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 /* timer values */ 89 VNET_DEFINE(int, nd6_prune) = 1; /* walk list every 1 seconds */ 90 VNET_DEFINE(int, nd6_delay) = 5; /* delay first probe time 5 second */ 91 VNET_DEFINE(int, nd6_umaxtries) = 3; /* maximum unicast query */ 92 VNET_DEFINE(int, nd6_mmaxtries) = 3; /* maximum multicast query */ 93 VNET_DEFINE(int, nd6_useloopback) = 1; /* use loopback interface for 94 * local traffic */ 95 VNET_DEFINE(int, nd6_gctimer) = (60 * 60 * 24); /* 1 day: garbage 96 * collection timer */ 97 98 /* preventing too many loops in ND option parsing */ 99 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */ 100 101 VNET_DEFINE(int, nd6_maxnudhint) = 0; /* max # of subsequent upper 102 * layer hints */ 103 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved 104 * ND entries */ 105 #define V_nd6_maxndopt VNET(nd6_maxndopt) 106 #define V_nd6_maxqueuelen VNET(nd6_maxqueuelen) 107 108 #ifdef ND6_DEBUG 109 VNET_DEFINE(int, nd6_debug) = 1; 110 #else 111 VNET_DEFINE(int, nd6_debug) = 0; 112 #endif 113 114 static eventhandler_tag lle_event_eh; 115 116 /* for debugging? */ 117 #if 0 118 static int nd6_inuse, nd6_allocated; 119 #endif 120 121 VNET_DEFINE(struct nd_drhead, nd_defrouter); 122 VNET_DEFINE(struct nd_prhead, nd_prefix); 123 124 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL; 125 #define V_nd6_recalc_reachtm_interval VNET(nd6_recalc_reachtm_interval) 126 127 int (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int); 128 129 static int nd6_is_new_addr_neighbor(struct sockaddr_in6 *, 130 struct ifnet *); 131 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); 132 static void nd6_slowtimo(void *); 133 static int regen_tmpaddr(struct in6_ifaddr *); 134 static struct llentry *nd6_free(struct llentry *, int); 135 static void nd6_llinfo_timer(void *); 136 static void clear_llinfo_pqueue(struct llentry *); 137 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *); 138 static int nd6_output_lle(struct ifnet *, struct ifnet *, struct mbuf *, 139 struct sockaddr_in6 *); 140 static int nd6_output_ifp(struct ifnet *, struct ifnet *, struct mbuf *, 141 struct sockaddr_in6 *); 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); 147 148 static void 149 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt) 150 { 151 struct rt_addrinfo rtinfo; 152 struct sockaddr_in6 dst; 153 struct sockaddr_dl gw; 154 struct ifnet *ifp; 155 int type; 156 157 LLE_WLOCK_ASSERT(lle); 158 159 if (lltable_get_af(lle->lle_tbl) != AF_INET6) 160 return; 161 162 switch (evt) { 163 case LLENTRY_RESOLVED: 164 type = RTM_ADD; 165 KASSERT(lle->la_flags & LLE_VALID, 166 ("%s: %p resolved but not valid?", __func__, lle)); 167 break; 168 case LLENTRY_EXPIRED: 169 type = RTM_DELETE; 170 break; 171 default: 172 return; 173 } 174 175 ifp = lltable_get_ifp(lle->lle_tbl); 176 177 bzero(&dst, sizeof(dst)); 178 bzero(&gw, sizeof(gw)); 179 bzero(&rtinfo, sizeof(rtinfo)); 180 lltable_fill_sa_entry(lle, (struct sockaddr *)&dst); 181 dst.sin6_scope_id = in6_getscopezone(ifp, 182 in6_addrscope(&dst.sin6_addr)); 183 gw.sdl_len = sizeof(struct sockaddr_dl); 184 gw.sdl_family = AF_LINK; 185 gw.sdl_alen = ifp->if_addrlen; 186 gw.sdl_index = ifp->if_index; 187 gw.sdl_type = ifp->if_type; 188 if (evt == LLENTRY_RESOLVED) 189 bcopy(&lle->ll_addr, gw.sdl_data, ifp->if_addrlen); 190 rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst; 191 rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw; 192 rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY; 193 rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | ( 194 type == RTM_ADD ? RTF_UP: 0), 0, RT_DEFAULT_FIB); 195 } 196 197 void 198 nd6_init(void) 199 { 200 201 LIST_INIT(&V_nd_prefix); 202 203 /* initialization of the default router list */ 204 TAILQ_INIT(&V_nd_defrouter); 205 206 /* start timer */ 207 callout_init(&V_nd6_slowtimo_ch, 0); 208 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 209 nd6_slowtimo, curvnet); 210 211 nd6_dad_init(); 212 if (IS_DEFAULT_VNET(curvnet)) 213 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event, 214 NULL, EVENTHANDLER_PRI_ANY); 215 } 216 217 #ifdef VIMAGE 218 void 219 nd6_destroy() 220 { 221 222 callout_drain(&V_nd6_slowtimo_ch); 223 callout_drain(&V_nd6_timer_ch); 224 if (IS_DEFAULT_VNET(curvnet)) 225 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh); 226 } 227 #endif 228 229 struct nd_ifinfo * 230 nd6_ifattach(struct ifnet *ifp) 231 { 232 struct nd_ifinfo *nd; 233 234 nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK|M_ZERO); 235 nd->initialized = 1; 236 237 nd->chlim = IPV6_DEFHLIM; 238 nd->basereachable = REACHABLE_TIME; 239 nd->reachable = ND_COMPUTE_RTIME(nd->basereachable); 240 nd->retrans = RETRANS_TIMER; 241 242 nd->flags = ND6_IFF_PERFORMNUD; 243 244 /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL. 245 * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by 246 * default regardless of the V_ip6_auto_linklocal configuration to 247 * give a reasonable default behavior. 248 */ 249 if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) || 250 (ifp->if_flags & IFF_LOOPBACK)) 251 nd->flags |= ND6_IFF_AUTO_LINKLOCAL; 252 /* 253 * A loopback interface does not need to accept RTADV. 254 * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by 255 * default regardless of the V_ip6_accept_rtadv configuration to 256 * prevent the interface from accepting RA messages arrived 257 * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV. 258 */ 259 if (V_ip6_accept_rtadv && 260 !(ifp->if_flags & IFF_LOOPBACK) && 261 (ifp->if_type != IFT_BRIDGE)) 262 nd->flags |= ND6_IFF_ACCEPT_RTADV; 263 if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK)) 264 nd->flags |= ND6_IFF_NO_RADR; 265 266 /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */ 267 nd6_setmtu0(ifp, nd); 268 269 return nd; 270 } 271 272 void 273 nd6_ifdetach(struct nd_ifinfo *nd) 274 { 275 276 free(nd, M_IP6NDP); 277 } 278 279 /* 280 * Reset ND level link MTU. This function is called when the physical MTU 281 * changes, which means we might have to adjust the ND level MTU. 282 */ 283 void 284 nd6_setmtu(struct ifnet *ifp) 285 { 286 287 nd6_setmtu0(ifp, ND_IFINFO(ifp)); 288 } 289 290 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */ 291 void 292 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi) 293 { 294 u_int32_t omaxmtu; 295 296 omaxmtu = ndi->maxmtu; 297 298 switch (ifp->if_type) { 299 case IFT_ARCNET: 300 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */ 301 break; 302 case IFT_FDDI: 303 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */ 304 break; 305 case IFT_ISO88025: 306 ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu); 307 break; 308 default: 309 ndi->maxmtu = ifp->if_mtu; 310 break; 311 } 312 313 /* 314 * Decreasing the interface MTU under IPV6 minimum MTU may cause 315 * undesirable situation. We thus notify the operator of the change 316 * explicitly. The check for omaxmtu is necessary to restrict the 317 * log to the case of changing the MTU, not initializing it. 318 */ 319 if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) { 320 log(LOG_NOTICE, "nd6_setmtu0: " 321 "new link MTU on %s (%lu) is too small for IPv6\n", 322 if_name(ifp), (unsigned long)ndi->maxmtu); 323 } 324 325 if (ndi->maxmtu > V_in6_maxmtu) 326 in6_setmaxmtu(); /* check all interfaces just in case */ 327 328 } 329 330 void 331 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts) 332 { 333 334 bzero(ndopts, sizeof(*ndopts)); 335 ndopts->nd_opts_search = (struct nd_opt_hdr *)opt; 336 ndopts->nd_opts_last 337 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len); 338 339 if (icmp6len == 0) { 340 ndopts->nd_opts_done = 1; 341 ndopts->nd_opts_search = NULL; 342 } 343 } 344 345 /* 346 * Take one ND option. 347 */ 348 struct nd_opt_hdr * 349 nd6_option(union nd_opts *ndopts) 350 { 351 struct nd_opt_hdr *nd_opt; 352 int olen; 353 354 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 355 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 356 __func__)); 357 if (ndopts->nd_opts_search == NULL) 358 return NULL; 359 if (ndopts->nd_opts_done) 360 return NULL; 361 362 nd_opt = ndopts->nd_opts_search; 363 364 /* make sure nd_opt_len is inside the buffer */ 365 if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) { 366 bzero(ndopts, sizeof(*ndopts)); 367 return NULL; 368 } 369 370 olen = nd_opt->nd_opt_len << 3; 371 if (olen == 0) { 372 /* 373 * Message validation requires that all included 374 * options have a length that is greater than zero. 375 */ 376 bzero(ndopts, sizeof(*ndopts)); 377 return NULL; 378 } 379 380 ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen); 381 if (ndopts->nd_opts_search > ndopts->nd_opts_last) { 382 /* option overruns the end of buffer, invalid */ 383 bzero(ndopts, sizeof(*ndopts)); 384 return NULL; 385 } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) { 386 /* reached the end of options chain */ 387 ndopts->nd_opts_done = 1; 388 ndopts->nd_opts_search = NULL; 389 } 390 return nd_opt; 391 } 392 393 /* 394 * Parse multiple ND options. 