1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/callout.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/sockio.h> 45 #include <sys/time.h> 46 #include <sys/kernel.h> 47 #include <sys/protosw.h> 48 #include <sys/errno.h> 49 #include <sys/syslog.h> 50 #include <sys/lock.h> 51 #include <sys/rwlock.h> 52 #include <sys/queue.h> 53 #include <sys/sdt.h> 54 #include <sys/sysctl.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/if_arc.h> 59 #include <net/if_dl.h> 60 #include <net/if_types.h> 61 #include <net/iso88025.h> 62 #include <net/fddi.h> 63 #include <net/route.h> 64 #include <net/vnet.h> 65 66 #include <netinet/in.h> 67 #include <netinet/in_kdtrace.h> 68 #include <net/if_llatbl.h> 69 #include <netinet/if_ether.h> 70 #include <netinet6/in6_var.h> 71 #include <netinet/ip6.h> 72 #include <netinet6/ip6_var.h> 73 #include <netinet6/scope6_var.h> 74 #include <netinet6/nd6.h> 75 #include <netinet6/in6_ifattach.h> 76 #include <netinet/icmp6.h> 77 #include <netinet6/send.h> 78 79 #include <sys/limits.h> 80 81 #include <security/mac/mac_framework.h> 82 83 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */ 84 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */ 85 86 #define SIN6(s) ((const struct sockaddr_in6 *)(s)) 87 88 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery"); 89 90 /* timer values */ 91 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 92 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 93 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 94 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 95 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 96 * local traffic */ 97 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 98 * collection timer */ 99 100 /* preventing too many loops in ND option parsing */ 101 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 102 103 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 104 * layer hints */ 105 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 106 * ND entries */ 107 #define V_nd6_maxndopt VNET(nd6_maxndopt) 108 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 109 110 #ifdef ND6_DEBUG 111 VNET_DEFINE(int, nd6_debug) = 1; 112 #else 113 VNET_DEFINE(int, nd6_debug) = 0; 114 #endif 115 116 static eventhandler_tag lle_event_eh, iflladdr_event_eh; 117 118 VNET_DEFINE(struct nd_drhead, nd_defrouter); 119 VNET_DEFINE(struct nd_prhead, nd_prefix); 120 VNET_DEFINE(struct rwlock, nd6_lock); 121 122 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 123 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 124 125 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 126 127 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *, 128 struct ifnet *); 129 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 130 static void nd6_slowtimo(void *); 131 static int regen_tmpaddr(struct in6_ifaddr *); 132 static void nd6_free(struct llentry **, int); 133 static void nd6_free_redirect(const struct llentry *); 134 static void nd6_llinfo_timer(void *); 135 static void nd6_llinfo_settimer_locked(struct llentry *, long); 136 static void clear_llinfo_pqueue(struct llentry *); 137 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 138 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *, 139 const struct sockaddr_in6 *, u_char *, uint32_t *); 140 static int nd6_need_cache(struct ifnet *); 141 142 143 static VNET_DEFINE(struct callout, nd6_slowtimo_ch); 144 #define V_nd6_slowtimo_ch VNET(nd6_slowtimo_ch) 145 146 VNET_DEFINE(struct callout, nd6_timer_ch); 147 #define V_nd6_timer_ch VNET(nd6_timer_ch) 148 149 static void 150 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 151 { 152 struct rt_addrinfo rtinfo; 153 struct sockaddr_in6 dst; 154 struct sockaddr_dl gw; 155 struct ifnet *ifp; 156 int type; 157 158 LLE_WLOCK_ASSERT(lle); 159 160 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 161 return; 162 163 switch (evt) { 164 case LLENTRY_RESOLVED: 165 type = RTM_ADD; 166 KASSERT(lle->la_flags & LLE_VALID, 167 ("%s: %p resolved but not valid?", __func__, lle)); 168 break; 169 case LLENTRY_EXPIRED: 170 type = RTM_DELETE; 171 break; 172 default: 173 return; 174 } 175 176 ifp = lltable_get_ifp(lle->lle_tbl); 177 178 bzero(&dst, sizeof(dst)); 179 bzero(&gw, sizeof(gw)); 180 bzero(&rtinfo, sizeof(rtinfo)); 181 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 182 dst.sin6_scope_id = in6_getscopezone(ifp, 183 in6_addrscope(&dst.sin6_addr)); 184 gw.sdl_len = sizeof(struct sockaddr_dl); 185 gw.sdl_family = AF_LINK; 186 gw.sdl_alen = ifp->if_addrlen; 187 gw.sdl_index = ifp->if_index; 188 gw.sdl_type = ifp->if_type; 189 if (evt == LLENTRY_RESOLVED) 190 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 191 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 192 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 193 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 194 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 195 type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB); 196 } 197 198 /* 199 * A handler for interface link layer address change event. 200 */ 201 static void 202 nd6_iflladdr(void *arg __unused, struct ifnet *ifp) 203 { 204 205 lltable_update_ifaddr(LLTABLE6(ifp)); 206 } 207 208 void 209 nd6_init(void) 210 { 211 212 rw_init(&V_nd6_lock, "nd6"); 213 214 LIST_INIT(&V_nd_prefix); 215 216 /* initialization of the default router list */ 217 TAILQ_INIT(&V_nd_defrouter); 218 219 /* Start timers. */ 220 callout_init(&V_nd6_slowtimo_ch, 0); 221 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 222 nd6_slowtimo, curvnet); 223 224 callout_init(&V_nd6_timer_ch, 0); 225 callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet); 226 227 nd6_dad_init(); 228 if (IS_DEFAULT_VNET(curvnet)) { 229 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 230 NULL, EVENTHANDLER_PRI_ANY); 231 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event, 232 nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY); 233 } 234 } 235 236 #ifdef VIMAGE 237 void 238 nd6_destroy() 239 { 240 241 callout_drain(&V_nd6_slowtimo_ch); 242 callout_drain(&V_nd6_timer_ch); 243 if (IS_DEFAULT_VNET(curvnet)) { 244 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 245 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh); 246 } 247 rw_destroy(&V_nd6_lock); 248 } 249 #endif 250 251 struct nd_ifinfo * 252 nd6_ifattach(struct ifnet *ifp) 253 { 254 struct nd_ifinfo *nd; 255 256 nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO); 257 nd->initialized = 1; 258 259 nd->chlim = IPV6_DEFHLIM; 260 nd->basereachable = REACHABLE_TIME; 261 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 262 nd->retrans = RETRANS_TIMER; 263 264 nd->flags = ND6_IFF_PERFORMNUD; 265 266 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 267 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 268 * default regardless of the V_ip6_auto_linklocal configuration to 269 * give a reasonable default behavior. 270 */ 271 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 272 (ifp->if_flags & IFF_LOOPBACK)) 273 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 274 /* 275 * A loopback interface does not need to accept RTADV. 276 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 277 * default regardless of the V_ip6_accept_rtadv configuration to 278 * prevent the interface from accepting RA messages arrived 279 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 280 */ 281 if (V_ip6_accept_rtadv && 282 !(ifp->if_flags & IFF_LOOPBACK) && 283 (ifp->if_type != IFT_BRIDGE)) 284 nd->flags |= ND6_IFF_ACCEPT_RTADV; 285 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 286 nd->flags |= ND6_IFF_NO_RADR; 287 288 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 289 nd6_setmtu0(ifp, nd); 290 291 return nd; 292 } 293 294 void 295 nd6_ifdetach(struct nd_ifinfo *nd) 296 { 297 298 free(nd, M_IP6NDP); 299 } 300 301 /* 302 * Reset ND level link MTU. This function is called when the physical MTU 303 * changes, which means we might have to adjust the ND level MTU. 304 */ 305 void 306 nd6_setmtu(struct ifnet *ifp) 307 { 308 if (ifp->if_afdata[AF_INET6] == NULL) 309 return; 310 311 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 312 } 313 314 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 315 void 316 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 317 { 318 u_int32_t omaxmtu; 319 320 omaxmtu = ndi->maxmtu; 321 322 switch (ifp->if_type) { 323 case IFT_ARCNET: 324 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 325 break; 326 case IFT_FDDI: 327 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 328 break; 329 case IFT_ISO88025: 330 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 331 break; 332 default: 333 ndi->maxmtu = ifp->if_mtu; 334 break; 335 } 336 337 /* 338 * Decreasing the interface MTU under IPV6 minimum MTU may cause 339 * undesirable situation. We thus notify the operator of the change 340 * explicitly. The check for omaxmtu is necessary to restrict the 341 * log to the case of changing the MTU, not initializing it. 342 */ 343 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 344 log(LOG_NOTICE, "nd6_setmtu0: " 345 "new link MTU on %s (%lu) is too small for IPv6\n", 346 if_name(ifp), (unsigned long)ndi->maxmtu); 347 } 348 349 if (ndi->maxmtu > V_in6_maxmtu) 350 in6_setmaxmtu(); /* check all interfaces just in case */ 351 352 } 353 354 void 355 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 356 { 357 358 bzero(ndopts, sizeof(*ndopts)); 359 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 360 ndopts->nd_opts_last 361 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 362 363 if (icmp6len == 0) { 364 ndopts->nd_opts_done = 1; 365 ndopts->nd_opts_search = NULL; 366 } 367 } 368 369 /* 370 * Take one ND option. 