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/eventhandler.h> 43 #include <sys/callout.h> 44 #include <sys/lock.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/mutex.h> 48 #include <sys/socket.h> 49 #include <sys/sockio.h> 50 #include <sys/time.h> 51 #include <sys/kernel.h> 52 #include <sys/protosw.h> 53 #include <sys/errno.h> 54 #include <sys/syslog.h> 55 #include <sys/rwlock.h> 56 #include <sys/queue.h> 57 #include <sys/sdt.h> 58 #include <sys/sysctl.h> 59 60 #include <net/if.h> 61 #include <net/if_var.h> 62 #include <net/if_dl.h> 63 #include <net/if_types.h> 64 #include <net/route.h> 65 #include <net/route/route_ctl.h> 66 #include <net/route/nhop.h> 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_kdtrace.h> 71 #include <net/if_llatbl.h> 72 #include <netinet/if_ether.h> 73 #include <netinet6/in6_var.h> 74 #include <netinet/ip6.h> 75 #include <netinet6/ip6_var.h> 76 #include <netinet6/scope6_var.h> 77 #include <netinet6/nd6.h> 78 #include <netinet6/in6_ifattach.h> 79 #include <netinet/icmp6.h> 80 #include <netinet6/send.h> 81 82 #include <sys/limits.h> 83 84 #include <security/mac/mac_framework.h> 85 86 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 87 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 88 89 #define SIN6(s) ((const struct sockaddr_in6 *)(s)) 90 91 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery"); 92 93 /* timer values */ 94 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 95 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 96 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 97 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 98 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 99 * local traffic */ 100 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 101 * collection timer */ 102 103 /* preventing too many loops in ND option parsing */ 104 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 105 106 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 107 * layer hints */ 108 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 109 * ND entries */ 110 #define V_nd6_maxndopt VNET(nd6_maxndopt) 111 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 112 113 #ifdef ND6_DEBUG 114 VNET_DEFINE(int, nd6_debug) = 1; 115 #else 116 VNET_DEFINE(int, nd6_debug) = 0; 117 #endif 118 119 static eventhandler_tag lle_event_eh, iflladdr_event_eh, ifnet_link_event_eh; 120 121 VNET_DEFINE(struct nd_prhead, nd_prefix); 122 VNET_DEFINE(struct rwlock, nd6_lock); 123 VNET_DEFINE(uint64_t, nd6_list_genid); 124 VNET_DEFINE(struct mtx, nd6_onlink_mtx); 125 126 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 127 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 128 129 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 130 131 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *, 132 struct ifnet *); 133 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 134 static void nd6_slowtimo(void *); 135 static int regen_tmpaddr(struct in6_ifaddr *); 136 static void nd6_free(struct llentry **, int); 137 static void nd6_free_redirect(const struct llentry *); 138 static void nd6_llinfo_timer(void *); 139 static void nd6_llinfo_settimer_locked(struct llentry *, long); 140 static void clear_llinfo_pqueue(struct llentry *); 141 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *, 142 const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **); 143 static int nd6_need_cache(struct ifnet *); 144 145 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch); 146 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 147 148 VNET_DEFINE_STATIC(struct callout, nd6_timer_ch); 149 #define V_nd6_timer_ch VNET(nd6_timer_ch) 150 151 SYSCTL_DECL(_net_inet6_icmp6); 152 153 static void 154 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 155 { 156 struct rt_addrinfo rtinfo; 157 struct sockaddr_in6 dst; 158 struct sockaddr_dl gw; 159 struct ifnet *ifp; 160 int type; 161 int fibnum; 162 163 LLE_WLOCK_ASSERT(lle); 164 165 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 166 return; 167 168 switch (evt) { 169 case LLENTRY_RESOLVED: 170 type = RTM_ADD; 171 KASSERT(lle->la_flags & LLE_VALID, 172 ("%s: %p resolved but not valid?", __func__, lle)); 173 break; 174 case LLENTRY_EXPIRED: 175 type = RTM_DELETE; 176 break; 177 default: 178 return; 179 } 180 181 ifp = lltable_get_ifp(lle->lle_tbl); 182 183 bzero(&dst, sizeof(dst)); 184 bzero(&gw, sizeof(gw)); 185 bzero(&rtinfo, sizeof(rtinfo)); 186 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 187 dst.sin6_scope_id = in6_getscopezone(ifp, 188 in6_addrscope(&dst.sin6_addr)); 189 gw.sdl_len = sizeof(struct sockaddr_dl); 190 gw.sdl_family = AF_LINK; 191 gw.sdl_alen = ifp->if_addrlen; 192 gw.sdl_index = ifp->if_index; 193 gw.sdl_type = ifp->if_type; 194 if (evt == LLENTRY_RESOLVED) 195 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 196 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 197 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 198 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 199 fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib; 200 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 201 type == RTM_ADD ? RTF_UP: 0), 0, fibnum); 202 } 203 204 /* 205 * A handler for interface link layer address change event. 206 */ 207 static void 208 nd6_iflladdr(void *arg __unused, struct ifnet *ifp) 209 { 210 211 lltable_update_ifaddr(LLTABLE6(ifp)); 212 } 213 214 void 215 nd6_init(void) 216 { 217 218 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF); 219 rw_init(&V_nd6_lock, "nd6 list"); 220 221 LIST_INIT(&V_nd_prefix); 222 nd6_defrouter_init(); 223 224 /* Start timers. */ 225 callout_init(&V_nd6_slowtimo_ch, 0); 226 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 227 nd6_slowtimo, curvnet); 228 229 callout_init(&V_nd6_timer_ch, 0); 230 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 231 232 nd6_dad_init(); 233 if (IS_DEFAULT_VNET(curvnet)) { 234 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 235 NULL, EVENTHANDLER_PRI_ANY); 236 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event, 237 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 238 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event, 239 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY); 240 } 241 } 242 243 #ifdef VIMAGE 244 void 245 nd6_destroy() 246 { 247 248 callout_drain(&V_nd6_slowtimo_ch); 249 callout_drain(&V_nd6_timer_ch); 250 if (IS_DEFAULT_VNET(curvnet)) { 251 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh); 252 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 253 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh); 254 } 255 rw_destroy(&V_nd6_lock); 256 mtx_destroy(&V_nd6_onlink_mtx); 257 } 258 #endif 259 260 struct nd_ifinfo * 261 nd6_ifattach(struct ifnet *ifp) 262 { 263 struct nd_ifinfo *nd; 264 265 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 266 nd->initialized = 1; 267 268 nd->chlim = IPV6_DEFHLIM; 269 nd->basereachable = REACHABLE_TIME; 270 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 271 nd->retrans = RETRANS_TIMER; 272 273 nd->flags = ND6_IFF_PERFORMNUD; 274 275 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 276 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 277 * default regardless of the V_ip6_auto_linklocal configuration to 278 * give a reasonable default behavior. 