395 * This function is much easier to use, for ND routines that do not need 396 * multiple options of the same type. 397 */ 398 int 399 nd6_options(union nd_opts *ndopts) 400 { 401 struct nd_opt_hdr *nd_opt; 402 int i = 0; 403 404 KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__)); 405 KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts", 406 __func__)); 407 if (ndopts->nd_opts_search == NULL) 408 return 0; 409 410 while (1) { 411 nd_opt = nd6_option(ndopts); 412 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) { 413 /* 414 * Message validation requires that all included 415 * options have a length that is greater than zero. 416 */ 417 ICMP6STAT_INC(icp6s_nd_badopt); 418 bzero(ndopts, sizeof(*ndopts)); 419 return -1; 420 } 421 422 if (nd_opt == NULL) 423 goto skip1; 424 425 switch (nd_opt->nd_opt_type) { 426 case ND_OPT_SOURCE_LINKADDR: 427 case ND_OPT_TARGET_LINKADDR: 428 case ND_OPT_MTU: 429 case ND_OPT_REDIRECTED_HEADER: 430 case ND_OPT_NONCE: 431 if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) { 432 nd6log((LOG_INFO, 433 "duplicated ND6 option found (type=%d)\n", 434 nd_opt->nd_opt_type)); 435 /* XXX bark? */ 436 } else { 437 ndopts->nd_opt_array[nd_opt->nd_opt_type] 438 = nd_opt; 439 } 440 break; 441 case ND_OPT_PREFIX_INFORMATION: 442 if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) { 443 ndopts->nd_opt_array[nd_opt->nd_opt_type] 444 = nd_opt; 445 } 446 ndopts->nd_opts_pi_end = 447 (struct nd_opt_prefix_info *)nd_opt; 448 break; 449 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */ 450 case ND_OPT_RDNSS: /* RFC 6106 */ 451 case ND_OPT_DNSSL: /* RFC 6106 */ 452 /* 453 * Silently ignore options we know and do not care about 454 * in the kernel. 455 */ 456 break; 457 default: 458 /* 459 * Unknown options must be silently ignored, 460 * to accomodate future extension to the protocol. 461 */ 462 nd6log((LOG_DEBUG, 463 "nd6_options: unsupported option %d - " 464 "option ignored\n", nd_opt->nd_opt_type)); 465 } 466 467 skip1: 468 i++; 469 if (i > V_nd6_maxndopt) { 470 ICMP6STAT_INC(icp6s_nd_toomanyopt); 471 nd6log((LOG_INFO, "too many loop in nd opt\n")); 472 break; 473 } 474 475 if (ndopts->nd_opts_done) 476 break; 477 } 478 479 return 0; 480 } 481 482 /* 483 * ND6 timer routine to handle ND6 entries 484 */ 485 void 486 nd6_llinfo_settimer_locked(struct llentry *ln, long tick) 487 { 488 int canceled; 489 490 LLE_WLOCK_ASSERT(ln); 491 492 if (tick < 0) { 493 ln->la_expire = 0; 494 ln->ln_ntick = 0; 495 canceled = callout_stop(&ln->lle_timer); 496 } else { 497 ln->la_expire = time_uptime + tick / hz; 498 LLE_ADDREF(ln); 499 if (tick > INT_MAX) { 500 ln->ln_ntick = tick - INT_MAX; 501 canceled = callout_reset(&ln->lle_timer, INT_MAX, 502 nd6_llinfo_timer, ln); 503 } else { 504 ln->ln_ntick = 0; 505 canceled = callout_reset(&ln->lle_timer, tick, 506 nd6_llinfo_timer, ln); 507 } 508 } 509 if (canceled) 510 LLE_REMREF(ln); 511 } 512 513 void 514 nd6_llinfo_settimer(struct llentry *ln, long tick) 515 { 516 517 LLE_WLOCK(ln); 518 nd6_llinfo_settimer_locked(ln, tick); 519 LLE_WUNLOCK(ln); 520 } 521 522 static void 523 nd6_llinfo_timer(void *arg) 524 { 525 struct llentry *ln; 526 struct in6_addr *dst; 527 struct ifnet *ifp; 528 struct nd_ifinfo *ndi = NULL; 529 530 KASSERT(arg != NULL, ("%s: arg NULL", __func__)); 531 ln = (struct llentry *)arg; 532 LLE_WLOCK(ln); 533 if (callout_pending(&ln->lle_timer)) { 534 /* 535 * Here we are a bit odd here in the treatment of 536 * active/pending. If the pending bit is set, it got 537 * rescheduled before I ran. The active 538 * bit we ignore, since if it was stopped 539 * in ll_tablefree() and was currently running 540 * it would have return 0 so the code would 541 * not have deleted it since the callout could 542 * not be stopped so we want to go through 543 * with the delete here now. If the callout 544 * was restarted, the pending bit will be back on and 545 * we just want to bail since the callout_reset would 546 * return 1 and our reference would have been removed 547 * by nd6_llinfo_settimer_locked above since canceled 548 * would have been 1. 549 */ 550 LLE_WUNLOCK(ln); 551 return; 552 } 553 ifp = ln->lle_tbl->llt_ifp; 554 CURVNET_SET(ifp->if_vnet); 555 556 if (ln->ln_ntick > 0) { 557 if (ln->ln_ntick > INT_MAX) { 558 ln->ln_ntick -= INT_MAX; 559 nd6_llinfo_settimer_locked(ln, INT_MAX); 560 } else { 561 ln->ln_ntick = 0; 562 nd6_llinfo_settimer_locked(ln, ln->ln_ntick); 563 } 564 goto done; 565 } 566 567 ndi = ND_IFINFO(ifp); 568 dst = &ln->r_l3addr.addr6; 569 if (ln->la_flags & LLE_STATIC) { 570 goto done; 571 } 572 573 if (ln->la_flags & LLE_DELETED) { 574 (void)nd6_free(ln, 0); 575 ln = NULL; 576 goto done; 577 } 578 579 switch (ln->ln_state) { 580 case ND6_LLINFO_INCOMPLETE: 581 if (ln->la_asked < V_nd6_mmaxtries) { 582 ln->la_asked++; 583 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 584 LLE_WUNLOCK(ln); 585 nd6_ns_output(ifp, NULL, dst, ln, NULL); 586 LLE_WLOCK(ln); 587 } else { 588 struct mbuf *m = ln->la_hold; 589 if (m) { 590 struct mbuf *m0; 591 592 /* 593 * assuming every packet in la_hold has the 594 * same IP header. Send error after unlock. 595 */ 596 m0 = m->m_nextpkt; 597 m->m_nextpkt = NULL; 598 ln->la_hold = m0; 599 clear_llinfo_pqueue(ln); 600 } 601 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_TIMEDOUT); 602 (void)nd6_free(ln, 0); 603 ln = NULL; 604 if (m != NULL) 605 icmp6_error2(m, ICMP6_DST_UNREACH, 606 ICMP6_DST_UNREACH_ADDR, 0, ifp); 607 } 608 break; 609 case ND6_LLINFO_REACHABLE: 610 if (!ND6_LLINFO_PERMANENT(ln)) { 611 ln->ln_state = ND6_LLINFO_STALE; 612 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 613 } 614 break; 615 616 case ND6_LLINFO_STALE: 617 /* Garbage Collection(RFC 2461 5.3) */ 618 if (!ND6_LLINFO_PERMANENT(ln)) { 619 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 620 (void)nd6_free(ln, 1); 621 ln = NULL; 622 } 623 break; 624 625 case ND6_LLINFO_DELAY: 626 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { 627 /* We need NUD */ 628 ln->la_asked = 1; 629 ln->ln_state = ND6_LLINFO_PROBE; 630 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 631 LLE_WUNLOCK(ln); 632 nd6_ns_output(ifp, dst, dst, ln, NULL); 633 LLE_WLOCK(ln); 634 } else { 635 ln->ln_state = ND6_LLINFO_STALE; /* XXX */ 636 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); 637 } 638 break; 639 case ND6_LLINFO_PROBE: 640 if (ln->la_asked < V_nd6_umaxtries) { 641 ln->la_asked++; 642 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000); 643 LLE_WUNLOCK(ln); 644 nd6_ns_output(ifp, dst, dst, ln, NULL); 645 LLE_WLOCK(ln); 646 } else { 647 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED); 648 (void)nd6_free(ln, 0); 649 ln = NULL; 650 } 651 break; 652 default: 653 panic("%s: paths in a dark night can be confusing: %d", 654 __func__, ln->ln_state); 655 } 656 done: 657 if (ln != NULL) 658 LLE_FREE_LOCKED(ln); 659 CURVNET_RESTORE(); 660 } 661 662 663 /* 664 * ND6 timer routine to expire default route list and prefix list 665 */ 666 void 667 nd6_timer(void *arg) 668 { 669 CURVNET_SET((struct vnet *) arg); 670 struct nd_defrouter *dr, *ndr; 671 struct nd_prefix *pr, *npr; 672 struct in6_ifaddr *ia6, *nia6; 673 674 callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz, 675 nd6_timer, curvnet); 676 677 /* expire default router list */ 678 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 679 if (dr->expire && dr->expire < time_uptime) 680 defrtrlist_del(dr); 681 } 682 683 /* 684 * expire interface addresses. 685 * in the past the loop was inside prefix expiry processing. 686 * However, from a stricter speci-confrmance standpoint, we should 687 * rather separate address lifetimes and prefix lifetimes. 688 * 689 * XXXRW: in6_ifaddrhead locking. 690 */ 691 addrloop: 692 TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) { 693 /* check address lifetime */ 694 if (IFA6_IS_INVALID(ia6)) { 695 int regen = 0; 696 697 /* 698 * If the expiring address is temporary, try 699 * regenerating a new one. This would be useful when 700 * we suspended a laptop PC, then turned it on after a 701 * period that could invalidate all temporary 702 * addresses. Although we may have to restart the 703 * loop (see below), it must be after purging the 704 * address. Otherwise, we'd see an infinite loop of 705 * regeneration. 