The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/netinet6/nd6.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
    5  * All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the project nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
   32  */
   33 
   34 #include <sys/cdefs.h>
   35 __FBSDID("$FreeBSD: releng/12.0/sys/netinet6/nd6.c 336676 2018-07-24 16:35:52Z andrew $");
   36 
   37 #include "opt_inet.h"
   38 #include "opt_inet6.h"
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/callout.h>
   43 #include <sys/lock.h>
   44 #include <sys/malloc.h>
   45 #include <sys/mbuf.h>
   46 #include <sys/mutex.h>
   47 #include <sys/socket.h>
   48 #include <sys/sockio.h>
   49 #include <sys/time.h>
   50 #include <sys/kernel.h>
   51 #include <sys/protosw.h>
   52 #include <sys/errno.h>
   53 #include <sys/syslog.h>
   54 #include <sys/rwlock.h>
   55 #include <sys/queue.h>
   56 #include <sys/sdt.h>
   57 #include <sys/sysctl.h>
   58 
   59 #include <net/if.h>
   60 #include <net/if_var.h>
   61 #include <net/if_dl.h>
   62 #include <net/if_types.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 VNET_DEFINE_STATIC(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 VNET_DEFINE_STATIC(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 VNET_DEFINE(uint64_t, nd6_list_genid);
  122 VNET_DEFINE(struct mtx, nd6_onlink_mtx);
  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(const 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 void nd6_free(struct llentry **, int);
  135 static void nd6_free_redirect(const struct llentry *);
  136 static void nd6_llinfo_timer(void *);
  137 static void nd6_llinfo_settimer_locked(struct llentry *, long);
  138 static void clear_llinfo_pqueue(struct llentry *);
  139 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  140 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *,
  141     const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **);
  142 static int nd6_need_cache(struct ifnet *);
  143  
  144 
  145 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch);
  146 #define V_nd6_slowtimo_ch               VNET(nd6_slowtimo_ch)
  147 
  148 VNET_DEFINE(struct callout, nd6_timer_ch);
  149 #define V_nd6_timer_ch                  VNET(nd6_timer_ch)
  150 
  151 static void
  152 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
  153 {
  154         struct rt_addrinfo rtinfo;
  155         struct sockaddr_in6 dst;
  156         struct sockaddr_dl gw;
  157         struct ifnet *ifp;
  158         int type;
  159         int fibnum;
  160 
  161         LLE_WLOCK_ASSERT(lle);
  162 
  163         if (lltable_get_af(lle->lle_tbl) != AF_INET6)
  164                 return;
  165 
  166         switch (evt) {
  167         case LLENTRY_RESOLVED:
  168                 type = RTM_ADD;
  169                 KASSERT(lle->la_flags & LLE_VALID,
  170                     ("%s: %p resolved but not valid?", __func__, lle));
  171                 break;
  172         case LLENTRY_EXPIRED:
  173                 type = RTM_DELETE;
  174                 break;
  175         default:
  176                 return;
  177         }
  178 
  179         ifp = lltable_get_ifp(lle->lle_tbl);
  180 
  181         bzero(&dst, sizeof(dst));
  182         bzero(&gw, sizeof(gw));
  183         bzero(&rtinfo, sizeof(rtinfo));
  184         lltable_fill_sa_entry(lle, (struct sockaddr *)&dst);
  185         dst.sin6_scope_id = in6_getscopezone(ifp,
  186             in6_addrscope(&dst.sin6_addr));
  187         gw.sdl_len = sizeof(struct sockaddr_dl);
  188         gw.sdl_family = AF_LINK;
  189         gw.sdl_alen = ifp->if_addrlen;
  190         gw.sdl_index = ifp->if_index;
  191         gw.sdl_type = ifp->if_type;
  192         if (evt == LLENTRY_RESOLVED)
  193                 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
  194         rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
  195         rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
  196         rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
  197         fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib;
  198         rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | (
  199             type == RTM_ADD ? RTF_UP: 0), 0, fibnum);
  200 }
  201 
  202 /*
  203  * A handler for interface link layer address change event.
  204  */
  205 static void
  206 nd6_iflladdr(void *arg __unused, struct ifnet *ifp)
  207 {
  208 
  209         lltable_update_ifaddr(LLTABLE6(ifp));
  210 }
  211 
  212 void
  213 nd6_init(void)
  214 {
  215 
  216         mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF);
  217         rw_init(&V_nd6_lock, "nd6 list");
  218 
  219         LIST_INIT(&V_nd_prefix);
  220         TAILQ_INIT(&V_nd_defrouter);
  221 
  222         /* Start timers. */
  223         callout_init(&V_nd6_slowtimo_ch, 0);
  224         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
  225             nd6_slowtimo, curvnet);
  226 
  227         callout_init(&V_nd6_timer_ch, 0);
  228         callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
  229 
  230         nd6_dad_init();
  231         if (IS_DEFAULT_VNET(curvnet)) {
  232                 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
  233                     NULL, EVENTHANDLER_PRI_ANY);
  234                 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event,
  235                     nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
  236         }
  237 }
  238 
  239 #ifdef VIMAGE
  240 void
  241 nd6_destroy()
  242 {
  243 
  244         callout_drain(&V_nd6_slowtimo_ch);
  245         callout_drain(&V_nd6_timer_ch);
  246         if (IS_DEFAULT_VNET(curvnet)) {
  247                 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
  248                 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
  249         }
  250         rw_destroy(&V_nd6_lock);
  251         mtx_destroy(&V_nd6_onlink_mtx);
  252 }
  253 #endif
  254 
  255 struct nd_ifinfo *
  256 nd6_ifattach(struct ifnet *ifp)
  257 {
  258         struct nd_ifinfo *nd;
  259 
  260         nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
  261         nd->initialized = 1;
  262 
  263         nd->chlim = IPV6_DEFHLIM;
  264         nd->basereachable = REACHABLE_TIME;
  265         nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
  266         nd->retrans = RETRANS_TIMER;
  267 
  268         nd->flags = ND6_IFF_PERFORMNUD;
  269 
  270         /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
  271          * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
  272          * default regardless of the V_ip6_auto_linklocal configuration to
  273          * give a reasonable default behavior.
  274          */
  275         if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
  276             (ifp->if_flags & IFF_LOOPBACK))
  277                 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
  278         /*
  279          * A loopback interface does not need to accept RTADV.
  280          * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
  281          * default regardless of the V_ip6_accept_rtadv configuration to
  282          * prevent the interface from accepting RA messages arrived
  283          * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
  284          */
  285         if (V_ip6_accept_rtadv &&
  286             !(ifp->if_flags & IFF_LOOPBACK) &&
  287             (ifp->if_type != IFT_BRIDGE))
  288                         nd->flags |= ND6_IFF_ACCEPT_RTADV;
  289         if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
  290                 nd->flags |= ND6_IFF_NO_RADR;
  291 
  292         /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
  293         nd6_setmtu0(ifp, nd);
  294 
  295         return nd;
  296 }
  297 
  298 void
  299 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd)
  300 {
  301         struct ifaddr *ifa, *next;
  302 
  303         IF_ADDR_RLOCK(ifp);
  304         CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  305                 if (ifa->ifa_addr->sa_family != AF_INET6)
  306                         continue;
  307 
  308                 /* stop DAD processing */
  309                 nd6_dad_stop(ifa);
  310         }
  311         IF_ADDR_RUNLOCK(ifp);
  312 
  313         free(nd, M_IP6NDP);
  314 }
  315 
  316 /*
  317  * Reset ND level link MTU. This function is called when the physical MTU
  318  * changes, which means we might have to adjust the ND level MTU.
  319  */
  320 void
  321 nd6_setmtu(struct ifnet *ifp)
  322 {
  323         if (ifp->if_afdata[AF_INET6] == NULL)
  324                 return;
  325 
  326         nd6_setmtu0(ifp, ND_IFINFO(ifp));
  327 }
  328 
  329 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
  330 void
  331 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
  332 {
  333         u_int32_t omaxmtu;
  334 
  335         omaxmtu = ndi->maxmtu;
  336         ndi->maxmtu = ifp->if_mtu;
  337 
  338         /*
  339          * Decreasing the interface MTU under IPV6 minimum MTU may cause
  340          * undesirable situation.  We thus notify the operator of the change
  341          * explicitly.  The check for omaxmtu is necessary to restrict the
  342          * log to the case of changing the MTU, not initializing it.
  343          */
  344         if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
  345                 log(LOG_NOTICE, "nd6_setmtu0: "
  346                     "new link MTU on %s (%lu) is too small for IPv6\n",
  347                     if_name(ifp), (unsigned long)ndi->maxmtu);
  348         }
  349 
  350         if (ndi->maxmtu > V_in6_maxmtu)
  351                 in6_setmaxmtu(); /* check all interfaces just in case */
  352 
  353 }
  354 
  355 void
  356 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
  357 {
  358 
  359         bzero(ndopts, sizeof(*ndopts));
  360         ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
  361         ndopts->nd_opts_last
  362                 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
  363 
  364         if (icmp6len == 0) {
  365                 ndopts->nd_opts_done = 1;
  366                 ndopts->nd_opts_search = NULL;
  367         }
  368 }
  369 
  370 /*
  371  * Take one ND option.
  372  */
  373 struct nd_opt_hdr *
  374 nd6_option(union nd_opts *ndopts)
  375 {
  376         struct nd_opt_hdr *nd_opt;
  377         int olen;
  378 
  379         KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
  380         KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
  381             __func__));
  382         if (ndopts->nd_opts_search == NULL)
  383                 return NULL;
  384         if (ndopts->nd_opts_done)
  385                 return NULL;
  386 
  387         nd_opt = ndopts->nd_opts_search;
  388 
  389         /* make sure nd_opt_len is inside the buffer */
  390         if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
  391                 bzero(ndopts, sizeof(*ndopts));
  392                 return NULL;
  393         }
  394 
  395         olen = nd_opt->nd_opt_len << 3;
  396         if (olen == 0) {
  397                 /*
  398                  * Message validation requires that all included
  399                  * options have a length that is greater than zero.
  400                  */
  401                 bzero(ndopts, sizeof(*ndopts));
  402                 return NULL;
  403         }
  404 
  405         ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
  406         if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
  407                 /* option overruns the end of buffer, invalid */
  408                 bzero(ndopts, sizeof(*ndopts));
  409                 return NULL;
  410         } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
  411                 /* reached the end of options chain */
  412                 ndopts->nd_opts_done = 1;
  413                 ndopts->nd_opts_search = NULL;
  414         }
  415         return nd_opt;
  416 }
  417 
  418 /*
  419  * Parse multiple ND options.
  420  * This function is much easier to use, for ND routines that do not need
  421  * multiple options of the same type.
  422  */
  423 int
  424 nd6_options(union nd_opts *ndopts)
  425 {
  426         struct nd_opt_hdr *nd_opt;
  427         int i = 0;
  428 
  429         KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
  430         KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
  431             __func__));
  432         if (ndopts->nd_opts_search == NULL)
  433                 return 0;
  434 
  435         while (1) {
  436                 nd_opt = nd6_option(ndopts);
  437                 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
  438                         /*
  439                          * Message validation requires that all included
  440                          * options have a length that is greater than zero.
