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


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

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

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

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
    5  * All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the project nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      $KAME: nd6.c,v 1.144 2001/05/24 07:44:00 itojun Exp $
   32  */
   33 
   34 #include <sys/cdefs.h>
   35 __FBSDID("$FreeBSD$");
   36 
   37 #include "opt_inet.h"
   38 #include "opt_inet6.h"
   39 
   40 #include <sys/param.h>
   41 #include <sys/systm.h>
   42 #include <sys/callout.h>
   43 #include <sys/lock.h>
   44 #include <sys/malloc.h>
   45 #include <sys/mbuf.h>
   46 #include <sys/mutex.h>
   47 #include <sys/socket.h>
   48 #include <sys/sockio.h>
   49 #include <sys/time.h>
   50 #include <sys/kernel.h>
   51 #include <sys/protosw.h>
   52 #include <sys/errno.h>
   53 #include <sys/syslog.h>
   54 #include <sys/rwlock.h>
   55 #include <sys/queue.h>
   56 #include <sys/sdt.h>
   57 #include <sys/sysctl.h>
   58 
   59 #include <net/if.h>
   60 #include <net/if_var.h>
   61 #include <net/if_dl.h>
   62 #include <net/if_types.h>
   63 #include <net/route.h>
   64 #include <net/vnet.h>
   65 
   66 #include <netinet/in.h>
   67 #include <netinet/in_kdtrace.h>
   68 #include <net/if_llatbl.h>
   69 #include <netinet/if_ether.h>
   70 #include <netinet6/in6_var.h>
   71 #include <netinet/ip6.h>
   72 #include <netinet6/ip6_var.h>
   73 #include <netinet6/scope6_var.h>
   74 #include <netinet6/nd6.h>
   75 #include <netinet6/in6_ifattach.h>
   76 #include <netinet/icmp6.h>
   77 #include <netinet6/send.h>
   78 
   79 #include <sys/limits.h>
   80 
   81 #include <security/mac/mac_framework.h>
   82 
   83 #define ND6_SLOWTIMER_INTERVAL (60 * 60) /* 1 hour */
   84 #define ND6_RECALC_REACHTM_INTERVAL (60 * 120) /* 2 hours */
   85 
   86 #define SIN6(s) ((const struct sockaddr_in6 *)(s))
   87 
   88 MALLOC_DEFINE(M_IP6NDP, "ip6ndp", "IPv6 Neighbor Discovery");
   89 
   90 /* timer values */
   91 VNET_DEFINE(int, nd6_prune)     = 1;    /* walk list every 1 seconds */
   92 VNET_DEFINE(int, nd6_delay)     = 5;    /* delay first probe time 5 second */
   93 VNET_DEFINE(int, nd6_umaxtries) = 3;    /* maximum unicast query */
   94 VNET_DEFINE(int, nd6_mmaxtries) = 3;    /* maximum multicast query */
   95 VNET_DEFINE(int, nd6_useloopback) = 1;  /* use loopback interface for
   96                                          * local traffic */
   97 VNET_DEFINE(int, nd6_gctimer)   = (60 * 60 * 24); /* 1 day: garbage
   98                                          * collection timer */
   99 
  100 /* preventing too many loops in ND option parsing */
  101 VNET_DEFINE_STATIC(int, nd6_maxndopt) = 10; /* max # of ND options allowed */
  102 
  103 VNET_DEFINE(int, nd6_maxnudhint) = 0;   /* max # of subsequent upper
  104                                          * layer hints */
  105 VNET_DEFINE_STATIC(int, nd6_maxqueuelen) = 1; /* max pkts cached in unresolved
  106                                          * ND entries */
  107 #define V_nd6_maxndopt                  VNET(nd6_maxndopt)
  108 #define V_nd6_maxqueuelen               VNET(nd6_maxqueuelen)
  109 
  110 #ifdef ND6_DEBUG
  111 VNET_DEFINE(int, nd6_debug) = 1;
  112 #else
  113 VNET_DEFINE(int, nd6_debug) = 0;
  114 #endif
  115 
  116 static eventhandler_tag lle_event_eh, iflladdr_event_eh;
  117 
  118 VNET_DEFINE(struct nd_prhead, nd_prefix);
  119 VNET_DEFINE(struct rwlock, nd6_lock);
  120 VNET_DEFINE(uint64_t, nd6_list_genid);
  121 VNET_DEFINE(struct mtx, nd6_onlink_mtx);
  122 
  123 VNET_DEFINE(int, nd6_recalc_reachtm_interval) = ND6_RECALC_REACHTM_INTERVAL;
  124 #define V_nd6_recalc_reachtm_interval   VNET(nd6_recalc_reachtm_interval)
  125 
  126 int     (*send_sendso_input_hook)(struct mbuf *, struct ifnet *, int, int);
  127 
  128 static int nd6_is_new_addr_neighbor(const struct sockaddr_in6 *,
  129         struct ifnet *);
  130 static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *);
  131 static void nd6_slowtimo(void *);
  132 static int regen_tmpaddr(struct in6_ifaddr *);
  133 static void nd6_free(struct llentry **, int);
  134 static void nd6_free_redirect(const struct llentry *);
  135 static void nd6_llinfo_timer(void *);
  136 static void nd6_llinfo_settimer_locked(struct llentry *, long);
  137 static void clear_llinfo_pqueue(struct llentry *);
  138 static void nd6_rtrequest(int, struct rtentry *, struct rt_addrinfo *);
  139 static int nd6_resolve_slow(struct ifnet *, int, struct mbuf *,
  140     const struct sockaddr_in6 *, u_char *, uint32_t *, struct llentry **);
  141 static int nd6_need_cache(struct ifnet *);
  142  
  143 
  144 VNET_DEFINE_STATIC(struct callout, nd6_slowtimo_ch);
  145 #define V_nd6_slowtimo_ch               VNET(nd6_slowtimo_ch)
  146 
  147 VNET_DEFINE_STATIC(struct callout, nd6_timer_ch);
  148 #define V_nd6_timer_ch                  VNET(nd6_timer_ch)
  149 
  150 SYSCTL_DECL(_net_inet6_icmp6);
  151 
  152 static void
  153 nd6_lle_event(void *arg __unused, struct llentry *lle, int evt)
  154 {
  155         struct rt_addrinfo rtinfo;
  156         struct sockaddr_in6 dst;
  157         struct sockaddr_dl gw;
  158         struct ifnet *ifp;
  159         int type;
  160         int fibnum;
  161 
  162         LLE_WLOCK_ASSERT(lle);
  163 
  164         if (lltable_get_af(lle->lle_tbl) != AF_INET6)
  165                 return;
  166 
  167         switch (evt) {
  168         case LLENTRY_RESOLVED:
  169                 type = RTM_ADD;
  170                 KASSERT(lle->la_flags & LLE_VALID,
  171                     ("%s: %p resolved but not valid?", __func__, lle));
  172                 break;
  173         case LLENTRY_EXPIRED:
  174                 type = RTM_DELETE;
  175                 break;
  176         default:
  177                 return;
  178         }
  179 
  180         ifp = lltable_get_ifp(lle->lle_tbl);
  181 
  182         bzero(&dst, sizeof(dst));
  183         bzero(&gw, sizeof(gw));
  184         bzero(&rtinfo, sizeof(rtinfo));
  185         lltable_fill_sa_entry(lle, (struct sockaddr *)&dst);
  186         dst.sin6_scope_id = in6_getscopezone(ifp,
  187             in6_addrscope(&dst.sin6_addr));
  188         gw.sdl_len = sizeof(struct sockaddr_dl);
  189         gw.sdl_family = AF_LINK;
  190         gw.sdl_alen = ifp->if_addrlen;
  191         gw.sdl_index = ifp->if_index;
  192         gw.sdl_type = ifp->if_type;
  193         if (evt == LLENTRY_RESOLVED)
  194                 bcopy(lle->ll_addr, gw.sdl_data, ifp->if_addrlen);
  195         rtinfo.rti_info[RTAX_DST] = (struct sockaddr *)&dst;
  196         rtinfo.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gw;
  197         rtinfo.rti_addrs = RTA_DST | RTA_GATEWAY;
  198         fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ifp->if_fib;
  199         rt_missmsg_fib(type, &rtinfo, RTF_HOST | RTF_LLDATA | (
  200             type == RTM_ADD ? RTF_UP: 0), 0, fibnum);
  201 }
  202 
  203 /*
  204  * A handler for interface link layer address change event.
  205  */
  206 static void
  207 nd6_iflladdr(void *arg __unused, struct ifnet *ifp)
  208 {
  209         if (ifp->if_afdata[AF_INET6] == NULL)
  210                 return;
  211 
  212         lltable_update_ifaddr(LLTABLE6(ifp));
  213 }
  214 
  215 void
  216 nd6_init(void)
  217 {
  218 
  219         mtx_init(&V_nd6_onlink_mtx, "nd6 onlink", NULL, MTX_DEF);
  220         rw_init(&V_nd6_lock, "nd6 list");
  221 
  222         LIST_INIT(&V_nd_prefix);
  223         nd6_defrouter_init();
  224 
  225         /* Start timers. */
  226         callout_init(&V_nd6_slowtimo_ch, 0);
  227         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
  228             nd6_slowtimo, curvnet);
  229 
  230         callout_init(&V_nd6_timer_ch, 0);
  231         callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
  232 
  233         nd6_dad_init();
  234         if (IS_DEFAULT_VNET(curvnet)) {
  235                 lle_event_eh = EVENTHANDLER_REGISTER(lle_event, nd6_lle_event,
  236                     NULL, EVENTHANDLER_PRI_ANY);
  237                 iflladdr_event_eh = EVENTHANDLER_REGISTER(iflladdr_event,
  238                     nd6_iflladdr, NULL, EVENTHANDLER_PRI_ANY);
  239         }
  240 }
  241 
  242 #ifdef VIMAGE
  243 void
  244 nd6_destroy(void)
  245 {
  246 
  247         callout_drain(&V_nd6_slowtimo_ch);
  248         callout_drain(&V_nd6_timer_ch);
  249         if (IS_DEFAULT_VNET(curvnet)) {
  250                 EVENTHANDLER_DEREGISTER(lle_event, lle_event_eh);
  251                 EVENTHANDLER_DEREGISTER(iflladdr_event, iflladdr_event_eh);
  252         }
  253         rw_destroy(&V_nd6_lock);
  254         mtx_destroy(&V_nd6_onlink_mtx);
  255 }
  256 #endif
  257 
  258 struct nd_ifinfo *
  259 nd6_ifattach(struct ifnet *ifp)
  260 {
  261         struct nd_ifinfo *nd;
  262 
  263         nd = malloc(sizeof(*nd), M_IP6NDP, M_WAITOK | M_ZERO);
  264         nd->initialized = 1;
  265 
  266         nd->chlim = IPV6_DEFHLIM;
  267         nd->basereachable = REACHABLE_TIME;
  268         nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
  269         nd->retrans = RETRANS_TIMER;
  270 
  271         nd->flags = ND6_IFF_PERFORMNUD;
  272 
  273         /* A loopback interface always has ND6_IFF_AUTO_LINKLOCAL.
  274          * XXXHRS: Clear ND6_IFF_AUTO_LINKLOCAL on an IFT_BRIDGE interface by
  275          * default regardless of the V_ip6_auto_linklocal configuration to
  276          * give a reasonable default behavior.
  277          */
  278         if ((V_ip6_auto_linklocal && ifp->if_type != IFT_BRIDGE) ||
  279             (ifp->if_flags & IFF_LOOPBACK))
  280                 nd->flags |= ND6_IFF_AUTO_LINKLOCAL;
  281         /*
  282          * A loopback interface does not need to accept RTADV.
  283          * XXXHRS: Clear ND6_IFF_ACCEPT_RTADV on an IFT_BRIDGE interface by
  284          * default regardless of the V_ip6_accept_rtadv configuration to
  285          * prevent the interface from accepting RA messages arrived
  286          * on one of the member interfaces with ND6_IFF_ACCEPT_RTADV.
  287          */
  288         if (V_ip6_accept_rtadv &&
  289             !(ifp->if_flags & IFF_LOOPBACK) &&
  290             (ifp->if_type != IFT_BRIDGE))
  291                         nd->flags |= ND6_IFF_ACCEPT_RTADV;
  292         if (V_ip6_no_radr && !(ifp->if_flags & IFF_LOOPBACK))
  293                 nd->flags |= ND6_IFF_NO_RADR;
  294 
  295         /* XXX: we cannot call nd6_setmtu since ifp is not fully initialized */
  296         nd6_setmtu0(ifp, nd);
  297 
  298         return nd;
  299 }
  300 
  301 void
  302 nd6_ifdetach(struct ifnet *ifp, struct nd_ifinfo *nd)
  303 {
  304         struct ifaddr *ifa, *next;
  305 
  306         IF_ADDR_RLOCK(ifp);
  307         CK_STAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, next) {
  308                 if (ifa->ifa_addr->sa_family != AF_INET6)
  309                         continue;
  310 
  311                 /* stop DAD processing */
  312                 nd6_dad_stop(ifa);
  313         }
  314         IF_ADDR_RUNLOCK(ifp);
  315 
  316         free(nd, M_IP6NDP);
  317 }
  318 
  319 /*
  320  * Reset ND level link MTU. This function is called when the physical MTU
  321  * changes, which means we might have to adjust the ND level MTU.
