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 /*      $FreeBSD: releng/5.2/sys/netinet6/nd6.c 122334 2003-11-08 23:36:32Z sam $       */
    2 /*      $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $   */
    3 
    4 /*
    5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
    6  * All rights reserved.
    7  *
    8  * Redistribution and use in source and binary forms, with or without
    9  * modification, are permitted provided that the following conditions
   10  * are met:
   11  * 1. Redistributions of source code must retain the above copyright
   12  *    notice, this list of conditions and the following disclaimer.
   13  * 2. Redistributions in binary form must reproduce the above copyright
   14  *    notice, this list of conditions and the following disclaimer in the
   15  *    documentation and/or other materials provided with the distribution.
   16  * 3. Neither the name of the project nor the names of its contributors
   17  *    may be used to endorse or promote products derived from this software
   18  *    without specific prior written permission.
   19  *
   20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
   21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
   24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   30  * SUCH DAMAGE.
   31  */
   32 
   33 #include "opt_inet.h"
   34 #include "opt_inet6.h"
   35 #include "opt_mac.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/systm.h>
   39 #include <sys/callout.h>
   40 #include <sys/mac.h>
   41 #include <sys/malloc.h>
   42 #include <sys/mbuf.h>
   43 #include <sys/socket.h>
   44 #include <sys/sockio.h>
   45 #include <sys/time.h>
   46 #include <sys/kernel.h>
   47 #include <sys/protosw.h>
   48 #include <sys/errno.h>
   49 #include <sys/syslog.h>
   50 #include <sys/queue.h>
   51 #include <sys/sysctl.h>
   52 
   53 #include <net/if.h>
   54 #include <net/if_arc.h>
   55 #include <net/if_dl.h>
   56 #include <net/if_types.h>
   57 #include <net/if_atm.h>
   58 #include <net/iso88025.h>
   59 #include <net/fddi.h>
   60 #include <net/route.h>
   61 
   62 #include <netinet/in.h>
   63 #include <netinet/if_ether.h>
   64 #include <netinet6/in6_var.h>
   65 #include <netinet/ip6.h>
   66 #include <netinet6/ip6_var.h>
   67 #include <netinet6/nd6.h>
   68 #include <netinet6/in6_prefix.h>
   69 #include <netinet/icmp6.h>
   70 
   71 #include <net/net_osdep.h>
   72 
   73 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
   74 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
   75 
   76 #define SIN6(s) ((struct sockaddr_in6 *)s)
   77 #define SDL(s) ((struct sockaddr_dl *)s)
   78 
   79 /* timer values */
   80 int     nd6_prune       = 1;    /* walk list every 1 seconds */
   81 int     nd6_delay       = 5;    /* delay first probe time 5 second */
   82 int     nd6_umaxtries   = 3;    /* maximum unicast query */
   83 int     nd6_mmaxtries   = 3;    /* maximum multicast query */
   84 int     nd6_useloopback = 1;    /* use loopback interface for local traffic */
   85 int     nd6_gctimer     = (60 * 60 * 24); /* 1 day: garbage collection timer */
   86 
   87 /* preventing too many loops in ND option parsing */
   88 int nd6_maxndopt = 10;  /* max # of ND options allowed */
   89 
   90 int nd6_maxnudhint = 0; /* max # of subsequent upper layer hints */
   91 
   92 #ifdef ND6_DEBUG
   93 int nd6_debug = 1;
   94 #else
   95 int nd6_debug = 0;
   96 #endif
   97 
   98 /* for debugging? */
   99 static int nd6_inuse, nd6_allocated;
  100 
  101 struct llinfo_nd6 llinfo_nd6 = {&llinfo_nd6, &llinfo_nd6};
  102 struct nd_drhead nd_defrouter;
  103 struct nd_prhead nd_prefix = { 0 };
  104 
  105 int nd6_recalc_reachtm_interval = ND6_RECALC_REACHTM_INTERVAL;
  106 static struct sockaddr_in6 all1_sa;
  107 
  108 static void nd6_setmtu0 __P((struct ifnet *, struct nd_ifinfo *));
  109 static void nd6_slowtimo __P((void *));
  110 static int regen_tmpaddr __P((struct in6_ifaddr *));
  111 
  112 struct callout nd6_slowtimo_ch;
  113 struct callout nd6_timer_ch;
  114 extern struct callout in6_tmpaddrtimer_ch;
  115 
  116 void
  117 nd6_init()
  118 {
  119         static int nd6_init_done = 0;
  120         int i;
  121 
  122         if (nd6_init_done) {
  123                 log(LOG_NOTICE, "nd6_init called more than once(ignored)\n");
  124                 return;
  125         }
  126 
  127         all1_sa.sin6_family = AF_INET6;
  128         all1_sa.sin6_len = sizeof(struct sockaddr_in6);
  129         for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
  130                 all1_sa.sin6_addr.s6_addr[i] = 0xff;
  131 
  132         /* initialization of the default router list */
  133         TAILQ_INIT(&nd_defrouter);
  134 
  135         nd6_init_done = 1;
  136 
  137         /* start timer */
  138         callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
  139             nd6_slowtimo, NULL);
  140 }
  141 
  142 struct nd_ifinfo *
  143 nd6_ifattach(ifp)
  144         struct ifnet *ifp;
  145 {
  146         struct nd_ifinfo *nd;
  147 
  148         nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
  149         bzero(nd, sizeof(*nd));
  150 
  151         nd->initialized = 1;
  152 
  153         nd->chlim = IPV6_DEFHLIM;
  154         nd->basereachable = REACHABLE_TIME;
  155         nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
  156         nd->retrans = RETRANS_TIMER;
  157         /*
  158          * Note that the default value of ip6_accept_rtadv is 0, which means
  159          * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
  160          * here.
  161          */
  162         nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
  163 
  164         /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
  165         nd6_setmtu0(ifp, nd);
  166 
  167         return nd;
  168 }
  169 
  170 void
  171 nd6_ifdetach(nd)
  172         struct nd_ifinfo *nd;
  173 {
  174 
  175         free(nd, M_IP6NDP);
  176 }
  177 
  178 /*
  179  * Reset ND level link MTU. This function is called when the physical MTU
  180  * changes, which means we might have to adjust the ND level MTU.
  181  */
  182 void
  183 nd6_setmtu(ifp)
  184         struct ifnet *ifp;
  185 {
  186 
  187         nd6_setmtu0(ifp, ND_IFINFO(ifp));
  188 }
  189 
  190 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
  191 void
  192 nd6_setmtu0(ifp, ndi)
  193         struct ifnet *ifp;
  194         struct nd_ifinfo *ndi;
  195 {
  196         u_int32_t omaxmtu;
  197 
  198         omaxmtu = ndi->maxmtu;
  199 
  200         switch (ifp->if_type) {
  201         case IFT_ARCNET:
  202                 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
  203                 break;
  204         case IFT_ETHER:
  205                 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
  206                 break;
  207         case IFT_FDDI:
  208                 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
  209                 break;
  210         case IFT_ATM:
  211                 ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
  212                 break;
  213         case IFT_IEEE1394:      /* XXX should be IEEE1394MTU(1500) */
  214                 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
  215                 break;
  216 #ifdef IFT_IEEE80211
  217         case IFT_IEEE80211:     /* XXX should be IEEE80211MTU(1500) */
  218                 ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
  219                 break;
  220 #endif
  221          case IFT_ISO88025:
  222                  ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
  223                  break;
  224         default:
  225                 ndi->maxmtu = ifp->if_mtu;
  226                 break;
  227         }
  228 
  229         /*
  230          * Decreasing the interface MTU under IPV6 minimum MTU may cause
  231          * undesirable situation.  We thus notify the operator of the change
  232          * explicitly.  The check for omaxmtu is necessary to restrict the
  233          * log to the case of changing the MTU, not initializing it.
  234          */
  235         if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
  236                 log(LOG_NOTICE, "nd6_setmtu0: "
  237                     "new link MTU on %s (%lu) is too small for IPv6\n",
  238                     if_name(ifp), (unsigned long)ndi->maxmtu);
  239         }
  240 
  241         if (ndi->maxmtu > in6_maxmtu)
  242                 in6_setmaxmtu(); /* check all interfaces just in case */
  243 
  244 #undef MIN
  245 }
  246 
  247 void
  248 nd6_option_init(opt, icmp6len, ndopts)
  249         void *opt;
  250         int icmp6len;
  251         union nd_opts *ndopts;
  252 {
  253 
  254         bzero(ndopts, sizeof(*ndopts));
  255         ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
  256         ndopts->nd_opts_last
  257                 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
  258 
  259         if (icmp6len == 0) {
  260                 ndopts->nd_opts_done = 1;
  261                 ndopts->nd_opts_search = NULL;
  262         }
  263 }
  264 
  265 /*
  266  * Take one ND option.
  267  */
  268 struct nd_opt_hdr *
  269 nd6_option(ndopts)
  270         union nd_opts *ndopts;
  271 {
  272         struct nd_opt_hdr *nd_opt;
  273         int olen;
  274 
  275         if (!ndopts)
  276                 panic("ndopts == NULL in nd6_option");
  277         if (!ndopts->nd_opts_last)
  278                 panic("uninitialized ndopts in nd6_option");
  279         if (!ndopts->nd_opts_search)
  280                 return NULL;
  281         if (ndopts->nd_opts_done)
  282                 return NULL;
  283 
  284         nd_opt = ndopts->nd_opts_search;
  285 
  286         /* make sure nd_opt_len is inside the buffer */
  287         if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
  288                 bzero(ndopts, sizeof(*ndopts));
  289                 return NULL;
  290         }
  291 
  292         olen = nd_opt->nd_opt_len << 3;
  293         if (olen == 0) {
  294                 /*
  295                  * Message validation requires that all included
  296                  * options have a length that is greater than zero.
  297                  */
  298                 bzero(ndopts, sizeof(*ndopts));
  299                 return NULL;
  300         }
  301 
  302         ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
  303         if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
  304                 /* option overruns the end of buffer, invalid */
  305                 bzero(ndopts, sizeof(*ndopts));
  306                 return NULL;
  307         } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
  308                 /* reached the end of options chain */
  309                 ndopts->nd_opts_done = 1;
  310                 ndopts->nd_opts_search = NULL;
  311         }
  312         return nd_opt;
  313 }
  314 
  315 /*
  316  * Parse multiple ND options.
  317  * This function is much easier to use, for ND routines that do not need
  318  * multiple options of the same type.
