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 *); 144 145 146 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 { 211 212 lltable_update_ifaddr(LLTABLE6(ifp)); 213 } 214 215 void 216 nd6_init(void) 217 { 218 219 mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF); 220 rw_init(&V_nd6_lock, "nd6 list"); 221 222 LIST_INIT(&V_nd_prefix); 223 nd6_defrouter_init(); 224 225 /* Start timers. */ 226 callout_init(&V_nd6_slowtimo_ch, 0); 227 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 228 nd6_slowtimo, curvnet); 229 230 callout_init(&V_nd6_timer_ch, 0); 231 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 232 233 nd6_dad_init(); 234 if (IS_DEFAULT_VNET(curvnet)) { 235 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 236 NULL, EVENTHANDLER_PRI_ANY); 237 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event, 238 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 239 ifnet_link_event_eh = EVENTHANDLER_REGISTER(ifnet_link_event, 240 nd6_ifnet_link_event, NULL, EVENTHANDLER_PRI_ANY); 241 } 242 } 243 244 #ifdef VIMAGE 245 void 246 nd6_destroy() 247 { 248 249 callout_drain(&V_nd6_slowtimo_ch); 250 callout_drain(&V_nd6_timer_ch); 251 if (IS_DEFAULT_VNET(curvnet)) { 252 EVENTHANDLER_DEREGISTER(ifnet_link_event, ifnet_link_event_eh); 253 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 254 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh); 255 } 256 rw_destroy(&V_nd6_lock); 257 mtx_destroy(&V_nd6_onlink_mtx); 258 } 259 #endif 260 261 struct nd_ifinfo * 262 nd6_ifattach(struct ifnet *ifp) 263 { 264 struct nd_ifinfo *nd; 265 266 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 267 nd->initialized = 1; 268 269 nd->chlim = IPV6_DEFHLIM; 270 nd->basereachable = REACHABLE_TIME; 271 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 272 nd->retrans = RETRANS_TIMER; 273 274 nd->flags = ND6_IFF_PERFORMNUD; 275 276 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 277 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 278 * default regardless of the V_ip6_auto_linklocal configuration to 279 * give a reasonable default behavior. 280 */ 281 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 282 (ifp->if_flags & IFF_LOOPBACK)) 283 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 284 /* 285 * A loopback interface does not need to accept RTADV. 286 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 287 * default regardless of the V_ip6_accept_rtadv configuration to 288 * prevent the interface from accepting RA messages arrived 289 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 290 */ 291 if (V_ip6_accept_rtadv && 292 !(ifp->if_flags & IFF_LOOPBACK) && 293 (ifp->if_type != IFT_BRIDGE)) 294 nd->flags |= ND6_IFF_ACCEPT_RTADV; 295 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 296 nd->flags |= ND6_IFF_NO_RADR; 297 298 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 299 nd6_setmtu0(ifp, nd); 300 301 return nd; 302 } 303 304 void 305 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd) 306 { 307 struct epoch_tracker et; 308 struct ifaddr *ifa, *next; 309 310 NET_EPOCH_ENTER(et); 311 CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) { 312 if (ifa->ifa_addr->sa_family != AF_INET6) 313 continue; 314 315 /* stop DAD processing */ 316 nd6_dad_stop(ifa); 317 } 318 NET_EPOCH_EXIT(et); 319 320 free(nd, M_IP6NDP); 321 } 322 323 /* 324 * Reset ND level link MTU. This function is called when the physical MTU 325 * changes, which means we might have to adjust the ND level MTU. 326 */ 327 void 328 nd6_setmtu(struct ifnet *ifp) 329 { 330 if (ifp->if_afdata[AF_INET6] == NULL) 331 return; 332 333 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 334 } 335 336 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 337 void 338 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 339 { 340 u_int32_t omaxmtu; 341 342 omaxmtu = ndi->maxmtu; 343 ndi->maxmtu = ifp->if_mtu; 344 345 /* 346 * Decreasing the interface MTU under IPV6 minimum MTU may cause 347 * undesirable situation. We thus notify the operator of the change 348 * explicitly. The check for omaxmtu is necessary to restrict the 349 * log to the case of changing the MTU, not initializing it. 350 */ 351 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 352 log(LOG_NOTICE, "nd6_setmtu0: " 353 "new link MTU on %s (%lu) is too small for IPv6\n", 354 if_name(ifp), (unsigned long)ndi->maxmtu); 355 } 356 357 if (ndi->maxmtu > V_in6_maxmtu) 358 in6_setmaxmtu(); /* check all interfaces just in case */ 359 360 } 361 362 void 363 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 364 { 365 366 bzero(ndopts, sizeof(*ndopts)); 367 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 368 ndopts->nd_opts_last 369 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 370 371 if (icmp6len == 0) { 372 ndopts->nd_opts_done = 1; 373 ndopts->nd_opts_search = NULL; 374 } 375 } 376 377 /* 378 * Take one ND option. 379 */ 380 struct nd_opt_hdr * 381 nd6_option(union nd_opts *ndopts) 382 { 383 struct nd_opt_hdr *nd_opt; 384 int olen; 385 386 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 387 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 388 __func__)); 389 if (ndopts->nd_opts_search == NULL) 390 return NULL; 391 if (ndopts->nd_opts_done) 392 return NULL; 393 394 nd_opt = ndopts->nd_opts_search; 395 396 /* make sure nd_opt_len is inside the buffer */ 397 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 398 bzero(ndopts, sizeof(*ndopts)); 399 return NULL; 400 } 401 402 olen = nd_opt->nd_opt_len << 3; 403 if (olen == 0) { 404 /* 405 * Message validation requires that all included 406 * options have a length that is greater than zero. 407 */ 408 bzero(ndopts, sizeof(*ndopts)); 409 return NULL; 410 } 411 412 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 413 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 414 /* option overruns the end of buffer, invalid */ 415 bzero(ndopts, sizeof(*ndopts)); 416 return NULL; 417 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 418 /* reached the end of options chain */ 419 ndopts->nd_opts_done = 1; 420 ndopts->nd_opts_search = NULL; 421 } 422 return nd_opt; 423 } 424 425 /* 426 * Parse multiple ND options. 427 * This function is much easier to use, for ND routines that do not need 428 * multiple options of the same type. 429 */ 430 int 431 nd6_options(union nd_opts *ndopts) 432 { 433 struct nd_opt_hdr *nd_opt; 434 int i = 0; 435 436 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 437 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 438 __func__)); 439 if (ndopts->nd_opts_search == NULL) 440 return 0; 441 442 while (1) { 443 nd_opt = nd6_option(ndopts); 444 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 445 /* 446 * Message validation requires that all included 447 * options have a length that is greater than zero. 448 */ 449 ICMP6STAT_INC(icp6s_nd_badopt); 450 bzero(ndopts, sizeof(*ndopts)); 451 return -1; 452 } 453 454 if (nd_opt == NULL) 455 goto skip1; 456 457 switch (nd_opt->nd_opt_type) { 458 case ND_OPT_SOURCE_LINKADDR: 459 case ND_OPT_TARGET_LINKADDR: 460 case ND_OPT_MTU: 461 case ND_OPT_REDIRECTED_HEADER: 462 case ND_OPT_NONCE: 463 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 464 nd6log((LOG_INFO, 465 "duplicated ND6 option found (type=%d)\n", 466 nd_opt->nd_opt_type)); 467 /* XXX bark? */ 468 } else { 469 ndopts->nd_opt_array[nd_opt->nd_opt_type] 470 = nd_opt; 471 } 472 break; 473 case ND_OPT_PREFIX_INFORMATION: 474 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 475 ndopts->nd_opt_array[nd_opt->nd_opt_type] 476 = nd_opt; 477 } 478 ndopts->nd_opts_pi_end = 479 (struct nd_opt_prefix_info *)nd_opt; 480 break; 481 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 482 case ND_OPT_RDNSS: /* RFC 6106 */ 483 case ND_OPT_DNSSL: /* RFC 6106 */ 484 /* 485 * Silently ignore options we know and do not care about 486 * in the kernel. 487 */ 488 break; 489 default: 490 /* 491 * Unknown options must be silently ignored, 492 * to accommodate future extension to the protocol. 