279 */ 280 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 281 (ifp->if_flags & IFF_LOOPBACK)) 282 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 283 /* 284 * A loopback interface does not need to accept RTADV. 285 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 286 * default regardless of the V_ip6_accept_rtadv configuration to 287 * prevent the interface from accepting RA messages arrived 288 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 289 */ 290 if (V_ip6_accept_rtadv && 291 !(ifp->if_flags & IFF_LOOPBACK) && 292 (ifp->if_type != IFT_BRIDGE)) 293 nd->flags |= ND6_IFF_ACCEPT_RTADV; 294 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 295 nd->flags |= ND6_IFF_NO_RADR; 296 297 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 298 nd6_setmtu0(ifp, nd); 299 300 return nd; 301 } 302 303 void 304 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd) 305 { 306 struct epoch_tracker et; 307 struct ifaddr *ifa, *next; 308 309 NET_EPOCH_ENTER(et); 310 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 311 if (ifa->ifa_addr->sa_family != AF_INET6) 312 continue; 313 314 /* stop DAD processing */ 315 nd6_dad_stop(ifa); 316 } 317 NET_EPOCH_EXIT(et); 318 319 free(nd, M_IP6NDP); 320 } 321 322 /* 323 * Reset ND level link MTU. This function is called when the physical MTU 324 * changes, which means we might have to adjust the ND level MTU. 325 */ 326 void 327 nd6_setmtu(struct ifnet *ifp) 328 { 329 if (ifp->if_afdata[AF_INET6] == NULL) 330 return; 331 332 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 333 } 334 335 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 336 void 337 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 338 { 339 u_int32_t omaxmtu; 340 341 omaxmtu = ndi->maxmtu; 342 ndi->maxmtu = ifp->if_mtu; 343 344 /* 345 * Decreasing the interface MTU under IPV6 minimum MTU may cause 346 * undesirable situation. We thus notify the operator of the change 347 * explicitly. The check for omaxmtu is necessary to restrict the 348 * log to the case of changing the MTU, not initializing it. 349 */ 350 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 351 log(LOG_NOTICE, "nd6_setmtu0: " 352 "new link MTU on %s (%lu) is too small for IPv6\n", 353 if_name(ifp), (unsigned long)ndi->maxmtu); 354 } 355 356 if (ndi->maxmtu > V_in6_maxmtu) 357 in6_setmaxmtu(); /* check all interfaces just in case */ 358 359 } 360 361 void 362 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 363 { 364 365 bzero(ndopts, sizeof(*ndopts)); 366 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 367 ndopts->nd_opts_last 368 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 369 370 if (icmp6len == 0) { 371 ndopts->nd_opts_done = 1; 372 ndopts->nd_opts_search = NULL; 373 } 374 } 375 376 /* 377 * Take one ND option. 378 */ 379 struct nd_opt_hdr * 380 nd6_option(union nd_opts *ndopts) 381 { 382 struct nd_opt_hdr *nd_opt; 383 int olen; 384 385 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 386 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 387 __func__)); 388 if (ndopts->nd_opts_search == NULL) 389 return NULL; 390 if (ndopts->nd_opts_done) 391 return NULL; 392 393 nd_opt = ndopts->nd_opts_search; 394 395 /* make sure nd_opt_len is inside the buffer */ 396 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 397 bzero(ndopts, sizeof(*ndopts)); 398 return NULL; 399 } 400 401 olen = nd_opt->nd_opt_len << 3; 402 if (olen == 0) { 403 /* 404 * Message validation requires that all included 405 * options have a length that is greater than zero. 406 */ 407 bzero(ndopts, sizeof(*ndopts)); 408 return NULL; 409 } 410 411 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 412 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 413 /* option overruns the end of buffer, invalid */ 414 bzero(ndopts, sizeof(*ndopts)); 415 return NULL; 416 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 417 /* reached the end of options chain */ 418 ndopts->nd_opts_done = 1; 419 ndopts->nd_opts_search = NULL; 420 } 421 return nd_opt; 422 } 423 424 /* 425 * Parse multiple ND options. 426 * This function is much easier to use, for ND routines that do not need 427 * multiple options of the same type. 428 */ 429 int 430 nd6_options(union nd_opts *ndopts) 431 { 432 struct nd_opt_hdr *nd_opt; 433 int i = 0; 434 435 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 436 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 437 __func__)); 438 if (ndopts->nd_opts_search == NULL) 439 return 0; 440 441 while (1) { 442 nd_opt = nd6_option(ndopts); 443 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 444 /* 445 * Message validation requires that all included 446 * options have a length that is greater than zero. 447 */ 448 ICMP6STAT_INC(icp6s_nd_badopt); 449 bzero(ndopts, sizeof(*ndopts)); 450 return -1; 451 } 452 453 if (nd_opt == NULL) 454 goto skip1; 455 456 switch (nd_opt->nd_opt_type) { 457 case ND_OPT_SOURCE_LINKADDR: 458 case ND_OPT_TARGET_LINKADDR: 459 case ND_OPT_MTU: 460 case ND_OPT_REDIRECTED_HEADER: 461 case ND_OPT_NONCE: 462 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 463 nd6log((LOG_INFO, 464 "duplicated ND6 option found (type=%d)\n", 465 nd_opt->nd_opt_type)); 466 /* XXX bark? */ 467 } else { 468 ndopts->nd_opt_array[nd_opt->nd_opt_type] 469 = nd_opt; 470 } 471 break; 472 case ND_OPT_PREFIX_INFORMATION: 473 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 474 ndopts->nd_opt_array[nd_opt->nd_opt_type] 475 = nd_opt; 476 } 477 ndopts->nd_opts_pi_end = 478 (struct nd_opt_prefix_info *)nd_opt; 479 break; 480 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 481 case ND_OPT_RDNSS: /* RFC 6106 */ 482 case ND_OPT_DNSSL: /* RFC 6106 */ 483 /* 484 * Silently ignore options we know and do not care about 485 * in the kernel. 486 */ 487 break; 488 default: 489 /* 490 * Unknown options must be silently ignored, 491 * to accommodate future extension to the protocol. 492 */ 493 nd6log((LOG_DEBUG, 494 "nd6_options: unsupported option %d - " 495 "option ignored\n", nd_opt->nd_opt_type)); 496 } 497 498 skip1: 499 i++; 500 if (i > V_nd6_maxndopt) { 501 ICMP6STAT_INC(icp6s_nd_toomanyopt); 502 nd6log((LOG_INFO, "too many loop in nd opt\n")); 503 break; 504 } 505 506 if (ndopts->nd_opts_done) 507 break; 508 } 509 510 return 0; 511 } 512 513 /* 514 * ND6 timer routine to handle ND6 entries 515 */ 516 static void 517 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 518 { 519 int canceled; 520 521 LLE_WLOCK_ASSERT(ln); 522 523 if (tick < 0) { 524 ln->la_expire = 0; 525 ln->ln_ntick = 0; 526 canceled = callout_stop(&ln->lle_timer); 527 } else { 528 ln->la_expire = time_uptime + tick / hz; 529 LLE_ADDREF(ln); 530 if (tick > INT_MAX) { 531 ln->ln_ntick = tick - INT_MAX; 532 canceled = callout_reset(&ln->lle_timer, INT_MAX, 533 nd6_llinfo_timer, ln); 534 } else { 535 ln->ln_ntick = 0; 536 canceled = callout_reset(&ln->lle_timer, tick, 537 nd6_llinfo_timer, ln); 538 } 539 } 540 if (canceled > 0) 541 LLE_REMREF(ln); 542 } 543 544 /* 545 * Gets source address of the first packet in hold queue 546 * and stores it in @src. 547 * Returns pointer to @src (if hold queue is not empty) or NULL. 548 * 549 * Set noinline to be dtrace-friendly 550 */ 551 static __noinline struct in6_addr * 552 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 553 { 554 struct ip6_hdr hdr; 555 struct mbuf *m; 556 557 if (ln->la_hold == NULL) 558 return (NULL); 559 560 /* 561 * assume every packet in la_hold has the same IP header 562 */ 563 m = ln->la_hold; 564 if (sizeof(hdr) > m->m_len) 565 return (NULL); 566 567 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 568 *src = hdr.ip6_src; 569 570 return (src); 571 } 572 573 /* 574 * Checks if we need to switch from STALE state. 