  319  */
  320 int
  321 nd6_options(ndopts)
  322         union nd_opts *ndopts;
  323 {
  324         struct nd_opt_hdr *nd_opt;
  325         int i = 0;
  326 
  327         if (!ndopts)
  328                 panic("ndopts == NULL in nd6_options");
  329         if (!ndopts->nd_opts_last)
  330                 panic("uninitialized ndopts in nd6_options");
  331         if (!ndopts->nd_opts_search)
  332                 return 0;
  333 
  334         while (1) {
  335                 nd_opt = nd6_option(ndopts);
  336                 if (!nd_opt && !ndopts->nd_opts_last) {
  337                         /*
  338                          * Message validation requires that all included
  339                          * options have a length that is greater than zero.
  340                          */
  341                         icmp6stat.icp6s_nd_badopt++;
  342                         bzero(ndopts, sizeof(*ndopts));
  343                         return -1;
  344                 }
  345 
  346                 if (!nd_opt)
  347                         goto skip1;
  348 
  349                 switch (nd_opt->nd_opt_type) {
  350                 case ND_OPT_SOURCE_LINKADDR:
  351                 case ND_OPT_TARGET_LINKADDR:
  352                 case ND_OPT_MTU:
  353                 case ND_OPT_REDIRECTED_HEADER:
  354                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
  355                                 nd6log((LOG_INFO,
  356                                     "duplicated ND6 option found (type=%d)\n",
  357                                     nd_opt->nd_opt_type));
  358                                 /* XXX bark? */
  359                         } else {
  360                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  361                                         = nd_opt;
  362                         }
  363                         break;
  364                 case ND_OPT_PREFIX_INFORMATION:
  365                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
  366                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  367                                         = nd_opt;
  368                         }
  369                         ndopts->nd_opts_pi_end =
  370                                 (struct nd_opt_prefix_info *)nd_opt;
  371                         break;
  372                 default:
  373                         /*
  374                          * Unknown options must be silently ignored,
  375                          * to accomodate future extension to the protocol.
  376                          */
  377                         nd6log((LOG_DEBUG,
  378                             "nd6_options: unsupported option %d - "
  379                             "option ignored\n", nd_opt->nd_opt_type));
  380                 }
  381 
  382 skip1:
  383                 i++;
  384                 if (i > nd6_maxndopt) {
  385                         icmp6stat.icp6s_nd_toomanyopt++;
  386                         nd6log((LOG_INFO, "too many loop in nd opt\n"));
  387                         break;
  388                 }
  389 
  390                 if (ndopts->nd_opts_done)
  391                         break;
  392         }
  393 
  394         return 0;
  395 }
  396 
  397 /*
  398  * ND6 timer routine to expire default route list and prefix list
  399  */
  400 void
  401 nd6_timer(ignored_arg)
  402         void    *ignored_arg;
  403 {
  404         int s;
  405         struct llinfo_nd6 *ln;
  406         struct nd_defrouter *dr;
  407         struct nd_prefix *pr;
  408         struct ifnet *ifp;
  409         struct in6_ifaddr *ia6, *nia6;
  410         struct in6_addrlifetime *lt6;
  411 
  412         s = splnet();
  413         callout_reset(&nd6_timer_ch, nd6_prune * hz,
  414             nd6_timer, NULL);
  415 
  416         ln = llinfo_nd6.ln_next;
  417         while (ln && ln != &llinfo_nd6) {
  418                 struct rtentry *rt;
  419                 struct sockaddr_in6 *dst;
  420                 struct llinfo_nd6 *next = ln->ln_next;
  421                 /* XXX: used for the DELAY case only: */
  422                 struct nd_ifinfo *ndi = NULL;
  423 
  424                 if ((rt = ln->ln_rt) == NULL) {
  425                         ln = next;
  426                         continue;
  427                 }
  428                 if ((ifp = rt->rt_ifp) == NULL) {
  429                         ln = next;
  430                         continue;
  431                 }
  432                 ndi = ND_IFINFO(ifp);
  433                 dst = (struct sockaddr_in6 *)rt_key(rt);
  434 
  435                 if (ln->ln_expire > time_second) {
  436                         ln = next;
  437                         continue;
  438                 }
  439 
  440                 /* sanity check */
  441                 if (!rt)
  442                         panic("rt=0 in nd6_timer(ln=%p)", ln);
  443                 if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln)
  444                         panic("rt_llinfo(%p) is not equal to ln(%p)",
  445                               rt->rt_llinfo, ln);
  446                 if (!dst)
  447                         panic("dst=0 in nd6_timer(ln=%p)", ln);
  448 
  449                 switch (ln->ln_state) {
  450                 case ND6_LLINFO_INCOMPLETE:
  451                         if (ln->ln_asked < nd6_mmaxtries) {
  452                                 ln->ln_asked++;
  453                                 ln->ln_expire = time_second +
  454                                         ND_IFINFO(ifp)->retrans / 1000;
  455                                 nd6_ns_output(ifp, NULL, &dst->sin6_addr,
  456                                         ln, 0);
  457                         } else {
  458                                 struct mbuf *m = ln->ln_hold;
  459                                 if (m) {
  460                                         if (rt->rt_ifp) {
  461                                                 /*
  462                                                  * Fake rcvif to make ICMP error
  463                                                  * more helpful in diagnosing
  464                                                  * for the receiver.
  465                                                  * XXX: should we consider
  466                                                  * older rcvif?
  467                                                  */
  468                                                 m->m_pkthdr.rcvif = rt->rt_ifp;
  469                                         }
  470                                         icmp6_error(m, ICMP6_DST_UNREACH,
  471                                                     ICMP6_DST_UNREACH_ADDR, 0);
  472                                         ln->ln_hold = NULL;
  473                                 }
  474                                 next = nd6_free(rt);
  475                         }
  476                         break;
  477                 case ND6_LLINFO_REACHABLE:
  478                         if (ln->ln_expire) {
  479                                 ln->ln_state = ND6_LLINFO_STALE;
  480                                 ln->ln_expire = time_second + nd6_gctimer;
  481                         }
  482                         break;
  483 
  484                 case ND6_LLINFO_STALE:
  485                         /* Garbage Collection(RFC 2461 5.3) */
  486                         if (ln->ln_expire)
  487                                 next = nd6_free(rt);
  488                         break;
  489 
  490                 case ND6_LLINFO_DELAY:
  491                         if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
  492                                 /* We need NUD */
  493                                 ln->ln_asked = 1;
  494                                 ln->ln_state = ND6_LLINFO_PROBE;
  495                                 ln->ln_expire = time_second +
  496                                         ndi->retrans / 1000;
  497                                 nd6_ns_output(ifp, &dst->sin6_addr,
  498                                               &dst->sin6_addr,
  499                                               ln, 0);
  500                         } else {
  501                                 ln->ln_state = ND6_LLINFO_STALE; /* XXX */
  502                                 ln->ln_expire = time_second + nd6_gctimer;
  503                         }
  504                         break;
  505                 case ND6_LLINFO_PROBE:
  506                         if (ln->ln_asked < nd6_umaxtries) {
  507                                 ln->ln_asked++;
  508                                 ln->ln_expire = time_second +
  509                                         ND_IFINFO(ifp)->retrans / 1000;
  510                                 nd6_ns_output(ifp, &dst->sin6_addr,
  511                                                &dst->sin6_addr, ln, 0);
  512                         } else {
  513                                 next = nd6_free(rt);
  514                         }
  515                         break;
  516                 }
  517                 ln = next;
  518         }
  519 
  520         /* expire default router list */
  521         dr = TAILQ_FIRST(&nd_defrouter);
  522         while (dr) {
  523                 if (dr->expire && dr->expire < time_second) {
  524                         struct nd_defrouter *t;
  525                         t = TAILQ_NEXT(dr, dr_entry);
  526                         defrtrlist_del(dr);
  527                         dr = t;
  528                 } else {
  529                         dr = TAILQ_NEXT(dr, dr_entry);
  530                 }
  531         }
  532 
  533         /*
  534          * expire interface addresses.
  535          * in the past the loop was inside prefix expiry processing.
  536          * However, from a stricter speci-confrmance standpoint, we should
  537          * rather separate address lifetimes and prefix lifetimes.
  538          */
  539   addrloop:
  540         for (ia6 = in6_ifaddr; ia6; ia6 = nia6) {
  541                 nia6 = ia6->ia_next;
  542                 /* check address lifetime */
  543                 lt6 = &ia6->ia6_lifetime;
  544                 if (IFA6_IS_INVALID(ia6)) {
  545                         int regen = 0;
  546 
  547                         /*
  548                          * If the expiring address is temporary, try
  549                          * regenerating a new one.  This would be useful when
  550                          * we suspended a laptop PC, then turned it on after a
  551                          * period that could invalidate all temporary
  552                          * addresses.  Although we may have to restart the
  553                          * loop (see below), it must be after purging the
  554                          * address.  Otherwise, we'd see an infinite loop of
  555                          * regeneration.
  556                          */
  557                         if (ip6_use_tempaddr &&
  558                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
  559                                 if (regen_tmpaddr(ia6) == 0)
  560                                         regen = 1;
  561                         }
  562 
  563                         in6_purgeaddr(&ia6->ia_ifa);
  564 
  565                         if (regen)
  566                                 goto addrloop; /* XXX: see below */
  567                 }
  568                 if (IFA6_IS_DEPRECATED(ia6)) {
  569                         int oldflags = ia6->ia6_flags;
  570 
  571                         ia6->ia6_flags |= IN6_IFF_DEPRECATED;
  572 
  573                         /*
  574                          * If a temporary address has just become deprecated,
  575                          * regenerate a new one if possible.
  576                          */
  577                         if (ip6_use_tempaddr &&
  578                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
  579                             (oldflags & IN6_IFF_DEPRECATED) == 0) {
  580 
  581                                 if (regen_tmpaddr(ia6) == 0) {
  582                                         /*
  583                                          * A new temporary address is
  584                                          * generated.
  585                                          * XXX: this means the address chain
  586                                          * has changed while we are still in
  587                                          * the loop.  Although the change
  588                                          * would not cause disaster (because
  589                                          * it's not a deletion, but an
  590                                          * addition,) we'd rather restart the
  591                                          * loop just for safety.  Or does this
  592                                          * significantly reduce performance??
  593                                          */
  594                                         goto addrloop;
  595                                 }
  596                         }
  597                 } else {
  598                         /*
  599                          * A new RA might have made a deprecated address
  600                          * preferred.
  601                          */
  602                         ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
  603                 }
  604         }
  605 
  606         /* expire prefix list */
  607         pr = nd_prefix.lh_first;
  608         while (pr) {
  609                 /*
  610                  * check prefix lifetime.