493 */ 494 nd6log((LOG_DEBUG, 495 "nd6_options: unsupported option %d - " 496 "option ignored\n", nd_opt->nd_opt_type)); 497 } 498 499 skip1: 500 i++; 501 if (i > V_nd6_maxndopt) { 502 ICMP6STAT_INC(icp6s_nd_toomanyopt); 503 nd6log((LOG_INFO, "too many loop in nd opt\n")); 504 break; 505 } 506 507 if (ndopts->nd_opts_done) 508 break; 509 } 510 511 return 0; 512 } 513 514 /* 515 * ND6 timer routine to handle ND6 entries 516 */ 517 static void 518 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 519 { 520 int canceled; 521 522 LLE_WLOCK_ASSERT(ln); 523 524 if (tick < 0) { 525 ln->la_expire = 0; 526 ln->ln_ntick = 0; 527 canceled = callout_stop(&ln->lle_timer); 528 } else { 529 ln->la_expire = time_uptime + tick / hz; 530 LLE_ADDREF(ln); 531 if (tick > INT_MAX) { 532 ln->ln_ntick = tick - INT_MAX; 533 canceled = callout_reset(&ln->lle_timer, INT_MAX, 534 nd6_llinfo_timer, ln); 535 } else { 536 ln->ln_ntick = 0; 537 canceled = callout_reset(&ln->lle_timer, tick, 538 nd6_llinfo_timer, ln); 539 } 540 } 541 if (canceled > 0) 542 LLE_REMREF(ln); 543 } 544 545 /* 546 * Gets source address of the first packet in hold queue 547 * and stores it in @src. 548 * Returns pointer to @src (if hold queue is not empty) or NULL. 549 * 550 * Set noinline to be dtrace-friendly 551 */ 552 static __noinline struct in6_addr * 553 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 554 { 555 struct ip6_hdr hdr; 556 struct mbuf *m; 557 558 if (ln->la_hold == NULL) 559 return (NULL); 560 561 /* 562 * assume every packet in la_hold has the same IP header 563 */ 564 m = ln->la_hold; 565 if (sizeof(hdr) > m->m_len) 566 return (NULL); 567 568 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 569 *src = hdr.ip6_src; 570 571 return (src); 572 } 573 574 /* 575 * Checks if we need to switch from STALE state. 576 * 577 * RFC 4861 requires switching from STALE to DELAY state 578 * on first packet matching entry, waiting V_nd6_delay and 579 * transition to PROBE state (if upper layer confirmation was 580 * not received). 581 * 582 * This code performs a bit differently: 583 * On packet hit we don't change state (but desired state 584 * can be guessed by control plane). However, after V_nd6_delay 585 * seconds code will transition to PROBE state (so DELAY state 586 * is kinda skipped in most situations). 587 * 588 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so 589 * we perform the following upon entering STALE state: 590 * 591 * 1) Arm timer to run each V_nd6_delay seconds to make sure that 592 * if packet was transmitted at the start of given interval, we 593 * would be able to switch to PROBE state in V_nd6_delay seconds 594 * as user expects. 595 * 596 * 2) Reschedule timer until original V_nd6_gctimer expires keeping 597 * lle in STALE state (remaining timer value stored in lle_remtime). 598 * 599 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay 600 * seconds ago. 601 * 602 * Returns non-zero value if the entry is still STALE (storing 603 * the next timer interval in @pdelay). 604 * 605 * Returns zero value if original timer expired or we need to switch to 606 * PROBE (store that in @do_switch variable). 607 */ 608 static int 609 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) 610 { 611 int nd_delay, nd_gctimer, r_skip_req; 612 time_t lle_hittime; 613 long delay; 614 615 *do_switch = 0; 616 nd_gctimer = V_nd6_gctimer; 617 nd_delay = V_nd6_delay; 618 619 LLE_REQ_LOCK(lle); 620 r_skip_req = lle->r_skip_req; 621 lle_hittime = lle->lle_hittime; 622 LLE_REQ_UNLOCK(lle); 623 624 if (r_skip_req > 0) { 625 626 /* 627 * Nonzero r_skip_req value was set upon entering 628 * STALE state. Since value was not changed, no 629 * packets were passed using this lle. Ask for 630 * timer reschedule and keep STALE state. 631 */ 632 delay = (long)(MIN(nd_gctimer, nd_delay)); 633 delay *= hz; 634 if (lle->lle_remtime > delay) 635 lle->lle_remtime -= delay; 636 else { 637 delay = lle->lle_remtime; 638 lle->lle_remtime = 0; 639 } 640 641 if (delay == 0) { 642 643 /* 644 * The original ng6_gctime timeout ended, 645 * no more rescheduling. 646 */ 647 return (0); 648 } 649 650 *pdelay = delay; 651 return (1); 652 } 653 654 /* 655 * Packet received. Verify timestamp 656 */ 657 delay = (long)(time_uptime - lle_hittime); 658 if (delay < nd_delay) { 659 660 /* 661 * V_nd6_delay still not passed since the first 662 * hit in STALE state. 663 * Reshedule timer and return. 664 */ 665 *pdelay = (long)(nd_delay - delay) * hz; 666 return (1); 667 } 668 669 /* Request switching to probe */ 670 *do_switch = 1; 671 return (0); 672 } 673 674 675 /* 676 * Switch @lle state to new state optionally arming timers. 677 * 678 * Set noinline to be dtrace-friendly 679 */ 680 __noinline void 681 nd6_llinfo_setstate(struct llentry *lle, int newstate) 682 { 683 struct ifnet *ifp; 684 int nd_gctimer, nd_delay; 685 long delay, remtime; 686 687 delay = 0; 688 remtime = 0; 689 690 switch (newstate) { 691 case ND6_LLINFO_INCOMPLETE: 692 ifp = lle->lle_tbl->llt_ifp; 693 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; 694 break; 695 case ND6_LLINFO_REACHABLE: 696 if (!ND6_LLINFO_PERMANENT(lle)) { 697 ifp = lle->lle_tbl->llt_ifp; 698 delay = (long)ND_IFINFO(ifp)->reachable * hz; 699 } 700 break; 701 case ND6_LLINFO_STALE: 702 703 /* 704 * Notify fast path that we want to know if any packet 705 * is transmitted by setting r_skip_req. 706 */ 707 LLE_REQ_LOCK(lle); 708 lle->r_skip_req = 1; 709 LLE_REQ_UNLOCK(lle); 710 nd_delay = V_nd6_delay; 711 nd_gctimer = V_nd6_gctimer; 712 713 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz; 714 remtime = (long)nd_gctimer * hz - delay; 715 break; 716 case ND6_LLINFO_DELAY: 717 lle->la_asked = 0; 718 delay = (long)V_nd6_delay * hz; 719 break; 720 } 721 722 if (delay > 0) 723 nd6_llinfo_settimer_locked(lle, delay); 724 725 lle->lle_remtime = remtime; 726 lle->ln_state = newstate; 727 } 728 729 /* 730 * Timer-dependent part of nd state machine. 731 * 732 * Set noinline to be dtrace-friendly 733 */ 734 static __noinline void 735 nd6_llinfo_timer(void *arg) 736 { 737 struct epoch_tracker et; 738 struct llentry *ln; 739 struct in6_addr *dst, *pdst, *psrc, src; 740 struct ifnet *ifp; 741 struct nd_ifinfo *ndi; 742 int do_switch, send_ns; 743 long delay; 744 745 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 746 ln = (struct llentry *)arg; 747 ifp = lltable_get_ifp(ln->lle_tbl); 748 CURVNET_SET(ifp->if_vnet); 749 750 ND6_RLOCK(); 751 LLE_WLOCK(ln); 752 if (callout_pending(&ln->lle_timer)) { 753 /* 754 * Here we are a bit odd here in the treatment of 755 * active/pending. If the pending bit is set, it got 756 * rescheduled before I ran. The active 757 * bit we ignore, since if it was stopped 758 * in ll_tablefree() and was currently running 759 * it would have return 0 so the code would 760 * not have deleted it since the callout could 761 * not be stopped so we want to go through 762 * with the delete here now. If the callout 763 * was restarted, the pending bit will be back on and 764 * we just want to bail since the callout_reset would 765 * return 1 and our reference would have been removed 766 * by nd6_llinfo_settimer_locked above since canceled 767 * would have been 1. 768 */ 769 LLE_WUNLOCK(ln); 770 ND6_RUNLOCK(); 771 CURVNET_RESTORE(); 772 return; 773 } 774 NET_EPOCH_ENTER(et); 775 ndi = ND_IFINFO(ifp); 776 send_ns = 0; 777 dst = &ln->r_l3addr.addr6; 778 pdst = dst; 779 780 if (ln->ln_ntick > 0) { 781 if (ln->ln_ntick > INT_MAX) { 782 ln->ln_ntick -= INT_MAX; 783 nd6_llinfo_settimer_locked(ln, INT_MAX); 784 } else { 785 ln->ln_ntick = 0; 786 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 787 } 788 goto done; 789 } 790 791 if (ln->la_flags & LLE_STATIC) { 792 goto done; 793 } 794 795 if (ln->la_flags & LLE_DELETED) { 796 nd6_free(&ln, 0); 797 goto done; 798 } 799 800 switch (ln->ln_state) { 801 case ND6_LLINFO_INCOMPLETE: 802 if (ln->la_asked < V_nd6_mmaxtries) { 803 ln->la_asked++; 804 send_ns = 1; 805 /* Send NS to multicast address */ 806 pdst = NULL; 807 } else { 808 struct mbuf *m = ln->la_hold; 809 if (m) { 810 struct mbuf *m0; 811 812 /* 813 * assuming every packet in la_hold has the 814 * same IP header. Send error after unlock. 815 */ 816 m0 = m->m_nextpkt; 817 m->m_nextpkt = NULL; 818 ln->la_hold = m0; 819 clear_llinfo_pqueue(ln); 820 } 821 nd6_free(&ln, 0); 822 if (m != NULL) { 823 struct mbuf *n = m; 824 825 /* 826 * if there are any ummapped mbufs, we 827 * must free them, rather than using 828 * them for an ICMP, as they cannot be 829 * checksummed. 830 */ 831 while ((n = n->m_next) != NULL) { 832 if (n->m_flags & M_EXTPG) 833 break; 834 } 835 if (n != NULL) { 836 m_freem(m); 837 m = NULL; 838 } else { 839 icmp6_error2(m, ICMP6_DST_UNREACH, 840 ICMP6_DST_UNREACH_ADDR, 0, ifp); 841 } 842 } 843 } 844 break; 845 case ND6_LLINFO_REACHABLE: 846 if (!ND6_LLINFO_PERMANENT(ln)) 847 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 848 break; 849 850 case ND6_LLINFO_STALE: 851 if (nd6_is_stale(ln, &delay, &do_switch) != 0) { 852 853 /* 854 * No packet has used this entry and GC timeout 855 * has not been passed. Reshedule timer and 856 * return. 