  441                          */
  442                         ICMP6STAT_INC(icp6s_nd_badopt);
  443                         bzero(ndopts, sizeof(*ndopts));
  444                         return -1;
  445                 }
  446 
  447                 if (nd_opt == NULL)
  448                         goto skip1;
  449 
  450                 switch (nd_opt->nd_opt_type) {
  451                 case ND_OPT_SOURCE_LINKADDR:
  452                 case ND_OPT_TARGET_LINKADDR:
  453                 case ND_OPT_MTU:
  454                 case ND_OPT_REDIRECTED_HEADER:
  455                 case ND_OPT_NONCE:
  456                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
  457                                 nd6log((LOG_INFO,
  458                                     "duplicated ND6 option found (type=%d)\n",
  459                                     nd_opt->nd_opt_type));
  460                                 /* XXX bark? */
  461                         } else {
  462                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  463                                         = nd_opt;
  464                         }
  465                         break;
  466                 case ND_OPT_PREFIX_INFORMATION:
  467                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
  468                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  469                                         = nd_opt;
  470                         }
  471                         ndopts->nd_opts_pi_end =
  472                                 (struct nd_opt_prefix_info *)nd_opt;
  473                         break;
  474                 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
  475                 case ND_OPT_RDNSS:      /* RFC 6106 */
  476                 case ND_OPT_DNSSL:      /* RFC 6106 */
  477                         /*
  478                          * Silently ignore options we know and do not care about
  479                          * in the kernel.
  480                          */
  481                         break;
  482                 default:
  483                         /*
  484                          * Unknown options must be silently ignored,
  485                          * to accommodate future extension to the protocol.
  486                          */
  487                         nd6log((LOG_DEBUG,
  488                             "nd6_options: unsupported option %d - "
  489                             "option ignored\n", nd_opt->nd_opt_type));
  490                 }
  491 
  492 skip1:
  493                 i++;
  494                 if (i > V_nd6_maxndopt) {
  495                         ICMP6STAT_INC(icp6s_nd_toomanyopt);
  496                         nd6log((LOG_INFO, "too many loop in nd opt\n"));
  497                         break;
  498                 }
  499 
  500                 if (ndopts->nd_opts_done)
  501                         break;
  502         }
  503 
  504         return 0;
  505 }
  506 
  507 /*
  508  * ND6 timer routine to handle ND6 entries
  509  */
  510 static void
  511 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
  512 {
  513         int canceled;
  514 
  515         LLE_WLOCK_ASSERT(ln);
  516 
  517         if (tick < 0) {
  518                 ln->la_expire = 0;
  519                 ln->ln_ntick = 0;
  520                 canceled = callout_stop(&ln->lle_timer);
  521         } else {
  522                 ln->la_expire = time_uptime + tick / hz;
  523                 LLE_ADDREF(ln);
  524                 if (tick > INT_MAX) {
  525                         ln->ln_ntick = tick - INT_MAX;
  526                         canceled = callout_reset(&ln->lle_timer, INT_MAX,
  527                             nd6_llinfo_timer, ln);
  528                 } else {
  529                         ln->ln_ntick = 0;
  530                         canceled = callout_reset(&ln->lle_timer, tick,
  531                             nd6_llinfo_timer, ln);
  532                 }
  533         }
  534         if (canceled > 0)
  535                 LLE_REMREF(ln);
  536 }
  537 
  538 /*
  539  * Gets source address of the first packet in hold queue
  540  * and stores it in @src.
  541  * Returns pointer to @src (if hold queue is not empty) or NULL.
  542  *
  543  * Set noinline to be dtrace-friendly
  544  */
  545 static __noinline struct in6_addr *
  546 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
  547 {
  548         struct ip6_hdr hdr;
  549         struct mbuf *m;
  550 
  551         if (ln->la_hold == NULL)
  552                 return (NULL);
  553 
  554         /*
  555          * assume every packet in la_hold has the same IP header
  556          */
  557         m = ln->la_hold;
  558         if (sizeof(hdr) > m->m_len)
  559                 return (NULL);
  560 
  561         m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
  562         *src = hdr.ip6_src;
  563 
  564         return (src);
  565 }
  566 
  567 /*
  568  * Checks if we need to switch from STALE state.
  569  *
  570  * RFC 4861 requires switching from STALE to DELAY state
  571  * on first packet matching entry, waiting V_nd6_delay and
  572  * transition to PROBE state (if upper layer confirmation was
  573  * not received).
  574  *
  575  * This code performs a bit differently:
  576  * On packet hit we don't change state (but desired state
  577  * can be guessed by control plane). However, after V_nd6_delay
  578  * seconds code will transition to PROBE state (so DELAY state
  579  * is kinda skipped in most situations).
  580  *
  581  * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so
  582  * we perform the following upon entering STALE state:
  583  *
  584  * 1) Arm timer to run each V_nd6_delay seconds to make sure that
  585  * if packet was transmitted at the start of given interval, we
  586  * would be able to switch to PROBE state in V_nd6_delay seconds
  587  * as user expects.
  588  *
  589  * 2) Reschedule timer until original V_nd6_gctimer expires keeping
  590  * lle in STALE state (remaining timer value stored in lle_remtime).
  591  *
  592  * 3) Reschedule timer if packet was transmitted less that V_nd6_delay
  593  * seconds ago.
  594  *
  595  * Returns non-zero value if the entry is still STALE (storing
  596  * the next timer interval in @pdelay).
  597  *
  598  * Returns zero value if original timer expired or we need to switch to
  599  * PROBE (store that in @do_switch variable).
  600  */
  601 static int
  602 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch)
  603 {
  604         int nd_delay, nd_gctimer, r_skip_req;
  605         time_t lle_hittime;
  606         long delay;
  607 
  608         *do_switch = 0;
  609         nd_gctimer = V_nd6_gctimer;
  610         nd_delay = V_nd6_delay;
  611 
  612         LLE_REQ_LOCK(lle);
  613         r_skip_req = lle->r_skip_req;
  614         lle_hittime = lle->lle_hittime;
  615         LLE_REQ_UNLOCK(lle);
  616 
  617         if (r_skip_req > 0) {
  618 
  619                 /*
  620                  * Nonzero r_skip_req value was set upon entering
  621                  * STALE state. Since value was not changed, no
  622                  * packets were passed using this lle. Ask for
  623                  * timer reschedule and keep STALE state.
  624                  */
  625                 delay = (long)(MIN(nd_gctimer, nd_delay));
  626                 delay *= hz;
  627                 if (lle->lle_remtime > delay)
  628                         lle->lle_remtime -= delay;
  629                 else {
  630                         delay = lle->lle_remtime;
  631                         lle->lle_remtime = 0;
  632                 }
  633 
  634                 if (delay == 0) {
  635 
  636                         /*
  637                          * The original ng6_gctime timeout ended,
  638                          * no more rescheduling.
  639                          */
  640                         return (0);
  641                 }
  642 
  643                 *pdelay = delay;
  644                 return (1);
  645         }
  646 
  647         /*
  648          * Packet received. Verify timestamp
  649          */
  650         delay = (long)(time_uptime - lle_hittime);
  651         if (delay < nd_delay) {
  652 
  653                 /*
  654                  * V_nd6_delay still not passed since the first
  655                  * hit in STALE state.
  656                  * Reshedule timer and return.
  657                  */
  658                 *pdelay = (long)(nd_delay - delay) * hz;
  659                 return (1);
  660         }
  661 
  662         /* Request switching to probe */
  663         *do_switch = 1;
  664         return (0);
  665 }
  666 
  667 
  668 /*
  669  * Switch @lle state to new state optionally arming timers.
  670  *
  671  * Set noinline to be dtrace-friendly
  672  */
  673 __noinline void
  674 nd6_llinfo_setstate(struct llentry *lle, int newstate)
  675 {
  676         struct ifnet *ifp;
  677         int nd_gctimer, nd_delay;
  678         long delay, remtime;
  679 
  680         delay = 0;
  681         remtime = 0;
  682 
  683         switch (newstate) {
  684         case ND6_LLINFO_INCOMPLETE:
  685                 ifp = lle->lle_tbl->llt_ifp;
  686                 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
  687                 break;
  688         case ND6_LLINFO_REACHABLE:
  689                 if (!ND6_LLINFO_PERMANENT(lle)) {
  690                         ifp = lle->lle_tbl->llt_ifp;
  691                         delay = (long)ND_IFINFO(ifp)->reachable * hz;
  692                 }
  693                 break;
  694         case ND6_LLINFO_STALE:
  695 
  696                 /*
  697                  * Notify fast path that we want to know if any packet
  698                  * is transmitted by setting r_skip_req.
  699                  */
  700                 LLE_REQ_LOCK(lle);
  701                 lle->r_skip_req = 1;
  702                 LLE_REQ_UNLOCK(lle);
  703                 nd_delay = V_nd6_delay;
  704                 nd_gctimer = V_nd6_gctimer;
  705 
  706                 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz;
  707                 remtime = (long)nd_gctimer * hz - delay;
  708                 break;
  709         case ND6_LLINFO_DELAY:
  710                 lle->la_asked = 0;
  711                 delay = (long)V_nd6_delay * hz;
  712                 break;
  713         }
  714 
  715         if (delay > 0)
  716                 nd6_llinfo_settimer_locked(lle, delay);
  717 
  718         lle->lle_remtime = remtime;
  719         lle->ln_state = newstate;
  720 }
  721 
  722 /*
  723  * Timer-dependent part of nd state machine.
  724  *
  725  * Set noinline to be dtrace-friendly
  726  */
  727 static __noinline void
  728 nd6_llinfo_timer(void *arg)
  729 {
  730         struct llentry *ln;
  731         struct in6_addr *dst, *pdst, *psrc, src;
  732         struct ifnet *ifp;
  733         struct nd_ifinfo *ndi;
  734         int do_switch, send_ns;
  735         long delay;
  736 
  737         KASSERT(arg != NULL, ("%s: arg NULL", __func__));
  738         ln = (struct llentry *)arg;
  739         ifp = lltable_get_ifp(ln->lle_tbl);
  740         CURVNET_SET(ifp->if_vnet);
  741 
  742         ND6_RLOCK();
  743         LLE_WLOCK(ln);
  744         if (callout_pending(&ln->lle_timer)) {
  745                 /*
  746                  * Here we are a bit odd here in the treatment of 
  747                  * active/pending. If the pending bit is set, it got
  748                  * rescheduled before I ran. The active
  749                  * bit we ignore, since if it was stopped
  750                  * in ll_tablefree() and was currently running
  751                  * it would have return 0 so the code would
  752                  * not have deleted it since the callout could
  753                  * not be stopped so we want to go through
  754                  * with the delete here now. If the callout
  755                  * was restarted, the pending bit will be back on and
  756                  * we just want to bail since the callout_reset would
  757                  * return 1 and our reference would have been removed
  758                  * by nd6_llinfo_settimer_locked above since canceled
  759                  * would have been 1.