575 * 576 * RFC 4861 requires switching from STALE to DELAY state 577 * on first packet matching entry, waiting V_nd6_delay and 578 * transition to PROBE state (if upper layer confirmation was 579 * not received). 580 * 581 * This code performs a bit differently: 582 * On packet hit we don't change state (but desired state 583 * can be guessed by control plane). However, after V_nd6_delay 584 * seconds code will transition to PROBE state (so DELAY state 585 * is kinda skipped in most situations). 586 * 587 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so 588 * we perform the following upon entering STALE state: 589 * 590 * 1) Arm timer to run each V_nd6_delay seconds to make sure that 591 * if packet was transmitted at the start of given interval, we 592 * would be able to switch to PROBE state in V_nd6_delay seconds 593 * as user expects. 594 * 595 * 2) Reschedule timer until original V_nd6_gctimer expires keeping 596 * lle in STALE state (remaining timer value stored in lle_remtime). 597 * 598 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay 599 * seconds ago. 600 * 601 * Returns non-zero value if the entry is still STALE (storing 602 * the next timer interval in @pdelay). 603 * 604 * Returns zero value if original timer expired or we need to switch to 605 * PROBE (store that in @do_switch variable). 606 */ 607 static int 608 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) 609 { 610 int nd_delay, nd_gctimer, r_skip_req; 611 time_t lle_hittime; 612 long delay; 613 614 *do_switch = 0; 615 nd_gctimer = V_nd6_gctimer; 616 nd_delay = V_nd6_delay; 617 618 LLE_REQ_LOCK(lle); 619 r_skip_req = lle->r_skip_req; 620 lle_hittime = lle->lle_hittime; 621 LLE_REQ_UNLOCK(lle); 622 623 if (r_skip_req > 0) { 624 /* 625 * Nonzero r_skip_req value was set upon entering 626 * STALE state. Since value was not changed, no 627 * packets were passed using this lle. Ask for 628 * timer reschedule and keep STALE state. 629 */ 630 delay = (long)(MIN(nd_gctimer, nd_delay)); 631 delay *= hz; 632 if (lle->lle_remtime > delay) 633 lle->lle_remtime -= delay; 634 else { 635 delay = lle->lle_remtime; 636 lle->lle_remtime = 0; 637 } 638 639 if (delay == 0) { 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 * V_nd6_delay still not passed since the first 658 * hit in STALE state. 659 * Reshedule timer and return. 660 */ 661 *pdelay = (long)(nd_delay - delay) * hz; 662 return (1); 663 } 664 665 /* Request switching to probe */ 666 *do_switch = 1; 667 return (0); 668 } 669 670 /* 671 * Switch @lle state to new state optionally arming timers. 672 * 673 * Set noinline to be dtrace-friendly 674 */ 675 __noinline void 676 nd6_llinfo_setstate(struct llentry *lle, int newstate) 677 { 678 struct ifnet *ifp; 679 int nd_gctimer, nd_delay; 680 long delay, remtime; 681 682 delay = 0; 683 remtime = 0; 684 685 switch (newstate) { 686 case ND6_LLINFO_INCOMPLETE: 687 ifp = lle->lle_tbl->llt_ifp; 688 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; 689 break; 690 case ND6_LLINFO_REACHABLE: 691 if (!ND6_LLINFO_PERMANENT(lle)) { 692 ifp = lle->lle_tbl->llt_ifp; 693 delay = (long)ND_IFINFO(ifp)->reachable * hz; 694 } 695 break; 696 case ND6_LLINFO_STALE: 697 698 /* 699 * Notify fast path that we want to know if any packet 700 * is transmitted by setting r_skip_req. 701 */ 702 LLE_REQ_LOCK(lle); 703 lle->r_skip_req = 1; 704 LLE_REQ_UNLOCK(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. In the case 1312 * where RADIX_MPATH is enabled, multiple route 1313 * entries (of the same rt_key value) will be 1314 * installed because the interface addresses all 1315 * differ. 1316 */ 1317 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1318 &rt_key.sin6_addr)) 1319 continue; 1320 } 1321 1322 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1323 &addr->sin6_addr, &pr->ndpr_mask)) { 1324 ND6_RUNLOCK(); 1325 return (1); 1326 } 1327 } 1328 ND6_RUNLOCK(); 1329 1330 /* 1331 * If the address is assigned on the node of the other side of 1332 * a p2p interface, the address should be a neighbor. 1333 */ 1334 if (ifp->if_flags & IFF_POINTOPOINT) { 1335 struct epoch_tracker et; 1336 1337 NET_EPOCH_ENTER(et); 1338 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1339 if (ifa->ifa_addr->sa_family != addr->sin6_family) 1340 continue; 1341 if (ifa->ifa_dstaddr != NULL && 1342 sa_equal(addr, ifa->ifa_dstaddr)) { 1343 NET_EPOCH_EXIT(et); 1344 return 1; 1345 } 1346 } 1347 NET_EPOCH_EXIT(et); 1348 } 1349 1350 /* 1351 * If the default router list is empty, all addresses are regarded 1352 * as on-link, and thus, as a neighbor. 1353 */ 1354 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1355 nd6_defrouter_list_empty() && 1356 V_nd6_defifindex == ifp->if_index) { 1357 return (1); 1358 } 1359 1360 return (0); 1361 } 1362 1363 /* 1364 * Detect if a given IPv6 address identifies a neighbor on a given link. 1365 * XXX: should take care of the destination of a p2p link? 1366 */ 1367 int 1368 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1369 { 1370 struct llentry *lle; 1371 int rc = 0; 1372 1373 NET_EPOCH_ASSERT(); 1374 IF_AFDATA_UNLOCK_ASSERT(ifp); 1375 if (nd6_is_new_addr_neighbor(addr, ifp)) 1376 return (1); 1377 1378 /* 1379 * Even if the address matches none of our addresses, it might be 1380 * in the neighbor cache. 1381 */ 1382 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1383 LLE_RUNLOCK(lle); 1384 rc = 1; 1385 } 1386 return (rc); 1387 } 1388 1389 /* 1390 * Free an nd6 llinfo entry. 1391 * Since the function would cause significant changes in the kernel, DO NOT 1392 * make it global, unless you have a strong reason for the change, and are sure 1393 * that the change is safe. 1394 * 1395 * Set noinline to be dtrace-friendly 1396 */ 1397 static __noinline void 1398 nd6_free(struct llentry **lnp, int gc) 1399 { 1400 struct ifnet *ifp; 1401 struct llentry *ln; 1402 struct nd_defrouter *dr; 1403 1404 ln = *lnp; 1405 *lnp = NULL; 1406 1407 LLE_WLOCK_ASSERT(ln); 1408 ND6_RLOCK_ASSERT(); 1409 1410 ifp = lltable_get_ifp(ln->lle_tbl); 1411 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0) 1412 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp); 1413 else 1414 dr = NULL; 1415 ND6_RUNLOCK(); 1416 1417 if ((ln->la_flags & LLE_DELETED) == 0) 1418 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 1419 1420 /* 1421 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1422 * even though it is not harmful, it was not really necessary. 