857 */ 858 nd6_llinfo_settimer_locked(ln, delay); 859 break; 860 } 861 862 if (do_switch == 0) { 863 864 /* 865 * GC timer has ended and entry hasn't been used. 866 * Run Garbage collector (RFC 4861, 5.3) 867 */ 868 if (!ND6_LLINFO_PERMANENT(ln)) 869 nd6_free(&ln, 1); 870 break; 871 } 872 873 /* Entry has been used AND delay timer has ended. */ 874 875 /* FALLTHROUGH */ 876 877 case ND6_LLINFO_DELAY: 878 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 879 /* We need NUD */ 880 ln->la_asked = 1; 881 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE); 882 send_ns = 1; 883 } else 884 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */ 885 break; 886 case ND6_LLINFO_PROBE: 887 if (ln->la_asked < V_nd6_umaxtries) { 888 ln->la_asked++; 889 send_ns = 1; 890 } else { 891 nd6_free(&ln, 0); 892 } 893 break; 894 default: 895 panic("%s: paths in a dark night can be confusing: %d", 896 __func__, ln->ln_state); 897 } 898 done: 899 if (ln != NULL) 900 ND6_RUNLOCK(); 901 if (send_ns != 0) { 902 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 903 psrc = nd6_llinfo_get_holdsrc(ln, &src); 904 LLE_FREE_LOCKED(ln); 905 ln = NULL; 906 nd6_ns_output(ifp, psrc, pdst, dst, NULL); 907 } 908 909 if (ln != NULL) 910 LLE_FREE_LOCKED(ln); 911 NET_EPOCH_EXIT(et); 912 CURVNET_RESTORE(); 913 } 914 915 916 /* 917 * ND6 timer routine to expire default route list and prefix list 918 */ 919 void 920 nd6_timer(void *arg) 921 { 922 CURVNET_SET((struct vnet *) arg); 923 struct epoch_tracker et; 924 struct nd_prhead prl; 925 struct nd_prefix *pr, *npr; 926 struct ifnet *ifp; 927 struct in6_ifaddr *ia6, *nia6; 928 uint64_t genid; 929 930 LIST_INIT(&prl); 931 932 NET_EPOCH_ENTER(et); 933 nd6_defrouter_timer(); 934 935 /* 936 * expire interface addresses. 937 * in the past the loop was inside prefix expiry processing. 938 * However, from a stricter speci-confrmance standpoint, we should 939 * rather separate address lifetimes and prefix lifetimes. 940 * 941 * XXXRW: in6_ifaddrhead locking. 942 */ 943 addrloop: 944 CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 945 /* check address lifetime */ 946 if (IFA6_IS_INVALID(ia6)) { 947 int regen = 0; 948 949 /* 950 * If the expiring address is temporary, try 951 * regenerating a new one. This would be useful when 952 * we suspended a laptop PC, then turned it on after a 953 * period that could invalidate all temporary 954 * addresses. Although we may have to restart the 955 * loop (see below), it must be after purging the 956 * address. Otherwise, we'd see an infinite loop of 957 * regeneration. 958 */ 959 if (V_ip6_use_tempaddr && 960 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 961 if (regen_tmpaddr(ia6) == 0) 962 regen = 1; 963 } 964 965 in6_purgeaddr(&ia6->ia_ifa); 966 967 if (regen) 968 goto addrloop; /* XXX: see below */ 969 } else if (IFA6_IS_DEPRECATED(ia6)) { 970 int oldflags = ia6->ia6_flags; 971 972 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 973 974 /* 975 * If a temporary address has just become deprecated, 976 * regenerate a new one if possible. 977 */ 978 if (V_ip6_use_tempaddr && 979 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 980 (oldflags & IN6_IFF_DEPRECATED) == 0) { 981 982 if (regen_tmpaddr(ia6) == 0) { 983 /* 984 * A new temporary address is 985 * generated. 986 * XXX: this means the address chain 987 * has changed while we are still in 988 * the loop. Although the change 989 * would not cause disaster (because 990 * it's not a deletion, but an 991 * addition,) we'd rather restart the 992 * loop just for safety. Or does this 993 * significantly reduce performance?? 994 */ 995 goto addrloop; 996 } 997 } 998 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 999 /* 1000 * Schedule DAD for a tentative address. This happens 1001 * if the interface was down or not running 1002 * when the address was configured. 1003 */ 1004 int delay; 1005 1006 delay = arc4random() % 1007 (MAX_RTR_SOLICITATION_DELAY * hz); 1008 nd6_dad_start((struct ifaddr *)ia6, delay); 1009 } else { 1010 /* 1011 * Check status of the interface. If it is down, 1012 * mark the address as tentative for future DAD. 1013 */ 1014 ifp = ia6->ia_ifp; 1015 if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 && 1016 ((ifp->if_flags & IFF_UP) == 0 || 1017 (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 || 1018 (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){ 1019 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 1020 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 1021 } 1022 1023 /* 1024 * A new RA might have made a deprecated address 1025 * preferred. 1026 */ 1027 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 1028 } 1029 } 1030 NET_EPOCH_EXIT(et); 1031 1032 ND6_WLOCK(); 1033 restart: 1034 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1035 /* 1036 * Expire prefixes. Since the pltime is only used for 1037 * autoconfigured addresses, pltime processing for prefixes is 1038 * not necessary. 1039 * 1040 * Only unlink after all derived addresses have expired. This 1041 * may not occur until two hours after the prefix has expired 1042 * per RFC 4862. If the prefix expires before its derived 1043 * addresses, mark it off-link. This will be done automatically 1044 * after unlinking if no address references remain. 1045 */ 1046 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME || 1047 time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime) 1048 continue; 1049 1050 if (pr->ndpr_addrcnt == 0) { 1051 nd6_prefix_unlink(pr, &prl); 1052 continue; 1053 } 1054 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { 1055 genid = V_nd6_list_genid; 1056 nd6_prefix_ref(pr); 1057 ND6_WUNLOCK(); 1058 ND6_ONLINK_LOCK(); 1059 (void)nd6_prefix_offlink(pr); 1060 ND6_ONLINK_UNLOCK(); 1061 ND6_WLOCK(); 1062 nd6_prefix_rele(pr); 1063 if (genid != V_nd6_list_genid) 1064 goto restart; 1065 } 1066 } 1067 ND6_WUNLOCK(); 1068 1069 while ((pr = LIST_FIRST(&prl)) != NULL) { 1070 LIST_REMOVE(pr, ndpr_entry); 1071 nd6_prefix_del(pr); 1072 } 1073 1074 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 1075 nd6_timer, curvnet); 1076 1077 CURVNET_RESTORE(); 1078 } 1079 1080 /* 1081 * ia6 - deprecated/invalidated temporary address 1082 */ 1083 static int 1084 regen_tmpaddr(struct in6_ifaddr *ia6) 1085 { 1086 struct ifaddr *ifa; 1087 struct ifnet *ifp; 1088 struct in6_ifaddr *public_ifa6 = NULL; 1089 1090 NET_EPOCH_ASSERT(); 1091 1092 ifp = ia6->ia_ifa.ifa_ifp; 1093 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1094 struct in6_ifaddr *it6; 1095 1096 if (ifa->ifa_addr->sa_family != AF_INET6) 1097 continue; 1098 1099 it6 = (struct in6_ifaddr *)ifa; 1100 1101 /* ignore no autoconf addresses. */ 1102 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1103 continue; 1104 1105 /* ignore autoconf addresses with different prefixes. */ 1106 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 1107 continue; 1108 1109 /* 1110 * Now we are looking at an autoconf address with the same 1111 * prefix as ours. If the address is temporary and is still 1112 * preferred, do not create another one. It would be rare, but 1113 * could happen, for example, when we resume a laptop PC after 1114 * a long period. 1115 */ 1116 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1117 !IFA6_IS_DEPRECATED(it6)) { 1118 public_ifa6 = NULL; 1119 break; 1120 } 1121 1122 /* 1123 * This is a public autoconf address that has the same prefix 1124 * as ours. If it is preferred, keep it. We can't break the 1125 * loop here, because there may be a still-preferred temporary 1126 * address with the prefix. 1127 */ 1128 if (!IFA6_IS_DEPRECATED(it6)) 1129 public_ifa6 = it6; 1130 } 1131 if (public_ifa6 != NULL) 1132 ifa_ref(&public_ifa6->ia_ifa); 1133 1134 if (public_ifa6 != NULL) { 1135 int e; 1136 1137 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 1138 ifa_free(&public_ifa6->ia_ifa); 1139 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 1140 " tmp addr,errno=%d\n", e); 1141 return (-1); 1142 } 1143 ifa_free(&public_ifa6->ia_ifa); 1144 return (0); 1145 } 1146 1147 return (-1); 1148 } 1149 1150 /* 1151 * Remove prefix and default router list entries corresponding to ifp. Neighbor 1152 * cache entries are freed in in6_domifdetach(). 