethernet and GIF. 2508 * 2509 * RFC2893 says: 2510 * - unidirectional tunnels needs no ND 2511 */ 2512 switch (ifp->if_type) { 2513 case IFT_ETHER: 2514 case IFT_IEEE1394: 2515 case IFT_L2VLAN: 2516 case IFT_INFINIBAND: 2517 case IFT_BRIDGE: 2518 case IFT_PROPVIRTUAL: 2519 return (1); 2520 default: 2521 return (0); 2522 } 2523 } 2524 2525 /* 2526 * Add pernament ND6 link-layer record for given 2527 * interface address. 2528 * 2529 * Very similar to IPv4 arp_ifinit(), but: 2530 * 1) IPv6 DAD is performed in different place 2531 * 2) It is called by IPv6 protocol stack in contrast to 2532 * arp_ifinit() which is typically called in SIOCSIFADDR 2533 * driver ioctl handler. 2534 * 2535 */ 2536 int 2537 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2538 { 2539 struct ifnet *ifp; 2540 struct llentry *ln, *ln_tmp; 2541 struct sockaddr *dst; 2542 2543 ifp = ia->ia_ifa.ifa_ifp; 2544 if (nd6_need_cache(ifp) == 0) 2545 return (0); 2546 2547 dst = (struct sockaddr *)&ia->ia_addr; 2548 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2549 if (ln == NULL) 2550 return (ENOBUFS); 2551 2552 IF_AFDATA_WLOCK(ifp); 2553 LLE_WLOCK(ln); 2554 /* Unlink any entry if exists */ 2555 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2556 if (ln_tmp != NULL) 2557 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2558 lltable_link_entry(LLTABLE6(ifp), ln); 2559 IF_AFDATA_WUNLOCK(ifp); 2560 2561 if (ln_tmp != NULL) 2562 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2563 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2564 2565 LLE_WUNLOCK(ln); 2566 if (ln_tmp != NULL) 2567 llentry_free(ln_tmp); 2568 2569 return (0); 2570 } 2571 2572 /* 2573 * Removes either all lle entries for given @ia, or lle 2574 * corresponding to @ia address. 2575 */ 2576 void 2577 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) 2578 { 2579 struct sockaddr_in6 mask, addr; 2580 struct sockaddr *saddr, *smask; 2581 struct ifnet *ifp; 2582 2583 ifp = ia->ia_ifa.ifa_ifp; 2584 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2585 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2586 saddr = (struct sockaddr *)&addr; 2587 smask = (struct sockaddr *)&mask; 2588 2589 if (all != 0) 2590 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); 2591 else 2592 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); 2593 } 2594 2595 static void 2596 clear_llinfo_pqueue(struct llentry *ln) 2597 { 2598 struct mbuf *m_hold, *m_hold_next; 2599 2600 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2601 m_hold_next = m_hold->m_nextpkt; 2602 m_freem(m_hold); 2603 } 2604 2605 ln->la_hold = NULL; 2606 } 2607 2608 static int 2609 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2610 { 2611 struct in6_prefix p; 2612 struct sockaddr_in6 s6; 2613 struct nd_prefix *pr; 2614 struct nd_pfxrouter *pfr; 2615 time_t maxexpire; 2616 int error; 2617 char ip6buf[INET6_ADDRSTRLEN]; 2618 2619 if (req->newptr) 2620 return (EPERM); 2621 2622 error = sysctl_wire_old_buffer(req, 0); 2623 if (error != 0) 2624 return (error); 2625 2626 bzero(&p, sizeof(p)); 2627 p.origin = PR_ORIG_RA; 2628 bzero(&s6, sizeof(s6)); 2629 s6.sin6_family = AF_INET6; 2630 s6.sin6_len = sizeof(s6); 2631 2632 ND6_RLOCK(); 2633 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2634 p.prefix = pr->ndpr_prefix; 2635 if (sa6_recoverscope(&p.prefix)) { 2636 log(LOG_ERR, "scope error in prefix list (%s)\n", 2637 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2638 /* XXX: press on... */ 2639 } 2640 p.raflags = pr->ndpr_raf; 2641 p.prefixlen = pr->ndpr_plen; 2642 p.vltime = pr->ndpr_vltime; 2643 p.pltime = pr->ndpr_pltime; 2644 p.if_index = pr->ndpr_ifp->if_index; 2645 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2646 p.expire = 0; 2647 else { 2648 /* XXX: we assume time_t is signed. */ 2649 maxexpire = (-1) & 2650 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2651 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2652 p.expire = pr->ndpr_lastupdate + 2653 pr->ndpr_vltime + 2654 (time_second - time_uptime); 2655 else 2656 p.expire = maxexpire; 2657 } 2658 p.refcnt = pr->ndpr_addrcnt; 2659 p.flags = pr->ndpr_stateflags; 2660 p.advrtrs = 0; 2661 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2662 p.advrtrs++; 2663 error = SYSCTL_OUT(req, &p, sizeof(p)); 2664 if (error != 0) 2665 break; 2666 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2667 s6.sin6_addr = pfr->router->rtaddr; 2668 if (sa6_recoverscope(&s6)) 2669 log(LOG_ERR, 2670 "scope error in prefix list (%s)\n", 2671 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2672 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2673 if (error != 0) 2674 goto out; 2675 } 2676 } 2677 out: 2678 ND6_RUNLOCK(); 2679 return (error); 2680 } 2681 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2682 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2683 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix", 2684 "NDP prefix list"); 2685 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2686 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2687 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2688 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2689