  611                  * since pltime is just for autoconf, pltime processing for
  612                  * prefix is not necessary.
  613                  */
  614                 if (pr->ndpr_expire && pr->ndpr_expire < time_second) {
  615                         struct nd_prefix *t;
  616                         t = pr->ndpr_next;
  617 
  618                         /*
  619                          * address expiration and prefix expiration are
  620                          * separate.  NEVER perform in6_purgeaddr here.
  621                          */
  622 
  623                         prelist_remove(pr);
  624                         pr = t;
  625                 } else
  626                         pr = pr->ndpr_next;
  627         }
  628         splx(s);
  629 }
  630 
  631 static int
  632 regen_tmpaddr(ia6)
  633         struct in6_ifaddr *ia6; /* deprecated/invalidated temporary address */
  634 {
  635         struct ifaddr *ifa;
  636         struct ifnet *ifp;
  637         struct in6_ifaddr *public_ifa6 = NULL;
  638 
  639         ifp = ia6->ia_ifa.ifa_ifp;
  640         for (ifa = ifp->if_addrlist.tqh_first; ifa;
  641              ifa = ifa->ifa_list.tqe_next) {
  642                 struct in6_ifaddr *it6;
  643 
  644                 if (ifa->ifa_addr->sa_family != AF_INET6)
  645                         continue;
  646 
  647                 it6 = (struct in6_ifaddr *)ifa;
  648 
  649                 /* ignore no autoconf addresses. */
  650                 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
  651                         continue;
  652 
  653                 /* ignore autoconf addresses with different prefixes. */
  654                 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
  655                         continue;
  656 
  657                 /*
  658                  * Now we are looking at an autoconf address with the same
  659                  * prefix as ours.  If the address is temporary and is still
  660                  * preferred, do not create another one.  It would be rare, but
  661                  * could happen, for example, when we resume a laptop PC after
  662                  * a long period.
  663                  */
  664                 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
  665                     !IFA6_IS_DEPRECATED(it6)) {
  666                         public_ifa6 = NULL;
  667                         break;
  668                 }
  669 
  670                 /*
  671                  * This is a public autoconf address that has the same prefix
  672                  * as ours.  If it is preferred, keep it.  We can't break the
  673                  * loop here, because there may be a still-preferred temporary
  674                  * address with the prefix.
  675                  */
  676                 if (!IFA6_IS_DEPRECATED(it6))
  677                     public_ifa6 = it6;
  678         }
  679 
  680         if (public_ifa6 != NULL) {
  681                 int e;
  682 
  683                 if ((e = in6_tmpifadd(public_ifa6, 0)) != 0) {
  684                         log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
  685                             " tmp addr,errno=%d\n", e);
  686                         return (-1);
  687                 }
  688                 return (0);
  689         }
  690 
  691         return (-1);
  692 }
  693 
  694 /*
  695  * Nuke neighbor cache/prefix/default router management table, right before
  696  * ifp goes away.
  697  */
  698 void
  699 nd6_purge(ifp)
  700         struct ifnet *ifp;
  701 {
  702         struct llinfo_nd6 *ln, *nln;
  703         struct nd_defrouter *dr, *ndr, drany;
  704         struct nd_prefix *pr, *npr;
  705 
  706         /* Nuke default router list entries toward ifp */
  707         if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
  708                 /*
  709                  * The first entry of the list may be stored in
  710                  * the routing table, so we'll delete it later.
  711                  */
  712                 for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = ndr) {
  713                         ndr = TAILQ_NEXT(dr, dr_entry);
  714                         if (dr->ifp == ifp)
  715                                 defrtrlist_del(dr);
  716                 }
  717                 dr = TAILQ_FIRST(&nd_defrouter);
  718                 if (dr->ifp == ifp)
  719                         defrtrlist_del(dr);
  720         }
  721 
  722         /* Nuke prefix list entries toward ifp */
  723         for (pr = nd_prefix.lh_first; pr; pr = npr) {
  724                 npr = pr->ndpr_next;
  725                 if (pr->ndpr_ifp == ifp) {
  726                         /*
  727                          * Previously, pr->ndpr_addr is removed as well,
  728                          * but I strongly believe we don't have to do it.
  729                          * nd6_purge() is only called from in6_ifdetach(),
  730                          * which removes all the associated interface addresses
  731                          * by itself.
  732                          * (jinmei@kame.net 20010129)
  733                          */
  734                         prelist_remove(pr);
  735                 }
  736         }
  737 
  738         /* cancel default outgoing interface setting */
  739         if (nd6_defifindex == ifp->if_index)
  740                 nd6_setdefaultiface(0);
  741 
  742         if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
  743                 /* refresh default router list */
  744                 bzero(&drany, sizeof(drany));
  745                 defrouter_delreq(&drany, 0);
  746                 defrouter_select();
  747         }
  748 
  749         /*
  750          * Nuke neighbor cache entries for the ifp.
  751          * Note that rt->rt_ifp may not be the same as ifp,
  752          * due to KAME goto ours hack.  See RTM_RESOLVE case in
  753          * nd6_rtrequest(), and ip6_input().
  754          */
  755         ln = llinfo_nd6.ln_next;
  756         while (ln && ln != &llinfo_nd6) {
  757                 struct rtentry *rt;
  758                 struct sockaddr_dl *sdl;
  759 
  760                 nln = ln->ln_next;
  761                 rt = ln->ln_rt;
  762                 if (rt && rt->rt_gateway &&
  763                     rt->rt_gateway->sa_family == AF_LINK) {
  764                         sdl = (struct sockaddr_dl *)rt->rt_gateway;
  765                         if (sdl->sdl_index == ifp->if_index)
  766                                 nln = nd6_free(rt);
  767                 }
  768                 ln = nln;
  769         }
  770 }
  771 
  772 struct rtentry *
  773 nd6_lookup(addr6, create, ifp)
  774         struct in6_addr *addr6;
  775         int create;
  776         struct ifnet *ifp;
  777 {
  778         struct rtentry *rt;
  779         struct sockaddr_in6 sin6;
  780 
  781         bzero(&sin6, sizeof(sin6));
  782         sin6.sin6_len = sizeof(struct sockaddr_in6);
  783         sin6.sin6_family = AF_INET6;
  784         sin6.sin6_addr = *addr6;
  785         rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL);
  786         if (rt) {
  787                 if ((rt->rt_flags & RTF_LLINFO) == 0 && create) {
  788                         /*
  789                          * This is the case for the default route.
  790                          * If we want to create a neighbor cache for the
  791                          * address, we should free the route for the
  792                          * destination and allocate an interface route.
  793                          */
  794                         RTFREE_LOCKED(rt);
  795                         rt = 0;
  796                 }
  797         }
  798         if (!rt) {
  799                 if (create && ifp) {
  800                         int e;
  801 
  802                         /*
  803                          * If no route is available and create is set,
  804                          * we allocate a host route for the destination
  805                          * and treat it like an interface route.
  806                          * This hack is necessary for a neighbor which can't
  807                          * be covered by our own prefix.
  808                          */
  809                         struct ifaddr *ifa =
  810                             ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp);
  811                         if (ifa == NULL)
  812                                 return (NULL);
  813 
  814                         /*
  815                          * Create a new route.  RTF_LLINFO is necessary
  816                          * to create a Neighbor Cache entry for the
  817                          * destination in nd6_rtrequest which will be
  818                          * called in rtrequest via ifa->ifa_rtrequest.
  819                          */
  820                         if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6,
  821                             ifa->ifa_addr, (struct sockaddr *)&all1_sa,
  822                             (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) &
  823                             ~RTF_CLONING, &rt)) != 0) {
  824                                 log(LOG_ERR,
  825                                     "nd6_lookup: failed to add route for a "
  826                                     "neighbor(%s), errno=%d\n",
  827                                     ip6_sprintf(addr6), e);
  828                         }
  829                         if (rt == NULL)
  830                                 return (NULL);
  831                         RT_LOCK(rt);
  832                         if (rt->rt_llinfo) {
  833                                 struct llinfo_nd6 *ln =
  834                                     (struct llinfo_nd6 *)rt->rt_llinfo;
  835                                 ln->ln_state = ND6_LLINFO_NOSTATE;
  836                         }
  837                 } else
  838                         return (NULL);
  839         }
  840         RT_LOCK_ASSERT(rt);
  841         RT_REMREF(rt);
  842         /*
  843          * Validation for the entry.
  844          * Note that the check for rt_llinfo is necessary because a cloned
  845          * route from a parent route that has the L flag (e.g. the default
  846          * route to a p2p interface) may have the flag, too, while the
  847          * destination is not actually a neighbor.
  848          * XXX: we can't use rt->rt_ifp to check for the interface, since
  849          *      it might be the loopback interface if the entry is for our
  850          *      own address on a non-loopback interface. Instead, we should
  851          *      use rt->rt_ifa->ifa_ifp, which would specify the REAL
  852          *      interface.
  853          */
  854         if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 ||
  855             rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL ||
  856             (ifp && rt->rt_ifa->ifa_ifp != ifp)) {
  857                 if (create) {
  858                         log(LOG_DEBUG,
  859                             "nd6_lookup: failed to lookup %s (if = %s)\n",
  860                             ip6_sprintf(addr6),
  861                             ifp ? if_name(ifp) : "unspec");
  862                         /* xxx more logs... kazu */
  863                 }
  864                 RT_UNLOCK(rt);
  865                 return (NULL);
  866         }
  867         RT_UNLOCK(rt);          /* XXX not ready to return rt locked */
  868         return (rt);
  869 }
  870 
  871 /*
  872  * Detect if a given IPv6 address identifies a neighbor on a given link.
  873  * XXX: should take care of the destination of a p2p link?
  874  */
  875 int
  876 nd6_is_addr_neighbor(addr, ifp)
  877         struct sockaddr_in6 *addr;
  878         struct ifnet *ifp;
  879 {
  880         struct ifaddr *ifa;
  881         int i;
  882 
  883 #define IFADDR6(a) ((((struct in6_ifaddr *)(a))->ia_addr).sin6_addr)
  884 #define IFMASK6(a) ((((struct in6_ifaddr *)(a))->ia_prefixmask).sin6_addr)
  885 
  886         /*
  887          * A link-local address is always a neighbor.
  888          * XXX: we should use the sin6_scope_id field rather than the embedded
  889          * interface index.
  890          */
  891         if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr) &&
  892             ntohs(*(u_int16_t *)&addr->sin6_addr.s6_addr[2]) == ifp->if_index)
  893                 return (1);
  894 
  895         /*
  896          * If the address matches one of our addresses,
  897          * it should be a neighbor.