1423 */ 1424 1425 /* cancel timer */ 1426 nd6_llinfo_settimer_locked(ln, -1); 1427 1428 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1429 if (dr != NULL && dr->expire && 1430 ln->ln_state == ND6_LLINFO_STALE && gc) { 1431 /* 1432 * If the reason for the deletion is just garbage 1433 * collection, and the neighbor is an active default 1434 * router, do not delete it. Instead, reset the GC 1435 * timer using the router's lifetime. 1436 * Simply deleting the entry would affect default 1437 * router selection, which is not necessarily a good 1438 * thing, especially when we're using router preference 1439 * values. 1440 * XXX: the check for ln_state would be redundant, 1441 * but we intentionally keep it just in case. 1442 */ 1443 if (dr->expire > time_uptime) 1444 nd6_llinfo_settimer_locked(ln, 1445 (dr->expire - time_uptime) * hz); 1446 else 1447 nd6_llinfo_settimer_locked(ln, 1448 (long)V_nd6_gctimer * hz); 1449 1450 LLE_REMREF(ln); 1451 LLE_WUNLOCK(ln); 1452 defrouter_rele(dr); 1453 return; 1454 } 1455 1456 if (dr) { 1457 /* 1458 * Unreachablity of a router might affect the default 1459 * router selection and on-link detection of advertised 1460 * prefixes. 1461 */ 1462 1463 /* 1464 * Temporarily fake the state to choose a new default 1465 * router and to perform on-link determination of 1466 * prefixes correctly. 1467 * Below the state will be set correctly, 1468 * or the entry itself will be deleted. 1469 */ 1470 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1471 } 1472 1473 if (ln->ln_router || dr) { 1474 /* 1475 * We need to unlock to avoid a LOR with rt6_flush() with the 1476 * rnh and for the calls to pfxlist_onlink_check() and 1477 * defrouter_select_fib() in the block further down for calls 1478 * into nd6_lookup(). We still hold a ref. 1479 */ 1480 LLE_WUNLOCK(ln); 1481 1482 /* 1483 * rt6_flush must be called whether or not the neighbor 1484 * is in the Default Router List. 1485 * See a corresponding comment in nd6_na_input(). 1486 */ 1487 rt6_flush(&ln->r_l3addr.addr6, ifp); 1488 } 1489 1490 if (dr) { 1491 /* 1492 * Since defrouter_select_fib() does not affect the 1493 * on-link determination and MIP6 needs the check 1494 * before the default router selection, we perform 1495 * the check now. 1496 */ 1497 pfxlist_onlink_check(); 1498 1499 /* 1500 * Refresh default router list. 1501 */ 1502 defrouter_select_fib(dr->ifp->if_fib); 1503 } 1504 1505 /* 1506 * If this entry was added by an on-link redirect, remove the 1507 * corresponding host route. 1508 */ 1509 if (ln->la_flags & LLE_REDIRECT) 1510 nd6_free_redirect(ln); 1511 1512 if (ln->ln_router || dr) 1513 LLE_WLOCK(ln); 1514 } 1515 1516 /* 1517 * Save to unlock. We still hold an extra reference and will not 1518 * free(9) in llentry_free() if someone else holds one as well. 1519 */ 1520 LLE_WUNLOCK(ln); 1521 IF_AFDATA_LOCK(ifp); 1522 LLE_WLOCK(ln); 1523 /* Guard against race with other llentry_free(). */ 1524 if (ln->la_flags & LLE_LINKED) { 1525 /* Remove callout reference */ 1526 LLE_REMREF(ln); 1527 lltable_unlink_entry(ln->lle_tbl, ln); 1528 } 1529 IF_AFDATA_UNLOCK(ifp); 1530 1531 llentry_free(ln); 1532 if (dr != NULL) 1533 defrouter_rele(dr); 1534 } 1535 1536 static int 1537 nd6_isdynrte(const struct rtentry *rt, const struct nhop_object *nh, void *xap) 1538 { 1539 1540 if (nh->nh_flags & NHF_REDIRECT) 1541 return (1); 1542 1543 return (0); 1544 } 1545 1546 /* 1547 * Remove the rtentry for the given llentry, 1548 * both of which were installed by a redirect. 1549 */ 1550 static void 1551 nd6_free_redirect(const struct llentry *ln) 1552 { 1553 int fibnum; 1554 struct sockaddr_in6 sin6; 1555 struct rt_addrinfo info; 1556 struct rib_cmd_info rc; 1557 struct epoch_tracker et; 1558 1559 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); 1560 memset(&info, 0, sizeof(info)); 1561 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; 1562 info.rti_filter = nd6_isdynrte; 1563 1564 NET_EPOCH_ENTER(et); 1565 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 1566 rib_action(fibnum, RTM_DELETE, &info, &rc); 1567 NET_EPOCH_EXIT(et); 1568 } 1569 1570 /* 1571 * Updates status of the default router route. 1572 */ 1573 void 1574 nd6_subscription_cb(struct rib_head *rnh, struct rib_cmd_info *rc, void *arg) 1575 { 1576 struct nd_defrouter *dr; 1577 struct nhop_object *nh; 1578 1579 if (rc->rc_cmd == RTM_DELETE) { 1580 nh = rc->rc_nh_old; 1581 1582 if (nh->nh_flags & NHF_DEFAULT) { 1583 dr = defrouter_lookup(&nh->gw6_sa.sin6_addr, nh->nh_ifp); 1584 if (dr != NULL) { 1585 dr->installed = 0; 1586 defrouter_rele(dr); 1587 } 1588 } 1589 } 1590 } 1591 1592 int 1593 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1594 { 1595 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1596 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1597 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1598 struct epoch_tracker et; 1599 int error = 0; 1600 1601 if (ifp->if_afdata[AF_INET6] == NULL) 1602 return (EPFNOSUPPORT); 1603 switch (cmd) { 1604 case OSIOCGIFINFO_IN6: 1605 #define ND ndi->ndi 1606 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1607 bzero(&ND, sizeof(ND)); 1608 ND.linkmtu = IN6_LINKMTU(ifp); 1609 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1610 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1611 ND.reachable = ND_IFINFO(ifp)->reachable; 1612 ND.retrans = ND_IFINFO(ifp)->retrans; 1613 ND.flags = ND_IFINFO(ifp)->flags; 1614 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1615 ND.chlim = ND_IFINFO(ifp)->chlim; 1616 break; 1617 case SIOCGIFINFO_IN6: 1618 ND = *ND_IFINFO(ifp); 1619 break; 1620 case SIOCSIFINFO_IN6: 1621 /* 1622 * used to change host variables from userland. 1623 * intended for a use on router to reflect RA configurations. 1624 */ 1625 /* 0 means 'unspecified' */ 1626 if (ND.linkmtu != 0) { 1627 if (ND.linkmtu < IPV6_MMTU || 1628 ND.linkmtu > IN6_LINKMTU(ifp)) { 1629 error = EINVAL; 1630 break; 1631 } 1632 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1633 } 1634 1635 if (ND.basereachable != 0) { 1636 int obasereachable = ND_IFINFO(ifp)->basereachable; 1637 1638 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1639 if (ND.basereachable != obasereachable) 1640 ND_IFINFO(ifp)->reachable = 1641 ND_COMPUTE_RTIME(ND.basereachable); 1642 } 1643 if (ND.retrans != 0) 1644 ND_IFINFO(ifp)->retrans = ND.retrans; 1645 if (ND.chlim != 0) 1646 ND_IFINFO(ifp)->chlim = ND.chlim; 1647 /* FALLTHROUGH */ 1648 case SIOCSIFINFO_FLAGS: 1649 { 1650 struct ifaddr *ifa; 1651 struct in6_ifaddr *ia; 1652 1653 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1654 !(ND.flags & ND6_IFF_IFDISABLED)) { 1655 /* ifdisabled 1->0 transision */ 1656 1657 /* 1658 * If the interface is marked as ND6_IFF_IFDISABLED and 1659 * has an link-local address with IN6_IFF_DUPLICATED, 1660 * do not clear ND6_IFF_IFDISABLED. 