  322  */
  323 void
  324 nd6_setmtu(struct ifnet *ifp)
  325 {
  326         if (ifp->if_afdata[AF_INET6] == NULL)
  327                 return;
  328 
  329         nd6_setmtu0(ifp, ND_IFINFO(ifp));
  330 }
  331 
  332 /* XXX todo: do not maintain copy of ifp->if_mtu in ndi->maxmtu */
  333 void
  334 nd6_setmtu0(struct ifnet *ifp, struct nd_ifinfo *ndi)
  335 {
  336         u_int32_t omaxmtu;
  337 
  338         omaxmtu = ndi->maxmtu;
  339         ndi->maxmtu = ifp->if_mtu;
  340 
  341         /*
  342          * Decreasing the interface MTU under IPV6 minimum MTU may cause
  343          * undesirable situation.  We thus notify the operator of the change
  344          * explicitly.  The check for omaxmtu is necessary to restrict the
  345          * log to the case of changing the MTU, not initializing it.
  346          */
  347         if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
  348                 log(LOG_NOTICE, "nd6_setmtu0: "
  349                     "new link MTU on %s (%lu) is too small for IPv6\n",
  350                     if_name(ifp), (unsigned long)ndi->maxmtu);
  351         }
  352 
  353         if (ndi->maxmtu > V_in6_maxmtu)
  354                 in6_setmaxmtu(); /* check all interfaces just in case */
  355 
  356 }
  357 
  358 void
  359 nd6_option_init(void *opt, int icmp6len, union nd_opts *ndopts)
  360 {
  361 
  362         bzero(ndopts, sizeof(*ndopts));
  363         ndopts->nd_opts_search = (struct nd_opt_hdr *)opt;
  364         ndopts->nd_opts_last
  365                 = (struct nd_opt_hdr *)(((u_char *)opt) + icmp6len);
  366 
  367         if (icmp6len == 0) {
  368                 ndopts->nd_opts_done = 1;
  369                 ndopts->nd_opts_search = NULL;
  370         }
  371 }
  372 
  373 /*
  374  * Take one ND option.
  375  */
  376 struct nd_opt_hdr *
  377 nd6_option(union nd_opts *ndopts)
  378 {
  379         struct nd_opt_hdr *nd_opt;
  380         int olen;
  381 
  382         KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
  383         KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
  384             __func__));
  385         if (ndopts->nd_opts_search == NULL)
  386                 return NULL;
  387         if (ndopts->nd_opts_done)
  388                 return NULL;
  389 
  390         nd_opt = ndopts->nd_opts_search;
  391 
  392         /* make sure nd_opt_len is inside the buffer */
  393         if ((caddr_t)&nd_opt->nd_opt_len >= (caddr_t)ndopts->nd_opts_last) {
  394                 bzero(ndopts, sizeof(*ndopts));
  395                 return NULL;
  396         }
  397 
  398         olen = nd_opt->nd_opt_len << 3;
  399         if (olen == 0) {
  400                 /*
  401                  * Message validation requires that all included
  402                  * options have a length that is greater than zero.
  403                  */
  404                 bzero(ndopts, sizeof(*ndopts));
  405                 return NULL;
  406         }
  407 
  408         ndopts->nd_opts_search = (struct nd_opt_hdr *)((caddr_t)nd_opt + olen);
  409         if (ndopts->nd_opts_search > ndopts->nd_opts_last) {
  410                 /* option overruns the end of buffer, invalid */
  411                 bzero(ndopts, sizeof(*ndopts));
  412                 return NULL;
  413         } else if (ndopts->nd_opts_search == ndopts->nd_opts_last) {
  414                 /* reached the end of options chain */
  415                 ndopts->nd_opts_done = 1;
  416                 ndopts->nd_opts_search = NULL;
  417         }
  418         return nd_opt;
  419 }
  420 
  421 /*
  422  * Parse multiple ND options.
  423  * This function is much easier to use, for ND routines that do not need
  424  * multiple options of the same type.
  425  */
  426 int
  427 nd6_options(union nd_opts *ndopts)
  428 {
  429         struct nd_opt_hdr *nd_opt;
  430         int i = 0;
  431 
  432         KASSERT(ndopts != NULL, ("%s: ndopts == NULL", __func__));
  433         KASSERT(ndopts->nd_opts_last != NULL, ("%s: uninitialized ndopts",
  434             __func__));
  435         if (ndopts->nd_opts_search == NULL)
  436                 return 0;
  437 
  438         while (1) {
  439                 nd_opt = nd6_option(ndopts);
  440                 if (nd_opt == NULL && ndopts->nd_opts_last == NULL) {
  441                         /*
  442                          * Message validation requires that all included
  443                          * options have a length that is greater than zero.
  444                          */
  445                         ICMP6STAT_INC(icp6s_nd_badopt);
  446                         bzero(ndopts, sizeof(*ndopts));
  447                         return -1;
  448                 }
  449 
  450                 if (nd_opt == NULL)
  451                         goto skip1;
  452 
  453                 switch (nd_opt->nd_opt_type) {
  454                 case ND_OPT_SOURCE_LINKADDR:
  455                 case ND_OPT_TARGET_LINKADDR:
  456                 case ND_OPT_MTU:
  457                 case ND_OPT_REDIRECTED_HEADER:
  458                 case ND_OPT_NONCE:
  459                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type]) {
  460                                 nd6log((LOG_INFO,
  461                                     "duplicated ND6 option found (type=%d)\n",
  462                                     nd_opt->nd_opt_type));
  463                                 /* XXX bark? */
  464                         } else {
  465                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  466                                         = nd_opt;
  467                         }
  468                         break;
  469                 case ND_OPT_PREFIX_INFORMATION:
  470                         if (ndopts->nd_opt_array[nd_opt->nd_opt_type] == 0) {
  471                                 ndopts->nd_opt_array[nd_opt->nd_opt_type]
  472                                         = nd_opt;
  473                         }
  474                         ndopts->nd_opts_pi_end =
  475                                 (struct nd_opt_prefix_info *)nd_opt;
  476                         break;
  477                 /* What about ND_OPT_ROUTE_INFO? RFC 4191 */
  478                 case ND_OPT_RDNSS:      /* RFC 6106 */
  479                 case ND_OPT_DNSSL:      /* RFC 6106 */
  480                         /*
  481                          * Silently ignore options we know and do not care about
  482                          * in the kernel.
  483                          */
  484                         break;
  485                 default:
  486                         /*
  487                          * Unknown options must be silently ignored,
  488                          * to accommodate future extension to the protocol.
  489                          */
  490                         nd6log((LOG_DEBUG,
  491                             "nd6_options: unsupported option %d - "
  492                             "option ignored\n", nd_opt->nd_opt_type));
  493                 }
  494 
  495 skip1:
  496                 i++;
  497                 if (i > V_nd6_maxndopt) {
  498                         ICMP6STAT_INC(icp6s_nd_toomanyopt);
  499                         nd6log((LOG_INFO, "too many loop in nd opt\n"));
  500                         break;
  501                 }
  502 
  503                 if (ndopts->nd_opts_done)
  504                         break;
  505         }
  506 
  507         return 0;
  508 }
  509 
  510 /*
  511  * ND6 timer routine to handle ND6 entries
  512  */
  513 static void
  514 nd6_llinfo_settimer_locked(struct llentry *ln, long tick)
  515 {
  516         int canceled;
  517 
  518         LLE_WLOCK_ASSERT(ln);
  519 
  520         if (tick < 0) {
  521                 ln->la_expire = 0;
  522                 ln->ln_ntick = 0;
  523                 canceled = callout_stop(&ln->lle_timer);
  524         } else {
  525                 ln->la_expire = time_uptime + tick / hz;
  526                 LLE_ADDREF(ln);
  527                 if (tick > INT_MAX) {
  528                         ln->ln_ntick = tick - INT_MAX;
  529                         canceled = callout_reset(&ln->lle_timer, INT_MAX,
  530                             nd6_llinfo_timer, ln);
  531                 } else {
  532                         ln->ln_ntick = 0;
  533                         canceled = callout_reset(&ln->lle_timer, tick,
  534                             nd6_llinfo_timer, ln);
  535                 }
  536         }
  537         if (canceled > 0)
  538                 LLE_REMREF(ln);
  539 }
  540 
  541 /*
  542  * Gets source address of the first packet in hold queue
  543  * and stores it in @src.
  544  * Returns pointer to @src (if hold queue is not empty) or NULL.
  545  *
  546  * Set noinline to be dtrace-friendly
  547  */
  548 static __noinline struct in6_addr *
  549 nd6_llinfo_get_holdsrc(struct llentry *ln, struct in6_addr *src)
  550 {
  551         struct ip6_hdr hdr;
  552         struct mbuf *m;
  553 
  554         if (ln->la_hold == NULL)
  555                 return (NULL);
  556 
  557         /*
  558          * assume every packet in la_hold has the same IP header
  559          */
  560         m = ln->la_hold;
  561         if (sizeof(hdr) > m->m_len)
  562                 return (NULL);
  563 
  564         m_copydata(m, 0, sizeof(hdr), (caddr_t)&hdr);
  565         *src = hdr.ip6_src;
  566 
  567         return (src);
  568 }
  569 
  570 /*
  571  * Checks if we need to switch from STALE state.
  572  *
  573  * RFC 4861 requires switching from STALE to DELAY state
  574  * on first packet matching entry, waiting V_nd6_delay and
  575  * transition to PROBE state (if upper layer confirmation was
  576  * not received).
  577  *
  578  * This code performs a bit differently:
  579  * On packet hit we don't change state (but desired state
  580  * can be guessed by control plane). However, after V_nd6_delay
  581  * seconds code will transition to PROBE state (so DELAY state
  582  * is kinda skipped in most situations).
  583  *
  584  * Typically, V_nd6_gctimer is bigger than V_nd6_delay, so
  585  * we perform the following upon entering STALE state:
  586  *
  587  * 1) Arm timer to run each V_nd6_delay seconds to make sure that
  588  * if packet was transmitted at the start of given interval, we
  589  * would be able to switch to PROBE state in V_nd6_delay seconds
  590  * as user expects.
  591  *
  592  * 2) Reschedule timer until original V_nd6_gctimer expires keeping
  593  * lle in STALE state (remaining timer value stored in lle_remtime).
  594  *
  595  * 3) Reschedule timer if packet was transmitted less that V_nd6_delay
  596  * seconds ago.
  597  *
  598  * Returns non-zero value if the entry is still STALE (storing
  599  * the next timer interval in @pdelay).
  600  *
  601  * Returns zero value if original timer expired or we need to switch to
  602  * PROBE (store that in @do_switch variable).
  603  */
  604 static int
  605 nd6_is_stale(struct llentry *lle, long *pdelay, int *do_switch)
  606 {
  607         int nd_delay, nd_gctimer, r_skip_req;
  608         time_t lle_hittime;
  609         long delay;
  610 
  611         *do_switch = 0;
  612         nd_gctimer = V_nd6_gctimer;
  613         nd_delay = V_nd6_delay;
  614 
  615         LLE_REQ_LOCK(lle);
  616         r_skip_req = lle->r_skip_req;
  617         lle_hittime = lle->lle_hittime;
  618         LLE_REQ_UNLOCK(lle);
  619 
  620         if (r_skip_req > 0) {
  621 
  622                 /*
  623                  * Nonzero r_skip_req value was set upon entering
  624                  * STALE state. Since value was not changed, no
  625                  * packets were passed using this lle. Ask for
  626                  * timer reschedule and keep STALE state.
  627                  */
  628                 delay = (long)(MIN(nd_gctimer, nd_delay));
  629                 delay *= hz;
  630                 if (lle->lle_remtime > delay)
  631                         lle->lle_remtime -= delay;
  632                 else {
  633                         delay = lle->lle_remtime;
  634                         lle->lle_remtime = 0;
  635                 }
  636 
  637                 if (delay == 0) {
  638 
  639                         /*
  640                          * The original ng6_gctime timeout ended,
  641                          * no more rescheduling.
  642                          */
  643                         return (0);
  644                 }
  645 
  646                 *pdelay = delay;
  647                 return (1);
  648         }
  649 
  650         /*
  651          * Packet received. Verify timestamp
  652          */
  653         delay = (long)(time_uptime - lle_hittime);
  654         if (delay < nd_delay) {
  655 
  656                 /*
  657                  * V_nd6_delay still not passed since the first
  658                  * hit in STALE state.
  659                  * Reschedule timer and return.
  660                  */
  661                 *pdelay = (long)(nd_delay - delay) * hz;
  662                 return (1);
  663         }
  664 
  665         /* Request switching to probe */
  666         *do_switch = 1;
  667         return (0);
  668 }
  669 
  670 
  671 /*
  672  * Switch @lle state to new state optionally arming timers.