1153 */ 1154 void 1155 nd6_purge(struct ifnet *ifp) 1156 { 1157 struct nd_prhead prl; 1158 struct nd_prefix *pr, *npr; 1159 1160 LIST_INIT(&prl); 1161 1162 /* Purge default router list entries toward ifp. */ 1163 nd6_defrouter_purge(ifp); 1164 1165 ND6_WLOCK(); 1166 /* 1167 * Remove prefixes on ifp. We should have already removed addresses on 1168 * this interface, so no addresses should be referencing these prefixes. 1169 */ 1170 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1171 if (pr->ndpr_ifp == ifp) 1172 nd6_prefix_unlink(pr, &prl); 1173 } 1174 ND6_WUNLOCK(); 1175 1176 /* Delete the unlinked prefix objects. */ 1177 while ((pr = LIST_FIRST(&prl)) != NULL) { 1178 LIST_REMOVE(pr, ndpr_entry); 1179 nd6_prefix_del(pr); 1180 } 1181 1182 /* cancel default outgoing interface setting */ 1183 if (V_nd6_defifindex == ifp->if_index) 1184 nd6_setdefaultiface(0); 1185 1186 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1187 /* Refresh default router list. */ 1188 defrouter_select_fib(ifp->if_fib); 1189 } 1190 } 1191 1192 /* 1193 * the caller acquires and releases the lock on the lltbls 1194 * Returns the llentry locked 1195 */ 1196 struct llentry * 1197 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1198 { 1199 struct sockaddr_in6 sin6; 1200 struct llentry *ln; 1201 1202 bzero(&sin6, sizeof(sin6)); 1203 sin6.sin6_len = sizeof(struct sockaddr_in6); 1204 sin6.sin6_family = AF_INET6; 1205 sin6.sin6_addr = *addr6; 1206 1207 IF_AFDATA_LOCK_ASSERT(ifp); 1208 1209 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6); 1210 1211 return (ln); 1212 } 1213 1214 static struct llentry * 1215 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1216 { 1217 struct sockaddr_in6 sin6; 1218 struct llentry *ln; 1219 1220 bzero(&sin6, sizeof(sin6)); 1221 sin6.sin6_len = sizeof(struct sockaddr_in6); 1222 sin6.sin6_family = AF_INET6; 1223 sin6.sin6_addr = *addr6; 1224 1225 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 1226 if (ln != NULL) 1227 ln->ln_state = ND6_LLINFO_NOSTATE; 1228 1229 return (ln); 1230 } 1231 1232 /* 1233 * Test whether a given IPv6 address is a neighbor or not, ignoring 1234 * the actual neighbor cache. The neighbor cache is ignored in order 1235 * to not reenter the routing code from within itself. 1236 */ 1237 static int 1238 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1239 { 1240 struct nd_prefix *pr; 1241 struct ifaddr *ifa; 1242 struct rt_addrinfo info; 1243 struct sockaddr_in6 rt_key; 1244 const struct sockaddr *dst6; 1245 uint64_t genid; 1246 int error, fibnum; 1247 1248 /* 1249 * A link-local address is always a neighbor. 1250 * XXX: a link does not necessarily specify a single interface. 1251 */ 1252 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 1253 struct sockaddr_in6 sin6_copy; 1254 u_int32_t zone; 1255 1256 /* 1257 * We need sin6_copy since sa6_recoverscope() may modify the 1258 * content (XXX). 1259 */ 1260 sin6_copy = *addr; 1261 if (sa6_recoverscope(&sin6_copy)) 1262 return (0); /* XXX: should be impossible */ 1263 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 1264 return (0); 1265 if (sin6_copy.sin6_scope_id == zone) 1266 return (1); 1267 else 1268 return (0); 1269 } 1270 1271 bzero(&rt_key, sizeof(rt_key)); 1272 bzero(&info, sizeof(info)); 1273 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1274 1275 /* 1276 * If the address matches one of our addresses, 1277 * it should be a neighbor. 1278 * If the address matches one of our on-link prefixes, it should be a 1279 * neighbor. 1280 */ 1281 ND6_RLOCK(); 1282 restart: 1283 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1284 if (pr->ndpr_ifp != ifp) 1285 continue; 1286 1287 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) { 1288 dst6 = (const struct sockaddr *)&pr->ndpr_prefix; 1289 1290 /* 1291 * We only need to check all FIBs if add_addr_allfibs 1292 * is unset. If set, checking any FIB will suffice. 1293 */ 1294 fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0; 1295 for (; fibnum < rt_numfibs; fibnum++) { 1296 genid = V_nd6_list_genid; 1297 ND6_RUNLOCK(); 1298 1299 /* 1300 * Restore length field before 1301 * retrying lookup 1302 */ 1303 rt_key.sin6_len = sizeof(rt_key); 1304 error = rib_lookup_info(fibnum, dst6, 0, 0, 1305 &info); 1306 1307 ND6_RLOCK(); 1308 if (genid != V_nd6_list_genid) 1309 goto restart; 1310 if (error == 0) 1311 break; 1312 } 1313 if (error != 0) 1314 continue; 1315 1316 /* 1317 * This is the case where multiple interfaces 1318 * have the same prefix, but only one is installed 1319 * into the routing table and that prefix entry 1320 * is not the one being examined here. In the case 1321 * where RADIX_MPATH is enabled, multiple route 1322 * entries (of the same rt_key value) will be 1323 * installed because the interface addresses all 1324 * differ. 1325 */ 1326 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1327 &rt_key.sin6_addr)) 1328 continue; 1329 } 1330 1331 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1332 &addr->sin6_addr, &pr->ndpr_mask)) { 1333 ND6_RUNLOCK(); 1334 return (1); 1335 } 1336 } 1337 ND6_RUNLOCK(); 1338 1339 /* 1340 * If the address is assigned on the node of the other side of 1341 * a p2p interface, the address should be a neighbor. 1342 */ 1343 if (ifp->if_flags & IFF_POINTOPOINT) { 1344 struct epoch_tracker et; 1345 1346 NET_EPOCH_ENTER(et); 1347 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1348 if (ifa->ifa_addr->sa_family != addr->sin6_family) 1349 continue; 1350 if (ifa->ifa_dstaddr != NULL && 1351 sa_equal(addr, ifa->ifa_dstaddr)) { 1352 NET_EPOCH_EXIT(et); 1353 return 1; 1354 } 1355 } 1356 NET_EPOCH_EXIT(et); 1357 } 1358 1359 /* 1360 * If the default router list is empty, all addresses are regarded 1361 * as on-link, and thus, as a neighbor. 1362 */ 1363 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1364 nd6_defrouter_list_empty() && 1365 V_nd6_defifindex == ifp->if_index) { 1366 return (1); 1367 } 1368 1369 return (0); 1370 } 1371 1372 1373 /* 1374 * Detect if a given IPv6 address identifies a neighbor on a given link. 1375 * XXX: should take care of the destination of a p2p link? 1376 */ 1377 int 1378 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1379 { 1380 struct llentry *lle; 1381 int rc = 0; 1382 1383 NET_EPOCH_ASSERT(); 1384 IF_AFDATA_UNLOCK_ASSERT(ifp); 1385 if (nd6_is_new_addr_neighbor(addr, ifp)) 1386 return (1); 1387 1388 /* 1389 * Even if the address matches none of our addresses, it might be 1390 * in the neighbor cache. 1391 */ 1392 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1393 LLE_RUNLOCK(lle); 1394 rc = 1; 1395 } 1396 return (rc); 1397 } 1398 1399 /* 1400 * Free an nd6 llinfo entry. 1401 * Since the function would cause significant changes in the kernel, DO NOT 1402 * make it global, unless you have a strong reason for the change, and are sure 1403 * that the change is safe. 1404 * 1405 * Set noinline to be dtrace-friendly 1406 */ 1407 static __noinline void 1408 nd6_free(struct llentry **lnp, int gc) 1409 { 1410 struct ifnet *ifp; 1411 struct llentry *ln; 1412 struct nd_defrouter *dr; 1413 1414 ln = *lnp; 1415 *lnp = NULL; 1416 1417 LLE_WLOCK_ASSERT(ln); 1418 ND6_RLOCK_ASSERT(); 1419 1420 ifp = lltable_get_ifp(ln->lle_tbl); 1421 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0) 1422 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp); 1423 else 1424 dr = NULL; 1425 ND6_RUNLOCK(); 1426 1427 if ((ln->la_flags & LLE_DELETED) == 0) 1428 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 1429 1430 /* 1431 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1432 * even though it is not harmful, it was not really necessary. 1433 */ 1434 1435 /* cancel timer */ 1436 nd6_llinfo_settimer_locked(ln, -1); 1437 1438 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1439 if (dr != NULL && dr->expire && 1440 ln->ln_state == ND6_LLINFO_STALE && gc) { 1441 /* 1442 * If the reason for the deletion is just garbage 1443 * collection, and the neighbor is an active default 1444 * router, do not delete it. Instead, reset the GC 1445 * timer using the router's lifetime. 1446 * Simply deleting the entry would affect default 1447 * router selection, which is not necessarily a good 1448 * thing, especially when we're using router preference 1449 * values. 1450 * XXX: the check for ln_state would be redundant, 1451 * but we intentionally keep it just in case. 1452 */ 1453 if (dr->expire > time_uptime) 1454 nd6_llinfo_settimer_locked(ln, 1455 (dr->expire - time_uptime) * hz); 1456 else 1457 nd6_llinfo_settimer_locked(ln, 1458 (long)V_nd6_gctimer * hz); 1459 1460 LLE_REMREF(ln); 1461 LLE_WUNLOCK(ln); 1462 defrouter_rele(dr); 1463 return; 1464 } 1465 1466 if (dr) { 1467 /* 1468 * Unreachablity of a router might affect the default 1469 * router selection and on-link detection of advertised 1470 * prefixes. 