  760                  */
  761                 LLE_WUNLOCK(ln);
  762                 ND6_RUNLOCK();
  763                 CURVNET_RESTORE();
  764                 return;
  765         }
  766         ndi = ND_IFINFO(ifp);
  767         send_ns = 0;
  768         dst = &ln->r_l3addr.addr6;
  769         pdst = dst;
  770 
  771         if (ln->ln_ntick > 0) {
  772                 if (ln->ln_ntick > INT_MAX) {
  773                         ln->ln_ntick -= INT_MAX;
  774                         nd6_llinfo_settimer_locked(ln, INT_MAX);
  775                 } else {
  776                         ln->ln_ntick = 0;
  777                         nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
  778                 }
  779                 goto done;
  780         }
  781 
  782         if (ln->la_flags & LLE_STATIC) {
  783                 goto done;
  784         }
  785 
  786         if (ln->la_flags & LLE_DELETED) {
  787                 nd6_free(&ln, 0);
  788                 goto done;
  789         }
  790 
  791         switch (ln->ln_state) {
  792         case ND6_LLINFO_INCOMPLETE:
  793                 if (ln->la_asked < V_nd6_mmaxtries) {
  794                         ln->la_asked++;
  795                         send_ns = 1;
  796                         /* Send NS to multicast address */
  797                         pdst = NULL;
  798                 } else {
  799                         struct mbuf *m = ln->la_hold;
  800                         if (m) {
  801                                 struct mbuf *m0;
  802 
  803                                 /*
  804                                  * assuming every packet in la_hold has the
  805                                  * same IP header.  Send error after unlock.
  806                                  */
  807                                 m0 = m->m_nextpkt;
  808                                 m->m_nextpkt = NULL;
  809                                 ln->la_hold = m0;
  810                                 clear_llinfo_pqueue(ln);
  811                         }
  812                         nd6_free(&ln, 0);
  813                         if (m != NULL)
  814                                 icmp6_error2(m, ICMP6_DST_UNREACH,
  815                                     ICMP6_DST_UNREACH_ADDR, 0, ifp);
  816                 }
  817                 break;
  818         case ND6_LLINFO_REACHABLE:
  819                 if (!ND6_LLINFO_PERMANENT(ln))
  820                         nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
  821                 break;
  822 
  823         case ND6_LLINFO_STALE:
  824                 if (nd6_is_stale(ln, &delay, &do_switch) != 0) {
  825 
  826                         /*
  827                          * No packet has used this entry and GC timeout
  828                          * has not been passed. Reshedule timer and
  829                          * return.
  830                          */
  831                         nd6_llinfo_settimer_locked(ln, delay);
  832                         break;
  833                 }
  834 
  835                 if (do_switch == 0) {
  836 
  837                         /*
  838                          * GC timer has ended and entry hasn't been used.
  839                          * Run Garbage collector (RFC 4861, 5.3)
  840                          */
  841                         if (!ND6_LLINFO_PERMANENT(ln))
  842                                 nd6_free(&ln, 1);
  843                         break;
  844                 }
  845 
  846                 /* Entry has been used AND delay timer has ended. */
  847 
  848                 /* FALLTHROUGH */
  849 
  850         case ND6_LLINFO_DELAY:
  851                 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
  852                         /* We need NUD */
  853                         ln->la_asked = 1;
  854                         nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
  855                         send_ns = 1;
  856                 } else
  857                         nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
  858                 break;
  859         case ND6_LLINFO_PROBE:
  860                 if (ln->la_asked < V_nd6_umaxtries) {
  861                         ln->la_asked++;
  862                         send_ns = 1;
  863                 } else {
  864                         nd6_free(&ln, 0);
  865                 }
  866                 break;
  867         default:
  868                 panic("%s: paths in a dark night can be confusing: %d",
  869                     __func__, ln->ln_state);
  870         }
  871 done:
  872         if (ln != NULL)
  873                 ND6_RUNLOCK();
  874         if (send_ns != 0) {
  875                 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
  876                 psrc = nd6_llinfo_get_holdsrc(ln, &src);
  877                 LLE_FREE_LOCKED(ln);
  878                 ln = NULL;
  879                 nd6_ns_output(ifp, psrc, pdst, dst, NULL);
  880         }
  881 
  882         if (ln != NULL)
  883                 LLE_FREE_LOCKED(ln);
  884         CURVNET_RESTORE();
  885 }
  886 
  887 
  888 /*
  889  * ND6 timer routine to expire default route list and prefix list
  890  */
  891 void
  892 nd6_timer(void *arg)
  893 {
  894         CURVNET_SET((struct vnet *) arg);
  895         struct nd_drhead drq;
  896         struct nd_prhead prl;
  897         struct nd_defrouter *dr, *ndr;
  898         struct nd_prefix *pr, *npr;
  899         struct in6_ifaddr *ia6, *nia6;
  900         uint64_t genid;
  901 
  902         TAILQ_INIT(&drq);
  903         LIST_INIT(&prl);
  904 
  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         CK_STAILQ_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         ND6_WLOCK();
 1012 restart:
 1013         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
 1014                 /*
 1015                  * Expire prefixes. Since the pltime is only used for
 1016                  * autoconfigured addresses, pltime processing for prefixes is
 1017                  * not necessary.
 1018                  *
 1019                  * Only unlink after all derived addresses have expired. This
 1020                  * may not occur until two hours after the prefix has expired
 1021                  * per RFC 4862. If the prefix expires before its derived
 1022                  * addresses, mark it off-link. This will be done automatically
 1023                  * after unlinking if no address references remain.
 1024                  */
 1025                 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME ||
 1026                     time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime)
 1027                         continue;
 1028 
 1029                 if (pr->ndpr_addrcnt == 0) {
 1030                         nd6_prefix_unlink(pr, &prl);
 1031                         continue;
 1032                 }
 1033                 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
 1034                         genid = V_nd6_list_genid;
 1035                         nd6_prefix_ref(pr);
 1036                         ND6_WUNLOCK();
 1037                         ND6_ONLINK_LOCK();
 1038                         (void)nd6_prefix_offlink(pr);
 1039                         ND6_ONLINK_UNLOCK();
 1040                         ND6_WLOCK();
 1041                         nd6_prefix_rele(pr);
 1042                         if (genid != V_nd6_list_genid)
 1043                                 goto restart;
 1044                 }
 1045         }
 1046         ND6_WUNLOCK();
 1047 
 1048         while ((pr = LIST_FIRST(&prl)) != NULL) {
 1049                 LIST_REMOVE(pr, ndpr_entry);
 1050                 nd6_prefix_del(pr);
 1051         }
 1052 
 1053         callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
 1054             nd6_timer, curvnet);
 1055 
 1056         CURVNET_RESTORE();
 1057 }
 1058 
 1059 /*
 1060  * ia6 - deprecated/invalidated temporary address
 1061  */
 1062 static int
 1063 regen_tmpaddr(struct in6_ifaddr *ia6)
 1064 {
 1065         struct ifaddr *ifa;
 1066         struct ifnet *ifp;
 1067         struct in6_ifaddr *public_ifa6 = NULL;
 1068 
 1069         ifp = ia6->ia_ifa.ifa_ifp;
 1070         IF_ADDR_RLOCK(ifp);
 1071         CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1072                 struct in6_ifaddr *it6;
 1073 
 1074                 if (ifa->ifa_addr->sa_family != AF_INET6)
 1075                         continue;
 1076 
 1077                 it6 = (struct in6_ifaddr *)ifa;
 1078 
 1079                 /* ignore no autoconf addresses. */
 1080                 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1081                         continue;
 1082 
 1083                 /* ignore autoconf addresses with different prefixes. */
 1084                 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
 1085                         continue;
 1086 
 1087                 /*
 1088                  * Now we are looking at an autoconf address with the same
 1089                  * prefix as ours.  If the address is temporary and is still
 1090                  * preferred, do not create another one.  It would be rare, but
 1091                  * could happen, for example, when we resume a laptop PC after
 1092                  * a long period.
 1093                  */
 1094                 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
 1095                     !IFA6_IS_DEPRECATED(it6)) {
 1096                         public_ifa6 = NULL;
 1097                         break;
 1098                 }
 1099 
 1100                 /*
 1101                  * This is a public autoconf address that has the same prefix
 1102                  * as ours.  If it is preferred, keep it.  We can't break the
 1103                  * loop here, because there may be a still-preferred temporary
 1104                  * address with the prefix.
 1105                  */
 1106                 if (!IFA6_IS_DEPRECATED(it6))
 1107                         public_ifa6 = it6;
 1108         }
 1109         if (public_ifa6 != NULL)
 1110                 ifa_ref(&public_ifa6->ia_ifa);
 1111         IF_ADDR_RUNLOCK(ifp);
 1112 
 1113         if (public_ifa6 != NULL) {
 1114                 int e;
 1115 
 1116                 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
 1117                         ifa_free(&public_ifa6->ia_ifa);
 1118                         log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
 1119                             " tmp addr,errno=%d\n", e);
 1120                         return (-1);
 1121                 }
 1122                 ifa_free(&public_ifa6->ia_ifa);
 1123                 return (0);
 1124         }
 1125 
 1126         return (-1);
 1127 }
 1128 
 1129 /*
 1130  * Remove prefix and default router list entries corresponding to ifp. Neighbor
 1131  * cache entries are freed in in6_domifdetach().
 1132  */
 1133 void
 1134 nd6_purge(struct ifnet *ifp)
 1135 {
 1136         struct nd_drhead drq;
 1137         struct nd_prhead prl;
 1138         struct nd_defrouter *dr, *ndr;
 1139         struct nd_prefix *pr, *npr;
 1140 
 1141         TAILQ_INIT(&drq);
 1142         LIST_INIT(&prl);
 1143 
 1144         /*
 1145          * Nuke default router list entries toward ifp.
 1146          * We defer removal of default router list entries that is installed
 1147          * in the routing table, in order to keep additional side effects as
 1148          * small as possible.
 1149          */
 1150         ND6_WLOCK();
 1151         TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
 1152                 if (dr->installed)
 1153                         continue;
 1154                 if (dr->ifp == ifp)
 1155                         defrouter_unlink(dr, &drq);
 1156         }
 1157         TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
 1158                 if (!dr->installed)
 1159                         continue;
 1160                 if (dr->ifp == ifp)
 1161                         defrouter_unlink(dr, &drq);
 1162         }
 1163 
 1164         /*
 1165          * Remove prefixes on ifp. We should have already removed addresses on
 1166          * this interface, so no addresses should be referencing these prefixes.