  673  *
  674  * Set noinline to be dtrace-friendly
  675  */
  676 __noinline void
  677 nd6_llinfo_setstate(struct llentry *lle, int newstate)
  678 {
  679         struct ifnet *ifp;
  680         int nd_gctimer, nd_delay;
  681         long delay, remtime;
  682 
  683         delay = 0;
  684         remtime = 0;
  685 
  686         switch (newstate) {
  687         case ND6_LLINFO_INCOMPLETE:
  688                 ifp = lle->lle_tbl->llt_ifp;
  689                 delay = (long)ND_IFINFO(ifp)->retrans * hz / 1000;
  690                 break;
  691         case ND6_LLINFO_REACHABLE:
  692                 if (!ND6_LLINFO_PERMANENT(lle)) {
  693                         ifp = lle->lle_tbl->llt_ifp;
  694                         delay = (long)ND_IFINFO(ifp)->reachable * hz;
  695                 }
  696                 break;
  697         case ND6_LLINFO_STALE:
  698 
  699                 /*
  700                  * Notify fast path that we want to know if any packet
  701                  * is transmitted by setting r_skip_req.
  702                  */
  703                 LLE_REQ_LOCK(lle);
  704                 lle->r_skip_req = 1;
  705                 LLE_REQ_UNLOCK(lle);
  706                 nd_delay = V_nd6_delay;
  707                 nd_gctimer = V_nd6_gctimer;
  708 
  709                 delay = (long)(MIN(nd_gctimer, nd_delay)) * hz;
  710                 remtime = (long)nd_gctimer * hz - delay;
  711                 break;
  712         case ND6_LLINFO_DELAY:
  713                 lle->la_asked = 0;
  714                 delay = (long)V_nd6_delay * hz;
  715                 break;
  716         }
  717 
  718         if (delay > 0)
  719                 nd6_llinfo_settimer_locked(lle, delay);
  720 
  721         lle->lle_remtime = remtime;
  722         lle->ln_state = newstate;
  723 }
  724 
  725 /*
  726  * Timer-dependent part of nd state machine.
  727  *
  728  * Set noinline to be dtrace-friendly
  729  */
  730 static __noinline void
  731 nd6_llinfo_timer(void *arg)
  732 {
  733         struct llentry *ln;
  734         struct in6_addr *dst, *pdst, *psrc, src;
  735         struct ifnet *ifp;
  736         struct nd_ifinfo *ndi;
  737         int do_switch, send_ns;
  738         long delay;
  739 
  740         KASSERT(arg != NULL, ("%s: arg NULL", __func__));
  741         ln = (struct llentry *)arg;
  742         ifp = lltable_get_ifp(ln->lle_tbl);
  743         CURVNET_SET(ifp->if_vnet);
  744 
  745         ND6_RLOCK();
  746         LLE_WLOCK(ln);
  747         if (callout_pending(&ln->lle_timer)) {
  748                 /*
  749                  * Here we are a bit odd here in the treatment of 
  750                  * active/pending. If the pending bit is set, it got
  751                  * rescheduled before I ran. The active
  752                  * bit we ignore, since if it was stopped
  753                  * in ll_tablefree() and was currently running
  754                  * it would have return 0 so the code would
  755                  * not have deleted it since the callout could
  756                  * not be stopped so we want to go through
  757                  * with the delete here now. If the callout
  758                  * was restarted, the pending bit will be back on and
  759                  * we just want to bail since the callout_reset would
  760                  * return 1 and our reference would have been removed
  761                  * by nd6_llinfo_settimer_locked above since canceled
  762                  * would have been 1.
  763                  */
  764                 LLE_WUNLOCK(ln);
  765                 ND6_RUNLOCK();
  766                 CURVNET_RESTORE();
  767                 return;
  768         }
  769         ndi = ND_IFINFO(ifp);
  770         send_ns = 0;
  771         dst = &ln->r_l3addr.addr6;
  772         pdst = dst;
  773 
  774         if (ln->ln_ntick > 0) {
  775                 if (ln->ln_ntick > INT_MAX) {
  776                         ln->ln_ntick -= INT_MAX;
  777                         nd6_llinfo_settimer_locked(ln, INT_MAX);
  778                 } else {
  779                         ln->ln_ntick = 0;
  780                         nd6_llinfo_settimer_locked(ln, ln->ln_ntick);
  781                 }
  782                 goto done;
  783         }
  784 
  785         if (ln->la_flags & LLE_STATIC) {
  786                 goto done;
  787         }
  788 
  789         if (ln->la_flags & LLE_DELETED) {
  790                 nd6_free(&ln, 0);
  791                 goto done;
  792         }
  793 
  794         switch (ln->ln_state) {
  795         case ND6_LLINFO_INCOMPLETE:
  796                 if (ln->la_asked < V_nd6_mmaxtries) {
  797                         ln->la_asked++;
  798                         send_ns = 1;
  799                         /* Send NS to multicast address */
  800                         pdst = NULL;
  801                 } else {
  802                         struct mbuf *m = ln->la_hold;
  803                         if (m) {
  804                                 struct mbuf *m0;
  805 
  806                                 /*
  807                                  * assuming every packet in la_hold has the
  808                                  * same IP header.  Send error after unlock.
  809                                  */
  810                                 m0 = m->m_nextpkt;
  811                                 m->m_nextpkt = NULL;
  812                                 ln->la_hold = m0;
  813                                 clear_llinfo_pqueue(ln);
  814                         }
  815                         nd6_free(&ln, 0);
  816                         if (m != NULL)
  817                                 icmp6_error2(m, ICMP6_DST_UNREACH,
  818                                     ICMP6_DST_UNREACH_ADDR, 0, ifp);
  819                 }
  820                 break;
  821         case ND6_LLINFO_REACHABLE:
  822                 if (!ND6_LLINFO_PERMANENT(ln))
  823                         nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
  824                 break;
  825 
  826         case ND6_LLINFO_STALE:
  827                 if (nd6_is_stale(ln, &delay, &do_switch) != 0) {
  828 
  829                         /*
  830                          * No packet has used this entry and GC timeout
  831                          * has not been passed. Reschedule timer and
  832                          * return.
  833                          */
  834                         nd6_llinfo_settimer_locked(ln, delay);
  835                         break;
  836                 }
  837 
  838                 if (do_switch == 0) {
  839 
  840                         /*
  841                          * GC timer has ended and entry hasn't been used.
  842                          * Run Garbage collector (RFC 4861, 5.3)
  843                          */
  844                         if (!ND6_LLINFO_PERMANENT(ln))
  845                                 nd6_free(&ln, 1);
  846                         break;
  847                 }
  848 
  849                 /* Entry has been used AND delay timer has ended. */
  850 
  851                 /* FALLTHROUGH */
  852 
  853         case ND6_LLINFO_DELAY:
  854                 if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) {
  855                         /* We need NUD */
  856                         ln->la_asked = 1;
  857                         nd6_llinfo_setstate(ln, ND6_LLINFO_PROBE);
  858                         send_ns = 1;
  859                 } else
  860                         nd6_llinfo_setstate(ln, ND6_LLINFO_STALE); /* XXX */
  861                 break;
  862         case ND6_LLINFO_PROBE:
  863                 if (ln->la_asked < V_nd6_umaxtries) {
  864                         ln->la_asked++;
  865                         send_ns = 1;
  866                 } else {
  867                         nd6_free(&ln, 0);
  868                 }
  869                 break;
  870         default:
  871                 panic("%s: paths in a dark night can be confusing: %d",
  872                     __func__, ln->ln_state);
  873         }
  874 done:
  875         if (ln != NULL)
  876                 ND6_RUNLOCK();
  877         if (send_ns != 0) {
  878                 nd6_llinfo_settimer_locked(ln, (long)ndi->retrans * hz / 1000);
  879                 psrc = nd6_llinfo_get_holdsrc(ln, &src);
  880                 LLE_FREE_LOCKED(ln);
  881                 ln = NULL;
  882                 nd6_ns_output(ifp, psrc, pdst, dst, NULL);
  883         }
  884 
  885         if (ln != NULL)
  886                 LLE_FREE_LOCKED(ln);
  887         CURVNET_RESTORE();
  888 }
  889 
  890 
  891 /*
  892  * ND6 timer routine to expire default route list and prefix list
  893  */
  894 void
  895 nd6_timer(void *arg)
  896 {
  897         CURVNET_SET((struct vnet *) arg);
  898         struct nd_prhead prl;
  899         struct nd_prefix *pr, *npr;
  900         struct ifnet *ifp;
  901         struct in6_ifaddr *ia6, *nia6;
  902         uint64_t genid;
  903 
  904         LIST_INIT(&prl);
  905 
  906         nd6_defrouter_timer();
  907 
  908         /*
  909          * expire interface addresses.
  910          * in the past the loop was inside prefix expiry processing.
  911          * However, from a stricter speci-confrmance standpoint, we should
  912          * rather separate address lifetimes and prefix lifetimes.
  913          *
  914          * XXXRW: in6_ifaddrhead locking.
  915          */
  916   addrloop:
  917         CK_STAILQ_FOREACH_SAFE(ia6, &V_in6_ifaddrhead, ia_link, nia6) {
  918                 /* check address lifetime */
  919                 if (IFA6_IS_INVALID(ia6)) {
  920                         int regen = 0;
  921 
  922                         /*
  923                          * If the expiring address is temporary, try
  924                          * regenerating a new one.  This would be useful when
  925                          * we suspended a laptop PC, then turned it on after a
  926                          * period that could invalidate all temporary
  927                          * addresses.  Although we may have to restart the
  928                          * loop (see below), it must be after purging the
  929                          * address.  Otherwise, we'd see an infinite loop of
  930                          * regeneration.
  931                          */
  932                         if (V_ip6_use_tempaddr &&
  933                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
  934                                 if (regen_tmpaddr(ia6) == 0)
  935                                         regen = 1;
  936                         }
  937 
  938                         in6_purgeaddr(&ia6->ia_ifa);
  939 
  940                         if (regen)
  941                                 goto addrloop; /* XXX: see below */
  942                 } else if (IFA6_IS_DEPRECATED(ia6)) {
  943                         int oldflags = ia6->ia6_flags;
  944 
  945                         ia6->ia6_flags |= IN6_IFF_DEPRECATED;
  946 
  947                         /*
  948                          * If a temporary address has just become deprecated,
  949                          * regenerate a new one if possible.
  950                          */
  951                         if (V_ip6_use_tempaddr &&
  952                             (ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
  953                             (oldflags & IN6_IFF_DEPRECATED) == 0) {
  954 
  955                                 if (regen_tmpaddr(ia6) == 0) {
  956                                         /*
  957                                          * A new temporary address is
  958                                          * generated.
  959                                          * XXX: this means the address chain
  960                                          * has changed while we are still in
  961                                          * the loop.  Although the change
  962                                          * would not cause disaster (because
  963                                          * it's not a deletion, but an
  964                                          * addition,) we'd rather restart the
  965                                          * loop just for safety.  Or does this
  966                                          * significantly reduce performance??
  967                                          */
  968                                         goto addrloop;
  969                                 }
  970                         }
  971                 } else if ((ia6->ia6_flags & IN6_IFF_TENTATIVE) != 0) {
  972                         /*
  973                          * Schedule DAD for a tentative address.  This happens
  974                          * if the interface was down or not running
  975                          * when the address was configured.
  976                          */
  977                         int delay;
  978 
  979                         delay = arc4random() %
  980                             (MAX_RTR_SOLICITATION_DELAY * hz);
  981                         nd6_dad_start((struct ifaddr *)ia6, delay);
  982                 } else {
  983                         /*
  984                          * Check status of the interface.  If it is down,
  985                          * mark the address as tentative for future DAD.
  986                          */
  987                         ifp = ia6->ia_ifp;
  988                         if ((ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0 &&
  989                             ((ifp->if_flags & IFF_UP) == 0 ||
  990                             (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0 ||
  991                             (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0)){
  992                                 ia6->ia6_flags &= ~IN6_IFF_DUPLICATED;
  993                                 ia6->ia6_flags |= IN6_IFF_TENTATIVE;
  994                         }
  995 
  996                         /*
  997                          * A new RA might have made a deprecated address
  998                          * preferred.
  999                          */
 1000                         ia6->ia6_flags &= ~IN6_IFF_DEPRECATED;
 1001                 }
 1002         }
 1003 
 1004         ND6_WLOCK();
 1005 restart:
 1006         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
 1007                 /*
 1008                  * Expire prefixes. Since the pltime is only used for
 1009                  * autoconfigured addresses, pltime processing for prefixes is
 1010                  * not necessary.
 1011                  *
 1012                  * Only unlink after all derived addresses have expired. This
 1013                  * may not occur until two hours after the prefix has expired
 1014                  * per RFC 4862. If the prefix expires before its derived
 1015                  * addresses, mark it off-link. This will be done automatically
 1016                  * after unlinking if no address references remain.