1471 */ 1472 1473 /* 1474 * Temporarily fake the state to choose a new default 1475 * router and to perform on-link determination of 1476 * prefixes correctly. 1477 * Below the state will be set correctly, 1478 * or the entry itself will be deleted. 1479 */ 1480 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1481 } 1482 1483 if (ln->ln_router || dr) { 1484 1485 /* 1486 * We need to unlock to avoid a LOR with rt6_flush() with the 1487 * rnh and for the calls to pfxlist_onlink_check() and 1488 * defrouter_select_fib() in the block further down for calls 1489 * into nd6_lookup(). We still hold a ref. 1490 */ 1491 LLE_WUNLOCK(ln); 1492 1493 /* 1494 * rt6_flush must be called whether or not the neighbor 1495 * is in the Default Router List. 1496 * See a corresponding comment in nd6_na_input(). 1497 */ 1498 rt6_flush(&ln->r_l3addr.addr6, ifp); 1499 } 1500 1501 if (dr) { 1502 /* 1503 * Since defrouter_select_fib() does not affect the 1504 * on-link determination and MIP6 needs the check 1505 * before the default router selection, we perform 1506 * the check now. 1507 */ 1508 pfxlist_onlink_check(); 1509 1510 /* 1511 * Refresh default router list. 1512 */ 1513 defrouter_select_fib(dr->ifp->if_fib); 1514 } 1515 1516 /* 1517 * If this entry was added by an on-link redirect, remove the 1518 * corresponding host route. 1519 */ 1520 if (ln->la_flags & LLE_REDIRECT) 1521 nd6_free_redirect(ln); 1522 1523 if (ln->ln_router || dr) 1524 LLE_WLOCK(ln); 1525 } 1526 1527 /* 1528 * Save to unlock. We still hold an extra reference and will not 1529 * free(9) in llentry_free() if someone else holds one as well. 1530 */ 1531 LLE_WUNLOCK(ln); 1532 IF_AFDATA_LOCK(ifp); 1533 LLE_WLOCK(ln); 1534 /* Guard against race with other llentry_free(). */ 1535 if (ln->la_flags & LLE_LINKED) { 1536 /* Remove callout reference */ 1537 LLE_REMREF(ln); 1538 lltable_unlink_entry(ln->lle_tbl, ln); 1539 } 1540 IF_AFDATA_UNLOCK(ifp); 1541 1542 llentry_free(ln); 1543 if (dr != NULL) 1544 defrouter_rele(dr); 1545 } 1546 1547 static int 1548 nd6_isdynrte(const struct rtentry *rt, const struct nhop_object *nh, void *xap) 1549 { 1550 1551 if (nh->nh_flags & NHF_REDIRECT) 1552 return (1); 1553 1554 return (0); 1555 } 1556 1557 /* 1558 * Remove the rtentry for the given llentry, 1559 * both of which were installed by a redirect. 1560 */ 1561 static void 1562 nd6_free_redirect(const struct llentry *ln) 1563 { 1564 int fibnum; 1565 struct sockaddr_in6 sin6; 1566 struct rt_addrinfo info; 1567 struct epoch_tracker et; 1568 1569 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); 1570 memset(&info, 0, sizeof(info)); 1571 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; 1572 info.rti_filter = nd6_isdynrte; 1573 1574 NET_EPOCH_ENTER(et); 1575 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 1576 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum); 1577 NET_EPOCH_EXIT(et); 1578 } 1579 1580 /* 1581 * Updates status of the default router route. 1582 */ 1583 void 1584 nd6_subscription_cb(struct rib_head *rnh, struct rib_cmd_info *rc, void *arg) 1585 { 1586 struct nd_defrouter *dr; 1587 struct nhop_object *nh; 1588 1589 if (rc->rc_cmd == RTM_DELETE) { 1590 nh = rc->rc_nh_old; 1591 1592 if (nh->nh_flags & NHF_DEFAULT) { 1593 dr = defrouter_lookup(&nh->gw6_sa.sin6_addr, nh->nh_ifp); 1594 if (dr != NULL) { 1595 dr->installed = 0; 1596 defrouter_rele(dr); 1597 } 1598 } 1599 } 1600 } 1601 1602 1603 int 1604 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1605 { 1606 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1607 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1608 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1609 struct epoch_tracker et; 1610 int error = 0; 1611 1612 if (ifp->if_afdata[AF_INET6] == NULL) 1613 return (EPFNOSUPPORT); 1614 switch (cmd) { 1615 case OSIOCGIFINFO_IN6: 1616 #define ND ndi->ndi 1617 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1618 bzero(&ND, sizeof(ND)); 1619 ND.linkmtu = IN6_LINKMTU(ifp); 1620 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1621 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1622 ND.reachable = ND_IFINFO(ifp)->reachable; 1623 ND.retrans = ND_IFINFO(ifp)->retrans; 1624 ND.flags = ND_IFINFO(ifp)->flags; 1625 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1626 ND.chlim = ND_IFINFO(ifp)->chlim; 1627 break; 1628 case SIOCGIFINFO_IN6: 1629 ND = *ND_IFINFO(ifp); 1630 break; 1631 case SIOCSIFINFO_IN6: 1632 /* 1633 * used to change host variables from userland. 1634 * intended for a use on router to reflect RA configurations. 1635 */ 1636 /* 0 means 'unspecified' */ 1637 if (ND.linkmtu != 0) { 1638 if (ND.linkmtu < IPV6_MMTU || 1639 ND.linkmtu > IN6_LINKMTU(ifp)) { 1640 error = EINVAL; 1641 break; 1642 } 1643 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1644 } 1645 1646 if (ND.basereachable != 0) { 1647 int obasereachable = ND_IFINFO(ifp)->basereachable; 1648 1649 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1650 if (ND.basereachable != obasereachable) 1651 ND_IFINFO(ifp)->reachable = 1652 ND_COMPUTE_RTIME(ND.basereachable); 1653 } 1654 if (ND.retrans != 0) 1655 ND_IFINFO(ifp)->retrans = ND.retrans; 1656 if (ND.chlim != 0) 1657 ND_IFINFO(ifp)->chlim = ND.chlim; 1658 /* FALLTHROUGH */ 1659 case SIOCSIFINFO_FLAGS: 1660 { 1661 struct ifaddr *ifa; 1662 struct in6_ifaddr *ia; 1663 1664 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1665 !(ND.flags & ND6_IFF_IFDISABLED)) { 1666 /* ifdisabled 1->0 transision */ 1667 1668 /* 1669 * If the interface is marked as ND6_IFF_IFDISABLED and 1670 * has an link-local address with IN6_IFF_DUPLICATED, 1671 * do not clear ND6_IFF_IFDISABLED. 1672 * See RFC 4862, Section 5.4.5. 1673 */ 1674 NET_EPOCH_ENTER(et); 1675 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1676 if (ifa->ifa_addr->sa_family != AF_INET6) 1677 continue; 1678 ia = (struct in6_ifaddr *)ifa; 1679 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1680 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1681 break; 1682 } 1683 NET_EPOCH_EXIT(et); 1684 1685 if (ifa != NULL) { 1686 /* LLA is duplicated. */ 1687 ND.flags |= ND6_IFF_IFDISABLED; 1688 log(LOG_ERR, "Cannot enable an interface" 1689 " with a link-local address marked" 1690 " duplicate.\n"); 1691 } else { 1692 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1693 if (ifp->if_flags & IFF_UP) 1694 in6_if_up(ifp); 1695 } 1696 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1697 (ND.flags & ND6_IFF_IFDISABLED)) { 1698 /* ifdisabled 0->1 transision */ 1699 /* Mark all IPv6 address as tentative. */ 1700 1701 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1702 if (V_ip6_dad_count > 0 && 1703 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1704 NET_EPOCH_ENTER(et); 1705 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1706 ifa_link) { 1707 if (ifa->ifa_addr->sa_family != 1708 AF_INET6) 1709 continue; 1710 ia = (struct in6_ifaddr *)ifa; 1711 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1712 } 1713 NET_EPOCH_EXIT(et); 1714 } 1715 } 1716 1717 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1718 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1719 /* auto_linklocal 0->1 transision */ 1720 1721 /* If no link-local address on ifp, configure */ 1722 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1723 in6_ifattach(ifp, NULL); 1724 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1725 ifp->if_flags & IFF_UP) { 1726 /* 1727 * When the IF already has 1728 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1729 * address is assigned, and IFF_UP, try to 1730 * assign one. 1731 */ 1732 NET_EPOCH_ENTER(et); 1733 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, 1734 ifa_link) { 1735 if (ifa->ifa_addr->sa_family != 1736 AF_INET6) 1737 continue; 1738 ia = (struct in6_ifaddr *)ifa; 1739 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1740 break; 1741 } 1742 NET_EPOCH_EXIT(et); 1743 if (ifa != NULL) 1744 /* No LLA is configured. */ 1745 in6_ifattach(ifp, NULL); 1746 } 1747 } 1748 ND_IFINFO(ifp)->flags = ND.flags; 1749 break; 1750 } 1751 #undef ND 1752 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1753 /* sync kernel routing table with the default router list */ 1754 defrouter_reset(); 1755 defrouter_select_fib(RT_ALL_FIBS); 1756 break; 1757 case SIOCSPFXFLUSH_IN6: 1758 { 1759 /* flush all the prefix advertised by routers */ 1760 struct in6_ifaddr *ia, *ia_next; 1761 struct nd_prefix *pr, *next; 1762 struct