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_rtr.c

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

    1 /*-
    2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
    3  * All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 3. Neither the name of the project nor the names of its contributors
   14  *    may be used to endorse or promote products derived from this software
   15  *    without specific prior written permission.
   16  *
   17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
   18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
   21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  *
   29  *      $KAME: nd6_rtr.c,v 1.111 2001/04/27 01:37:15 jinmei Exp $
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD: releng/8.3/sys/netinet6/nd6_rtr.c 231685 2012-02-14 15:17:58Z bz $");
   34 
   35 #include "opt_inet.h"
   36 #include "opt_inet6.h"
   37 
   38 #include <sys/param.h>
   39 #include <sys/systm.h>
   40 #include <sys/malloc.h>
   41 #include <sys/mbuf.h>
   42 #include <sys/socket.h>
   43 #include <sys/sockio.h>
   44 #include <sys/time.h>
   45 #include <sys/kernel.h>
   46 #include <sys/lock.h>
   47 #include <sys/errno.h>
   48 #include <sys/rwlock.h>
   49 #include <sys/syslog.h>
   50 #include <sys/queue.h>
   51 
   52 #include <net/if.h>
   53 #include <net/if_types.h>
   54 #include <net/if_dl.h>
   55 #include <net/route.h>
   56 #include <net/radix.h>
   57 #include <net/vnet.h>
   58 
   59 #include <netinet/in.h>
   60 #include <net/if_llatbl.h>
   61 #include <netinet6/in6_var.h>
   62 #include <netinet6/in6_ifattach.h>
   63 #include <netinet/ip6.h>
   64 #include <netinet6/ip6_var.h>
   65 #include <netinet6/nd6.h>
   66 #include <netinet/icmp6.h>
   67 #include <netinet6/scope6_var.h>
   68 
   69 static int rtpref(struct nd_defrouter *);
   70 static struct nd_defrouter *defrtrlist_update(struct nd_defrouter *);
   71 static int prelist_update __P((struct nd_prefixctl *, struct nd_defrouter *,
   72     struct mbuf *, int));
   73 static struct in6_ifaddr *in6_ifadd(struct nd_prefixctl *,      int);
   74 static struct nd_pfxrouter *pfxrtr_lookup __P((struct nd_prefix *,
   75         struct nd_defrouter *));
   76 static void pfxrtr_add(struct nd_prefix *, struct nd_defrouter *);
   77 static void pfxrtr_del(struct nd_pfxrouter *);
   78 static struct nd_pfxrouter *find_pfxlist_reachable_router
   79 (struct nd_prefix *);
   80 static void defrouter_delreq(struct nd_defrouter *);
   81 static void nd6_rtmsg(int, struct rtentry *);
   82 
   83 static int in6_init_prefix_ltimes(struct nd_prefix *);
   84 static void in6_init_address_ltimes __P((struct nd_prefix *,
   85         struct in6_addrlifetime *));
   86 
   87 static int nd6_prefix_onlink(struct nd_prefix *);
   88 static int nd6_prefix_offlink(struct nd_prefix *);
   89 
   90 static int rt6_deleteroute(struct radix_node *, void *);
   91 
   92 VNET_DECLARE(int, nd6_recalc_reachtm_interval);
   93 #define V_nd6_recalc_reachtm_interval   VNET(nd6_recalc_reachtm_interval)
   94 
   95 static VNET_DEFINE(struct ifnet *, nd6_defifp);
   96 VNET_DEFINE(int, nd6_defifindex);
   97 #define V_nd6_defifp                    VNET(nd6_defifp)
   98 
   99 VNET_DEFINE(int, ip6_use_tempaddr) = 0;
  100 
  101 VNET_DEFINE(int, ip6_desync_factor);
  102 VNET_DEFINE(u_int32_t, ip6_temp_preferred_lifetime) = DEF_TEMP_PREFERRED_LIFETIME;
  103 VNET_DEFINE(u_int32_t, ip6_temp_valid_lifetime) = DEF_TEMP_VALID_LIFETIME;
  104 
  105 VNET_DEFINE(int, ip6_temp_regen_advance) = TEMPADDR_REGEN_ADVANCE;
  106 
  107 /* RTPREF_MEDIUM has to be 0! */
  108 #define RTPREF_HIGH     1
  109 #define RTPREF_MEDIUM   0
  110 #define RTPREF_LOW      (-1)
  111 #define RTPREF_RESERVED (-2)
  112 #define RTPREF_INVALID  (-3)    /* internal */
  113 
  114 /*
  115  * Receive Router Solicitation Message - just for routers.
  116  * Router solicitation/advertisement is mostly managed by userland program
  117  * (rtadvd) so here we have no function like nd6_ra_output().
  118  *
  119  * Based on RFC 2461
  120  */
  121 void
  122 nd6_rs_input(struct mbuf *m, int off, int icmp6len)
  123 {
  124         struct ifnet *ifp = m->m_pkthdr.rcvif;
  125         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
  126         struct nd_router_solicit *nd_rs;
  127         struct in6_addr saddr6 = ip6->ip6_src;
  128         char *lladdr = NULL;
  129         int lladdrlen = 0;
  130         union nd_opts ndopts;
  131         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
  132 
  133         /* If I'm not a router, ignore it. */
  134         if (V_ip6_accept_rtadv != 0 || V_ip6_forwarding != 1)
  135                 goto freeit;
  136 
  137         /* Sanity checks */
  138         if (ip6->ip6_hlim != 255) {
  139                 nd6log((LOG_ERR,
  140                     "nd6_rs_input: invalid hlim (%d) from %s to %s on %s\n",
  141                     ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
  142                     ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
  143                 goto bad;
  144         }
  145 
  146         /*
  147          * Don't update the neighbor cache, if src = ::.
  148          * This indicates that the src has no IP address assigned yet.
  149          */
  150         if (IN6_IS_ADDR_UNSPECIFIED(&saddr6))
  151                 goto freeit;
  152 
  153 #ifndef PULLDOWN_TEST
  154         IP6_EXTHDR_CHECK(m, off, icmp6len,);
  155         nd_rs = (struct nd_router_solicit *)((caddr_t)ip6 + off);
  156 #else
  157         IP6_EXTHDR_GET(nd_rs, struct nd_router_solicit *, m, off, icmp6len);
  158         if (nd_rs == NULL) {
  159                 ICMP6STAT_INC(icp6s_tooshort);
  160                 return;
  161         }
  162 #endif
  163 
  164         icmp6len -= sizeof(*nd_rs);
  165         nd6_option_init(nd_rs + 1, icmp6len, &ndopts);
  166         if (nd6_options(&ndopts) < 0) {
  167                 nd6log((LOG_INFO,
  168                     "nd6_rs_input: invalid ND option, ignored\n"));
  169                 /* nd6_options have incremented stats */
  170                 goto freeit;
  171         }
  172 
  173         if (ndopts.nd_opts_src_lladdr) {
  174                 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
  175                 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
  176         }
  177 
  178         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
  179                 nd6log((LOG_INFO,
  180                     "nd6_rs_input: lladdrlen mismatch for %s "
  181                     "(if %d, RS packet %d)\n",
  182                     ip6_sprintf(ip6bufs, &saddr6),
  183                     ifp->if_addrlen, lladdrlen - 2));
  184                 goto bad;
  185         }
  186 
  187         nd6_cache_lladdr(ifp, &saddr6, lladdr, lladdrlen, ND_ROUTER_SOLICIT, 0);
  188 
  189  freeit:
  190         m_freem(m);
  191         return;
  192 
  193  bad:
  194         ICMP6STAT_INC(icp6s_badrs);
  195         m_freem(m);
  196 }
  197 
  198 /*
  199  * Receive Router Advertisement Message.
  200  *
  201  * Based on RFC 2461
  202  * TODO: on-link bit on prefix information
  203  * TODO: ND_RA_FLAG_{OTHER,MANAGED} processing
  204  */
  205 void
  206 nd6_ra_input(struct mbuf *m, int off, int icmp6len)
  207 {
  208         struct ifnet *ifp = m->m_pkthdr.rcvif;
  209         struct nd_ifinfo *ndi = ND_IFINFO(ifp);
  210         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
  211         struct nd_router_advert *nd_ra;
  212         struct in6_addr saddr6 = ip6->ip6_src;
  213         int mcast = 0;
  214         union nd_opts ndopts;
  215         struct nd_defrouter *dr;
  216         char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
  217 
  218         /*
  219          * We only accept RAs only when
  220          * the system-wide variable allows the acceptance, and
  221          * per-interface variable allows RAs on the receiving interface.