 1017                  */
 1018                 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME ||
 1019                     time_uptime - pr->ndpr_lastupdate <= pr->ndpr_vltime)
 1020                         continue;
 1021 
 1022                 if (pr->ndpr_addrcnt == 0) {
 1023                         nd6_prefix_unlink(pr, &prl);
 1024                         continue;
 1025                 }
 1026                 if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
 1027                         genid = V_nd6_list_genid;
 1028                         nd6_prefix_ref(pr);
 1029                         ND6_WUNLOCK();
 1030                         ND6_ONLINK_LOCK();
 1031                         (void)nd6_prefix_offlink(pr);
 1032                         ND6_ONLINK_UNLOCK();
 1033                         ND6_WLOCK();
 1034                         nd6_prefix_rele(pr);
 1035                         if (genid != V_nd6_list_genid)
 1036                                 goto restart;
 1037                 }
 1038         }
 1039         ND6_WUNLOCK();
 1040 
 1041         while ((pr = LIST_FIRST(&prl)) != NULL) {
 1042                 LIST_REMOVE(pr, ndpr_entry);
 1043                 nd6_prefix_del(pr);
 1044         }
 1045 
 1046         callout_reset(&V_nd6_timer_ch, V_nd6_prune * hz,
 1047             nd6_timer, curvnet);
 1048 
 1049         CURVNET_RESTORE();
 1050 }
 1051 
 1052 /*
 1053  * ia6 - deprecated/invalidated temporary address
 1054  */
 1055 static int
 1056 regen_tmpaddr(struct in6_ifaddr *ia6)
 1057 {
 1058         struct ifaddr *ifa;
 1059         struct ifnet *ifp;
 1060         struct in6_ifaddr *public_ifa6 = NULL;
 1061 
 1062         ifp = ia6->ia_ifa.ifa_ifp;
 1063         IF_ADDR_RLOCK(ifp);
 1064         CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1065                 struct in6_ifaddr *it6;
 1066 
 1067                 if (ifa->ifa_addr->sa_family != AF_INET6)
 1068                         continue;
 1069 
 1070                 it6 = (struct in6_ifaddr *)ifa;
 1071 
 1072                 /* ignore no autoconf addresses. */
 1073                 if ((it6->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1074                         continue;
 1075 
 1076                 /* ignore autoconf addresses with different prefixes. */
 1077                 if (it6->ia6_ndpr == NULL || it6->ia6_ndpr != ia6->ia6_ndpr)
 1078                         continue;
 1079 
 1080                 /*
 1081                  * Now we are looking at an autoconf address with the same
 1082                  * prefix as ours.  If the address is temporary and is still
 1083                  * preferred, do not create another one.  It would be rare, but
 1084                  * could happen, for example, when we resume a laptop PC after
 1085                  * a long period.
 1086                  */
 1087                 if ((it6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
 1088                     !IFA6_IS_DEPRECATED(it6)) {
 1089                         public_ifa6 = NULL;
 1090                         break;
 1091                 }
 1092 
 1093                 /*
 1094                  * This is a public autoconf address that has the same prefix
 1095                  * as ours.  If it is preferred, keep it.  We can't break the
 1096                  * loop here, because there may be a still-preferred temporary
 1097                  * address with the prefix.
 1098                  */
 1099                 if (!IFA6_IS_DEPRECATED(it6))
 1100                         public_ifa6 = it6;
 1101         }
 1102         if (public_ifa6 != NULL)
 1103                 ifa_ref(&public_ifa6->ia_ifa);
 1104         IF_ADDR_RUNLOCK(ifp);
 1105 
 1106         if (public_ifa6 != NULL) {
 1107                 int e;
 1108 
 1109                 if ((e = in6_tmpifadd(public_ifa6, 0, 0)) != 0) {
 1110                         ifa_free(&public_ifa6->ia_ifa);
 1111                         log(LOG_NOTICE, "regen_tmpaddr: failed to create a new"
 1112                             " tmp addr,errno=%d\n", e);
 1113                         return (-1);
 1114                 }
 1115                 ifa_free(&public_ifa6->ia_ifa);
 1116                 return (0);
 1117         }
 1118 
 1119         return (-1);
 1120 }
 1121 
 1122 /*
 1123  * Remove prefix and default router list entries corresponding to ifp. Neighbor
 1124  * cache entries are freed in in6_domifdetach().
 1125  */
 1126 void
 1127 nd6_purge(struct ifnet *ifp)
 1128 {
 1129         struct nd_prhead prl;
 1130         struct nd_prefix *pr, *npr;
 1131 
 1132         LIST_INIT(&prl);
 1133 
 1134         /* Purge default router list entries toward ifp. */
 1135         nd6_defrouter_purge(ifp);
 1136 
 1137         ND6_WLOCK();
 1138         /*
 1139          * Remove prefixes on ifp. We should have already removed addresses on
 1140          * this interface, so no addresses should be referencing these prefixes.
 1141          */
 1142         LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, npr) {
 1143                 if (pr->ndpr_ifp == ifp)
 1144                         nd6_prefix_unlink(pr, &prl);
 1145         }
 1146         ND6_WUNLOCK();
 1147 
 1148         /* Delete the unlinked prefix objects. */
 1149         while ((pr = LIST_FIRST(&prl)) != NULL) {
 1150                 LIST_REMOVE(pr, ndpr_entry);
 1151                 nd6_prefix_del(pr);
 1152         }
 1153 
 1154         /* cancel default outgoing interface setting */
 1155         if (V_nd6_defifindex == ifp->if_index)
 1156                 nd6_setdefaultiface(0);
 1157 
 1158         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 1159                 /* Refresh default router list. */
 1160                 defrouter_select_fib(ifp->if_fib);
 1161         }
 1162 }
 1163 
 1164 /* 
 1165  * the caller acquires and releases the lock on the lltbls
 1166  * Returns the llentry locked
 1167  */
 1168 struct llentry *
 1169 nd6_lookup(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
 1170 {
 1171         struct sockaddr_in6 sin6;
 1172         struct llentry *ln;
 1173         
 1174         bzero(&sin6, sizeof(sin6));
 1175         sin6.sin6_len = sizeof(struct sockaddr_in6);
 1176         sin6.sin6_family = AF_INET6;
 1177         sin6.sin6_addr = *addr6;
 1178 
 1179         IF_AFDATA_LOCK_ASSERT(ifp);
 1180 
 1181         ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)&sin6);
 1182 
 1183         return (ln);
 1184 }
 1185 
 1186 struct llentry *
 1187 nd6_alloc(const struct in6_addr *addr6, int flags, struct ifnet *ifp)
 1188 {
 1189         struct sockaddr_in6 sin6;
 1190         struct llentry *ln;
 1191 
 1192         bzero(&sin6, sizeof(sin6));
 1193         sin6.sin6_len = sizeof(struct sockaddr_in6);
 1194         sin6.sin6_family = AF_INET6;
 1195         sin6.sin6_addr = *addr6;
 1196 
 1197         ln = lltable_alloc_entry(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6);
 1198         if (ln != NULL)
 1199                 ln->ln_state = ND6_LLINFO_NOSTATE;
 1200 
 1201         return (ln);
 1202 }
 1203 
 1204 /*
 1205  * Test whether a given IPv6 address is a neighbor or not, ignoring
 1206  * the actual neighbor cache.  The neighbor cache is ignored in order
 1207  * to not reenter the routing code from within itself.
 1208  */
 1209 static int
 1210 nd6_is_new_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
 1211 {
 1212         struct nd_prefix *pr;
 1213         struct ifaddr *ifa;
 1214         struct rt_addrinfo info;
 1215         struct sockaddr_in6 rt_key;
 1216         const struct sockaddr *dst6;
 1217         uint64_t genid;
 1218         int error, fibnum;
 1219 
 1220         /*
 1221          * A link-local address is always a neighbor.
 1222          * XXX: a link does not necessarily specify a single interface.
 1223          */
 1224         if (IN6_IS_ADDR_LINKLOCAL(&addr->sin6_addr)) {
 1225                 struct sockaddr_in6 sin6_copy;
 1226                 u_int32_t zone;
 1227 
 1228                 /*
 1229                  * We need sin6_copy since sa6_recoverscope() may modify the
 1230                  * content (XXX).
 1231                  */
 1232                 sin6_copy = *addr;
 1233                 if (sa6_recoverscope(&sin6_copy))
 1234                         return (0); /* XXX: should be impossible */
 1235                 if (in6_setscope(&sin6_copy.sin6_addr, ifp, &zone))
 1236                         return (0);
 1237                 if (sin6_copy.sin6_scope_id == zone)
 1238                         return (1);
 1239                 else
 1240                         return (0);
 1241         }
 1242 
 1243         bzero(&rt_key, sizeof(rt_key));
 1244         bzero(&info, sizeof(info));
 1245         info.rti_info[RTAX_DST] = (struct sockaddr *)&rt_key;
 1246 
 1247         /*
 1248          * If the address matches one of our addresses,
 1249          * it should be a neighbor.
 1250          * If the address matches one of our on-link prefixes, it should be a
 1251          * neighbor.
 1252          */
 1253         ND6_RLOCK();
 1254 restart:
 1255         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 1256                 if (pr->ndpr_ifp != ifp)
 1257                         continue;
 1258 
 1259                 if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
 1260                         dst6 = (const struct sockaddr *)&pr->ndpr_prefix;
 1261 
 1262                         /*
 1263                          * We only need to check all FIBs if add_addr_allfibs
 1264                          * is unset. If set, checking any FIB will suffice.
 1265                          */
 1266                         fibnum = V_rt_add_addr_allfibs ? rt_numfibs - 1 : 0;
 1267                         for (; fibnum < rt_numfibs; fibnum++) {
 1268                                 genid = V_nd6_list_genid;
 1269                                 ND6_RUNLOCK();
 1270 
 1271                                 /*
 1272                                  * Restore length field before
 1273                                  * retrying lookup
 1274                                  */
 1275                                 rt_key.sin6_len = sizeof(rt_key);
 1276                                 error = rib_lookup_info(fibnum, dst6, 0, 0,
 1277                                                         &info);
 1278 
 1279                                 ND6_RLOCK();
 1280                                 if (genid != V_nd6_list_genid)
 1281                                         goto restart;
 1282                                 if (error == 0)
 1283                                         break;
 1284                         }
 1285                         if (error != 0)
 1286                                 continue;
 1287 
 1288                         /*
 1289                          * This is the case where multiple interfaces
 1290                          * have the same prefix, but only one is installed 
 1291                          * into the routing table and that prefix entry
 1292                          * is not the one being examined here. In the case
 1293                          * where RADIX_MPATH is enabled, multiple route
 1294                          * entries (of the same rt_key value) will be 
 1295                          * installed because the interface addresses all
 1296                          * differ.
 1297                          */
 1298                         if (!IN6_ARE_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
 1299                             &rt_key.sin6_addr))
 1300                                 continue;
 1301                 }
 1302 
 1303                 if (IN6_ARE_MASKED_ADDR_EQUAL(&pr->ndpr_prefix.sin6_addr,
 1304                     &addr->sin6_addr, &pr->ndpr_mask)) {
 1305                         ND6_RUNLOCK();
 1306                         return (1);
 1307                 }
 1308         }
 1309         ND6_RUNLOCK();
 1310 
 1311         /*
 1312          * If the address is assigned on the node of the other side of
 1313          * a p2p interface, the address should be a neighbor.
 1314          */
 1315         if (ifp->if_flags & IFF_POINTOPOINT) {
 1316                 IF_ADDR_RLOCK(ifp);
 1317                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1318                         if (ifa->ifa_addr->sa_family != addr->sin6_family)
 1319                                 continue;
 1320                         if (ifa->ifa_dstaddr != NULL &&
 1321                             sa_equal(addr, ifa->ifa_dstaddr)) {
 1322                                 IF_ADDR_RUNLOCK(ifp);
 1323                                 return 1;
 1324                         }
 1325                 }
 1326                 IF_ADDR_RUNLOCK(ifp);
 1327         }
 1328 
 1329         /*
 1330          * If the default router list is empty, all addresses are regarded
 1331          * as on-link, and thus, as a neighbor.
 1332          */
 1333         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV &&
 1334             nd6_defrouter_list_empty() &&
 1335             V_nd6_defifindex == ifp->if_index) {
 1336                 return (1);
 1337         }
 1338 
 1339         return (0);
 1340 }
 1341 
 1342 
 1343 /*
 1344  * Detect if a given IPv6 address identifies a neighbor on a given link.
 1345  * XXX: should take care of the destination of a p2p link?
 1346  */
 1347 int
 1348 nd6_is_addr_neighbor(const struct sockaddr_in6 *addr, struct ifnet *ifp)
 1349 {
 1350         struct llentry *lle;
 1351         int rc = 0;
 1352 
 1353         IF_AFDATA_UNLOCK_ASSERT(ifp);
 1354         if (nd6_is_new_addr_neighbor(addr, ifp))
 1355                 return (1);
 1356 
 1357         /*
 1358          * Even if the address matches none of our addresses, it might be
 1359          * in the neighbor cache.
 1360          */
 1361         IF_AFDATA_RLOCK(ifp);
 1362         if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) {
 1363                 LLE_RUNLOCK(lle);
 1364                 rc = 1;
 1365         }
 1366         IF_AFDATA_RUNLOCK(ifp);
 1367         return (rc);
 1368 }
 1369 
 1370 /*
 1371  * Free an nd6 llinfo entry.
 1372  * Since the function would cause significant changes in the kernel, DO NOT
 1373  * make it global, unless you have a strong reason for the change, and are sure
 1374  * that the change is safe.
