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


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

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

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

    1 /*-
    2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
    3  * All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 3. Neither the name of the project nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD$");
   34 
   35 #include "opt_inet.h"
   36 #include "opt_inet6.h"
   37 
   38 #include <sys/param.h>
   39 #include <sys/systm.h>
   40 #include <sys/callout.h>
   41 #include <sys/malloc.h>
   42 #include <sys/mbuf.h>
   43 #include <sys/socket.h>
   44 #include <sys/sockio.h>
   45 #include <sys/time.h>
   46 #include <sys/kernel.h>
   47 #include <sys/protosw.h>
   48 #include <sys/errno.h>
   49 #include <sys/syslog.h>
   50 #include <sys/lock.h>
   51 #include <sys/rwlock.h>
   52 #include <sys/queue.h>
   53 #include <sys/sysctl.h>
   54 
   55 #include <net/if.h>
   56 #include <net/if_arc.h>
   57 #include <net/if_dl.h>
   58 #include <net/if_types.h>
   59 #include <net/iso88025.h>
   60 #include <net/fddi.h>
   61 #include <net/route.h>
   62 #include <net/vnet.h>
   63 
   64 #include <netinet/in.h>
   65 #include <net/if_llatbl.h>
   66 #define L3_ADDR_SIN6(le)        ((struct sockaddr_in6 *) L3_ADDR(le))
   67 #include <netinet/if_ether.h>
   68 #include <netinet6/in6_var.h>
   69 #include <netinet/ip6.h>
   70 #include <netinet6/ip6_var.h>
   71 #include <netinet6/scope6_var.h>
   72 #include <netinet6/nd6.h>
   73 #include <netinet/icmp6.h>
   74 
   75 #include <sys/limits.h>
   76 
   77 #include <security/mac/mac_framework.h>
   78 
   79 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
   80 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
   81 
   82 #define SIN6(s) ((struct sockaddr_in6 *)s)
   83 
   84 /* timer values */
   85 VNET_DEFINE(int, nd6_prune)     = 1;    /* walk list every 1 seconds */
   86 VNET_DEFINE(int, nd6_delay)     = 5;    /* delay first probe time 5 second */
   87 VNET_DEFINE(int, nd6_umaxtries) = 3;    /* maximum unicast query */
   88 VNET_DEFINE(int, nd6_mmaxtries) = 3;    /* maximum multicast query */
   89 VNET_DEFINE(int, nd6_useloopback) = 1;  /* use loopback interface for
   90                                          * local traffic */
   91 VNET_DEFINE(int, nd6_gctimer)   = (60 * 60 * 24); /* 1 day: garbage
   92                                          * collection timer */
   93 
   94 /* preventing too many loops in ND option parsing */
   95 static VNET_DEFINE(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
   96 
   97 VNET_DEFINE(int, nd6_maxnudhint) = 0;   /* max # of subsequent upper
   98                                          * layer hints */
   99 static VNET_DEFINE(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
  100                                          * ND entries */
  101 #define V_nd6_maxndopt                  VNET(nd6_maxndopt)
  102 #define V_nd6_maxqueuelen               VNET(nd6_maxqueuelen)
  103 
  104 #ifdef ND6_DEBUG
  105 VNET_DEFINE(int, nd6_debug) = 1;
  106 #else
  107 VNET_DEFINE(int, nd6_debug) = 0;
  108 #endif
  109 
  110 /* for debugging? */
  111 #if 0
  112 static int nd6_inuse, nd6_allocated;
  113 #endif
  114 
  115 VNET_DEFINE(struct nd_drhead, nd_defrouter);
  116 VNET_DEFINE(struct nd_prhead, nd_prefix);
  117 
  118 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
  119 #define V_nd6_recalc_reachtm_interval   VNET(nd6_recalc_reachtm_interval)
  120 
  121 static struct sockaddr_in6 all1_sa;
  122 
  123 static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *,
  124         struct ifnet *));
  125 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
  126 static void nd6_slowtimo(void *);
  127 static int regen_tmpaddr(struct in6_ifaddr *);
  128 static struct llentry *nd6_free(struct llentry *, int);
  129 static void nd6_llinfo_timer(void *);
  130 static void clear_llinfo_pqueue(struct llentry *);
  131 
  132 static VNET_DEFINE(struct callout, nd6_slowtimo_ch);
  133 #define V_nd6_slowtimo_ch               VNET(nd6_slowtimo_ch)
  134 
  135 VNET_DEFINE(struct callout, nd6_timer_ch);
  136 
  137 void
  138 nd6_init(void)
  139 {
  140         int i;
  141 
  142         LIST_INIT(&V_nd_prefix);
  143 
  144         all1_sa.sin6_family = AF_INET6;
  145         all1_sa.sin6_len = sizeof(struct sockaddr_in6);
  146         for (i = 0; i < sizeof(all1_sa.sin6_addr); i++)
  147                 all1_sa.sin6_addr.s6_addr[i] = 0xff;
  148 
  149         /* initialization of the default router list */
  150         TAILQ_INIT(&V_nd_defrouter);
  151 
  152         /* start timer */
  153         callout_init(&V_nd6_slowtimo_ch, 0);
  154         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
  155             nd6_slowtimo, curvnet);
  156 }
  157 
  158 #ifdef VIMAGE
  159 void
  160 nd6_destroy()
  161 {
  162 
  163         callout_drain(&V_nd6_slowtimo_ch);
  164         callout_drain(&V_nd6_timer_ch);
  165 }
  166 #endif
  167 
  168 struct nd_ifinfo *
  169 nd6_ifattach(struct ifnet *ifp)
  170 {
  171         struct nd_ifinfo *nd;
  172 
  173         nd = (struct nd_ifinfo *)malloc(sizeof(*nd), M_IP6NDP, M_WAITOK);
  174         bzero(nd, sizeof(*nd));
  175 
  176         nd->initialized = 1;
  177 
  178         nd->chlim = IPV6_DEFHLIM;
  179         nd->basereachable = REACHABLE_TIME;
  180         nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
  181         nd->retrans = RETRANS_TIMER;
  182         /*
  183          * Note that the default value of ip6_accept_rtadv is 0, which means
  184          * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
  185          * here.
  186          */
  187         nd->flags = (ND6_IFF_PERFORMNUD | ND6_IFF_ACCEPT_RTADV);
  188 
  189         /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
  190         nd6_setmtu0(ifp, nd);
  191 
  192         return nd;
  193 }
  194 
  195 void
  196 nd6_ifdetach(struct nd_ifinfo *nd)
  197 {
  198 
  199         free(nd, M_IP6NDP);
  200 }
  201 
  202 /*
  203  * Reset ND level link MTU. This function is called when the physical MTU
  204  * changes, which means we might have to adjust the ND level MTU.
  205  */
  206 void
  207 nd6_setmtu(struct ifnet *ifp)
  208 {
  209 
  210         nd6_setmtu0(ifp, ND_IFINFO(ifp));
  211 }
  212 
  213 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
  214 void
  215 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
  216 {
  217         u_int32_t omaxmtu;
  218 
  219         omaxmtu = ndi->maxmtu;
  220 
  221         switch (ifp->if_type) {
  222         case IFT_ARCNET:
  223                 ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
  224                 break;
  225         case IFT_FDDI:
  226                 ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
  227                 break;
  228         case IFT_ISO88025:
  229                  ndi->maxmtu = MIN(ISO88025_MAX_MTU, ifp->if_mtu);
  230                  break;
  231         default:
  232                 ndi->maxmtu = ifp->if_mtu;
  233                 break;
  234         }
  235 
  236         /*
  237          * Decreasing the interface MTU under IPV6 minimum MTU may cause
  238          * undesirable situation.  We thus notify the operator of the change
  239          * explicitly.  The check for omaxmtu is necessary to restrict the
  240          * log to the case of changing the MTU, not initializing it.
  241          */
  242         if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
  243                 log(LOG_NOTICE, "nd6_setmtu0: "
  244                     "new link MTU on %s (%lu) is too small for IPv6\n",
  245                     if_name(ifp), (unsigned long)ndi->maxmtu);
  246         }
  247 
  248         if (ndi->maxmtu > V_in6_maxmtu)
  249                 in6_setmaxmtu(); /* check all interfaces just in case */
  250 
  251 }
  252 
  253 void
  254 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
  255 {
  256 
  257         bzero(ndopts, sizeof(*ndopts));
  258         ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
  259         ndopts->nd_opts_last
  260                 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
  261 
  262         if (icmp6len == 0) {
  263                 ndopts->nd_opts_done = 1;
  264                 ndopts->nd_opts_search = NULL;
  265         }
  266 }
  267 
  268 /*
  269  * Take one ND option.
  270  */
  271 struct nd_opt_hdr *
  272 nd6_option(union nd_opts *ndopts)
  273 {
  274         struct nd_opt_hdr *nd_opt;
  275         int olen;
  276 
  277         if (ndopts == NULL)
  278                 panic("ndopts == NULL in nd6_option");
  279         if (ndopts->nd_opts_last == NULL)
  280                 panic("uninitialized ndopts in nd6_option");
  281         if (ndopts->nd_opts_search == NULL)
  282                 return NULL;
  283         if (ndopts->nd_opts_done)
  284                 return NULL;
  285 
  286         nd_opt = ndopts->nd_opts_search;
  287 
  288         /* make sure nd_opt_len is inside the buffer */
  289         if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
  290                 bzero(ndopts, sizeof(*ndopts));
  291                 return NULL;
  292         }
  293 
  294         olen = nd_opt->nd_opt_len << 3;
  295         if (olen == 0) {
  296                 /*
  297                  * Message validation requires that all included
  298                  * options have a length that is greater than zero.
  299                  */
  300                 bzero(ndopts, sizeof(*ndopts));
  301                 return NULL;
  302         }
  303 
  304         ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
  305         if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
  306                 /* option overruns the end of buffer, invalid */
  307                 bzero(ndopts, sizeof(*ndopts));
  308                 return NULL;
  309         } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
  310                 /* reached the end of options chain */
  311                 ndopts->nd_opts_done = 1;
  312                 ndopts->nd_opts_search = NULL;
  313         }
  314         return nd_opt;
  315 }
  316 
  317 /*
  318  * Parse multiple ND options.