  222          */
  223         if (V_ip6_accept_rtadv == 0)
  224                 goto freeit;
  225         if (!(ndi->flags & ND6_IFF_ACCEPT_RTADV))
  226                 goto freeit;
  227 
  228         if (ip6->ip6_hlim != 255) {
  229                 nd6log((LOG_ERR,
  230                     "nd6_ra_input: invalid hlim (%d) from %s to %s on %s\n",
  231                     ip6->ip6_hlim, ip6_sprintf(ip6bufs, &ip6->ip6_src),
  232                     ip6_sprintf(ip6bufd, &ip6->ip6_dst), if_name(ifp)));
  233                 goto bad;
  234         }
  235 
  236         if (!IN6_IS_ADDR_LINKLOCAL(&saddr6)) {
  237                 nd6log((LOG_ERR,
  238                     "nd6_ra_input: src %s is not link-local\n",
  239                     ip6_sprintf(ip6bufs, &saddr6)));
  240                 goto bad;
  241         }
  242 
  243 #ifndef PULLDOWN_TEST
  244         IP6_EXTHDR_CHECK(m, off, icmp6len,);
  245         nd_ra = (struct nd_router_advert *)((caddr_t)ip6 + off);
  246 #else
  247         IP6_EXTHDR_GET(nd_ra, struct nd_router_advert *, m, off, icmp6len);
  248         if (nd_ra == NULL) {
  249                 ICMP6STAT_INC(icp6s_tooshort);
  250                 return;
  251         }
  252 #endif
  253 
  254         icmp6len -= sizeof(*nd_ra);
  255         nd6_option_init(nd_ra + 1, icmp6len, &ndopts);
  256         if (nd6_options(&ndopts) < 0) {
  257                 nd6log((LOG_INFO,
  258                     "nd6_ra_input: invalid ND option, ignored\n"));
  259                 /* nd6_options have incremented stats */
  260                 goto freeit;
  261         }
  262 
  263     {
  264         struct nd_defrouter dr0;
  265         u_int32_t advreachable = nd_ra->nd_ra_reachable;
  266 
  267         /* remember if this is a multicasted advertisement */
  268         if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
  269                 mcast = 1;
  270 
  271         bzero(&dr0, sizeof(dr0));
  272         dr0.rtaddr = saddr6;
  273         dr0.flags  = nd_ra->nd_ra_flags_reserved;
  274         dr0.rtlifetime = ntohs(nd_ra->nd_ra_router_lifetime);
  275         dr0.expire = time_second + dr0.rtlifetime;
  276         dr0.ifp = ifp;
  277         /* unspecified or not? (RFC 2461 6.3.4) */
  278         if (advreachable) {
  279                 advreachable = ntohl(advreachable);
  280                 if (advreachable <= MAX_REACHABLE_TIME &&
  281                     ndi->basereachable != advreachable) {
  282                         ndi->basereachable = advreachable;
  283                         ndi->reachable = ND_COMPUTE_RTIME(ndi->basereachable);
  284                         ndi->recalctm = V_nd6_recalc_reachtm_interval; /* reset */
  285                 }
  286         }
  287         if (nd_ra->nd_ra_retransmit)
  288                 ndi->retrans = ntohl(nd_ra->nd_ra_retransmit);
  289         if (nd_ra->nd_ra_curhoplimit)
  290                 ndi->chlim = nd_ra->nd_ra_curhoplimit;
  291         dr = defrtrlist_update(&dr0);
  292     }
  293 
  294         /*
  295          * prefix
  296          */
  297         if (ndopts.nd_opts_pi) {
  298                 struct nd_opt_hdr *pt;
  299                 struct nd_opt_prefix_info *pi = NULL;
  300                 struct nd_prefixctl pr;
  301 
  302                 for (pt = (struct nd_opt_hdr *)ndopts.nd_opts_pi;
  303                      pt <= (struct nd_opt_hdr *)ndopts.nd_opts_pi_end;
  304                      pt = (struct nd_opt_hdr *)((caddr_t)pt +
  305                                                 (pt->nd_opt_len << 3))) {
  306                         if (pt->nd_opt_type != ND_OPT_PREFIX_INFORMATION)
  307                                 continue;
  308                         pi = (struct nd_opt_prefix_info *)pt;
  309 
  310                         if (pi->nd_opt_pi_len != 4) {
  311                                 nd6log((LOG_INFO,
  312                                     "nd6_ra_input: invalid option "
  313                                     "len %d for prefix information option, "
  314                                     "ignored\n", pi->nd_opt_pi_len));
  315                                 continue;
  316                         }
  317 
  318                         if (128 < pi->nd_opt_pi_prefix_len) {
  319                                 nd6log((LOG_INFO,
  320                                     "nd6_ra_input: invalid prefix "
  321                                     "len %d for prefix information option, "
  322                                     "ignored\n", pi->nd_opt_pi_prefix_len));
  323                                 continue;
  324                         }
  325 
  326                         if (IN6_IS_ADDR_MULTICAST(&pi->nd_opt_pi_prefix)
  327                          || IN6_IS_ADDR_LINKLOCAL(&pi->nd_opt_pi_prefix)) {
  328                                 nd6log((LOG_INFO,
  329                                     "nd6_ra_input: invalid prefix "
  330                                     "%s, ignored\n",
  331                                     ip6_sprintf(ip6bufs,
  332                                         &pi->nd_opt_pi_prefix)));
  333                                 continue;
  334                         }
  335 
  336                         bzero(&pr, sizeof(pr));
  337                         pr.ndpr_prefix.sin6_family = AF_INET6;
  338                         pr.ndpr_prefix.sin6_len = sizeof(pr.ndpr_prefix);
  339                         pr.ndpr_prefix.sin6_addr = pi->nd_opt_pi_prefix;
  340                         pr.ndpr_ifp = (struct ifnet *)m->m_pkthdr.rcvif;
  341 
  342                         pr.ndpr_raf_onlink = (pi->nd_opt_pi_flags_reserved &
  343                             ND_OPT_PI_FLAG_ONLINK) ? 1 : 0;
  344                         pr.ndpr_raf_auto = (pi->nd_opt_pi_flags_reserved &
  345                             ND_OPT_PI_FLAG_AUTO) ? 1 : 0;
  346                         pr.ndpr_plen = pi->nd_opt_pi_prefix_len;
  347                         pr.ndpr_vltime = ntohl(pi->nd_opt_pi_valid_time);
  348                         pr.ndpr_pltime = ntohl(pi->nd_opt_pi_preferred_time);
  349                         (void)prelist_update(&pr, dr, m, mcast);
  350                 }
  351         }
  352 
  353         /*
  354          * MTU
  355          */
  356         if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
  357                 u_long mtu;
  358                 u_long maxmtu;
  359 
  360                 mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
  361 
  362                 /* lower bound */
  363                 if (mtu < IPV6_MMTU) {
  364                         nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
  365                             "mtu=%lu sent from %s, ignoring\n",
  366                             mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src)));
  367                         goto skip;
  368                 }
  369 
  370                 /* upper bound */
  371                 maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
  372                     ? ndi->maxmtu : ifp->if_mtu;
  373                 if (mtu <= maxmtu) {
  374                         int change = (ndi->linkmtu != mtu);
  375 
  376                         ndi->linkmtu = mtu;
  377                         if (change) /* in6_maxmtu may change */
  378                                 in6_setmaxmtu();
  379                 } else {
  380                         nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
  381                             "mtu=%lu sent from %s; "
  382                             "exceeds maxmtu %lu, ignoring\n",
  383                             mtu, ip6_sprintf(ip6bufs, &ip6->ip6_src), maxmtu));
  384                 }
  385         }
  386 
  387  skip:
  388 
  389         /*
  390          * Source link layer address
  391          */
  392     {
  393         char *lladdr = NULL;
  394         int lladdrlen = 0;
  395 
  396         if (ndopts.nd_opts_src_lladdr) {
  397                 lladdr = (char *)(ndopts.nd_opts_src_lladdr + 1);
  398                 lladdrlen = ndopts.nd_opts_src_lladdr->nd_opt_len << 3;
  399         }
  400 
  401         if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
  402                 nd6log((LOG_INFO,
  403                     "nd6_ra_input: lladdrlen mismatch for %s "
  404                     "(if %d, RA packet %d)\n", ip6_sprintf(ip6bufs, &saddr6),
  405                     ifp->if_addrlen, lladdrlen - 2));
  406                 goto bad;
  407         }
  408 
  409         nd6_cache_lladdr(ifp, &saddr6, lladdr,
  410             lladdrlen, ND_ROUTER_ADVERT, 0);
  411 
  412         /*
  413          * Installing a link-layer address might change the state of the
  414          * router's neighbor cache, which might also affect our on-link
  415          * detection of adveritsed prefixes.
  416          */
  417         pfxlist_onlink_check();
  418     }
  419 
  420  freeit:
  421         m_freem(m);
  422         return;
  423 
  424  bad:
  425         ICMP6STAT_INC(icp6s_badra);
  426         m_freem(m);
  427 }
  428 
  429 /*
  430  * default router list proccessing sub routines
  431  */
  432 
  433 /* tell the change to user processes watching the routing socket. */
  434 static void
  435 nd6_rtmsg(int cmd, struct rtentry *rt)
  436 {
  437         struct rt_addrinfo info;
  438         struct ifnet *ifp;
  439         struct ifaddr *ifa;
  440 
  441         bzero((caddr_t)&info, sizeof(info));
  442         info.rti_info[RTAX_DST] = rt_key(rt);
  443         info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
  444         info.rti_info[RTAX_NETMASK] = rt_mask(rt);
  445         ifp = rt->rt_ifp;
  446         if (ifp != NULL) {
  447                 IF_ADDR_LOCK(ifp);
  448                 ifa = TAILQ_FIRST(&ifp->if_addrhead);
  449                 info.rti_info[RTAX_IFP] = ifa->ifa_addr;
  450                 ifa_ref(ifa);
  451                 IF_ADDR_UNLOCK(ifp);
  452                 info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
  453         } else
  454                 ifa = NULL;
  455 
  456         rt_missmsg(cmd, &info, rt->rt_flags, 0);
  457         if (ifa != NULL)
  458                 ifa_free(ifa);
  459 }
  460 
  461 static void
  462 defrouter_addreq(struct nd_defrouter *new)
  463 {
  464         struct sockaddr_in6 def, mask, gate;
  465         struct rtentry *newrt = NULL;
  466         int s;
  467         int error;
  468 
  469         bzero(&def, sizeof(def));
  470         bzero(&mask, sizeof(mask));
  471         bzero(&gate, sizeof(gate));
  472 
  473         def.sin6_len = mask.sin6_len = gate.sin6_len =
  474             sizeof(struct sockaddr_in6);
  475         def.sin6_family = gate.sin6_family = AF_INET6;
  476         gate.sin6_addr = new->rtaddr;
  477 
  478         s = splnet();
  479         error = rtrequest(RTM_ADD, (struct sockaddr *)&def,
  480             (struct sockaddr *)&gate, (struct sockaddr *)&mask,
  481             RTF_GATEWAY, &newrt);
  482         if (newrt) {
  483                 nd6_rtmsg(RTM_ADD, newrt); /* tell user process */
  484                 RTFREE(newrt);
  485         }
  486         if (error == 0)
  487                 new->installed = 1;
  488         splx(s);
  489         return;
  490 }
  491 
  492 struct nd_defrouter *
  493 defrouter_lookup(struct in6_addr *addr, struct ifnet *ifp)
  494 {
  495         struct nd_defrouter *dr;
  496 
  497         for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
  498              dr = TAILQ_NEXT(dr, dr_entry)) {
  499                 if (dr->ifp == ifp && IN6_ARE_ADDR_EQUAL(addr, &dr->rtaddr))
  500                         return (dr);
  501         }
  502 
  503         return (NULL);          /* search failed */
  504 }
  505 
  506 /*
  507  * Remove the default route for a given router.
  508  * This is just a subroutine function for defrouter_select(), and should
  509  * not be called from anywhere else.
  510  */
  511 static void
  512 defrouter_delreq(struct nd_defrouter *dr)
  513 {
  514         struct sockaddr_in6 def, mask, gate;
  515         struct rtentry *oldrt = NULL;
  516 
  517         bzero(&def, sizeof(def));
  518         bzero(&mask, sizeof(mask));
  519         bzero(&gate, sizeof(gate));
  520 
  521         def.sin6_len = mask.sin6_len = gate.sin6_len =
  522             sizeof(struct sockaddr_in6);
  523         def.sin6_family = gate.sin6_family = AF_INET6;
  524         gate.sin6_addr = dr->rtaddr;
  525 
  526         rtrequest(RTM_DELETE, (struct sockaddr *)&def,
  527             (struct sockaddr *)&gate,
  528             (struct sockaddr *)&mask, RTF_GATEWAY, &oldrt);
  529         if (oldrt) {
  530                 nd6_rtmsg(RTM_DELETE, oldrt);
  531                 RTFREE(oldrt);
  532         }
  533 
  534         dr->installed = 0;
  535 }
  536 
  537 /*
  538  * remove all default routes from default router list
  539  */
  540 void
  541 defrouter_reset(void)
  542 {
  543         struct nd_defrouter *dr;
  544 
  545         for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
  546              dr = TAILQ_NEXT(dr, dr_entry))
  547                 defrouter_delreq(dr);
  548 
  549         /*
  550          * XXX should we also nuke any default routers in the kernel, by
  551          * going through them by rtalloc1()?
  552          */
  553 }
  554 
  555 void
  556 defrtrlist_del(struct nd_defrouter *dr)
  557 {
  558         struct nd_defrouter *deldr = NULL;
  559         struct nd_prefix *pr;
  560 
  561         /*
  562          * Flush all the routing table entries that use the router
  563          * as a next hop.