706 */ 707 if (V_ip6_use_tempaddr && 708 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) { 709 if (regen_tmpaddr(ia6) == 0) 710 regen = 1; 711 } 712 713 in6_purgeaddr(&ia6->ia_ifa); 714 715 if (regen) 716 goto addrloop; /* XXX: see below */ 717 } else if (IFA6_IS_DEPRECATED(ia6)) { 718 int oldflags = ia6->ia6_flags; 719 720 ia6->ia6_flags |= IN6_IFF_DEPRECATED; 721 722 /* 723 * If a temporary address has just become deprecated, 724 * regenerate a new one if possible. 725 */ 726 if (V_ip6_use_tempaddr && 727 (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 728 (oldflags & IN6_IFF_DEPRECATED) == 0) { 729 730 if (regen_tmpaddr(ia6) == 0) { 731 /* 732 * A new temporary address is 733 * generated. 734 * XXX: this means the address chain 735 * has changed while we are still in 736 * the loop. Although the change 737 * would not cause disaster (because 738 * it's not a deletion, but an 739 * addition,) we'd rather restart the 740 * loop just for safety. Or does this 741 * significantly reduce performance?? 742 */ 743 goto addrloop; 744 } 745 } 746 } else { 747 /* 748 * A new RA might have made a deprecated address 749 * preferred. 750 */ 751 ia6->ia6_flags &= ~IN6_IFF_DEPRECATED; 752 } 753 } 754 755 /* expire prefix list */ 756 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 757 /* 758 * check prefix lifetime. 759 * since pltime is just for autoconf, pltime processing for 760 * prefix is not necessary. 761 */ 762 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME && 763 time_uptime - pr->ndpr_lastupdate > pr->ndpr_vltime) { 764 765 /* 766 * address expiration and prefix expiration are 767 * separate. NEVER perform in6_purgeaddr here. 768 */ 769 prelist_remove(pr); 770 } 771 } 772 CURVNET_RESTORE(); 773 } 774 775 /* 776 * ia6 - deprecated/invalidated temporary address 777 */ 778 static int 779 regen_tmpaddr(struct in6_ifaddr *ia6) 780 { 781 struct ifaddr *ifa; 782 struct ifnet *ifp; 783 struct in6_ifaddr *public_ifa6 = NULL; 784 785 ifp = ia6->ia_ifa.ifa_ifp; 786 IF_ADDR_RLOCK(ifp); 787 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 788 struct in6_ifaddr *it6; 789 790 if (ifa->ifa_addr->sa_family != AF_INET6) 791 continue; 792 793 it6 = (struct in6_ifaddr *)ifa; 794 795 /* ignore no autoconf addresses. */ 796 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0) 797 continue; 798 799 /* ignore autoconf addresses with different prefixes. */ 800 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr) 801 continue; 802 803 /* 804 * Now we are looking at an autoconf address with the same 805 * prefix as ours. If the address is temporary and is still 806 * preferred, do not create another one. It would be rare, but 807 * could happen, for example, when we resume a laptop PC after 808 * a long period. 809 */ 810 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 && 811 !IFA6_IS_DEPRECATED(it6)) { 812 public_ifa6 = NULL; 813 break; 814 } 815 816 /* 817 * This is a public autoconf address that has the same prefix 818 * as ours. If it is preferred, keep it. We can't break the 819 * loop here, because there may be a still-preferred temporary 820 * address with the prefix. 821 */ 822 if (!IFA6_IS_DEPRECATED(it6)) 823 public_ifa6 = it6; 824 } 825 if (public_ifa6 != NULL) 826 ifa_ref(&public_ifa6->ia_ifa); 827 IF_ADDR_RUNLOCK(ifp); 828 829 if (public_ifa6 != NULL) { 830 int e; 831 832 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) { 833 ifa_free(&public_ifa6->ia_ifa); 834 log(LOG_NOTICE, "regen_tmpaddr: failed to create a new" 835 " tmp addr,errno=%d\n", e); 836 return (-1); 837 } 838 ifa_free(&public_ifa6->ia_ifa); 839 return (0); 840 } 841 842 return (-1); 843 } 844 845 /* 846 * Nuke neighbor cache/prefix/default router management table, right before 847 * ifp goes away. 848 */ 849 void 850 nd6_purge(struct ifnet *ifp) 851 { 852 struct nd_defrouter *dr, *ndr; 853 struct nd_prefix *pr, *npr; 854 855 /* 856 * Nuke default router list entries toward ifp. 857 * We defer removal of default router list entries that is installed 858 * in the routing table, in order to keep additional side effects as 859 * small as possible. 860 */ 861 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 862 if (dr->installed) 863 continue; 864 865 if (dr->ifp == ifp) 866 defrtrlist_del(dr); 867 } 868 869 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) { 870 if (!dr->installed) 871 continue; 872 873 if (dr->ifp == ifp) 874 defrtrlist_del(dr); 875 } 876 877 /* Nuke prefix list entries toward ifp */ 878 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) { 879 if (pr->ndpr_ifp == ifp) { 880 /* 881 * Because if_detach() does *not* release prefixes 882 * while purging addresses the reference count will 883 * still be above zero. We therefore reset it to 884 * make sure that the prefix really gets purged. 885 */ 886 pr->ndpr_refcnt = 0; 887 888 /* 889 * Previously, pr->ndpr_addr is removed as well, 890 * but I strongly believe we don't have to do it. 891 * nd6_purge() is only called from in6_ifdetach(), 892 * which removes all the associated interface addresses 893 * by itself. 894 * (jinmei@kame.net 20010129) 895 */ 896 prelist_remove(pr); 897 } 898 } 899 900 /* cancel default outgoing interface setting */ 901 if (V_nd6_defifindex == ifp->if_index) 902 nd6_setdefaultiface(0); 903 904 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 905 /* Refresh default router list. */ 906 defrouter_select(); 907 } 908 909 /* XXXXX 910 * We do not nuke the neighbor cache entries here any more 911 * because the neighbor cache is kept in if_afdata[AF_INET6]. 912 * nd6_purge() is invoked by in6_ifdetach() which is called 913 * from if_detach() where everything gets purged. So let 914 * in6_domifdetach() do the actual L2 table purging work. 915 */ 916 } 917 918 /* 919 * the caller acquires and releases the lock on the lltbls 920 * Returns the llentry locked 921 */ 922 struct llentry * 923 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) 924 { 925 struct sockaddr_in6 sin6; 926 struct llentry *ln; 927 int llflags; 928 929 bzero(&sin6, sizeof(sin6)); 930 sin6.sin6_len = sizeof(struct sockaddr_in6); 931 sin6.sin6_family = AF_INET6; 932 sin6.sin6_addr = *addr6; 933 934 IF_AFDATA_LOCK_ASSERT(ifp); 935 936 llflags = (flags & ND6_EXCLUSIVE) ? LLE_EXCLUSIVE : 0; 937 ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); 938 939 return (ln); 940 } 941 942 struct llentry * 943 nd6_alloc(struct in6_addr *addr6, int flags, struct ifnet *ifp) 944 { 945 struct sockaddr_in6 sin6; 946 struct llentry *ln; 947 948 bzero(&sin6, sizeof(sin6)); 949 sin6.sin6_len = sizeof(struct sockaddr_in6); 950 sin6.sin6_family = AF_INET6; 951 sin6.sin6_addr = *addr6; 952 953 ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6); 954 if (ln != NULL) 955 ln->ln_state = ND6_LLINFO_NOSTATE; 956 957 return (ln); 958 } 959 960 /* 961 * Test whether a given IPv6 address is a neighbor or not, ignoring 962 * the actual neighbor cache. The neighbor cache is ignored in order 963 * to not reenter the routing code from within itself. 964 */ 965 static int 966 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 967 { 968 struct nd_prefix *pr; 969 struct ifaddr *dstaddr; 970 971 /* 972 * A link-local address is always a neighbor. 973 * XXX: a link does not necessarily specify a single interface. 974 */ 975 if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) { 976 struct sockaddr_in6 sin6_copy; 977 u_int32_t zone; 978 979 /* 980 * We need sin6_copy since sa6_recoverscope() may modify the 981 * content (XXX). 982 */ 983 sin6_copy = *addr; 984 if (sa6_recoverscope(&sin6_copy)) 985 return (0); /* XXX: should be impossible */ 986 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone)) 987 return (0); 988 if (sin6_copy.sin6_scope_id == zone) 989 return (1); 990 else 991 return (0); 992 } 993 994 /* 995 * If the address matches one of our addresses, 996 * it should be a neighbor. 997 * If the address matches one of our on-link prefixes, it should be a 998 * neighbor. 