371 */ 372 struct nd_opt_hdr * 373 nd6_option(union nd_opts *ndopts) 374 { 375 struct nd_opt_hdr *nd_opt; 376 int olen; 377 378 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 379 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 380 __func__)); 381 if (ndopts->nd_opts_search == NULL) 382 return NULL; 383 if (ndopts->nd_opts_done) 384 return NULL; 385 386 nd_opt = ndopts->nd_opts_search; 387 388 /* make sure nd_opt_len is inside the buffer */ 389 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 390 bzero(ndopts, sizeof(*ndopts)); 391 return NULL; 392 } 393 394 olen = nd_opt->nd_opt_len << 3; 395 if (olen == 0) { 396 /* 397 * Message validation requires that all included 398 * options have a length that is greater than zero. 399 */ 400 bzero(ndopts, sizeof(*ndopts)); 401 return NULL; 402 } 403 404 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 405 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 406 /* option overruns the end of buffer, invalid */ 407 bzero(ndopts, sizeof(*ndopts)); 408 return NULL; 409 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 410 /* reached the end of options chain */ 411 ndopts->nd_opts_done = 1; 412 ndopts->nd_opts_search = NULL; 413 } 414 return nd_opt; 415 } 416 417 /* 418 * Parse multiple ND options. 419 * This function is much easier to use, for ND routines that do not need 420 * multiple options of the same type. 421 */ 422 int 423 nd6_options(union nd_opts *ndopts) 424 { 425 struct nd_opt_hdr *nd_opt; 426 int i = 0; 427 428 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 429 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 430 __func__)); 431 if (ndopts->nd_opts_search == NULL) 432 return 0; 433 434 while (1) { 435 nd_opt = nd6_option(ndopts); 436 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 437 /* 438 * Message validation requires that all included 439 * options have a length that is greater than zero. 440 */ 441 ICMP6STAT_INC(icp6s_nd_badopt); 442 bzero(ndopts, sizeof(*ndopts)); 443 return -1; 444 } 445 446 if (nd_opt == NULL) 447 goto skip1; 448 449 switch (nd_opt->nd_opt_type) { 450 case ND_OPT_SOURCE_LINKADDR: 451 case ND_OPT_TARGET_LINKADDR: 452 case ND_OPT_MTU: 453 case ND_OPT_REDIRECTED_HEADER: 454 case ND_OPT_NONCE: 455 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 456 nd6log((LOG_INFO, 457 "duplicated ND6 option found (type=%d)\n", 458 nd_opt->nd_opt_type)); 459 /* XXX bark? */ 460 } else { 461 ndopts->nd_opt_array[nd_opt->nd_opt_type] 462 = nd_opt; 463 } 464 break; 465 case ND_OPT_PREFIX_INFORMATION: 466 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 467 ndopts->nd_opt_array[nd_opt->nd_opt_type] 468 = nd_opt; 469 } 470 ndopts->nd_opts_pi_end = 471 (struct nd_opt_prefix_info *)nd_opt; 472 break; 473 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 474 case ND_OPT_RDNSS: /* RFC 6106 */ 475 case ND_OPT_DNSSL: /* RFC 6106 */ 476 /* 477 * Silently ignore options we know and do not care about 478 * in the kernel. 479 */ 480 break; 481 default: 482 /* 483 * Unknown options must be silently ignored, 484 * to accommodate future extension to the protocol. 485 */ 486 nd6log((LOG_DEBUG, 487 "nd6_options: unsupported option %d - " 488 "option ignored\n", nd_opt->nd_opt_type)); 489 } 490 491 skip1: 492 i++; 493 if (i > V_nd6_maxndopt) { 494 ICMP6STAT_INC(icp6s_nd_toomanyopt); 495 nd6log((LOG_INFO, "too many loop in nd opt\n")); 496 break; 497 } 498 499 if (ndopts->nd_opts_done) 500 break; 501 } 502 503 return 0; 504 } 505 506 /* 507 * ND6 timer routine to handle ND6 entries 508 */ 509 static void 510 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 511 { 512 int canceled; 513 514 LLE_WLOCK_ASSERT(ln); 515 516 if (tick < 0) { 517 ln->la_expire = 0; 518 ln->ln_ntick = 0; 519 canceled = callout_stop(&ln->lle_timer); 520 } else { 521 ln->la_expire = time_uptime + tick / hz; 522 LLE_ADDREF(ln); 523 if (tick > INT_MAX) { 524 ln->ln_ntick = tick - INT_MAX; 525 canceled = callout_reset(&ln->lle_timer, INT_MAX, 526 nd6_llinfo_timer, ln); 527 } else { 528 ln->ln_ntick = 0; 529 canceled = callout_reset(&ln->lle_timer, tick, 530 nd6_llinfo_timer, ln); 531 } 532 } 533 if (canceled > 0) 534 LLE_REMREF(ln); 535 } 536 537 /* 538 * Gets source address of the first packet in hold queue 539 * and stores it in @src. 540 * Returns pointer to @src (if hold queue is not empty) or NULL. 541 * 542 * Set noinline to be dtrace-friendly 543 */ 544 static __noinline struct in6_addr * 545 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src) 546 { 547 struct ip6_hdr hdr; 548 struct mbuf *m; 549 550 if (ln->la_hold == NULL) 551 return (NULL); 552 553 /* 554 * assume every packet in la_hold has the same IP header 555 */ 556 m = ln->la_hold; 557 if (sizeof(hdr) > m->m_len) 558 return (NULL); 559 560 m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr); 561 *src = hdr.ip6_src; 562 563 return (src); 564 } 565 566 /* 567 * Checks if we need to switch from STALE state. 568 * 569 * RFC 4861 requires switching from STALE to DELAY state 570 * on first packet matching entry, waiting V_nd6_delay and 571 * transition to PROBE state (if upper layer confirmation was 572 * not received). 573 * 574 * This code performs a bit differently: 575 * On packet hit we don't change state (but desired state 576 * can be guessed by control plane). However, after V_nd6_delay 577 * seconds code will transition to PROBE state (so DELAY state 578 * is kinda skipped in most situations). 579 * 580 * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so 581 * we perform the following upon entering STALE state: 582 * 583 * 1) Arm timer to run each V_nd6_delay seconds to make sure that 584 * if packet was transmitted at the start of given interval, we 585 * would be able to switch to PROBE state in V_nd6_delay seconds 586 * as user expects. 587 * 588 * 2) Reschedule timer until original V_nd6_gctimer expires keeping 589 * lle in STALE state (remaining timer value stored in lle_remtime). 590 * 591 * 3) Reschedule timer if packet was transmitted less that V_nd6_delay 592 * seconds ago. 593 * 594 * Returns non-zero value if the entry is still STALE (storing 595 * the next timer interval in @pdelay). 596 * 597 * Returns zero value if original timer expired or we need to switch to 598 * PROBE (store that in @do_switch variable). 599 */ 600 static int 601 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch) 602 { 603 int nd_delay, nd_gctimer, r_skip_req; 604 time_t lle_hittime; 605 long delay; 606 607 *do_switch = 0; 608 nd_gctimer = V_nd6_gctimer; 609 nd_delay = V_nd6_delay; 610 611 LLE_REQ_LOCK(lle); 612 r_skip_req = lle->r_skip_req; 613 lle_hittime = lle->lle_hittime; 614 LLE_REQ_UNLOCK(lle); 615 616 if (r_skip_req > 0) { 617 618 /* 619 * Nonzero r_skip_req value was set upon entering 620 * STALE state. Since value was not changed, no 621 * packets were passed using this lle. Ask for 622 * timer reschedule and keep STALE state. 623 */ 624 delay = (long)(MIN(nd_gctimer, nd_delay)); 625 delay *= hz; 626 if (lle->lle_remtime > delay) 627 lle->lle_remtime -= delay; 628 else { 629 delay = lle->lle_remtime; 630 lle->lle_remtime = 0; 631 } 632 633 if (delay == 0) { 634 635 /* 636 * The original ng6_gctime timeout ended, 637 * no more rescheduling. 638 */ 639 return (0); 640 } 641 642 *pdelay = delay; 643 return (1); 644 } 645 646 /* 647 * Packet received. Verify timestamp 648 */ 649 delay = (long)(time_uptime - lle_hittime); 650 if (delay < nd_delay) { 651 652 /* 653 * V_nd6_delay still not passed since the first 654 * hit in STALE state. 655 * Reshedule timer and return. 656 */ 657 *pdelay = (long)(nd_delay - delay) * hz; 658 return (1); 659 } 660 661 /* Request switching to probe */ 662 *do_switch = 1; 663 return (0); 664 } 665 666 667 /* 668 * Switch @lle state to new state optionally arming timers. 669 * 670 * Set noinline to be dtrace-friendly 671 */ 672 __noinline void 673 nd6_llinfo_setstate(struct llentry *lle, int newstate) 674 { 675 struct ifnet *ifp; 676 int nd_gctimer, nd_delay; 677 long delay, remtime; 678 679 delay = 0; 680 remtime = 0; 681 682 switch (newstate) { 683 case ND6_LLINFO_INCOMPLETE: 684 ifp = lle->lle_tbl->llt_ifp; 685 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000; 686 break; 687 case ND6_LLINFO_REACHABLE: 688 if (!ND6_LLINFO_PERMANENT(lle)) { 689 ifp = lle->lle_tbl->llt_ifp; 690 delay = (long)ND_IFINFO(ifp)->reachable * hz; 691 } 692 break; 693 case ND6_LLINFO_STALE: 694 695 /* 696 * Notify fast path that we want to know if any packet 697 * is transmitted by setting r_skip_req. 698 */ 699 LLE_REQ_LOCK(lle); 700 lle->r_skip_req = 1; 701 LLE_REQ_UNLOCK(lle); 702 nd_delay = V_nd6_delay; 703 nd_gctimer = V_nd6_gctimer; 704 705 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz; 706 remtime = (long)nd_gctimer * hz - delay; 707 break; 708 case ND6_LLINFO_DELAY: 709 lle->la_asked = 0; 710 delay = (long)V_nd6_delay * hz; 711 break; 712 } 713 714 if (delay > 0) 715 nd6_llinfo_settimer_locked(lle, delay); 716 717 lle->lle_remtime = remtime; 718 lle->ln_state = newstate; 719 } 720 721 /* 722 * Timer-dependent part of nd state machine. 