  564          */
  565         if (!V_ip6_forwarding && V_ip6_accept_rtadv) /* XXX: better condition? */
  566                 rt6_flush(&dr->rtaddr, dr->ifp);
  567 
  568         if (dr->installed) {
  569                 deldr = dr;
  570                 defrouter_delreq(dr);
  571         }
  572         TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
  573 
  574         /*
  575          * Also delete all the pointers to the router in each prefix lists.
  576          */
  577         for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
  578                 struct nd_pfxrouter *pfxrtr;
  579                 if ((pfxrtr = pfxrtr_lookup(pr, dr)) != NULL)
  580                         pfxrtr_del(pfxrtr);
  581         }
  582         pfxlist_onlink_check();
  583 
  584         /*
  585          * If the router is the primary one, choose a new one.
  586          * Note that defrouter_select() will remove the current gateway
  587          * from the routing table.
  588          */
  589         if (deldr)
  590                 defrouter_select();
  591 
  592         free(dr, M_IP6NDP);
  593 }
  594 
  595 /*
  596  * Default Router Selection according to Section 6.3.6 of RFC 2461 and
  597  * draft-ietf-ipngwg-router-selection:
  598  * 1) Routers that are reachable or probably reachable should be preferred.
  599  *    If we have more than one (probably) reachable router, prefer ones
  600  *    with the highest router preference.
  601  * 2) When no routers on the list are known to be reachable or
  602  *    probably reachable, routers SHOULD be selected in a round-robin
  603  *    fashion, regardless of router preference values.
  604  * 3) If the Default Router List is empty, assume that all
  605  *    destinations are on-link.
  606  *
  607  * We assume nd_defrouter is sorted by router preference value.
  608  * Since the code below covers both with and without router preference cases,
  609  * we do not need to classify the cases by ifdef.
  610  *
  611  * At this moment, we do not try to install more than one default router,
  612  * even when the multipath routing is available, because we're not sure about
  613  * the benefits for stub hosts comparing to the risk of making the code
  614  * complicated and the possibility of introducing bugs.
  615  */
  616 void
  617 defrouter_select(void)
  618 {
  619         int s = splnet();
  620         struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL;
  621         struct llentry *ln = NULL;
  622 
  623         /*
  624          * This function should be called only when acting as an autoconfigured
  625          * host.  Although the remaining part of this function is not effective
  626          * if the node is not an autoconfigured host, we explicitly exclude
  627          * such cases here for safety.
  628          */
  629         if (V_ip6_forwarding || !V_ip6_accept_rtadv) {
  630                 nd6log((LOG_WARNING,
  631                     "defrouter_select: called unexpectedly (forwarding=%d, "
  632                     "accept_rtadv=%d)\n", V_ip6_forwarding, V_ip6_accept_rtadv));
  633                 splx(s);
  634                 return;
  635         }
  636 
  637         /*
  638          * Let's handle easy case (3) first:
  639          * If default router list is empty, there's nothing to be done.
  640          */
  641         if (!TAILQ_FIRST(&V_nd_defrouter)) {
  642                 splx(s);
  643                 return;
  644         }
  645 
  646         /*
  647          * Search for a (probably) reachable router from the list.
  648          * We just pick up the first reachable one (if any), assuming that
  649          * the ordering rule of the list described in defrtrlist_update().
  650          */
  651         for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
  652              dr = TAILQ_NEXT(dr, dr_entry)) {
  653                 IF_AFDATA_LOCK(dr->ifp);
  654                 if (selected_dr == NULL &&
  655                     (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) &&
  656                     ND6_IS_LLINFO_PROBREACH(ln)) {
  657                         selected_dr = dr;
  658                 }
  659                 IF_AFDATA_UNLOCK(dr->ifp);
  660                 if (ln != NULL) {
  661                         LLE_RUNLOCK(ln);
  662                         ln = NULL;
  663                 }
  664 
  665                 if (dr->installed && installed_dr == NULL)
  666                         installed_dr = dr;
  667                 else if (dr->installed && installed_dr) {
  668                         /* this should not happen.  warn for diagnosis. */
  669                         log(LOG_ERR, "defrouter_select: more than one router"
  670                             " is installed\n");
  671                 }
  672         }
  673         /*
  674          * If none of the default routers was found to be reachable,
  675          * round-robin the list regardless of preference.
  676          * Otherwise, if we have an installed router, check if the selected
  677          * (reachable) router should really be preferred to the installed one.
  678          * We only prefer the new router when the old one is not reachable
  679          * or when the new one has a really higher preference value.
  680          */
  681         if (selected_dr == NULL) {
  682                 if (installed_dr == NULL || !TAILQ_NEXT(installed_dr, dr_entry))
  683                         selected_dr = TAILQ_FIRST(&V_nd_defrouter);
  684                 else
  685                         selected_dr = TAILQ_NEXT(installed_dr, dr_entry);
  686         } else if (installed_dr) {
  687                 IF_AFDATA_LOCK(installed_dr->ifp);
  688                 if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) &&
  689                     ND6_IS_LLINFO_PROBREACH(ln) &&
  690                     rtpref(selected_dr) <= rtpref(installed_dr)) {
  691                         selected_dr = installed_dr;
  692                 }
  693                 IF_AFDATA_UNLOCK(installed_dr->ifp);
  694                 if (ln != NULL)
  695                         LLE_RUNLOCK(ln);
  696         }
  697 
  698         /*
  699          * If the selected router is different than the installed one,
  700          * remove the installed router and install the selected one.
  701          * Note that the selected router is never NULL here.
  702          */
  703         if (installed_dr != selected_dr) {
  704                 if (installed_dr)
  705                         defrouter_delreq(installed_dr);
  706                 defrouter_addreq(selected_dr);
  707         }
  708 
  709         splx(s);
  710         return;
  711 }
  712 
  713 /*
  714  * for default router selection
  715  * regards router-preference field as a 2-bit signed integer
  716  */
  717 static int
  718 rtpref(struct nd_defrouter *dr)
  719 {
  720         switch (dr->flags & ND_RA_FLAG_RTPREF_MASK) {
  721         case ND_RA_FLAG_RTPREF_HIGH:
  722                 return (RTPREF_HIGH);
  723         case ND_RA_FLAG_RTPREF_MEDIUM:
  724         case ND_RA_FLAG_RTPREF_RSV:
  725                 return (RTPREF_MEDIUM);
  726         case ND_RA_FLAG_RTPREF_LOW:
  727                 return (RTPREF_LOW);
  728         default:
  729                 /*
  730                  * This case should never happen.  If it did, it would mean a
  731                  * serious bug of kernel internal.  We thus always bark here.
  732                  * Or, can we even panic?
  733                  */
  734                 log(LOG_ERR, "rtpref: impossible RA flag %x\n", dr->flags);
  735                 return (RTPREF_INVALID);
  736         }
  737         /* NOTREACHED */
  738 }
  739 
  740 static struct nd_defrouter *
  741 defrtrlist_update(struct nd_defrouter *new)
  742 {
  743         struct nd_defrouter *dr, *n;
  744         int s = splnet();
  745 
  746         if ((dr = defrouter_lookup(&new->rtaddr, new->ifp)) != NULL) {
  747                 /* entry exists */
  748                 if (new->rtlifetime == 0) {
  749                         defrtrlist_del(dr);
  750                         dr = NULL;
  751                 } else {
  752                         int oldpref = rtpref(dr);
  753 
  754                         /* override */
  755                         dr->flags = new->flags; /* xxx flag check */
  756                         dr->rtlifetime = new->rtlifetime;
  757                         dr->expire = new->expire;
  758 
  759                         /*
  760                          * If the preference does not change, there's no need
  761                          * to sort the entries. Also make sure the selected
  762                          * router is still installed in the kernel.
  763                          */
  764                         if (dr->installed && rtpref(new) == oldpref) {
  765                                 splx(s);
  766                                 return (dr);
  767                         }
  768 
  769                         /*
  770                          * preferred router may be changed, so relocate
  771                          * this router.
  772                          * XXX: calling TAILQ_REMOVE directly is a bad manner.
  773                          * However, since defrtrlist_del() has many side
  774                          * effects, we intentionally do so here.
  775                          * defrouter_select() below will handle routing
  776                          * changes later.
  777                          */
  778                         TAILQ_REMOVE(&V_nd_defrouter, dr, dr_entry);
  779                         n = dr;
  780                         goto insert;
  781                 }
  782                 splx(s);
  783                 return (dr);
  784         }
  785 
  786         /* entry does not exist */
  787         if (new->rtlifetime == 0) {
  788                 splx(s);
  789                 return (NULL);
  790         }
  791 
  792         n = (struct nd_defrouter *)malloc(sizeof(*n), M_IP6NDP, M_NOWAIT);
  793         if (n == NULL) {
  794                 splx(s);
  795                 return (NULL);
  796         }
  797         bzero(n, sizeof(*n));
  798         *n = *new;
  799 
  800 insert:
  801         /*
  802          * Insert the new router in the Default Router List;
  803          * The Default Router List should be in the descending order
  804          * of router-preferece.  Routers with the same preference are
  805          * sorted in the arriving time order.