  319  * This function is much easier to use, for ND routines that do not need
  320  * multiple options of the same type.
  321  */
  322 int
  323 nd6_options(union nd_opts *ndopts)
  324 {
  325         struct nd_opt_hdr *nd_opt;
  326         int i = 0;
  327 
  328         if (ndopts == NULL)
  329                 panic("ndopts == NULL in nd6_options");
  330         if (ndopts->nd_opts_last == NULL)
  331                 panic("uninitialized ndopts in nd6_options");
  332         if (ndopts->nd_opts_search == NULL)
  333                 return 0;
  334 
  335         while (1) {
  336                 nd_opt = nd6_option(ndopts);
  337                 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
  338                         /*
  339                          * Message validation requires that all included
  340                          * options have a length that is greater than zero.
  341                          */
  342                         ICMP6STAT_INC(icp6s_nd_badopt);
  343                         bzero(ndopts, sizeof(*ndopts));
  344                         return -1;
  345                 }
  346 
  347                 if (nd_opt == NULL)
  348                         goto skip1;
  349 
  350                 switch (nd_opt->nd_opt_type) {
  351                 case ND_OPT_SOURCE_LINKADDR:
  352                 case ND_OPT_TARGET_LINKADDR:
  353                 case ND_OPT_MTU:
  354                 case ND_OPT_REDIRECTED_HEADER:
  355                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
  356                                 nd6log((LOG_INFO,
  357                                     "duplicated ND6 option found (type=%d)\n",
  358                                     nd_opt->nd_opt_type));
  359                                 /* XXX bark? */
  360                         } else {
  361                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  362                                         = nd_opt;
  363                         }
  364                         break;
  365                 case ND_OPT_PREFIX_INFORMATION:
  366                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
  367                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  368                                         = nd_opt;
  369                         }
  370                         ndopts->nd_opts_pi_end =
  371                                 (struct nd_opt_prefix_info *)nd_opt;
  372                         break;
  373                 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
  374                 case ND_OPT_RDNSS:      /* RFC 6106 */
  375                 case ND_OPT_DNSSL:      /* RFC 6106 */
  376                         /*
  377                          * Silently ignore options we know and do not care about
  378                          * in the kernel.
  379                          */
  380                         break;
  381                 default:
  382                         /*
  383                          * Unknown options must be silently ignored,
  384                          * to accomodate future extension to the protocol.
  385                          */
  386                         nd6log((LOG_DEBUG,
  387                             "nd6_options: unsupported option %d - "
  388                             "option ignored\n", nd_opt->nd_opt_type));
  389                 }
  390 
  391 skip1:
  392                 i++;
  393                 if (i > V_nd6_maxndopt) {
  394                         ICMP6STAT_INC(icp6s_nd_toomanyopt);
  395                         nd6log((LOG_INFO, "too many loop in nd opt\n"));
  396                         break;
  397                 }
  398 
  399                 if (ndopts->nd_opts_done)
  400                         break;
  401         }
  402 
  403         return 0;
  404 }
  405 
  406 /*
  407  * ND6 timer routine to handle ND6 entries
  408  */
  409 void
  410 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
  411 {
  412         int canceled;
  413 
  414         LLE_WLOCK_ASSERT(ln);
  415 
  416         if (tick < 0) {
  417                 ln->la_expire = 0;
  418                 ln->ln_ntick = 0;
  419                 canceled = callout_stop(&ln->ln_timer_ch);
  420         } else {
  421                 ln->la_expire = time_second + tick / hz;
  422                 LLE_ADDREF(ln);
  423                 if (tick > INT_MAX) {
  424                         ln->ln_ntick = tick - INT_MAX;
  425                         canceled = callout_reset(&ln->ln_timer_ch, INT_MAX,
  426                             nd6_llinfo_timer, ln);
  427                 } else {
  428                         ln->ln_ntick = 0;
  429                         canceled = callout_reset(&ln->ln_timer_ch, tick,
  430                             nd6_llinfo_timer, ln);
  431                 }
  432         }
  433         if (canceled)
  434                 LLE_REMREF(ln);
  435 }
  436 
  437 void
  438 nd6_llinfo_settimer(struct llentry *ln, long tick)
  439 {
  440 
  441         LLE_WLOCK(ln);
  442         nd6_llinfo_settimer_locked(ln, tick);
  443         LLE_WUNLOCK(ln);
  444 }
  445 
  446 static void
  447 nd6_llinfo_timer(void *arg)
  448 {
  449         struct llentry *ln;
  450         struct in6_addr *dst;
  451         struct ifnet *ifp;
  452         struct nd_ifinfo *ndi = NULL;
  453 
  454         KASSERT(arg != NULL, ("%s: arg NULL", __func__));
  455         ln = (struct llentry *)arg;
  456         LLE_WLOCK_ASSERT(ln);
  457         ifp = ln->lle_tbl->llt_ifp;
  458 
  459         CURVNET_SET(ifp->if_vnet);
  460 
  461         if (ln->ln_ntick > 0) {
  462                 if (ln->ln_ntick > INT_MAX) {
  463                         ln->ln_ntick -= INT_MAX;
  464                         nd6_llinfo_settimer_locked(ln, INT_MAX);
  465                 } else {
  466                         ln->ln_ntick = 0;
  467                         nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
  468                 }
  469                 goto done;
  470         }
  471 
  472         ndi = ND_IFINFO(ifp);
  473         dst = &L3_ADDR_SIN6(ln)->sin6_addr;
  474         if (ln->la_flags & LLE_STATIC) {
  475                 goto done;
  476         }
  477 
  478         if (ln->la_flags & LLE_DELETED) {
  479                 (void)nd6_free(ln, 0);
  480                 ln = NULL;
  481                 goto done;
  482         }
  483 
  484         switch (ln->ln_state) {
  485         case ND6_LLINFO_INCOMPLETE:
  486                 if (ln->la_asked < V_nd6_mmaxtries) {
  487                         ln->la_asked++;
  488                         nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
  489                         LLE_WUNLOCK(ln);
  490                         nd6_ns_output(ifp, NULL, dst, ln, 0);
  491                         LLE_WLOCK(ln);
  492                 } else {
  493                         struct mbuf *m = ln->la_hold;
  494                         if (m) {
  495                                 struct mbuf *m0;
  496 
  497                                 /*
  498                                  * assuming every packet in la_hold has the
  499                                  * same IP header.  Send error after unlock.
  500                                  */
  501                                 m0 = m->m_nextpkt;
  502                                 m->m_nextpkt = NULL;
  503                                 ln->la_hold = m0;
  504                                 clear_llinfo_pqueue(ln);
  505                         }
  506                         (void)nd6_free(ln, 0);
  507                         ln = NULL;
  508                         if (m != NULL)
  509                                 icmp6_error2(m, ICMP6_DST_UNREACH,
  510                                     ICMP6_DST_UNREACH_ADDR, 0, ifp);
  511                 }
  512                 break;
  513         case ND6_LLINFO_REACHABLE:
  514                 if (!ND6_LLINFO_PERMANENT(ln)) {
  515                         ln->ln_state = ND6_LLINFO_STALE;
  516                         nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
  517                 }
  518                 break;
  519 
  520         case ND6_LLINFO_STALE:
  521                 /* Garbage Collection(RFC 2461 5.3) */
  522                 if (!ND6_LLINFO_PERMANENT(ln)) {
  523                         (void)nd6_free(ln, 1);
  524                         ln = NULL;
  525                 }
  526                 break;
  527 
  528         case ND6_LLINFO_DELAY:
  529                 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
  530                         /* We need NUD */
  531                         ln->la_asked = 1;
  532                         ln->ln_state = ND6_LLINFO_PROBE;
  533                         nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
  534                         LLE_WUNLOCK(ln);
  535                         nd6_ns_output(ifp, dst, dst, ln, 0);
  536                         LLE_WLOCK(ln);
  537                 } else {
  538                         ln->ln_state = ND6_LLINFO_STALE; /* XXX */
  539                         nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
  540                 }
  541                 break;
  542         case ND6_LLINFO_PROBE:
  543                 if (ln->la_asked < V_nd6_umaxtries) {
  544                         ln->la_asked++;
  545                         nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
  546                         LLE_WUNLOCK(ln);
  547                         nd6_ns_output(ifp, dst, dst, ln, 0);
  548                         LLE_WLOCK(ln);
  549                 } else {
  550                         (void)nd6_free(ln, 0);
  551                         ln = NULL;
  552                 }
  553                 break;
  554         default:
  555                 panic("%s: paths in a dark night can be confusing: %d",
  556                     __func__, ln->ln_state);
  557         }
  558 done:
  559         if (ln != NULL)
  560                 LLE_FREE_LOCKED(ln);
  561         CURVNET_RESTORE();
  562 }
  563 
  564 
  565 /*
  566  * ND6 timer routine to expire default route list and prefix list
  567  */
  568 void
  569 nd6_timer(void *arg)
  570 {
  571         CURVNET_SET((struct vnet *) arg);
  572         int s;
  573         struct nd_defrouter *dr, *ndr;
  574         struct nd_prefix *pr, *npr;
  575         struct in6_ifaddr *ia6, *nia6;
  576 
  577         callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
  578             nd6_timer, curvnet);
  579 
  580         /* expire default router list */
  581         s = splnet();
  582         TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
  583                 if (dr->expire && dr->expire < time_second)
  584                         defrtrlist_del(dr);
  585         }
  586 
  587         /*
  588          * expire interface addresses.
  589          * in the past the loop was inside prefix expiry processing.
  590          * However, from a stricter speci-confrmance standpoint, we should
  591          * rather separate address lifetimes and prefix lifetimes.
  592          *
  593          * XXXRW: in6_ifaddrhead locking.