723 * 724 * Set noinline to be dtrace-friendly 725 */ 726 static __noinline void 727 nd6_llinfo_timer(void *arg) 728 { 729 struct llentry *ln; 730 struct in6_addr *dst, *pdst, *psrc, src; 731 struct ifnet *ifp; 732 struct nd_ifinfo *ndi; 733 int do_switch, send_ns; 734 long delay; 735 736 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 737 ln = (struct llentry *)arg; 738 ifp = lltable_get_ifp(ln->lle_tbl); 739 CURVNET_SET(ifp->if_vnet); 740 741 ND6_RLOCK(); 742 LLE_WLOCK(ln); 743 if (callout_pending(&ln->lle_timer)) { 744 /* 745 * Here we are a bit odd here in the treatment of 746 * active/pending. If the pending bit is set, it got 747 * rescheduled before I ran. The active 748 * bit we ignore, since if it was stopped 749 * in ll_tablefree() and was currently running 750 * it would have return 0 so the code would 751 * not have deleted it since the callout could 752 * not be stopped so we want to go through 753 * with the delete here now. If the callout 754 * was restarted, the pending bit will be back on and 755 * we just want to bail since the callout_reset would 756 * return 1 and our reference would have been removed 757 * by nd6_llinfo_settimer_locked above since canceled 758 * would have been 1. 759 */ 760 LLE_WUNLOCK(ln); 761 ND6_RUNLOCK(); 762 CURVNET_RESTORE(); 763 return; 764 } 765 ndi = ND_IFINFO(ifp); 766 send_ns = 0; 767 dst = &ln->r_l3addr.addr6; 768 pdst = dst; 769 770 if (ln->ln_ntick > 0) { 771 if (ln->ln_ntick > INT_MAX) { 772 ln->ln_ntick -= INT_MAX; 773 nd6_llinfo_settimer_locked(ln, INT_MAX); 774 } else { 775 ln->ln_ntick = 0; 776 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 777 } 778 goto done; 779 } 780 781 if (ln->la_flags & LLE_STATIC) { 782 goto done; 783 } 784 785 if (ln->la_flags & LLE_DELETED) { 786 nd6_free(&ln, 0); 787 goto done; 788 } 789 790 switch (ln->ln_state) { 791 case ND6_LLINFO_INCOMPLETE: 792 if (ln->la_asked < V_nd6_mmaxtries) { 793 ln->la_asked++; 794 send_ns = 1; 795 /* Send NS to multicast address */ 796 pdst = NULL; 797 } else { 798 struct mbuf *m = ln->la_hold; 799 if (m) { 800 struct mbuf *m0; 801 802 /* 803 * assuming every packet in la_hold has the 804 * same IP header. Send error after unlock. 805 */ 806 m0 = m->m_nextpkt; 807 m->m_nextpkt = NULL; 808 ln->la_hold = m0; 809 clear_llinfo_pqueue(ln); 810 } 811 nd6_free(&ln, 0); 812 if (m != NULL) 813 icmp6_error2(m, ICMP6_DST_UNREACH, 814 ICMP6_DST_UNREACH_ADDR, 0, ifp); 815 } 816 break; 817 case ND6_LLINFO_REACHABLE: 818 if (!ND6_LLINFO_PERMANENT(ln)) 819 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 820 break; 821 822 case ND6_LLINFO_STALE: 823 if (nd6_is_stale(ln, &delay, &do_switch) != 0) { 824 825 /* 826 * No packet has used this entry and GC timeout 827 * has not been passed. Reshedule timer and 828 * return. 829 */ 830 nd6_llinfo_settimer_locked(ln, delay); 831 break; 832 } 833 834 if (do_switch == 0) { 835 836 /* 837 * GC timer has ended and entry hasn't been used. 838 * Run Garbage collector (RFC 4861, 5.3) 839 */ 840 if (!ND6_LLINFO_PERMANENT(ln)) 841 nd6_free(&ln, 1); 842 break; 843 } 844 845 /* Entry has been used AND delay timer has ended. */ 846 847 /* FALLTHROUGH */ 848 849 case ND6_LLINFO_DELAY: 850 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 851 /* We need NUD */ 852 ln->la_asked = 1; 853 nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE); 854 send_ns = 1; 855 } else 856 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */ 857 break; 858 case ND6_LLINFO_PROBE: 859 if (ln->la_asked < V_nd6_umaxtries) { 860 ln->la_asked++; 861 send_ns = 1; 862 } else { 863 nd6_free(&ln, 0); 864 } 865 break; 866 default: 867 panic("%s: paths in a dark night can be confusing: %d", 868 __func__, ln->ln_state); 869 } 870 done: 871 if (ln != NULL) 872 ND6_RUNLOCK(); 873 if (send_ns != 0) { 874 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 875 psrc = nd6_llinfo_get_holdsrc(ln, &src); 876 LLE_FREE_LOCKED(ln); 877 ln = NULL; 878 nd6_ns_output(ifp, psrc, pdst, dst, NULL); 879 } 880 881 if (ln != NULL) 882 LLE_FREE_LOCKED(ln); 883 CURVNET_RESTORE(); 884 } 885 886 887 /* 888 * ND6 timer routine to expire default route list and prefix list 889 */ 890 void 891 nd6_timer(void *arg) 892 { 893 CURVNET_SET((struct vnet *) arg); 894 struct nd_drhead drq; 895 struct nd_defrouter *dr, *ndr; 896 struct nd_prefix *pr, *npr; 897 struct in6_ifaddr *ia6, *nia6; 898 899 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 900 nd6_timer, curvnet); 901 902 TAILQ_INIT(&drq); 903 904 /* expire default router list */ 905 ND6_WLOCK(); 906 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) 907 if (dr->expire && dr->expire < time_uptime) 908 defrouter_unlink(dr, &drq); 909 ND6_WUNLOCK(); 910 911 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 912 TAILQ_REMOVE(&drq, dr, dr_entry); 913 defrouter_del(dr); 914 } 915 916 /* 917 * expire interface addresses. 918 * in the past the loop was inside prefix expiry processing. 919 * However, from a stricter speci-confrmance standpoint, we should 920 * rather separate address lifetimes and prefix lifetimes. 921 * 922 * XXXRW: in6_ifaddrhead locking. 923 */ 924 addrloop: 925 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 926 /* check address lifetime */ 927 if (IFA6_IS_INVALID(ia6)) { 928 int regen = 0; 929 930 /* 931 * If the expiring address is temporary, try 932 * regenerating a new one. This would be useful when 933 * we suspended a laptop PC, then turned it on after a 934 * period that could invalidate all temporary 935 * addresses. Although we may have to restart the 936 * loop (see below), it must be after purging the 937 * address. Otherwise, we'd see an infinite loop of 938 * regeneration. 939 */ 940 if (V_ip6_use_tempaddr && 941 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 942 if (regen_tmpaddr(ia6) == 0) 943 regen = 1; 944 } 945 946 in6_purgeaddr(&ia6->ia_ifa); 947 948 if (regen) 949 goto addrloop; /* XXX: see below */ 950 } else if (IFA6_IS_DEPRECATED(ia6)) { 951 int oldflags = ia6->ia6_flags; 952 953 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 954 955 /* 956 * If a temporary address has just become deprecated, 957 * regenerate a new one if possible. 958 */ 959 if (V_ip6_use_tempaddr && 960 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 961 (oldflags & IN6_IFF_DEPRECATED) == 0) { 962 963 if (regen_tmpaddr(ia6) == 0) { 964 /* 965 * A new temporary address is 966 * generated. 967 * XXX: this means the address chain 968 * has changed while we are still in 969 * the loop. Although the change 970 * would not cause disaster (because 971 * it's not a deletion, but an 972 * addition,) we'd rather restart the 973 * loop just for safety. Or does this 974 * significantly reduce performance?? 975 */ 976 goto addrloop; 977 } 978 } 979 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) { 980 /* 981 * Schedule DAD for a tentative address. This happens 982 * if the interface was down or not running 983 * when the address was configured. 984 */ 985 int delay; 986 987 delay = arc4random() % 988 (MAX_RTR_SOLICITATION_DELAY * hz); 989 nd6_dad_start((struct ifaddr *)ia6, delay); 990 } else { 991 /* 992 * Check status of the interface. If it is down, 993 * mark the address as tentative for future DAD. 994 */ 995 if ((ia6->ia_ifp->if_flags & IFF_UP) == 0 || 996 (ia6->ia_ifp->if_drv_flags & IFF_DRV_RUNNING) 997 == 0 || 998 (ND_IFINFO(ia6->ia_ifp)->flags & 999 ND6_IFF_IFDISABLED) != 0) { 1000 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED; 1001 ia6->ia6_flags |= IN6_IFF_TENTATIVE; 1002 } 1003 /* 1004 * A new RA might have made a deprecated address 1005 * preferred. 1006 */ 1007 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 1008 } 1009 } 1010 1011 /* expire prefix list */ 1012 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1013 /* 1014 * check prefix lifetime. 1015 * since pltime is just for autoconf, pltime processing for 1016 * prefix is not necessary. 1017 */ 1018 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 1019 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { 1020 1021 /* 1022 * address expiration and prefix expiration are 1023 * separate. NEVER perform in6_purgeaddr here. 1024 */ 1025 prelist_remove(pr); 1026 } 1027 } 1028 CURVNET_RESTORE(); 1029 } 1030 1031 /* 1032 * ia6 - deprecated/invalidated temporary address 1033 */ 1034 static int 1035 regen_tmpaddr(struct in6_ifaddr *ia6) 1036 { 1037 struct ifaddr *ifa; 1038 struct ifnet *ifp; 1039 struct in6_ifaddr *public_ifa6 = NULL; 1040 1041 ifp = ia6->ia_ifa.ifa_ifp; 1042 IF_ADDR_RLOCK(ifp); 1043 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1044 struct in6_ifaddr *it6; 1045 1046 if (ifa->ifa_addr->sa_family != AF_INET6) 1047 continue; 1048 1049 it6 = (struct in6_ifaddr *)ifa; 1050 1051 /* ignore no autoconf addresses. */ 1052 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1053 continue; 1054 1055 /* ignore autoconf addresses with different prefixes. */ 1056 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 1057 continue; 1058 1059 /* 1060 * Now we are looking at an autoconf address with the same 1061 * prefix as ours. If the address is temporary and is still 1062 * preferred, do not create another one. It would be rare, but 1063 * could happen, for example, when we resume a laptop PC after 1064 * a long period. 