  806          */
  807 
  808         /* insert at the end of the group */
  809         for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
  810              dr = TAILQ_NEXT(dr, dr_entry)) {
  811                 if (rtpref(n) > rtpref(dr))
  812                         break;
  813         }
  814         if (dr)
  815                 TAILQ_INSERT_BEFORE(dr, n, dr_entry);
  816         else
  817                 TAILQ_INSERT_TAIL(&V_nd_defrouter, n, dr_entry);
  818 
  819         defrouter_select();
  820 
  821         splx(s);
  822 
  823         return (n);
  824 }
  825 
  826 static struct nd_pfxrouter *
  827 pfxrtr_lookup(struct nd_prefix *pr, struct nd_defrouter *dr)
  828 {
  829         struct nd_pfxrouter *search;
  830 
  831         for (search = pr->ndpr_advrtrs.lh_first; search; search = search->pfr_next) {
  832                 if (search->router == dr)
  833                         break;
  834         }
  835 
  836         return (search);
  837 }
  838 
  839 static void
  840 pfxrtr_add(struct nd_prefix *pr, struct nd_defrouter *dr)
  841 {
  842         struct nd_pfxrouter *new;
  843 
  844         new = (struct nd_pfxrouter *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
  845         if (new == NULL)
  846                 return;
  847         bzero(new, sizeof(*new));
  848         new->router = dr;
  849 
  850         LIST_INSERT_HEAD(&pr->ndpr_advrtrs, new, pfr_entry);
  851 
  852         pfxlist_onlink_check();
  853 }
  854 
  855 static void
  856 pfxrtr_del(struct nd_pfxrouter *pfr)
  857 {
  858         LIST_REMOVE(pfr, pfr_entry);
  859         free(pfr, M_IP6NDP);
  860 }
  861 
  862 struct nd_prefix *
  863 nd6_prefix_lookup(struct nd_prefixctl *key)
  864 {
  865         struct nd_prefix *search;
  866 
  867         for (search = V_nd_prefix.lh_first;
  868             search; search = search->ndpr_next) {
  869                 if (key->ndpr_ifp == search->ndpr_ifp &&
  870                     key->ndpr_plen == search->ndpr_plen &&
  871                     in6_are_prefix_equal(&key->ndpr_prefix.sin6_addr,
  872                     &search->ndpr_prefix.sin6_addr, key->ndpr_plen)) {
  873                         break;
  874                 }
  875         }
  876 
  877         return (search);
  878 }
  879 
  880 int
  881 nd6_prelist_add(struct nd_prefixctl *pr, struct nd_defrouter *dr,
  882     struct nd_prefix **newp)
  883 {
  884         struct nd_prefix *new = NULL;
  885         int error = 0;
  886         int i, s;
  887         char ip6buf[INET6_ADDRSTRLEN];
  888 
  889         new = (struct nd_prefix *)malloc(sizeof(*new), M_IP6NDP, M_NOWAIT);
  890         if (new == NULL)
  891                 return(ENOMEM);
  892         bzero(new, sizeof(*new));
  893         new->ndpr_ifp = pr->ndpr_ifp;
  894         new->ndpr_prefix = pr->ndpr_prefix;
  895         new->ndpr_plen = pr->ndpr_plen;
  896         new->ndpr_vltime = pr->ndpr_vltime;
  897         new->ndpr_pltime = pr->ndpr_pltime;
  898         new->ndpr_flags = pr->ndpr_flags;
  899         if ((error = in6_init_prefix_ltimes(new)) != 0) {
  900                 free(new, M_IP6NDP);
  901                 return(error);
  902         }
  903         new->ndpr_lastupdate = time_second;
  904         if (newp != NULL)
  905                 *newp = new;
  906 
  907         /* initialization */
  908         LIST_INIT(&new->ndpr_advrtrs);
  909         in6_prefixlen2mask(&new->ndpr_mask, new->ndpr_plen);
  910         /* make prefix in the canonical form */
  911         for (i = 0; i < 4; i++)
  912                 new->ndpr_prefix.sin6_addr.s6_addr32[i] &=
  913                     new->ndpr_mask.s6_addr32[i];
  914 
  915         s = splnet();
  916         /* link ndpr_entry to nd_prefix list */
  917         LIST_INSERT_HEAD(&V_nd_prefix, new, ndpr_entry);
  918         splx(s);
  919 
  920         /* ND_OPT_PI_FLAG_ONLINK processing */
  921         if (new->ndpr_raf_onlink) {
  922                 int e;
  923 
  924                 if ((e = nd6_prefix_onlink(new)) != 0) {
  925                         nd6log((LOG_ERR, "nd6_prelist_add: failed to make "
  926                             "the prefix %s/%d on-link on %s (errno=%d)\n",
  927                             ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
  928                             pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
  929                         /* proceed anyway. XXX: is it correct? */
  930                 }
  931         }
  932 
  933         if (dr)
  934                 pfxrtr_add(new, dr);
  935 
  936         return 0;
  937 }
  938 
  939 void
  940 prelist_remove(struct nd_prefix *pr)
  941 {
  942         struct nd_pfxrouter *pfr, *next;
  943         int e, s;
  944         char ip6buf[INET6_ADDRSTRLEN];
  945 
  946         /* make sure to invalidate the prefix until it is really freed. */
  947         pr->ndpr_vltime = 0;
  948         pr->ndpr_pltime = 0;
  949 
  950         /*
  951          * Though these flags are now meaningless, we'd rather keep the value
  952          * of pr->ndpr_raf_onlink and pr->ndpr_raf_auto not to confuse users
  953          * when executing "ndp -p".
  954          */
  955 
  956         if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0 &&
  957             (e = nd6_prefix_offlink(pr)) != 0) {
  958                 nd6log((LOG_ERR, "prelist_remove: failed to make %s/%d offlink "
  959                     "on %s, errno=%d\n",
  960                     ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
  961                     pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
  962                 /* what should we do? */
  963         }
  964 
  965         if (pr->ndpr_refcnt > 0)
  966                 return;         /* notice here? */
  967 
  968         s = splnet();
  969 
  970         /* unlink ndpr_entry from nd_prefix list */
  971         LIST_REMOVE(pr, ndpr_entry);
  972 
  973         /* free list of routers that adversed the prefix */
  974         for (pfr = pr->ndpr_advrtrs.lh_first; pfr; pfr = next) {
  975                 next = pfr->pfr_next;
  976 
  977                 free(pfr, M_IP6NDP);
  978         }
  979         splx(s);
  980 
  981         free(pr, M_IP6NDP);
  982 
  983         pfxlist_onlink_check();
  984 }
  985 
  986 /*
  987  * dr - may be NULL
  988  */
  989 
  990 static int
  991 prelist_update(struct nd_prefixctl *new, struct nd_defrouter *dr,
  992     struct mbuf *m, int mcast)
  993 {
  994         struct in6_ifaddr *ia6 = NULL, *ia6_match = NULL;
  995         struct ifaddr *ifa;
  996         struct ifnet *ifp = new->ndpr_ifp;
  997         struct nd_prefix *pr;
  998         int s = splnet();
  999         int error = 0;
 1000         int newprefix = 0;
 1001         int auth;
 1002         struct in6_addrlifetime lt6_tmp;
 1003         char ip6buf[INET6_ADDRSTRLEN];
 1004 
 1005         auth = 0;
 1006         if (m) {
 1007                 /*
 1008                  * Authenticity for NA consists authentication for
 1009                  * both IP header and IP datagrams, doesn't it ?
 1010                  */
 1011 #if defined(M_AUTHIPHDR) && defined(M_AUTHIPDGM)
 1012                 auth = ((m->m_flags & M_AUTHIPHDR) &&
 1013                     (m->m_flags & M_AUTHIPDGM));
 1014 #endif
 1015         }
 1016 
 1017         if ((pr = nd6_prefix_lookup(new)) != NULL) {
 1018                 /*
 1019                  * nd6_prefix_lookup() ensures that pr and new have the same
 1020                  * prefix on a same interface.
 1021                  */
 1022 
 1023                 /*
 1024                  * Update prefix information.  Note that the on-link (L) bit
 1025                  * and the autonomous (A) bit should NOT be changed from 1
 1026                  * to 0.
 1027                  */
 1028                 if (new->ndpr_raf_onlink == 1)
 1029                         pr->ndpr_raf_onlink = 1;
 1030                 if (new->ndpr_raf_auto == 1)
 1031                         pr->ndpr_raf_auto = 1;
 1032                 if (new->ndpr_raf_onlink) {
 1033                         pr->ndpr_vltime = new->ndpr_vltime;
 1034                         pr->ndpr_pltime = new->ndpr_pltime;
 1035                         (void)in6_init_prefix_ltimes(pr); /* XXX error case? */
 1036                         pr->ndpr_lastupdate = time_second;
 1037                 }
 1038 
 1039                 if (new->ndpr_raf_onlink &&
 1040                     (pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
 1041                         int e;
 1042 
 1043                         if ((e = nd6_prefix_onlink(pr)) != 0) {
 1044                                 nd6log((LOG_ERR,
 1045                                     "prelist_update: failed to make "
 1046                                     "the prefix %s/%d on-link on %s "
 1047                                     "(errno=%d)\n",
 1048                                     ip6_sprintf(ip6buf,
 1049                                             &pr->ndpr_prefix.sin6_addr),
 1050                                     pr->ndpr_plen, if_name(pr->ndpr_ifp), e));
 1051                                 /* proceed anyway. XXX: is it correct? */
 1052                         }
 1053                 }
 1054 
 1055                 if (dr && pfxrtr_lookup(pr, dr) == NULL)
 1056                         pfxrtr_add(pr, dr);
 1057         } else {
 1058                 struct nd_prefix *newpr = NULL;
 1059 
 1060                 newprefix = 1;
 1061 
 1062                 if (new->ndpr_vltime == 0)
 1063                         goto end;
 1064                 if (new->ndpr_raf_onlink == 0 && new->ndpr_raf_auto == 0)
 1065                         goto end;
 1066 
 1067                 error = nd6_prelist_add(new, dr, &newpr);
 1068                 if (error != 0 || newpr == NULL) {
 1069                         nd6log((LOG_NOTICE, "prelist_update: "
 1070                             "nd6_prelist_add failed for %s/%d on %s "
 1071                             "errno=%d, returnpr=%p\n",
 1072                             ip6_sprintf(ip6buf, &new->ndpr_prefix.sin6_addr),
 1073                             new->ndpr_plen, if_name(new->ndpr_ifp),
 1074                             error, newpr));
 1075                         goto end; /* we should just give up in this case. */
 1076                 }
 1077 
 1078                 /*
 1079                  * XXX: from the ND point of view, we can ignore a prefix
 1080                  * with the on-link bit being zero.  However, we need a
 1081                  * prefix structure for references from autoconfigured
 1082                  * addresses.  Thus, we explicitly make sure that the prefix
 1083                  * itself expires now.
 1084                  */
 1085                 if (newpr->ndpr_raf_onlink == 0) {
 1086                         newpr->ndpr_vltime = 0;
 1087                         newpr->ndpr_pltime = 0;
 1088                         in6_init_prefix_ltimes(newpr);
 1089                 }
 1090 
 1091                 pr = newpr;
 1092         }
 1093 
 1094         /*
 1095          * Address autoconfiguration based on Section 5.5.3 of RFC 2462.
 1096          * Note that pr must be non NULL at this point.
 1097          */
 1098 
 1099         /* 5.5.3 (a). Ignore the prefix without the A bit set. */
 1100         if (!new->ndpr_raf_auto)
 1101                 goto end;
 1102 
 1103         /*
 1104          * 5.5.3 (b). the link-local prefix should have been ignored in
 1105          * nd6_ra_input.
 1106          */
 1107 
 1108         /* 5.5.3 (c). Consistency check on lifetimes: pltime <= vltime. */
 1109         if (new->ndpr_pltime > new->ndpr_vltime) {
 1110                 error = EINVAL; /* XXX: won't be used */
 1111                 goto end;
 1112         }
 1113 
 1114         /*
 1115          * 5.5.3 (d).  If the prefix advertised is not equal to the prefix of
 1116          * an address configured by stateless autoconfiguration already in the
 1117          * list of addresses associated with the interface, and the Valid
 1118          * Lifetime is not 0, form an address.  We first check if we have
 1119          * a matching prefix.
 1120          * Note: we apply a clarification in rfc2462bis-02 here.  We only
 1121          * consider autoconfigured addresses while RFC2462 simply said
 1122          * "address".
 1123          */
 1124         IF_ADDR_LOCK(ifp);
 1125         TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1126                 struct in6_ifaddr *ifa6;
 1127                 u_int32_t remaininglifetime;
 1128 
 1129                 if (ifa->ifa_addr->sa_family != AF_INET6)
 1130                         continue;
 1131 
 1132                 ifa6 = (struct in6_ifaddr *)ifa;
 1133 
 1134                 /*
 1135                  * We only consider autoconfigured addresses as per rfc2462bis.
 1136                  */
 1137                 if (!(ifa6->ia6_flags & IN6_IFF_AUTOCONF))
 1138                         continue;
 1139 
 1140                 /*
 1141                  * Spec is not clear here, but I believe we should concentrate
 1142                  * on unicast (i.e. not anycast) addresses.
 1143                  * XXX: other ia6_flags? detached or duplicated?