  594          */
  595   addrloop:
  596         TAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
  597                 /* check address lifetime */
  598                 if (IFA6_IS_INVALID(ia6)) {
  599                         int regen = 0;
  600 
  601                         /*
  602                          * If the expiring address is temporary, try
  603                          * regenerating a new one.  This would be useful when
  604                          * we suspended a laptop PC, then turned it on after a
  605                          * period that could invalidate all temporary
  606                          * addresses.  Although we may have to restart the
  607                          * loop (see below), it must be after purging the
  608                          * address.  Otherwise, we'd see an infinite loop of
  609                          * regeneration.
  610                          */
  611                         if (V_ip6_use_tempaddr &&
  612                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
  613                                 if (regen_tmpaddr(ia6) == 0)
  614                                         regen = 1;
  615                         }
  616 
  617                         in6_purgeaddr(&ia6->ia_ifa);
  618 
  619                         if (regen)
  620                                 goto addrloop; /* XXX: see below */
  621                 } else if (IFA6_IS_DEPRECATED(ia6)) {
  622                         int oldflags = ia6->ia6_flags;
  623 
  624                         ia6->ia6_flags |= IN6_IFF_DEPRECATED;
  625 
  626                         /*
  627                          * If a temporary address has just become deprecated,
  628                          * regenerate a new one if possible.
  629                          */
  630                         if (V_ip6_use_tempaddr &&
  631                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
  632                             (oldflags & IN6_IFF_DEPRECATED) == 0) {
  633 
  634                                 if (regen_tmpaddr(ia6) == 0) {
  635                                         /*
  636                                          * A new temporary address is
  637                                          * generated.
  638                                          * XXX: this means the address chain
  639                                          * has changed while we are still in
  640                                          * the loop.  Although the change
  641                                          * would not cause disaster (because
  642                                          * it's not a deletion, but an
  643                                          * addition,) we'd rather restart the
  644                                          * loop just for safety.  Or does this
  645                                          * significantly reduce performance??
  646                                          */
  647                                         goto addrloop;
  648                                 }
  649                         }
  650                 } else {
  651                         /*
  652                          * A new RA might have made a deprecated address
  653                          * preferred.
  654                          */
  655                         ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
  656                 }
  657         }
  658 
  659         /* expire prefix list */
  660         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
  661                 /*
  662                  * check prefix lifetime.
  663                  * since pltime is just for autoconf, pltime processing for
  664                  * prefix is not necessary.
  665                  */
  666                 if (pr->ndpr_vltime != ND6_INFINITE_LIFETIME &&
  667                     time_second - pr->ndpr_lastupdate > pr->ndpr_vltime) {
  668 
  669                         /*
  670                          * address expiration and prefix expiration are
  671                          * separate.  NEVER perform in6_purgeaddr here.
  672                          */
  673                         prelist_remove(pr);
  674                 }
  675         }
  676         splx(s);
  677         CURVNET_RESTORE();
  678 }
  679 
  680 /*
  681  * ia6 - deprecated/invalidated temporary address
  682  */
  683 static int
  684 regen_tmpaddr(struct in6_ifaddr *ia6)
  685 {
  686         struct ifaddr *ifa;
  687         struct ifnet *ifp;
  688         struct in6_ifaddr *public_ifa6 = NULL;
  689 
  690         ifp = ia6->ia_ifa.ifa_ifp;
  691         IF_ADDR_RLOCK(ifp);
  692         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
  693                 struct in6_ifaddr *it6;
  694 
  695                 if (ifa->ifa_addr->sa_family != AF_INET6)
  696                         continue;
  697 
  698                 it6 = (struct in6_ifaddr *)ifa;
  699 
  700                 /* ignore no autoconf addresses. */
  701                 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
  702                         continue;
  703 
  704                 /* ignore autoconf addresses with different prefixes. */
  705                 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
  706                         continue;
  707 
  708                 /*
  709                  * Now we are looking at an autoconf address with the same
  710                  * prefix as ours.  If the address is temporary and is still
  711                  * preferred, do not create another one.  It would be rare, but
  712                  * could happen, for example, when we resume a laptop PC after
  713                  * a long period.
  714                  */
  715                 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
  716                     !IFA6_IS_DEPRECATED(it6)) {
  717                         public_ifa6 = NULL;
  718                         break;
  719                 }
  720 
  721                 /*
  722                  * This is a public autoconf address that has the same prefix
  723                  * as ours.  If it is preferred, keep it.  We can't break the
  724                  * loop here, because there may be a still-preferred temporary
  725                  * address with the prefix.
  726                  */
  727                 if (!IFA6_IS_DEPRECATED(it6))
  728                     public_ifa6 = it6;
  729 
  730                 if (public_ifa6 != NULL)
  731                         ifa_ref(&public_ifa6->ia_ifa);
  732         }
  733         IF_ADDR_RUNLOCK(ifp);
  734 
  735         if (public_ifa6 != NULL) {
  736                 int e;
  737 
  738                 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
  739                         ifa_free(&public_ifa6->ia_ifa);
  740                         log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
  741                             " tmp addr,errno=%d\n", e);
  742                         return (-1);
  743                 }
  744                 ifa_free(&public_ifa6->ia_ifa);
  745                 return (0);
  746         }
  747 
  748         return (-1);
  749 }
  750 
  751 /*
  752  * Nuke neighbor cache/prefix/default router management table, right before
  753  * ifp goes away.
  754  */
  755 void
  756 nd6_purge(struct ifnet *ifp)
  757 {
  758         struct nd_defrouter *dr, *ndr;
  759         struct nd_prefix *pr, *npr;
  760 
  761         /*
  762          * Nuke default router list entries toward ifp.
  763          * We defer removal of default router list entries that is installed
  764          * in the routing table, in order to keep additional side effects as
  765          * small as possible.
  766          */
  767         TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
  768                 if (dr->installed)
  769                         continue;
  770 
  771                 if (dr->ifp == ifp)
  772                         defrtrlist_del(dr);
  773         }
  774 
  775         TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, ndr) {
  776                 if (!dr->installed)
  777                         continue;
  778 
  779                 if (dr->ifp == ifp)
  780                         defrtrlist_del(dr);
  781         }
  782 
  783         /* Nuke prefix list entries toward ifp */
  784         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
  785                 if (pr->ndpr_ifp == ifp) {
  786                         /*
  787                          * Because if_detach() does *not* release prefixes
  788                          * while purging addresses the reference count will
  789                          * still be above zero. We therefore reset it to
  790                          * make sure that the prefix really gets purged.
  791                          */
  792                         pr->ndpr_refcnt = 0;
  793 
  794                         /*
  795                          * Previously, pr->ndpr_addr is removed as well,
  796                          * but I strongly believe we don't have to do it.
  797                          * nd6_purge() is only called from in6_ifdetach(),
  798                          * which removes all the associated interface addresses
  799                          * by itself.
  800                          * (jinmei@kame.net 20010129)
  801                          */
  802                         prelist_remove(pr);
  803                 }
  804         }
  805 
  806         /* cancel default outgoing interface setting */
  807         if (V_nd6_defifindex == ifp->if_index)
  808                 nd6_setdefaultiface(0);
  809 
  810         if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */
  811                 /* refresh default router list
  812                  *
  813                  * 
  814                  */
  815                 defrouter_select();
  816 
  817         }
  818 
  819         /* XXXXX
  820          * We do not nuke the neighbor cache entries here any more
  821          * because the neighbor cache is kept in if_afdata[AF_INET6].
  822          * nd6_purge() is invoked by in6_ifdetach() which is called
  823          * from if_detach() where everything gets purged. So let
  824          * in6_domifdetach() do the actual L2 table purging work.
  825          */
  826 }
  827 
  828 /* 
  829  * the caller acquires and releases the lock on the lltbls
  830  * Returns the llentry locked
  831  */
  832 struct llentry *
  833 nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp)
  834 {
  835         struct sockaddr_in6 sin6;
  836         struct llentry *ln;
  837         int llflags;
  838         
  839         bzero(&sin6, sizeof(sin6));
  840         sin6.sin6_len = sizeof(struct sockaddr_in6);
  841         sin6.sin6_family = AF_INET6;
  842         sin6.sin6_addr = *addr6;
  843 
  844         IF_AFDATA_LOCK_ASSERT(ifp);
  845 
  846         llflags = 0;
  847         if (flags & ND6_CREATE)
  848             llflags |= LLE_CREATE;
  849         if (flags & ND6_EXCLUSIVE)
  850             llflags |= LLE_EXCLUSIVE;   
  851         
  852         ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6);
  853         if ((ln != NULL) && (llflags & LLE_CREATE))
  854                 ln->ln_state = ND6_LLINFO_NOSTATE;
  855         
  856         return (ln);
  857 }
  858 
  859 /*
  860  * Test whether a given IPv6 address is a neighbor or not, ignoring
  861  * the actual neighbor cache.  The neighbor cache is ignored in order
  862  * to not reenter the routing code from within itself.
  863  */
  864 static int
  865 nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
  866 {
  867         struct nd_prefix *pr;
  868         struct ifaddr *dstaddr;
  869 
  870         /*
  871          * A link-local address is always a neighbor.
  872          * XXX: a link does not necessarily specify a single interface.
  873          */
  874         if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
  875                 struct sockaddr_in6 sin6_copy;
  876                 u_int32_t zone;
  877 
  878                 /*
  879                  * We need sin6_copy since sa6_recoverscope() may modify the
  880                  * content (XXX).
  881                  */
  882                 sin6_copy = *addr;
  883                 if (sa6_recoverscope(&sin6_copy))
  884                         return (0); /* XXX: should be impossible */
  885                 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
  886                         return (0);
  887                 if (sin6_copy.sin6_scope_id == zone)
  888                         return (1);
  889                 else
  890                         return (0);
  891         }
  892 
  893         /*
  894          * If the address matches one of our addresses,
  895          * it should be a neighbor.
  896          * If the address matches one of our on-link prefixes, it should be a
  897          * neighbor.