nd_prhead prl; 1763 1764 LIST_INIT(&prl); 1765 1766 ND6_WLOCK(); 1767 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1768 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1769 continue; /* XXX */ 1770 nd6_prefix_unlink(pr, &prl); 1771 } 1772 ND6_WUNLOCK(); 1773 1774 while ((pr = LIST_FIRST(&prl)) != NULL) { 1775 LIST_REMOVE(pr, ndpr_entry); 1776 /* XXXRW: in6_ifaddrhead locking. */ 1777 CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1778 ia_next) { 1779 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1780 continue; 1781 1782 if (ia->ia6_ndpr == pr) 1783 in6_purgeaddr(&ia->ia_ifa); 1784 } 1785 nd6_prefix_del(pr); 1786 } 1787 break; 1788 } 1789 case SIOCSRTRFLUSH_IN6: 1790 { 1791 /* flush all the default routers */ 1792 1793 defrouter_reset(); 1794 nd6_defrouter_flush_all(); 1795 defrouter_select_fib(RT_ALL_FIBS); 1796 break; 1797 } 1798 case SIOCGNBRINFO_IN6: 1799 { 1800 struct llentry *ln; 1801 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1802 1803 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1804 return (error); 1805 1806 NET_EPOCH_ENTER(et); 1807 ln = nd6_lookup(&nb_addr, 0, ifp); 1808 NET_EPOCH_EXIT(et); 1809 1810 if (ln == NULL) { 1811 error = EINVAL; 1812 break; 1813 } 1814 nbi->state = ln->ln_state; 1815 nbi->asked = ln->la_asked; 1816 nbi->isrouter = ln->ln_router; 1817 if (ln->la_expire == 0) 1818 nbi->expire = 0; 1819 else 1820 nbi->expire = ln->la_expire + ln->lle_remtime / hz + 1821 (time_second - time_uptime); 1822 LLE_RUNLOCK(ln); 1823 break; 1824 } 1825 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1826 ndif->ifindex = V_nd6_defifindex; 1827 break; 1828 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1829 return (nd6_setdefaultiface(ndif->ifindex)); 1830 } 1831 return (error); 1832 } 1833 1834 /* 1835 * Calculates new isRouter value based on provided parameters and 1836 * returns it. 1837 */ 1838 static int 1839 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, 1840 int ln_router) 1841 { 1842 1843 /* 1844 * ICMP6 type dependent behavior. 1845 * 1846 * NS: clear IsRouter if new entry 1847 * RS: clear IsRouter 1848 * RA: set IsRouter if there's lladdr 1849 * redir: clear IsRouter if new entry 1850 * 1851 * RA case, (1): 1852 * The spec says that we must set IsRouter in the following cases: 1853 * - If lladdr exist, set IsRouter. This means (1-5). 1854 * - If it is old entry (!newentry), set IsRouter. This means (7). 1855 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1856 * A quetion arises for (1) case. (1) case has no lladdr in the 1857 * neighbor cache, this is similar to (6). 1858 * This case is rare but we figured that we MUST NOT set IsRouter. 1859 * 1860 * is_new old_addr new_addr NS RS RA redir 1861 * D R 1862 * 0 n n (1) c ? s 1863 * 0 y n (2) c s s 1864 * 0 n y (3) c s s 1865 * 0 y y (4) c s s 1866 * 0 y y (5) c s s 1867 * 1 -- n (6) c c c s 1868 * 1 -- y (7) c c s c s 1869 * 1870 * (c=clear s=set) 1871 */ 1872 switch (type & 0xff) { 1873 case ND_NEIGHBOR_SOLICIT: 1874 /* 1875 * New entry must have is_router flag cleared. 1876 */ 1877 if (is_new) /* (6-7) */ 1878 ln_router = 0; 1879 break; 1880 case ND_REDIRECT: 1881 /* 1882 * If the icmp is a redirect to a better router, always set the 1883 * is_router flag. Otherwise, if the entry is newly created, 1884 * clear the flag. [RFC 2461, sec 8.3] 1885 */ 1886 if (code == ND_REDIRECT_ROUTER) 1887 ln_router = 1; 1888 else { 1889 if (is_new) /* (6-7) */ 1890 ln_router = 0; 1891 } 1892 break; 1893 case ND_ROUTER_SOLICIT: 1894 /* 1895 * is_router flag must always be cleared. 1896 */ 1897 ln_router = 0; 1898 break; 1899 case ND_ROUTER_ADVERT: 1900 /* 1901 * Mark an entry with lladdr as a router. 1902 */ 1903 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ 1904 (is_new && new_addr)) { /* (7) */ 1905 ln_router = 1; 1906 } 1907 break; 1908 } 1909 1910 return (ln_router); 1911 } 1912 1913 /* 1914 * Create neighbor cache entry and cache link-layer address, 1915 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1916 * 1917 * type - ICMP6 type 1918 * code - type dependent information 1919 * 1920 */ 1921 void 1922 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1923 int lladdrlen, int type, int code) 1924 { 1925 struct llentry *ln = NULL, *ln_tmp; 1926 int is_newentry; 1927 int do_update; 1928 int olladdr; 1929 int llchange; 1930 int flags; 1931 uint16_t router = 0; 1932 struct sockaddr_in6 sin6; 1933 struct mbuf *chain = NULL; 1934 u_char linkhdr[LLE_MAX_LINKHDR]; 1935 size_t linkhdrsize; 1936 int lladdr_off; 1937 1938 NET_EPOCH_ASSERT(); 1939 IF_AFDATA_UNLOCK_ASSERT(ifp); 1940 1941 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1942 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1943 1944 /* nothing must be updated for unspecified address */ 1945 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1946 return; 1947 1948 /* 1949 * Validation about ifp->if_addrlen and lladdrlen must be done in 1950 * the caller. 1951 * 1952 * XXX If the link does not have link-layer adderss, what should 1953 * we do? (ifp->if_addrlen == 0) 1954 * Spec says nothing in sections for RA, RS and NA. There's small 1955 * description on it in NS section (RFC 2461 7.2.3). 1956 */ 1957 flags = lladdr ? LLE_EXCLUSIVE : 0; 1958 ln = nd6_lookup(from, flags, ifp); 1959 is_newentry = 0; 1960 if (ln == NULL) { 1961 flags |= LLE_EXCLUSIVE; 1962 ln = nd6_alloc(from, 0, ifp); 1963 if (ln == NULL) 1964 return; 1965 1966 /* 1967 * Since we already know all the data for the new entry, 1968 * fill it before insertion. 1969 */ 1970 if (lladdr != NULL) { 1971 linkhdrsize = sizeof(linkhdr); 1972 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 1973 linkhdr, &linkhdrsize, &lladdr_off) != 0) 1974 return; 1975 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 1976 lladdr_off); 1977 } 1978 1979 IF_AFDATA_WLOCK(ifp); 1980 LLE_WLOCK(ln); 1981 /* Prefer any existing lle over newly-created one */ 1982 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); 1983 if (ln_tmp == NULL) 1984 lltable_link_entry(LLTABLE6(ifp), ln); 1985 IF_AFDATA_WUNLOCK(ifp); 1986 if (ln_tmp == NULL) { 1987 /* No existing lle, mark as new entry (6,7) */ 1988 is_newentry = 1; 1989 if (lladdr != NULL) { /* (7) */ 1990 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 1991 EVENTHANDLER_INVOKE(lle_event, ln, 1992 LLENTRY_RESOLVED); 1993 } 1994 } else { 1995 lltable_free_entry(LLTABLE6(ifp), ln); 1996 ln = ln_tmp; 1997 ln_tmp = NULL; 1998 } 1999 } 2000 /* do nothing if static ndp is set */ 2001 if ((ln->la_flags & LLE_STATIC)) { 2002 if (flags & LLE_EXCLUSIVE) 2003 LLE_WUNLOCK(ln); 2004 else 2005 LLE_RUNLOCK(ln); 2006 return; 2007 } 2008 2009 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 2010 if (olladdr && lladdr) { 2011 llchange = bcmp(lladdr, ln->ll_addr, 2012 ifp->if_addrlen); 2013 } else if (!olladdr && lladdr) 2014 llchange = 1; 2015 else 2016 llchange = 0; 2017 2018 /* 2019 * newentry olladdr lladdr llchange (*=record) 2020 * 0 n n -- (1) 2021 * 0 y n -- (2) 2022 * 0 n y y (3) * STALE 2023 * 0 y y n (4) * 2024 * 0 y y y (5) * STALE 2025 * 1 -- n -- (6) NOSTATE(= PASSIVE) 2026 * 1 -- y -- (7) * STALE 2027 */ 2028 2029 do_update = 0; 2030 if (is_newentry == 0 && llchange != 0) { 2031 do_update = 1; /* (3,5) */ 2032 2033 /* 2034 * Record source link-layer address 2035 * XXX is it dependent to ifp->if_type? 2036 */ 2037 linkhdrsize = sizeof(linkhdr); 2038 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 2039 linkhdr, &linkhdrsize, &lladdr_off) != 0) 2040 return; 2041 2042 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 2043 lladdr_off) == 0) { 2044 /* Entry was deleted */ 2045 return; 2046 } 2047 2048 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2049 2050 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2051 2052 if (ln->la_hold != NULL) 2053 nd6_grab_holdchain(ln, &chain, &sin6); 2054 } 2055 2056 /* Calculates new router status */ 2057 router = nd6_is_router(type, code, is_newentry, olladdr, 2058 lladdr != NULL ? 1 : 0, ln->ln_router); 2059 2060 ln->ln_router = router; 2061 /* Mark non-router redirects with special flag */ 2062 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) 2063 ln->la_flags |= LLE_REDIRECT; 2064 2065 if (flags & LLE_EXCLUSIVE) 2066 LLE_WUNLOCK(ln); 2067 else 2068 LLE_RUNLOCK(ln); 2069 2070 if (chain != NULL) 2071 nd6_flush_holdchain(ifp, chain, &sin6); 2072 2073 /* 2074 * When the link-layer address of a router changes, select the 2075 * best router again. In particular, when the neighbor entry is newly 2076 * created, it might affect the selection policy. 