 1167          */
 1168         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
 1169                 if (pr->ndpr_ifp == ifp)
 1170                         nd6_prefix_unlink(pr, &prl);
 1171         }
 1172         ND6_WUNLOCK();
 1173 
 1174         /* Delete the unlinked router and prefix objects. */
 1175         while ((dr = TAILQ_FIRST(&drq)) != NULL) {
 1176                 TAILQ_REMOVE(&drq, dr, dr_entry);
 1177                 defrouter_del(dr);
 1178         }
 1179         while ((pr = LIST_FIRST(&prl)) != NULL) {
 1180                 LIST_REMOVE(pr, ndpr_entry);
 1181                 nd6_prefix_del(pr);
 1182         }
 1183 
 1184         /* cancel default outgoing interface setting */
 1185         if (V_nd6_defifindex == ifp->if_index)
 1186                 nd6_setdefaultiface(0);
 1187 
 1188         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 1189                 /* Refresh default router list. */
 1190                 defrouter_select_fib(ifp->if_fib);
 1191         }
 1192 }
 1193 
 1194 /* 
 1195  * the caller acquires and releases the lock on the lltbls
 1196  * Returns the llentry locked
 1197  */
 1198 struct llentry *
 1199 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
 1200 {
 1201         struct sockaddr_in6 sin6;
 1202         struct llentry *ln;
 1203         
 1204         bzero(&sin6, sizeof(sin6));
 1205         sin6.sin6_len = sizeof(struct sockaddr_in6);
 1206         sin6.sin6_family = AF_INET6;
 1207         sin6.sin6_addr = *addr6;
 1208 
 1209         IF_AFDATA_LOCK_ASSERT(ifp);
 1210 
 1211         ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6);
 1212 
 1213         return (ln);
 1214 }
 1215 
 1216 struct llentry *
 1217 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
 1218 {
 1219         struct sockaddr_in6 sin6;
 1220         struct llentry *ln;
 1221 
 1222         bzero(&sin6, sizeof(sin6));
 1223         sin6.sin6_len = sizeof(struct sockaddr_in6);
 1224         sin6.sin6_family = AF_INET6;
 1225         sin6.sin6_addr = *addr6;
 1226 
 1227         ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
 1228         if (ln != NULL)
 1229                 ln->ln_state = ND6_LLINFO_NOSTATE;
 1230 
 1231         return (ln);
 1232 }
 1233 
 1234 /*
 1235  * Test whether a given IPv6 address is a neighbor or not, ignoring
 1236  * the actual neighbor cache.  The neighbor cache is ignored in order
 1237  * to not reenter the routing code from within itself.
 1238  */
 1239 static int
 1240 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
 1241 {
 1242         struct nd_prefix *pr;
 1243         struct ifaddr *ifa;
 1244         struct rt_addrinfo info;
 1245         struct sockaddr_in6 rt_key;
 1246         const struct sockaddr *dst6;
 1247         uint64_t genid;
 1248         int error, fibnum;
 1249 
 1250         /*
 1251          * A link-local address is always a neighbor.
 1252          * XXX: a link does not necessarily specify a single interface.
 1253          */
 1254         if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
 1255                 struct sockaddr_in6 sin6_copy;
 1256                 u_int32_t zone;
 1257 
 1258                 /*
 1259                  * We need sin6_copy since sa6_recoverscope() may modify the
 1260                  * content (XXX).
 1261                  */
 1262                 sin6_copy = *addr;
 1263                 if (sa6_recoverscope(&sin6_copy))
 1264                         return (0); /* XXX: should be impossible */
 1265                 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
 1266                         return (0);
 1267                 if (sin6_copy.sin6_scope_id == zone)
 1268                         return (1);
 1269                 else
 1270                         return (0);
 1271         }
 1272 
 1273         bzero(&rt_key, sizeof(rt_key));
 1274         bzero(&info, sizeof(info));
 1275         info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
 1276 
 1277         /*
 1278          * If the address matches one of our addresses,
 1279          * it should be a neighbor.
 1280          * If the address matches one of our on-link prefixes, it should be a
 1281          * neighbor.
 1282          */
 1283         ND6_RLOCK();
 1284 restart:
 1285         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 1286                 if (pr->ndpr_ifp != ifp)
 1287                         continue;
 1288 
 1289                 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
 1290                         dst6 = (const struct sockaddr *)&pr->ndpr_prefix;
 1291 
 1292                         /*
 1293                          * We only need to check all FIBs if add_addr_allfibs
 1294                          * is unset. If set, checking any FIB will suffice.
 1295                          */
 1296                         fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0;
 1297                         for (; fibnum < rt_numfibs; fibnum++) {
 1298                                 genid = V_nd6_list_genid;
 1299                                 ND6_RUNLOCK();
 1300 
 1301                                 /*
 1302                                  * Restore length field before
 1303                                  * retrying lookup
 1304                                  */
 1305                                 rt_key.sin6_len = sizeof(rt_key);
 1306                                 error = rib_lookup_info(fibnum, dst6, 0, 0,
 1307                                                         &info);
 1308 
 1309                                 ND6_RLOCK();
 1310                                 if (genid != V_nd6_list_genid)
 1311                                         goto restart;
 1312                                 if (error == 0)
 1313                                         break;
 1314                         }
 1315                         if (error != 0)
 1316                                 continue;
 1317 
 1318                         /*
 1319                          * This is the case where multiple interfaces
 1320                          * have the same prefix, but only one is installed 
 1321                          * into the routing table and that prefix entry
 1322                          * is not the one being examined here. In the case
 1323                          * where RADIX_MPATH is enabled, multiple route
 1324                          * entries (of the same rt_key value) will be 
 1325                          * installed because the interface addresses all
 1326                          * differ.
 1327                          */
 1328                         if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
 1329                             &rt_key.sin6_addr))
 1330                                 continue;
 1331                 }
 1332 
 1333                 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
 1334                     &addr->sin6_addr, &pr->ndpr_mask)) {
 1335                         ND6_RUNLOCK();
 1336                         return (1);
 1337                 }
 1338         }
 1339         ND6_RUNLOCK();
 1340 
 1341         /*
 1342          * If the address is assigned on the node of the other side of
 1343          * a p2p interface, the address should be a neighbor.
 1344          */
 1345         if (ifp->if_flags & IFF_POINTOPOINT) {
 1346                 IF_ADDR_RLOCK(ifp);
 1347                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1348                         if (ifa->ifa_addr->sa_family != addr->sin6_family)
 1349                                 continue;
 1350                         if (ifa->ifa_dstaddr != NULL &&
 1351                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1352                                 IF_ADDR_RUNLOCK(ifp);
 1353                                 return 1;
 1354                         }
 1355                 }
 1356                 IF_ADDR_RUNLOCK(ifp);
 1357         }
 1358 
 1359         /*
 1360          * If the default router list is empty, all addresses are regarded
 1361          * as on-link, and thus, as a neighbor.
 1362          */
 1363         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
 1364             TAILQ_EMPTY(&V_nd_defrouter) &&
 1365             V_nd6_defifindex == ifp->if_index) {
 1366                 return (1);
 1367         }
 1368 
 1369         return (0);
 1370 }
 1371 
 1372 
 1373 /*
 1374  * Detect if a given IPv6 address identifies a neighbor on a given link.
 1375  * XXX: should take care of the destination of a p2p link?
 1376  */
 1377 int
 1378 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
 1379 {
 1380         struct llentry *lle;
 1381         int rc = 0;
 1382 
 1383         IF_AFDATA_UNLOCK_ASSERT(ifp);
 1384         if (nd6_is_new_addr_neighbor(addr, ifp))
 1385                 return (1);
 1386 
 1387         /*
 1388          * Even if the address matches none of our addresses, it might be
 1389          * in the neighbor cache.
 1390          */
 1391         IF_AFDATA_RLOCK(ifp);
 1392         if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
 1393                 LLE_RUNLOCK(lle);
 1394                 rc = 1;
 1395         }
 1396         IF_AFDATA_RUNLOCK(ifp);
 1397         return (rc);
 1398 }
 1399 
 1400 /*
 1401  * Free an nd6 llinfo entry.
 1402  * Since the function would cause significant changes in the kernel, DO NOT
 1403  * make it global, unless you have a strong reason for the change, and are sure
 1404  * that the change is safe.
 1405  *
 1406  * Set noinline to be dtrace-friendly
 1407  */
 1408 static __noinline void
 1409 nd6_free(struct llentry **lnp, int gc)
 1410 {
 1411         struct ifnet *ifp;
 1412         struct llentry *ln;
 1413         struct nd_defrouter *dr;
 1414 
 1415         ln = *lnp;
 1416         *lnp = NULL;
 1417 
 1418         LLE_WLOCK_ASSERT(ln);
 1419         ND6_RLOCK_ASSERT();
 1420 
 1421         ifp = lltable_get_ifp(ln->lle_tbl);
 1422         if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0)
 1423                 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp);
 1424         else
 1425                 dr = NULL;
 1426         ND6_RUNLOCK();
 1427 
 1428         if ((ln->la_flags & LLE_DELETED) == 0)
 1429                 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
 1430 
 1431         /*
 1432          * we used to have pfctlinput(PRC_HOSTDEAD) here.
 1433          * even though it is not harmful, it was not really necessary.
 1434          */
 1435 
 1436         /* cancel timer */
 1437         nd6_llinfo_settimer_locked(ln, -1);
 1438 
 1439         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 1440                 if (dr != NULL && dr->expire &&
 1441                     ln->ln_state == ND6_LLINFO_STALE && gc) {
 1442                         /*
 1443                          * If the reason for the deletion is just garbage
 1444                          * collection, and the neighbor is an active default
 1445                          * router, do not delete it.  Instead, reset the GC
 1446                          * timer using the router's lifetime.
 1447                          * Simply deleting the entry would affect default
 1448                          * router selection, which is not necessarily a good
 1449                          * thing, especially when we're using router preference
 1450                          * values.
 1451                          * XXX: the check for ln_state would be redundant,
 1452                          *      but we intentionally keep it just in case.
 1453                          */
 1454                         if (dr->expire > time_uptime)
 1455                                 nd6_llinfo_settimer_locked(ln,
 1456                                     (dr->expire - time_uptime) * hz);
 1457                         else
 1458                                 nd6_llinfo_settimer_locked(ln,
 1459                                     (long)V_nd6_gctimer * hz);
 1460 
 1461                         LLE_REMREF(ln);
 1462                         LLE_WUNLOCK(ln);
 1463                         defrouter_rele(dr);
 1464                         return;
 1465                 }
 1466 
 1467                 if (dr) {
 1468                         /*
 1469                          * Unreachablity of a router might affect the default
 1470                          * router selection and on-link detection of advertised
 1471                          * prefixes.
 1472                          */
 1473 
 1474                         /*
 1475                          * Temporarily fake the state to choose a new default
 1476                          * router and to perform on-link determination of
 1477                          * prefixes correctly.
 1478                          * Below the state will be set correctly,
 1479                          * or the entry itself will be deleted.
 1480                          */
 1481                         ln->ln_state = ND6_LLINFO_INCOMPLETE;
 1482                 }
 1483 
 1484                 if (ln->ln_router || dr) {
 1485 
 1486                         /*
 1487                          * We need to unlock to avoid a LOR with rt6_flush() with the
 1488                          * rnh and for the calls to pfxlist_onlink_check() and
 1489                          * defrouter_select_fib() in the block further down for calls
 1490                          * into nd6_lookup().  We still hold a ref.
 1491                          */
 1492                         LLE_WUNLOCK(ln);
 1493 
 1494                         /*
 1495                          * rt6_flush must be called whether or not the neighbor
 1496                          * is in the Default Router List.
 1497                          * See a corresponding comment in nd6_na_input().
 1498                          */
 1499                         rt6_flush(&ln->r_l3addr.addr6, ifp);
 1500                 }
 1501 
 1502                 if (dr) {
 1503                         /*
 1504                          * Since defrouter_select_fib() does not affect the
 1505                          * on-link determination and MIP6 needs the check
 1506                          * before the default router selection, we perform
 1507                          * the check now.
 1508                          */
 1509                         pfxlist_onlink_check();
 1510 
 1511                         /*
 1512                          * Refresh default router list.