  898          */
  899         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
  900                 if (pr->ndpr_ifp != ifp)
  901                         continue;
  902 
  903                 if (!(pr->ndpr_stateflags & NDPRF_ONLINK)) {
  904                         struct rtentry *rt;
  905 
  906                         /* Always use the default FIB here. */
  907                         rt = in6_rtalloc1((struct sockaddr *)&pr->ndpr_prefix,
  908                             0, 0, RT_DEFAULT_FIB);
  909                         if (rt == NULL)
  910                                 continue;
  911                         /*
  912                          * This is the case where multiple interfaces
  913                          * have the same prefix, but only one is installed 
  914                          * into the routing table and that prefix entry
  915                          * is not the one being examined here. In the case
  916                          * where RADIX_MPATH is enabled, multiple route
  917                          * entries (of the same rt_key value) will be 
  918                          * installed because the interface addresses all
  919                          * differ.
  920                          */
  921                         if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
  922                                &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr)) {
  923                                 RTFREE_LOCKED(rt);
  924                                 continue;
  925                         }
  926                         RTFREE_LOCKED(rt);
  927                 }
  928 
  929                 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
  930                     &addr->sin6_addr, &pr->ndpr_mask))
  931                         return (1);
  932         }
  933 
  934         /*
  935          * If the address is assigned on the node of the other side of
  936          * a p2p interface, the address should be a neighbor.
  937          */
  938         dstaddr = ifa_ifwithdstaddr((struct sockaddr *)addr);
  939         if (dstaddr != NULL) {
  940                 if (dstaddr->ifa_ifp == ifp) {
  941                         ifa_free(dstaddr);
  942                         return (1);
  943                 }
  944                 ifa_free(dstaddr);
  945         }
  946 
  947         /*
  948          * If the default router list is empty, all addresses are regarded
  949          * as on-link, and thus, as a neighbor.
  950          * XXX: we restrict the condition to hosts, because routers usually do
  951          * not have the "default router list".
  952          */
  953         if (!V_ip6_forwarding && TAILQ_EMPTY(&V_nd_defrouter) &&
  954             V_nd6_defifindex == ifp->if_index) {
  955                 return (1);
  956         }
  957 
  958         return (0);
  959 }
  960 
  961 
  962 /*
  963  * Detect if a given IPv6 address identifies a neighbor on a given link.
  964  * XXX: should take care of the destination of a p2p link?
  965  */
  966 int
  967 nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp)
  968 {
  969         struct llentry *lle;
  970         int rc = 0;
  971 
  972         IF_AFDATA_UNLOCK_ASSERT(ifp);
  973         if (nd6_is_new_addr_neighbor(addr, ifp))
  974                 return (1);
  975 
  976         /*
  977          * Even if the address matches none of our addresses, it might be
  978          * in the neighbor cache.
  979          */
  980         IF_AFDATA_RLOCK(ifp);
  981         if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
  982                 LLE_RUNLOCK(lle);
  983                 rc = 1;
  984         }
  985         IF_AFDATA_RUNLOCK(ifp);
  986         return (rc);
  987 }
  988 
  989 /*
  990  * Free an nd6 llinfo entry.
  991  * Since the function would cause significant changes in the kernel, DO NOT
  992  * make it global, unless you have a strong reason for the change, and are sure
  993  * that the change is safe.
  994  */
  995 static struct llentry *
  996 nd6_free(struct llentry *ln, int gc)
  997 {
  998         struct llentry *next;
  999         struct nd_defrouter *dr;
 1000         struct ifnet *ifp;
 1001 
 1002         LLE_WLOCK_ASSERT(ln);
 1003 
 1004         /*
 1005          * we used to have pfctlinput(PRC_HOSTDEAD) here.
 1006          * even though it is not harmful, it was not really necessary.
 1007          */
 1008 
 1009         /* cancel timer */
 1010         nd6_llinfo_settimer_locked(ln, -1);
 1011 
 1012         ifp = ln->lle_tbl->llt_ifp;
 1013 
 1014         if (!V_ip6_forwarding) {
 1015 
 1016                 dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
 1017 
 1018                 if (dr != NULL && dr->expire &&
 1019                     ln->ln_state == ND6_LLINFO_STALE && gc) {
 1020                         /*
 1021                          * If the reason for the deletion is just garbage
 1022                          * collection, and the neighbor is an active default
 1023                          * router, do not delete it.  Instead, reset the GC
 1024                          * timer using the router's lifetime.
 1025                          * Simply deleting the entry would affect default
 1026                          * router selection, which is not necessarily a good
 1027                          * thing, especially when we're using router preference
 1028                          * values.
 1029                          * XXX: the check for ln_state would be redundant,
 1030                          *      but we intentionally keep it just in case.
 1031                          */
 1032                         if (dr->expire > time_second)
 1033                                 nd6_llinfo_settimer_locked(ln,
 1034                                     (dr->expire - time_second) * hz);
 1035                         else
 1036                                 nd6_llinfo_settimer_locked(ln,
 1037                                     (long)V_nd6_gctimer * hz);
 1038 
 1039                         next = LIST_NEXT(ln, lle_next);
 1040                         LLE_REMREF(ln);
 1041                         LLE_WUNLOCK(ln);
 1042                         return (next);
 1043                 }
 1044 
 1045                 if (dr) {
 1046                         /*
 1047                          * Unreachablity of a router might affect the default
 1048                          * router selection and on-link detection of advertised
 1049                          * prefixes.
 1050                          */
 1051 
 1052                         /*
 1053                          * Temporarily fake the state to choose a new default
 1054                          * router and to perform on-link determination of
 1055                          * prefixes correctly.
 1056                          * Below the state will be set correctly,
 1057                          * or the entry itself will be deleted.
 1058                          */
 1059                         ln->ln_state = ND6_LLINFO_INCOMPLETE;
 1060                 }
 1061 
 1062                 if (ln->ln_router || dr) {
 1063 
 1064                         /*
 1065                          * We need to unlock to avoid a LOR with rt6_flush() with the
 1066                          * rnh and for the calls to pfxlist_onlink_check() and
 1067                          * defrouter_select() in the block further down for calls
 1068                          * into nd6_lookup().  We still hold a ref.
 1069                          */
 1070                         LLE_WUNLOCK(ln);
 1071 
 1072                         /*
 1073                          * rt6_flush must be called whether or not the neighbor
 1074                          * is in the Default Router List.
 1075                          * See a corresponding comment in nd6_na_input().
 1076                          */
 1077                         rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ifp);
 1078                 }
 1079 
 1080                 if (dr) {
 1081                         /*
 1082                          * Since defrouter_select() does not affect the
 1083                          * on-link determination and MIP6 needs the check
 1084                          * before the default router selection, we perform
 1085                          * the check now.
 1086                          */
 1087                         pfxlist_onlink_check();
 1088 
 1089                         /*
 1090                          * Refresh default router list.
 1091                          */
 1092                         defrouter_select();
 1093                 }
 1094 
 1095                 if (ln->ln_router || dr)
 1096                         LLE_WLOCK(ln);
 1097         }
 1098 
 1099         /*
 1100          * Before deleting the entry, remember the next entry as the
 1101          * return value.  We need this because pfxlist_onlink_check() above
 1102          * might have freed other entries (particularly the old next entry) as
 1103          * a side effect (XXX).
 1104          */
 1105         next = LIST_NEXT(ln, lle_next);
 1106 
 1107         /*
 1108          * Save to unlock. We still hold an extra reference and will not
 1109          * free(9) in llentry_free() if someone else holds one as well.
 1110          */
 1111         LLE_WUNLOCK(ln);
 1112         IF_AFDATA_LOCK(ifp);
 1113         LLE_WLOCK(ln);
 1114         LLE_REMREF(ln);
 1115         llentry_free(ln);
 1116         IF_AFDATA_UNLOCK(ifp);
 1117 
 1118         return (next);
 1119 }
 1120 
 1121 /*
 1122  * Upper-layer reachability hint for Neighbor Unreachability Detection.
 1123  *
 1124  * XXX cost-effective methods?
 1125  */
 1126 void
 1127 nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force)
 1128 {
 1129         struct llentry *ln;
 1130         struct ifnet *ifp;
 1131 
 1132         if ((dst6 == NULL) || (rt == NULL))
 1133                 return;
 1134 
 1135         ifp = rt->rt_ifp;
 1136         IF_AFDATA_LOCK(ifp);
 1137         ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL);
 1138         IF_AFDATA_UNLOCK(ifp);
 1139         if (ln == NULL)
 1140                 return;
 1141 
 1142         if (ln->ln_state < ND6_LLINFO_REACHABLE)
 1143                 goto done;
 1144 
 1145         /*
 1146          * if we get upper-layer reachability confirmation many times,
 1147          * it is possible we have false information.
 1148          */
 1149         if (!force) {
 1150                 ln->ln_byhint++;
 1151                 if (ln->ln_byhint > V_nd6_maxnudhint) {
 1152                         goto done;
 1153                 }
 1154         }
 1155 
 1156         ln->ln_state = ND6_LLINFO_REACHABLE;
 1157         if (!ND6_LLINFO_PERMANENT(ln)) {
 1158                 nd6_llinfo_settimer_locked(ln,
 1159                     (long)ND_IFINFO(rt->rt_ifp)->reachable * hz);
 1160         }
 1161 done:
 1162         LLE_WUNLOCK(ln);
 1163 }
 1164 
 1165 
 1166 /*
 1167  * Rejuvenate this function for routing operations related
 1168  * processing.
 1169  */
 1170 void
 1171 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
 1172 {
 1173         struct sockaddr_in6 *gateway = (struct sockaddr_in6 *)rt->rt_gateway;
 1174         struct nd_defrouter *dr;
 1175         struct ifnet *ifp = rt->rt_ifp;
 1176 
 1177         RT_LOCK_ASSERT(rt);
 1178 
 1179         switch (req) {
 1180         case RTM_ADD:
 1181                 break;
 1182 
 1183         case RTM_DELETE:
 1184                 if (!ifp)
 1185                         return;
 1186                 /*
 1187                  * Only indirect routes are interesting.