2077 * Question: can we restrict the first condition to the "is_newentry" 2078 * case? 2079 * XXX: when we hear an RA from a new router with the link-layer 2080 * address option, defrouter_select_fib() is called twice, since 2081 * defrtrlist_update called the function as well. However, I believe 2082 * we can compromise the overhead, since it only happens the first 2083 * time. 2084 * XXX: although defrouter_select_fib() should not have a bad effect 2085 * for those are not autoconfigured hosts, we explicitly avoid such 2086 * cases for safety. 2087 */ 2088 if ((do_update || is_newentry) && router && 2089 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 2090 /* 2091 * guaranteed recursion 2092 */ 2093 defrouter_select_fib(ifp->if_fib); 2094 } 2095 } 2096 2097 static void 2098 nd6_slowtimo(void *arg) 2099 { 2100 struct epoch_tracker et; 2101 CURVNET_SET((struct vnet *) arg); 2102 struct nd_ifinfo *nd6if; 2103 struct ifnet *ifp; 2104 2105 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 2106 nd6_slowtimo, curvnet); 2107 NET_EPOCH_ENTER(et); 2108 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2109 if (ifp->if_afdata[AF_INET6] == NULL) 2110 continue; 2111 nd6if = ND_IFINFO(ifp); 2112 if (nd6if->basereachable && /* already initialized */ 2113 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 2114 /* 2115 * Since reachable time rarely changes by router 2116 * advertisements, we SHOULD insure that a new random 2117 * value gets recomputed at least once every few hours. 2118 * (RFC 2461, 6.3.4) 2119 */ 2120 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 2121 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 2122 } 2123 } 2124 NET_EPOCH_EXIT(et); 2125 CURVNET_RESTORE(); 2126 } 2127 2128 void 2129 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 2130 struct sockaddr_in6 *sin6) 2131 { 2132 2133 LLE_WLOCK_ASSERT(ln); 2134 2135 *chain = ln->la_hold; 2136 ln->la_hold = NULL; 2137 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 2138 2139 if (ln->ln_state == ND6_LLINFO_STALE) { 2140 2141 /* 2142 * The first time we send a packet to a 2143 * neighbor whose entry is STALE, we have 2144 * to change the state to DELAY and a sets 2145 * a timer to expire in DELAY_FIRST_PROBE_TIME 2146 * seconds to ensure do neighbor unreachability 2147 * detection on expiration. 2148 * (RFC 2461 7.3.3) 2149 */ 2150 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); 2151 } 2152 } 2153 2154 int 2155 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2156 struct sockaddr_in6 *dst, struct route *ro) 2157 { 2158 int error; 2159 int ip6len; 2160 struct ip6_hdr *ip6; 2161 struct m_tag *mtag; 2162 2163 #ifdef MAC 2164 mac_netinet6_nd6_send(ifp, m); 2165 #endif 2166 2167 /* 2168 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2169 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2170 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2171 * to be diverted to user space. When re-injected into the kernel, 2172 * send_output() will directly dispatch them to the outgoing interface. 2173 */ 2174 if (send_sendso_input_hook != NULL) { 2175 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2176 if (mtag != NULL) { 2177 ip6 = mtod(m, struct ip6_hdr *); 2178 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2179 /* Use the SEND socket */ 2180 error = send_sendso_input_hook(m, ifp, SND_OUT, 2181 ip6len); 2182 /* -1 == no app on SEND socket */ 2183 if (error == 0 || error != -1) 2184 return (error); 2185 } 2186 } 2187 2188 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 2189 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 2190 mtod(m, struct ip6_hdr *)); 2191 2192 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2193 origifp = ifp; 2194 2195 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); 2196 return (error); 2197 } 2198 2199 /* 2200 * Lookup link headerfor @sa_dst address. Stores found 2201 * data in @desten buffer. Copy of lle ln_flags can be also 2202 * saved in @pflags if @pflags is non-NULL. 2203 * 2204 * If destination LLE does not exists or lle state modification 2205 * is required, call "slow" version. 2206 * 2207 * Return values: 2208 * - 0 on success (address copied to buffer). 2209 * - EWOULDBLOCK (no local error, but address is still unresolved) 2210 * - other errors (alloc failure, etc) 2211 */ 2212 int 2213 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 2214 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags, 2215 struct llentry **plle) 2216 { 2217 struct llentry *ln = NULL; 2218 const struct sockaddr_in6 *dst6; 2219 2220 NET_EPOCH_ASSERT(); 2221 2222 if (pflags != NULL) 2223 *pflags = 0; 2224 2225 dst6 = (const struct sockaddr_in6 *)sa_dst; 2226 2227 /* discard the packet if IPv6 operation is disabled on the interface */ 2228 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2229 m_freem(m); 2230 return (ENETDOWN); /* better error? */ 2231 } 2232 2233 if (m != NULL && m->m_flags & M_MCAST) { 2234 switch (ifp->if_type) { 2235 case IFT_ETHER: 2236 case IFT_L2VLAN: 2237 case IFT_BRIDGE: 2238 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, 2239 desten); 2240 return (0); 2241 default: 2242 m_freem(m); 2243 return (EAFNOSUPPORT); 2244 } 2245 } 2246 2247 ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED, 2248 ifp); 2249 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { 2250 /* Entry found, let's copy lle info */ 2251 bcopy(ln->r_linkdata, desten, ln->r_hdrlen); 2252 if (pflags != NULL) 2253 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); 2254 /* Check if we have feedback request from nd6 timer */ 2255 if (ln->r_skip_req != 0) { 2256 LLE_REQ_LOCK(ln); 2257 ln->r_skip_req = 0; /* Notify that entry was used */ 2258 ln->lle_hittime = time_uptime; 2259 LLE_REQ_UNLOCK(ln); 2260 } 2261 if (plle) { 2262 LLE_ADDREF(ln); 2263 *plle = ln; 2264 LLE_WUNLOCK(ln); 2265 } 2266 return (0); 2267 } else if (plle && ln) 2268 LLE_WUNLOCK(ln); 2269 2270 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle)); 2271 } 2272 2273 2274 /* 2275 * Do L2 address resolution for @sa_dst address. Stores found 2276 * address in @desten buffer. Copy of lle ln_flags can be also 2277 * saved in @pflags if @pflags is non-NULL. 2278 * 2279 * Heavy version. 2280 * Function assume that destination LLE does not exist, 2281 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. 2282 * 2283 * Set noinline to be dtrace-friendly 2284 */ 2285 static __noinline int 2286 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m, 2287 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags, 2288 struct llentry **plle) 2289 { 2290 struct llentry *lle = NULL, *lle_tmp; 2291 struct in6_addr *psrc, src; 2292 int send_ns, ll_len; 2293 char *lladdr; 2294 2295 NET_EPOCH_ASSERT(); 2296 2297 /* 2298 * Address resolution or Neighbor Unreachability Detection 2299 * for the next hop. 2300 * At this point, the destination of the packet must be a unicast 2301 * or an anycast address(i.e. not a multicast). 2302 */ 2303 if (lle == NULL) { 2304 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2305 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2306 /* 2307 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2308 * the condition below is not very efficient. But we believe 2309 * it is tolerable, because this should be a rare case. 2310 */ 2311 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2312 if (lle == NULL) { 2313 char ip6buf[INET6_ADDRSTRLEN]; 2314 log(LOG_DEBUG, 2315 "nd6_output: can't allocate llinfo for %s " 2316 "(ln=%p)\n", 2317 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2318 m_freem(m); 2319 return (ENOBUFS); 2320 } 2321 2322 IF_AFDATA_WLOCK(ifp); 2323 LLE_WLOCK(lle); 2324 /* Prefer any existing entry over newly-created one */ 2325 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2326 if (lle_tmp == NULL) 2327 lltable_link_entry(LLTABLE6(ifp), lle); 2328 IF_AFDATA_WUNLOCK(ifp); 2329 if (lle_tmp != NULL) { 2330 lltable_free_entry(LLTABLE6(ifp), lle); 2331 lle = lle_tmp; 2332 lle_tmp = NULL; 2333 } 2334 } 2335 } 2336 if (lle == NULL) { 2337 m_freem(m); 2338 return (ENOBUFS); 2339 } 2340 2341 LLE_WLOCK_ASSERT(lle); 2342 2343 /* 2344 * The first time we send a packet to a neighbor whose entry is 2345 * STALE, we have to change the state to DELAY and a sets a timer to 2346 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2347 * neighbor unreachability detection on expiration. 