999 */ 1000 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1001 if (pr->ndpr_ifp != ifp) 1002 continue; 1003 1004 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) { 1005 struct rtentry *rt; 1006 1007 /* Always use the default FIB here. */ 1008 rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix, 1009 0, 0, RT_DEFAULT_FIB); 1010 if (rt == NULL) 1011 continue; 1012 /* 1013 * This is the case where multiple interfaces 1014 * have the same prefix, but only one is installed 1015 * into the routing table and that prefix entry 1016 * is not the one being examined here. In the case 1017 * where RADIX_MPATH is enabled, multiple route 1018 * entries (of the same rt_key value) will be 1019 * installed because the interface addresses all 1020 * differ. 1021 */ 1022 if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1023 &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) { 1024 RTFREE_LOCKED(rt); 1025 continue; 1026 } 1027 RTFREE_LOCKED(rt); 1028 } 1029 1030 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr, 1031 &addr->sin6_addr, &pr->ndpr_mask)) 1032 return (1); 1033 } 1034 1035 /* 1036 * If the address is assigned on the node of the other side of 1037 * a p2p interface, the address should be a neighbor. 1038 */ 1039 dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr, RT_ALL_FIBS); 1040 if (dstaddr != NULL) { 1041 if (dstaddr->ifa_ifp == ifp) { 1042 ifa_free(dstaddr); 1043 return (1); 1044 } 1045 ifa_free(dstaddr); 1046 } 1047 1048 /* 1049 * If the default router list is empty, all addresses are regarded 1050 * as on-link, and thus, as a neighbor. 1051 */ 1052 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV && 1053 TAILQ_EMPTY(&V_nd_defrouter) && 1054 V_nd6_defifindex == ifp->if_index) { 1055 return (1); 1056 } 1057 1058 return (0); 1059 } 1060 1061 1062 /* 1063 * Detect if a given IPv6 address identifies a neighbor on a given link. 1064 * XXX: should take care of the destination of a p2p link? 1065 */ 1066 int 1067 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) 1068 { 1069 struct llentry *lle; 1070 int rc = 0; 1071 1072 IF_AFDATA_UNLOCK_ASSERT(ifp); 1073 if (nd6_is_new_addr_neighbor(addr, ifp)) 1074 return (1); 1075 1076 /* 1077 * Even if the address matches none of our addresses, it might be 1078 * in the neighbor cache. 1079 */ 1080 IF_AFDATA_RLOCK(ifp); 1081 if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { 1082 LLE_RUNLOCK(lle); 1083 rc = 1; 1084 } 1085 IF_AFDATA_RUNLOCK(ifp); 1086 return (rc); 1087 } 1088 1089 /* 1090 * Free an nd6 llinfo entry. 1091 * Since the function would cause significant changes in the kernel, DO NOT 1092 * make it global, unless you have a strong reason for the change, and are sure 1093 * that the change is safe. 1094 */ 1095 static struct llentry * 1096 nd6_free(struct llentry *ln, int gc) 1097 { 1098 struct llentry *next; 1099 struct nd_defrouter *dr; 1100 struct ifnet *ifp; 1101 1102 LLE_WLOCK_ASSERT(ln); 1103 1104 /* 1105 * we used to have pfctlinput(PRC_HOSTDEAD) here. 1106 * even though it is not harmful, it was not really necessary. 1107 */ 1108 1109 /* cancel timer */ 1110 nd6_llinfo_settimer_locked(ln, -1); 1111 1112 ifp = ln->lle_tbl->llt_ifp; 1113 1114 if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1115 dr = defrouter_lookup(&ln->r_l3addr.addr6, ifp); 1116 1117 if (dr != NULL && dr->expire && 1118 ln->ln_state == ND6_LLINFO_STALE && gc) { 1119 /* 1120 * If the reason for the deletion is just garbage 1121 * collection, and the neighbor is an active default 1122 * router, do not delete it. Instead, reset the GC 1123 * timer using the router's lifetime. 1124 * Simply deleting the entry would affect default 1125 * router selection, which is not necessarily a good 1126 * thing, especially when we're using router preference 1127 * values. 1128 * XXX: the check for ln_state would be redundant, 1129 * but we intentionally keep it just in case. 1130 */ 1131 if (dr->expire > time_uptime) 1132 nd6_llinfo_settimer_locked(ln, 1133 (dr->expire - time_uptime) * hz); 1134 else 1135 nd6_llinfo_settimer_locked(ln, 1136 (long)V_nd6_gctimer * hz); 1137 1138 next = LIST_NEXT(ln, lle_next); 1139 LLE_REMREF(ln); 1140 LLE_WUNLOCK(ln); 1141 return (next); 1142 } 1143 1144 if (dr) { 1145 /* 1146 * Unreachablity of a router might affect the default 1147 * router selection and on-link detection of advertised 1148 * prefixes. 1149 */ 1150 1151 /* 1152 * Temporarily fake the state to choose a new default 1153 * router and to perform on-link determination of 1154 * prefixes correctly. 1155 * Below the state will be set correctly, 1156 * or the entry itself will be deleted. 1157 */ 1158 ln->ln_state = ND6_LLINFO_INCOMPLETE; 1159 } 1160 1161 if (ln->ln_router || dr) { 1162 1163 /* 1164 * We need to unlock to avoid a LOR with rt6_flush() with the 1165 * rnh and for the calls to pfxlist_onlink_check() and 1166 * defrouter_select() in the block further down for calls 1167 * into nd6_lookup(). We still hold a ref. 1168 */ 1169 LLE_WUNLOCK(ln); 1170 1171 /* 1172 * rt6_flush must be called whether or not the neighbor 1173 * is in the Default Router List. 1174 * See a corresponding comment in nd6_na_input(). 1175 */ 1176 rt6_flush(&ln->r_l3addr.addr6, ifp); 1177 } 1178 1179 if (dr) { 1180 /* 1181 * Since defrouter_select() does not affect the 1182 * on-link determination and MIP6 needs the check 1183 * before the default router selection, we perform 1184 * the check now. 1185 */ 1186 pfxlist_onlink_check(); 1187 1188 /* 1189 * Refresh default router list. 1190 */ 1191 defrouter_select(); 1192 } 1193 1194 if (ln->ln_router || dr) 1195 LLE_WLOCK(ln); 1196 } 1197 1198 /* 1199 * Before deleting the entry, remember the next entry as the 1200 * return value. We need this because pfxlist_onlink_check() above 1201 * might have freed other entries (particularly the old next entry) as 1202 * a side effect (XXX). 1203 */ 1204 next = LIST_NEXT(ln, lle_next); 1205 1206 /* 1207 * Save to unlock. We still hold an extra reference and will not 1208 * free(9) in llentry_free() if someone else holds one as well. 1209 */ 1210 LLE_WUNLOCK(ln); 1211 IF_AFDATA_LOCK(ifp); 1212 LLE_WLOCK(ln); 1213 1214 /* Guard against race with other llentry_free(). */ 1215 if (ln->la_flags & LLE_LINKED) { 1216 LLE_REMREF(ln); 1217 llentry_free(ln); 1218 } else 1219 LLE_FREE_LOCKED(ln); 1220 1221 IF_AFDATA_UNLOCK(ifp); 1222 1223 return (next); 1224 } 1225 1226 /* 1227 * Upper-layer reachability hint for Neighbor Unreachability Detection. 1228 * 1229 * XXX cost-effective methods? 1230 */ 1231 void 1232 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) 1233 { 1234 struct llentry *ln; 1235 struct ifnet *ifp; 1236 1237 if ((dst6 == NULL) || (rt == NULL)) 1238 return; 1239 1240 ifp = rt->rt_ifp; 1241 IF_AFDATA_RLOCK(ifp); 1242 ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); 1243 IF_AFDATA_RUNLOCK(ifp); 1244 if (ln == NULL) 1245 return; 1246 1247 if (ln->ln_state < ND6_LLINFO_REACHABLE) 1248 goto done; 1249 1250 /* 1251 * if we get upper-layer reachability confirmation many times, 1252 * it is possible we have false information. 1253 */ 1254 if (!force) { 1255 ln->ln_byhint++; 1256 if (ln->ln_byhint > V_nd6_maxnudhint) { 1257 goto done; 1258 } 1259 } 1260 1261 ln->ln_state = ND6_LLINFO_REACHABLE; 1262 if (!ND6_LLINFO_PERMANENT(ln)) { 1263 nd6_llinfo_settimer_locked(ln, 1264 (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); 1265 } 1266 done: 1267 LLE_WUNLOCK(ln); 1268 } 1269 1270 1271 /* 1272 * Rejuvenate this function for routing operations related 1273 * processing. 1274 */ 1275 void 1276 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) 1277 { 1278 struct sockaddr_in6 *gateway; 1279 struct nd_defrouter *dr; 1280 struct ifnet *ifp; 1281 1282 gateway = (struct sockaddr_in6 *)rt->rt_gateway; 1283 ifp = rt->rt_ifp; 1284 1285 switch (req) { 1286 case RTM_ADD: 1287 break; 1288 1289 case RTM_DELETE: 1290 if (!ifp) 1291 return; 1292 /* 1293 * Only indirect routes are interesting. 