1661 * See RFC 4862, Section 5.4.5. 1662 */ 1663 NET_EPOCH_ENTER(et); 1664 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1665 if (ifa->ifa_addr->sa_family != AF_INET6) 1666 continue; 1667 ia = (struct in6_ifaddr *)ifa; 1668 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1669 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1670 break; 1671 } 1672 NET_EPOCH_EXIT(et); 1673 1674 if (ifa != NULL) { 1675 /* LLA is duplicated. */ 1676 ND.flags |= ND6_IFF_IFDISABLED; 1677 log(LOG_ERR, "Cannot enable an interface" 1678 " with a link-local address marked" 1679 " duplicate.\n"); 1680 } else { 1681 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1682 if (ifp->if_flags & IFF_UP) 1683 in6_if_up(ifp); 1684 } 1685 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1686 (ND.flags & ND6_IFF_IFDISABLED)) { 1687 /* ifdisabled 0->1 transision */ 1688 /* Mark all IPv6 address as tentative. */ 1689 1690 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1691 if (V_ip6_dad_count > 0 && 1692 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1693 NET_EPOCH_ENTER(et); 1694 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1695 ifa_link) { 1696 if (ifa->ifa_addr->sa_family != 1697 AF_INET6) 1698 continue; 1699 ia = (struct in6_ifaddr *)ifa; 1700 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1701 } 1702 NET_EPOCH_EXIT(et); 1703 } 1704 } 1705 1706 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1707 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1708 /* auto_linklocal 0->1 transision */ 1709 1710 /* If no link-local address on ifp, configure */ 1711 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1712 in6_ifattach(ifp, NULL); 1713 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1714 ifp->if_flags & IFF_UP) { 1715 /* 1716 * When the IF already has 1717 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1718 * address is assigned, and IFF_UP, try to 1719 * assign one. 1720 */ 1721 NET_EPOCH_ENTER(et); 1722 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1723 ifa_link) { 1724 if (ifa->ifa_addr->sa_family != 1725 AF_INET6) 1726 continue; 1727 ia = (struct in6_ifaddr *)ifa; 1728 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1729 break; 1730 } 1731 NET_EPOCH_EXIT(et); 1732 if (ifa != NULL) 1733 /* No LLA is configured. */ 1734 in6_ifattach(ifp, NULL); 1735 } 1736 } 1737 ND_IFINFO(ifp)->flags = ND.flags; 1738 break; 1739 } 1740 #undef ND 1741 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1742 /* sync kernel routing table with the default router list */ 1743 defrouter_reset(); 1744 defrouter_select_fib(RT_ALL_FIBS); 1745 break; 1746 case SIOCSPFXFLUSH_IN6: 1747 { 1748 /* flush all the prefix advertised by routers */ 1749 struct in6_ifaddr *ia, *ia_next; 1750 struct nd_prefix *pr, *next; 1751 struct nd_prhead prl; 1752 1753 LIST_INIT(&prl); 1754 1755 ND6_WLOCK(); 1756 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1757 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1758 continue; /* XXX */ 1759 nd6_prefix_unlink(pr, &prl); 1760 } 1761 ND6_WUNLOCK(); 1762 1763 while ((pr = LIST_FIRST(&prl)) != NULL) { 1764 LIST_REMOVE(pr, ndpr_entry); 1765 /* XXXRW: in6_ifaddrhead locking. */ 1766 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1767 ia_next) { 1768 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1769 continue; 1770 1771 if (ia->ia6_ndpr == pr) 1772 in6_purgeaddr(&ia->ia_ifa); 1773 } 1774 nd6_prefix_del(pr); 1775 } 1776 break; 1777 } 1778 case SIOCSRTRFLUSH_IN6: 1779 { 1780 /* flush all the default routers */ 1781 1782 defrouter_reset(); 1783 nd6_defrouter_flush_all(); 1784 defrouter_select_fib(RT_ALL_FIBS); 1785 break; 1786 } 1787 case SIOCGNBRINFO_IN6: 1788 { 1789 struct llentry *ln; 1790 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1791 1792 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1793 return (error); 1794 1795 NET_EPOCH_ENTER(et); 1796 ln = nd6_lookup(&nb_addr, 0, ifp); 1797 NET_EPOCH_EXIT(et); 1798 1799 if (ln == NULL) { 1800 error = EINVAL; 1801 break; 1802 } 1803 nbi->state = ln->ln_state; 1804 nbi->asked = ln->la_asked; 1805 nbi->isrouter = ln->ln_router; 1806 if (ln->la_expire == 0) 1807 nbi->expire = 0; 1808 else 1809 nbi->expire = ln->la_expire + ln->lle_remtime / hz + 1810 (time_second - time_uptime); 1811 LLE_RUNLOCK(ln); 1812 break; 1813 } 1814 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1815 ndif->ifindex = V_nd6_defifindex; 1816 break; 1817 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1818 return (nd6_setdefaultiface(ndif->ifindex)); 1819 } 1820 return (error); 1821 } 1822 1823 /* 1824 * Calculates new isRouter value based on provided parameters and 1825 * returns it. 1826 */ 1827 static int 1828 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, 1829 int ln_router) 1830 { 1831 1832 /* 1833 * ICMP6 type dependent behavior. 1834 * 1835 * NS: clear IsRouter if new entry 1836 * RS: clear IsRouter 1837 * RA: set IsRouter if there's lladdr 1838 * redir: clear IsRouter if new entry 1839 * 1840 * RA case, (1): 1841 * The spec says that we must set IsRouter in the following cases: 1842 * - If lladdr exist, set IsRouter. This means (1-5). 1843 * - If it is old entry (!newentry), set IsRouter. This means (7). 1844 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1845 * A quetion arises for (1) case. (1) case has no lladdr in the 1846 * neighbor cache, this is similar to (6). 1847 * This case is rare but we figured that we MUST NOT set IsRouter. 1848 * 1849 * is_new old_addr new_addr NS RS RA redir 1850 * D R 1851 * 0 n n (1) c ? s 1852 * 0 y n (2) c s s 1853 * 0 n y (3) c s s 1854 * 0 y y (4) c s s 1855 * 0 y y (5) c s s 1856 * 1 -- n (6) c c c s 1857 * 1 -- y (7) c c s c s 1858 * 1859 * (c=clear s=set) 1860 */ 1861 switch (type & 0xff) { 1862 case ND_NEIGHBOR_SOLICIT: 1863 /* 1864 * New entry must have is_router flag cleared. 1865 */ 1866 if (is_new) /* (6-7) */ 1867 ln_router = 0; 1868 break; 1869 case ND_REDIRECT: 1870 /* 1871 * If the icmp is a redirect to a better router, always set the 1872 * is_router flag. Otherwise, if the entry is newly created, 1873 * clear the flag. [RFC 2461, sec 8.3] 1874 */ 1875 if (code == ND_REDIRECT_ROUTER) 1876 ln_router = 1; 1877 else { 1878 if (is_new) /* (6-7) */ 1879 ln_router = 0; 1880 } 1881 break; 1882 case ND_ROUTER_SOLICIT: 1883 /* 1884 * is_router flag must always be cleared. 1885 */ 1886 ln_router = 0; 1887 break; 1888 case ND_ROUTER_ADVERT: 1889 /* 1890 * Mark an entry with lladdr as a router. 