 1188                  */
 1189                 if ((rt->rt_flags & RTF_GATEWAY) == 0)
 1190                         return;
 1191                 /*
 1192                  * check for default route
 1193                  */
 1194                 if (IN6_ARE_ADDR_EQUAL(&in6addr_any, 
 1195                                        &SIN6(rt_key(rt))->sin6_addr)) {
 1196 
 1197                         dr = defrouter_lookup(&gateway->sin6_addr, ifp);
 1198                         if (dr != NULL)
 1199                                 dr->installed = 0;
 1200                 }
 1201                 break;
 1202         }
 1203 }
 1204 
 1205 
 1206 int
 1207 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
 1208 {
 1209         struct in6_drlist *drl = (struct in6_drlist *)data;
 1210         struct in6_oprlist *oprl = (struct in6_oprlist *)data;
 1211         struct in6_ndireq *ndi = (struct in6_ndireq *)data;
 1212         struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
 1213         struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
 1214         struct nd_defrouter *dr;
 1215         struct nd_prefix *pr;
 1216         int i = 0, error = 0;
 1217         int s;
 1218 
 1219         switch (cmd) {
 1220         case SIOCGDRLST_IN6:
 1221                 /*
 1222                  * obsolete API, use sysctl under net.inet6.icmp6
 1223                  */
 1224                 bzero(drl, sizeof(*drl));
 1225                 s = splnet();
 1226                 TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
 1227                         if (i >= DRLSTSIZ)
 1228                                 break;
 1229                         drl->defrouter[i].rtaddr = dr->rtaddr;
 1230                         in6_clearscope(&drl->defrouter[i].rtaddr);
 1231 
 1232                         drl->defrouter[i].flags = dr->flags;
 1233                         drl->defrouter[i].rtlifetime = dr->rtlifetime;
 1234                         drl->defrouter[i].expire = dr->expire;
 1235                         drl->defrouter[i].if_index = dr->ifp->if_index;
 1236                         i++;
 1237                 }
 1238                 splx(s);
 1239                 break;
 1240         case SIOCGPRLST_IN6:
 1241                 /*
 1242                  * obsolete API, use sysctl under net.inet6.icmp6
 1243                  *
 1244                  * XXX the structure in6_prlist was changed in backward-
 1245                  * incompatible manner.  in6_oprlist is used for SIOCGPRLST_IN6,
 1246                  * in6_prlist is used for nd6_sysctl() - fill_prlist().
 1247                  */
 1248                 /*
 1249                  * XXX meaning of fields, especialy "raflags", is very
 1250                  * differnet between RA prefix list and RR/static prefix list.
 1251                  * how about separating ioctls into two?
 1252                  */
 1253                 bzero(oprl, sizeof(*oprl));
 1254                 s = splnet();
 1255                 LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 1256                         struct nd_pfxrouter *pfr;
 1257                         int j;
 1258 
 1259                         if (i >= PRLSTSIZ)
 1260                                 break;
 1261                         oprl->prefix[i].prefix = pr->ndpr_prefix.sin6_addr;
 1262                         oprl->prefix[i].raflags = pr->ndpr_raf;
 1263                         oprl->prefix[i].prefixlen = pr->ndpr_plen;
 1264                         oprl->prefix[i].vltime = pr->ndpr_vltime;
 1265                         oprl->prefix[i].pltime = pr->ndpr_pltime;
 1266                         oprl->prefix[i].if_index = pr->ndpr_ifp->if_index;
 1267                         if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
 1268                                 oprl->prefix[i].expire = 0;
 1269                         else {
 1270                                 time_t maxexpire;
 1271 
 1272                                 /* XXX: we assume time_t is signed. */
 1273                                 maxexpire = (-1) &
 1274                                     ~((time_t)1 <<
 1275                                     ((sizeof(maxexpire) * 8) - 1));
 1276                                 if (pr->ndpr_vltime <
 1277                                     maxexpire - pr->ndpr_lastupdate) {
 1278                                         oprl->prefix[i].expire =
 1279                                             pr->ndpr_lastupdate +
 1280                                             pr->ndpr_vltime;
 1281                                 } else
 1282                                         oprl->prefix[i].expire = maxexpire;
 1283                         }
 1284 
 1285                         j = 0;
 1286                         LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
 1287                                 if (j < DRLSTSIZ) {
 1288 #define RTRADDR oprl->prefix[i].advrtr[j]
 1289                                         RTRADDR = pfr->router->rtaddr;
 1290                                         in6_clearscope(&RTRADDR);
 1291 #undef RTRADDR
 1292                                 }
 1293                                 j++;
 1294                         }
 1295                         oprl->prefix[i].advrtrs = j;
 1296                         oprl->prefix[i].origin = PR_ORIG_RA;
 1297 
 1298                         i++;
 1299                 }
 1300                 splx(s);
 1301 
 1302                 break;
 1303         case OSIOCGIFINFO_IN6:
 1304 #define ND      ndi->ndi
 1305                 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
 1306                 bzero(&ND, sizeof(ND));
 1307                 ND.linkmtu = IN6_LINKMTU(ifp);
 1308                 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
 1309                 ND.basereachable = ND_IFINFO(ifp)->basereachable;
 1310                 ND.reachable = ND_IFINFO(ifp)->reachable;
 1311                 ND.retrans = ND_IFINFO(ifp)->retrans;
 1312                 ND.flags = ND_IFINFO(ifp)->flags;
 1313                 ND.recalctm = ND_IFINFO(ifp)->recalctm;
 1314                 ND.chlim = ND_IFINFO(ifp)->chlim;
 1315                 break;
 1316         case SIOCGIFINFO_IN6:
 1317                 ND = *ND_IFINFO(ifp);
 1318                 break;
 1319         case SIOCSIFINFO_IN6:
 1320                 /*
 1321                  * used to change host variables from userland.
 1322                  * intented for a use on router to reflect RA configurations.
 1323                  */
 1324                 /* 0 means 'unspecified' */
 1325                 if (ND.linkmtu != 0) {
 1326                         if (ND.linkmtu < IPV6_MMTU ||
 1327                             ND.linkmtu > IN6_LINKMTU(ifp)) {
 1328                                 error = EINVAL;
 1329                                 break;
 1330                         }
 1331                         ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
 1332                 }
 1333 
 1334                 if (ND.basereachable != 0) {
 1335                         int obasereachable = ND_IFINFO(ifp)->basereachable;
 1336 
 1337                         ND_IFINFO(ifp)->basereachable = ND.basereachable;
 1338                         if (ND.basereachable != obasereachable)
 1339                                 ND_IFINFO(ifp)->reachable =
 1340                                     ND_COMPUTE_RTIME(ND.basereachable);
 1341                 }
 1342                 if (ND.retrans != 0)
 1343                         ND_IFINFO(ifp)->retrans = ND.retrans;
 1344                 if (ND.chlim != 0)
 1345                         ND_IFINFO(ifp)->chlim = ND.chlim;
 1346                 /* FALLTHROUGH */
 1347         case SIOCSIFINFO_FLAGS:
 1348                 ND_IFINFO(ifp)->flags = ND.flags;
 1349                 break;
 1350 #undef ND
 1351         case SIOCSNDFLUSH_IN6:  /* XXX: the ioctl name is confusing... */
 1352                 /* sync kernel routing table with the default router list */
 1353                 defrouter_reset();
 1354                 defrouter_select();
 1355                 break;
 1356         case SIOCSPFXFLUSH_IN6:
 1357         {
 1358                 /* flush all the prefix advertised by routers */
 1359                 struct nd_prefix *pr, *next;
 1360 
 1361                 s = splnet();
 1362                 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
 1363                         struct in6_ifaddr *ia, *ia_next;
 1364 
 1365                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1366                                 continue; /* XXX */
 1367 
 1368                         /* do we really have to remove addresses as well? */
 1369                         /* XXXRW: in6_ifaddrhead locking. */
 1370                         TAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
 1371                             ia_next) {
 1372                                 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1373                                         continue;
 1374 
 1375                                 if (ia->ia6_ndpr == pr)
 1376                                         in6_purgeaddr(&ia->ia_ifa);
 1377                         }
 1378                         prelist_remove(pr);
 1379                 }
 1380                 splx(s);
 1381                 break;
 1382         }
 1383         case SIOCSRTRFLUSH_IN6:
 1384         {
 1385                 /* flush all the default routers */
 1386                 struct nd_defrouter *dr, *next;
 1387 
 1388                 s = splnet();
 1389                 defrouter_reset();
 1390                 TAILQ_FOREACH_SAFE(dr, &V_nd_defrouter, dr_entry, next) {
 1391                         defrtrlist_del(dr);
 1392                 }
 1393                 defrouter_select();
 1394                 splx(s);
 1395                 break;
 1396         }
 1397         case SIOCGNBRINFO_IN6:
 1398         {
 1399                 struct llentry *ln;
 1400                 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
 1401 
 1402                 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
 1403                         return (error);
 1404 
 1405                 IF_AFDATA_RLOCK(ifp);
 1406                 ln = nd6_lookup(&nb_addr, 0, ifp);
 1407                 IF_AFDATA_RUNLOCK(ifp);
 1408 
 1409                 if (ln == NULL) {
 1410                         error = EINVAL;
 1411                         break;
 1412                 }
 1413                 nbi->state = ln->ln_state;
 1414                 nbi->asked = ln->la_asked;
 1415                 nbi->isrouter = ln->ln_router;
 1416                 nbi->expire = ln->la_expire;
 1417                 LLE_RUNLOCK(ln);
 1418                 break;
 1419         }
 1420         case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1421                 ndif->ifindex = V_nd6_defifindex;
 1422                 break;
 1423         case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1424                 return (nd6_setdefaultiface(ndif->ifindex));
 1425         }
 1426         return (error);
 1427 }
 1428 
 1429 /*
 1430  * Create neighbor cache entry and cache link-layer address,
 1431  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
 1432  *
 1433  * type - ICMP6 type
 1434  * code - type dependent information
 1435  *
 1436  * XXXXX
 1437  *  The caller of this function already acquired the ndp 
 1438  *  cache table lock because the cache entry is returned.