1065 */ 1066 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 1067 !IFA6_IS_DEPRECATED(it6)) { 1068 public_ifa6 = NULL; 1069 break; 1070 } 1071 1072 /* 1073 * This is a public autoconf address that has the same prefix 1074 * as ours. If it is preferred, keep it. We can't break the 1075 * loop here, because there may be a still-preferred temporary 1076 * address with the prefix. 1077 */ 1078 if (!IFA6_IS_DEPRECATED(it6)) 1079 public_ifa6 = it6; 1080 } 1081 if (public_ifa6 != NULL) 1082 ifa_ref(&public_ifa6->ia_ifa); 1083 IF_ADDR_RUNLOCK(ifp); 1084 1085 if (public_ifa6 != NULL) { 1086 int e; 1087 1088 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 1089 ifa_free(&public_ifa6->ia_ifa); 1090 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 1091 " tmp addr,errno=%d\n", e); 1092 return (-1); 1093 } 1094 ifa_free(&public_ifa6->ia_ifa); 1095 return (0); 1096 } 1097 1098 return (-1); 1099 } 1100 1101 /* 1102 * Remove prefix and default router list entries corresponding to ifp. Neighbor 1103 * cache entries are freed in in6_domifdetach(). 1104 */ 1105 void 1106 nd6_purge(struct ifnet *ifp) 1107 { 1108 struct nd_drhead drq; 1109 struct nd_defrouter *dr, *ndr; 1110 struct nd_prefix *pr, *npr; 1111 1112 TAILQ_INIT(&drq); 1113 1114 /* 1115 * Nuke default router list entries toward ifp. 1116 * We defer removal of default router list entries that is installed 1117 * in the routing table, in order to keep additional side effects as 1118 * small as possible. 1119 */ 1120 ND6_WLOCK(); 1121 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 1122 if (dr->installed) 1123 continue; 1124 if (dr->ifp == ifp) 1125 defrouter_unlink(dr, &drq); 1126 } 1127 1128 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 1129 if (!dr->installed) 1130 continue; 1131 if (dr->ifp == ifp) 1132 defrouter_unlink(dr, &drq); 1133 } 1134 ND6_WUNLOCK(); 1135 1136 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 1137 TAILQ_REMOVE(&drq, dr, dr_entry); 1138 defrouter_del(dr); 1139 } 1140 1141 /* Nuke prefix list entries toward ifp */ 1142 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 1143 if (pr->ndpr_ifp == ifp) { 1144 /* 1145 * Because if_detach() does *not* release prefixes 1146 * while purging addresses the reference count will 1147 * still be above zero. We therefore reset it to 1148 * make sure that the prefix really gets purged. 1149 */ 1150 pr->ndpr_refcnt = 0; 1151 1152 prelist_remove(pr); 1153 } 1154 } 1155 1156 /* cancel default outgoing interface setting */ 1157 if (V_nd6_defifindex == ifp->if_index) 1158 nd6_setdefaultiface(0); 1159 1160 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1161 /* Refresh default router list. */ 1162 defrouter_select(); 1163 } 1164 } 1165 1166 /* 1167 * the caller acquires and releases the lock on the lltbls 1168 * Returns the llentry locked 1169 */ 1170 struct llentry * 1171 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1172 { 1173 struct sockaddr_in6 sin6; 1174 struct llentry *ln; 1175 1176 bzero(&sin6, sizeof(sin6)); 1177 sin6.sin6_len = sizeof(struct sockaddr_in6); 1178 sin6.sin6_family = AF_INET6; 1179 sin6.sin6_addr = *addr6; 1180 1181 IF_AFDATA_LOCK_ASSERT(ifp); 1182 1183 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6); 1184 1185 return (ln); 1186 } 1187 1188 struct llentry * 1189 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp) 1190 { 1191 struct sockaddr_in6 sin6; 1192 struct llentry *ln; 1193 1194 bzero(&sin6, sizeof(sin6)); 1195 sin6.sin6_len = sizeof(struct sockaddr_in6); 1196 sin6.sin6_family = AF_INET6; 1197 sin6.sin6_addr = *addr6; 1198 1199 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 1200 if (ln != NULL) 1201 ln->ln_state = ND6_LLINFO_NOSTATE; 1202 1203 return (ln); 1204 } 1205 1206 /* 1207 * Test whether a given IPv6 address is a neighbor or not, ignoring 1208 * the actual neighbor cache. The neighbor cache is ignored in order 1209 * to not reenter the routing code from within itself. 1210 */ 1211 static int 1212 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1213 { 1214 struct nd_prefix *pr; 1215 struct ifaddr *dstaddr; 1216 struct rt_addrinfo info; 1217 struct sockaddr_in6 rt_key; 1218 struct sockaddr *dst6; 1219 int fibnum; 1220 1221 /* 1222 * A link-local address is always a neighbor. 1223 * XXX: a link does not necessarily specify a single interface. 1224 */ 1225 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 1226 struct sockaddr_in6 sin6_copy; 1227 u_int32_t zone; 1228 1229 /* 1230 * We need sin6_copy since sa6_recoverscope() may modify the 1231 * content (XXX). 1232 */ 1233 sin6_copy = *addr; 1234 if (sa6_recoverscope(&sin6_copy)) 1235 return (0); /* XXX: should be impossible */ 1236 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 1237 return (0); 1238 if (sin6_copy.sin6_scope_id == zone) 1239 return (1); 1240 else 1241 return (0); 1242 } 1243 1244 bzero(&rt_key, sizeof(rt_key)); 1245 bzero(&info, sizeof(info)); 1246 info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key; 1247 1248 /* Always use the default FIB here. XXME - why? */ 1249 fibnum = RT_DEFAULT_FIB; 1250 1251 /* 1252 * If the address matches one of our addresses, 1253 * it should be a neighbor. 1254 * If the address matches one of our on-link prefixes, it should be a 1255 * neighbor. 1256 */ 1257 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1258 if (pr->ndpr_ifp != ifp) 1259 continue; 1260 1261 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 1262 1263 /* Always use the default FIB here. */ 1264 dst6 = (struct sockaddr *)&pr->ndpr_prefix; 1265 1266 /* Restore length field before retrying lookup */ 1267 rt_key.sin6_len = sizeof(rt_key); 1268 if (rib_lookup_info(fibnum, dst6, 0, 0, &info) != 0) 1269 continue; 1270 /* 1271 * This is the case where multiple interfaces 1272 * have the same prefix, but only one is installed 1273 * into the routing table and that prefix entry 1274 * is not the one being examined here. In the case 1275 * where RADIX_MPATH is enabled, multiple route 1276 * entries (of the same rt_key value) will be 1277 * installed because the interface addresses all 1278 * differ. 1279 */ 1280 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1281 &rt_key.sin6_addr)) 1282 continue; 1283 } 1284 1285 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1286 &addr->sin6_addr, &pr->ndpr_mask)) 1287 return (1); 1288 } 1289 1290 /* 1291 * If the address is assigned on the node of the other side of 1292 * a p2p interface, the address should be a neighbor. 1293 */ 1294 dstaddr = ifa_ifwithdstaddr((const struct sockaddr *)addr, RT_ALL_FIBS); 1295 if (dstaddr != NULL) { 1296 if (dstaddr->ifa_ifp == ifp) { 1297 ifa_free(dstaddr); 1298 return (1); 1299 } 1300 ifa_free(dstaddr); 1301 } 1302 1303 /* 1304 * If the default router list is empty, all addresses are regarded 1305 * as on-link, and thus, as a neighbor. 1306 */ 1307 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1308 TAILQ_EMPTY(&V_nd_defrouter) && 1309 V_nd6_defifindex == ifp->if_index) { 1310 return (1); 1311 } 1312 1313 return (0); 1314 } 1315 1316 1317 /* 1318 * Detect if a given IPv6 address identifies a neighbor on a given link. 1319 * XXX: should take care of the destination of a p2p link? 1320 */ 1321 int 1322 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp) 1323 { 1324 struct llentry *lle; 1325 int rc = 0; 1326 1327 IF_AFDATA_UNLOCK_ASSERT(ifp); 1328 if (nd6_is_new_addr_neighbor(addr, ifp)) 1329 return (1); 1330 1331 /* 1332 * Even if the address matches none of our addresses, it might be 1333 * in the neighbor cache. 1334 */ 1335 IF_AFDATA_RLOCK(ifp); 1336 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1337 LLE_RUNLOCK(lle); 1338 rc = 1; 1339 } 1340 IF_AFDATA_RUNLOCK(ifp); 1341 return (rc); 1342 } 1343 1344 /* 1345 * Free an nd6 llinfo entry. 1346 * Since the function would cause significant changes in the kernel, DO NOT 1347 * make it global, unless you have a strong reason for the change, and are sure 1348 * that the change is safe. 1349 * 1350 * Set noinline to be dtrace-friendly 1351 */ 1352 static __noinline void 1353 nd6_free(struct llentry **lnp, int gc) 1354 { 1355 struct ifnet *ifp; 1356 struct llentry *ln; 1357 struct nd_defrouter *dr; 1358 1359 ln = *lnp; 1360 *lnp = NULL; 1361 1362 LLE_WLOCK_ASSERT(ln); 1363 ND6_RLOCK_ASSERT(); 1364 1365 ifp = lltable_get_ifp(ln->lle_tbl); 1366 if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0) 1367 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp); 1368 else 1369 dr = NULL; 1370 ND6_RUNLOCK(); 1371 1372 if ((ln->la_flags & LLE_DELETED) == 0) 1373 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 1374 1375 /* 1376 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1377 * even though it is not harmful, it was not really necessary. 