1891 */ 1892 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ 1893 (is_new && new_addr)) { /* (7) */ 1894 ln_router = 1; 1895 } 1896 break; 1897 } 1898 1899 return (ln_router); 1900 } 1901 1902 /* 1903 * Create neighbor cache entry and cache link-layer address, 1904 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1905 * 1906 * type - ICMP6 type 1907 * code - type dependent information 1908 * 1909 */ 1910 void 1911 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1912 int lladdrlen, int type, int code) 1913 { 1914 struct llentry *ln = NULL, *ln_tmp; 1915 int is_newentry; 1916 int do_update; 1917 int olladdr; 1918 int llchange; 1919 int flags; 1920 uint16_t router = 0; 1921 struct sockaddr_in6 sin6; 1922 struct mbuf *chain = NULL; 1923 u_char linkhdr[LLE_MAX_LINKHDR]; 1924 size_t linkhdrsize; 1925 int lladdr_off; 1926 1927 NET_EPOCH_ASSERT(); 1928 IF_AFDATA_UNLOCK_ASSERT(ifp); 1929 1930 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1931 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1932 1933 /* nothing must be updated for unspecified address */ 1934 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1935 return; 1936 1937 /* 1938 * Validation about ifp->if_addrlen and lladdrlen must be done in 1939 * the caller. 1940 * 1941 * XXX If the link does not have link-layer adderss, what should 1942 * we do? (ifp->if_addrlen == 0) 1943 * Spec says nothing in sections for RA, RS and NA. There's small 1944 * description on it in NS section (RFC 2461 7.2.3). 1945 */ 1946 flags = lladdr ? LLE_EXCLUSIVE : 0; 1947 ln = nd6_lookup(from, flags, ifp); 1948 is_newentry = 0; 1949 if (ln == NULL) { 1950 flags |= LLE_EXCLUSIVE; 1951 ln = nd6_alloc(from, 0, ifp); 1952 if (ln == NULL) 1953 return; 1954 1955 /* 1956 * Since we already know all the data for the new entry, 1957 * fill it before insertion. 1958 */ 1959 if (lladdr != NULL) { 1960 linkhdrsize = sizeof(linkhdr); 1961 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 1962 linkhdr, &linkhdrsize, &lladdr_off) != 0) 1963 return; 1964 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 1965 lladdr_off); 1966 } 1967 1968 IF_AFDATA_WLOCK(ifp); 1969 LLE_WLOCK(ln); 1970 /* Prefer any existing lle over newly-created one */ 1971 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); 1972 if (ln_tmp == NULL) 1973 lltable_link_entry(LLTABLE6(ifp), ln); 1974 IF_AFDATA_WUNLOCK(ifp); 1975 if (ln_tmp == NULL) { 1976 /* No existing lle, mark as new entry (6,7) */ 1977 is_newentry = 1; 1978 if (lladdr != NULL) { /* (7) */ 1979 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 1980 EVENTHANDLER_INVOKE(lle_event, ln, 1981 LLENTRY_RESOLVED); 1982 } 1983 } else { 1984 lltable_free_entry(LLTABLE6(ifp), ln); 1985 ln = ln_tmp; 1986 ln_tmp = NULL; 1987 } 1988 } 1989 /* do nothing if static ndp is set */ 1990 if ((ln->la_flags & LLE_STATIC)) { 1991 if (flags & LLE_EXCLUSIVE) 1992 LLE_WUNLOCK(ln); 1993 else 1994 LLE_RUNLOCK(ln); 1995 return; 1996 } 1997 1998 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1999 if (olladdr && lladdr) { 2000 llchange = bcmp(lladdr, ln->ll_addr, 2001 ifp->if_addrlen); 2002 } else if (!olladdr && lladdr) 2003 llchange = 1; 2004 else 2005 llchange = 0; 2006 2007 /* 2008 * newentry olladdr lladdr llchange (*=record) 2009 * 0 n n -- (1) 2010 * 0 y n -- (2) 2011 * 0 n y y (3) * STALE 2012 * 0 y y n (4) * 2013 * 0 y y y (5) * STALE 2014 * 1 -- n -- (6) NOSTATE(= PASSIVE) 2015 * 1 -- y -- (7) * STALE 2016 */ 2017 2018 do_update = 0; 2019 if (is_newentry == 0 && llchange != 0) { 2020 do_update = 1; /* (3,5) */ 2021 2022 /* 2023 * Record source link-layer address 2024 * XXX is it dependent to ifp->if_type? 2025 */ 2026 linkhdrsize = sizeof(linkhdr); 2027 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 2028 linkhdr, &linkhdrsize, &lladdr_off) != 0) 2029 return; 2030 2031 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 2032 lladdr_off) == 0) { 2033 /* Entry was deleted */ 2034 return; 2035 } 2036 2037 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2038 2039 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2040 2041 if (ln->la_hold != NULL) 2042 nd6_grab_holdchain(ln, &chain, &sin6); 2043 } 2044 2045 /* Calculates new router status */ 2046 router = nd6_is_router(type, code, is_newentry, olladdr, 2047 lladdr != NULL ? 1 : 0, ln->ln_router); 2048 2049 ln->ln_router = router; 2050 /* Mark non-router redirects with special flag */ 2051 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) 2052 ln->la_flags |= LLE_REDIRECT; 2053 2054 if (flags & LLE_EXCLUSIVE) 2055 LLE_WUNLOCK(ln); 2056 else 2057 LLE_RUNLOCK(ln); 2058 2059 if (chain != NULL) 2060 nd6_flush_holdchain(ifp, chain, &sin6); 2061 2062 /* 2063 * When the link-layer address of a router changes, select the 2064 * best router again. In particular, when the neighbor entry is newly 2065 * created, it might affect the selection policy. 2066 * Question: can we restrict the first condition to the "is_newentry" 2067 * case? 2068 * XXX: when we hear an RA from a new router with the link-layer 2069 * address option, defrouter_select_fib() is called twice, since 2070 * defrtrlist_update called the function as well. However, I believe 2071 * we can compromise the overhead, since it only happens the first 2072 * time. 2073 * XXX: although defrouter_select_fib() should not have a bad effect 2074 * for those are not autoconfigured hosts, we explicitly avoid such 2075 * cases for safety. 2076 */ 2077 if ((do_update || is_newentry) && router && 2078 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 2079 /* 2080 * guaranteed recursion 2081 */ 2082 defrouter_select_fib(ifp->if_fib); 2083 } 2084 } 2085 2086 static void 2087 nd6_slowtimo(void *arg) 2088 { 2089 struct epoch_tracker et; 2090 CURVNET_SET((struct vnet *) arg); 2091 struct nd_ifinfo *nd6if; 2092 struct ifnet *ifp; 2093 2094 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 2095 nd6_slowtimo, curvnet); 2096 NET_EPOCH_ENTER(et); 2097 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2098 if (ifp->if_afdata[AF_INET6] == NULL) 2099 continue; 2100 nd6if = ND_IFINFO(ifp); 2101 if (nd6if->basereachable && /* already initialized */ 2102 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 2103 /* 2104 * Since reachable time rarely changes by router 2105 * advertisements, we SHOULD insure that a new random 2106 * value gets recomputed at least once every few hours. 2107 * (RFC 2461, 6.3.4) 2108 */ 2109 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 2110 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 2111 } 2112 } 2113 NET_EPOCH_EXIT(et); 2114 CURVNET_RESTORE(); 2115 } 2116 2117 void 2118 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 2119 struct sockaddr_in6 *sin6) 2120 { 2121 2122 LLE_WLOCK_ASSERT(ln); 2123 2124 *chain = ln->la_hold; 2125 ln->la_hold = NULL; 2126 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 2127 2128 if (ln->ln_state == ND6_LLINFO_STALE) { 2129 /* 2130 * The first time we send a packet to a 2131 * neighbor whose entry is STALE, we have 2132 * to change the state to DELAY and a sets 2133 * a timer to expire in DELAY_FIRST_PROBE_TIME 2134 * seconds to ensure do neighbor unreachability 2135 * detection on expiration. 2136 * (RFC 2461 7.3.3) 2137 */ 2138 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); 2139 } 2140 } 2141 2142 int 2143 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2144 struct sockaddr_in6 *dst, struct route *ro) 2145 { 2146 int error; 2147 int ip6len; 2148 struct ip6_hdr *ip6; 2149 struct m_tag *mtag; 2150 2151 #ifdef MAC 2152 mac_netinet6_nd6_send(ifp, m); 2153 #endif 2154 2155 /* 2156 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2157 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2158 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2159 * to be diverted to user space. When re-injected into the kernel, 2160 * send_output() will directly dispatch them to the outgoing interface. 