 1439  */
 1440 struct llentry *
 1441 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 1442     int lladdrlen, int type, int code)
 1443 {
 1444         struct llentry *ln = NULL;
 1445         int is_newentry;
 1446         int do_update;
 1447         int olladdr;
 1448         int llchange;
 1449         int flags;
 1450         int newstate = 0;
 1451         uint16_t router = 0;
 1452         struct sockaddr_in6 sin6;
 1453         struct mbuf *chain = NULL;
 1454         int static_route = 0;
 1455 
 1456         IF_AFDATA_UNLOCK_ASSERT(ifp);
 1457 
 1458         if (ifp == NULL)
 1459                 panic("ifp == NULL in nd6_cache_lladdr");
 1460         if (from == NULL)
 1461                 panic("from == NULL in nd6_cache_lladdr");
 1462 
 1463         /* nothing must be updated for unspecified address */
 1464         if (IN6_IS_ADDR_UNSPECIFIED(from))
 1465                 return NULL;
 1466 
 1467         /*
 1468          * Validation about ifp->if_addrlen and lladdrlen must be done in
 1469          * the caller.
 1470          *
 1471          * XXX If the link does not have link-layer adderss, what should
 1472          * we do? (ifp->if_addrlen == 0)
 1473          * Spec says nothing in sections for RA, RS and NA.  There's small
 1474          * description on it in NS section (RFC 2461 7.2.3).
 1475          */
 1476         flags = lladdr ? ND6_EXCLUSIVE : 0;
 1477         IF_AFDATA_LOCK(ifp);
 1478         ln = nd6_lookup(from, flags, ifp);
 1479 
 1480         if (ln == NULL) {
 1481                 flags |= ND6_EXCLUSIVE;
 1482                 ln = nd6_lookup(from, flags | ND6_CREATE, ifp);
 1483                 IF_AFDATA_UNLOCK(ifp);
 1484                 is_newentry = 1;
 1485         } else {
 1486                 IF_AFDATA_UNLOCK(ifp);          
 1487                 /* do nothing if static ndp is set */
 1488                 if (ln->la_flags & LLE_STATIC) {
 1489                         static_route = 1;
 1490                         goto done;
 1491                 }
 1492                 is_newentry = 0;
 1493         }
 1494         if (ln == NULL)
 1495                 return (NULL);
 1496 
 1497         olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
 1498         if (olladdr && lladdr) {
 1499                 llchange = bcmp(lladdr, &ln->ll_addr,
 1500                     ifp->if_addrlen);
 1501         } else
 1502                 llchange = 0;
 1503 
 1504         /*
 1505          * newentry olladdr  lladdr  llchange   (*=record)
 1506          *      0       n       n       --      (1)
 1507          *      0       y       n       --      (2)
 1508          *      0       n       y       --      (3) * STALE
 1509          *      0       y       y       n       (4) *
 1510          *      0       y       y       y       (5) * STALE
 1511          *      1       --      n       --      (6)   NOSTATE(= PASSIVE)
 1512          *      1       --      y       --      (7) * STALE
 1513          */
 1514 
 1515         if (lladdr) {           /* (3-5) and (7) */
 1516                 /*
 1517                  * Record source link-layer address
 1518                  * XXX is it dependent to ifp->if_type?
 1519                  */
 1520                 bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen);
 1521                 ln->la_flags |= LLE_VALID;
 1522         }
 1523 
 1524         if (!is_newentry) {
 1525                 if ((!olladdr && lladdr != NULL) ||     /* (3) */
 1526                     (olladdr && lladdr != NULL && llchange)) {  /* (5) */
 1527                         do_update = 1;
 1528                         newstate = ND6_LLINFO_STALE;
 1529                 } else                                  /* (1-2,4) */
 1530                         do_update = 0;
 1531         } else {
 1532                 do_update = 1;
 1533                 if (lladdr == NULL)                     /* (6) */
 1534                         newstate = ND6_LLINFO_NOSTATE;
 1535                 else                                    /* (7) */
 1536                         newstate = ND6_LLINFO_STALE;
 1537         }
 1538 
 1539         if (do_update) {
 1540                 /*
 1541                  * Update the state of the neighbor cache.
 1542                  */
 1543                 ln->ln_state = newstate;
 1544 
 1545                 if (ln->ln_state == ND6_LLINFO_STALE) {
 1546                         /*
 1547                          * XXX: since nd6_output() below will cause
 1548                          * state tansition to DELAY and reset the timer,
 1549                          * we must set the timer now, although it is actually
 1550                          * meaningless.
 1551                          */
 1552                         nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
 1553 
 1554                         if (ln->la_hold) {
 1555                                 struct mbuf *m_hold, *m_hold_next;
 1556 
 1557                                 /*
 1558                                  * reset the la_hold in advance, to explicitly
 1559                                  * prevent a la_hold lookup in nd6_output()
 1560                                  * (wouldn't happen, though...)
 1561                                  */
 1562                                 for (m_hold = ln->la_hold, ln->la_hold = NULL;
 1563                                     m_hold; m_hold = m_hold_next) {
 1564                                         m_hold_next = m_hold->m_nextpkt;
 1565                                         m_hold->m_nextpkt = NULL;
 1566 
 1567                                         /*
 1568                                          * we assume ifp is not a p2p here, so
 1569                                          * just set the 2nd argument as the
 1570                                          * 1st one.
 1571                                          */
 1572                                         nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain);
 1573                                 }
 1574                                 /*
 1575                                  * If we have mbufs in the chain we need to do
 1576                                  * deferred transmit. Copy the address from the
 1577                                  * llentry before dropping the lock down below.
 1578                                  */
 1579                                 if (chain != NULL)
 1580                                         memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6));
 1581                         }
 1582                 } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) {
 1583                         /* probe right away */
 1584                         nd6_llinfo_settimer_locked((void *)ln, 0);
 1585                 }
 1586         }
 1587 
 1588         /*
 1589          * ICMP6 type dependent behavior.
 1590          *
 1591          * NS: clear IsRouter if new entry
 1592          * RS: clear IsRouter
 1593          * RA: set IsRouter if there's lladdr
 1594          * redir: clear IsRouter if new entry
 1595          *
 1596          * RA case, (1):
 1597          * The spec says that we must set IsRouter in the following cases:
 1598          * - If lladdr exist, set IsRouter.  This means (1-5).
 1599          * - If it is old entry (!newentry), set IsRouter.  This means (7).
 1600          * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
 1601          * A quetion arises for (1) case.  (1) case has no lladdr in the
 1602          * neighbor cache, this is similar to (6).
 1603          * This case is rare but we figured that we MUST NOT set IsRouter.
 1604          *
 1605          * newentry olladdr  lladdr  llchange       NS  RS  RA  redir
 1606          *                                                      D R
 1607          *      0       n       n       --      (1)     c   ?     s
 1608          *      0       y       n       --      (2)     c   s     s
 1609          *      0       n       y       --      (3)     c   s     s
 1610          *      0       y       y       n       (4)     c   s     s
 1611          *      0       y       y       y       (5)     c   s     s
 1612          *      1       --      n       --      (6) c   c       c s
 1613          *      1       --      y       --      (7) c   c   s   c s
 1614          *
 1615          *                                      (c=clear s=set)
 1616          */
 1617         switch (type & 0xff) {
 1618         case ND_NEIGHBOR_SOLICIT:
 1619                 /*
 1620                  * New entry must have is_router flag cleared.
 1621                  */
 1622                 if (is_newentry)        /* (6-7) */
 1623                         ln->ln_router = 0;
 1624                 break;
 1625         case ND_REDIRECT:
 1626                 /*
 1627                  * If the icmp is a redirect to a better router, always set the
 1628                  * is_router flag.  Otherwise, if the entry is newly created,
 1629                  * clear the flag.  [RFC 2461, sec 8.3]
 1630                  */
 1631                 if (code == ND_REDIRECT_ROUTER)
 1632                         ln->ln_router = 1;
 1633                 else if (is_newentry) /* (6-7) */
 1634                         ln->ln_router = 0;
 1635                 break;
 1636         case ND_ROUTER_SOLICIT:
 1637                 /*
 1638                  * is_router flag must always be cleared.
 1639                  */
 1640                 ln->ln_router = 0;
 1641                 break;
 1642         case ND_ROUTER_ADVERT:
 1643                 /*
 1644                  * Mark an entry with lladdr as a router.
 1645                  */
 1646                 if ((!is_newentry && (olladdr || lladdr)) ||    /* (2-5) */
 1647                     (is_newentry && lladdr)) {                  /* (7) */
 1648                         ln->ln_router = 1;
 1649                 }
 1650                 break;
 1651         }
 1652 
 1653         if (ln != NULL) {
 1654                 static_route = (ln->la_flags & LLE_STATIC);
 1655                 router = ln->ln_router;
 1656 
 1657                 if (flags & ND6_EXCLUSIVE)
 1658                         LLE_WUNLOCK(ln);
 1659                 else
 1660                         LLE_RUNLOCK(ln);
 1661                 if (static_route)
 1662                         ln = NULL;
 1663         }
 1664         if (chain)
 1665                 nd6_output_flush(ifp, ifp, chain, &sin6, NULL);
 1666         
 1667         /*
 1668          * When the link-layer address of a router changes, select the
 1669          * best router again.  In particular, when the neighbor entry is newly
 1670          * created, it might affect the selection policy.
 1671          * Question: can we restrict the first condition to the "is_newentry"
 1672          * case?
 1673          * XXX: when we hear an RA from a new router with the link-layer
 1674          * address option, defrouter_select() is called twice, since
 1675          * defrtrlist_update called the function as well.  However, I believe
 1676          * we can compromise the overhead, since it only happens the first
 1677          * time.
 1678          * XXX: although defrouter_select() should not have a bad effect
 1679          * for those are not autoconfigured hosts, we explicitly avoid such
 1680          * cases for safety.