  898          */
  899         for (ifa = ifp->if_addrlist.tqh_first; ifa;
  900              ifa = ifa->ifa_list.tqe_next) {
  901                 if (ifa->ifa_addr->sa_family != AF_INET6)
  902                         next: continue;
  903 
  904                 for (i = 0; i < 4; i++) {
  905                         if ((IFADDR6(ifa).s6_addr32[i] ^
  906                              addr->sin6_addr.s6_addr32[i]) &
  907                             IFMASK6(ifa).s6_addr32[i])
  908                                 goto next;
  909                 }
  910                 return (1);
  911         }
  912 
  913         /*
  914          * Even if the address matches none of our addresses, it might be
  915          * in the neighbor cache.
  916          */
  917         if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL)
  918                 return (1);
  919 
  920         return (0);
  921 #undef IFADDR6
  922 #undef IFMASK6
  923 }
  924 
  925 /*
  926  * Free an nd6 llinfo entry.
  927  */
  928 struct llinfo_nd6 *
  929 nd6_free(rt)
  930         struct rtentry *rt;
  931 {
  932         struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next;
  933         struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr;
  934         struct nd_defrouter *dr;
  935 
  936         /*
  937          * we used to have pfctlinput(PRC_HOSTDEAD) here.
  938          * even though it is not harmful, it was not really necessary.
  939          */
  940 
  941         if (!ip6_forwarding && ip6_accept_rtadv) { /* XXX: too restrictive? */
  942                 int s;
  943                 s = splnet();
  944                 dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
  945                     rt->rt_ifp);
  946 
  947                 if (ln->ln_router || dr) {
  948                         /*
  949                          * rt6_flush must be called whether or not the neighbor
  950                          * is in the Default Router List.
  951                          * See a corresponding comment in nd6_na_input().
  952                          */
  953                         rt6_flush(&in6, rt->rt_ifp);
  954                 }
  955 
  956                 if (dr) {
  957                         /*
  958                          * Unreachablity of a router might affect the default
  959                          * router selection and on-link detection of advertised
  960                          * prefixes.
  961                          */
  962 
  963                         /*
  964                          * Temporarily fake the state to choose a new default
  965                          * router and to perform on-link determination of
  966                          * prefixes correctly.
  967                          * Below the state will be set correctly,
  968                          * or the entry itself will be deleted.
  969                          */
  970                         ln->ln_state = ND6_LLINFO_INCOMPLETE;
  971 
  972                         /*
  973                          * Since defrouter_select() does not affect the
  974                          * on-link determination and MIP6 needs the check
  975                          * before the default router selection, we perform
  976                          * the check now.
  977                          */
  978                         pfxlist_onlink_check();
  979 
  980                         if (dr == TAILQ_FIRST(&nd_defrouter)) {
  981                                 /*
  982                                  * It is used as the current default router,
  983                                  * so we have to move it to the end of the
  984                                  * list and choose a new one.
  985                                  * XXX: it is not very efficient if this is
  986                                  *      the only router.
  987                                  */
  988                                 TAILQ_REMOVE(&nd_defrouter, dr, dr_entry);
  989                                 TAILQ_INSERT_TAIL(&nd_defrouter, dr, dr_entry);
  990 
  991                                 defrouter_select();
  992                         }
  993                 }
  994                 splx(s);
  995         }
  996 
  997         /*
  998          * Before deleting the entry, remember the next entry as the
  999          * return value.  We need this because pfxlist_onlink_check() above
 1000          * might have freed other entries (particularly the old next entry) as
 1001          * a side effect (XXX).
 1002          */
 1003         next = ln->ln_next;
 1004 
 1005         /*
 1006          * Detach the route from the routing tree and the list of neighbor
 1007          * caches, and disable the route entry not to be used in already
 1008          * cached routes.
 1009          */
 1010         rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0,
 1011             rt_mask(rt), 0, (struct rtentry **)0);
 1012 
 1013         return (next);
 1014 }
 1015 
 1016 /*
 1017  * Upper-layer reachability hint for Neighbor Unreachability Detection.
 1018  *
 1019  * XXX cost-effective metods?
 1020  */
 1021 void
 1022 nd6_nud_hint(rt, dst6, force)
 1023         struct rtentry *rt;
 1024         struct in6_addr *dst6;
 1025         int force;
 1026 {
 1027         struct llinfo_nd6 *ln;
 1028 
 1029         /*
 1030          * If the caller specified "rt", use that.  Otherwise, resolve the
 1031          * routing table by supplied "dst6".
 1032          */
 1033         if (!rt) {
 1034                 if (!dst6)
 1035                         return;
 1036                 if (!(rt = nd6_lookup(dst6, 0, NULL)))
 1037                         return;
 1038         }
 1039 
 1040         if ((rt->rt_flags & RTF_GATEWAY) != 0 ||
 1041             (rt->rt_flags & RTF_LLINFO) == 0 ||
 1042             !rt->rt_llinfo || !rt->rt_gateway ||
 1043             rt->rt_gateway->sa_family != AF_LINK) {
 1044                 /* This is not a host route. */
 1045                 return;
 1046         }
 1047 
 1048         ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1049         if (ln->ln_state < ND6_LLINFO_REACHABLE)
 1050                 return;
 1051 
 1052         /*
 1053          * if we get upper-layer reachability confirmation many times,
 1054          * it is possible we have false information.
 1055          */
 1056         if (!force) {
 1057                 ln->ln_byhint++;
 1058                 if (ln->ln_byhint > nd6_maxnudhint)
 1059                         return;
 1060         }
 1061 
 1062         ln->ln_state = ND6_LLINFO_REACHABLE;
 1063         if (ln->ln_expire)
 1064                 ln->ln_expire = time_second +
 1065                         ND_IFINFO(rt->rt_ifp)->reachable;
 1066 }
 1067 
 1068 void
 1069 nd6_rtrequest(req, rt, info)
 1070         int     req;
 1071         struct rtentry *rt;
 1072         struct rt_addrinfo *info; /* xxx unused */
 1073 {
 1074         struct sockaddr *gate = rt->rt_gateway;
 1075         struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1076         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1077         struct ifnet *ifp = rt->rt_ifp;
 1078         struct ifaddr *ifa;
 1079 
 1080         RT_LOCK_ASSERT(rt);
 1081 
 1082         if ((rt->rt_flags & RTF_GATEWAY) != 0)
 1083                 return;
 1084 
 1085         if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) {
 1086                 /*
 1087                  * This is probably an interface direct route for a link
 1088                  * which does not need neighbor caches (e.g. fe80::%lo0/64).
 1089                  * We do not need special treatment below for such a route.
 1090                  * Moreover, the RTF_LLINFO flag which would be set below
 1091                  * would annoy the ndp(8) command.
 1092                  */
 1093                 return;
 1094         }
 1095 
 1096         if (req == RTM_RESOLVE &&
 1097             (nd6_need_cache(ifp) == 0 || /* stf case */
 1098              !nd6_is_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), ifp))) {
 1099                 /*
 1100                  * FreeBSD and BSD/OS often make a cloned host route based
 1101                  * on a less-specific route (e.g. the default route).
 1102                  * If the less specific route does not have a "gateway"
 1103                  * (this is the case when the route just goes to a p2p or an
 1104                  * stf interface), we'll mistakenly make a neighbor cache for
 1105                  * the host route, and will see strange neighbor solicitation
 1106                  * for the corresponding destination.  In order to avoid the
 1107                  * confusion, we check if the destination of the route is
 1108                  * a neighbor in terms of neighbor discovery, and stop the
 1109                  * process if not.  Additionally, we remove the LLINFO flag
 1110                  * so that ndp(8) will not try to get the neighbor information
 1111                  * of the destination.
 1112                  */
 1113                 rt->rt_flags &= ~RTF_LLINFO;
 1114                 return;
 1115         }
 1116 
 1117         switch (req) {
 1118         case RTM_ADD:
 1119                 /*
 1120                  * There is no backward compatibility :)
 1121                  *
 1122                  * if ((rt->rt_flags & RTF_HOST) == 0 &&
 1123                  *     SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff)
 1124                  *         rt->rt_flags |= RTF_CLONING;
 1125                  */
 1126                 if (rt->rt_flags & (RTF_CLONING | RTF_LLINFO)) {
 1127                         /*
 1128                          * Case 1: This route should come from
 1129                          * a route to interface.  RTF_LLINFO flag is set
 1130                          * for a host route whose destination should be
 1131                          * treated as on-link.
 1132                          */
 1133                         rt_setgate(rt, rt_key(rt),
 1134                                    (struct sockaddr *)&null_sdl);
 1135                         gate = rt->rt_gateway;
 1136                         SDL(gate)->sdl_type = ifp->if_type;
 1137                         SDL(gate)->sdl_index = ifp->if_index;
 1138                         if (ln)
 1139                                 ln->ln_expire = time_second;
 1140                         if (ln && ln->ln_expire == 0) {
 1141                                 /* kludge for desktops */
 1142                                 ln->ln_expire = 1;
 1143                         }
 1144                         if ((rt->rt_flags & RTF_CLONING) != 0)
 1145                                 break;
 1146                 }
 1147                 /*
 1148                  * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here.
 1149                  * We don't do that here since llinfo is not ready yet.
 1150                  *
 1151                  * There are also couple of other things to be discussed:
 1152                  * - unsolicited NA code needs improvement beforehand
 1153                  * - RFC2461 says we MAY send multicast unsolicited NA
 1154                  *   (7.2.6 paragraph 4), however, it also says that we
 1155                  *   SHOULD provide a mechanism to prevent multicast NA storm.
 1156                  *   we don't have anything like it right now.
 1157                  *   note that the mechanism needs a mutual agreement
 1158                  *   between proxies, which means that we need to implement
 1159                  *   a new protocol, or a new kludge.
 1160                  * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA.
 1161                  *   we need to check ip6forwarding before sending it.
 1162                  *   (or should we allow proxy ND configuration only for
 1163                  *   routers?  there's no mention about proxy ND from hosts)
 1164                  */
 1165 #if 0
 1166                 /* XXX it does not work */
 1167                 if (rt->rt_flags & RTF_ANNOUNCE)
 1168                         nd6_na_output(ifp,
 1169                               &SIN6(rt_key(rt))->sin6_addr,
 1170                               &SIN6(rt_key(rt))->sin6_addr,
 1171                               ip6_forwarding ? ND_NA_FLAG_ROUTER : 0,
 1172                               1, NULL);
 1173 #endif
 1174                 /* FALLTHROUGH */
 1175         case RTM_RESOLVE:
 1176                 if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
 1177                         /*
 1178                          * Address resolution isn't necessary for a point to
 1179                          * point link, so we can skip this test for a p2p link.