 1513                          */
 1514                         defrouter_select_fib(dr->ifp->if_fib);
 1515                 }
 1516 
 1517                 /*
 1518                  * If this entry was added by an on-link redirect, remove the
 1519                  * corresponding host route.
 1520                  */
 1521                 if (ln->la_flags & LLE_REDIRECT)
 1522                         nd6_free_redirect(ln);
 1523 
 1524                 if (ln->ln_router || dr)
 1525                         LLE_WLOCK(ln);
 1526         }
 1527 
 1528         /*
 1529          * Save to unlock. We still hold an extra reference and will not
 1530          * free(9) in llentry_free() if someone else holds one as well.
 1531          */
 1532         LLE_WUNLOCK(ln);
 1533         IF_AFDATA_LOCK(ifp);
 1534         LLE_WLOCK(ln);
 1535         /* Guard against race with other llentry_free(). */
 1536         if (ln->la_flags & LLE_LINKED) {
 1537                 /* Remove callout reference */
 1538                 LLE_REMREF(ln);
 1539                 lltable_unlink_entry(ln->lle_tbl, ln);
 1540         }
 1541         IF_AFDATA_UNLOCK(ifp);
 1542 
 1543         llentry_free(ln);
 1544         if (dr != NULL)
 1545                 defrouter_rele(dr);
 1546 }
 1547 
 1548 static int
 1549 nd6_isdynrte(const struct rtentry *rt, void *xap)
 1550 {
 1551 
 1552         if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC))
 1553                 return (1);
 1554 
 1555         return (0);
 1556 }
 1557 /*
 1558  * Remove the rtentry for the given llentry,
 1559  * both of which were installed by a redirect.
 1560  */
 1561 static void
 1562 nd6_free_redirect(const struct llentry *ln)
 1563 {
 1564         int fibnum;
 1565         struct sockaddr_in6 sin6;
 1566         struct rt_addrinfo info;
 1567 
 1568         lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
 1569         memset(&info, 0, sizeof(info));
 1570         info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6;
 1571         info.rti_filter = nd6_isdynrte;
 1572 
 1573         for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
 1574                 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum);
 1575 }
 1576 
 1577 /*
 1578  * Rejuvenate this function for routing operations related
 1579  * processing.
 1580  */
 1581 void
 1582 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
 1583 {
 1584         struct sockaddr_in6 *gateway;
 1585         struct nd_defrouter *dr;
 1586         struct ifnet *ifp;
 1587 
 1588         gateway = (struct sockaddr_in6 *)rt->rt_gateway;
 1589         ifp = rt->rt_ifp;
 1590 
 1591         switch (req) {
 1592         case RTM_ADD:
 1593                 break;
 1594 
 1595         case RTM_DELETE:
 1596                 if (!ifp)
 1597                         return;
 1598                 /*
 1599                  * Only indirect routes are interesting.
 1600                  */
 1601                 if ((rt->rt_flags & RTF_GATEWAY) == 0)
 1602                         return;
 1603                 /*
 1604                  * check for default route
 1605                  */
 1606                 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
 1607                     &SIN6(rt_key(rt))->sin6_addr)) {
 1608                         dr = defrouter_lookup(&gateway->sin6_addr, ifp);
 1609                         if (dr != NULL) {
 1610                                 dr->installed = 0;
 1611                                 defrouter_rele(dr);
 1612                         }
 1613                 }
 1614                 break;
 1615         }
 1616 }
 1617 
 1618 
 1619 int
 1620 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
 1621 {
 1622         struct in6_ndireq *ndi = (struct in6_ndireq *)data;
 1623         struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
 1624         struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
 1625         int error = 0;
 1626 
 1627         if (ifp->if_afdata[AF_INET6] == NULL)
 1628                 return (EPFNOSUPPORT);
 1629         switch (cmd) {
 1630         case OSIOCGIFINFO_IN6:
 1631 #define ND      ndi->ndi
 1632                 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
 1633                 bzero(&ND, sizeof(ND));
 1634                 ND.linkmtu = IN6_LINKMTU(ifp);
 1635                 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
 1636                 ND.basereachable = ND_IFINFO(ifp)->basereachable;
 1637                 ND.reachable = ND_IFINFO(ifp)->reachable;
 1638                 ND.retrans = ND_IFINFO(ifp)->retrans;
 1639                 ND.flags = ND_IFINFO(ifp)->flags;
 1640                 ND.recalctm = ND_IFINFO(ifp)->recalctm;
 1641                 ND.chlim = ND_IFINFO(ifp)->chlim;
 1642                 break;
 1643         case SIOCGIFINFO_IN6:
 1644                 ND = *ND_IFINFO(ifp);
 1645                 break;
 1646         case SIOCSIFINFO_IN6:
 1647                 /*
 1648                  * used to change host variables from userland.
 1649                  * intended for a use on router to reflect RA configurations.
 1650                  */
 1651                 /* 0 means 'unspecified' */
 1652                 if (ND.linkmtu != 0) {
 1653                         if (ND.linkmtu < IPV6_MMTU ||
 1654                             ND.linkmtu > IN6_LINKMTU(ifp)) {
 1655                                 error = EINVAL;
 1656                                 break;
 1657                         }
 1658                         ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
 1659                 }
 1660 
 1661                 if (ND.basereachable != 0) {
 1662                         int obasereachable = ND_IFINFO(ifp)->basereachable;
 1663 
 1664                         ND_IFINFO(ifp)->basereachable = ND.basereachable;
 1665                         if (ND.basereachable != obasereachable)
 1666                                 ND_IFINFO(ifp)->reachable =
 1667                                     ND_COMPUTE_RTIME(ND.basereachable);
 1668                 }
 1669                 if (ND.retrans != 0)
 1670                         ND_IFINFO(ifp)->retrans = ND.retrans;
 1671                 if (ND.chlim != 0)
 1672                         ND_IFINFO(ifp)->chlim = ND.chlim;
 1673                 /* FALLTHROUGH */
 1674         case SIOCSIFINFO_FLAGS:
 1675         {
 1676                 struct ifaddr *ifa;
 1677                 struct in6_ifaddr *ia;
 1678 
 1679                 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
 1680                     !(ND.flags & ND6_IFF_IFDISABLED)) {
 1681                         /* ifdisabled 1->0 transision */
 1682 
 1683                         /*
 1684                          * If the interface is marked as ND6_IFF_IFDISABLED and
 1685                          * has an link-local address with IN6_IFF_DUPLICATED,
 1686                          * do not clear ND6_IFF_IFDISABLED.
 1687                          * See RFC 4862, Section 5.4.5.
 1688                          */
 1689                         IF_ADDR_RLOCK(ifp);
 1690                         CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1691                                 if (ifa->ifa_addr->sa_family != AF_INET6)
 1692                                         continue;
 1693                                 ia = (struct in6_ifaddr *)ifa;
 1694                                 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
 1695                                     IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
 1696                                         break;
 1697                         }
 1698                         IF_ADDR_RUNLOCK(ifp);
 1699 
 1700                         if (ifa != NULL) {
 1701                                 /* LLA is duplicated. */
 1702                                 ND.flags |= ND6_IFF_IFDISABLED;
 1703                                 log(LOG_ERR, "Cannot enable an interface"
 1704                                     " with a link-local address marked"
 1705                                     " duplicate.\n");
 1706                         } else {
 1707                                 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
 1708                                 if (ifp->if_flags & IFF_UP)
 1709                                         in6_if_up(ifp);
 1710                         }
 1711                 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
 1712                             (ND.flags & ND6_IFF_IFDISABLED)) {
 1713                         /* ifdisabled 0->1 transision */
 1714                         /* Mark all IPv6 address as tentative. */
 1715 
 1716                         ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
 1717                         if (V_ip6_dad_count > 0 &&
 1718                             (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
 1719                                 IF_ADDR_RLOCK(ifp);
 1720                                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
 1721                                     ifa_link) {
 1722                                         if (ifa->ifa_addr->sa_family !=
 1723                                             AF_INET6)
 1724                                                 continue;
 1725                                         ia = (struct in6_ifaddr *)ifa;
 1726                                         ia->ia6_flags |= IN6_IFF_TENTATIVE;
 1727                                 }
 1728                                 IF_ADDR_RUNLOCK(ifp);
 1729                         }
 1730                 }
 1731 
 1732                 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
 1733                         if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
 1734                                 /* auto_linklocal 0->1 transision */
 1735 
 1736                                 /* If no link-local address on ifp, configure */
 1737                                 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
 1738                                 in6_ifattach(ifp, NULL);
 1739                         } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
 1740                             ifp->if_flags & IFF_UP) {
 1741                                 /*
 1742                                  * When the IF already has
 1743                                  * ND6_IFF_AUTO_LINKLOCAL, no link-local
 1744                                  * address is assigned, and IFF_UP, try to
 1745                                  * assign one.
 1746                                  */
 1747                                 IF_ADDR_RLOCK(ifp);
 1748                                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
 1749                                     ifa_link) {
 1750                                         if (ifa->ifa_addr->sa_family !=
 1751                                             AF_INET6)
 1752                                                 continue;
 1753                                         ia = (struct in6_ifaddr *)ifa;
 1754                                         if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
 1755                                                 break;
 1756                                 }
 1757                                 IF_ADDR_RUNLOCK(ifp);
 1758                                 if (ifa != NULL)
 1759                                         /* No LLA is configured. */
 1760                                         in6_ifattach(ifp, NULL);
 1761                         }
 1762                 }
 1763         }
 1764                 ND_IFINFO(ifp)->flags = ND.flags;
 1765                 break;
 1766 #undef ND
 1767         case SIOCSNDFLUSH_IN6:  /* XXX: the ioctl name is confusing... */
 1768                 /* sync kernel routing table with the default router list */
 1769                 defrouter_reset();
 1770                 defrouter_select();
 1771                 break;
 1772         case SIOCSPFXFLUSH_IN6:
 1773         {
 1774                 /* flush all the prefix advertised by routers */
 1775                 struct in6_ifaddr *ia, *ia_next;
 1776                 struct nd_prefix *pr, *next;
 1777                 struct nd_prhead prl;
 1778 
 1779                 LIST_INIT(&prl);
 1780 
 1781                 ND6_WLOCK();
 1782                 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
 1783                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1784                                 continue; /* XXX */
 1785                         nd6_prefix_unlink(pr, &prl);
 1786                 }
 1787                 ND6_WUNLOCK();
 1788 
 1789                 while ((pr = LIST_FIRST(&prl)) != NULL) {
 1790                         LIST_REMOVE(pr, ndpr_entry);
 1791                         /* XXXRW: in6_ifaddrhead locking. */
 1792                         CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
 1793                             ia_next) {
 1794                                 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1795                                         continue;
 1796 
 1797                                 if (ia->ia6_ndpr == pr)
 1798                                         in6_purgeaddr(&ia->ia_ifa);
 1799                         }
 1800                         nd6_prefix_del(pr);
 1801                 }
 1802                 break;
 1803         }
 1804         case SIOCSRTRFLUSH_IN6:
 1805         {
 1806                 /* flush all the default routers */
 1807                 struct nd_drhead drq;
 1808                 struct nd_defrouter *dr;
 1809 
 1810                 TAILQ_INIT(&drq);
 1811 
 1812                 defrouter_reset();
 1813 
 1814                 ND6_WLOCK();
 1815                 while ((dr = TAILQ_FIRST(&V_nd_defrouter)) != NULL)
 1816                         defrouter_unlink(dr, &drq);
 1817                 ND6_WUNLOCK();
 1818                 while ((dr = TAILQ_FIRST(&drq)) != NULL) {
 1819                         TAILQ_REMOVE(&drq, dr, dr_entry);
 1820                         defrouter_del(dr);
 1821                 }
 1822 
 1823                 defrouter_select();
 1824                 break;
 1825         }
 1826         case SIOCGNBRINFO_IN6:
 1827         {
 1828                 struct llentry *ln;
 1829                 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
 1830 
 1831                 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
 1832                         return (error);
 1833 
 1834                 IF_AFDATA_RLOCK(ifp);
 1835                 ln = nd6_lookup(&nb_addr, 0, ifp);
 1836                 IF_AFDATA_RUNLOCK(ifp);
 1837 
 1838                 if (ln == NULL) {
 1839                         error = EINVAL;
 1840                         break;
 1841                 }
 1842                 nbi->state = ln->ln_state;
 1843                 nbi->asked = ln->la_asked;
 1844                 nbi->isrouter = ln->ln_router;
 1845                 if (ln->la_expire == 0)
 1846                         nbi->expire = 0;
 1847                 else
 1848                         nbi->expire = ln->la_expire + ln->lle_remtime / hz +
 1849                             (time_second - time_uptime);
 1850                 LLE_RUNLOCK(ln);
 1851                 break;
 1852         }
 1853         case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1854                 ndif->ifindex = V_nd6_defifindex;
 1855                 break;
 1856         case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1857                 return (nd6_setdefaultiface(ndif->ifindex));
 1858         }
 1859         return (error);
 1860 }
 1861 
 1862 /*
 1863  * Calculates new isRouter value based on provided parameters and
 1864  * returns it.