 1375  *
 1376  * Set noinline to be dtrace-friendly
 1377  */
 1378 static __noinline void
 1379 nd6_free(struct llentry **lnp, int gc)
 1380 {
 1381         struct ifnet *ifp;
 1382         struct llentry *ln;
 1383         struct nd_defrouter *dr;
 1384 
 1385         ln = *lnp;
 1386         *lnp = NULL;
 1387 
 1388         LLE_WLOCK_ASSERT(ln);
 1389         ND6_RLOCK_ASSERT();
 1390 
 1391         ifp = lltable_get_ifp(ln->lle_tbl);
 1392         if ((ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) != 0)
 1393                 dr = defrouter_lookup_locked(&ln->r_l3addr.addr6, ifp);
 1394         else
 1395                 dr = NULL;
 1396         ND6_RUNLOCK();
 1397 
 1398         if ((ln->la_flags & LLE_DELETED) == 0)
 1399                 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_EXPIRED);
 1400 
 1401         /*
 1402          * we used to have pfctlinput(PRC_HOSTDEAD) here.
 1403          * even though it is not harmful, it was not really necessary.
 1404          */
 1405 
 1406         /* cancel timer */
 1407         nd6_llinfo_settimer_locked(ln, -1);
 1408 
 1409         if (ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 1410                 if (dr != NULL && dr->expire &&
 1411                     ln->ln_state == ND6_LLINFO_STALE && gc) {
 1412                         /*
 1413                          * If the reason for the deletion is just garbage
 1414                          * collection, and the neighbor is an active default
 1415                          * router, do not delete it.  Instead, reset the GC
 1416                          * timer using the router's lifetime.
 1417                          * Simply deleting the entry would affect default
 1418                          * router selection, which is not necessarily a good
 1419                          * thing, especially when we're using router preference
 1420                          * values.
 1421                          * XXX: the check for ln_state would be redundant,
 1422                          *      but we intentionally keep it just in case.
 1423                          */
 1424                         if (dr->expire > time_uptime)
 1425                                 nd6_llinfo_settimer_locked(ln,
 1426                                     (dr->expire - time_uptime) * hz);
 1427                         else
 1428                                 nd6_llinfo_settimer_locked(ln,
 1429                                     (long)V_nd6_gctimer * hz);
 1430 
 1431                         LLE_REMREF(ln);
 1432                         LLE_WUNLOCK(ln);
 1433                         defrouter_rele(dr);
 1434                         return;
 1435                 }
 1436 
 1437                 if (dr) {
 1438                         /*
 1439                          * Unreachability of a router might affect the default
 1440                          * router selection and on-link detection of advertised
 1441                          * prefixes.
 1442                          */
 1443 
 1444                         /*
 1445                          * Temporarily fake the state to choose a new default
 1446                          * router and to perform on-link determination of
 1447                          * prefixes correctly.
 1448                          * Below the state will be set correctly,
 1449                          * or the entry itself will be deleted.
 1450                          */
 1451                         ln->ln_state = ND6_LLINFO_INCOMPLETE;
 1452                 }
 1453 
 1454                 if (ln->ln_router || dr) {
 1455 
 1456                         /*
 1457                          * We need to unlock to avoid a LOR with rt6_flush() with the
 1458                          * rnh and for the calls to pfxlist_onlink_check() and
 1459                          * defrouter_select_fib() in the block further down for calls
 1460                          * into nd6_lookup().  We still hold a ref.
 1461                          */
 1462                         LLE_WUNLOCK(ln);
 1463 
 1464                         /*
 1465                          * rt6_flush must be called whether or not the neighbor
 1466                          * is in the Default Router List.
 1467                          * See a corresponding comment in nd6_na_input().
 1468                          */
 1469                         rt6_flush(&ln->r_l3addr.addr6, ifp);
 1470                 }
 1471 
 1472                 if (dr) {
 1473                         /*
 1474                          * Since defrouter_select_fib() does not affect the
 1475                          * on-link determination and MIP6 needs the check
 1476                          * before the default router selection, we perform
 1477                          * the check now.
 1478                          */
 1479                         pfxlist_onlink_check();
 1480 
 1481                         /*
 1482                          * Refresh default router list.
 1483                          */
 1484                         defrouter_select_fib(dr->ifp->if_fib);
 1485                 }
 1486 
 1487                 /*
 1488                  * If this entry was added by an on-link redirect, remove the
 1489                  * corresponding host route.
 1490                  */
 1491                 if (ln->la_flags & LLE_REDIRECT)
 1492                         nd6_free_redirect(ln);
 1493 
 1494                 if (ln->ln_router || dr)
 1495                         LLE_WLOCK(ln);
 1496         }
 1497 
 1498         /*
 1499          * Save to unlock. We still hold an extra reference and will not
 1500          * free(9) in llentry_free() if someone else holds one as well.
 1501          */
 1502         LLE_WUNLOCK(ln);
 1503         IF_AFDATA_LOCK(ifp);
 1504         LLE_WLOCK(ln);
 1505         /* Guard against race with other llentry_free(). */
 1506         if (ln->la_flags & LLE_LINKED) {
 1507                 /* Remove callout reference */
 1508                 LLE_REMREF(ln);
 1509                 lltable_unlink_entry(ln->lle_tbl, ln);
 1510         }
 1511         IF_AFDATA_UNLOCK(ifp);
 1512 
 1513         llentry_free(ln);
 1514         if (dr != NULL)
 1515                 defrouter_rele(dr);
 1516 }
 1517 
 1518 static int
 1519 nd6_isdynrte(const struct rtentry *rt, void *xap)
 1520 {
 1521 
 1522         if (rt->rt_flags == (RTF_UP | RTF_HOST | RTF_DYNAMIC))
 1523                 return (1);
 1524 
 1525         return (0);
 1526 }
 1527 /*
 1528  * Remove the rtentry for the given llentry,
 1529  * both of which were installed by a redirect.
 1530  */
 1531 static void
 1532 nd6_free_redirect(const struct llentry *ln)
 1533 {
 1534         int fibnum;
 1535         struct sockaddr_in6 sin6;
 1536         struct rt_addrinfo info;
 1537 
 1538         lltable_fill_sa_entry(ln, (struct sockaddr *)&sin6);
 1539         memset(&info, 0, sizeof(info));
 1540         info.rti_info[RTAX_DST] = (struct sockaddr *)&sin6;
 1541         info.rti_filter = nd6_isdynrte;
 1542 
 1543         for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
 1544                 rtrequest1_fib(RTM_DELETE, &info, NULL, fibnum);
 1545 }
 1546 
 1547 /*
 1548  * Rejuvenate this function for routing operations related
 1549  * processing.
 1550  */
 1551 void
 1552 nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info)
 1553 {
 1554         struct sockaddr_in6 *gateway;
 1555         struct nd_defrouter *dr;
 1556         struct ifnet *ifp;
 1557 
 1558         gateway = (struct sockaddr_in6 *)rt->rt_gateway;
 1559         ifp = rt->rt_ifp;
 1560 
 1561         switch (req) {
 1562         case RTM_ADD:
 1563                 break;
 1564 
 1565         case RTM_DELETE:
 1566                 if (!ifp)
 1567                         return;
 1568                 /*
 1569                  * Only indirect routes are interesting.
 1570                  */
 1571                 if ((rt->rt_flags & RTF_GATEWAY) == 0)
 1572                         return;
 1573                 /*
 1574                  * check for default route
 1575                  */
 1576                 if (IN6_ARE_ADDR_EQUAL(&in6addr_any,
 1577                     &SIN6(rt_key(rt))->sin6_addr)) {
 1578                         dr = defrouter_lookup(&gateway->sin6_addr, ifp);
 1579                         if (dr != NULL) {
 1580                                 dr->installed = 0;
 1581                                 defrouter_rele(dr);
 1582                         }
 1583                 }
 1584                 break;
 1585         }
 1586 }
 1587 
 1588 
 1589 int
 1590 nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp)
 1591 {
 1592         struct in6_ndireq *ndi = (struct in6_ndireq *)data;
 1593         struct in6_nbrinfo *nbi = (struct in6_nbrinfo *)data;
 1594         struct in6_ndifreq *ndif = (struct in6_ndifreq *)data;
 1595         int error = 0;
 1596 
 1597         if (ifp->if_afdata[AF_INET6] == NULL)
 1598                 return (EPFNOSUPPORT);
 1599         switch (cmd) {
 1600         case OSIOCGIFINFO_IN6:
 1601 #define ND      ndi->ndi
 1602                 /* XXX: old ndp(8) assumes a positive value for linkmtu. */
 1603                 bzero(&ND, sizeof(ND));
 1604                 ND.linkmtu = IN6_LINKMTU(ifp);
 1605                 ND.maxmtu = ND_IFINFO(ifp)->maxmtu;
 1606                 ND.basereachable = ND_IFINFO(ifp)->basereachable;
 1607                 ND.reachable = ND_IFINFO(ifp)->reachable;
 1608                 ND.retrans = ND_IFINFO(ifp)->retrans;
 1609                 ND.flags = ND_IFINFO(ifp)->flags;
 1610                 ND.recalctm = ND_IFINFO(ifp)->recalctm;
 1611                 ND.chlim = ND_IFINFO(ifp)->chlim;
 1612                 break;
 1613         case SIOCGIFINFO_IN6:
 1614                 ND = *ND_IFINFO(ifp);
 1615                 break;
 1616         case SIOCSIFINFO_IN6:
 1617                 /*
 1618                  * used to change host variables from userland.
 1619                  * intended for a use on router to reflect RA configurations.
 1620                  */
 1621                 /* 0 means 'unspecified' */
 1622                 if (ND.linkmtu != 0) {
 1623                         if (ND.linkmtu < IPV6_MMTU ||
 1624                             ND.linkmtu > IN6_LINKMTU(ifp)) {
 1625                                 error = EINVAL;
 1626                                 break;
 1627                         }
 1628                         ND_IFINFO(ifp)->linkmtu = ND.linkmtu;
 1629                 }
 1630 
 1631                 if (ND.basereachable != 0) {
 1632                         int obasereachable = ND_IFINFO(ifp)->basereachable;
 1633 
 1634                         ND_IFINFO(ifp)->basereachable = ND.basereachable;
 1635                         if (ND.basereachable != obasereachable)
 1636                                 ND_IFINFO(ifp)->reachable =
 1637                                     ND_COMPUTE_RTIME(ND.basereachable);
 1638                 }
 1639                 if (ND.retrans != 0)
 1640                         ND_IFINFO(ifp)->retrans = ND.retrans;
 1641                 if (ND.chlim != 0)
 1642                         ND_IFINFO(ifp)->chlim = ND.chlim;
 1643                 /* FALLTHROUGH */
 1644         case SIOCSIFINFO_FLAGS:
 1645         {
 1646                 struct ifaddr *ifa;
 1647                 struct in6_ifaddr *ia;
 1648 
 1649                 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
 1650                     !(ND.flags & ND6_IFF_IFDISABLED)) {
 1651                         /* ifdisabled 1->0 transision */
 1652 
 1653                         /*
 1654                          * If the interface is marked as ND6_IFF_IFDISABLED and
 1655                          * has an link-local address with IN6_IFF_DUPLICATED,
 1656                          * do not clear ND6_IFF_IFDISABLED.
 1657                          * See RFC 4862, Section 5.4.5.
 1658                          */
 1659                         IF_ADDR_RLOCK(ifp);
 1660                         CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1661                                 if (ifa->ifa_addr->sa_family != AF_INET6)
 1662                                         continue;
 1663                                 ia = (struct in6_ifaddr *)ifa;
 1664                                 if ((ia->ia6_flags & IN6_IFF_DUPLICATED) &&
 1665                                     IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
 1666                                         break;
 1667                         }
 1668                         IF_ADDR_RUNLOCK(ifp);
 1669 
 1670                         if (ifa != NULL) {
 1671                                 /* LLA is duplicated. */
 1672                                 ND.flags |= ND6_IFF_IFDISABLED;
 1673                                 log(LOG_ERR, "Cannot enable an interface"
 1674                                     " with a link-local address marked"
 1675                                     " duplicate.\n");
 1676                         } else {
 1677                                 ND_IFINFO(ifp)->flags &= ~ND6_IFF_IFDISABLED;
 1678                                 if (ifp->if_flags & IFF_UP)
 1679                                         in6_if_up(ifp);
 1680                         }
 1681                 } else if (!(ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) &&
 1682                             (ND.flags & ND6_IFF_IFDISABLED)) {
 1683                         /* ifdisabled 0->1 transision */
 1684                         /* Mark all IPv6 address as tentative. */
 1685 
 1686                         ND_IFINFO(ifp)->flags |= ND6_IFF_IFDISABLED;
 1687                         if (V_ip6_dad_count > 0 &&
 1688                             (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD) == 0) {
 1689                                 IF_ADDR_RLOCK(ifp);
 1690                                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
 1691                                     ifa_link) {
 1692                                         if (ifa->ifa_addr->sa_family !=
 1693                                             AF_INET6)
 1694                                                 continue;
 1695                                         ia = (struct in6_ifaddr *)ifa;
 1696                                         ia->ia6_flags |= IN6_IFF_TENTATIVE;
 1697                                 }
 1698                                 IF_ADDR_RUNLOCK(ifp);
 1699                         }
 1700                 }
 1701 
 1702                 if (ND.flags & ND6_IFF_AUTO_LINKLOCAL) {
 1703                         if (!(ND_IFINFO(ifp)->flags & ND6_IFF_AUTO_LINKLOCAL)) {
 1704                                 /* auto_linklocal 0->1 transision */
 1705 
 1706                                 /* If no link-local address on ifp, configure */
 1707                                 ND_IFINFO(ifp)->flags |= ND6_IFF_AUTO_LINKLOCAL;
 1708                                 in6_ifattach(ifp, NULL);
 1709                         } else if (!(ND.flags & ND6_IFF_IFDISABLED) &&
 1710                             ifp->if_flags & IFF_UP) {
 1711                                 /*
 1712                                  * When the IF already has
 1713                                  * ND6_IFF_AUTO_LINKLOCAL, no link-local
 1714                                  * address is assigned, and IFF_UP, try to
 1715                                  * assign one.