 1180                          */
 1181                         if (gate->sa_family != AF_LINK ||
 1182                             gate->sa_len < sizeof(null_sdl)) {
 1183                                 log(LOG_DEBUG,
 1184                                     "nd6_rtrequest: bad gateway value: %s\n",
 1185                                     if_name(ifp));
 1186                                 break;
 1187                         }
 1188                         SDL(gate)->sdl_type = ifp->if_type;
 1189                         SDL(gate)->sdl_index = ifp->if_index;
 1190                 }
 1191                 if (ln != NULL)
 1192                         break;  /* This happens on a route change */
 1193                 /*
 1194                  * Case 2: This route may come from cloning, or a manual route
 1195                  * add with a LL address.
 1196                  */
 1197                 R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln));
 1198                 rt->rt_llinfo = (caddr_t)ln;
 1199                 if (!ln) {
 1200                         log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n");
 1201                         break;
 1202                 }
 1203                 nd6_inuse++;
 1204                 nd6_allocated++;
 1205                 Bzero(ln, sizeof(*ln));
 1206                 ln->ln_rt = rt;
 1207                 /* this is required for "ndp" command. - shin */
 1208                 if (req == RTM_ADD) {
 1209                         /*
 1210                          * gate should have some valid AF_LINK entry,
 1211                          * and ln->ln_expire should have some lifetime
 1212                          * which is specified by ndp command.
 1213                          */
 1214                         ln->ln_state = ND6_LLINFO_REACHABLE;
 1215                         ln->ln_byhint = 0;
 1216                 } else {
 1217                         /*
 1218                          * When req == RTM_RESOLVE, rt is created and
 1219                          * initialized in rtrequest(), so rt_expire is 0.
 1220                          */
 1221                         ln->ln_state = ND6_LLINFO_NOSTATE;
 1222                         ln->ln_expire = time_second;
 1223                 }
 1224                 rt->rt_flags |= RTF_LLINFO;
 1225                 ln->ln_next = llinfo_nd6.ln_next;
 1226                 llinfo_nd6.ln_next = ln;
 1227                 ln->ln_prev = &llinfo_nd6;
 1228                 ln->ln_next->ln_prev = ln;
 1229 
 1230                 /*
 1231                  * check if rt_key(rt) is one of my address assigned
 1232                  * to the interface.
 1233                  */
 1234                 ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp,
 1235                     &SIN6(rt_key(rt))->sin6_addr);
 1236                 if (ifa) {
 1237                         caddr_t macp = nd6_ifptomac(ifp);
 1238                         ln->ln_expire = 0;
 1239                         ln->ln_state = ND6_LLINFO_REACHABLE;
 1240                         ln->ln_byhint = 0;
 1241                         if (macp) {
 1242                                 Bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen);
 1243                                 SDL(gate)->sdl_alen = ifp->if_addrlen;
 1244                         }
 1245                         if (nd6_useloopback) {
 1246                                 rt->rt_ifp = &loif[0];  /* XXX */
 1247                                 /*
 1248                                  * Make sure rt_ifa be equal to the ifaddr
 1249                                  * corresponding to the address.
 1250                                  * We need this because when we refer
 1251                                  * rt_ifa->ia6_flags in ip6_input, we assume
 1252                                  * that the rt_ifa points to the address instead
 1253                                  * of the loopback address.
 1254                                  */
 1255                                 if (ifa != rt->rt_ifa) {
 1256                                         IFAFREE(rt->rt_ifa);
 1257                                         IFAREF(ifa);
 1258                                         rt->rt_ifa = ifa;
 1259                                 }
 1260                         }
 1261                 } else if (rt->rt_flags & RTF_ANNOUNCE) {
 1262                         ln->ln_expire = 0;
 1263                         ln->ln_state = ND6_LLINFO_REACHABLE;
 1264                         ln->ln_byhint = 0;
 1265 
 1266                         /* join solicited node multicast for proxy ND */
 1267                         if (ifp->if_flags & IFF_MULTICAST) {
 1268                                 struct in6_addr llsol;
 1269                                 int error;
 1270 
 1271                                 llsol = SIN6(rt_key(rt))->sin6_addr;
 1272                                 llsol.s6_addr16[0] = htons(0xff02);
 1273                                 llsol.s6_addr16[1] = htons(ifp->if_index);
 1274                                 llsol.s6_addr32[1] = 0;
 1275                                 llsol.s6_addr32[2] = htonl(1);
 1276                                 llsol.s6_addr8[12] = 0xff;
 1277 
 1278                                 if (!in6_addmulti(&llsol, ifp, &error)) {
 1279                                         nd6log((LOG_ERR, "%s: failed to join "
 1280                                             "%s (errno=%d)\n", if_name(ifp),
 1281                                             ip6_sprintf(&llsol), error));
 1282                                 }
 1283                         }
 1284                 }
 1285                 break;
 1286 
 1287         case RTM_DELETE:
 1288                 if (!ln)
 1289                         break;
 1290                 /* leave from solicited node multicast for proxy ND */
 1291                 if ((rt->rt_flags & RTF_ANNOUNCE) != 0 &&
 1292                     (ifp->if_flags & IFF_MULTICAST) != 0) {
 1293                         struct in6_addr llsol;
 1294                         struct in6_multi *in6m;
 1295 
 1296                         llsol = SIN6(rt_key(rt))->sin6_addr;
 1297                         llsol.s6_addr16[0] = htons(0xff02);
 1298                         llsol.s6_addr16[1] = htons(ifp->if_index);
 1299                         llsol.s6_addr32[1] = 0;
 1300                         llsol.s6_addr32[2] = htonl(1);
 1301                         llsol.s6_addr8[12] = 0xff;
 1302 
 1303                         IN6_LOOKUP_MULTI(llsol, ifp, in6m);
 1304                         if (in6m)
 1305                                 in6_delmulti(in6m);
 1306                 }
 1307                 nd6_inuse--;
 1308                 ln->ln_next->ln_prev = ln->ln_prev;
 1309                 ln->ln_prev->ln_next = ln->ln_next;
 1310                 ln->ln_prev = NULL;
 1311                 rt->rt_llinfo = 0;
 1312                 rt->rt_flags &= ~RTF_LLINFO;
 1313                 if (ln->ln_hold)
 1314                         m_freem(ln->ln_hold);
 1315                 Free((caddr_t)ln);
 1316         }
 1317 }
 1318 
 1319 int
 1320 nd6_ioctl(cmd, data, ifp)
 1321         u_long cmd;
 1322         caddr_t data;
 1323         struct ifnet *ifp;
 1324 {
 1325         struct in6_drlist *drl = (struct in6_drlist *)data;
 1326         struct in6_oprlist *oprl = (struct in6_oprlist *)data;
 1327         struct in6_ndireq *ndi = (struct in6_ndireq *)data;
 1328         struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
 1329         struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
 1330         struct nd_defrouter *dr, any;
 1331         struct nd_prefix *pr;
 1332         struct rtentry *rt;
 1333         int i = 0, error = 0;
 1334         int s;
 1335 
 1336         switch (cmd) {
 1337         case SIOCGDRLST_IN6:
 1338                 /*
 1339                  * obsolete API, use sysctl under net.inet6.icmp6
 1340                  */
 1341                 bzero(drl, sizeof(*drl));
 1342                 s = splnet();
 1343                 dr = TAILQ_FIRST(&nd_defrouter);
 1344                 while (dr && i < DRLSTSIZ) {
 1345                         drl->defrouter[i].rtaddr = dr->rtaddr;
 1346                         in6_clearscope(&drl->defrouter[i].rtaddr);
 1347 
 1348                         drl->defrouter[i].flags = dr->flags;
 1349                         drl->defrouter[i].rtlifetime = dr->rtlifetime;
 1350                         drl->defrouter[i].expire = dr->expire;
 1351                         drl->defrouter[i].if_index = dr->ifp->if_index;
 1352                         i++;
 1353                         dr = TAILQ_NEXT(dr, dr_entry);
 1354                 }
 1355                 splx(s);
 1356                 break;
 1357         case SIOCGPRLST_IN6:
 1358                 /*
 1359                  * obsolete API, use sysctl under net.inet6.icmp6
 1360                  *
 1361                  * XXX the structure in6_prlist was changed in backward-
 1362                  * incompatible manner.  in6_oprlist is used for SIOCGPRLST_IN6,
 1363                  * in6_prlist is used for nd6_sysctl() - fill_prlist().
 1364                  */
 1365                 /*
 1366                  * XXX meaning of fields, especialy "raflags", is very
 1367                  * differnet between RA prefix list and RR/static prefix list.
 1368                  * how about separating ioctls into two?