 1865  */
 1866 static int
 1867 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
 1868     int ln_router)
 1869 {
 1870 
 1871         /*
 1872          * ICMP6 type dependent behavior.
 1873          *
 1874          * NS: clear IsRouter if new entry
 1875          * RS: clear IsRouter
 1876          * RA: set IsRouter if there's lladdr
 1877          * redir: clear IsRouter if new entry
 1878          *
 1879          * RA case, (1):
 1880          * The spec says that we must set IsRouter in the following cases:
 1881          * - If lladdr exist, set IsRouter.  This means (1-5).
 1882          * - If it is old entry (!newentry), set IsRouter.  This means (7).
 1883          * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
 1884          * A quetion arises for (1) case.  (1) case has no lladdr in the
 1885          * neighbor cache, this is similar to (6).
 1886          * This case is rare but we figured that we MUST NOT set IsRouter.
 1887          *
 1888          *   is_new  old_addr new_addr      NS  RS  RA  redir
 1889          *                                                      D R
 1890          *      0       n       n       (1)     c   ?     s
 1891          *      0       y       n       (2)     c   s     s
 1892          *      0       n       y       (3)     c   s     s
 1893          *      0       y       y       (4)     c   s     s
 1894          *      0       y       y       (5)     c   s     s
 1895          *      1       --      n       (6) c   c       c s
 1896          *      1       --      y       (7) c   c   s   c s
 1897          *
 1898          *                                      (c=clear s=set)
 1899          */
 1900         switch (type & 0xff) {
 1901         case ND_NEIGHBOR_SOLICIT:
 1902                 /*
 1903                  * New entry must have is_router flag cleared.
 1904                  */
 1905                 if (is_new)                                     /* (6-7) */
 1906                         ln_router = 0;
 1907                 break;
 1908         case ND_REDIRECT:
 1909                 /*
 1910                  * If the icmp is a redirect to a better router, always set the
 1911                  * is_router flag.  Otherwise, if the entry is newly created,
 1912                  * clear the flag.  [RFC 2461, sec 8.3]
 1913                  */
 1914                 if (code == ND_REDIRECT_ROUTER)
 1915                         ln_router = 1;
 1916                 else {
 1917                         if (is_new)                             /* (6-7) */
 1918                                 ln_router = 0;
 1919                 }
 1920                 break;
 1921         case ND_ROUTER_SOLICIT:
 1922                 /*
 1923                  * is_router flag must always be cleared.
 1924                  */
 1925                 ln_router = 0;
 1926                 break;
 1927         case ND_ROUTER_ADVERT:
 1928                 /*
 1929                  * Mark an entry with lladdr as a router.
 1930                  */
 1931                 if ((!is_new && (old_addr || new_addr)) ||      /* (2-5) */
 1932                     (is_new && new_addr)) {                     /* (7) */
 1933                         ln_router = 1;
 1934                 }
 1935                 break;
 1936         }
 1937 
 1938         return (ln_router);
 1939 }
 1940 
 1941 /*
 1942  * Create neighbor cache entry and cache link-layer address,
 1943  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
 1944  *
 1945  * type - ICMP6 type
 1946  * code - type dependent information
 1947  *
 1948  */
 1949 void
 1950 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 1951     int lladdrlen, int type, int code)
 1952 {
 1953         struct llentry *ln = NULL, *ln_tmp;
 1954         int is_newentry;
 1955         int do_update;
 1956         int olladdr;
 1957         int llchange;
 1958         int flags;
 1959         uint16_t router = 0;
 1960         struct sockaddr_in6 sin6;
 1961         struct mbuf *chain = NULL;
 1962         u_char linkhdr[LLE_MAX_LINKHDR];
 1963         size_t linkhdrsize;
 1964         int lladdr_off;
 1965 
 1966         IF_AFDATA_UNLOCK_ASSERT(ifp);
 1967 
 1968         KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
 1969         KASSERT(from != NULL, ("%s: from == NULL", __func__));
 1970 
 1971         /* nothing must be updated for unspecified address */
 1972         if (IN6_IS_ADDR_UNSPECIFIED(from))
 1973                 return;
 1974 
 1975         /*
 1976          * Validation about ifp->if_addrlen and lladdrlen must be done in
 1977          * the caller.
 1978          *
 1979          * XXX If the link does not have link-layer adderss, what should
 1980          * we do? (ifp->if_addrlen == 0)
 1981          * Spec says nothing in sections for RA, RS and NA.  There's small
 1982          * description on it in NS section (RFC 2461 7.2.3).
 1983          */
 1984         flags = lladdr ? LLE_EXCLUSIVE : 0;
 1985         IF_AFDATA_RLOCK(ifp);
 1986         ln = nd6_lookup(from, flags, ifp);
 1987         IF_AFDATA_RUNLOCK(ifp);
 1988         is_newentry = 0;
 1989         if (ln == NULL) {
 1990                 flags |= LLE_EXCLUSIVE;
 1991                 ln = nd6_alloc(from, 0, ifp);
 1992                 if (ln == NULL)
 1993                         return;
 1994 
 1995                 /*
 1996                  * Since we already know all the data for the new entry,
 1997                  * fill it before insertion.
 1998                  */
 1999                 if (lladdr != NULL) {
 2000                         linkhdrsize = sizeof(linkhdr);
 2001                         if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
 2002                             linkhdr, &linkhdrsize, &lladdr_off) != 0)
 2003                                 return;
 2004                         lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
 2005                             lladdr_off);
 2006                 }
 2007 
 2008                 IF_AFDATA_WLOCK(ifp);
 2009                 LLE_WLOCK(ln);
 2010                 /* Prefer any existing lle over newly-created one */
 2011                 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp);
 2012                 if (ln_tmp == NULL)
 2013                         lltable_link_entry(LLTABLE6(ifp), ln);
 2014                 IF_AFDATA_WUNLOCK(ifp);
 2015                 if (ln_tmp == NULL) {
 2016                         /* No existing lle, mark as new entry (6,7) */
 2017                         is_newentry = 1;
 2018                         if (lladdr != NULL) {   /* (7) */
 2019                                 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
 2020                                 EVENTHANDLER_INVOKE(lle_event, ln,
 2021                                     LLENTRY_RESOLVED);
 2022                         }
 2023                 } else {
 2024                         lltable_free_entry(LLTABLE6(ifp), ln);
 2025                         ln = ln_tmp;
 2026                         ln_tmp = NULL;
 2027                 }
 2028         } 
 2029         /* do nothing if static ndp is set */
 2030         if ((ln->la_flags & LLE_STATIC)) {
 2031                 if (flags & LLE_EXCLUSIVE)
 2032                         LLE_WUNLOCK(ln);
 2033                 else
 2034                         LLE_RUNLOCK(ln);
 2035                 return;
 2036         }
 2037 
 2038         olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
 2039         if (olladdr && lladdr) {
 2040                 llchange = bcmp(lladdr, ln->ll_addr,
 2041                     ifp->if_addrlen);
 2042         } else if (!olladdr && lladdr)
 2043                 llchange = 1;
 2044         else
 2045                 llchange = 0;
 2046 
 2047         /*
 2048          * newentry olladdr  lladdr  llchange   (*=record)
 2049          *      0       n       n       --      (1)
 2050          *      0       y       n       --      (2)
 2051          *      0       n       y       y       (3) * STALE
 2052          *      0       y       y       n       (4) *
 2053          *      0       y       y       y       (5) * STALE
 2054          *      1       --      n       --      (6)   NOSTATE(= PASSIVE)
 2055          *      1       --      y       --      (7) * STALE
 2056          */
 2057 
 2058         do_update = 0;
 2059         if (is_newentry == 0 && llchange != 0) {
 2060                 do_update = 1;  /* (3,5) */
 2061 
 2062                 /*
 2063                  * Record source link-layer address
 2064                  * XXX is it dependent to ifp->if_type?
 2065                  */
 2066                 linkhdrsize = sizeof(linkhdr);
 2067                 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
 2068                     linkhdr, &linkhdrsize, &lladdr_off) != 0)
 2069                         return;
 2070 
 2071                 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
 2072                     lladdr_off) == 0) {
 2073                         /* Entry was deleted */
 2074                         return;
 2075                 }
 2076 
 2077                 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
 2078 
 2079                 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
 2080 
 2081                 if (ln->la_hold != NULL)
 2082                         nd6_grab_holdchain(ln, &chain, &sin6);
 2083         }
 2084 
 2085         /* Calculates new router status */
 2086         router = nd6_is_router(type, code, is_newentry, olladdr,
 2087             lladdr != NULL ? 1 : 0, ln->ln_router);
 2088 
 2089         ln->ln_router = router;
 2090         /* Mark non-router redirects with special flag */
 2091         if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
 2092                 ln->la_flags |= LLE_REDIRECT;
 2093 
 2094         if (flags & LLE_EXCLUSIVE)
 2095                 LLE_WUNLOCK(ln);
 2096         else
 2097                 LLE_RUNLOCK(ln);
 2098 
 2099         if (chain != NULL)
 2100                 nd6_flush_holdchain(ifp, chain, &sin6);
 2101         
 2102         /*
 2103          * When the link-layer address of a router changes, select the
 2104          * best router again.  In particular, when the neighbor entry is newly
 2105          * created, it might affect the selection policy.
 2106          * Question: can we restrict the first condition to the "is_newentry"
 2107          * case?