 1144                  */
 1145                 if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0)
 1146                         continue;
 1147 
 1148                 /*
 1149                  * Ignore the address if it is not associated with a prefix
 1150                  * or is associated with a prefix that is different from this
 1151                  * one.  (pr is never NULL here)
 1152                  */
 1153                 if (ifa6->ia6_ndpr != pr)
 1154                         continue;
 1155 
 1156                 if (ia6_match == NULL) /* remember the first one */
 1157                         ia6_match = ifa6;
 1158 
 1159                 /*
 1160                  * An already autoconfigured address matched.  Now that we
 1161                  * are sure there is at least one matched address, we can
 1162                  * proceed to 5.5.3. (e): update the lifetimes according to the
 1163                  * "two hours" rule and the privacy extension.
 1164                  * We apply some clarifications in rfc2462bis:
 1165                  * - use remaininglifetime instead of storedlifetime as a
 1166                  *   variable name
 1167                  * - remove the dead code in the "two-hour" rule
 1168                  */
 1169 #define TWOHOUR         (120*60)
 1170                 lt6_tmp = ifa6->ia6_lifetime;
 1171 
 1172                 if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME)
 1173                         remaininglifetime = ND6_INFINITE_LIFETIME;
 1174                 else if (time_second - ifa6->ia6_updatetime >
 1175                          lt6_tmp.ia6t_vltime) {
 1176                         /*
 1177                          * The case of "invalid" address.  We should usually
 1178                          * not see this case.
 1179                          */
 1180                         remaininglifetime = 0;
 1181                 } else
 1182                         remaininglifetime = lt6_tmp.ia6t_vltime -
 1183                             (time_second - ifa6->ia6_updatetime);
 1184 
 1185                 /* when not updating, keep the current stored lifetime. */
 1186                 lt6_tmp.ia6t_vltime = remaininglifetime;
 1187 
 1188                 if (TWOHOUR < new->ndpr_vltime ||
 1189                     remaininglifetime < new->ndpr_vltime) {
 1190                         lt6_tmp.ia6t_vltime = new->ndpr_vltime;
 1191                 } else if (remaininglifetime <= TWOHOUR) {
 1192                         if (auth) {
 1193                                 lt6_tmp.ia6t_vltime = new->ndpr_vltime;
 1194                         }
 1195                 } else {
 1196                         /*
 1197                          * new->ndpr_vltime <= TWOHOUR &&
 1198                          * TWOHOUR < remaininglifetime
 1199                          */
 1200                         lt6_tmp.ia6t_vltime = TWOHOUR;
 1201                 }
 1202 
 1203                 /* The 2 hour rule is not imposed for preferred lifetime. */
 1204                 lt6_tmp.ia6t_pltime = new->ndpr_pltime;
 1205 
 1206                 in6_init_address_ltimes(pr, &lt6_tmp);
 1207 
 1208                 /*
 1209                  * We need to treat lifetimes for temporary addresses
 1210                  * differently, according to
 1211                  * draft-ietf-ipv6-privacy-addrs-v2-01.txt 3.3 (1);
 1212                  * we only update the lifetimes when they are in the maximum
 1213                  * intervals.
 1214                  */
 1215                 if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0) {
 1216                         u_int32_t maxvltime, maxpltime;
 1217 
 1218                         if (V_ip6_temp_valid_lifetime >
 1219                             (u_int32_t)((time_second - ifa6->ia6_createtime) +
 1220                             V_ip6_desync_factor)) {
 1221                                 maxvltime = V_ip6_temp_valid_lifetime -
 1222                                     (time_second - ifa6->ia6_createtime) -
 1223                                     V_ip6_desync_factor;
 1224                         } else
 1225                                 maxvltime = 0;
 1226                         if (V_ip6_temp_preferred_lifetime >
 1227                             (u_int32_t)((time_second - ifa6->ia6_createtime) +
 1228                             V_ip6_desync_factor)) {
 1229                                 maxpltime = V_ip6_temp_preferred_lifetime -
 1230                                     (time_second - ifa6->ia6_createtime) -
 1231                                     V_ip6_desync_factor;
 1232                         } else
 1233                                 maxpltime = 0;
 1234 
 1235                         if (lt6_tmp.ia6t_vltime == ND6_INFINITE_LIFETIME ||
 1236                             lt6_tmp.ia6t_vltime > maxvltime) {
 1237                                 lt6_tmp.ia6t_vltime = maxvltime;
 1238                         }
 1239                         if (lt6_tmp.ia6t_pltime == ND6_INFINITE_LIFETIME ||
 1240                             lt6_tmp.ia6t_pltime > maxpltime) {
 1241                                 lt6_tmp.ia6t_pltime = maxpltime;
 1242                         }
 1243                 }
 1244                 ifa6->ia6_lifetime = lt6_tmp;
 1245                 ifa6->ia6_updatetime = time_second;
 1246         }
 1247         IF_ADDR_UNLOCK(ifp);
 1248         if (ia6_match == NULL && new->ndpr_vltime) {
 1249                 int ifidlen;
 1250 
 1251                 /*
 1252                  * 5.5.3 (d) (continued)
 1253                  * No address matched and the valid lifetime is non-zero.
 1254                  * Create a new address.
 1255                  */
 1256 
 1257                 /*
 1258                  * Prefix Length check:
 1259                  * If the sum of the prefix length and interface identifier
 1260                  * length does not equal 128 bits, the Prefix Information
 1261                  * option MUST be ignored.  The length of the interface
 1262                  * identifier is defined in a separate link-type specific
 1263                  * document.
 1264                  */
 1265                 ifidlen = in6_if2idlen(ifp);
 1266                 if (ifidlen < 0) {
 1267                         /* this should not happen, so we always log it. */
 1268                         log(LOG_ERR, "prelist_update: IFID undefined (%s)\n",
 1269                             if_name(ifp));
 1270                         goto end;
 1271                 }
 1272                 if (ifidlen + pr->ndpr_plen != 128) {
 1273                         nd6log((LOG_INFO,
 1274                             "prelist_update: invalid prefixlen "
 1275                             "%d for %s, ignored\n",
 1276                             pr->ndpr_plen, if_name(ifp)));
 1277                         goto end;
 1278                 }
 1279 
 1280                 if ((ia6 = in6_ifadd(new, mcast)) != NULL) {
 1281                         /*
 1282                          * note that we should use pr (not new) for reference.
 1283                          */
 1284                         pr->ndpr_refcnt++;
 1285                         ia6->ia6_ndpr = pr;
 1286 
 1287                         /*
 1288                          * RFC 3041 3.3 (2).
 1289                          * When a new public address is created as described
 1290                          * in RFC2462, also create a new temporary address.
 1291                          *
 1292                          * RFC 3041 3.5.
 1293                          * When an interface connects to a new link, a new
 1294                          * randomized interface identifier should be generated
 1295                          * immediately together with a new set of temporary
 1296                          * addresses.  Thus, we specifiy 1 as the 2nd arg of
 1297                          * in6_tmpifadd().
 1298                          */
 1299                         if (V_ip6_use_tempaddr) {
 1300                                 int e;
 1301                                 if ((e = in6_tmpifadd(ia6, 1, 1)) != 0) {
 1302                                         nd6log((LOG_NOTICE, "prelist_update: "
 1303                                             "failed to create a temporary "
 1304                                             "address, errno=%d\n",
 1305                                             e));
 1306                                 }
 1307                         }
 1308                         ifa_free(&ia6->ia_ifa);
 1309 
 1310                         /*
 1311                          * A newly added address might affect the status
 1312                          * of other addresses, so we check and update it.
 1313                          * XXX: what if address duplication happens?
 1314                          */
 1315                         pfxlist_onlink_check();
 1316                 } else {
 1317                         /* just set an error. do not bark here. */
 1318                         error = EADDRNOTAVAIL; /* XXX: might be unused. */
 1319                 }
 1320         }
 1321 
 1322  end:
 1323         splx(s);
 1324         return error;
 1325 }
 1326 
 1327 /*
 1328  * A supplement function used in the on-link detection below;
 1329  * detect if a given prefix has a (probably) reachable advertising router.
 1330  * XXX: lengthy function name...
 1331  */
 1332 static struct nd_pfxrouter *
 1333 find_pfxlist_reachable_router(struct nd_prefix *pr)
 1334 {
 1335         struct nd_pfxrouter *pfxrtr;
 1336         struct llentry *ln;
 1337         int canreach;
 1338 
 1339         for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr != NULL;
 1340              pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) {
 1341                 IF_AFDATA_LOCK(pfxrtr->router->ifp);
 1342                 ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp);
 1343                 IF_AFDATA_UNLOCK(pfxrtr->router->ifp);
 1344                 if (ln == NULL)
 1345                         continue;
 1346                 canreach = ND6_IS_LLINFO_PROBREACH(ln);
 1347                 LLE_RUNLOCK(ln);
 1348                 if (canreach)
 1349                         break;
 1350         }
 1351         return (pfxrtr);
 1352 }
 1353 
 1354 /*
 1355  * Check if each prefix in the prefix list has at least one available router
 1356  * that advertised the prefix (a router is "available" if its neighbor cache
 1357  * entry is reachable or probably reachable).
 1358  * If the check fails, the prefix may be off-link, because, for example,
 1359  * we have moved from the network but the lifetime of the prefix has not
 1360  * expired yet.  So we should not use the prefix if there is another prefix
 1361  * that has an available router.
 1362  * But, if there is no prefix that has an available router, we still regards
 1363  * all the prefixes as on-link.  This is because we can't tell if all the
 1364  * routers are simply dead or if we really moved from the network and there
 1365  * is no router around us.
 1366  */
 1367 void
 1368 pfxlist_onlink_check()
 1369 {
 1370         struct nd_prefix *pr;
 1371         struct in6_ifaddr *ifa;
 1372         struct nd_defrouter *dr;
 1373         struct nd_pfxrouter *pfxrtr = NULL;
 1374 
 1375         /*
 1376          * Check if there is a prefix that has a reachable advertising
 1377          * router.
 1378          */
 1379         for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
 1380                 if (pr->ndpr_raf_onlink && find_pfxlist_reachable_router(pr))
 1381                         break;
 1382         }
 1383 
 1384         /*
 1385          * If we have no such prefix, check whether we still have a router
 1386          * that does not advertise any prefixes.
 1387          */
 1388         if (pr == NULL) {
 1389                 for (dr = TAILQ_FIRST(&V_nd_defrouter); dr;
 1390                     dr = TAILQ_NEXT(dr, dr_entry)) {
 1391                         struct nd_prefix *pr0;
 1392 
 1393                         for (pr0 = V_nd_prefix.lh_first; pr0;
 1394                             pr0 = pr0->ndpr_next) {
 1395                                 if ((pfxrtr = pfxrtr_lookup(pr0, dr)) != NULL)
 1396                                         break;
 1397                         }
 1398                         if (pfxrtr != NULL)
 1399                                 break;
 1400                 }
 1401         }
 1402         if (pr != NULL || (TAILQ_FIRST(&V_nd_defrouter) && pfxrtr == NULL)) {
 1403                 /*
 1404                  * There is at least one prefix that has a reachable router,
 1405                  * or at least a router which probably does not advertise
 1406                  * any prefixes.  The latter would be the case when we move
 1407                  * to a new link where we have a router that does not provide
 1408                  * prefixes and we configure an address by hand.
 1409                  * Detach prefixes which have no reachable advertising
 1410                  * router, and attach other prefixes.