1378 */ 1379 1380 /* cancel timer */ 1381 nd6_llinfo_settimer_locked(ln, -1); 1382 1383 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1384 if (dr != NULL && dr->expire && 1385 ln->ln_state == ND6_LLINFO_STALE && gc) { 1386 /* 1387 * If the reason for the deletion is just garbage 1388 * collection, and the neighbor is an active default 1389 * router, do not delete it. Instead, reset the GC 1390 * timer using the router's lifetime. 1391 * Simply deleting the entry would affect default 1392 * router selection, which is not necessarily a good 1393 * thing, especially when we're using router preference 1394 * values. 1395 * XXX: the check for ln_state would be redundant, 1396 * but we intentionally keep it just in case. 1397 */ 1398 if (dr->expire > time_uptime) 1399 nd6_llinfo_settimer_locked(ln, 1400 (dr->expire - time_uptime) * hz); 1401 else 1402 nd6_llinfo_settimer_locked(ln, 1403 (long)V_nd6_gctimer * hz); 1404 1405 LLE_REMREF(ln); 1406 LLE_WUNLOCK(ln); 1407 defrouter_rele(dr); 1408 return; 1409 } 1410 1411 if (dr) { 1412 /* 1413 * Unreachablity of a router might affect the default 1414 * router selection and on-link detection of advertised 1415 * prefixes. 1416 */ 1417 1418 /* 1419 * Temporarily fake the state to choose a new default 1420 * router and to perform on-link determination of 1421 * prefixes correctly. 1422 * Below the state will be set correctly, 1423 * or the entry itself will be deleted. 1424 */ 1425 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1426 } 1427 1428 if (ln->ln_router || dr) { 1429 1430 /* 1431 * We need to unlock to avoid a LOR with rt6_flush() with the 1432 * rnh and for the calls to pfxlist_onlink_check() and 1433 * defrouter_select() in the block further down for calls 1434 * into nd6_lookup(). We still hold a ref. 1435 */ 1436 LLE_WUNLOCK(ln); 1437 1438 /* 1439 * rt6_flush must be called whether or not the neighbor 1440 * is in the Default Router List. 1441 * See a corresponding comment in nd6_na_input(). 1442 */ 1443 rt6_flush(&ln->r_l3addr.addr6, ifp); 1444 } 1445 1446 if (dr) { 1447 /* 1448 * Since defrouter_select() does not affect the 1449 * on-link determination and MIP6 needs the check 1450 * before the default router selection, we perform 1451 * the check now. 1452 */ 1453 pfxlist_onlink_check(); 1454 1455 /* 1456 * Refresh default router list. 1457 */ 1458 defrouter_select(); 1459 } 1460 1461 /* 1462 * If this entry was added by an on-link redirect, remove the 1463 * corresponding host route. 1464 */ 1465 if (ln->la_flags & LLE_REDIRECT) 1466 nd6_free_redirect(ln); 1467 1468 if (ln->ln_router || dr) 1469 LLE_WLOCK(ln); 1470 } 1471 1472 /* 1473 * Save to unlock. We still hold an extra reference and will not 1474 * free(9) in llentry_free() if someone else holds one as well. 1475 */ 1476 LLE_WUNLOCK(ln); 1477 IF_AFDATA_LOCK(ifp); 1478 LLE_WLOCK(ln); 1479 /* Guard against race with other llentry_free(). */ 1480 if (ln->la_flags & LLE_LINKED) { 1481 /* Remove callout reference */ 1482 LLE_REMREF(ln); 1483 lltable_unlink_entry(ln->lle_tbl, ln); 1484 } 1485 IF_AFDATA_UNLOCK(ifp); 1486 1487 llentry_free(ln); 1488 if (dr != NULL) 1489 defrouter_rele(dr); 1490 } 1491 1492 static int 1493 nd6_isdynrte(const struct rtentry *rt, void *xap) 1494 { 1495 1496 if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC)) 1497 return (1); 1498 1499 return (0); 1500 } 1501 /* 1502 * Remove the rtentry for the given llentry, 1503 * both of which were installed by a redirect. 1504 */ 1505 static void 1506 nd6_free_redirect(const struct llentry *ln) 1507 { 1508 int fibnum; 1509 struct sockaddr_in6 sin6; 1510 struct rt_addrinfo info; 1511 1512 lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6); 1513 memset(&info, 0, sizeof(info)); 1514 info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6; 1515 info.rti_filter = nd6_isdynrte; 1516 1517 for (fibnum = 0; fibnum < rt_numfibs; fibnum++) 1518 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum); 1519 } 1520 1521 /* 1522 * Rejuvenate this function for routing operations related 1523 * processing. 1524 */ 1525 void 1526 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1527 { 1528 struct sockaddr_in6 *gateway; 1529 struct nd_defrouter *dr; 1530 struct ifnet *ifp; 1531 1532 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1533 ifp = rt->rt_ifp; 1534 1535 switch (req) { 1536 case RTM_ADD: 1537 break; 1538 1539 case RTM_DELETE: 1540 if (!ifp) 1541 return; 1542 /* 1543 * Only indirect routes are interesting. 1544 */ 1545 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1546 return; 1547 /* 1548 * check for default route 1549 */ 1550 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1551 &SIN6(rt_key(rt))->sin6_addr)) { 1552 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1553 if (dr != NULL) { 1554 dr->installed = 0; 1555 defrouter_rele(dr); 1556 } 1557 } 1558 break; 1559 } 1560 } 1561 1562 1563 int 1564 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1565 { 1566 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1567 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1568 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1569 int error = 0; 1570 1571 if (ifp->if_afdata[AF_INET6] == NULL) 1572 return (EPFNOSUPPORT); 1573 switch (cmd) { 1574 case OSIOCGIFINFO_IN6: 1575 #define ND ndi->ndi 1576 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1577 bzero(&ND, sizeof(ND)); 1578 ND.linkmtu = IN6_LINKMTU(ifp); 1579 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1580 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1581 ND.reachable = ND_IFINFO(ifp)->reachable; 1582 ND.retrans = ND_IFINFO(ifp)->retrans; 1583 ND.flags = ND_IFINFO(ifp)->flags; 1584 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1585 ND.chlim = ND_IFINFO(ifp)->chlim; 1586 break; 1587 case SIOCGIFINFO_IN6: 1588 ND = *ND_IFINFO(ifp); 1589 break; 1590 case SIOCSIFINFO_IN6: 1591 /* 1592 * used to change host variables from userland. 1593 * intended for a use on router to reflect RA configurations. 1594 */ 1595 /* 0 means 'unspecified' */ 1596 if (ND.linkmtu != 0) { 1597 if (ND.linkmtu < IPV6_MMTU || 1598 ND.linkmtu > IN6_LINKMTU(ifp)) { 1599 error = EINVAL; 1600 break; 1601 } 1602 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1603 } 1604 1605 if (ND.basereachable != 0) { 1606 int obasereachable = ND_IFINFO(ifp)->basereachable; 1607 1608 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1609 if (ND.basereachable != obasereachable) 1610 ND_IFINFO(ifp)->reachable = 1611 ND_COMPUTE_RTIME(ND.basereachable); 1612 } 1613 if (ND.retrans != 0) 1614 ND_IFINFO(ifp)->retrans = ND.retrans; 1615 if (ND.chlim != 0) 1616 ND_IFINFO(ifp)->chlim = ND.chlim; 1617 /* FALLTHROUGH */ 1618 case SIOCSIFINFO_FLAGS: 1619 { 1620 struct ifaddr *ifa; 1621 struct in6_ifaddr *ia; 1622 1623 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1624 !(ND.flags & ND6_IFF_IFDISABLED)) { 1625 /* ifdisabled 1->0 transision */ 1626 1627 /* 1628 * If the interface is marked as ND6_IFF_IFDISABLED and 1629 * has an link-local address with IN6_IFF_DUPLICATED, 1630 * do not clear ND6_IFF_IFDISABLED. 1631 * See RFC 4862, Section 5.4.5. 1632 */ 1633 IF_ADDR_RLOCK(ifp); 1634 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1635 if (ifa->ifa_addr->sa_family != AF_INET6) 1636 continue; 1637 ia = (struct in6_ifaddr *)ifa; 1638 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1639 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1640 break; 1641 } 1642 IF_ADDR_RUNLOCK(ifp); 1643 1644 if (ifa != NULL) { 1645 /* LLA is duplicated. */ 1646 ND.flags |= ND6_IFF_IFDISABLED; 1647 log(LOG_ERR, "Cannot enable an interface" 1648 " with a link-local address marked" 1649 " duplicate.\n"); 1650 } else { 1651 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1652 if (ifp->if_flags & IFF_UP) 1653 in6_if_up(ifp); 1654 } 1655 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1656 (ND.flags & ND6_IFF_IFDISABLED)) { 1657 /* ifdisabled 0->1 transision */ 1658 /* Mark all IPv6 address as tentative. */ 1659 1660 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1661 if (V_ip6_dad_count > 0 && 1662 (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) { 1663 IF_ADDR_RLOCK(ifp); 1664 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1665 ifa_link) { 1666 if (ifa->ifa_addr->sa_family != 1667 AF_INET6) 1668 continue; 1669 ia = (struct in6_ifaddr *)ifa; 1670 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1671 } 1672 IF_ADDR_RUNLOCK(ifp); 1673 } 1674 } 1675 1676 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1677 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1678 /* auto_linklocal 0->1 transision */ 1679 1680 /* If no link-local address on ifp, configure */ 1681 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1682 in6_ifattach(ifp, NULL); 1683 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1684 ifp->if_flags & IFF_UP) { 1685 /* 1686 * When the IF already has 1687 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1688 * address is assigned, and IFF_UP, try to 1689 * assign one. 1690 */ 1691 IF_ADDR_RLOCK(ifp); 1692 TAILQ_FOREACH(ifa, &ifp->if_addrhead, 1693 ifa_link) { 1694 if (ifa->ifa_addr->sa_family != 1695 AF_INET6) 1696 continue; 1697 ia = (struct in6_ifaddr *)ifa; 1698 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) 1699 break; 1700 } 1701 IF_ADDR_RUNLOCK(ifp); 1702 if (ifa != NULL) 1703 /* No LLA is configured. */ 1704 in6_ifattach(ifp, NULL); 1705 } 1706 } 1707 } 1708 