 1716                                  */
 1717                                 IF_ADDR_RLOCK(ifp);
 1718                                 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead,
 1719                                     ifa_link) {
 1720                                         if (ifa->ifa_addr->sa_family !=
 1721                                             AF_INET6)
 1722                                                 continue;
 1723                                         ia = (struct in6_ifaddr *)ifa;
 1724                                         if (IN6_IS_ADDR_LINKLOCAL(IA6_IN6(ia)))
 1725                                                 break;
 1726                                 }
 1727                                 IF_ADDR_RUNLOCK(ifp);
 1728                                 if (ifa != NULL)
 1729                                         /* No LLA is configured. */
 1730                                         in6_ifattach(ifp, NULL);
 1731                         }
 1732                 }
 1733         }
 1734                 ND_IFINFO(ifp)->flags = ND.flags;
 1735                 break;
 1736 #undef ND
 1737         case SIOCSNDFLUSH_IN6:  /* XXX: the ioctl name is confusing... */
 1738                 /* sync kernel routing table with the default router list */
 1739                 defrouter_reset();
 1740                 defrouter_select();
 1741                 break;
 1742         case SIOCSPFXFLUSH_IN6:
 1743         {
 1744                 /* flush all the prefix advertised by routers */
 1745                 struct in6_ifaddr *ia, *ia_next;
 1746                 struct nd_prefix *pr, *next;
 1747                 struct nd_prhead prl;
 1748 
 1749                 LIST_INIT(&prl);
 1750 
 1751                 ND6_WLOCK();
 1752                 LIST_FOREACH_SAFE(pr, &V_nd_prefix, ndpr_entry, next) {
 1753                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1754                                 continue; /* XXX */
 1755                         nd6_prefix_unlink(pr, &prl);
 1756                 }
 1757                 ND6_WUNLOCK();
 1758 
 1759                 while ((pr = LIST_FIRST(&prl)) != NULL) {
 1760                         LIST_REMOVE(pr, ndpr_entry);
 1761                         /* XXXRW: in6_ifaddrhead locking. */
 1762                         CK_STAILQ_FOREACH_SAFE(ia, &V_in6_ifaddrhead, ia_link,
 1763                             ia_next) {
 1764                                 if ((ia->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1765                                         continue;
 1766 
 1767                                 if (ia->ia6_ndpr == pr)
 1768                                         in6_purgeaddr(&ia->ia_ifa);
 1769                         }
 1770                         nd6_prefix_del(pr);
 1771                 }
 1772                 break;
 1773         }
 1774         case SIOCSRTRFLUSH_IN6:
 1775         {
 1776                 /* flush all the default routers */
 1777 
 1778                 defrouter_reset();
 1779                 nd6_defrouter_flush_all();
 1780                 defrouter_select();
 1781                 break;
 1782         }
 1783         case SIOCGNBRINFO_IN6:
 1784         {
 1785                 struct llentry *ln;
 1786                 struct in6_addr nb_addr = nbi->addr; /* make local for safety */
 1787 
 1788                 if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0)
 1789                         return (error);
 1790 
 1791                 IF_AFDATA_RLOCK(ifp);
 1792                 ln = nd6_lookup(&nb_addr, 0, ifp);
 1793                 IF_AFDATA_RUNLOCK(ifp);
 1794 
 1795                 if (ln == NULL) {
 1796                         error = EINVAL;
 1797                         break;
 1798                 }
 1799                 nbi->state = ln->ln_state;
 1800                 nbi->asked = ln->la_asked;
 1801                 nbi->isrouter = ln->ln_router;
 1802                 if (ln->la_expire == 0)
 1803                         nbi->expire = 0;
 1804                 else
 1805                         nbi->expire = ln->la_expire + ln->lle_remtime / hz +
 1806                             (time_second - time_uptime);
 1807                 LLE_RUNLOCK(ln);
 1808                 break;
 1809         }
 1810         case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1811                 ndif->ifindex = V_nd6_defifindex;
 1812                 break;
 1813         case SIOCSDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */
 1814                 return (nd6_setdefaultiface(ndif->ifindex));
 1815         }
 1816         return (error);
 1817 }
 1818 
 1819 /*
 1820  * Calculates new isRouter value based on provided parameters and
 1821  * returns it.
 1822  */
 1823 static int
 1824 nd6_is_router(int type, int code, int is_new, int old_addr, int new_addr,
 1825     int ln_router)
 1826 {
 1827 
 1828         /*
 1829          * ICMP6 type dependent behavior.
 1830          *
 1831          * NS: clear IsRouter if new entry
 1832          * RS: clear IsRouter
 1833          * RA: set IsRouter if there's lladdr
 1834          * redir: clear IsRouter if new entry
 1835          *
 1836          * RA case, (1):
 1837          * The spec says that we must set IsRouter in the following cases:
 1838          * - If lladdr exist, set IsRouter.  This means (1-5).
 1839          * - If it is old entry (!newentry), set IsRouter.  This means (7).
 1840          * So, based on the spec, in (1-5) and (7) cases we must set IsRouter.
 1841          * A quetion arises for (1) case.  (1) case has no lladdr in the
 1842          * neighbor cache, this is similar to (6).
 1843          * This case is rare but we figured that we MUST NOT set IsRouter.
 1844          *
 1845          *   is_new  old_addr new_addr      NS  RS  RA  redir
 1846          *                                                      D R
 1847          *      0       n       n       (1)     c   ?     s
 1848          *      0       y       n       (2)     c   s     s
 1849          *      0       n       y       (3)     c   s     s
 1850          *      0       y       y       (4)     c   s     s
 1851          *      0       y       y       (5)     c   s     s
 1852          *      1       --      n       (6) c   c       c s
 1853          *      1       --      y       (7) c   c   s   c s
 1854          *
 1855          *                                      (c=clear s=set)
 1856          */
 1857         switch (type & 0xff) {
 1858         case ND_NEIGHBOR_SOLICIT:
 1859                 /*
 1860                  * New entry must have is_router flag cleared.
 1861                  */
 1862                 if (is_new)                                     /* (6-7) */
 1863                         ln_router = 0;
 1864                 break;
 1865         case ND_REDIRECT:
 1866                 /*
 1867                  * If the icmp is a redirect to a better router, always set the
 1868                  * is_router flag.  Otherwise, if the entry is newly created,
 1869                  * clear the flag.  [RFC 2461, sec 8.3]
 1870                  */
 1871                 if (code == ND_REDIRECT_ROUTER)
 1872                         ln_router = 1;
 1873                 else {
 1874                         if (is_new)                             /* (6-7) */
 1875                                 ln_router = 0;
 1876                 }
 1877                 break;
 1878         case ND_ROUTER_SOLICIT:
 1879                 /*
 1880                  * is_router flag must always be cleared.
 1881                  */
 1882                 ln_router = 0;
 1883                 break;
 1884         case ND_ROUTER_ADVERT:
 1885                 /*
 1886                  * Mark an entry with lladdr as a router.
 1887                  */
 1888                 if ((!is_new && (old_addr || new_addr)) ||      /* (2-5) */
 1889                     (is_new && new_addr)) {                     /* (7) */
 1890                         ln_router = 1;
 1891                 }
 1892                 break;
 1893         }
 1894 
 1895         return (ln_router);
 1896 }
 1897 
 1898 /*
 1899  * Create neighbor cache entry and cache link-layer address,
 1900  * on reception of inbound ND6 packets.  (RS/RA/NS/redirect)
 1901  *
 1902  * type - ICMP6 type
 1903  * code - type dependent information
 1904  *
 1905  */
 1906 void
 1907 nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr,
 1908     int lladdrlen, int type, int code)
 1909 {
 1910         struct llentry *ln = NULL, *ln_tmp;
 1911         int is_newentry;
 1912         int do_update;
 1913         int olladdr;
 1914         int llchange;
 1915         int flags;
 1916         uint16_t router = 0;
 1917         struct sockaddr_in6 sin6;
 1918         struct mbuf *chain = NULL;
 1919         u_char linkhdr[LLE_MAX_LINKHDR];
 1920         size_t linkhdrsize;
 1921         int lladdr_off;
 1922 
 1923         IF_AFDATA_UNLOCK_ASSERT(ifp);
 1924 
 1925         KASSERT(ifp != NULL, ("%s: ifp == NULL", __func__));
 1926         KASSERT(from != NULL, ("%s: from == NULL", __func__));
 1927 
 1928         /* nothing must be updated for unspecified address */
 1929         if (IN6_IS_ADDR_UNSPECIFIED(from))
 1930                 return;
 1931 
 1932         /*
 1933          * Validation about ifp->if_addrlen and lladdrlen must be done in
 1934          * the caller.
 1935          *
 1936          * XXX If the link does not have link-layer adderss, what should
 1937          * we do? (ifp->if_addrlen == 0)
 1938          * Spec says nothing in sections for RA, RS and NA.  There's small
 1939          * description on it in NS section (RFC 2461 7.2.3).
 1940          */
 1941         flags = lladdr ? LLE_EXCLUSIVE : 0;
 1942         IF_AFDATA_RLOCK(ifp);
 1943         ln = nd6_lookup(from, flags, ifp);
 1944         IF_AFDATA_RUNLOCK(ifp);
 1945         is_newentry = 0;
 1946         if (ln == NULL) {
 1947                 flags |= LLE_EXCLUSIVE;
 1948                 ln = nd6_alloc(from, 0, ifp);
 1949                 if (ln == NULL)
 1950                         return;
 1951 
 1952                 /*
 1953                  * Since we already know all the data for the new entry,
 1954                  * fill it before insertion.
 1955                  */
 1956                 if (lladdr != NULL) {
 1957                         linkhdrsize = sizeof(linkhdr);
 1958                         if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
 1959                             linkhdr, &linkhdrsize, &lladdr_off) != 0)
 1960                                 return;
 1961                         lltable_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
 1962                             lladdr_off);
 1963                 }
 1964 
 1965                 IF_AFDATA_WLOCK(ifp);
 1966                 LLE_WLOCK(ln);
 1967                 /* Prefer any existing lle over newly-created one */
 1968                 ln_tmp = nd6_lookup(from, LLE_EXCLUSIVE, ifp);
 1969                 if (ln_tmp == NULL)
 1970                         lltable_link_entry(LLTABLE6(ifp), ln);
 1971                 IF_AFDATA_WUNLOCK(ifp);
 1972                 if (ln_tmp == NULL) {
 1973                         /* No existing lle, mark as new entry (6,7) */
 1974                         is_newentry = 1;
 1975                         if (lladdr != NULL) {   /* (7) */
 1976                                 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
 1977                                 EVENTHANDLER_INVOKE(lle_event, ln,
 1978                                     LLENTRY_RESOLVED);
 1979                         }
 1980                 } else {
 1981                         lltable_free_entry(LLTABLE6(ifp), ln);
 1982                         ln = ln_tmp;
 1983                         ln_tmp = NULL;
 1984                 }
 1985         } 
 1986         /* do nothing if static ndp is set */
 1987         if ((ln->la_flags & LLE_STATIC)) {
 1988                 if (flags & LLE_EXCLUSIVE)
 1989                         LLE_WUNLOCK(ln);
 1990                 else
 1991                         LLE_RUNLOCK(ln);
 1992                 return;
 1993         }
 1994 
 1995         olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0;
 1996         if (olladdr && lladdr) {
 1997                 llchange = bcmp(lladdr, ln->ll_addr,
 1998                     ifp->if_addrlen);
 1999         } else if (!olladdr && lladdr)
 2000                 llchange = 1;
 2001         else
 2002                 llchange = 0;
 2003 
 2004         /*
 2005          * newentry olladdr  lladdr  llchange   (*=record)
 2006          *      0       n       n       --      (1)
 2007          *      0       y       n       --      (2)
 2008          *      0       n       y       y       (3) * STALE
 2009          *      0       y       y       n       (4) *
 2010          *      0       y       y       y       (5) * STALE
 2011          *      1       --      n       --      (6)   NOSTATE(= PASSIVE)
 2012          *      1       --      y       --      (7) * STALE
 2013          */
 2014 
 2015         do_update = 0;
 2016         if (is_newentry == 0 && llchange != 0) {
 2017                 do_update = 1;  /* (3,5) */
 2018 
 2019                 /*
 2020                  * Record source link-layer address
 2021                  * XXX is it dependent to ifp->if_type?