1294 */ 1295 if ((rt->rt_flags & RTF_GATEWAY) == 0) 1296 return; 1297 /* 1298 * check for default route 1299 */ 1300 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 1301 &SIN6(rt_key(rt))->sin6_addr)) { 1302 1303 dr = defrouter_lookup(&gateway->sin6_addr, ifp); 1304 if (dr != NULL) 1305 dr->installed = 0; 1306 } 1307 break; 1308 } 1309 } 1310 1311 1312 int 1313 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) 1314 { 1315 struct in6_drlist *drl = (struct in6_drlist *)data; 1316 struct in6_oprlist *oprl = (struct in6_oprlist *)data; 1317 struct in6_ndireq *ndi = (struct in6_ndireq *)data; 1318 struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data; 1319 struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; 1320 struct nd_defrouter *dr; 1321 struct nd_prefix *pr; 1322 int i = 0, error = 0; 1323 1324 if (ifp->if_afdata[AF_INET6] == NULL) 1325 return (EPFNOSUPPORT); 1326 switch (cmd) { 1327 case SIOCGDRLST_IN6: 1328 /* 1329 * obsolete API, use sysctl under net.inet6.icmp6 1330 */ 1331 bzero(drl, sizeof(*drl)); 1332 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 1333 if (i >= DRLSTSIZ) 1334 break; 1335 drl->defrouter[i].rtaddr = dr->rtaddr; 1336 in6_clearscope(&drl->defrouter[i].rtaddr); 1337 1338 drl->defrouter[i].flags = dr->flags; 1339 drl->defrouter[i].rtlifetime = dr->rtlifetime; 1340 drl->defrouter[i].expire = dr->expire + 1341 (time_second - time_uptime); 1342 drl->defrouter[i].if_index = dr->ifp->if_index; 1343 i++; 1344 } 1345 break; 1346 case SIOCGPRLST_IN6: 1347 /* 1348 * obsolete API, use sysctl under net.inet6.icmp6 1349 * 1350 * XXX the structure in6_prlist was changed in backward- 1351 * incompatible manner. in6_oprlist is used for SIOCGPRLST_IN6, 1352 * in6_prlist is used for nd6_sysctl() - fill_prlist(). 1353 */ 1354 /* 1355 * XXX meaning of fields, especialy "raflags", is very 1356 * differnet between RA prefix list and RR/static prefix list. 1357 * how about separating ioctls into two? 1358 */ 1359 bzero(oprl, sizeof(*oprl)); 1360 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 1361 struct nd_pfxrouter *pfr; 1362 int j; 1363 1364 if (i >= PRLSTSIZ) 1365 break; 1366 oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr; 1367 oprl->prefix[i].raflags = pr->ndpr_raf; 1368 oprl->prefix[i].prefixlen = pr->ndpr_plen; 1369 oprl->prefix[i].vltime = pr->ndpr_vltime; 1370 oprl->prefix[i].pltime = pr->ndpr_pltime; 1371 oprl->prefix[i].if_index = pr->ndpr_ifp->if_index; 1372 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 1373 oprl->prefix[i].expire = 0; 1374 else { 1375 time_t maxexpire; 1376 1377 /* XXX: we assume time_t is signed. */ 1378 maxexpire = (-1) & 1379 ~((time_t)1 << 1380 ((sizeof(maxexpire) * 8) - 1)); 1381 if (pr->ndpr_vltime < 1382 maxexpire - pr->ndpr_lastupdate) { 1383 oprl->prefix[i].expire = 1384 pr->ndpr_lastupdate + 1385 pr->ndpr_vltime + 1386 (time_second - time_uptime); 1387 } else 1388 oprl->prefix[i].expire = maxexpire; 1389 } 1390 1391 j = 0; 1392 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 1393 if (j < DRLSTSIZ) { 1394 #define RTRADDR oprl->prefix[i].advrtr[j] 1395 RTRADDR = pfr->router->rtaddr; 1396 in6_clearscope(&RTRADDR); 1397 #undef RTRADDR 1398 } 1399 j++; 1400 } 1401 oprl->prefix[i].advrtrs = j; 1402 oprl->prefix[i].origin = PR_ORIG_RA; 1403 1404 i++; 1405 } 1406 1407 break; 1408 case OSIOCGIFINFO_IN6: 1409 #define ND ndi->ndi 1410 /* XXX: old ndp(8) assumes a positive value for linkmtu. */ 1411 bzero(&ND, sizeof(ND)); 1412 ND.linkmtu = IN6_LINKMTU(ifp); 1413 ND.maxmtu = ND_IFINFO(ifp)->maxmtu; 1414 ND.basereachable = ND_IFINFO(ifp)->basereachable; 1415 ND.reachable = ND_IFINFO(ifp)->reachable; 1416 ND.retrans = ND_IFINFO(ifp)->retrans; 1417 ND.flags = ND_IFINFO(ifp)->flags; 1418 ND.recalctm = ND_IFINFO(ifp)->recalctm; 1419 ND.chlim = ND_IFINFO(ifp)->chlim; 1420 break; 1421 case SIOCGIFINFO_IN6: 1422 ND = *ND_IFINFO(ifp); 1423 break; 1424 case SIOCSIFINFO_IN6: 1425 /* 1426 * used to change host variables from userland. 1427 * intented for a use on router to reflect RA configurations. 1428 */ 1429 /* 0 means 'unspecified' */ 1430 if (ND.linkmtu != 0) { 1431 if (ND.linkmtu < IPV6_MMTU || 1432 ND.linkmtu > IN6_LINKMTU(ifp)) { 1433 error = EINVAL; 1434 break; 1435 } 1436 ND_IFINFO(ifp)->linkmtu = ND.linkmtu; 1437 } 1438 1439 if (ND.basereachable != 0) { 1440 int obasereachable = ND_IFINFO(ifp)->basereachable; 1441 1442 ND_IFINFO(ifp)->basereachable = ND.basereachable; 1443 if (ND.basereachable != obasereachable) 1444 ND_IFINFO(ifp)->reachable = 1445 ND_COMPUTE_RTIME(ND.basereachable); 1446 } 1447 if (ND.retrans != 0) 1448 ND_IFINFO(ifp)->retrans = ND.retrans; 1449 if (ND.chlim != 0) 1450 ND_IFINFO(ifp)->chlim = ND.chlim; 1451 /* FALLTHROUGH */ 1452 case SIOCSIFINFO_FLAGS: 1453 { 1454 struct ifaddr *ifa; 1455 struct in6_ifaddr *ia; 1456 1457 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1458 !(ND.flags & ND6_IFF_IFDISABLED)) { 1459 /* ifdisabled 1->0 transision */ 1460 1461 /* 1462 * If the interface is marked as ND6_IFF_IFDISABLED and 1463 * has an link-local address with IN6_IFF_DUPLICATED, 1464 * do not clear ND6_IFF_IFDISABLED. 1465 * See RFC 4862, Section 5.4.5. 1466 */ 1467 int duplicated_linklocal = 0; 1468 1469 IF_ADDR_RLOCK(ifp); 1470 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1471 if (ifa->ifa_addr->sa_family != AF_INET6) 1472 continue; 1473 ia = (struct in6_ifaddr *)ifa; 1474 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) && 1475 IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1476 duplicated_linklocal = 1; 1477 break; 1478 } 1479 } 1480 IF_ADDR_RUNLOCK(ifp); 1481 1482 if (duplicated_linklocal) { 1483 ND.flags |= ND6_IFF_IFDISABLED; 1484 log(LOG_ERR, "Cannot enable an interface" 1485 " with a link-local address marked" 1486 " duplicate.\n"); 1487 } else { 1488 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED; 1489 if (ifp->if_flags & IFF_UP) 1490 in6_if_up(ifp); 1491 } 1492 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) && 1493 (ND.flags & ND6_IFF_IFDISABLED)) { 1494 /* ifdisabled 0->1 transision */ 1495 /* Mark all IPv6 address as tentative. */ 1496 1497 ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED; 1498 IF_ADDR_RLOCK(ifp); 1499 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1500 if (ifa->ifa_addr->sa_family != AF_INET6) 1501 continue; 1502 ia = (struct in6_ifaddr *)ifa; 1503 ia->ia6_flags |= IN6_IFF_TENTATIVE; 1504 } 1505 IF_ADDR_RUNLOCK(ifp); 1506 } 1507 1508 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) { 1509 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) { 1510 /* auto_linklocal 0->1 transision */ 1511 1512 /* If no link-local address on ifp, configure */ 1513 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL; 1514 in6_ifattach(ifp, NULL); 1515 } else if (!(ND.flags & ND6_IFF_IFDISABLED) && 1516 ifp->if_flags & IFF_UP) { 1517 /* 1518 * When the IF already has 1519 * ND6_IFF_AUTO_LINKLOCAL, no link-local 1520 * address is assigned, and IFF_UP, try to 1521 * assign one. 1522 */ 1523 int haslinklocal = 0; 1524 1525 IF_ADDR_RLOCK(ifp); 1526 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 1527 if (ifa->ifa_addr->sa_family != AF_INET6) 1528 continue; 1529 ia = (struct in6_ifaddr *)ifa; 1530 if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia))) { 1531 haslinklocal = 1; 1532 break; 1533 } 1534 } 1535 IF_ADDR_RUNLOCK(ifp); 1536 if (!haslinklocal) 1537 in6_ifattach(ifp, NULL); 1538 } 1539 } 1540 } 1541 ND_IFINFO(ifp)->flags = ND.flags; 1542 break; 1543 #undef ND 1544 case SIOCSNDFLUSH_IN6: /* XXX: the ioctl name is confusing... */ 1545 /* sync kernel routing table with the default router list */ 1546 defrouter_reset(); 1547 defrouter_select(); 1548 break; 1549 case SIOCSPFXFLUSH_IN6: 1550 { 1551 /* flush all the prefix advertised by routers */ 1552 struct nd_prefix *pr, *next; 1553 1554 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) { 1555 struct in6_ifaddr *ia, *ia_next; 1556 1557 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr)) 1558 continue; /* XXX */ 1559 1560 /* do we really have to remove addresses as well? */ 1561 /* XXXRW: in6_ifaddrhead locking. */ 1562 TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link, 1563 ia_next) { 1564 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0) 1565 continue; 1566 1567 if (ia->ia6_ndpr == pr) 1568 in6_purgeaddr(&ia->ia_ifa); 1569 } 1570 prelist_remove(pr); 1571 } 1572 break; 1573 } 1574 case SIOCSRTRFLUSH_IN6: 1575 { 1576 /* flush all the default routers */ 1577 struct nd_defrouter *dr, *next; 1578 1579 defrouter_reset(); 1580 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) { 1581 defrtrlist_del(dr); 1582 } 1583 defrouter_select(); 1584 break; 1585 } 1586 case SIOCGNBRINFO_IN6: 1587 { 1588 struct llentry *ln; 1589 struct in6_addr nb_addr = nbi->addr; /* make local for safety */ 1590 1591 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) 1592 return (error); 1593 1594 IF_AFDATA_RLOCK(ifp); 1595 ln = nd6_lookup(&nb_addr, 0, ifp); 1596 IF_AFDATA_RUNLOCK(ifp); 1597 1598 if (ln == NULL) { 1599 error = EINVAL; 1600 break; 1601 } 1602 nbi->state = ln->ln_state; 1603 nbi->asked = ln->la_asked; 1604 nbi->isrouter = ln->ln_router; 1605 if (ln->la_expire == 0) 1606 nbi->expire = 0; 1607 else 1608 nbi->expire = ln->la_expire + 1609 (time_second - time_uptime); 1610 LLE_RUNLOCK(ln); 1611 break; 1612 } 1613 case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1614 ndif->ifindex = V_nd6_defifindex; 1615 break; 1616 case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ 1617 return (nd6_setdefaultiface(ndif->ifindex)); 1618 } 1619 return (error); 1620 } 1621 1622 /* 1623 * Create neighbor cache entry and cache link-layer address, 1624 * on reception of inbound ND6 packets. (RS/RA/NS/redirect) 1625 * 1626 * type - ICMP6 type 1627 * code - type dependent information 1628 * 1629 * XXXXX 1630 * The caller of this function already acquired the ndp 1631 * cache table lock because the cache entry is returned. 