ND_IFINFO(ifp)->flags = ND.flags; 1709 break; 1710 #undef ND 1711 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1712 /* sync kernel routing table with the default router list */ 1713 defrouter_reset(); 1714 defrouter_select(); 1715 break; 1716 case SIOCSPFXFLUSH_IN6: 1717 { 1718 /* flush all the prefix advertised by routers */ 1719 struct nd_prefix *pr, *next; 1720 1721 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1722 struct in6_ifaddr *ia, *ia_next; 1723 1724 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1725 continue; /* XXX */ 1726 1727 /* do we really have to remove addresses as well? */ 1728 /* XXXRW: in6_ifaddrhead locking. */ 1729 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1730 ia_next) { 1731 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1732 continue; 1733 1734 if (ia->ia6_ndpr == pr) 1735 in6_purgeaddr(&ia->ia_ifa); 1736 } 1737 prelist_remove(pr); 1738 } 1739 break; 1740 } 1741 case SIOCSRTRFLUSH_IN6: 1742 { 1743 /* flush all the default routers */ 1744 struct nd_drhead drq; 1745 struct nd_defrouter *dr; 1746 1747 TAILQ_INIT(&drq); 1748 1749 defrouter_reset(); 1750 1751 ND6_WLOCK(); 1752 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL) 1753 defrouter_unlink(dr, &drq); 1754 ND6_WUNLOCK(); 1755 while ((dr = TAILQ_FIRST(&drq)) != NULL) { 1756 TAILQ_REMOVE(&drq, dr, dr_entry); 1757 defrouter_del(dr); 1758 } 1759 1760 defrouter_select(); 1761 break; 1762 } 1763 case SIOCGNBRINFO_IN6: 1764 { 1765 struct llentry *ln; 1766 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1767 1768 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1769 return (error); 1770 1771 IF_AFDATA_RLOCK(ifp); 1772 ln = nd6_lookup(&nb_addr, 0, ifp); 1773 IF_AFDATA_RUNLOCK(ifp); 1774 1775 if (ln == NULL) { 1776 error = EINVAL; 1777 break; 1778 } 1779 nbi->state = ln->ln_state; 1780 nbi->asked = ln->la_asked; 1781 nbi->isrouter = ln->ln_router; 1782 if (ln->la_expire == 0) 1783 nbi->expire = 0; 1784 else 1785 nbi->expire = ln->la_expire + ln->lle_remtime / hz + 1786 (time_second - time_uptime); 1787 LLE_RUNLOCK(ln); 1788 break; 1789 } 1790 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1791 ndif->ifindex = V_nd6_defifindex; 1792 break; 1793 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1794 return (nd6_setdefaultiface(ndif->ifindex)); 1795 } 1796 return (error); 1797 } 1798 1799 /* 1800 * Calculates new isRouter value based on provided parameters and 1801 * returns it. 1802 */ 1803 static int 1804 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr, 1805 int ln_router) 1806 { 1807 1808 /* 1809 * ICMP6 type dependent behavior. 1810 * 1811 * NS: clear IsRouter if new entry 1812 * RS: clear IsRouter 1813 * RA: set IsRouter if there's lladdr 1814 * redir: clear IsRouter if new entry 1815 * 1816 * RA case, (1): 1817 * The spec says that we must set IsRouter in the following cases: 1818 * - If lladdr exist, set IsRouter. This means (1-5). 1819 * - If it is old entry (!newentry), set IsRouter. This means (7). 1820 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1821 * A quetion arises for (1) case. (1) case has no lladdr in the 1822 * neighbor cache, this is similar to (6). 1823 * This case is rare but we figured that we MUST NOT set IsRouter. 1824 * 1825 * is_new old_addr new_addr NS RS RA redir 1826 * D R 1827 * 0 n n (1) c ? s 1828 * 0 y n (2) c s s 1829 * 0 n y (3) c s s 1830 * 0 y y (4) c s s 1831 * 0 y y (5) c s s 1832 * 1 -- n (6) c c c s 1833 * 1 -- y (7) c c s c s 1834 * 1835 * (c=clear s=set) 1836 */ 1837 switch (type & 0xff) { 1838 case ND_NEIGHBOR_SOLICIT: 1839 /* 1840 * New entry must have is_router flag cleared. 1841 */ 1842 if (is_new) /* (6-7) */ 1843 ln_router = 0; 1844 break; 1845 case ND_REDIRECT: 1846 /* 1847 * If the icmp is a redirect to a better router, always set the 1848 * is_router flag. Otherwise, if the entry is newly created, 1849 * clear the flag. [RFC 2461, sec 8.3] 1850 */ 1851 if (code == ND_REDIRECT_ROUTER) 1852 ln_router = 1; 1853 else { 1854 if (is_new) /* (6-7) */ 1855 ln_router = 0; 1856 } 1857 break; 1858 case ND_ROUTER_SOLICIT: 1859 /* 1860 * is_router flag must always be cleared. 1861 */ 1862 ln_router = 0; 1863 break; 1864 case ND_ROUTER_ADVERT: 1865 /* 1866 * Mark an entry with lladdr as a router. 1867 */ 1868 if ((!is_new && (old_addr || new_addr)) || /* (2-5) */ 1869 (is_new && new_addr)) { /* (7) */ 1870 ln_router = 1; 1871 } 1872 break; 1873 } 1874 1875 return (ln_router); 1876 } 1877 1878 /* 1879 * Create neighbor cache entry and cache link-layer address, 1880 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1881 * 1882 * type - ICMP6 type 1883 * code - type dependent information 1884 * 1885 */ 1886 void 1887 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1888 int lladdrlen, int type, int code) 1889 { 1890 struct llentry *ln = NULL, *ln_tmp; 1891 int is_newentry; 1892 int do_update; 1893 int olladdr; 1894 int llchange; 1895 int flags; 1896 uint16_t router = 0; 1897 struct sockaddr_in6 sin6; 1898 struct mbuf *chain = NULL; 1899 u_char linkhdr[LLE_MAX_LINKHDR]; 1900 size_t linkhdrsize; 1901 int lladdr_off; 1902 1903 IF_AFDATA_UNLOCK_ASSERT(ifp); 1904 1905 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1906 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1907 1908 /* nothing must be updated for unspecified address */ 1909 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1910 return; 1911 1912 /* 1913 * Validation about ifp->if_addrlen and lladdrlen must be done in 1914 * the caller. 1915 * 1916 * XXX If the link does not have link-layer adderss, what should 1917 * we do? (ifp->if_addrlen == 0) 1918 * Spec says nothing in sections for RA, RS and NA. There's small 1919 * description on it in NS section (RFC 2461 7.2.3). 1920 */ 1921 flags = lladdr ? LLE_EXCLUSIVE : 0; 1922 IF_AFDATA_RLOCK(ifp); 1923 ln = nd6_lookup(from, flags, ifp); 1924 IF_AFDATA_RUNLOCK(ifp); 1925 is_newentry = 0; 1926 if (ln == NULL) { 1927 flags |= LLE_EXCLUSIVE; 1928 ln = nd6_alloc(from, 0, ifp); 1929 if (ln == NULL) 1930 return; 1931 1932 /* 1933 * Since we already know all the data for the new entry, 1934 * fill it before insertion. 1935 */ 1936 if (lladdr != NULL) { 1937 linkhdrsize = sizeof(linkhdr); 1938 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 1939 linkhdr, &linkhdrsize, &lladdr_off) != 0) 1940 return; 1941 lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 1942 lladdr_off); 1943 } 1944 1945 IF_AFDATA_WLOCK(ifp); 1946 LLE_WLOCK(ln); 1947 /* Prefer any existing lle over newly-created one */ 1948 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp); 1949 if (ln_tmp == NULL) 1950 lltable_link_entry(LLTABLE6(ifp), ln); 1951 IF_AFDATA_WUNLOCK(ifp); 1952 if (ln_tmp == NULL) { 1953 /* No existing lle, mark as new entry (6,7) */ 1954 is_newentry = 1; 1955 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 1956 if (lladdr != NULL) /* (7) */ 1957 EVENTHANDLER_INVOKE(lle_event, ln, 1958 LLENTRY_RESOLVED); 1959 } else { 1960 lltable_free_entry(LLTABLE6(ifp), ln); 1961 ln = ln_tmp; 1962 ln_tmp = NULL; 1963 } 1964 } 1965 /* do nothing if static ndp is set */ 1966 if ((ln->la_flags & LLE_STATIC)) { 1967 if (flags & LLE_EXCLUSIVE) 1968 LLE_WUNLOCK(ln); 1969 else 1970 LLE_RUNLOCK(ln); 1971 return; 1972 } 1973 1974 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1975 if (olladdr && lladdr) { 1976 llchange = bcmp(lladdr, ln->ll_addr, 1977 ifp->if_addrlen); 1978 } else if (!olladdr && lladdr) 1979 llchange = 1; 1980 else 1981 llchange = 0; 1982 1983 /* 1984 * newentry olladdr lladdr llchange (*=record) 1985 * 0 n n -- (1) 1986 * 0 y n -- (2) 1987 * 0 n y y (3) * STALE 1988 * 0 y y n (4) * 1989 * 0 y y y (5) * STALE 1990 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1991 * 1 -- y -- (7) * STALE 1992 */ 1993 1994 do_update = 0; 1995 if (is_newentry == 0 && llchange != 0) { 1996 do_update = 1; /* (3,5) */ 1997 1998 /* 1999 * Record source link-layer address 2000 * XXX is it dependent to ifp->if_type? 2001 */ 2002 linkhdrsize = sizeof(linkhdr); 2003 if (lltable_calc_llheader(ifp, AF_INET6, lladdr, 2004 linkhdr, &linkhdrsize, &lladdr_off) != 0) 2005 return; 2006 2007 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize, 2008 lladdr_off) == 0) { 2009 /* Entry was deleted */ 2010 return; 2011 } 2012 2013 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); 2014 2015 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2016 2017 if (ln->la_hold != NULL) 2018 nd6_grab_holdchain(ln, &chain, &sin6); 2019 } 2020 2021 /* Calculates new router status */ 2022 router = nd6_is_router(type, code, is_newentry, olladdr, 2023 lladdr != NULL ? 1 : 0, ln->ln_router); 2024 2025 ln->ln_router = router; 2026 /* Mark non-router redirects with special flag */ 2027 if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER) 2028 ln->la_flags |= LLE_REDIRECT; 2029 2030 if (flags & LLE_EXCLUSIVE) 2031 LLE_WUNLOCK(ln); 2032 else 2033 LLE_RUNLOCK(ln); 2034 2035 if (chain != NULL) 2036 nd6_flush_holdchain(ifp, ifp, chain, &sin6); 2037 2038 /* 2039 * When the link-layer address of a router changes, select the 2040 * best router again. In particular, when the neighbor entry is newly 2041 * created, it might affect the selection policy. 