 2022                  */
 2023                 linkhdrsize = sizeof(linkhdr);
 2024                 if (lltable_calc_llheader(ifp, AF_INET6, lladdr,
 2025                     linkhdr, &linkhdrsize, &lladdr_off) != 0)
 2026                         return;
 2027 
 2028                 if (lltable_try_set_entry_addr(ifp, ln, linkhdr, linkhdrsize,
 2029                     lladdr_off) == 0) {
 2030                         /* Entry was deleted */
 2031                         return;
 2032                 }
 2033 
 2034                 nd6_llinfo_setstate(ln, ND6_LLINFO_STALE);
 2035 
 2036                 EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
 2037 
 2038                 if (ln->la_hold != NULL)
 2039                         nd6_grab_holdchain(ln, &chain, &sin6);
 2040         }
 2041 
 2042         /* Calculates new router status */
 2043         router = nd6_is_router(type, code, is_newentry, olladdr,
 2044             lladdr != NULL ? 1 : 0, ln->ln_router);
 2045 
 2046         ln->ln_router = router;
 2047         /* Mark non-router redirects with special flag */
 2048         if ((type & 0xFF) == ND_REDIRECT && code != ND_REDIRECT_ROUTER)
 2049                 ln->la_flags |= LLE_REDIRECT;
 2050 
 2051         if (flags & LLE_EXCLUSIVE)
 2052                 LLE_WUNLOCK(ln);
 2053         else
 2054                 LLE_RUNLOCK(ln);
 2055 
 2056         if (chain != NULL)
 2057                 nd6_flush_holdchain(ifp, chain, &sin6);
 2058         
 2059         /*
 2060          * When the link-layer address of a router changes, select the
 2061          * best router again.  In particular, when the neighbor entry is newly
 2062          * created, it might affect the selection policy.
 2063          * Question: can we restrict the first condition to the "is_newentry"
 2064          * case?
 2065          * XXX: when we hear an RA from a new router with the link-layer
 2066          * address option, defrouter_select_fib() is called twice, since
 2067          * defrtrlist_update called the function as well.  However, I believe
 2068          * we can compromise the overhead, since it only happens the first
 2069          * time.
 2070          * XXX: although defrouter_select_fib() should not have a bad effect
 2071          * for those are not autoconfigured hosts, we explicitly avoid such
 2072          * cases for safety.
 2073          */
 2074         if ((do_update || is_newentry) && router &&
 2075             ND_IFINFO(ifp)->flags & ND6_IFF_ACCEPT_RTADV) {
 2076                 /*
 2077                  * guaranteed recursion
 2078                  */
 2079                 defrouter_select_fib(ifp->if_fib);
 2080         }
 2081 }
 2082 
 2083 static void
 2084 nd6_slowtimo(void *arg)
 2085 {
 2086         CURVNET_SET((struct vnet *) arg);
 2087         struct nd_ifinfo *nd6if;
 2088         struct ifnet *ifp;
 2089 
 2090         callout_reset(&V_nd6_slowtimo_ch, ND6_SLOWTIMER_INTERVAL * hz,
 2091             nd6_slowtimo, curvnet);
 2092         IFNET_RLOCK_NOSLEEP();
 2093         CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
 2094                 if (ifp->if_afdata[AF_INET6] == NULL)
 2095                         continue;
 2096                 nd6if = ND_IFINFO(ifp);
 2097                 if (nd6if->basereachable && /* already initialized */
 2098                     (nd6if->recalctm -= ND6_SLOWTIMER_INTERVAL) <= 0) {
 2099                         /*
 2100                          * Since reachable time rarely changes by router
 2101                          * advertisements, we SHOULD insure that a new random
 2102                          * value gets recomputed at least once every few hours.
 2103                          * (RFC 2461, 6.3.4)
 2104                          */
 2105                         nd6if->recalctm = V_nd6_recalc_reachtm_interval;
 2106                         nd6if->reachable = ND_COMPUTE_RTIME(nd6if->basereachable);
 2107                 }
 2108         }
 2109         IFNET_RUNLOCK_NOSLEEP();
 2110         CURVNET_RESTORE();
 2111 }
 2112 
 2113 void
 2114 nd6_grab_holdchain(struct llentry *ln, struct mbuf **chain,
 2115     struct sockaddr_in6 *sin6)
 2116 {
 2117 
 2118         LLE_WLOCK_ASSERT(ln);
 2119 
 2120         *chain = ln->la_hold;
 2121         ln->la_hold = NULL;
 2122         lltable_fill_sa_entry(ln, (struct sockaddr *)sin6);
 2123 
 2124         if (ln->ln_state == ND6_LLINFO_STALE) {
 2125 
 2126                 /*
 2127                  * The first time we send a packet to a
 2128                  * neighbor whose entry is STALE, we have
 2129                  * to change the state to DELAY and a sets
 2130                  * a timer to expire in DELAY_FIRST_PROBE_TIME
 2131                  * seconds to ensure do neighbor unreachability
 2132                  * detection on expiration.
 2133                  * (RFC 2461 7.3.3)
 2134                  */
 2135                 nd6_llinfo_setstate(ln, ND6_LLINFO_DELAY);
 2136         }
 2137 }
 2138 
 2139 int
 2140 nd6_output_ifp(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m,
 2141     struct sockaddr_in6 *dst, struct route *ro)
 2142 {
 2143         int error;
 2144         int ip6len;
 2145         struct ip6_hdr *ip6;
 2146         struct m_tag *mtag;
 2147 
 2148 #ifdef MAC
 2149         mac_netinet6_nd6_send(ifp, m);
 2150 #endif
 2151 
 2152         /*
 2153          * If called from nd6_ns_output() (NS), nd6_na_output() (NA),
 2154          * icmp6_redirect_output() (REDIRECT) or from rip6_output() (RS, RA
 2155          * as handled by rtsol and rtadvd), mbufs will be tagged for SeND
 2156          * to be diverted to user space.  When re-injected into the kernel,
 2157          * send_output() will directly dispatch them to the outgoing interface.
 2158          */
 2159         if (send_sendso_input_hook != NULL) {
 2160                 mtag = m_tag_find(m, PACKET_TAG_ND_OUTGOING, NULL);
 2161                 if (mtag != NULL) {
 2162                         ip6 = mtod(m, struct ip6_hdr *);
 2163                         ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
 2164                         /* Use the SEND socket */
 2165                         error = send_sendso_input_hook(m, ifp, SND_OUT,
 2166                             ip6len);
 2167                         /* -1 == no app on SEND socket */
 2168                         if (error == 0 || error != -1)
 2169                             return (error);
 2170                 }
 2171         }
 2172 
 2173         m_clrprotoflags(m);     /* Avoid confusing lower layers. */
 2174         IP_PROBE(send, NULL, NULL, mtod(m, struct ip6_hdr *), ifp, NULL,
 2175             mtod(m, struct ip6_hdr *));
 2176 
 2177         if ((ifp->if_flags & IFF_LOOPBACK) == 0)
 2178                 origifp = ifp;
 2179 
 2180         error = (*ifp->if_output)(origifp, m, (struct sockaddr *)dst, ro);
 2181         return (error);
 2182 }
 2183 
 2184 /*
 2185  * Lookup link headerfor @sa_dst address. Stores found
 2186  * data in @desten buffer. Copy of lle ln_flags can be also
 2187  * saved in @pflags if @pflags is non-NULL.
 2188  *
 2189  * If destination LLE does not exists or lle state modification
 2190  * is required, call "slow" version.
 2191  *
 2192  * Return values:
 2193  * - 0 on success (address copied to buffer).
 2194  * - EWOULDBLOCK (no local error, but address is still unresolved)
 2195  * - other errors (alloc failure, etc)
 2196  */
 2197 int
 2198 nd6_resolve(struct ifnet *ifp, int is_gw, struct mbuf *m,
 2199     const struct sockaddr *sa_dst, u_char *desten, uint32_t *pflags,
 2200     struct llentry **plle)
 2201 {
 2202         struct llentry *ln = NULL;
 2203         const struct sockaddr_in6 *dst6;
 2204 
 2205         if (pflags != NULL)
 2206                 *pflags = 0;
 2207 
 2208         dst6 = (const struct sockaddr_in6 *)sa_dst;
 2209 
 2210         /* discard the packet if IPv6 operation is disabled on the interface */
 2211         if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)) {
 2212                 m_freem(m);
 2213                 return (ENETDOWN); /* better error? */
 2214         }
 2215 
 2216         if (m != NULL && m->m_flags & M_MCAST) {
 2217                 switch (ifp->if_type) {
 2218                 case IFT_ETHER:
 2219                 case IFT_L2VLAN:
 2220                 case IFT_BRIDGE:
 2221                         ETHER_MAP_IPV6_MULTICAST(&dst6->sin6_addr,
 2222                                                  desten);
 2223                         return (0);
 2224                 default:
 2225                         m_freem(m);
 2226                         return (EAFNOSUPPORT);
 2227                 }
 2228         }
 2229 
 2230         IF_AFDATA_RLOCK(ifp);
 2231         ln = nd6_lookup(&dst6->sin6_addr, plle ? LLE_EXCLUSIVE : LLE_UNLOCKED,
 2232             ifp);
 2233         if (ln != NULL && (ln->r_flags & RLLE_VALID) != 0) {
 2234                 /* Entry found, let's copy lle info */
 2235                 bcopy(ln->r_linkdata, desten, ln->r_hdrlen);
 2236                 if (pflags != NULL)
 2237                         *pflags = LLE_VALID | (ln->r_flags & RLLE_IFADDR);
 2238                 /* Check if we have feedback request from nd6 timer */
 2239                 if (ln->r_skip_req != 0) {
 2240                         LLE_REQ_LOCK(ln);
 2241                         ln->r_skip_req = 0; /* Notify that entry was used */
 2242                         ln->lle_hittime = time_uptime;
 2243                         LLE_REQ_UNLOCK(ln);
 2244                 }
 2245                 if (plle) {
 2246                         LLE_ADDREF(ln);
 2247                         *plle = ln;
 2248                         LLE_WUNLOCK(ln);
 2249                 }
 2250                 IF_AFDATA_RUNLOCK(ifp);
 2251                 return (0);
 2252         } else if (plle && ln)
 2253                 LLE_WUNLOCK(ln);
 2254         IF_AFDATA_RUNLOCK(ifp);
 2255 
 2256         return (nd6_resolve_slow(ifp, 0, m, dst6, desten, pflags, plle));
 2257 }
 2258 
 2259 
 2260 /*
 2261  * Do L2 address resolution for @sa_dst address. Stores found
 2262  * address in @desten buffer. Copy of lle ln_flags can be also
 2263  * saved in @pflags if @pflags is non-NULL.
 2264  *
 2265  * Heavy version.
 2266  * Function assume that destination LLE does not exist,
 2267  * is invalid or stale, so LLE_EXCLUSIVE lock needs to be acquired.
 2268  *
 2269  * Set noinline to be dtrace-friendly
 2270  */
 2271 static __noinline int
 2272 nd6_resolve_slow(struct ifnet *ifp, int flags, struct mbuf *m,
 2273     const struct sockaddr_in6 *dst, u_char *desten, uint32_t *pflags,
 2274     struct llentry **plle)
 2275 {
 2276         struct llentry *lle = NULL, *lle_tmp;
 2277         struct in6_addr *psrc, src;
 2278         int send_ns, ll_len;
 2279         char *lladdr;
 2280 
 2281         /*
 2282          * Address resolution or Neighbor Unreachability Detection
 2283          * for the next hop.
 2284          * At this point, the destination of the packet must be a unicast
 2285          * or an anycast address(i.e. not a multicast).
 2286          */
 2287         if (lle == NULL) {
 2288                 IF_AFDATA_RLOCK(ifp);
 2289                 lle = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
 2290                 IF_AFDATA_RUNLOCK(ifp);
 2291                 if ((lle == NULL) && nd6_is_addr_neighbor(dst, ifp))  {
 2292                         /*
 2293                          * Since nd6_is_addr_neighbor() internally calls nd6_lookup(),
 2294                          * the condition below is not very efficient.  But we believe
 2295                          * it is tolerable, because this should be a rare case.
 2296                          */
 2297                         lle = nd6_alloc(&dst->sin6_addr, 0, ifp);
 2298                         if (lle == NULL) {
 2299                                 char ip6buf[INET6_ADDRSTRLEN];
 2300                                 log(LOG_DEBUG,
 2301                                     "nd6_output: can't allocate llinfo for %s "
 2302                                     "(ln=%p)\n",
 2303                                     ip6_sprintf(ip6buf, &dst->sin6_addr), lle);
 2304                                 m_freem(m);
 2305                                 return (ENOBUFS);
 2306                         }
 2307 
 2308                         IF_AFDATA_WLOCK(ifp);
 2309                         LLE_WLOCK(lle);
 2310                         /* Prefer any existing entry over newly-created one */
 2311                         lle_tmp = nd6_lookup(&dst->sin6_addr, LLE_EXCLUSIVE, ifp);
 2312                         if (lle_tmp == NULL)
 2313                                 lltable_link_entry(LLTABLE6(ifp), lle);
 2314                         IF_AFDATA_WUNLOCK(ifp);
 2315                         if (lle_tmp != NULL) {
 2316                                 lltable_free_entry(LLTABLE6(ifp), lle);
 2317                                 lle = lle_tmp;
 2318                                 lle_tmp = NULL;
 2319                         }
 2320                 }
 2321         } 
 2322         if (lle == NULL) {
 2323                 m_freem(m);
 2324                 return (ENOBUFS);
 2325         }
 2326 
 2327         LLE_WLOCK_ASSERT(lle);
 2328 
 2329         /*
 2330          * The first time we send a packet to a neighbor whose entry is
 2331          * STALE, we have to change the state to DELAY and a sets a timer to
 2332          * expire in DELAY_FIRST_PROBE_TIME seconds to ensure do
 2333          * neighbor unreachability detection on expiration.