 1369                  */
 1370                 bzero(oprl, sizeof(*oprl));
 1371                 s = splnet();
 1372                 pr = nd_prefix.lh_first;
 1373                 while (pr && i < PRLSTSIZ) {
 1374                         struct nd_pfxrouter *pfr;
 1375                         int j;
 1376 
 1377                         (void)in6_embedscope(&oprl->prefix[i].prefix,
 1378                             &pr->ndpr_prefix, NULL, NULL);
 1379                         oprl->prefix[i].raflags = pr->ndpr_raf;
 1380                         oprl->prefix[i].prefixlen = pr->ndpr_plen;
 1381                         oprl->prefix[i].vltime = pr->ndpr_vltime;
 1382                         oprl->prefix[i].pltime = pr->ndpr_pltime;
 1383                         oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
 1384                         oprl->prefix[i].expire = pr->ndpr_expire;
 1385 
 1386                         pfr = pr->ndpr_advrtrs.lh_first;
 1387                         j = 0;
 1388                         while (pfr) {
 1389                                 if (j < DRLSTSIZ) {
 1390 #define RTRADDR oprl->prefix[i].advrtr[j]
 1391                                         RTRADDR = pfr->router->rtaddr;
 1392                                         in6_clearscope(&RTRADDR);
 1393 #undef RTRADDR
 1394                                 }
 1395                                 j++;
 1396                                 pfr = pfr->pfr_next;
 1397                         }
 1398                         oprl->prefix[i].advrtrs = j;
 1399                         oprl->prefix[i].origin = PR_ORIG_RA;
 1400 
 1401                         i++;
 1402                         pr = pr->ndpr_next;
 1403                 }
 1404               {
 1405                 struct rr_prefix *rpp;
 1406 
 1407                 for (rpp = LIST_FIRST(&rr_prefix); rpp;
 1408                      rpp = LIST_NEXT(rpp, rp_entry)) {
 1409                         if (i >= PRLSTSIZ)
 1410                                 break;
 1411                         (void)in6_embedscope(&oprl->prefix[i].prefix,
 1412                             &pr->ndpr_prefix, NULL, NULL);
 1413                         oprl->prefix[i].raflags = rpp->rp_raf;
 1414                         oprl->prefix[i].prefixlen = rpp->rp_plen;
 1415                         oprl->prefix[i].vltime = rpp->rp_vltime;
 1416                         oprl->prefix[i].pltime = rpp->rp_pltime;
 1417                         oprl->prefix[i].if_index = rpp->rp_ifp->if_index;
 1418                         oprl->prefix[i].expire = rpp->rp_expire;
 1419                         oprl->prefix[i].advrtrs = 0;
 1420                         oprl->prefix[i].origin = rpp->rp_origin;
 1421                         i++;
 1422                 }
 1423               }
 1424                 splx(s);
 1425 
 1426                 break;
 1427         case OSIOCGIFINFO_IN6:
 1428                 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
 1429                 bzero(&ndi->ndi, sizeof(ndi->ndi));
 1430                 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
 1431                 ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu;
 1432                 ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable;
 1433                 ndi->ndi.reachable = ND_IFINFO(ifp)->reachable;
 1434                 ndi->ndi.retrans = ND_IFINFO(ifp)->retrans;
 1435                 ndi->ndi.flags = ND_IFINFO(ifp)->flags;
 1436                 ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm;
 1437                 ndi->ndi.chlim = ND_IFINFO(ifp)->chlim;
 1438                 break;
 1439         case SIOCGIFINFO_IN6:
 1440                 ndi->ndi = *ND_IFINFO(ifp);
 1441                 ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
 1442                 break;
 1443         case SIOCSIFINFO_FLAGS:
 1444                 ND_IFINFO(ifp)->flags = ndi->ndi.flags;
 1445                 break;
 1446         case SIOCSNDFLUSH_IN6:  /* XXX: the ioctl name is confusing... */
 1447                 /* flush default router list */
 1448                 /*
 1449                  * xxx sumikawa: should not delete route if default
 1450                  * route equals to the top of default router list
 1451                  */
 1452                 bzero(&any, sizeof(any));
 1453                 defrouter_delreq(&any, 0);
 1454                 defrouter_select();
 1455                 /* xxx sumikawa: flush prefix list */
 1456                 break;
 1457         case SIOCSPFXFLUSH_IN6:
 1458         {
 1459                 /* flush all the prefix advertised by routers */
 1460                 struct nd_prefix *pr, *next;
 1461 
 1462                 s = splnet();
 1463                 for (pr = nd_prefix.lh_first; pr; pr = next) {
 1464                         struct in6_ifaddr *ia, *ia_next;
 1465 
 1466                         next = pr->ndpr_next;
 1467 
 1468                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1469                                 continue; /* XXX */
 1470 
 1471                         /* do we really have to remove addresses as well? */
 1472                         for (ia = in6_ifaddr; ia; ia = ia_next) {
 1473                                 /* ia might be removed.  keep the next ptr. */
 1474                                 ia_next = ia->ia_next;
 1475 
 1476                                 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1477                                         continue;
 1478 
 1479                                 if (ia->ia6_ndpr == pr)
 1480                                         in6_purgeaddr(&ia->ia_ifa);
 1481                         }
 1482                         prelist_remove(pr);
 1483                 }
 1484                 splx(s);
 1485                 break;
 1486         }
 1487         case SIOCSRTRFLUSH_IN6:
 1488         {
 1489                 /* flush all the default routers */
 1490                 struct nd_defrouter *dr, *next;
 1491 
 1492                 s = splnet();
 1493                 if ((dr = TAILQ_FIRST(&nd_defrouter)) != NULL) {
 1494                         /*
 1495                          * The first entry of the list may be stored in
 1496                          * the routing table, so we'll delete it later.
 1497                          */
 1498                         for (dr = TAILQ_NEXT(dr, dr_entry); dr; dr = next) {
 1499                                 next = TAILQ_NEXT(dr, dr_entry);
 1500                                 defrtrlist_del(dr);
 1501                         }
 1502                         defrtrlist_del(TAILQ_FIRST(&nd_defrouter));
 1503                 }
 1504                 splx(s);
 1505                 break;
 1506         }
 1507         case SIOCGNBRINFO_IN6:
 1508         {
 1509                 struct llinfo_nd6 *ln;
 1510                 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
 1511 
 1512                 /*
 1513                  * XXX: KAME specific hack for scoped addresses
 1514                  *      XXXX: for other scopes than link-local?
 1515                  */
 1516                 if (IN6_IS_ADDR_LINKLOCAL(&nbi->addr) ||
 1517                     IN6_IS_ADDR_MC_LINKLOCAL(&nbi->addr)) {
 1518                         u_int16_t *idp = (u_int16_t *)&nb_addr.s6_addr[2];
 1519 
 1520                         if (*idp == 0)
 1521                                 *idp = htons(ifp->if_index);
 1522                 }
 1523 
 1524                 s = splnet();
 1525                 if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) {
 1526                         error = EINVAL;
 1527                         splx(s);
 1528                         break;
 1529                 }
 1530                 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1531                 nbi->state = ln->ln_state;
 1532                 nbi->asked = ln->ln_asked;
 1533                 nbi->isrouter = ln->ln_router;
 1534                 nbi->expire = ln->ln_expire;
 1535                 splx(s);
 1536 
 1537                 break;
 1538         }
 1539         case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1540                 ndif->ifindex = nd6_defifindex;
 1541                 break;
 1542         case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1543                 return (nd6_setdefaultiface(ndif->ifindex));
 1544         }
 1545         return (error);
 1546 }
 1547 
 1548 /*
 1549  * Create neighbor cache entry and cache link-layer address,
 1550  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
 1551  */
 1552 struct rtentry *
 1553 nd6_cache_lladdr(ifp, from, lladdr, lladdrlen, type, code)
 1554         struct ifnet *ifp;
 1555         struct in6_addr *from;
 1556         char *lladdr;
 1557         int lladdrlen;
 1558         int type;       /* ICMP6 type */
 1559         int code;       /* type dependent information */
 1560 {
 1561         struct rtentry *rt = NULL;
 1562         struct llinfo_nd6 *ln = NULL;
 1563         int is_newentry;
 1564         struct sockaddr_dl *sdl = NULL;
 1565         int do_update;
 1566         int olladdr;
 1567         int llchange;
 1568         int newstate = 0;
 1569 
 1570         if (!ifp)
 1571                 panic("ifp == NULL in nd6_cache_lladdr");
 1572         if (!from)
 1573                 panic("from == NULL in nd6_cache_lladdr");
 1574 
 1575         /* nothing must be updated for unspecified address */
 1576         if (IN6_IS_ADDR_UNSPECIFIED(from))
 1577                 return NULL;
 1578 
 1579         /*
 1580          * Validation about ifp->if_addrlen and lladdrlen must be done in
 1581          * the caller.
 1582          *
 1583          * XXX If the link does not have link-layer adderss, what should
 1584          * we do? (ifp->if_addrlen == 0)
 1585          * Spec says nothing in sections for RA, RS and NA.  There's small
 1586          * description on it in NS section (RFC 2461 7.2.3).
 1587          */
 1588 
 1589         rt = nd6_lookup(from, 0, ifp);
 1590         if (!rt) {
 1591 #if 0
 1592                 /* nothing must be done if there's no lladdr */
 1593                 if (!lladdr || !lladdrlen)
 1594                         return NULL;
 1595 #endif
 1596 
 1597                 rt = nd6_lookup(from, 1, ifp);
 1598                 is_newentry = 1;
 1599         } else {
 1600                 /* do nothing if static ndp is set */
 1601                 if (rt->rt_flags & RTF_STATIC)
 1602                         return NULL;
 1603                 is_newentry = 0;
 1604         }
 1605 
 1606         if (!rt)
 1607                 return NULL;
 1608         if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) {
 1609 fail:
 1610                 (void)nd6_free(rt);
 1611                 return NULL;
 1612         }
 1613         ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1614         if (!ln)
 1615                 goto fail;
 1616         if (!rt->rt_gateway)
 1617                 goto fail;
 1618         if (rt->rt_gateway->sa_family != AF_LINK)
 1619                 goto fail;
 1620         sdl = SDL(rt->rt_gateway);
 1621 
 1622         olladdr = (sdl->sdl_alen) ? 1 : 0;
 1623         if (olladdr && lladdr) {
 1624                 if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen))
 1625                         llchange = 1;
 1626                 else
 1627                         llchange = 0;
 1628         } else
 1629                 llchange = 0;
 1630 
 1631         /*
 1632          * newentry olladdr  lladdr  llchange   (*=record)
 1633          *      0       n       n       --      (1)
 1634          *      0       y       n       --      (2)
 1635          *      0       n       y       --      (3) * STALE
 1636          *      0       y       y       n       (4) *
 1637          *      0       y       y       y       (5) * STALE
 1638          *      1       --      n       --      (6)   NOSTATE(= PASSIVE)
 1639          *      1       --      y       --      (7) * STALE
 1640          */
 1641 
 1642         if (lladdr) {           /* (3-5) and (7) */
 1643                 /*
 1644                  * Record source link-layer address
 1645                  * XXX is it dependent to ifp->if_type?
 1646                  */
 1647                 sdl->sdl_alen = ifp->if_addrlen;
 1648                 bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen);
 1649         }
 1650 
 1651         if (!is_newentry) {
 1652                 if ((!olladdr && lladdr) ||             /* (3) */
 1653                     (olladdr && lladdr && llchange)) {  /* (5) */
 1654                         do_update = 1;
 1655                         newstate = ND6_LLINFO_STALE;
 1656                 } else                                  /* (1-2,4) */
 1657                         do_update = 0;
 1658         } else {
 1659                 do_update = 1;
 1660                 if (!lladdr)                            /* (6) */
 1661                         newstate = ND6_LLINFO_NOSTATE;
 1662                 else                                    /* (7) */
 1663                         newstate = ND6_LLINFO_STALE;
 1664         }
 1665 
 1666         if (do_update) {
 1667                 /*
 1668                  * Update the state of the neighbor cache.