1632 */ 1633 struct llentry * 1634 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, 1635 int lladdrlen, int type, int code) 1636 { 1637 struct llentry *ln = NULL, *ln_tmp; 1638 int is_newentry; 1639 int do_update; 1640 int olladdr; 1641 int llchange; 1642 int flags; 1643 int newstate = 0; 1644 uint16_t router = 0; 1645 struct sockaddr_in6 sin6; 1646 struct mbuf *chain = NULL; 1647 int static_route = 0; 1648 1649 IF_AFDATA_UNLOCK_ASSERT(ifp); 1650 1651 KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__)); 1652 KASSERT(from != NULL, ("%s: from == NULL", __func__)); 1653 1654 /* nothing must be updated for unspecified address */ 1655 if (IN6_IS_ADDR_UNSPECIFIED(from)) 1656 return NULL; 1657 1658 /* 1659 * Validation about ifp->if_addrlen and lladdrlen must be done in 1660 * the caller. 1661 * 1662 * XXX If the link does not have link-layer adderss, what should 1663 * we do? (ifp->if_addrlen == 0) 1664 * Spec says nothing in sections for RA, RS and NA. There's small 1665 * description on it in NS section (RFC 2461 7.2.3). 1666 */ 1667 flags = lladdr ? ND6_EXCLUSIVE : 0; 1668 IF_AFDATA_RLOCK(ifp); 1669 ln = nd6_lookup(from, flags, ifp); 1670 IF_AFDATA_RUNLOCK(ifp); 1671 is_newentry = 0; 1672 if (ln == NULL) { 1673 flags |= ND6_EXCLUSIVE; 1674 ln = nd6_alloc(from, 0, ifp); 1675 if (ln == NULL) 1676 return (NULL); 1677 IF_AFDATA_WLOCK(ifp); 1678 LLE_WLOCK(ln); 1679 /* Prefer any existing lle over newly-created one */ 1680 ln_tmp = nd6_lookup(from, ND6_EXCLUSIVE, ifp); 1681 if (ln_tmp == NULL) 1682 lltable_link_entry(LLTABLE6(ifp), ln); 1683 IF_AFDATA_WUNLOCK(ifp); 1684 if (ln_tmp == NULL) 1685 /* No existing lle, mark as new entry */ 1686 is_newentry = 1; 1687 else { 1688 lltable_free_entry(LLTABLE6(ifp), ln); 1689 ln = ln_tmp; 1690 ln_tmp = NULL; 1691 } 1692 } 1693 /* do nothing if static ndp is set */ 1694 if ((ln->la_flags & LLE_STATIC)) { 1695 static_route = 1; 1696 goto done; 1697 } 1698 1699 olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; 1700 if (olladdr && lladdr) { 1701 llchange = bcmp(lladdr, &ln->ll_addr, 1702 ifp->if_addrlen); 1703 } else 1704 llchange = 0; 1705 1706 /* 1707 * newentry olladdr lladdr llchange (*=record) 1708 * 0 n n -- (1) 1709 * 0 y n -- (2) 1710 * 0 n y -- (3) * STALE 1711 * 0 y y n (4) * 1712 * 0 y y y (5) * STALE 1713 * 1 -- n -- (6) NOSTATE(= PASSIVE) 1714 * 1 -- y -- (7) * STALE 1715 */ 1716 1717 if (lladdr) { /* (3-5) and (7) */ 1718 /* 1719 * Record source link-layer address 1720 * XXX is it dependent to ifp->if_type? 1721 */ 1722 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); 1723 ln->la_flags |= LLE_VALID; 1724 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 1725 } 1726 1727 if (!is_newentry) { 1728 if ((!olladdr && lladdr != NULL) || /* (3) */ 1729 (olladdr && lladdr != NULL && llchange)) { /* (5) */ 1730 do_update = 1; 1731 newstate = ND6_LLINFO_STALE; 1732 } else /* (1-2,4) */ 1733 do_update = 0; 1734 } else { 1735 do_update = 1; 1736 if (lladdr == NULL) /* (6) */ 1737 newstate = ND6_LLINFO_NOSTATE; 1738 else /* (7) */ 1739 newstate = ND6_LLINFO_STALE; 1740 } 1741 1742 if (do_update) { 1743 /* 1744 * Update the state of the neighbor cache. 1745 */ 1746 ln->ln_state = newstate; 1747 1748 if (ln->ln_state == ND6_LLINFO_STALE) { 1749 if (ln->la_hold != NULL) 1750 nd6_grab_holdchain(ln, &chain, &sin6); 1751 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { 1752 /* probe right away */ 1753 nd6_llinfo_settimer_locked((void *)ln, 0); 1754 } 1755 } 1756 1757 /* 1758 * ICMP6 type dependent behavior. 1759 * 1760 * NS: clear IsRouter if new entry 1761 * RS: clear IsRouter 1762 * RA: set IsRouter if there's lladdr 1763 * redir: clear IsRouter if new entry 1764 * 1765 * RA case, (1): 1766 * The spec says that we must set IsRouter in the following cases: 1767 * - If lladdr exist, set IsRouter. This means (1-5). 1768 * - If it is old entry (!newentry), set IsRouter. This means (7). 1769 * So, based on the spec, in (1-5) and (7) cases we must set IsRouter. 1770 * A quetion arises for (1) case. (1) case has no lladdr in the 1771 * neighbor cache, this is similar to (6). 1772 * This case is rare but we figured that we MUST NOT set IsRouter. 1773 * 1774 * newentry olladdr lladdr llchange NS RS RA redir 1775 * D R 1776 * 0 n n -- (1) c ? s 1777 * 0 y n -- (2) c s s 1778 * 0 n y -- (3) c s s 1779 * 0 y y n (4) c s s 1780 * 0 y y y (5) c s s 1781 * 1 -- n -- (6) c c c s 1782 * 1 -- y -- (7) c c s c s 1783 * 1784 * (c=clear s=set) 1785 */ 1786 switch (type & 0xff) { 1787 case ND_NEIGHBOR_SOLICIT: 1788 /* 1789 * New entry must have is_router flag cleared. 1790 */ 1791 if (is_newentry) /* (6-7) */ 1792 ln->ln_router = 0; 1793 break; 1794 case ND_REDIRECT: 1795 /* 1796 * If the icmp is a redirect to a better router, always set the 1797 * is_router flag. Otherwise, if the entry is newly created, 1798 * clear the flag. [RFC 2461, sec 8.3] 1799 */ 1800 if (code == ND_REDIRECT_ROUTER) 1801 ln->ln_router = 1; 1802 else if (is_newentry) /* (6-7) */ 1803 ln->ln_router = 0; 1804 break; 1805 case ND_ROUTER_SOLICIT: 1806 /* 1807 * is_router flag must always be cleared. 1808 */ 1809 ln->ln_router = 0; 1810 break; 1811 case ND_ROUTER_ADVERT: 1812 /* 1813 * Mark an entry with lladdr as a router. 1814 */ 1815 if ((!is_newentry && (olladdr || lladdr)) || /* (2-5) */ 1816 (is_newentry && lladdr)) { /* (7) */ 1817 ln->ln_router = 1; 1818 } 1819 break; 1820 } 1821 1822 if (ln != NULL) { 1823 static_route = (ln->la_flags & LLE_STATIC); 1824 router = ln->ln_router; 1825 1826 if (flags & ND6_EXCLUSIVE) 1827 LLE_WUNLOCK(ln); 1828 else 1829 LLE_RUNLOCK(ln); 1830 if (static_route) 1831 ln = NULL; 1832 } 1833 if (chain != NULL) 1834 nd6_flush_holdchain(ifp, ifp, chain, &sin6); 1835 1836 /* 1837 * When the link-layer address of a router changes, select the 1838 * best router again. In particular, when the neighbor entry is newly 1839 * created, it might affect the selection policy. 1840 * Question: can we restrict the first condition to the "is_newentry" 1841 * case? 1842 * XXX: when we hear an RA from a new router with the link-layer 1843 * address option, defrouter_select() is called twice, since 1844 * defrtrlist_update called the function as well. However, I believe 1845 * we can compromise the overhead, since it only happens the first 1846 * time. 1847 * XXX: although defrouter_select() should not have a bad effect 1848 * for those are not autoconfigured hosts, we explicitly avoid such 1849 * cases for safety. 1850 */ 1851 if (do_update && router && 1852 ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) { 1853 /* 1854 * guaranteed recursion 1855 */ 1856 defrouter_select(); 1857 } 1858 1859 return (ln); 1860 done: 1861 if (ln != NULL) { 1862 if (flags & ND6_EXCLUSIVE) 1863 LLE_WUNLOCK(ln); 1864 else 1865 LLE_RUNLOCK(ln); 1866 if (static_route) 1867 ln = NULL; 1868 } 1869 return (ln); 1870 } 1871 1872 static void 1873 nd6_slowtimo(void *arg) 1874 { 1875 CURVNET_SET((struct vnet *) arg); 1876 struct nd_ifinfo *nd6if; 1877 struct ifnet *ifp; 1878 1879 callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz, 1880 nd6_slowtimo, curvnet); 1881 IFNET_RLOCK_NOSLEEP(); 1882 TAILQ_FOREACH(ifp, &V_ifnet, if_link) { 1883 if (ifp->if_afdata[AF_INET6] == NULL) 1884 continue; 1885 nd6if = ND_IFINFO(ifp); 1886 if (nd6if->basereachable && /* already initialized */ 1887 (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) { 1888 /* 1889 * Since reachable time rarely changes by router 1890 * advertisements, we SHOULD insure that a new random 1891 * value gets recomputed at least once every few hours. 1892 * (RFC 2461, 6.3.4) 1893 */ 1894 nd6if->recalctm = V_nd6_recalc_reachtm_interval; 1895 nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable); 1896 } 1897 } 1898 IFNET_RUNLOCK_NOSLEEP(); 1899 CURVNET_RESTORE(); 1900 } 1901 1902 void 1903 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain, 1904 struct sockaddr_in6 *sin6) 1905 { 1906 1907 LLE_WLOCK_ASSERT(ln); 1908 1909 *chain = ln->la_hold; 1910 ln->la_hold = NULL; 1911 lltable_fill_sa_entry(ln, (struct sockaddr *)sin6); 1912 1913 if (ln->ln_state == ND6_LLINFO_STALE) { 1914 1915 /* 1916 * The first time we send a packet to a 1917 * neighbor whose entry is STALE, we have 1918 * to change the state to DELAY and a sets 1919 * a timer to expire in DELAY_FIRST_PROBE_TIME 1920 * seconds to ensure do neighbor unreachability 1921 * detection on expiration. 