2042 * Question: can we restrict the first condition to the "is_newentry" 2043 * case? 2044 * XXX: when we hear an RA from a new router with the link-layer 2045 * address option, defrouter_select() is called twice, since 2046 * defrtrlist_update called the function as well. However, I believe 2047 * we can compromise the overhead, since it only happens the first 2048 * time. 2049 * XXX: although defrouter_select() should not have a bad effect 2050 * for those are not autoconfigured hosts, we explicitly avoid such 2051 * cases for safety. 2052 */ 2053 if ((do_update || is_newentry) && router && 2054 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 2055 /* 2056 * guaranteed recursion 2057 */ 2058 defrouter_select(); 2059 } 2060 } 2061 2062 static void 2063 nd6_slowtimo(void *arg) 2064 { 2065 CURVNET_SET((struct vnet *) arg); 2066 struct nd_ifinfo *nd6if; 2067 struct ifnet *ifp; 2068 2069 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 2070 nd6_slowtimo, curvnet); 2071 IFNET_RLOCK_NOSLEEP(); 2072 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 2073 if (ifp->if_afdata[AF_INET6] == NULL) 2074 continue; 2075 nd6if = ND_IFINFO(ifp); 2076 if (nd6if->basereachable && /* already initialized */ 2077 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 2078 /* 2079 * Since reachable time rarely changes by router 2080 * advertisements, we SHOULD insure that a new random 2081 * value gets recomputed at least once every few hours. 2082 * (RFC 2461, 6.3.4) 2083 */ 2084 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 2085 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 2086 } 2087 } 2088 IFNET_RUNLOCK_NOSLEEP(); 2089 CURVNET_RESTORE(); 2090 } 2091 2092 void 2093 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 2094 struct sockaddr_in6 *sin6) 2095 { 2096 2097 LLE_WLOCK_ASSERT(ln); 2098 2099 *chain = ln->la_hold; 2100 ln->la_hold = NULL; 2101 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 2102 2103 if (ln->ln_state == ND6_LLINFO_STALE) { 2104 2105 /* 2106 * The first time we send a packet to a 2107 * neighbor whose entry is STALE, we have 2108 * to change the state to DELAY and a sets 2109 * a timer to expire in DELAY_FIRST_PROBE_TIME 2110 * seconds to ensure do neighbor unreachability 2111 * detection on expiration. 2112 * (RFC 2461 7.3.3) 2113 */ 2114 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY); 2115 } 2116 } 2117 2118 int 2119 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2120 struct sockaddr_in6 *dst, struct route *ro) 2121 { 2122 int error; 2123 int ip6len; 2124 struct ip6_hdr *ip6; 2125 struct m_tag *mtag; 2126 2127 #ifdef MAC 2128 mac_netinet6_nd6_send(ifp, m); 2129 #endif 2130 2131 /* 2132 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 2133 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 2134 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 2135 * to be diverted to user space. When re-injected into the kernel, 2136 * send_output() will directly dispatch them to the outgoing interface. 2137 */ 2138 if (send_sendso_input_hook != NULL) { 2139 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 2140 if (mtag != NULL) { 2141 ip6 = mtod(m, struct ip6_hdr *); 2142 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 2143 /* Use the SEND socket */ 2144 error = send_sendso_input_hook(m, ifp, SND_OUT, 2145 ip6len); 2146 /* -1 == no app on SEND socket */ 2147 if (error == 0 || error != -1) 2148 return (error); 2149 } 2150 } 2151 2152 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 2153 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 2154 mtod(m, struct ip6_hdr *)); 2155 2156 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 2157 origifp = ifp; 2158 2159 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro); 2160 return (error); 2161 } 2162 2163 /* 2164 * Lookup link headerfor @sa_dst address. Stores found 2165 * data in @desten buffer. Copy of lle ln_flags can be also 2166 * saved in @pflags if @pflags is non-NULL. 2167 * 2168 * If destination LLE does not exists or lle state modification 2169 * is required, call "slow" version. 2170 * 2171 * Return values: 2172 * - 0 on success (address copied to buffer). 2173 * - EWOULDBLOCK (no local error, but address is still unresolved) 2174 * - other errors (alloc failure, etc) 2175 */ 2176 int 2177 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m, 2178 const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags) 2179 { 2180 struct llentry *ln = NULL; 2181 const struct sockaddr_in6 *dst6; 2182 2183 if (pflags != NULL) 2184 *pflags = 0; 2185 2186 dst6 = (const struct sockaddr_in6 *)sa_dst; 2187 2188 /* discard the packet if IPv6 operation is disabled on the interface */ 2189 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2190 m_freem(m); 2191 return (ENETDOWN); /* better error? */ 2192 } 2193 2194 if (m != NULL && m->m_flags & M_MCAST) { 2195 switch (ifp->if_type) { 2196 case IFT_ETHER: 2197 case IFT_FDDI: 2198 case IFT_L2VLAN: 2199 case IFT_IEEE80211: 2200 case IFT_BRIDGE: 2201 case IFT_ISO88025: 2202 ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr, 2203 desten); 2204 return (0); 2205 default: 2206 m_freem(m); 2207 return (EAFNOSUPPORT); 2208 } 2209 } 2210 2211 IF_AFDATA_RLOCK(ifp); 2212 ln = nd6_lookup(&dst6->sin6_addr, LLE_UNLOCKED, ifp); 2213 if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) { 2214 /* Entry found, let's copy lle info */ 2215 bcopy(ln->r_linkdata, desten, ln->r_hdrlen); 2216 if (pflags != NULL) 2217 *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR); 2218 /* Check if we have feedback request from nd6 timer */ 2219 if (ln->r_skip_req != 0) { 2220 LLE_REQ_LOCK(ln); 2221 ln->r_skip_req = 0; /* Notify that entry was used */ 2222 ln->lle_hittime = time_uptime; 2223 LLE_REQ_UNLOCK(ln); 2224 } 2225 IF_AFDATA_RUNLOCK(ifp); 2226 return (0); 2227 } 2228 IF_AFDATA_RUNLOCK(ifp); 2229 2230 return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags)); 2231 } 2232 2233 2234 /* 2235 * Do L2 address resolution for @sa_dst address. Stores found 2236 * address in @desten buffer. Copy of lle ln_flags can be also 2237 * saved in @pflags if @pflags is non-NULL. 2238 * 2239 * Heavy version. 2240 * Function assume that destination LLE does not exist, 2241 * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired. 2242 * 2243 * Set noinline to be dtrace-friendly 2244 */ 2245 static __noinline int 2246 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m, 2247 const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags) 2248 { 2249 struct llentry *lle = NULL, *lle_tmp; 2250 struct in6_addr *psrc, src; 2251 int send_ns, ll_len; 2252 char *lladdr; 2253 2254 /* 2255 * Address resolution or Neighbor Unreachability Detection 2256 * for the next hop. 2257 * At this point, the destination of the packet must be a unicast 2258 * or an anycast address(i.e. not a multicast). 2259 */ 2260 if (lle == NULL) { 2261 IF_AFDATA_RLOCK(ifp); 2262 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2263 IF_AFDATA_RUNLOCK(ifp); 2264 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2265 /* 2266 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2267 * the condition below is not very efficient. But we believe 2268 * it is tolerable, because this should be a rare case. 2269 */ 2270 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2271 if (lle == NULL) { 2272 char ip6buf[INET6_ADDRSTRLEN]; 2273 log(LOG_DEBUG, 2274 "nd6_output: can't allocate llinfo for %s " 2275 "(ln=%p)\n", 2276 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2277 m_freem(m); 2278 return (ENOBUFS); 2279 } 2280 2281 IF_AFDATA_WLOCK(ifp); 2282 LLE_WLOCK(lle); 2283 /* Prefer any existing entry over newly-created one */ 2284 lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp); 2285 if (lle_tmp == NULL) 2286 lltable_link_entry(LLTABLE6(ifp), lle); 2287 IF_AFDATA_WUNLOCK(ifp); 2288 if (lle_tmp != NULL) { 2289 lltable_free_entry(LLTABLE6(ifp), lle); 2290 lle = lle_tmp; 2291 lle_tmp = NULL; 2292 } 2293 } 2294 } 2295 if (lle == NULL) { 2296 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2297 m_freem(m); 2298 return (ENOBUFS); 2299 } 2300 2301 if (m != NULL) 2302 m_freem(m); 2303 return (ENOBUFS); 2304 } 2305 2306 LLE_WLOCK_ASSERT(lle); 2307 2308 /* 2309 * The first time we send a packet to a neighbor whose entry is 2310 * STALE, we have to change the state to DELAY and a sets a timer to 2311 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2312 * neighbor unreachability detection on expiration. 2313 * (RFC 2461 7.3.3) 2314 */ 2315 if (lle->ln_state == ND6_LLINFO_STALE) 2316 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY); 2317 2318 /* 2319 * If the neighbor cache entry has a state other than INCOMPLETE 2320 * (i.e. its link-layer address is already resolved), just 2321 * send the packet. 2322 */ 2323 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) { 2324 if (flags & LLE_ADDRONLY) { 2325 lladdr = lle->ll_addr; 2326 ll_len = ifp->if_addrlen; 2327 } else { 2328 lladdr = lle->r_linkdata; 2329 ll_len = lle->r_hdrlen; 2330 } 2331 bcopy(lladdr, desten, ll_len); 2332 if (pflags != NULL) 2333 *pflags = lle->la_flags; 2334 LLE_WUNLOCK(lle); 2335 return (0); 2336 } 2337 2338 /* 2339 * There is a neighbor cache entry, but no ethernet address 2340 * response yet. Append this latest packet to the end of the 2341 * packet queue in the mbuf. When it exceeds nd6_maxqueuelen, 2342 * the oldest packet in the queue will be removed. 