 1411                  */
 1412                 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
 1413                         /* XXX: a link-local prefix should never be detached */
 1414                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1415                                 continue;
 1416 
 1417                         /*
 1418                          * we aren't interested in prefixes without the L bit
 1419                          * set.
 1420                          */
 1421                         if (pr->ndpr_raf_onlink == 0)
 1422                                 continue;
 1423 
 1424                         if (pr->ndpr_raf_auto == 0)
 1425                                 continue;
 1426 
 1427                         if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
 1428                             find_pfxlist_reachable_router(pr) == NULL)
 1429                                 pr->ndpr_stateflags |= NDPRF_DETACHED;
 1430                         if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
 1431                             find_pfxlist_reachable_router(pr) != 0)
 1432                                 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
 1433                 }
 1434         } else {
 1435                 /* there is no prefix that has a reachable router */
 1436                 for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
 1437                         if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1438                                 continue;
 1439 
 1440                         if (pr->ndpr_raf_onlink == 0)
 1441                                 continue;
 1442 
 1443                         if (pr->ndpr_raf_auto == 0)
 1444                                 continue;
 1445 
 1446                         if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0)
 1447                                 pr->ndpr_stateflags &= ~NDPRF_DETACHED;
 1448                 }
 1449         }
 1450 
 1451         /*
 1452          * Remove each interface route associated with a (just) detached
 1453          * prefix, and reinstall the interface route for a (just) attached
 1454          * prefix.  Note that all attempt of reinstallation does not
 1455          * necessarily success, when a same prefix is shared among multiple
 1456          * interfaces.  Such cases will be handled in nd6_prefix_onlink,
 1457          * so we don't have to care about them.
 1458          */
 1459         for (pr = V_nd_prefix.lh_first; pr; pr = pr->ndpr_next) {
 1460                 int e;
 1461                 char ip6buf[INET6_ADDRSTRLEN];
 1462 
 1463                 if (IN6_IS_ADDR_LINKLOCAL(&pr->ndpr_prefix.sin6_addr))
 1464                         continue;
 1465 
 1466                 if (pr->ndpr_raf_onlink == 0)
 1467                         continue;
 1468 
 1469                 if (pr->ndpr_raf_auto == 0)
 1470                         continue;
 1471 
 1472                 if ((pr->ndpr_stateflags & NDPRF_DETACHED) != 0 &&
 1473                     (pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
 1474                         if ((e = nd6_prefix_offlink(pr)) != 0) {
 1475                                 nd6log((LOG_ERR,
 1476                                     "pfxlist_onlink_check: failed to "
 1477                                     "make %s/%d offlink, errno=%d\n",
 1478                                     ip6_sprintf(ip6buf,
 1479                                             &pr->ndpr_prefix.sin6_addr),
 1480                                             pr->ndpr_plen, e));
 1481                         }
 1482                 }
 1483                 if ((pr->ndpr_stateflags & NDPRF_DETACHED) == 0 &&
 1484                     (pr->ndpr_stateflags & NDPRF_ONLINK) == 0 &&
 1485                     pr->ndpr_raf_onlink) {
 1486                         if ((e = nd6_prefix_onlink(pr)) != 0) {
 1487                                 nd6log((LOG_ERR,
 1488                                     "pfxlist_onlink_check: failed to "
 1489                                     "make %s/%d onlink, errno=%d\n",
 1490                                     ip6_sprintf(ip6buf,
 1491                                             &pr->ndpr_prefix.sin6_addr),
 1492                                             pr->ndpr_plen, e));
 1493                         }
 1494                 }
 1495         }
 1496 
 1497         /*
 1498          * Changes on the prefix status might affect address status as well.
 1499          * Make sure that all addresses derived from an attached prefix are
 1500          * attached, and that all addresses derived from a detached prefix are
 1501          * detached.  Note, however, that a manually configured address should
 1502          * always be attached.
 1503          * The precise detection logic is same as the one for prefixes.
 1504          *
 1505          * XXXRW: in6_ifaddrhead locking.
 1506          */
 1507         TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
 1508                 if (!(ifa->ia6_flags & IN6_IFF_AUTOCONF))
 1509                         continue;
 1510 
 1511                 if (ifa->ia6_ndpr == NULL) {
 1512                         /*
 1513                          * This can happen when we first configure the address
 1514                          * (i.e. the address exists, but the prefix does not).
 1515                          * XXX: complicated relationships...
 1516                          */
 1517                         continue;
 1518                 }
 1519 
 1520                 if (find_pfxlist_reachable_router(ifa->ia6_ndpr))
 1521                         break;
 1522         }
 1523         if (ifa) {
 1524                 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
 1525                         if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1526                                 continue;
 1527 
 1528                         if (ifa->ia6_ndpr == NULL) /* XXX: see above. */
 1529                                 continue;
 1530 
 1531                         if (find_pfxlist_reachable_router(ifa->ia6_ndpr)) {
 1532                                 if (ifa->ia6_flags & IN6_IFF_DETACHED) {
 1533                                         ifa->ia6_flags &= ~IN6_IFF_DETACHED;
 1534                                         ifa->ia6_flags |= IN6_IFF_TENTATIVE;
 1535                                         nd6_dad_start((struct ifaddr *)ifa, 0);
 1536                                 }
 1537                         } else {
 1538                                 ifa->ia6_flags |= IN6_IFF_DETACHED;
 1539                         }
 1540                 }
 1541         }
 1542         else {
 1543                 TAILQ_FOREACH(ifa, &V_in6_ifaddrhead, ia_link) {
 1544                         if ((ifa->ia6_flags & IN6_IFF_AUTOCONF) == 0)
 1545                                 continue;
 1546 
 1547                         if (ifa->ia6_flags & IN6_IFF_DETACHED) {
 1548                                 ifa->ia6_flags &= ~IN6_IFF_DETACHED;
 1549                                 ifa->ia6_flags |= IN6_IFF_TENTATIVE;
 1550                                 /* Do we need a delay in this case? */
 1551                                 nd6_dad_start((struct ifaddr *)ifa, 0);
 1552                         }
 1553                 }
 1554         }
 1555 }
 1556 
 1557 static int
 1558 nd6_prefix_onlink(struct nd_prefix *pr)
 1559 {
 1560         struct ifaddr *ifa;
 1561         struct ifnet *ifp = pr->ndpr_ifp;
 1562         struct sockaddr_in6 mask6;
 1563         struct nd_prefix *opr;
 1564         u_long rtflags;
 1565         int error = 0;
 1566         struct radix_node_head *rnh;
 1567         struct rtentry *rt = NULL;
 1568         char ip6buf[INET6_ADDRSTRLEN];
 1569         struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1570 
 1571         /* sanity check */
 1572         if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) {
 1573                 nd6log((LOG_ERR,
 1574                     "nd6_prefix_onlink: %s/%d is already on-link\n",
 1575                     ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
 1576                     pr->ndpr_plen));
 1577                 return (EEXIST);
 1578         }
 1579 
 1580         /*
 1581          * Add the interface route associated with the prefix.  Before
 1582          * installing the route, check if there's the same prefix on another
 1583          * interface, and the prefix has already installed the interface route.
 1584          * Although such a configuration is expected to be rare, we explicitly
 1585          * allow it.
 1586          */
 1587         for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
 1588                 if (opr == pr)
 1589                         continue;
 1590 
 1591                 if ((opr->ndpr_stateflags & NDPRF_ONLINK) == 0)
 1592                         continue;
 1593 
 1594                 if (opr->ndpr_plen == pr->ndpr_plen &&
 1595                     in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
 1596                     &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen))
 1597                         return (0);
 1598         }
 1599 
 1600         /*
 1601          * We prefer link-local addresses as the associated interface address.
 1602          */
 1603         /* search for a link-local addr */
 1604         ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp,
 1605             IN6_IFF_NOTREADY | IN6_IFF_ANYCAST);
 1606         if (ifa == NULL) {
 1607                 /* XXX: freebsd does not have ifa_ifwithaf */
 1608                 IF_ADDR_LOCK(ifp);
 1609                 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
 1610                         if (ifa->ifa_addr->sa_family == AF_INET6)
 1611                                 break;
 1612                 }
 1613                 if (ifa != NULL)
 1614                         ifa_ref(ifa);
 1615                 IF_ADDR_UNLOCK(ifp);
 1616                 /* should we care about ia6_flags? */
 1617         }
 1618         if (ifa == NULL) {
 1619                 /*
 1620                  * This can still happen, when, for example, we receive an RA
 1621                  * containing a prefix with the L bit set and the A bit clear,
 1622                  * after removing all IPv6 addresses on the receiving
 1623                  * interface.  This should, of course, be rare though.
 1624                  */
 1625                 nd6log((LOG_NOTICE,
 1626                     "nd6_prefix_onlink: failed to find any ifaddr"
 1627                     " to add route for a prefix(%s/%d) on %s\n",
 1628                     ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
 1629                     pr->ndpr_plen, if_name(ifp)));
 1630                 return (0);
 1631         }
 1632 
 1633         /*
 1634          * in6_ifinit() sets nd6_rtrequest to ifa_rtrequest for all ifaddrs.
 1635          * ifa->ifa_rtrequest = nd6_rtrequest;
 1636          */
 1637         bzero(&mask6, sizeof(mask6));
 1638         mask6.sin6_len = sizeof(mask6);
 1639         mask6.sin6_addr = pr->ndpr_mask;
 1640         rtflags = (ifa->ifa_flags & ~IFA_RTSELF) | RTF_UP;
 1641         error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix,
 1642             ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt);
 1643         if (error == 0) {
 1644                 if (rt != NULL) /* this should be non NULL, though */ {
 1645                         rnh = rt_tables_get_rnh(rt->rt_fibnum, AF_INET6);
 1646                         /* XXX what if rhn == NULL? */
 1647                         RADIX_NODE_HEAD_LOCK(rnh);
 1648                         RT_LOCK(rt);
 1649                         if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) {
 1650                                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type =
 1651                                         rt->rt_ifp->if_type;
 1652                                 ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 1653                                         rt->rt_ifp->if_index;
 1654                         }
 1655                         RADIX_NODE_HEAD_UNLOCK(rnh);
 1656                         nd6_rtmsg(RTM_ADD, rt);
 1657                         RT_UNLOCK(rt);
 1658                 }
 1659                 pr->ndpr_stateflags |= NDPRF_ONLINK;
 1660         } else {
 1661                 char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN];
 1662                 nd6log((LOG_ERR, "nd6_prefix_onlink: failed to add route for a"
 1663                     " prefix (%s/%d) on %s, gw=%s, mask=%s, flags=%lx "
 1664                     "errno = %d\n",
 1665                     ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
 1666                     pr->ndpr_plen, if_name(ifp),
 1667                     ip6_sprintf(ip6bufg, &((struct sockaddr_in6 *)ifa->ifa_addr)->sin6_addr),
 1668                     ip6_sprintf(ip6bufm, &mask6.sin6_addr), rtflags, error));
 1669         }
 1670 
 1671         if (rt != NULL) {
 1672                 RT_LOCK(rt);
 1673                 RT_REMREF(rt);
 1674                 RT_UNLOCK(rt);
 1675         }
 1676         if (ifa != NULL)
 1677                 ifa_free(ifa);
 1678 
 1679         return (error);
 1680 }
 1681 
 1682 static int
 1683 nd6_prefix_offlink(struct nd_prefix *pr)
 1684 {
 1685         int error = 0;
 1686         struct ifnet *ifp = pr->ndpr_ifp;
 1687         struct nd_prefix *opr;
 1688         struct sockaddr_in6 sa6, mask6;
 1689         struct rtentry *rt = NULL;
 1690         char ip6buf[INET6_ADDRSTRLEN];
 1691 
 1692         /* sanity check */
 1693         if ((pr->ndpr_stateflags & NDPRF_ONLINK) == 0) {
 1694                 nd6log((LOG_ERR,
 1695                     "nd6_prefix_offlink: %s/%d is already off-link\n",
 1696                     ip6_sprintf(ip6buf, &pr->ndpr_prefix.sin6_addr),
 1697                     pr->ndpr_plen));
 1698                 return (EEXIST);
 1699         }
 1700 
 1701         bzero(&sa6, sizeof(sa6));
 1702         sa6.sin6_family = AF_INET6;
 1703         sa6.sin6_len = sizeof(sa6);
 1704         bcopy(&pr->ndpr_prefix.sin6_addr, &sa6.sin6_addr,
 1705             sizeof(struct in6_addr));
 1706         bzero(&mask6, sizeof(mask6));
 1707         mask6.sin6_family = AF_INET6;
 1708         mask6.sin6_len = sizeof(sa6);
 1709         bcopy(&pr->ndpr_mask, &mask6.sin6_addr, sizeof(struct in6_addr));
 1710         error = rtrequest(RTM_DELETE, (struct sockaddr *)&sa6, NULL,
 1711             (struct sockaddr *)&mask6, 0, &rt);
 1712         if (error == 0) {
 1713                 pr->ndpr_stateflags &= ~NDPRF_ONLINK;
 1714 
 1715                 /* report the route deletion to the routing socket. */
 1716                 if (rt != NULL)
 1717                         nd6_rtmsg(RTM_DELETE, rt);
 1718 
 1719                 /*
 1720                  * There might be the same prefix on another interface,
 1721                  * the prefix which could not be on-link just because we have
 1722                  * the interface route (see comments in nd6_prefix_onlink).