 2108          * XXX: when we hear an RA from a new router with the link-layer
 2109          * address option, defrouter_select_fib() is called twice, since
 2110          * defrtrlist_update called the function as well.  However, I believe
 2111          * we can compromise the overhead, since it only happens the first
 2112          * time.
 2113          * XXX: although defrouter_select_fib() should not have a bad effect
 2114          * for those are not autoconfigured hosts, we explicitly avoid such
 2115          * cases for safety.
 2116          */
 2117         if ((do_update || is_newentry) && router &&
 2118             ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 2119                 /*
 2120                  * guaranteed recursion
 2121                  */
 2122                 defrouter_select_fib(ifp->if_fib);
 2123         }
 2124 }
 2125 
 2126 static void
 2127 nd6_slowtimo(void *arg)
 2128 {
 2129         CURVNET_SET((struct vnet *) arg);
 2130         struct nd_ifinfo *nd6if;
 2131         struct ifnet *ifp;
 2132 
 2133         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
 2134             nd6_slowtimo, curvnet);
 2135         IFNET_RLOCK_NOSLEEP();
 2136         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2137                 if (ifp->if_afdata[AF_INET6] == NULL)
 2138                         continue;
 2139                 nd6if = ND_IFINFO(ifp);
 2140                 if (nd6if->basereachable && /* already initialized */
 2141                     (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
 2142                         /*
 2143                          * Since reachable time rarely changes by router
 2144                          * advertisements, we SHOULD insure that a new random
 2145                          * value gets recomputed at least once every few hours.
 2146                          * (RFC 2461, 6.3.4)
 2147                          */
 2148                         nd6if->recalctm = V_nd6_recalc_reachtm_interval;
 2149                         nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
 2150                 }
 2151         }
 2152         IFNET_RUNLOCK_NOSLEEP();
 2153         CURVNET_RESTORE();
 2154 }
 2155 
 2156 void
 2157 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
 2158     struct sockaddr_in6 *sin6)
 2159 {
 2160 
 2161         LLE_WLOCK_ASSERT(ln);
 2162 
 2163         *chain = ln->la_hold;
 2164         ln->la_hold = NULL;
 2165         lltable_fill_sa_entry(ln, (struct sockaddr *)sin6);
 2166 
 2167         if (ln->ln_state == ND6_LLINFO_STALE) {
 2168 
 2169                 /*
 2170                  * The first time we send a packet to a
 2171                  * neighbor whose entry is STALE, we have
 2172                  * to change the state to DELAY and a sets
 2173                  * a timer to expire in DELAY_FIRST_PROBE_TIME
 2174                  * seconds to ensure do neighbor unreachability
 2175                  * detection on expiration.
 2176                  * (RFC 2461 7.3.3)
 2177                  */
 2178                 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
 2179         }
 2180 }
 2181 
 2182 int
 2183 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
 2184     struct sockaddr_in6 *dst, struct route *ro)
 2185 {
 2186         int error;
 2187         int ip6len;
 2188         struct ip6_hdr *ip6;
 2189         struct m_tag *mtag;
 2190 
 2191 #ifdef MAC
 2192         mac_netinet6_nd6_send(ifp, m);
 2193 #endif
 2194 
 2195         /*
 2196          * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
 2197          * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
 2198          * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
 2199          * to be diverted to user space.  When re-injected into the kernel,
 2200          * send_output() will directly dispatch them to the outgoing interface.
 2201          */
 2202         if (send_sendso_input_hook != NULL) {
 2203                 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
 2204                 if (mtag != NULL) {
 2205                         ip6 = mtod(m, struct ip6_hdr *);
 2206                         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
 2207                         /* Use the SEND socket */
 2208                         error = send_sendso_input_hook(m, ifp, SND_OUT,
 2209                             ip6len);
 2210                         /* -1 == no app on SEND socket */
 2211                         if (error == 0 || error != -1)
 2212                             return (error);
 2213                 }
 2214         }
 2215 
 2216         m_clrprotoflags(m);     /* Avoid confusing lower layers. */
 2217         IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
 2218             mtod(m, struct ip6_hdr *));
 2219 
 2220         if ((ifp->if_flags & IFF_LOOPBACK) == 0)
 2221                 origifp = ifp;
 2222 
 2223         error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro);
 2224         return (error);
 2225 }
 2226 
 2227 /*
 2228  * Lookup link headerfor @sa_dst address. Stores found
 2229  * data in @desten buffer. Copy of lle ln_flags can be also
 2230  * saved in @pflags if @pflags is non-NULL.
 2231  *
 2232  * If destination LLE does not exists or lle state modification
 2233  * is required, call "slow" version.
 2234  *
 2235  * Return values:
 2236  * - 0 on success (address copied to buffer).
 2237  * - EWOULDBLOCK (no local error, but address is still unresolved)
 2238  * - other errors (alloc failure, etc)
 2239  */
 2240 int
 2241 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
 2242     const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags,
 2243     struct llentry **plle)
 2244 {
 2245         struct llentry *ln = NULL;
 2246         const struct sockaddr_in6 *dst6;
 2247 
 2248         if (pflags != NULL)
 2249                 *pflags = 0;
 2250 
 2251         dst6 = (const struct sockaddr_in6 *)sa_dst;
 2252 
 2253         /* discard the packet if IPv6 operation is disabled on the interface */
 2254         if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
 2255                 m_freem(m);
 2256                 return (ENETDOWN); /* better error? */
 2257         }
 2258 
 2259         if (m != NULL && m->m_flags & M_MCAST) {
 2260                 switch (ifp->if_type) {
 2261                 case IFT_ETHER:
 2262                 case IFT_L2VLAN:
 2263                 case IFT_BRIDGE:
 2264                         ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
 2265                                                  desten);
 2266                         return (0);
 2267                 default:
 2268                         m_freem(m);
 2269                         return (EAFNOSUPPORT);
 2270                 }
 2271         }
 2272 
 2273         IF_AFDATA_RLOCK(ifp);
 2274         ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED,
 2275             ifp);
 2276         if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
 2277                 /* Entry found, let's copy lle info */
 2278                 bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
 2279                 if (pflags != NULL)
 2280                         *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
 2281                 /* Check if we have feedback request from nd6 timer */
 2282                 if (ln->r_skip_req != 0) {
 2283                         LLE_REQ_LOCK(ln);
 2284                         ln->r_skip_req = 0; /* Notify that entry was used */
 2285                         ln->lle_hittime = time_uptime;
 2286                         LLE_REQ_UNLOCK(ln);
 2287                 }
 2288                 if (plle) {
 2289                         LLE_ADDREF(ln);
 2290                         *plle = ln;
 2291                         LLE_WUNLOCK(ln);
 2292                 }
 2293                 IF_AFDATA_RUNLOCK(ifp);
 2294                 return (0);
 2295         } else if (plle && ln)
 2296                 LLE_WUNLOCK(ln);
 2297         IF_AFDATA_RUNLOCK(ifp);
 2298 
 2299         return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle));
 2300 }
 2301 
 2302 
 2303 /*
 2304  * Do L2 address resolution for @sa_dst address. Stores found
 2305  * address in @desten buffer. Copy of lle ln_flags can be also
 2306  * saved in @pflags if @pflags is non-NULL.
 2307  *
 2308  * Heavy version.
 2309  * Function assume that destination LLE does not exist,
 2310  * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired.
 2311  *
 2312  * Set noinline to be dtrace-friendly
 2313  */
 2314 static __noinline int
 2315 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m,
 2316     const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags,
 2317     struct llentry **plle)
 2318 {
 2319         struct llentry *lle = NULL, *lle_tmp;
 2320         struct in6_addr *psrc, src;
 2321         int send_ns, ll_len;
 2322         char *lladdr;
 2323 
 2324         /*
 2325          * Address resolution or Neighbor Unreachability Detection
 2326          * for the next hop.
 2327          * At this point, the destination of the packet must be a unicast
 2328          * or an anycast address(i.e. not a multicast).
 2329          */
 2330         if (lle == NULL) {
 2331                 IF_AFDATA_RLOCK(ifp);
 2332                 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
 2333                 IF_AFDATA_RUNLOCK(ifp);
 2334                 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp))  {
 2335                         /*
 2336                          * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
 2337                          * the condition below is not very efficient.  But we believe
 2338                          * it is tolerable, because this should be a rare case.
 2339                          */
 2340                         lle = nd6_alloc(&dst->sin6_addr, 0, ifp);
 2341                         if (lle == NULL) {
 2342                                 char ip6buf[INET6_ADDRSTRLEN];
 2343                                 log(LOG_DEBUG,
 2344                                     "nd6_output: can't allocate llinfo for %s "
 2345                                     "(ln=%p)\n",
 2346                                     ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
 2347                                 m_freem(m);
 2348                                 return (ENOBUFS);
 2349                         }
 2350 
 2351                         IF_AFDATA_WLOCK(ifp);
 2352                         LLE_WLOCK(lle);
 2353                         /* Prefer any existing entry over newly-created one */
 2354                         lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
 2355                         if (lle_tmp == NULL)
 2356                                 lltable_link_entry(LLTABLE6(ifp), lle);
 2357                         IF_AFDATA_WUNLOCK(ifp);
 2358                         if (lle_tmp != NULL) {
 2359                                 lltable_free_entry(LLTABLE6(ifp), lle);
 2360                                 lle = lle_tmp;
 2361                                 lle_tmp = NULL;
 2362                         }
 2363                 }
 2364         } 
 2365         if (lle == NULL) {
 2366                 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
 2367                         m_freem(m);
 2368                         return (ENOBUFS);
 2369                 }
 2370 
 2371                 if (m != NULL)
 2372                         m_freem(m);
 2373                 return (ENOBUFS);
 2374         }
 2375 
 2376         LLE_WLOCK_ASSERT(lle);
 2377 
 2378         /*
 2379          * The first time we send a packet to a neighbor whose entry is
 2380          * STALE, we have to change the state to DELAY and a sets a timer to
 2381          * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
 2382          * neighbor unreachability detection on expiration.
 2383          * (RFC 2461 7.3.3)
 2384          */
 2385         if (lle->ln_state == ND6_LLINFO_STALE)
 2386                 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
 2387 
 2388         /*
 2389          * If the neighbor cache entry has a state other than INCOMPLETE
 2390          * (i.e. its link-layer address is already resolved), just
 2391          * send the packet.
 2392          */
 2393         if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
 2394                 if (flags & LLE_ADDRONLY) {
 2395                         lladdr = lle->ll_addr;
 2396                         ll_len = ifp->if_addrlen;
 2397                 } else {
 2398                         lladdr = lle->r_linkdata;
 2399                         ll_len = lle->r_hdrlen;
 2400                 }
 2401                 bcopy(lladdr, desten, ll_len);
 2402                 if (pflags != NULL)
 2403                         *pflags = lle->la_flags;
 2404                 if (plle) {
 2405                         LLE_ADDREF(lle);
 2406                         *plle = lle;
 2407                 }
 2408                 LLE_WUNLOCK(lle);
 2409                 return (0);
 2410         }
 2411 
 2412         /*
 2413          * There is a neighbor cache entry, but no ethernet address
 2414          * response yet.  Append this latest packet to the end of the
 2415          * packet queue in the mbuf.  When it exceeds nd6_maxqueuelen,
 2416          * the oldest packet in the queue will be removed.