2161 */ 2162 if (send_sendso_input_hook != NULL) { 2163 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2164 if (mtag != NULL) { 2165 ip6 = mtod(m, struct ip6_hdr *); 2166 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2167 /* Use the SEND socket */ 2168 error = send_sendso_input_hook(m, ifp, SND_OUT, 2169 ip6len); 2170 /* -1 == no app on SEND socket */ 2171 if (error == 0 || error != -1) 2172 return (error); 2173 } 2174 } 2175 2176 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 2177 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 2178 mtod(m, struct ip6_hdr *)); 2179 2180 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2181 origifp = ifp; 2182 2183 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); 2184 return (error); 2185 } 2186 2187 /* 2188 * Lookup link headerfor @sa_dst address. Stores found 2189 * data in @desten buffer. Copy of lle ln_flags can be also 2190 * saved in @pflags if @pflags is non-NULL. 2191 * 2192 * If destination LLE does not exists or lle state modification 2193 * is required, call "slow" version. 2194 * 2195 * Return values: 2196 * - 0 on success (address copied to buffer). 2197 * - EWOULDBLOCK (no local error, but address is still unresolved) 2198 * - other errors (alloc failure, etc) 2199 */ 2200 int 2201 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 2202 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags, 2203 struct llentry **plle) 2204 { 2205 struct llentry *ln = NULL; 2206 const struct sockaddr_in6 *dst6; 2207 2208 NET_EPOCH_ASSERT(); 2209 2210 if (pflags != NULL) 2211 *pflags = 0; 2212 2213 dst6 = (const struct sockaddr_in6 *)sa_dst; 2214 2215 /* discard the packet if IPv6 operation is disabled on the interface */ 2216 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2217 m_freem(m); 2218 return (ENETDOWN); /* better error? */ 2219 } 2220 2221 if (m != NULL && m->m_flags & M_MCAST) { 2222 switch (ifp->if_type) { 2223 case IFT_ETHER: 2224 case IFT_L2VLAN: 2225 case IFT_BRIDGE: 2226 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, 2227 desten); 2228 return (0); 2229 default: 2230 m_freem(m); 2231 return (EAFNOSUPPORT); 2232 } 2233 } 2234 2235 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, 2236 ifp); 2237 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { 2238 /* Entry found, let's copy lle info */ 2239 bcopy(ln->r_linkdata, desten, ln->r_hdrlen); 2240 if (pflags != NULL) 2241 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); 2242 /* Check if we have feedback request from nd6 timer */ 2243 if (ln->r_skip_req != 0) { 2244 LLE_REQ_LOCK(ln); 2245 ln->r_skip_req = 0; /* Notify that entry was used */ 2246 ln->lle_hittime = time_uptime; 2247 LLE_REQ_UNLOCK(ln); 2248 } 2249 if (plle) { 2250 LLE_ADDREF(ln); 2251 *plle = ln; 2252 LLE_WUNLOCK(ln); 2253 } 2254 return (0); 2255 } else if (plle && ln) 2256 LLE_WUNLOCK(ln); 2257 2258 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle)); 2259 } 2260 2261 /* 2262 * Do L2 address resolution for @sa_dst address. Stores found 2263 * address in @desten buffer. Copy of lle ln_flags can be also 2264 * saved in @pflags if @pflags is non-NULL. 2265 * 2266 * Heavy version. 2267 * Function assume that destination LLE does not exist, 2268 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. 2269 * 2270 * Set noinline to be dtrace-friendly 2271 */ 2272 static __noinline int 2273 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m, 2274 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags, 2275 struct llentry **plle) 2276 { 2277 struct llentry *lle = NULL, *lle_tmp; 2278 struct in6_addr *psrc, src; 2279 int send_ns, ll_len; 2280 char *lladdr; 2281 2282 NET_EPOCH_ASSERT(); 2283 2284 /* 2285 * Address resolution or Neighbor Unreachability Detection 2286 * for the next hop. 2287 * At this point, the destination of the packet must be a unicast 2288 * or an anycast address(i.e. not a multicast). 2289 */ 2290 if (lle == NULL) { 2291 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2292 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2293 /* 2294 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2295 * the condition below is not very efficient. But we believe 2296 * it is tolerable, because this should be a rare case. 2297 */ 2298 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2299 if (lle == NULL) { 2300 char ip6buf[INET6_ADDRSTRLEN]; 2301 log(LOG_DEBUG, 2302 "nd6_output: can't allocate llinfo for %s " 2303 "(ln=%p)\n", 2304 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2305 m_freem(m); 2306 return (ENOBUFS); 2307 } 2308 2309 IF_AFDATA_WLOCK(ifp); 2310 LLE_WLOCK(lle); 2311 /* Prefer any existing entry over newly-created one */ 2312 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2313 if (lle_tmp == NULL) 2314 lltable_link_entry(LLTABLE6(ifp), lle); 2315 IF_AFDATA_WUNLOCK(ifp); 2316 if (lle_tmp != NULL) { 2317 lltable_free_entry(LLTABLE6(ifp), lle); 2318 lle = lle_tmp; 2319 lle_tmp = NULL; 2320 } 2321 } 2322 } 2323 if (lle == NULL) { 2324 m_freem(m); 2325 return (ENOBUFS); 2326 } 2327 2328 LLE_WLOCK_ASSERT(lle); 2329 2330 /* 2331 * The first time we send a packet to a neighbor whose entry is 2332 * STALE, we have to change the state to DELAY and a sets a timer to 2333 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2334 * neighbor unreachability detection on expiration. 2335 * (RFC 2461 7.3.3) 2336 */ 2337 if (lle->ln_state == ND6_LLINFO_STALE) 2338 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); 2339 2340 /* 2341 * If the neighbor cache entry has a state other than INCOMPLETE 2342 * (i.e. its link-layer address is already resolved), just 2343 * send the packet. 2344 */ 2345 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { 2346 if (flags & LLE_ADDRONLY) { 2347 lladdr = lle->ll_addr; 2348 ll_len = ifp->if_addrlen; 2349 } else { 2350 lladdr = lle->r_linkdata; 2351 ll_len = lle->r_hdrlen; 2352 } 2353 bcopy(lladdr, desten, ll_len); 2354 if (pflags != NULL) 2355 *pflags = lle->la_flags; 2356 if (plle) { 2357 LLE_ADDREF(lle); 2358 *plle = lle; 2359 } 2360 LLE_WUNLOCK(lle); 2361 return (0); 2362 } 2363 2364 /* 2365 * There is a neighbor cache entry, but no ethernet address 2366 * response yet. Append this latest packet to the end of the 2367 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen, 2368 * the oldest packet in the queue will be removed. 2369 */ 2370 2371 if (lle->la_hold != NULL) { 2372 struct mbuf *m_hold; 2373 int i; 2374 2375 i = 0; 2376 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2377 i++; 2378 if (m_hold->m_nextpkt == NULL) { 2379 m_hold->m_nextpkt = m; 2380 break; 2381 } 2382 } 2383 while (i >= V_nd6_maxqueuelen) { 2384 m_hold = lle->la_hold; 2385 lle->la_hold = lle->la_hold->m_nextpkt; 2386 m_freem(m_hold); 2387 i--; 2388 } 2389 } else { 2390 lle->la_hold = m; 2391 } 2392 2393 /* 2394 * If there has been no NS for the neighbor after entering the 2395 * INCOMPLETE state, send the first solicitation. 