 1669                  */
 1670                 ln->ln_state = newstate;
 1671 
 1672                 if (ln->ln_state == ND6_LLINFO_STALE) {
 1673                         /*
 1674                          * XXX: since nd6_output() below will cause
 1675                          * state tansition to DELAY and reset the timer,
 1676                          * we must set the timer now, although it is actually
 1677                          * meaningless.
 1678                          */
 1679                         ln->ln_expire = time_second + nd6_gctimer;
 1680 
 1681                         if (ln->ln_hold) {
 1682                                 /*
 1683                                  * we assume ifp is not a p2p here, so just
 1684                                  * set the 2nd argument as the 1st one.
 1685                                  */
 1686                                 nd6_output(ifp, ifp, ln->ln_hold,
 1687                                     (struct sockaddr_in6 *)rt_key(rt), rt);
 1688                                 ln->ln_hold = NULL;
 1689                         }
 1690                 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
 1691                         /* probe right away */
 1692                         ln->ln_expire = time_second;
 1693                 }
 1694         }
 1695 
 1696         /*
 1697          * ICMP6 type dependent behavior.
 1698          *
 1699          * NS: clear IsRouter if new entry
 1700          * RS: clear IsRouter
 1701          * RA: set IsRouter if there's lladdr
 1702          * redir: clear IsRouter if new entry
 1703          *
 1704          * RA case, (1):
 1705          * The spec says that we must set IsRouter in the following cases:
 1706          * - If lladdr exist, set IsRouter.  This means (1-5).
 1707          * - If it is old entry (!newentry), set IsRouter.  This means (7).
 1708          * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
 1709          * A quetion arises for (1) case.  (1) case has no lladdr in the
 1710          * neighbor cache, this is similar to (6).
 1711          * This case is rare but we figured that we MUST NOT set IsRouter.
 1712          *
 1713          * newentry olladdr  lladdr  llchange       NS  RS  RA  redir
 1714          *                                                      D R
 1715          *      0       n       n       --      (1)     c   ?     s
 1716          *      0       y       n       --      (2)     c   s     s
 1717          *      0       n       y       --      (3)     c   s     s
 1718          *      0       y       y       n       (4)     c   s     s
 1719          *      0       y       y       y       (5)     c   s     s
 1720          *      1       --      n       --      (6) c   c       c s
 1721          *      1       --      y       --      (7) c   c   s   c s
 1722          *
 1723          *                                      (c=clear s=set)
 1724          */
 1725         switch (type & 0xff) {
 1726         case ND_NEIGHBOR_SOLICIT:
 1727                 /*
 1728                  * New entry must have is_router flag cleared.
 1729                  */
 1730                 if (is_newentry)        /* (6-7) */
 1731                         ln->ln_router = 0;
 1732                 break;
 1733         case ND_REDIRECT:
 1734                 /*
 1735                  * If the icmp is a redirect to a better router, always set the
 1736                  * is_router flag.  Otherwise, if the entry is newly created,
 1737                  * clear the flag.  [RFC 2461, sec 8.3]
 1738                  */
 1739                 if (code == ND_REDIRECT_ROUTER)
 1740                         ln->ln_router = 1;
 1741                 else if (is_newentry) /* (6-7) */
 1742                         ln->ln_router = 0;
 1743                 break;
 1744         case ND_ROUTER_SOLICIT:
 1745                 /*
 1746                  * is_router flag must always be cleared.
 1747                  */
 1748                 ln->ln_router = 0;
 1749                 break;
 1750         case ND_ROUTER_ADVERT:
 1751                 /*
 1752                  * Mark an entry with lladdr as a router.
 1753                  */
 1754                 if ((!is_newentry && (olladdr || lladdr)) ||    /* (2-5) */
 1755                     (is_newentry && lladdr)) {                  /* (7) */
 1756                         ln->ln_router = 1;
 1757                 }
 1758                 break;
 1759         }
 1760 
 1761         /*
 1762          * When the link-layer address of a router changes, select the
 1763          * best router again.  In particular, when the neighbor entry is newly
 1764          * created, it might affect the selection policy.
 1765          * Question: can we restrict the first condition to the "is_newentry"
 1766          * case?
 1767          * XXX: when we hear an RA from a new router with the link-layer
 1768          * address option, defrouter_select() is called twice, since
 1769          * defrtrlist_update called the function as well.  However, I believe
 1770          * we can compromise the overhead, since it only happens the first
 1771          * time.
 1772          * XXX: although defrouter_select() should not have a bad effect
 1773          * for those are not autoconfigured hosts, we explicitly avoid such
 1774          * cases for safety.
 1775          */
 1776         if (do_update && ln->ln_router && !ip6_forwarding && ip6_accept_rtadv)
 1777                 defrouter_select();
 1778 
 1779         return rt;
 1780 }
 1781 
 1782 static void
 1783 nd6_slowtimo(ignored_arg)
 1784     void *ignored_arg;
 1785 {
 1786         int s = splnet();
 1787         struct nd_ifinfo *nd6if;
 1788         struct ifnet *ifp;
 1789 
 1790         callout_reset(&nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
 1791             nd6_slowtimo, NULL);
 1792         for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
 1793                 nd6if = ND_IFINFO(ifp);
 1794                 if (nd6if->basereachable && /* already initialized */
 1795                     (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
 1796                         /*
 1797                          * Since reachable time rarely changes by router
 1798                          * advertisements, we SHOULD insure that a new random
 1799                          * value gets recomputed at least once every few hours.
 1800                          * (RFC 2461, 6.3.4)
 1801                          */
 1802                         nd6if->recalctm = nd6_recalc_reachtm_interval;
 1803                         nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
 1804                 }
 1805         }
 1806         splx(s);
 1807 }
 1808 
 1809 #define senderr(e) { error = (e); goto bad;}
 1810 int
 1811 nd6_output(ifp, origifp, m0, dst, rt0)
 1812         struct ifnet *ifp;
 1813         struct ifnet *origifp;
 1814         struct mbuf *m0;
 1815         struct sockaddr_in6 *dst;
 1816         struct rtentry *rt0;
 1817 {
 1818         struct mbuf *m = m0;
 1819         struct rtentry *rt = rt0;
 1820         struct sockaddr_in6 *gw6 = NULL;
 1821         struct llinfo_nd6 *ln = NULL;
 1822         int error = 0;
 1823 
 1824         if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
 1825                 goto sendpkt;
 1826 
 1827         if (nd6_need_cache(ifp) == 0)
 1828                 goto sendpkt;
 1829 
 1830         /*
 1831          * next hop determination.  This routine is derived from ether_outpout.
 1832          */
 1833         if (rt) {
 1834                 if ((rt->rt_flags & RTF_UP) == 0) {
 1835                         rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL);
 1836                         if (rt != NULL) {
 1837                                 RT_REMREF(rt);
 1838                                 RT_UNLOCK(rt);
 1839                                 if (rt->rt_ifp != ifp) {
 1840                                         /* XXX: loop care? */
 1841                                         return nd6_output(ifp, origifp, m0,
 1842                                             dst, rt);
 1843                                 }
 1844                         } else
 1845                                 senderr(EHOSTUNREACH);
 1846                 }
 1847 
 1848                 if (rt->rt_flags & RTF_GATEWAY) {
 1849                         gw6 = (struct sockaddr_in6 *)rt->rt_gateway;
 1850 
 1851                         /*
 1852                          * We skip link-layer address resolution and NUD
 1853                          * if the gateway is not a neighbor from ND point
 1854                          * of view, regardless of the value of nd_ifinfo.flags.
 1855                          * The second condition is a bit tricky; we skip
 1856                          * if the gateway is our own address, which is
 1857                          * sometimes used to install a route to a p2p link.
 1858                          */
 1859                         if (!nd6_is_addr_neighbor(gw6, ifp) ||
 1860                             in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) {
 1861                                 /*
 1862                                  * We allow this kind of tricky route only
 1863                                  * when the outgoing interface is p2p.
 1864                                  * XXX: we may need a more generic rule here.
 1865                                  */
 1866                                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
 1867                                         senderr(EHOSTUNREACH);
 1868 
 1869                                 goto sendpkt;
 1870                         }
 1871 
 1872                         if (rt->rt_gwroute == 0)
 1873                                 goto lookup;
 1874                         if (((rt = rt->rt_gwroute)->rt_flags & RTF_UP) == 0) {
 1875                                 RT_LOCK(rt);
 1876                                 rtfree(rt); rt = rt0;
 1877                         lookup:
 1878                                 rt->rt_gwroute = rtalloc1(rt->rt_gateway, 1, 0UL);
 1879                                 if ((rt = rt->rt_gwroute) == 0)
 1880                                         senderr(EHOSTUNREACH);
 1881                                 RT_UNLOCK(rt);
 1882                         }
 1883                 }
 1884         }
 1885 
 1886         /*
 1887          * Address resolution or Neighbor Unreachability Detection
 1888          * for the next hop.
 1889          * At this point, the destination of the packet must be a unicast
 1890          * or an anycast address(i.e. not a multicast).
 1891          */
 1892 
 1893         /* Look up the neighbor cache for the nexthop */
 1894         if (rt && (rt->rt_flags & RTF_LLINFO) != 0)
 1895                 ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1896         else {
 1897                 /*
 1898                  * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
 1899                  * the condition below is not very efficient.  But we believe
 1900                  * it is tolerable, because this should be a rare case.
 1901                  */
 1902                 if (nd6_is_addr_neighbor(dst, ifp) &&
 1903                     (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL)
 1904                         ln = (struct llinfo_nd6 *)rt->rt_llinfo;
 1905         }
 1906         if (!ln || !rt) {
 1907                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
 1908                     !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
 1909                         log(LOG_DEBUG,
 1910                             "nd6_output: can't allocate llinfo for %s "
 1911                             "(ln=%p, rt=%p)\n",
 1912                             ip6_sprintf(&dst->sin6_addr), ln, rt);
 1913                         senderr(EIO);   /* XXX: good error? */
 1914                 }
 1915 
 1916                 goto sendpkt;   /* send anyway */
 1917         }
 1918 
 1919         /* We don't have to do link-layer address resolution on a p2p link. */
 1920         if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
 1921             ln->ln_state < ND6_LLINFO_REACHABLE) {
 1922                 ln->ln_state = ND6_LLINFO_STALE;
 1923                 ln->ln_expire = time_second + nd6_gctimer;
 1924         }
 1925 
 1926         /*
 1927          * The first time we send a packet to a neighbor whose entry is
 1928          * STALE, we have to change the state to DELAY and a sets a timer to
 1929          * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
 1930          * neighbor unreachability detection on expiration.