2343 */ 2344 2345 if (lle->la_hold != NULL) { 2346 struct mbuf *m_hold; 2347 int i; 2348 2349 i = 0; 2350 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2351 i++; 2352 if (m_hold->m_nextpkt == NULL) { 2353 m_hold->m_nextpkt = m; 2354 break; 2355 } 2356 } 2357 while (i >= V_nd6_maxqueuelen) { 2358 m_hold = lle->la_hold; 2359 lle->la_hold = lle->la_hold->m_nextpkt; 2360 m_freem(m_hold); 2361 i--; 2362 } 2363 } else { 2364 lle->la_hold = m; 2365 } 2366 2367 /* 2368 * If there has been no NS for the neighbor after entering the 2369 * INCOMPLETE state, send the first solicitation. 2370 * Note that for newly-created lle la_asked will be 0, 2371 * so we will transition from ND6_LLINFO_NOSTATE to 2372 * ND6_LLINFO_INCOMPLETE state here. 2373 */ 2374 psrc = NULL; 2375 send_ns = 0; 2376 if (lle->la_asked == 0) { 2377 lle->la_asked++; 2378 send_ns = 1; 2379 psrc = nd6_llinfo_get_holdsrc(lle, &src); 2380 2381 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE); 2382 } 2383 LLE_WUNLOCK(lle); 2384 if (send_ns != 0) 2385 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL); 2386 2387 return (EWOULDBLOCK); 2388 } 2389 2390 /* 2391 * Do L2 address resolution for @sa_dst address. Stores found 2392 * address in @desten buffer. Copy of lle ln_flags can be also 2393 * saved in @pflags if @pflags is non-NULL. 2394 * 2395 * Return values: 2396 * - 0 on success (address copied to buffer). 2397 * - EWOULDBLOCK (no local error, but address is still unresolved) 2398 * - other errors (alloc failure, etc) 2399 */ 2400 int 2401 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst, 2402 char *desten, uint32_t *pflags) 2403 { 2404 int error; 2405 2406 flags |= LLE_ADDRONLY; 2407 error = nd6_resolve_slow(ifp, flags, NULL, 2408 (const struct sockaddr_in6 *)dst, desten, pflags); 2409 return (error); 2410 } 2411 2412 int 2413 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2414 struct sockaddr_in6 *dst) 2415 { 2416 struct mbuf *m, *m_head; 2417 struct ifnet *outifp; 2418 int error = 0; 2419 2420 m_head = chain; 2421 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2422 outifp = origifp; 2423 else 2424 outifp = ifp; 2425 2426 while (m_head) { 2427 m = m_head; 2428 m_head = m_head->m_nextpkt; 2429 error = nd6_output_ifp(ifp, origifp, m, dst, NULL); 2430 } 2431 2432 /* 2433 * XXX 2434 * note that intermediate errors are blindly ignored 2435 */ 2436 return (error); 2437 } 2438 2439 static int 2440 nd6_need_cache(struct ifnet *ifp) 2441 { 2442 /* 2443 * XXX: we currently do not make neighbor cache on any interface 2444 * other than ARCnet, Ethernet, FDDI and GIF. 2445 * 2446 * RFC2893 says: 2447 * - unidirectional tunnels needs no ND 2448 */ 2449 switch (ifp->if_type) { 2450 case IFT_ARCNET: 2451 case IFT_ETHER: 2452 case IFT_FDDI: 2453 case IFT_IEEE1394: 2454 case IFT_L2VLAN: 2455 case IFT_IEEE80211: 2456 case IFT_INFINIBAND: 2457 case IFT_BRIDGE: 2458 case IFT_PROPVIRTUAL: 2459 return (1); 2460 default: 2461 return (0); 2462 } 2463 } 2464 2465 /* 2466 * Add pernament ND6 link-layer record for given 2467 * interface address. 2468 * 2469 * Very similar to IPv4 arp_ifinit(), but: 2470 * 1) IPv6 DAD is performed in different place 2471 * 2) It is called by IPv6 protocol stack in contrast to 2472 * arp_ifinit() which is typically called in SIOCSIFADDR 2473 * driver ioctl handler. 2474 * 2475 */ 2476 int 2477 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2478 { 2479 struct ifnet *ifp; 2480 struct llentry *ln, *ln_tmp; 2481 struct sockaddr *dst; 2482 2483 ifp = ia->ia_ifa.ifa_ifp; 2484 if (nd6_need_cache(ifp) == 0) 2485 return (0); 2486 2487 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; 2488 dst = (struct sockaddr *)&ia->ia_addr; 2489 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2490 if (ln == NULL) 2491 return (ENOBUFS); 2492 2493 IF_AFDATA_WLOCK(ifp); 2494 LLE_WLOCK(ln); 2495 /* Unlink any entry if exists */ 2496 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2497 if (ln_tmp != NULL) 2498 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2499 lltable_link_entry(LLTABLE6(ifp), ln); 2500 IF_AFDATA_WUNLOCK(ifp); 2501 2502 if (ln_tmp != NULL) 2503 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2504 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2505 2506 LLE_WUNLOCK(ln); 2507 if (ln_tmp != NULL) 2508 llentry_free(ln_tmp); 2509 2510 return (0); 2511 } 2512 2513 /* 2514 * Removes either all lle entries for given @ia, or lle 2515 * corresponding to @ia address. 2516 */ 2517 void 2518 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all) 2519 { 2520 struct sockaddr_in6 mask, addr; 2521 struct sockaddr *saddr, *smask; 2522 struct ifnet *ifp; 2523 2524 ifp = ia->ia_ifa.ifa_ifp; 2525 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2526 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2527 saddr = (struct sockaddr *)&addr; 2528 smask = (struct sockaddr *)&mask; 2529 2530 if (all != 0) 2531 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC); 2532 else 2533 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr); 2534 } 2535 2536 static void 2537 clear_llinfo_pqueue(struct llentry *ln) 2538 { 2539 struct mbuf *m_hold, *m_hold_next; 2540 2541 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2542 m_hold_next = m_hold->m_nextpkt; 2543 m_freem(m_hold); 2544 } 2545 2546 ln->la_hold = NULL; 2547 } 2548 2549 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2550 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2551 2552 SYSCTL_DECL(_net_inet6_icmp6); 2553 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2554 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2555 NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter", 2556 "NDP default router list"); 2557 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2558 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, 2559 NULL, 0, nd6_sysctl_prlist, "S,in6_prefix", 2560 "NDP prefix list"); 2561 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2562 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2563 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2564 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2565 2566 static int 2567 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2568 { 2569 struct in6_defrouter d; 2570 struct nd_defrouter *dr; 2571 int error; 2572 2573 if (req->newptr != NULL) 2574 return (EPERM); 2575 2576 error = sysctl_wire_old_buffer(req, 0); 2577 if (error != 0) 2578 return (error); 2579 2580 bzero(&d, sizeof(d)); 2581 d.rtaddr.sin6_family = AF_INET6; 2582 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2583 2584 ND6_RLOCK(); 2585 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2586 d.rtaddr.sin6_addr = dr->rtaddr; 2587 error = sa6_recoverscope(&d.rtaddr); 2588 if (error != 0) 2589 break; 2590 d.flags = dr->raflags; 2591 d.rtlifetime = dr->rtlifetime; 2592 d.expire = dr->expire + (time_second - time_uptime); 2593 d.if_index = dr->ifp->if_index; 2594 error = SYSCTL_OUT(req, &d, sizeof(d)); 2595 if (error != 0) 2596 break; 2597 } 2598 ND6_RUNLOCK(); 2599 return (error); 2600 } 2601 2602 static int 2603 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2604 { 2605 struct in6_prefix p; 2606 struct sockaddr_in6 s6; 2607 struct nd_prefix *pr; 2608 struct nd_pfxrouter *pfr; 2609 time_t maxexpire; 2610 int error; 2611 char ip6buf[INET6_ADDRSTRLEN]; 2612 2613 if (req->newptr) 2614 return (EPERM); 2615 2616 error = sysctl_wire_old_buffer(req, 0); 2617 if (error != 0) 2618 return (error); 2619 2620 bzero(&p, sizeof(p)); 2621 p.origin = PR_ORIG_RA; 2622 bzero(&s6, sizeof(s6)); 2623 s6.sin6_family = AF_INET6; 2624 s6.sin6_len = sizeof(s6); 2625 2626 ND6_RLOCK(); 2627 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2628 p.prefix = pr->ndpr_prefix; 2629 if (sa6_recoverscope(&p.prefix)) { 2630 log(LOG_ERR, "scope error in prefix list (%s)\n", 2631 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2632 /* XXX: press on... */ 2633 } 2634 p.raflags = pr->ndpr_raf; 2635 p.prefixlen = pr->ndpr_plen; 2636 p.vltime = pr->ndpr_vltime; 2637 p.pltime = pr->ndpr_pltime; 2638 p.if_index = pr->ndpr_ifp->if_index; 2639 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2640 p.expire = 0; 2641 else { 2642 /* XXX: we assume time_t is signed. */ 2643 maxexpire = (-1) & 2644 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2645 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2646 p.expire = pr->ndpr_lastupdate + 2647 pr->ndpr_vltime + 2648 (time_second - time_uptime); 2649 else 2650 p.expire = maxexpire; 2651 } 2652 p.refcnt = pr->ndpr_refcnt; 2653 p.flags = pr->ndpr_stateflags; 2654 p.advrtrs = 0; 2655 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2656 p.advrtrs++; 2657 error = SYSCTL_OUT(req, &p, sizeof(p)); 2658 if (error != 0) 2659 break; 2660 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2661 s6.sin6_addr = pfr->router->rtaddr; 2662 if (sa6_recoverscope(&s6)) 2663 log(LOG_ERR, 2664 "scope error in prefix list (%s)\n", 2665 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2666 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2667 if (error != 0) 2668 break; 2669 } 2670 } 2671 ND6_RUNLOCK(); 2672 return (error); 2673 } 2674