2348 * (RFC 2461 7.3.3) 2349 */ 2350 if (lle->ln_state == ND6_LLINFO_STALE) 2351 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); 2352 2353 /* 2354 * If the neighbor cache entry has a state other than INCOMPLETE 2355 * (i.e. its link-layer address is already resolved), just 2356 * send the packet. 2357 */ 2358 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { 2359 if (flags & LLE_ADDRONLY) { 2360 lladdr = lle->ll_addr; 2361 ll_len = ifp->if_addrlen; 2362 } else { 2363 lladdr = lle->r_linkdata; 2364 ll_len = lle->r_hdrlen; 2365 } 2366 bcopy(lladdr, desten, ll_len); 2367 if (pflags != NULL) 2368 *pflags = lle->la_flags; 2369 if (plle) { 2370 LLE_ADDREF(lle); 2371 *plle = lle; 2372 } 2373 LLE_WUNLOCK(lle); 2374 return (0); 2375 } 2376 2377 /* 2378 * There is a neighbor cache entry, but no ethernet address 2379 * response yet. Append this latest packet to the end of the 2380 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen, 2381 * the oldest packet in the queue will be removed. 2382 */ 2383 2384 if (lle->la_hold != NULL) { 2385 struct mbuf *m_hold; 2386 int i; 2387 2388 i = 0; 2389 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2390 i++; 2391 if (m_hold->m_nextpkt == NULL) { 2392 m_hold->m_nextpkt = m; 2393 break; 2394 } 2395 } 2396 while (i >= V_nd6_maxqueuelen) { 2397 m_hold = lle->la_hold; 2398 lle->la_hold = lle->la_hold->m_nextpkt; 2399 m_freem(m_hold); 2400 i--; 2401 } 2402 } else { 2403 lle->la_hold = m; 2404 } 2405 2406 /* 2407 * If there has been no NS for the neighbor after entering the 2408 * INCOMPLETE state, send the first solicitation. 2409 * Note that for newly-created lle la_asked will be 0, 2410 * so we will transition from ND6_LLINFO_NOSTATE to 2411 * ND6_LLINFO_INCOMPLETE state here. 2412 */ 2413 psrc = NULL; 2414 send_ns = 0; 2415 if (lle->la_asked == 0) { 2416 lle->la_asked++; 2417 send_ns = 1; 2418 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2419 2420 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); 2421 } 2422 LLE_WUNLOCK(lle); 2423 if (send_ns != 0) 2424 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2425 2426 return (EWOULDBLOCK); 2427 } 2428 2429 /* 2430 * Do L2 address resolution for @sa_dst address. Stores found 2431 * address in @desten buffer. Copy of lle ln_flags can be also 2432 * saved in @pflags if @pflags is non-NULL. 2433 * 2434 * Return values: 2435 * - 0 on success (address copied to buffer). 2436 * - EWOULDBLOCK (no local error, but address is still unresolved) 2437 * - other errors (alloc failure, etc) 2438 */ 2439 int 2440 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 2441 char *desten, uint32_t *pflags) 2442 { 2443 int error; 2444 2445 flags |= LLE_ADDRONLY; 2446 error = nd6_resolve_slow(ifp, flags, NULL, 2447 (const struct sockaddr_in6 *)dst, desten, pflags, NULL); 2448 return (error); 2449 } 2450 2451 int 2452 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain, 2453 struct sockaddr_in6 *dst) 2454 { 2455 struct mbuf *m, *m_head; 2456 int error = 0; 2457 2458 m_head = chain; 2459 2460 while (m_head) { 2461 m = m_head; 2462 m_head = m_head->m_nextpkt; 2463 error = nd6_output_ifp(ifp, ifp, m, dst, NULL); 2464 } 2465 2466 /* 2467 * XXX 2468 * note that intermediate errors are blindly ignored 2469 */ 2470 return (error); 2471 } 2472 2473 static int 2474 nd6_need_cache(struct ifnet *ifp) 2475 { 2476 /* 2477 * XXX: we currently do not make neighbor cache on any interface 2478 * other than Ethernet and GIF. 2479 * 2480 * RFC2893 says: 2481 * - unidirectional tunnels needs no ND 2482 */ 2483 switch (ifp->if_type) { 2484 case IFT_ETHER: 2485 case IFT_IEEE1394: 2486 case IFT_L2VLAN: 2487 case IFT_INFINIBAND: 2488 case IFT_BRIDGE: 2489 case IFT_PROPVIRTUAL: 2490 return (1); 2491 default: 2492 return (0); 2493 } 2494 } 2495 2496 /* 2497 * Add pernament ND6 link-layer record for given 2498 * interface address. 2499 * 2500 * Very similar to IPv4 arp_ifinit(), but: 2501 * 1) IPv6 DAD is performed in different place 2502 * 2) It is called by IPv6 protocol stack in contrast to 2503 * arp_ifinit() which is typically called in SIOCSIFADDR 2504 * driver ioctl handler. 2505 * 2506 */ 2507 int 2508 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2509 { 2510 struct ifnet *ifp; 2511 struct llentry *ln, *ln_tmp; 2512 struct sockaddr *dst; 2513 2514 ifp = ia->ia_ifa.ifa_ifp; 2515 if (nd6_need_cache(ifp) == 0) 2516 return (0); 2517 2518 dst = (struct sockaddr *)&ia->ia_addr; 2519 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2520 if (ln == NULL) 2521 return (ENOBUFS); 2522 2523 IF_AFDATA_WLOCK(ifp); 2524 LLE_WLOCK(ln); 2525 /* Unlink any entry if exists */ 2526 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2527 if (ln_tmp != NULL) 2528 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2529 lltable_link_entry(LLTABLE6(ifp), ln); 2530 IF_AFDATA_WUNLOCK(ifp); 2531 2532 if (ln_tmp != NULL) 2533 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2534 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2535 2536 LLE_WUNLOCK(ln); 2537 if (ln_tmp != NULL) 2538 llentry_free(ln_tmp); 2539 2540 return (0); 2541 } 2542 2543 /* 2544 * Removes either all lle entries for given @ia, or lle 2545 * corresponding to @ia address. 2546 */ 2547 void 2548 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) 2549 { 2550 struct sockaddr_in6 mask, addr; 2551 struct sockaddr *saddr, *smask; 2552 struct ifnet *ifp; 2553 2554 ifp = ia->ia_ifa.ifa_ifp; 2555 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2556 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2557 saddr = (struct sockaddr *)&addr; 2558 smask = (struct sockaddr *)&mask; 2559 2560 if (all != 0) 2561 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); 2562 else 2563 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); 2564 } 2565 2566 static void 2567 clear_llinfo_pqueue(struct llentry *ln) 2568 { 2569 struct mbuf *m_hold, *m_hold_next; 2570 2571 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2572 m_hold_next = m_hold->m_nextpkt; 2573 m_freem(m_hold); 2574 } 2575 2576 ln->la_hold = NULL; 2577 } 2578 2579 static int 2580 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2581 { 2582 struct in6_prefix p; 2583 struct sockaddr_in6 s6; 2584 struct nd_prefix *pr; 2585 struct nd_pfxrouter *pfr; 2586 time_t maxexpire; 2587 int error; 2588 char ip6buf[INET6_ADDRSTRLEN]; 2589 2590 if (req->newptr) 2591 return (EPERM); 2592 2593 error = sysctl_wire_old_buffer(req, 0); 2594 if (error != 0) 2595 return (error); 2596 2597 bzero(&p, sizeof(p)); 2598 p.origin = PR_ORIG_RA; 2599 bzero(&s6, sizeof(s6)); 2600 s6.sin6_family = AF_INET6; 2601 s6.sin6_len = sizeof(s6); 2602 2603 ND6_RLOCK(); 2604 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2605 p.prefix = pr->ndpr_prefix; 2606 if (sa6_recoverscope(&p.prefix)) { 2607 log(LOG_ERR, "scope error in prefix list (%s)\n", 2608 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2609 /* XXX: press on... */ 2610 } 2611 p.raflags = pr->ndpr_raf; 2612 p.prefixlen = pr->ndpr_plen; 2613 p.vltime = pr->ndpr_vltime; 2614 p.pltime = pr->ndpr_pltime; 2615 p.if_index = pr->ndpr_ifp->if_index; 2616 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2617 p.expire = 0; 2618 else { 2619 /* XXX: we assume time_t is signed. */ 2620 maxexpire = (-1) & 2621 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2622 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2623 p.expire = pr->ndpr_lastupdate + 2624 pr->ndpr_vltime + 2625 (time_second - time_uptime); 2626 else 2627 p.expire = maxexpire; 2628 } 2629 p.refcnt = pr->ndpr_addrcnt; 2630 p.flags = pr->ndpr_stateflags; 2631 p.advrtrs = 0; 2632 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2633 p.advrtrs++; 2634 error = SYSCTL_OUT(req, &p, sizeof(p)); 2635 if (error != 0) 2636 break; 2637 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2638 s6.sin6_addr = pfr->router->rtaddr; 2639 if (sa6_recoverscope(&s6)) 2640 log(LOG_ERR, 2641 "scope error in prefix list (%s)\n", 2642 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2643 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2644 if (error != 0) 2645 goto out; 2646 } 2647 } 2648 out: 2649 ND6_RUNLOCK(); 2650 return (error); 2651 } 2652 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2653 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2654 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix", 2655 "NDP prefix list"); 2656 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2657 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2658 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2659 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2660