 2334          * (RFC 2461 7.3.3)
 2335          */
 2336         if (lle->ln_state == ND6_LLINFO_STALE)
 2337                 nd6_llinfo_setstate(lle, ND6_LLINFO_DELAY);
 2338 
 2339         /*
 2340          * If the neighbor cache entry has a state other than INCOMPLETE
 2341          * (i.e. its link-layer address is already resolved), just
 2342          * send the packet.
 2343          */
 2344         if (lle->ln_state > ND6_LLINFO_INCOMPLETE) {
 2345                 if (flags & LLE_ADDRONLY) {
 2346                         lladdr = lle->ll_addr;
 2347                         ll_len = ifp->if_addrlen;
 2348                 } else {
 2349                         lladdr = lle->r_linkdata;
 2350                         ll_len = lle->r_hdrlen;
 2351                 }
 2352                 bcopy(lladdr, desten, ll_len);
 2353                 if (pflags != NULL)
 2354                         *pflags = lle->la_flags;
 2355                 if (plle) {
 2356                         LLE_ADDREF(lle);
 2357                         *plle = lle;
 2358                 }
 2359                 LLE_WUNLOCK(lle);
 2360                 return (0);
 2361         }
 2362 
 2363         /*
 2364          * There is a neighbor cache entry, but no ethernet address
 2365          * response yet.  Append this latest packet to the end of the
 2366          * packet queue in the mbuf.  When it exceeds nd6_maxqueuelen,
 2367          * the oldest packet in the queue will be removed.
 2368          */
 2369 
 2370         if (lle->la_hold != NULL) {
 2371                 struct mbuf *m_hold;
 2372                 int i;
 2373                 
 2374                 i = 0;
 2375                 for (m_hold = lle->la_hold; m_hold; m_hold = m_hold->m_nextpkt){
 2376                         i++;
 2377                         if (m_hold->m_nextpkt == NULL) {
 2378                                 m_hold->m_nextpkt = m;
 2379                                 break;
 2380                         }
 2381                 }
 2382                 while (i >= V_nd6_maxqueuelen) {
 2383                         m_hold = lle->la_hold;
 2384                         lle->la_hold = lle->la_hold->m_nextpkt;
 2385                         m_freem(m_hold);
 2386                         i--;
 2387                 }
 2388         } else {
 2389                 lle->la_hold = m;
 2390         }
 2391 
 2392         /*
 2393          * If there has been no NS for the neighbor after entering the
 2394          * INCOMPLETE state, send the first solicitation.
 2395          * Note that for newly-created lle la_asked will be 0,
 2396          * so we will transition from ND6_LLINFO_NOSTATE to
 2397          * ND6_LLINFO_INCOMPLETE state here.
 2398          */
 2399         psrc = NULL;
 2400         send_ns = 0;
 2401         if (lle->la_asked == 0) {
 2402                 lle->la_asked++;
 2403                 send_ns = 1;
 2404                 psrc = nd6_llinfo_get_holdsrc(lle, &src);
 2405 
 2406                 nd6_llinfo_setstate(lle, ND6_LLINFO_INCOMPLETE);
 2407         }
 2408         LLE_WUNLOCK(lle);
 2409         if (send_ns != 0)
 2410                 nd6_ns_output(ifp, psrc, NULL, &dst->sin6_addr, NULL);
 2411 
 2412         return (EWOULDBLOCK);
 2413 }
 2414 
 2415 /*
 2416  * Do L2 address resolution for @sa_dst address. Stores found
 2417  * address in @desten buffer. Copy of lle ln_flags can be also
 2418  * saved in @pflags if @pflags is non-NULL.
 2419  *
 2420  * Return values:
 2421  * - 0 on success (address copied to buffer).
 2422  * - EWOULDBLOCK (no local error, but address is still unresolved)
 2423  * - other errors (alloc failure, etc)
 2424  */
 2425 int
 2426 nd6_resolve_addr(struct ifnet *ifp, int flags, const struct sockaddr *dst,
 2427     char *desten, uint32_t *pflags)
 2428 {
 2429         int error;
 2430 
 2431         flags |= LLE_ADDRONLY;
 2432         error = nd6_resolve_slow(ifp, flags, NULL,
 2433             (const struct sockaddr_in6 *)dst, desten, pflags, NULL);
 2434         return (error);
 2435 }
 2436 
 2437 int
 2438 nd6_flush_holdchain(struct ifnet *ifp, struct mbuf *chain,
 2439     struct sockaddr_in6 *dst)
 2440 {
 2441         struct mbuf *m, *m_head;
 2442         int error = 0;
 2443 
 2444         m_head = chain;
 2445 
 2446         while (m_head) {
 2447                 m = m_head;
 2448                 m_head = m_head->m_nextpkt;
 2449                 m->m_nextpkt = NULL;
 2450                 error = nd6_output_ifp(ifp, ifp, m, dst, NULL);
 2451         }
 2452 
 2453         /*
 2454          * XXX
 2455          * note that intermediate errors are blindly ignored
 2456          */
 2457         return (error);
 2458 }
 2459 
 2460 static int
 2461 nd6_need_cache(struct ifnet *ifp)
 2462 {
 2463         /*
 2464          * XXX: we currently do not make neighbor cache on any interface
 2465          * other than Ethernet and GIF.
 2466          *
 2467          * RFC2893 says:
 2468          * - unidirectional tunnels needs no ND
 2469          */
 2470         switch (ifp->if_type) {
 2471         case IFT_ETHER:
 2472         case IFT_IEEE1394:
 2473         case IFT_L2VLAN:
 2474         case IFT_INFINIBAND:
 2475         case IFT_BRIDGE:
 2476         case IFT_PROPVIRTUAL:
 2477                 return (1);
 2478         default:
 2479                 return (0);
 2480         }
 2481 }
 2482 
 2483 /*
 2484  * Add pernament ND6 link-layer record for given
 2485  * interface address.
 2486  *
 2487  * Very similar to IPv4 arp_ifinit(), but:
 2488  * 1) IPv6 DAD is performed in different place
 2489  * 2) It is called by IPv6 protocol stack in contrast to
 2490  * arp_ifinit() which is typically called in SIOCSIFADDR
 2491  * driver ioctl handler.
 2492  *
 2493  */
 2494 int
 2495 nd6_add_ifa_lle(struct in6_ifaddr *ia)
 2496 {
 2497         struct ifnet *ifp;
 2498         struct llentry *ln, *ln_tmp;
 2499         struct sockaddr *dst;
 2500 
 2501         ifp = ia->ia_ifa.ifa_ifp;
 2502         if (nd6_need_cache(ifp) == 0)
 2503                 return (0);
 2504 
 2505         ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
 2506         dst = (struct sockaddr *)&ia->ia_addr;
 2507         ln = lltable_alloc_entry(LLTABLE6(ifp), LLE_IFADDR, dst);
 2508         if (ln == NULL)
 2509                 return (ENOBUFS);
 2510 
 2511         IF_AFDATA_WLOCK(ifp);
 2512         LLE_WLOCK(ln);
 2513         /* Unlink any entry if exists */
 2514         ln_tmp = lla_lookup(LLTABLE6(ifp), LLE_EXCLUSIVE, dst);
 2515         if (ln_tmp != NULL)
 2516                 lltable_unlink_entry(LLTABLE6(ifp), ln_tmp);
 2517         lltable_link_entry(LLTABLE6(ifp), ln);
 2518         IF_AFDATA_WUNLOCK(ifp);
 2519 
 2520         if (ln_tmp != NULL)
 2521                 EVENTHANDLER_INVOKE(lle_event, ln_tmp, LLENTRY_EXPIRED);
 2522         EVENTHANDLER_INVOKE(lle_event, ln, LLENTRY_RESOLVED);
 2523 
 2524         LLE_WUNLOCK(ln);
 2525         if (ln_tmp != NULL)
 2526                 llentry_free(ln_tmp);
 2527 
 2528         return (0);
 2529 }
 2530 
 2531 /*
 2532  * Removes either all lle entries for given @ia, or lle
 2533  * corresponding to @ia address.
 2534  */
 2535 void
 2536 nd6_rem_ifa_lle(struct in6_ifaddr *ia, int all)
 2537 {
 2538         struct sockaddr_in6 mask, addr;
 2539         struct sockaddr *saddr, *smask;
 2540         struct ifnet *ifp;
 2541 
 2542         ifp = ia->ia_ifa.ifa_ifp;
 2543         memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
 2544         memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
 2545         saddr = (struct sockaddr *)&addr;
 2546         smask = (struct sockaddr *)&mask;
 2547 
 2548         if (all != 0)
 2549                 lltable_prefix_free(AF_INET6, saddr, smask, LLE_STATIC);
 2550         else
 2551                 lltable_delete_addr(LLTABLE6(ifp), LLE_IFADDR, saddr);
 2552 }
 2553 
 2554 static void 
 2555 clear_llinfo_pqueue(struct llentry *ln)
 2556 {
 2557         struct mbuf *m_hold, *m_hold_next;
 2558 
 2559         for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) {
 2560                 m_hold_next = m_hold->m_nextpkt;
 2561                 m_freem(m_hold);
 2562         }
 2563 
 2564         ln->la_hold = NULL;
 2565 }
 2566 
 2567 static int
 2568 nd6_sysctl_prlist(SYSCTL_HANDLER_ARGS)
 2569 {
 2570         struct in6_prefix p;
 2571         struct sockaddr_in6 s6;
 2572         struct nd_prefix *pr;
 2573         struct nd_pfxrouter *pfr;
 2574         time_t maxexpire;
 2575         int error;
 2576         char ip6buf[INET6_ADDRSTRLEN];
 2577 
 2578         if (req->newptr)
 2579                 return (EPERM);
 2580 
 2581         error = sysctl_wire_old_buffer(req, 0);
 2582         if (error != 0)
 2583                 return (error);
 2584 
 2585         bzero(&p, sizeof(p));
 2586         p.origin = PR_ORIG_RA;
 2587         bzero(&s6, sizeof(s6));
 2588         s6.sin6_family = AF_INET6;
 2589         s6.sin6_len = sizeof(s6);
 2590 
 2591         ND6_RLOCK();
 2592         LIST_FOREACH(pr, &V_nd_prefix, ndpr_entry) {
 2593                 p.prefix = pr->ndpr_prefix;
 2594                 if (sa6_recoverscope(&p.prefix)) {
 2595                         log(LOG_ERR, "scope error in prefix list (%s)\n",
 2596                             ip6_sprintf(ip6buf, &p.prefix.sin6_addr));
 2597                         /* XXX: press on... */
 2598                 }
 2599                 p.raflags = pr->ndpr_raf;
 2600                 p.prefixlen = pr->ndpr_plen;
 2601                 p.vltime = pr->ndpr_vltime;
 2602                 p.pltime = pr->ndpr_pltime;
 2603                 p.if_index = pr->ndpr_ifp->if_index;
 2604                 if (pr->ndpr_vltime == ND6_INFINITE_LIFETIME)
 2605                         p.expire = 0;
 2606                 else {
 2607                         /* XXX: we assume time_t is signed. */
 2608                         maxexpire = (-1) &
 2609                             ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
 2610                         if (pr->ndpr_vltime < maxexpire - pr->ndpr_lastupdate)
 2611                                 p.expire = pr->ndpr_lastupdate +
 2612                                     pr->ndpr_vltime +
 2613                                     (time_second - time_uptime);
 2614                         else
 2615                                 p.expire = maxexpire;
 2616                 }
 2617                 p.refcnt = pr->ndpr_addrcnt;
 2618                 p.flags = pr->ndpr_stateflags;
 2619                 p.advrtrs = 0;
 2620                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry)
 2621                         p.advrtrs++;
 2622                 error = SYSCTL_OUT(req, &p, sizeof(p));
 2623                 if (error != 0)
 2624                         break;
 2625                 LIST_FOREACH(pfr, &pr->ndpr_advrtrs, pfr_entry) {
 2626                         s6.sin6_addr = pfr->router->rtaddr;
 2627                         if (sa6_recoverscope(&s6))
 2628                                 log(LOG_ERR,
 2629                                     "scope error in prefix list (%s)\n",
 2630                                     ip6_sprintf(ip6buf, &pfr->router->rtaddr));
 2631                         error = SYSCTL_OUT(req, &s6, sizeof(s6));
 2632                         if (error != 0)
 2633                                 goto out;
 2634                 }
 2635         }
 2636 out:
 2637         ND6_RUNLOCK();
 2638         return (error);
 2639 }
 2640 SYSCTL_PROC(_net_inet6_icmp6, ICMPV6CTL_ND6_PRLIST, nd6_prlist,
 2641         CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
 2642         NULL, 0, nd6_sysctl_prlist, "S,in6_prefix",
 2643         "NDP prefix list");
 2644 SYSCTL_INT(_net_inet6_icmp6, ICMPV6CTL_ND6_MAXQLEN, nd6_maxqueuelen,
 2645         CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_maxqueuelen), 1, "");
 2646 SYSCTL_INT(_net_inet6_icmp6, OID_AUTO, nd6_gctimer,
 2647         CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(nd6_gctimer), (60 * 60 * 24), "");

Cache object: 3822f56aea19e559f6ee2e1e485e5794


[ 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.