 1723                  * If there's one, try to make the prefix on-link on the
 1724                  * interface.
 1725                  */
 1726                 for (opr = V_nd_prefix.lh_first; opr; opr = opr->ndpr_next) {
 1727                         if (opr == pr)
 1728                                 continue;
 1729 
 1730                         if ((opr->ndpr_stateflags & NDPRF_ONLINK) != 0)
 1731                                 continue;
 1732 
 1733                         /*
 1734                          * KAME specific: detached prefixes should not be
 1735                          * on-link.
 1736                          */
 1737                         if ((opr->ndpr_stateflags & NDPRF_DETACHED) != 0)
 1738                                 continue;
 1739 
 1740                         if (opr->ndpr_plen == pr->ndpr_plen &&
 1741                             in6_are_prefix_equal(&pr->ndpr_prefix.sin6_addr,
 1742                             &opr->ndpr_prefix.sin6_addr, pr->ndpr_plen)) {
 1743                                 int e;
 1744 
 1745                                 if ((e = nd6_prefix_onlink(opr)) != 0) {
 1746                                         nd6log((LOG_ERR,
 1747                                             "nd6_prefix_offlink: failed to "
 1748                                             "recover a prefix %s/%d from %s "
 1749                                             "to %s (errno = %d)\n",
 1750                                             ip6_sprintf(ip6buf,
 1751                                                 &opr->ndpr_prefix.sin6_addr),
 1752                                             opr->ndpr_plen, if_name(ifp),
 1753                                             if_name(opr->ndpr_ifp), e));
 1754                                 }
 1755                         }
 1756                 }
 1757         } else {
 1758                 /* XXX: can we still set the NDPRF_ONLINK flag? */
 1759                 nd6log((LOG_ERR,
 1760                     "nd6_prefix_offlink: failed to delete route: "
 1761                     "%s/%d on %s (errno = %d)\n",
 1762                     ip6_sprintf(ip6buf, &sa6.sin6_addr), pr->ndpr_plen,
 1763                     if_name(ifp), error));
 1764         }
 1765 
 1766         if (rt != NULL) {
 1767                 RTFREE(rt);
 1768         }
 1769 
 1770         return (error);
 1771 }
 1772 
 1773 static struct in6_ifaddr *
 1774 in6_ifadd(struct nd_prefixctl *pr, int mcast)
 1775 {
 1776         struct ifnet *ifp = pr->ndpr_ifp;
 1777         struct ifaddr *ifa;
 1778         struct in6_aliasreq ifra;
 1779         struct in6_ifaddr *ia, *ib;
 1780         int error, plen0;
 1781         struct in6_addr mask;
 1782         int prefixlen = pr->ndpr_plen;
 1783         int updateflags;
 1784         char ip6buf[INET6_ADDRSTRLEN];
 1785 
 1786         in6_prefixlen2mask(&mask, prefixlen);
 1787 
 1788         /*
 1789          * find a link-local address (will be interface ID).
 1790          * Is it really mandatory? Theoretically, a global or a site-local
 1791          * address can be configured without a link-local address, if we
 1792          * have a unique interface identifier...
 1793          *
 1794          * it is not mandatory to have a link-local address, we can generate
 1795          * interface identifier on the fly.  we do this because:
 1796          * (1) it should be the easiest way to find interface identifier.
 1797          * (2) RFC2462 5.4 suggesting the use of the same interface identifier
 1798          * for multiple addresses on a single interface, and possible shortcut
 1799          * of DAD.  we omitted DAD for this reason in the past.
 1800          * (3) a user can prevent autoconfiguration of global address
 1801          * by removing link-local address by hand (this is partly because we
 1802          * don't have other way to control the use of IPv6 on an interface.
 1803          * this has been our design choice - cf. NRL's "ifconfig auto").
 1804          * (4) it is easier to manage when an interface has addresses
 1805          * with the same interface identifier, than to have multiple addresses
 1806          * with different interface identifiers.
 1807          */
 1808         ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0); /* 0 is OK? */
 1809         if (ifa)
 1810                 ib = (struct in6_ifaddr *)ifa;
 1811         else
 1812                 return NULL;
 1813 
 1814         /* prefixlen + ifidlen must be equal to 128 */
 1815         plen0 = in6_mask2len(&ib->ia_prefixmask.sin6_addr, NULL);
 1816         if (prefixlen != plen0) {
 1817                 ifa_free(ifa);
 1818                 nd6log((LOG_INFO, "in6_ifadd: wrong prefixlen for %s "
 1819                     "(prefix=%d ifid=%d)\n",
 1820                     if_name(ifp), prefixlen, 128 - plen0));
 1821                 return NULL;
 1822         }
 1823 
 1824         /* make ifaddr */
 1825 
 1826         bzero(&ifra, sizeof(ifra));
 1827         /*
 1828          * in6_update_ifa() does not use ifra_name, but we accurately set it
 1829          * for safety.
 1830          */
 1831         strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
 1832         ifra.ifra_addr.sin6_family = AF_INET6;
 1833         ifra.ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
 1834         /* prefix */
 1835         ifra.ifra_addr.sin6_addr = pr->ndpr_prefix.sin6_addr;
 1836         ifra.ifra_addr.sin6_addr.s6_addr32[0] &= mask.s6_addr32[0];
 1837         ifra.ifra_addr.sin6_addr.s6_addr32[1] &= mask.s6_addr32[1];
 1838         ifra.ifra_addr.sin6_addr.s6_addr32[2] &= mask.s6_addr32[2];
 1839         ifra.ifra_addr.sin6_addr.s6_addr32[3] &= mask.s6_addr32[3];
 1840 
 1841         /* interface ID */
 1842         ifra.ifra_addr.sin6_addr.s6_addr32[0] |=
 1843             (ib->ia_addr.sin6_addr.s6_addr32[0] & ~mask.s6_addr32[0]);
 1844         ifra.ifra_addr.sin6_addr.s6_addr32[1] |=
 1845             (ib->ia_addr.sin6_addr.s6_addr32[1] & ~mask.s6_addr32[1]);
 1846         ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
 1847             (ib->ia_addr.sin6_addr.s6_addr32[2] & ~mask.s6_addr32[2]);
 1848         ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
 1849             (ib->ia_addr.sin6_addr.s6_addr32[3] & ~mask.s6_addr32[3]);
 1850         ifa_free(ifa);
 1851 
 1852         /* new prefix mask. */
 1853         ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
 1854         ifra.ifra_prefixmask.sin6_family = AF_INET6;
 1855         bcopy(&mask, &ifra.ifra_prefixmask.sin6_addr,
 1856             sizeof(ifra.ifra_prefixmask.sin6_addr));
 1857 
 1858         /* lifetimes. */
 1859         ifra.ifra_lifetime.ia6t_vltime = pr->ndpr_vltime;
 1860         ifra.ifra_lifetime.ia6t_pltime = pr->ndpr_pltime;
 1861 
 1862         /* XXX: scope zone ID? */
 1863 
 1864         ifra.ifra_flags |= IN6_IFF_AUTOCONF; /* obey autoconf */
 1865 
 1866         /*
 1867          * Make sure that we do not have this address already.  This should
 1868          * usually not happen, but we can still see this case, e.g., if we
 1869          * have manually configured the exact address to be configured.
 1870          */
 1871         ifa = (struct ifaddr *)in6ifa_ifpwithaddr(ifp,
 1872             &ifra.ifra_addr.sin6_addr);
 1873         if (ifa != NULL) {
 1874                 ifa_free(ifa);
 1875                 /* this should be rare enough to make an explicit log */
 1876                 log(LOG_INFO, "in6_ifadd: %s is already configured\n",
 1877                     ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr));
 1878                 return (NULL);
 1879         }
 1880 
 1881         /*
 1882          * Allocate ifaddr structure, link into chain, etc.
 1883          * If we are going to create a new address upon receiving a multicasted
 1884          * RA, we need to impose a random delay before starting DAD.
 1885          * [draft-ietf-ipv6-rfc2462bis-02.txt, Section 5.4.2]
 1886          */
 1887         updateflags = 0;
 1888         if (mcast)
 1889                 updateflags |= IN6_IFAUPDATE_DADDELAY;
 1890         if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0) {
 1891                 nd6log((LOG_ERR,
 1892                     "in6_ifadd: failed to make ifaddr %s on %s (errno=%d)\n",
 1893                     ip6_sprintf(ip6buf, &ifra.ifra_addr.sin6_addr),
 1894                     if_name(ifp), error));
 1895                 return (NULL);  /* ifaddr must not have been allocated. */
 1896         }
 1897 
 1898         ia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
 1899         /*
 1900          * XXXRW: Assumption of non-NULLness here might not be true with
 1901          * fine-grained locking -- should we validate it?  Or just return
 1902          * earlier ifa rather than looking it up again?