2396 * Note that for newly-created lle la_asked will be 0, 2397 * so we will transition from ND6_LLINFO_NOSTATE to 2398 * ND6_LLINFO_INCOMPLETE state here. 2399 */ 2400 psrc = NULL; 2401 send_ns = 0; 2402 if (lle->la_asked == 0) { 2403 lle->la_asked++; 2404 send_ns = 1; 2405 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2406 2407 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); 2408 } 2409 LLE_WUNLOCK(lle); 2410 if (send_ns != 0) 2411 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2412 2413 return (EWOULDBLOCK); 2414 } 2415 2416 /* 2417 * Do L2 address resolution for @sa_dst address. Stores found 2418 * address in @desten buffer. Copy of lle ln_flags can be also 2419 * saved in @pflags if @pflags is non-NULL. 2420 * 2421 * Return values: 2422 * - 0 on success (address copied to buffer). 2423 * - EWOULDBLOCK (no local error, but address is still unresolved) 2424 * - other errors (alloc failure, etc) 2425 */ 2426 int 2427 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 2428 char *desten, uint32_t *pflags) 2429 { 2430 int error; 2431 2432 flags |= LLE_ADDRONLY; 2433 error = nd6_resolve_slow(ifp, flags, NULL, 2434 (const struct sockaddr_in6 *)dst, desten, pflags, NULL); 2435 return (error); 2436 } 2437 2438 int 2439 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain, 2440 struct sockaddr_in6 *dst) 2441 { 2442 struct mbuf *m, *m_head; 2443 int error = 0; 2444 2445 m_head = chain; 2446 2447 while (m_head) { 2448 m = m_head; 2449 m_head = m_head->m_nextpkt; 2450 error = nd6_output_ifp(ifp, ifp, m, dst, NULL); 2451 } 2452 2453 /* 2454 * XXX 2455 * note that intermediate errors are blindly ignored 2456 */ 2457 return (error); 2458 } 2459 2460 static int 2461 nd6_need_cache(struct ifnet *ifp) 2462 { 2463 /* 2464 * XXX: we currently do not make neighbor cache on any interface 2465 * other than Ethernet and GIF. 2466 * 2467 * RFC2893 says: 2468 * - unidirectional tunnels needs no ND 2469 */ 2470 switch (ifp->if_type) { 2471 case IFT_ETHER: 2472 case IFT_IEEE1394: 2473 case IFT_L2VLAN: 2474 case IFT_INFINIBAND: 2475 case IFT_BRIDGE: 2476 case IFT_PROPVIRTUAL: 2477 return (1); 2478 default: 2479 return (0); 2480 } 2481 } 2482 2483 /* 2484 * Add pernament ND6 link-layer record for given 2485 * interface address. 2486 * 2487 * Very similar to IPv4 arp_ifinit(), but: 2488 * 1) IPv6 DAD is performed in different place 2489 * 2) It is called by IPv6 protocol stack in contrast to 2490 * arp_ifinit() which is typically called in SIOCSIFADDR 2491 * driver ioctl handler. 2492 * 2493 */ 2494 int 2495 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2496 { 2497 struct ifnet *ifp; 2498 struct llentry *ln, *ln_tmp; 2499 struct sockaddr *dst; 2500 2501 ifp = ia->ia_ifa.ifa_ifp; 2502 if (nd6_need_cache(ifp) == 0) 2503 return (0); 2504 2505 dst = (struct sockaddr *)&ia->ia_addr; 2506 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2507 if (ln == NULL) 2508 return (ENOBUFS); 2509 2510 IF_AFDATA_WLOCK(ifp); 2511 LLE_WLOCK(ln); 2512 /* Unlink any entry if exists */ 2513 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2514 if (ln_tmp != NULL) 2515 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2516 lltable_link_entry(LLTABLE6(ifp), ln); 2517 IF_AFDATA_WUNLOCK(ifp); 2518 2519 if (ln_tmp != NULL) 2520 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2521 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2522 2523 LLE_WUNLOCK(ln); 2524 if (ln_tmp != NULL) 2525 llentry_free(ln_tmp); 2526 2527 return (0); 2528 } 2529 2530 /* 2531 * Removes either all lle entries for given @ia, or lle 2532 * corresponding to @ia address. 2533 */ 2534 void 2535 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) 2536 { 2537 struct sockaddr_in6 mask, addr; 2538 struct sockaddr *saddr, *smask; 2539 struct ifnet *ifp; 2540 2541 ifp = ia->ia_ifa.ifa_ifp; 2542 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2543 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2544 saddr = (struct sockaddr *)&addr; 2545 smask = (struct sockaddr *)&mask; 2546 2547 if (all != 0) 2548 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); 2549 else 2550 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); 2551 } 2552 2553 static void 2554 clear_llinfo_pqueue(struct llentry *ln) 2555 { 2556 struct mbuf *m_hold, *m_hold_next; 2557 2558 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2559 m_hold_next = m_hold->m_nextpkt; 2560 m_freem(m_hold); 2561 } 2562 2563 ln->la_hold = NULL; 2564 } 2565 2566 static int 2567 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2568 { 2569 struct in6_prefix p; 2570 struct sockaddr_in6 s6; 2571 struct nd_prefix *pr; 2572 struct nd_pfxrouter *pfr; 2573 time_t maxexpire; 2574 int error; 2575 char ip6buf[INET6_ADDRSTRLEN]; 2576 2577 if (req->newptr) 2578 return (EPERM); 2579 2580 error = sysctl_wire_old_buffer(req, 0); 2581 if (error != 0) 2582 return (error); 2583 2584 bzero(&p, sizeof(p)); 2585 p.origin = PR_ORIG_RA; 2586 bzero(&s6, sizeof(s6)); 2587 s6.sin6_family = AF_INET6; 2588 s6.sin6_len = sizeof(s6); 2589 2590 ND6_RLOCK(); 2591 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2592 p.prefix = pr->ndpr_prefix; 2593 if (sa6_recoverscope(&p.prefix)) { 2594 log(LOG_ERR, "scope error in prefix list (%s)\n", 2595 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2596 /* XXX: press on... */ 2597 } 2598 p.raflags = pr->ndpr_raf; 2599 p.prefixlen = pr->ndpr_plen; 2600 p.vltime = pr->ndpr_vltime; 2601 p.pltime = pr->ndpr_pltime; 2602 p.if_index = pr->ndpr_ifp->if_index; 2603 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2604 p.expire = 0; 2605 else { 2606 /* XXX: we assume time_t is signed. */ 2607 maxexpire = (-1) & 2608 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2609 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2610 p.expire = pr->ndpr_lastupdate + 2611 pr->ndpr_vltime + 2612 (time_second - time_uptime); 2613 else 2614 p.expire = maxexpire; 2615 } 2616 p.refcnt = pr->ndpr_addrcnt; 2617 p.flags = pr->ndpr_stateflags; 2618 p.advrtrs = 0; 2619 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2620 p.advrtrs++; 2621 error = SYSCTL_OUT(req, &p, sizeof(p)); 2622 if (error != 0) 2623 break; 2624 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2625 s6.sin6_addr = pfr->router->rtaddr; 2626 if (sa6_recoverscope(&s6)) 2627 log(LOG_ERR, 2628 "scope error in prefix list (%s)\n", 2629 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2630 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2631 if (error != 0) 2632 goto out; 2633 } 2634 } 2635 out: 2636 ND6_RUNLOCK(); 2637 return (error); 2638 } 2639 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2640 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2641 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix", 2642 "NDP prefix list"); 2643 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2644 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2645 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2646 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2647