 1681          */
 1682         if (do_update && router && !V_ip6_forwarding && V_ip6_accept_rtadv) {
 1683                 /*
 1684                  * guaranteed recursion
 1685                  */
 1686                 defrouter_select();
 1687         }
 1688         
 1689         return (ln);
 1690 done:   
 1691         if (ln != NULL) {
 1692                 if (flags & ND6_EXCLUSIVE)
 1693                         LLE_WUNLOCK(ln);
 1694                 else
 1695                         LLE_RUNLOCK(ln);
 1696                 if (static_route)
 1697                         ln = NULL;
 1698         }
 1699         return (ln);
 1700 }
 1701 
 1702 static void
 1703 nd6_slowtimo(void *arg)
 1704 {
 1705         CURVNET_SET((struct vnet *) arg);
 1706         struct nd_ifinfo *nd6if;
 1707         struct ifnet *ifp;
 1708 
 1709         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
 1710             nd6_slowtimo, curvnet);
 1711         IFNET_RLOCK_NOSLEEP();
 1712         TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
 1713                 nd6if = ND_IFINFO(ifp);
 1714                 if (nd6if->basereachable && /* already initialized */
 1715                     (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
 1716                         /*
 1717                          * Since reachable time rarely changes by router
 1718                          * advertisements, we SHOULD insure that a new random
 1719                          * value gets recomputed at least once every few hours.
 1720                          * (RFC 2461, 6.3.4)
 1721                          */
 1722                         nd6if->recalctm = V_nd6_recalc_reachtm_interval;
 1723                         nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
 1724                 }
 1725         }
 1726         IFNET_RUNLOCK_NOSLEEP();
 1727         CURVNET_RESTORE();
 1728 }
 1729 
 1730 int
 1731 nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
 1732     struct sockaddr_in6 *dst, struct rtentry *rt0)
 1733 {
 1734 
 1735         return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL));
 1736 }
 1737 
 1738 
 1739 /*
 1740  * Note that I'm not enforcing any global serialization
 1741  * lle state or asked changes here as the logic is too
 1742  * complicated to avoid having to always acquire an exclusive
 1743  * lock
 1744  * KMM
 1745  *
 1746  */
 1747 #define senderr(e) { error = (e); goto bad;}
 1748 
 1749 int
 1750 nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0,
 1751     struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle,
 1752         struct mbuf **chain)
 1753 {
 1754         struct mbuf *m = m0;
 1755         struct llentry *ln = lle;
 1756         int error = 0;
 1757         int flags = 0;
 1758 
 1759 #ifdef INVARIANTS
 1760         if (lle != NULL) {
 1761                 
 1762                 LLE_WLOCK_ASSERT(lle);
 1763 
 1764                 KASSERT(chain != NULL, (" lle locked but no mbuf chain pointer passed"));
 1765         }
 1766 #endif
 1767         if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr))
 1768                 goto sendpkt;
 1769 
 1770         if (nd6_need_cache(ifp) == 0)
 1771                 goto sendpkt;
 1772 
 1773         /*
 1774          * next hop determination.  This routine is derived from ether_output.
 1775          */
 1776 
 1777         /*
 1778          * Address resolution or Neighbor Unreachability Detection
 1779          * for the next hop.
 1780          * At this point, the destination of the packet must be a unicast
 1781          * or an anycast address(i.e. not a multicast).
 1782          */
 1783 
 1784         flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0;
 1785         if (ln == NULL) {
 1786         retry:
 1787                 IF_AFDATA_LOCK(ifp);
 1788                 ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst);
 1789                 IF_AFDATA_UNLOCK(ifp);
 1790                 if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp))  {
 1791                         /*
 1792                          * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
 1793                          * the condition below is not very efficient.  But we believe
 1794                          * it is tolerable, because this should be a rare case.
 1795                          */
 1796                         flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0);
 1797                         IF_AFDATA_LOCK(ifp);
 1798                         ln = nd6_lookup(&dst->sin6_addr, flags, ifp);
 1799                         IF_AFDATA_UNLOCK(ifp);
 1800                 }
 1801         } 
 1802         if (ln == NULL) {
 1803                 if ((ifp->if_flags & IFF_POINTOPOINT) == 0 &&
 1804                     !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) {
 1805                         char ip6buf[INET6_ADDRSTRLEN];
 1806                         log(LOG_DEBUG,
 1807                             "nd6_output: can't allocate llinfo for %s "
 1808                             "(ln=%p)\n",
 1809                             ip6_sprintf(ip6buf, &dst->sin6_addr), ln);
 1810                         senderr(EIO);   /* XXX: good error? */
 1811                 }
 1812                 goto sendpkt;   /* send anyway */
 1813         }
 1814 
 1815         /* We don't have to do link-layer address resolution on a p2p link. */
 1816         if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
 1817             ln->ln_state < ND6_LLINFO_REACHABLE) {
 1818                 if ((flags & LLE_EXCLUSIVE) == 0) {
 1819                         flags |= LLE_EXCLUSIVE;
 1820                         goto retry;
 1821                 }
 1822                 ln->ln_state = ND6_LLINFO_STALE;
 1823                 nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz);
 1824         }
 1825 
 1826         /*
 1827          * The first time we send a packet to a neighbor whose entry is
 1828          * STALE, we have to change the state to DELAY and a sets a timer to
 1829          * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
 1830          * neighbor unreachability detection on expiration.
 1831          * (RFC 2461 7.3.3)
 1832          */
 1833         if (ln->ln_state == ND6_LLINFO_STALE) {
 1834                 if ((flags & LLE_EXCLUSIVE) == 0) {
 1835                         flags |= LLE_EXCLUSIVE;
 1836                         LLE_RUNLOCK(ln);
 1837                         goto retry;
 1838                 }
 1839                 ln->la_asked = 0;
 1840                 ln->ln_state = ND6_LLINFO_DELAY;
 1841                 nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz);
 1842         }
 1843 
 1844         /*
 1845          * If the neighbor cache entry has a state other than INCOMPLETE
 1846          * (i.e. its link-layer address is already resolved), just
 1847          * send the packet.
 1848          */
 1849         if (ln->ln_state > ND6_LLINFO_INCOMPLETE)
 1850                 goto sendpkt;
 1851 
 1852         /*
 1853          * There is a neighbor cache entry, but no ethernet address
 1854          * response yet.  Append this latest packet to the end of the
 1855          * packet queue in the mbuf, unless the number of the packet
 1856          * does not exceed nd6_maxqueuelen.  When it exceeds nd6_maxqueuelen,
 1857          * the oldest packet in the queue will be removed.
 1858          */
 1859         if (ln->ln_state == ND6_LLINFO_NOSTATE)
 1860                 ln->ln_state = ND6_LLINFO_INCOMPLETE;
 1861 
 1862         if ((flags & LLE_EXCLUSIVE) == 0) {
 1863                 flags |= LLE_EXCLUSIVE;
 1864                 LLE_RUNLOCK(ln);
 1865                 goto retry;
 1866         }
 1867 
 1868         LLE_WLOCK_ASSERT(ln);
 1869 
 1870         if (ln->la_hold) {
 1871                 struct mbuf *m_hold;
 1872                 int i;
 1873                 
 1874                 i = 0;
 1875                 for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) {
 1876                         i++;
 1877                         if (m_hold->m_nextpkt == NULL) {
 1878                                 m_hold->m_nextpkt = m;
 1879                                 break;
 1880                         }
 1881                 }
 1882                 while (i >= V_nd6_maxqueuelen) {
 1883                         m_hold = ln->la_hold;
 1884                         ln->la_hold = ln->la_hold->m_nextpkt;
 1885                         m_freem(m_hold);
 1886                         i--;
 1887                 }
 1888         } else {
 1889                 ln->la_hold = m;
 1890         }
 1891 
 1892         /*
 1893          * If there has been no NS for the neighbor after entering the
 1894          * INCOMPLETE state, send the first solicitation.
 1895          */
 1896         if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) {
 1897                 ln->la_asked++;
 1898                 
 1899                 nd6_llinfo_settimer_locked(ln,
 1900                     (long)ND_IFINFO(ifp)->retrans * hz / 1000);
 1901                 LLE_WUNLOCK(ln);
 1902                 nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0);
 1903                 if (lle != NULL && ln == lle)
 1904                         LLE_WLOCK(lle);
 1905 
 1906         } else if (lle == NULL || ln != lle) {
 1907                 /*
 1908                  * We did the lookup (no lle arg) so we
 1909                  * need to do the unlock here.
 1910                  */
 1911                 LLE_WUNLOCK(ln);
 1912         }
 1913 
 1914         return (0);
 1915 
 1916   sendpkt:
 1917         /* discard the packet if IPv6 operation is disabled on the interface */
 1918         if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
 1919                 error = ENETDOWN; /* better error? */
 1920                 goto bad;
 1921         }
 1922         /*
 1923          * ln is valid and the caller did not pass in 
 1924          * an llentry
 1925          */
 1926         if ((ln != NULL) && (lle == NULL)) {
 1927                 if (flags & LLE_EXCLUSIVE)
 1928                         LLE_WUNLOCK(ln);
 1929                 else
 1930                         LLE_RUNLOCK(ln);
 1931         }
 1932 
 1933 #ifdef MAC
 1934         mac_netinet6_nd6_send(ifp, m);
 1935 #endif
 1936         /*
 1937          * We were passed in a pointer to an lle with the lock held 
 1938          * this means that we can't call if_output as we will
 1939          * recurse on the lle lock - so what we do is we create
 1940          * a list of mbufs to send and transmit them in the caller
 1941          * after the lock is dropped
 1942          */
 1943         if (lle != NULL) {
 1944                 if (*chain == NULL)
 1945                         *chain = m;
 1946                 else {
 1947                         struct mbuf *mb;
 1948 
 1949                         /*
 1950                          * append mbuf to end of deferred chain
 1951                          */
 1952                         mb = *chain;
 1953                         while (mb->m_nextpkt != NULL)
 1954                                 mb = mb->m_nextpkt;
 1955                         mb->m_nextpkt = m;
 1956                 }
 1957                 return (error);
 1958         }
 1959         if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
 1960                 return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst,
 1961                     NULL));
 1962         }
 1963         error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, NULL);
 1964         return (error);
 1965 
 1966   bad:
 1967         /*
 1968          * ln is valid and the caller did not pass in 
 1969          * an llentry
 1970          */
 1971         if ((ln != NULL) && (lle == NULL)) {
 1972                 if (flags & LLE_EXCLUSIVE)
 1973                         LLE_WUNLOCK(ln);
 1974                 else
 1975                         LLE_RUNLOCK(ln);
 1976         }
 1977         if (m)
 1978                 m_freem(m);
 1979         return (error);
 1980 }
 1981 #undef senderr
 1982 
 1983 
 1984 int
 1985 nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain,
 1986     struct sockaddr_in6 *dst, struct route *ro)
 1987 {
 1988         struct mbuf *m, *m_head;
 1989         struct ifnet *outifp;
 1990         int error = 0;
 1991 
 1992         m_head = chain;
 1993         if ((ifp->if_flags & IFF_LOOPBACK) != 0)
 1994                 outifp = origifp;
 1995         else
 1996                 outifp = ifp;
 1997         
 1998         while (m_head) {
 1999                 m = m_head;
 2000                 m_head = m_head->m_nextpkt;
 2001                 error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro);                         
 2002         }
 2003 
 2004         /*
 2005          * XXX
 2006          * note that intermediate errors are blindly ignored - but this is 
 2007          * the same convention as used with nd6_output when called by
 2008          * nd6_cache_lladdr
 2009          */
 2010         return (error);
 2011 }       
 2012 
 2013 
 2014 int
 2015 nd6_need_cache(struct ifnet *ifp)
 2016 {
 2017         /*
 2018          * XXX: we currently do not make neighbor cache on any interface
 2019          * other than ARCnet, Ethernet, FDDI and GIF.