 1903          */
 1904         return (ia);            /* this is always non-NULL  and referenced. */
 1905 }
 1906 
 1907 /*
 1908  * ia0 - corresponding public address
 1909  */
 1910 int
 1911 in6_tmpifadd(const struct in6_ifaddr *ia0, int forcegen, int delay)
 1912 {
 1913         struct ifnet *ifp = ia0->ia_ifa.ifa_ifp;
 1914         struct in6_ifaddr *newia, *ia;
 1915         struct in6_aliasreq ifra;
 1916         int i, error;
 1917         int trylimit = 3;       /* XXX: adhoc value */
 1918         int updateflags;
 1919         u_int32_t randid[2];
 1920         time_t vltime0, pltime0;
 1921 
 1922         bzero(&ifra, sizeof(ifra));
 1923         strncpy(ifra.ifra_name, if_name(ifp), sizeof(ifra.ifra_name));
 1924         ifra.ifra_addr = ia0->ia_addr;
 1925         /* copy prefix mask */
 1926         ifra.ifra_prefixmask = ia0->ia_prefixmask;
 1927         /* clear the old IFID */
 1928         for (i = 0; i < 4; i++) {
 1929                 ifra.ifra_addr.sin6_addr.s6_addr32[i] &=
 1930                     ifra.ifra_prefixmask.sin6_addr.s6_addr32[i];
 1931         }
 1932 
 1933   again:
 1934         if (in6_get_tmpifid(ifp, (u_int8_t *)randid,
 1935             (const u_int8_t *)&ia0->ia_addr.sin6_addr.s6_addr[8], forcegen)) {
 1936                 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to find a good "
 1937                     "random IFID\n"));
 1938                 return (EINVAL);
 1939         }
 1940         ifra.ifra_addr.sin6_addr.s6_addr32[2] |=
 1941             (randid[0] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[2]));
 1942         ifra.ifra_addr.sin6_addr.s6_addr32[3] |=
 1943             (randid[1] & ~(ifra.ifra_prefixmask.sin6_addr.s6_addr32[3]));
 1944 
 1945         /*
 1946          * in6_get_tmpifid() quite likely provided a unique interface ID.
 1947          * However, we may still have a chance to see collision, because
 1948          * there may be a time lag between generation of the ID and generation
 1949          * of the address.  So, we'll do one more sanity check.
 1950          */
 1951         IN6_IFADDR_RLOCK();
 1952         TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
 1953                 if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
 1954                     &ifra.ifra_addr.sin6_addr)) {
 1955                         if (trylimit-- == 0) {
 1956                                 IN6_IFADDR_RUNLOCK();
 1957                                 /*
 1958                                  * Give up.  Something strange should have
 1959                                  * happened.
 1960                                  */
 1961                                 nd6log((LOG_NOTICE, "in6_tmpifadd: failed to "
 1962                                     "find a unique random IFID\n"));
 1963                                 return (EEXIST);
 1964                         }
 1965                         IN6_IFADDR_RUNLOCK();
 1966                         forcegen = 1;
 1967                         goto again;
 1968                 }
 1969         }
 1970         IN6_IFADDR_RUNLOCK();
 1971 
 1972         /*
 1973          * The Valid Lifetime is the lower of the Valid Lifetime of the
 1974          * public address or TEMP_VALID_LIFETIME.
 1975          * The Preferred Lifetime is the lower of the Preferred Lifetime
 1976          * of the public address or TEMP_PREFERRED_LIFETIME -
 1977          * DESYNC_FACTOR.
 1978          */
 1979         if (ia0->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
 1980                 vltime0 = IFA6_IS_INVALID(ia0) ? 0 :
 1981                     (ia0->ia6_lifetime.ia6t_vltime -
 1982                     (time_second - ia0->ia6_updatetime));
 1983                 if (vltime0 > V_ip6_temp_valid_lifetime)
 1984                         vltime0 = V_ip6_temp_valid_lifetime;
 1985         } else
 1986                 vltime0 = V_ip6_temp_valid_lifetime;
 1987         if (ia0->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
 1988                 pltime0 = IFA6_IS_DEPRECATED(ia0) ? 0 :
 1989                     (ia0->ia6_lifetime.ia6t_pltime -
 1990                     (time_second - ia0->ia6_updatetime));
 1991                 if (pltime0 > V_ip6_temp_preferred_lifetime - V_ip6_desync_factor){
 1992                         pltime0 = V_ip6_temp_preferred_lifetime -
 1993                             V_ip6_desync_factor;
 1994                 }
 1995         } else
 1996                 pltime0 = V_ip6_temp_preferred_lifetime - V_ip6_desync_factor;
 1997         ifra.ifra_lifetime.ia6t_vltime = vltime0;
 1998         ifra.ifra_lifetime.ia6t_pltime = pltime0;
 1999 
 2000         /*
 2001          * A temporary address is created only if this calculated Preferred
 2002          * Lifetime is greater than REGEN_ADVANCE time units.
 2003          */
 2004         if (ifra.ifra_lifetime.ia6t_pltime <= V_ip6_temp_regen_advance)
 2005                 return (0);
 2006 
 2007         /* XXX: scope zone ID? */
 2008 
 2009         ifra.ifra_flags |= (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY);
 2010 
 2011         /* allocate ifaddr structure, link into chain, etc. */
 2012         updateflags = 0;
 2013         if (delay)
 2014                 updateflags |= IN6_IFAUPDATE_DADDELAY;
 2015         if ((error = in6_update_ifa(ifp, &ifra, NULL, updateflags)) != 0)
 2016                 return (error);
 2017 
 2018         newia = in6ifa_ifpwithaddr(ifp, &ifra.ifra_addr.sin6_addr);
 2019         if (newia == NULL) {    /* XXX: can it happen? */
 2020                 nd6log((LOG_ERR,
 2021                     "in6_tmpifadd: ifa update succeeded, but we got "
 2022                     "no ifaddr\n"));
 2023                 return (EINVAL); /* XXX */
 2024         }
 2025         newia->ia6_ndpr = ia0->ia6_ndpr;
 2026         newia->ia6_ndpr->ndpr_refcnt++;
 2027         ifa_free(&newia->ia_ifa);
 2028 
 2029         /*
 2030          * A newly added address might affect the status of other addresses.
 2031          * XXX: when the temporary address is generated with a new public
 2032          * address, the onlink check is redundant.  However, it would be safe
 2033          * to do the check explicitly everywhere a new address is generated,
 2034          * and, in fact, we surely need the check when we create a new
 2035          * temporary address due to deprecation of an old temporary address.
 2036          */
 2037         pfxlist_onlink_check();
 2038 
 2039         return (0);
 2040 }
 2041 
 2042 static int
 2043 in6_init_prefix_ltimes(struct nd_prefix *ndpr)
 2044 {
 2045         if (ndpr->ndpr_pltime == ND6_INFINITE_LIFETIME)
 2046                 ndpr->ndpr_preferred = 0;
 2047         else
 2048                 ndpr->ndpr_preferred = time_second + ndpr->ndpr_pltime;
 2049         if (ndpr->ndpr_vltime == ND6_INFINITE_LIFETIME)
 2050                 ndpr->ndpr_expire = 0;
 2051         else
 2052                 ndpr->ndpr_expire = time_second + ndpr->ndpr_vltime;
 2053 
 2054         return 0;
 2055 }
 2056 
 2057 static void
 2058 in6_init_address_ltimes(struct nd_prefix *new, struct in6_addrlifetime *lt6)
 2059 {
 2060         /* init ia6t_expire */
 2061         if (lt6->ia6t_vltime == ND6_INFINITE_LIFETIME)
 2062                 lt6->ia6t_expire = 0;
 2063         else {
 2064                 lt6->ia6t_expire = time_second;
 2065                 lt6->ia6t_expire += lt6->ia6t_vltime;
 2066         }
 2067 
 2068         /* init ia6t_preferred */
 2069         if (lt6->ia6t_pltime == ND6_INFINITE_LIFETIME)
 2070                 lt6->ia6t_preferred = 0;
 2071         else {
 2072                 lt6->ia6t_preferred = time_second;
 2073                 lt6->ia6t_preferred += lt6->ia6t_pltime;
 2074         }
 2075 }
 2076 
 2077 /*
 2078  * Delete all the routing table entries that use the specified gateway.
 2079  * XXX: this function causes search through all entries of routing table, so
 2080  * it shouldn't be called when acting as a router.
 2081  */
 2082 void
 2083 rt6_flush(struct in6_addr *gateway, struct ifnet *ifp)
 2084 {
 2085         struct radix_node_head *rnh;
 2086         int s = splnet();
 2087 
 2088         /* We'll care only link-local addresses */
 2089         if (!IN6_IS_ADDR_LINKLOCAL(gateway)) {
 2090                 splx(s);
 2091                 return;
 2092         }
 2093 
 2094         rnh = rt_tables_get_rnh(0, AF_INET6);
 2095         if (rnh == NULL)
 2096                 return;
 2097 
 2098         RADIX_NODE_HEAD_LOCK(rnh);
 2099         rnh->rnh_walktree(rnh, rt6_deleteroute, (void *)gateway);
 2100         RADIX_NODE_HEAD_UNLOCK(rnh);
 2101         splx(s);
 2102 }
 2103 
 2104 static int
 2105 rt6_deleteroute(struct radix_node *rn, void *arg)
 2106 {
 2107 #define SIN6(s) ((struct sockaddr_in6 *)s)
 2108         struct rtentry *rt = (struct rtentry *)rn;
 2109         struct in6_addr *gate = (struct in6_addr *)arg;
 2110 
 2111         if (rt->rt_gateway == NULL || rt->rt_gateway->sa_family != AF_INET6)
 2112                 return (0);
 2113 
 2114         if (!IN6_ARE_ADDR_EQUAL(gate, &SIN6(rt->rt_gateway)->sin6_addr)) {
 2115                 return (0);
 2116         }
 2117 
 2118         /*
 2119          * Do not delete a static route.
 2120          * XXX: this seems to be a bit ad-hoc. Should we consider the
 2121          * 'cloned' bit instead?
 2122          */
 2123         if ((rt->rt_flags & RTF_STATIC) != 0)
 2124                 return (0);
 2125 
 2126         /*
 2127          * We delete only host route. This means, in particular, we don't
 2128          * delete default route.
 2129          */
 2130         if ((rt->rt_flags & RTF_HOST) == 0)
 2131                 return (0);
 2132 
 2133         return (rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
 2134             rt_mask(rt), rt->rt_flags, 0));
 2135 #undef SIN6
 2136 }
 2137 
 2138 int
 2139 nd6_setdefaultiface(int ifindex)
 2140 {
 2141         int error = 0;
 2142 
 2143         if (ifindex < 0 || V_if_index < ifindex)
 2144                 return (EINVAL);
 2145         if (ifindex != 0 && !ifnet_byindex(ifindex))
 2146                 return (EINVAL);
 2147 
 2148         if (V_nd6_defifindex != ifindex) {
 2149                 V_nd6_defifindex = ifindex;
 2150                 if (V_nd6_defifindex > 0)
 2151                         V_nd6_defifp = ifnet_byindex(V_nd6_defifindex);
 2152                 else
 2153                         V_nd6_defifp = NULL;
 2154 
 2155                 /*
 2156                  * Our current implementation assumes one-to-one maping between
 2157                  * interfaces and links, so it would be natural to use the
 2158                  * default interface as the default link.
 2159                  */
 2160                 scope6_setdefault(V_nd6_defifp);
 2161         }
 2162 
 2163         return (error);
 2164 }

Cache object: 3b6480559144918c4d81c1ab93cf7f6a


[ 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.