 1931          * (RFC 2461 7.3.3)
 1932          */
 1933         if (ln->ln_state == ND6_LLINFO_STALE) {
 1934                 ln->ln_asked = 0;
 1935                 ln->ln_state = ND6_LLINFO_DELAY;
 1936                 ln->ln_expire = time_second + nd6_delay;
 1937         }
 1938 
 1939         /*
 1940          * If the neighbor cache entry has a state other than INCOMPLETE
 1941          * (i.e. its link-layer address is already resolved), just
 1942          * send the packet.
 1943          */
 1944         if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
 1945                 goto sendpkt;
 1946 
 1947         /*
 1948          * There is a neighbor cache entry, but no ethernet address
 1949          * response yet.  Replace the held mbuf (if any) with this
 1950          * latest one.
 1951          *
 1952          * This code conforms to the rate-limiting rule described in Section
 1953          * 7.2.2 of RFC 2461, because the timer is set correctly after sending
 1954          * an NS below.
 1955          */
 1956         if (ln->ln_state == ND6_LLINFO_NOSTATE)
 1957                 ln->ln_state = ND6_LLINFO_INCOMPLETE;
 1958         if (ln->ln_hold)
 1959                 m_freem(ln->ln_hold);
 1960         ln->ln_hold = m;
 1961         if (ln->ln_expire) {
 1962                 if (ln->ln_asked < nd6_mmaxtries &&
 1963                     ln->ln_expire < time_second) {
 1964                         ln->ln_asked++;
 1965                         ln->ln_expire = time_second +
 1966                                 ND_IFINFO(ifp)->retrans / 1000;
 1967                         nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
 1968                 }
 1969         }
 1970         return (0);
 1971 
 1972   sendpkt:
 1973 #ifdef IPSEC
 1974         /* clean ipsec history once it goes out of the node */
 1975         ipsec_delaux(m);
 1976 #endif
 1977 
 1978 #ifdef MAC
 1979         mac_create_mbuf_linklayer(ifp, m);
 1980 #endif
 1981         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 1982                 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
 1983                     rt));
 1984         }
 1985         return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt));
 1986 
 1987   bad:
 1988         if (m)
 1989                 m_freem(m);
 1990         return (error);
 1991 }
 1992 #undef senderr
 1993 
 1994 int
 1995 nd6_need_cache(ifp)
 1996         struct ifnet *ifp;
 1997 {
 1998         /*
 1999          * XXX: we currently do not make neighbor cache on any interface
 2000          * other than ARCnet, Ethernet, FDDI and GIF.
 2001          *
 2002          * RFC2893 says:
 2003          * - unidirectional tunnels needs no ND
 2004          */
 2005         switch (ifp->if_type) {
 2006         case IFT_ARCNET:
 2007         case IFT_ETHER:
 2008         case IFT_FDDI:
 2009         case IFT_IEEE1394:
 2010 #ifdef IFT_L2VLAN
 2011         case IFT_L2VLAN:
 2012 #endif
 2013 #ifdef IFT_IEEE80211
 2014         case IFT_IEEE80211:
 2015 #endif
 2016         case IFT_GIF:           /* XXX need more cases? */
 2017                 return (1);
 2018         default:
 2019                 return (0);
 2020         }
 2021 }
 2022 
 2023 int
 2024 nd6_storelladdr(ifp, rt, m, dst, desten)
 2025         struct ifnet *ifp;
 2026         struct rtentry *rt;
 2027         struct mbuf *m;
 2028         struct sockaddr *dst;
 2029         u_char *desten;
 2030 {
 2031         int i;
 2032         struct sockaddr_dl *sdl;
 2033 
 2034         if (m->m_flags & M_MCAST) {
 2035                 switch (ifp->if_type) {
 2036                 case IFT_ETHER:
 2037                 case IFT_FDDI:
 2038 #ifdef IFT_L2VLAN
 2039         case IFT_L2VLAN:
 2040 #endif
 2041 #ifdef IFT_IEEE80211
 2042                 case IFT_IEEE80211:
 2043 #endif
 2044                 case IFT_ISO88025:
 2045                         ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
 2046                                                  desten);
 2047                         return (1);
 2048                 case IFT_IEEE1394:
 2049                         /*
 2050                          * netbsd can use if_broadcastaddr, but we don't do so
 2051                          * to reduce # of ifdef.
 2052                          */
 2053                         for (i = 0; i < ifp->if_addrlen; i++)
 2054                                 desten[i] = ~0;
 2055                         return (1);
 2056                 case IFT_ARCNET:
 2057                         *desten = 0;
 2058                         return (1);
 2059                 default:
 2060                         m_freem(m);
 2061                         return (0);
 2062                 }
 2063         }
 2064 
 2065         if (rt == NULL) {
 2066                 /* this could happen, if we could not allocate memory */
 2067                 m_freem(m);
 2068                 return (0);
 2069         }
 2070         if (rt->rt_gateway->sa_family != AF_LINK) {
 2071                 printf("nd6_storelladdr: something odd happens\n");
 2072                 m_freem(m);
 2073                 return (0);
 2074         }
 2075         sdl = SDL(rt->rt_gateway);
 2076         if (sdl->sdl_alen == 0) {
 2077                 /* this should be impossible, but we bark here for debugging */
 2078                 printf("nd6_storelladdr: sdl_alen == 0\n");
 2079                 m_freem(m);
 2080                 return (0);
 2081         }
 2082 
 2083         bcopy(LLADDR(sdl), desten, sdl->sdl_alen);
 2084         return (1);
 2085 }
 2086 
 2087 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
 2088 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
 2089 #ifdef SYSCTL_DECL
 2090 SYSCTL_DECL(_net_inet6_icmp6);
 2091 #endif
 2092 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
 2093         CTLFLAG_RD, nd6_sysctl_drlist, "");
 2094 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
 2095         CTLFLAG_RD, nd6_sysctl_prlist, "");
 2096 
 2097 static int
 2098 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
 2099 {
 2100         int error;
 2101         char buf[1024];
 2102         struct in6_defrouter *d, *de;
 2103         struct nd_defrouter *dr;
 2104 
 2105         if (req->newptr)
 2106                 return EPERM;
 2107         error = 0;
 2108 
 2109         for (dr = TAILQ_FIRST(&nd_defrouter); dr;
 2110              dr = TAILQ_NEXT(dr, dr_entry)) {
 2111                 d = (struct in6_defrouter *)buf;
 2112                 de = (struct in6_defrouter *)(buf + sizeof(buf));
 2113 
 2114                 if (d + 1 <= de) {
 2115                         bzero(d, sizeof(*d));
 2116                         d->rtaddr.sin6_family = AF_INET6;
 2117                         d->rtaddr.sin6_len = sizeof(d->rtaddr);
 2118                         if (in6_recoverscope(&d->rtaddr, &dr->rtaddr,
 2119                             dr->ifp) != 0)
 2120                                 log(LOG_ERR,
 2121                                     "scope error in "
 2122                                     "default router list (%s)\n",
 2123                                     ip6_sprintf(&dr->rtaddr));
 2124                         d->flags = dr->flags;
 2125                         d->rtlifetime = dr->rtlifetime;
 2126                         d->expire = dr->expire;
 2127                         d->if_index = dr->ifp->if_index;
 2128                 } else
 2129                         panic("buffer too short");
 2130 
 2131                 error = SYSCTL_OUT(req, buf, sizeof(*d));
 2132                 if (error)
 2133                         break;
 2134         }
 2135 
 2136         return (error);
 2137 }
 2138 
 2139 static int
 2140 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
 2141 {
 2142         int error;
 2143         char buf[1024];
 2144         struct in6_prefix *p, *pe;
 2145         struct nd_prefix *pr;
 2146 
 2147         if (req->newptr)
 2148                 return EPERM;
 2149         error = 0;
 2150 
 2151         for (pr = nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
 2152                 u_short advrtrs;
 2153                 size_t advance;
 2154                 struct sockaddr_in6 *sin6, *s6;
 2155                 struct nd_pfxrouter *pfr;
 2156 
 2157                 p = (struct in6_prefix *)buf;
 2158                 pe = (struct in6_prefix *)(buf + sizeof(buf));
 2159 
 2160                 if (p + 1 <= pe) {
 2161                         bzero(p, sizeof(*p));
 2162                         sin6 = (struct sockaddr_in6 *)(p + 1);
 2163 
 2164                         p->prefix = pr->ndpr_prefix;
 2165                         if (in6_recoverscope(&p->prefix,
 2166                             &p->prefix.sin6_addr, pr->ndpr_ifp) != 0)
 2167                                 log(LOG_ERR,
 2168                                     "scope error in prefix list (%s)\n",
 2169                                     ip6_sprintf(&p->prefix.sin6_addr));
 2170                         p->raflags = pr->ndpr_raf;
 2171                         p->prefixlen = pr->ndpr_plen;
 2172                         p->vltime = pr->ndpr_vltime;
 2173                         p->pltime = pr->ndpr_pltime;
 2174                         p->if_index = pr->ndpr_ifp->if_index;
 2175                         p->expire = pr->ndpr_expire;
 2176                         p->refcnt = pr->ndpr_refcnt;
 2177                         p->flags = pr->ndpr_stateflags;
 2178                         p->origin = PR_ORIG_RA;
 2179                         advrtrs = 0;
 2180                         for (pfr = pr->ndpr_advrtrs.lh_first; pfr;
 2181                              pfr = pfr->pfr_next) {
 2182                                 if ((void *)&sin6[advrtrs + 1] > (void *)pe) {
 2183                                         advrtrs++;
 2184                                         continue;
 2185                                 }
 2186                                 s6 = &sin6[advrtrs];
 2187                                 bzero(s6, sizeof(*s6));
 2188                                 s6->sin6_family = AF_INET6;
 2189                                 s6->sin6_len = sizeof(*sin6);
 2190                                 if (in6_recoverscope(s6, &pfr->router->rtaddr,
 2191                                     pfr->router->ifp) != 0)
 2192                                         log(LOG_ERR,
 2193                                             "scope error in "
 2194                                             "prefix list (%s)\n",
 2195                                             ip6_sprintf(&pfr->router->rtaddr));
 2196                                 advrtrs++;
 2197                         }
 2198                         p->advrtrs = advrtrs;
 2199                 } else
 2200                         panic("buffer too short");
 2201 
 2202                 advance = sizeof(*p) + sizeof(*sin6) * advrtrs;
 2203                 error = SYSCTL_OUT(req, buf, advance);
 2204                 if (error)
 2205                         break;
 2206         }
 2207 
 2208         return (error);
 2209 }

Cache object: d2ff5ea4b5e954d96eeaaf91c3af0840


[ 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.