 2417          */
 2418 
 2419         if (lle->la_hold != NULL) {
 2420                 struct mbuf *m_hold;
 2421                 int i;
 2422                 
 2423                 i = 0;
 2424                 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
 2425                         i++;
 2426                         if (m_hold->m_nextpkt == NULL) {
 2427                                 m_hold->m_nextpkt = m;
 2428                                 break;
 2429                         }
 2430                 }
 2431                 while (i >= V_nd6_maxqueuelen) {
 2432                         m_hold = lle->la_hold;
 2433                         lle->la_hold = lle->la_hold->m_nextpkt;
 2434                         m_freem(m_hold);
 2435                         i--;
 2436                 }
 2437         } else {
 2438                 lle->la_hold = m;
 2439         }
 2440 
 2441         /*
 2442          * If there has been no NS for the neighbor after entering the
 2443          * INCOMPLETE state, send the first solicitation.
 2444          * Note that for newly-created lle la_asked will be 0,
 2445          * so we will transition from ND6_LLINFO_NOSTATE to
 2446          * ND6_LLINFO_INCOMPLETE state here.
 2447          */
 2448         psrc = NULL;
 2449         send_ns = 0;
 2450         if (lle->la_asked == 0) {
 2451                 lle->la_asked++;
 2452                 send_ns = 1;
 2453                 psrc = nd6_llinfo_get_holdsrc(lle, &src);
 2454 
 2455                 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
 2456         }
 2457         LLE_WUNLOCK(lle);
 2458         if (send_ns != 0)
 2459                 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
 2460 
 2461         return (EWOULDBLOCK);
 2462 }
 2463 
 2464 /*
 2465  * Do L2 address resolution for @sa_dst address. Stores found
 2466  * address in @desten buffer. Copy of lle ln_flags can be also
 2467  * saved in @pflags if @pflags is non-NULL.
 2468  *
 2469  * Return values:
 2470  * - 0 on success (address copied to buffer).
 2471  * - EWOULDBLOCK (no local error, but address is still unresolved)
 2472  * - other errors (alloc failure, etc)
 2473  */
 2474 int
 2475 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
 2476     char *desten, uint32_t *pflags)
 2477 {
 2478         int error;
 2479 
 2480         flags |= LLE_ADDRONLY;
 2481         error = nd6_resolve_slow(ifp, flags, NULL,
 2482             (const struct sockaddr_in6 *)dst, desten, pflags, NULL);
 2483         return (error);
 2484 }
 2485 
 2486 int
 2487 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain,
 2488     struct sockaddr_in6 *dst)
 2489 {
 2490         struct mbuf *m, *m_head;
 2491         int error = 0;
 2492 
 2493         m_head = chain;
 2494 
 2495         while (m_head) {
 2496                 m = m_head;
 2497                 m_head = m_head->m_nextpkt;
 2498                 error = nd6_output_ifp(ifp, ifp, m, dst, NULL);
 2499         }
 2500 
 2501         /*
 2502          * XXX
 2503          * note that intermediate errors are blindly ignored
 2504          */
 2505         return (error);
 2506 }
 2507 
 2508 static int
 2509 nd6_need_cache(struct ifnet *ifp)
 2510 {
 2511         /*
 2512          * XXX: we currently do not make neighbor cache on any interface
 2513          * other than Ethernet and GIF.
 2514          *
 2515          * RFC2893 says:
 2516          * - unidirectional tunnels needs no ND
 2517          */
 2518         switch (ifp->if_type) {
 2519         case IFT_ETHER:
 2520         case IFT_IEEE1394:
 2521         case IFT_L2VLAN:
 2522         case IFT_INFINIBAND:
 2523         case IFT_BRIDGE:
 2524         case IFT_PROPVIRTUAL:
 2525                 return (1);
 2526         default:
 2527                 return (0);
 2528         }
 2529 }
 2530 
 2531 /*
 2532  * Add pernament ND6 link-layer record for given
 2533  * interface address.
 2534  *
 2535  * Very similar to IPv4 arp_ifinit(), but:
 2536  * 1) IPv6 DAD is performed in different place
 2537  * 2) It is called by IPv6 protocol stack in contrast to
 2538  * arp_ifinit() which is typically called in SIOCSIFADDR
 2539  * driver ioctl handler.
 2540  *
 2541  */
 2542 int
 2543 nd6_add_ifa_lle(struct in6_ifaddr *ia)
 2544 {
 2545         struct ifnet *ifp;
 2546         struct llentry *ln, *ln_tmp;
 2547         struct sockaddr *dst;
 2548 
 2549         ifp = ia->ia_ifa.ifa_ifp;
 2550         if (nd6_need_cache(ifp) == 0)
 2551                 return (0);
 2552 
 2553         ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
 2554         dst = (struct sockaddr *)&ia->ia_addr;
 2555         ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
 2556         if (ln == NULL)
 2557                 return (ENOBUFS);
 2558 
 2559         IF_AFDATA_WLOCK(ifp);
 2560         LLE_WLOCK(ln);
 2561         /* Unlink any entry if exists */
 2562         ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst);
 2563         if (ln_tmp != NULL)
 2564                 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
 2565         lltable_link_entry(LLTABLE6(ifp), ln);
 2566         IF_AFDATA_WUNLOCK(ifp);
 2567 
 2568         if (ln_tmp != NULL)
 2569                 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
 2570         EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
 2571 
 2572         LLE_WUNLOCK(ln);
 2573         if (ln_tmp != NULL)
 2574                 llentry_free(ln_tmp);
 2575 
 2576         return (0);
 2577 }
 2578 
 2579 /*
 2580  * Removes either all lle entries for given @ia, or lle
 2581  * corresponding to @ia address.
 2582  */
 2583 void
 2584 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
 2585 {
 2586         struct sockaddr_in6 mask, addr;
 2587         struct sockaddr *saddr, *smask;
 2588         struct ifnet *ifp;
 2589 
 2590         ifp = ia->ia_ifa.ifa_ifp;
 2591         memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
 2592         memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
 2593         saddr = (struct sockaddr *)&addr;
 2594         smask = (struct sockaddr *)&mask;
 2595 
 2596         if (all != 0)
 2597                 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
 2598         else
 2599                 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
 2600 }
 2601 
 2602 static void 
 2603 clear_llinfo_pqueue(struct llentry *ln)
 2604 {
 2605         struct mbuf *m_hold, *m_hold_next;
 2606 
 2607         for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
 2608                 m_hold_next = m_hold->m_nextpkt;
 2609                 m_freem(m_hold);
 2610         }
 2611 
 2612         ln->la_hold = NULL;
 2613 }
 2614 
 2615 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
 2616 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
 2617 
 2618 SYSCTL_DECL(_net_inet6_icmp6);
 2619 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
 2620         CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
 2621         NULL, 0, nd6_sysctl_drlist, "S,in6_defrouter",
 2622         "NDP default router list");
 2623 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
 2624         CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
 2625         NULL, 0, nd6_sysctl_prlist, "S,in6_prefix",
 2626         "NDP prefix list");
 2627 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
 2628         CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
 2629 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
 2630         CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");
 2631 
 2632 static int
 2633 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
 2634 {
 2635         struct in6_defrouter d;
 2636         struct nd_defrouter *dr;
 2637         int error;
 2638 
 2639         if (req->newptr != NULL)
 2640                 return (EPERM);
 2641 
 2642         error = sysctl_wire_old_buffer(req, 0);
 2643         if (error != 0)
 2644                 return (error);
 2645 
 2646         bzero(&d, sizeof(d));
 2647         d.rtaddr.sin6_family = AF_INET6;
 2648         d.rtaddr.sin6_len = sizeof(d.rtaddr);
 2649 
 2650         ND6_RLOCK();
 2651         TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
 2652                 d.rtaddr.sin6_addr = dr->rtaddr;
 2653                 error = sa6_recoverscope(&d.rtaddr);
 2654                 if (error != 0)
 2655                         break;
 2656                 d.flags = dr->raflags;
 2657                 d.rtlifetime = dr->rtlifetime;
 2658                 d.expire = dr->expire + (time_second - time_uptime);
 2659                 d.if_index = dr->ifp->if_index;
 2660                 error = SYSCTL_OUT(req, &d, sizeof(d));
 2661                 if (error != 0)
 2662                         break;
 2663         }
 2664         ND6_RUNLOCK();
 2665         return (error);
 2666 }
 2667 
 2668 static int
 2669 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
 2670 {
 2671         struct in6_prefix p;
 2672         struct sockaddr_in6 s6;
 2673         struct nd_prefix *pr;
 2674         struct nd_pfxrouter *pfr;
 2675         time_t maxexpire;
 2676         int error;
 2677         char ip6buf[INET6_ADDRSTRLEN];
 2678 
 2679         if (req->newptr)
 2680                 return (EPERM);
 2681 
 2682         error = sysctl_wire_old_buffer(req, 0);
 2683         if (error != 0)
 2684                 return (error);
 2685 
 2686         bzero(&p, sizeof(p));
 2687         p.origin = PR_ORIG_RA;
 2688         bzero(&s6, sizeof(s6));
 2689         s6.sin6_family = AF_INET6;
 2690         s6.sin6_len = sizeof(s6);
 2691 
 2692         ND6_RLOCK();
 2693         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 2694                 p.prefix = pr->ndpr_prefix;
 2695                 if (sa6_recoverscope(&p.prefix)) {
 2696                         log(LOG_ERR, "scope error in prefix list (%s)\n",
 2697                             ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
 2698                         /* XXX: press on... */
 2699                 }
 2700                 p.raflags = pr->ndpr_raf;
 2701                 p.prefixlen = pr->ndpr_plen;
 2702                 p.vltime = pr->ndpr_vltime;
 2703                 p.pltime = pr->ndpr_pltime;
 2704                 p.if_index = pr->ndpr_ifp->if_index;
 2705                 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
 2706                         p.expire = 0;
 2707                 else {
 2708                         /* XXX: we assume time_t is signed. */
 2709                         maxexpire = (-1) &
 2710                             ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
 2711                         if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
 2712                                 p.expire = pr->ndpr_lastupdate +
 2713                                     pr->ndpr_vltime +
 2714                                     (time_second - time_uptime);
 2715                         else
 2716                                 p.expire = maxexpire;
 2717                 }
 2718                 p.refcnt = pr->ndpr_addrcnt;
 2719                 p.flags = pr->ndpr_stateflags;
 2720                 p.advrtrs = 0;
 2721                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
 2722                         p.advrtrs++;
 2723                 error = SYSCTL_OUT(req, &p, sizeof(p));
 2724                 if (error != 0)
 2725                         break;
 2726                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
 2727                         s6.sin6_addr = pfr->router->rtaddr;
 2728                         if (sa6_recoverscope(&s6))
 2729                                 log(LOG_ERR,
 2730                                     "scope error in prefix list (%s)\n",
 2731                                     ip6_sprintf(ip6buf, &pfr->router->rtaddr));
 2732                         error = SYSCTL_OUT(req, &s6, sizeof(s6));
 2733                         if (error != 0)
 2734                                 goto out;
 2735                 }
 2736         }
 2737 out:
 2738         ND6_RUNLOCK();
 2739         return (error);
 2740 }

Cache object: 18b95332daad55ac75013f4b635ddee1


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.