1922 * (RFC 2461 7.3.3) 1923 */ 1924 ln->la_asked = 0; 1925 ln->ln_state = ND6_LLINFO_DELAY; 1926 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); 1927 } 1928 } 1929 1930 static int 1931 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1932 struct sockaddr_in6 *dst) 1933 { 1934 int error; 1935 int ip6len; 1936 struct ip6_hdr *ip6; 1937 struct m_tag *mtag; 1938 1939 #ifdef MAC 1940 mac_netinet6_nd6_send(ifp, m); 1941 #endif 1942 1943 /* 1944 * If called from nd6_ns_output() (NS), nd6_na_output() (NA), 1945 * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA 1946 * as handled by rtsol and rtadvd), mbufs will be tagged for SeND 1947 * to be diverted to user space. When re-injected into the kernel, 1948 * send_output() will directly dispatch them to the outgoing interface. 1949 */ 1950 if (send_sendso_input_hook != NULL) { 1951 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL); 1952 if (mtag != NULL) { 1953 ip6 = mtod(m, struct ip6_hdr *); 1954 ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen); 1955 /* Use the SEND socket */ 1956 error = send_sendso_input_hook(m, ifp, SND_OUT, 1957 ip6len); 1958 /* -1 == no app on SEND socket */ 1959 if (error == 0 || error != -1) 1960 return (error); 1961 } 1962 } 1963 1964 m_clrprotoflags(m); /* Avoid confusing lower layers. */ 1965 IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL, 1966 mtod(m, struct ip6_hdr *)); 1967 1968 if ((ifp->if_flags & IFF_LOOPBACK) == 0) 1969 origifp = ifp; 1970 1971 error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, NULL); 1972 return (error); 1973 } 1974 1975 /* 1976 * IPv6 packet output - light version. 1977 * Checks if destination LLE exists and is in proper state 1978 * (e.g no modification required). If not true, fall back to 1979 * "heavy" version. 1980 */ 1981 int 1982 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 1983 struct sockaddr_in6 *dst, struct rtentry *rt0) 1984 { 1985 struct llentry *ln = NULL; 1986 1987 /* discard the packet if IPv6 operation is disabled on the interface */ 1988 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 1989 m_freem(m); 1990 return (ENETDOWN); /* better error? */ 1991 } 1992 1993 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 1994 goto sendpkt; 1995 1996 if (nd6_need_cache(ifp) == 0) 1997 goto sendpkt; 1998 1999 IF_AFDATA_RLOCK(ifp); 2000 ln = nd6_lookup(&dst->sin6_addr, 0, ifp); 2001 IF_AFDATA_RUNLOCK(ifp); 2002 2003 /* 2004 * Perform fast path for the following cases: 2005 * 1) lle state is REACHABLE 2006 * 2) lle state is DELAY (NS message sentNS message sent) 2007 * 2008 * Every other case involves lle modification, so we handle 2009 * them separately. 2010 */ 2011 if (ln == NULL || (ln->ln_state != ND6_LLINFO_REACHABLE && 2012 ln->ln_state != ND6_LLINFO_DELAY)) { 2013 /* Fall back to slow processing path */ 2014 if (ln != NULL) 2015 LLE_RUNLOCK(ln); 2016 return (nd6_output_lle(ifp, origifp, m, dst)); 2017 } 2018 2019 sendpkt: 2020 if (ln != NULL) 2021 LLE_RUNLOCK(ln); 2022 2023 return (nd6_output_ifp(ifp, origifp, m, dst)); 2024 } 2025 2026 2027 /* 2028 * Output IPv6 packet - heavy version. 2029 * Function assume that either 2030 * 1) destination LLE does not exist, is invalid or stale, so 2031 * ND6_EXCLUSIVE lock needs to be acquired 2032 * 2) destination lle is provided (with ND6_EXCLUSIVE lock), 2033 * in that case packets are queued in &chain. 2034 * 2035 */ 2036 static int 2037 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m, 2038 struct sockaddr_in6 *dst) 2039 { 2040 struct llentry *lle = NULL, *lle_tmp; 2041 2042 KASSERT(m != NULL, ("NULL mbuf, nothing to send")); 2043 /* discard the packet if IPv6 operation is disabled on the interface */ 2044 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) { 2045 m_freem(m); 2046 return (ENETDOWN); /* better error? */ 2047 } 2048 2049 if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) 2050 goto sendpkt; 2051 2052 if (nd6_need_cache(ifp) == 0) 2053 goto sendpkt; 2054 2055 /* 2056 * Address resolution or Neighbor Unreachability Detection 2057 * for the next hop. 2058 * At this point, the destination of the packet must be a unicast 2059 * or an anycast address(i.e. not a multicast). 2060 */ 2061 if (lle == NULL) { 2062 IF_AFDATA_RLOCK(ifp); 2063 lle = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 2064 IF_AFDATA_RUNLOCK(ifp); 2065 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp)) { 2066 /* 2067 * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), 2068 * the condition below is not very efficient. But we believe 2069 * it is tolerable, because this should be a rare case. 2070 */ 2071 lle = nd6_alloc(&dst->sin6_addr, 0, ifp); 2072 if (lle == NULL) { 2073 char ip6buf[INET6_ADDRSTRLEN]; 2074 log(LOG_DEBUG, 2075 "nd6_output: can't allocate llinfo for %s " 2076 "(ln=%p)\n", 2077 ip6_sprintf(ip6buf, &dst->sin6_addr), lle); 2078 m_freem(m); 2079 return (ENOBUFS); 2080 } 2081 lle->ln_state = ND6_LLINFO_NOSTATE; 2082 2083 IF_AFDATA_WLOCK(ifp); 2084 LLE_WLOCK(lle); 2085 /* Prefer any existing entry over newly-created one */ 2086 lle_tmp = nd6_lookup(&dst->sin6_addr, ND6_EXCLUSIVE, ifp); 2087 if (lle_tmp == NULL) 2088 lltable_link_entry(LLTABLE6(ifp), lle); 2089 IF_AFDATA_WUNLOCK(ifp); 2090 if (lle_tmp != NULL) { 2091 lltable_free_entry(LLTABLE6(ifp), lle); 2092 lle = lle_tmp; 2093 lle_tmp = NULL; 2094 } 2095 } 2096 } 2097 if (lle == NULL) { 2098 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && 2099 !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { 2100 m_freem(m); 2101 return (ENOBUFS); 2102 } 2103 goto sendpkt; /* send anyway */ 2104 } 2105 2106 LLE_WLOCK_ASSERT(lle); 2107 2108 /* We don't have to do link-layer address resolution on a p2p link. */ 2109 if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && 2110 lle->ln_state < ND6_LLINFO_REACHABLE) { 2111 lle->ln_state = ND6_LLINFO_STALE; 2112 nd6_llinfo_settimer_locked(lle, (long)V_nd6_gctimer * hz); 2113 } 2114 2115 /* 2116 * The first time we send a packet to a neighbor whose entry is 2117 * STALE, we have to change the state to DELAY and a sets a timer to 2118 * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do 2119 * neighbor unreachability detection on expiration. 2120 * (RFC 2461 7.3.3) 2121 */ 2122 if (lle->ln_state == ND6_LLINFO_STALE) { 2123 lle->la_asked = 0; 2124 lle->ln_state = ND6_LLINFO_DELAY; 2125 nd6_llinfo_settimer_locked(lle, (long)V_nd6_delay * hz); 2126 } 2127 2128 /* 2129 * If the neighbor cache entry has a state other than INCOMPLETE 2130 * (i.e. its link-layer address is already resolved), just 2131 * send the packet. 2132 */ 2133 if (lle->ln_state > ND6_LLINFO_INCOMPLETE) 2134 goto sendpkt; 2135 2136 /* 2137 * There is a neighbor cache entry, but no ethernet address 2138 * response yet. Append this latest packet to the end of the 2139 * packet queue in the mbuf, unless the number of the packet 2140 * does not exceed nd6_maxqueuelen. When it exceeds nd6_maxqueuelen, 2141 * the oldest packet in the queue will be removed. 2142 */ 2143 if (lle->ln_state == ND6_LLINFO_NOSTATE) 2144 lle->ln_state = ND6_LLINFO_INCOMPLETE; 2145 2146 if (lle->la_hold != NULL) { 2147 struct mbuf *m_hold; 2148 int i; 2149 2150 i = 0; 2151 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){ 2152 i++; 2153 if (m_hold->m_nextpkt == NULL) { 2154 m_hold->m_nextpkt = m; 2155 break; 2156 } 2157 } 2158 while (i >= V_nd6_maxqueuelen) { 2159 m_hold = lle->la_hold; 2160 lle->la_hold = lle->la_hold->m_nextpkt; 2161 m_freem(m_hold); 2162 i--; 2163 } 2164 } else { 2165 lle->la_hold = m; 2166 } 2167 2168 /* 2169 * If there has been no NS for the neighbor after entering the 2170 * INCOMPLETE state, send the first solicitation. 