 2020          *
 2021          * RFC2893 says:
 2022          * - unidirectional tunnels needs no ND
 2023          */
 2024         switch (ifp->if_type) {
 2025         case IFT_ARCNET:
 2026         case IFT_ETHER:
 2027         case IFT_FDDI:
 2028         case IFT_IEEE1394:
 2029 #ifdef IFT_L2VLAN
 2030         case IFT_L2VLAN:
 2031 #endif
 2032 #ifdef IFT_IEEE80211
 2033         case IFT_IEEE80211:
 2034 #endif
 2035 #ifdef IFT_CARP
 2036         case IFT_CARP:
 2037 #endif
 2038         case IFT_GIF:           /* XXX need more cases? */
 2039         case IFT_PPP:
 2040         case IFT_TUNNEL:
 2041         case IFT_BRIDGE:
 2042         case IFT_PROPVIRTUAL:
 2043                 return (1);
 2044         default:
 2045                 return (0);
 2046         }
 2047 }
 2048 
 2049 /*
 2050  * the callers of this function need to be re-worked to drop
 2051  * the lle lock, drop here for now
 2052  */
 2053 int
 2054 nd6_storelladdr(struct ifnet *ifp, struct mbuf *m,
 2055     struct sockaddr *dst, u_char *desten, struct llentry **lle)
 2056 {
 2057         struct llentry *ln;
 2058 
 2059         *lle = NULL;
 2060         IF_AFDATA_UNLOCK_ASSERT(ifp);
 2061         if (m->m_flags & M_MCAST) {
 2062                 int i;
 2063 
 2064                 switch (ifp->if_type) {
 2065                 case IFT_ETHER:
 2066                 case IFT_FDDI:
 2067 #ifdef IFT_L2VLAN
 2068                 case IFT_L2VLAN:
 2069 #endif
 2070 #ifdef IFT_IEEE80211
 2071                 case IFT_IEEE80211:
 2072 #endif
 2073                 case IFT_BRIDGE:
 2074                 case IFT_ISO88025:
 2075                         ETHER_MAP_IPV6_MULTICAST(&SIN6(dst)->sin6_addr,
 2076                                                  desten);
 2077                         return (0);
 2078                 case IFT_IEEE1394:
 2079                         /*
 2080                          * netbsd can use if_broadcastaddr, but we don't do so
 2081                          * to reduce # of ifdef.
 2082                          */
 2083                         for (i = 0; i < ifp->if_addrlen; i++)
 2084                                 desten[i] = ~0;
 2085                         return (0);
 2086                 case IFT_ARCNET:
 2087                         *desten = 0;
 2088                         return (0);
 2089                 default:
 2090                         m_freem(m);
 2091                         return (EAFNOSUPPORT);
 2092                 }
 2093         }
 2094 
 2095 
 2096         /*
 2097          * the entry should have been created in nd6_store_lladdr
 2098          */
 2099         IF_AFDATA_RLOCK(ifp);
 2100         ln = lla_lookup(LLTABLE6(ifp), 0, dst);
 2101         IF_AFDATA_RUNLOCK(ifp);
 2102         if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) {
 2103                 if (ln != NULL)
 2104                         LLE_RUNLOCK(ln);
 2105                 /* this could happen, if we could not allocate memory */
 2106                 m_freem(m);
 2107                 return (1);
 2108         }
 2109 
 2110         bcopy(&ln->ll_addr, desten, ifp->if_addrlen);
 2111         *lle = ln;
 2112         LLE_RUNLOCK(ln);
 2113         /*
 2114          * A *small* use after free race exists here
 2115          */
 2116         return (0);
 2117 }
 2118 
 2119 static void 
 2120 clear_llinfo_pqueue(struct llentry *ln)
 2121 {
 2122         struct mbuf *m_hold, *m_hold_next;
 2123 
 2124         for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
 2125                 m_hold_next = m_hold->m_nextpkt;
 2126                 m_hold->m_nextpkt = NULL;
 2127                 m_freem(m_hold);
 2128         }
 2129 
 2130         ln->la_hold = NULL;
 2131         return;
 2132 }
 2133 
 2134 static int nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS);
 2135 static int nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS);
 2136 #ifdef SYSCTL_DECL
 2137 SYSCTL_DECL(_net_inet6_icmp6);
 2138 #endif
 2139 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_DRLIST, nd6_drlist,
 2140         CTLFLAG_RD, nd6_sysctl_drlist, "");
 2141 SYSCTL_NODE(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
 2142         CTLFLAG_RD, nd6_sysctl_prlist, "");
 2143 SYSCTL_VNET_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
 2144         CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
 2145 
 2146 static int
 2147 nd6_sysctl_drlist(SYSCTL_HANDLER_ARGS)
 2148 {
 2149         struct in6_defrouter d;
 2150         struct nd_defrouter *dr;
 2151         int error;
 2152 
 2153         if (req->newptr)
 2154                 return (EPERM);
 2155 
 2156         bzero(&d, sizeof(d));
 2157         d.rtaddr.sin6_family = AF_INET6;
 2158         d.rtaddr.sin6_len = sizeof(d.rtaddr);
 2159 
 2160         /*
 2161          * XXX locking
 2162          */
 2163         TAILQ_FOREACH(dr, &V_nd_defrouter, dr_entry) {
 2164                 d.rtaddr.sin6_addr = dr->rtaddr;
 2165                 error = sa6_recoverscope(&d.rtaddr);
 2166                 if (error != 0)
 2167                         return (error);
 2168                 d.flags = dr->flags;
 2169                 d.rtlifetime = dr->rtlifetime;
 2170                 d.expire = dr->expire;
 2171                 d.if_index = dr->ifp->if_index;
 2172                 error = SYSCTL_OUT(req, &d, sizeof(d));
 2173                 if (error != 0)
 2174                         return (error);
 2175         }
 2176         return (0);
 2177 }
 2178 
 2179 static int
 2180 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
 2181 {
 2182         struct in6_prefix p;
 2183         struct sockaddr_in6 s6;
 2184         struct nd_prefix *pr;
 2185         struct nd_pfxrouter *pfr;
 2186         time_t maxexpire;
 2187         int error;
 2188         char ip6buf[INET6_ADDRSTRLEN];
 2189 
 2190         if (req->newptr)
 2191                 return (EPERM);
 2192 
 2193         bzero(&p, sizeof(p));
 2194         p.origin = PR_ORIG_RA;
 2195         bzero(&s6, sizeof(s6));
 2196         s6.sin6_family = AF_INET6;
 2197         s6.sin6_len = sizeof(s6);
 2198 
 2199         /*
 2200          * XXX locking
 2201          */
 2202         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 2203                 p.prefix = pr->ndpr_prefix;
 2204                 if (sa6_recoverscope(&p.prefix)) {
 2205                         log(LOG_ERR, "scope error in prefix list (%s)\n",
 2206                             ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
 2207                         /* XXX: press on... */
 2208                 }
 2209                 p.raflags = pr->ndpr_raf;
 2210                 p.prefixlen = pr->ndpr_plen;
 2211                 p.vltime = pr->ndpr_vltime;
 2212                 p.pltime = pr->ndpr_pltime;
 2213                 p.if_index = pr->ndpr_ifp->if_index;
 2214                 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
 2215                         p.expire = 0;
 2216                 else {
 2217                         /* XXX: we assume time_t is signed. */
 2218                         maxexpire = (-1) &
 2219                             ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
 2220                         if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
 2221                                 p.expire = pr->ndpr_lastupdate +
 2222                                     pr->ndpr_vltime;
 2223                         else
 2224                                 p.expire = maxexpire;
 2225                 }
 2226                 p.refcnt = pr->ndpr_refcnt;
 2227                 p.flags = pr->ndpr_stateflags;
 2228                 p.advrtrs = 0;
 2229                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
 2230                         p.advrtrs++;
 2231                 error = SYSCTL_OUT(req, &p, sizeof(p));
 2232                 if (error != 0)
 2233                         return (error);
 2234                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
 2235                         s6.sin6_addr = pfr->router->rtaddr;
 2236                         if (sa6_recoverscope(&s6))
 2237                                 log(LOG_ERR,
 2238                                     "scope error in prefix list (%s)\n",
 2239                                     ip6_sprintf(ip6buf, &pfr->router->rtaddr));
 2240                         error = SYSCTL_OUT(req, &s6, sizeof(s6));
 2241                         if (error != 0)
 2242                                 return (error);
 2243                 }
 2244         }
 2245         return (0);
 2246 }

Cache object: ea879cb498c5a505da6160fc2e325a46


[ 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.