2171 */ 2172 if (!ND6_LLINFO_PERMANENT(lle) && lle->la_asked == 0) { 2173 lle->la_asked++; 2174 2175 nd6_llinfo_settimer_locked(lle, 2176 (long)ND_IFINFO(ifp)->retrans * hz / 1000); 2177 LLE_WUNLOCK(lle); 2178 nd6_ns_output(ifp, NULL, &dst->sin6_addr, lle, NULL); 2179 } else { 2180 /* We did the lookup so we need to do the unlock here. */ 2181 LLE_WUNLOCK(lle); 2182 } 2183 2184 return (0); 2185 2186 sendpkt: 2187 if (lle != NULL) 2188 LLE_WUNLOCK(lle); 2189 2190 return (nd6_output_ifp(ifp, origifp, m, dst)); 2191 } 2192 2193 2194 int 2195 nd6_flush_holdchain(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, 2196 struct sockaddr_in6 *dst) 2197 { 2198 struct mbuf *m, *m_head; 2199 struct ifnet *outifp; 2200 int error = 0; 2201 2202 m_head = chain; 2203 if ((ifp->if_flags & IFF_LOOPBACK) != 0) 2204 outifp = origifp; 2205 else 2206 outifp = ifp; 2207 2208 while (m_head) { 2209 m = m_head; 2210 m_head = m_head->m_nextpkt; 2211 error = nd6_output_ifp(ifp, origifp, m, dst); 2212 } 2213 2214 /* 2215 * XXX 2216 * note that intermediate errors are blindly ignored - but this is 2217 * the same convention as used with nd6_output when called by 2218 * nd6_cache_lladdr 2219 */ 2220 return (error); 2221 } 2222 2223 2224 int 2225 nd6_need_cache(struct ifnet *ifp) 2226 { 2227 /* 2228 * XXX: we currently do not make neighbor cache on any interface 2229 * other than ARCnet, Ethernet, FDDI and GIF. 2230 * 2231 * RFC2893 says: 2232 * - unidirectional tunnels needs no ND 2233 */ 2234 switch (ifp->if_type) { 2235 case IFT_ARCNET: 2236 case IFT_ETHER: 2237 case IFT_FDDI: 2238 case IFT_IEEE1394: 2239 case IFT_L2VLAN: 2240 case IFT_IEEE80211: 2241 case IFT_INFINIBAND: 2242 case IFT_BRIDGE: 2243 case IFT_PROPVIRTUAL: 2244 return (1); 2245 default: 2246 return (0); 2247 } 2248 } 2249 2250 /* 2251 * Add pernament ND6 link-layer record for given 2252 * interface address. 2253 * 2254 * Very similar to IPv4 arp_ifinit(), but: 2255 * 1) IPv6 DAD is performed in different place 2256 * 2) It is called by IPv6 protocol stack in contrast to 2257 * arp_ifinit() which is typically called in SIOCSIFADDR 2258 * driver ioctl handler. 2259 * 2260 */ 2261 int 2262 nd6_add_ifa_lle(struct in6_ifaddr *ia) 2263 { 2264 struct ifnet *ifp; 2265 struct llentry *ln, *ln_tmp; 2266 struct sockaddr *dst; 2267 2268 ifp = ia->ia_ifa.ifa_ifp; 2269 if (nd6_need_cache(ifp) == 0) 2270 return (0); 2271 2272 ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; 2273 dst = (struct sockaddr *)&ia->ia_addr; 2274 ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst); 2275 if (ln == NULL) 2276 return (ENOBUFS); 2277 2278 IF_AFDATA_WLOCK(ifp); 2279 LLE_WLOCK(ln); 2280 /* Unlink any entry if exists */ 2281 ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst); 2282 if (ln_tmp != NULL) 2283 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp); 2284 lltable_link_entry(LLTABLE6(ifp), ln); 2285 IF_AFDATA_WUNLOCK(ifp); 2286 2287 if (ln_tmp != NULL) 2288 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED); 2289 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED); 2290 2291 LLE_WUNLOCK(ln); 2292 if (ln_tmp != NULL) 2293 llentry_free(ln_tmp); 2294 2295 return (0); 2296 } 2297 2298 /* 2299 * Removes ALL lle records for interface address prefix. 2300 * XXXME: That's probably not we really want to do, we need 2301 * to remove address record only and keep other records 2302 * until we determine if given prefix is really going 2303 * to be removed. 2304 */ 2305 void 2306 nd6_rem_ifa_lle(struct in6_ifaddr *ia) 2307 { 2308 struct sockaddr_in6 mask, addr; 2309 struct ifnet *ifp; 2310 2311 ifp = ia->ia_ifa.ifa_ifp; 2312 memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr)); 2313 memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask)); 2314 lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr, 2315 (struct sockaddr *)&mask, LLE_STATIC); 2316 } 2317 2318 /* 2319 * the callers of this function need to be re-worked to drop 2320 * the lle lock, drop here for now 2321 */ 2322 int 2323 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m, 2324 const struct sockaddr *dst, u_char *desten, uint32_t *pflags) 2325 { 2326 struct llentry *ln; 2327 2328 if (pflags != NULL) 2329 *pflags = 0; 2330 IF_AFDATA_UNLOCK_ASSERT(ifp); 2331 if (m != NULL && m->m_flags & M_MCAST) { 2332 switch (ifp->if_type) { 2333 case IFT_ETHER: 2334 case IFT_FDDI: 2335 case IFT_L2VLAN: 2336 case IFT_IEEE80211: 2337 case IFT_BRIDGE: 2338 case IFT_ISO88025: 2339 ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr, 2340 desten); 2341 return (0); 2342 default: 2343 m_freem(m); 2344 return (EAFNOSUPPORT); 2345 } 2346 } 2347 2348 2349 /* 2350 * the entry should have been created in nd6_store_lladdr 2351 */ 2352 IF_AFDATA_RLOCK(ifp); 2353 ln = lla_lookup(LLTABLE6(ifp), 0, dst); 2354 IF_AFDATA_RUNLOCK(ifp); 2355 if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { 2356 if (ln != NULL) 2357 LLE_RUNLOCK(ln); 2358 /* this could happen, if we could not allocate memory */ 2359 m_freem(m); 2360 return (1); 2361 } 2362 2363 bcopy(&ln->ll_addr, desten, ifp->if_addrlen); 2364 if (pflags != NULL) 2365 *pflags = ln->la_flags; 2366 LLE_RUNLOCK(ln); 2367 /* 2368 * A *small* use after free race exists here 2369 */ 2370 return (0); 2371 } 2372 2373 static void 2374 clear_llinfo_pqueue(struct llentry *ln) 2375 { 2376 struct mbuf *m_hold, *m_hold_next; 2377 2378 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { 2379 m_hold_next = m_hold->m_nextpkt; 2380 m_freem(m_hold); 2381 } 2382 2383 ln->la_hold = NULL; 2384 return; 2385 } 2386 2387 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS); 2388 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS); 2389 #ifdef SYSCTL_DECL 2390 SYSCTL_DECL(_net_inet6_icmp6); 2391 #endif 2392 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist, 2393 CTLFLAG_RD, nd6_sysctl_drlist, ""); 2394 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist, 2395 CTLFLAG_RD, nd6_sysctl_prlist, ""); 2396 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen, 2397 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, ""); 2398 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer, 2399 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), ""); 2400 2401 static int 2402 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS) 2403 { 2404 struct in6_defrouter d; 2405 struct nd_defrouter *dr; 2406 int error; 2407 2408 if (req->newptr) 2409 return (EPERM); 2410 2411 bzero(&d, sizeof(d)); 2412 d.rtaddr.sin6_family = AF_INET6; 2413 d.rtaddr.sin6_len = sizeof(d.rtaddr); 2414 2415 /* 2416 * XXX locking 2417 */ 2418 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) { 2419 d.rtaddr.sin6_addr = dr->rtaddr; 2420 error = sa6_recoverscope(&d.rtaddr); 2421 if (error != 0) 2422 return (error); 2423 d.flags = dr->flags; 2424 d.rtlifetime = dr->rtlifetime; 2425 d.expire = dr->expire + (time_second - time_uptime); 2426 d.if_index = dr->ifp->if_index; 2427 error = SYSCTL_OUT(req, &d, sizeof(d)); 2428 if (error != 0) 2429 return (error); 2430 } 2431 return (0); 2432 } 2433 2434 static int 2435 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS) 2436 { 2437 struct in6_prefix p; 2438 struct sockaddr_in6 s6; 2439 struct nd_prefix *pr; 2440 struct nd_pfxrouter *pfr; 2441 time_t maxexpire; 2442 int error; 2443 char ip6buf[INET6_ADDRSTRLEN]; 2444 2445 if (req->newptr) 2446 return (EPERM); 2447 2448 bzero(&p, sizeof(p)); 2449 p.origin = PR_ORIG_RA; 2450 bzero(&s6, sizeof(s6)); 2451 s6.sin6_family = AF_INET6; 2452 s6.sin6_len = sizeof(s6); 2453 2454 /* 2455 * XXX locking 2456 */ 2457 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) { 2458 p.prefix = pr->ndpr_prefix; 2459 if (sa6_recoverscope(&p.prefix)) { 2460 log(LOG_ERR, "scope error in prefix list (%s)\n", 2461 ip6_sprintf(ip6buf, &p.prefix.sin6_addr)); 2462 /* XXX: press on... */ 2463 } 2464 p.raflags = pr->ndpr_raf; 2465 p.prefixlen = pr->ndpr_plen; 2466 p.vltime = pr->ndpr_vltime; 2467 p.pltime = pr->ndpr_pltime; 2468 p.if_index = pr->ndpr_ifp->if_index; 2469 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME) 2470 p.expire = 0; 2471 else { 2472 /* XXX: we assume time_t is signed. */ 2473 maxexpire = (-1) & 2474 ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1)); 2475 if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate) 2476 p.expire = pr->ndpr_lastupdate + 2477 pr->ndpr_vltime + 2478 (time_second - time_uptime); 2479 else 2480 p.expire = maxexpire; 2481 } 2482 p.refcnt = pr->ndpr_refcnt; 2483 p.flags = pr->ndpr_stateflags; 2484 p.advrtrs = 0; 2485 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) 2486 p.advrtrs++; 2487 error = SYSCTL_OUT(req, &p, sizeof(p)); 2488 if (error != 0) 2489 return (error); 2490 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) { 2491 s6.sin6_addr = pfr->router->rtaddr; 2492 if (sa6_recoverscope(&s6)) 2493 log(LOG_ERR, 2494 "scope error in prefix list (%s)\n", 2495 ip6_sprintf(ip6buf, &pfr->router->rtaddr)); 2496 error = SYSCTL_OUT(req, &s6, sizeof(s6)); 2497 if (error != 0) 2498 return (error); 2499 } 2500 } 2501 return (0); 2502 } 2503