SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 104 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 105 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 106
 107 #ifdef CONFIG_SYSCTL
 108 static void addrconf_sysctl_register(struct inet6_dev *idev);
 109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 110 #else
 111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 112 {
 113 }
 114
 115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 116 {
 117 }
 118 #endif
 119
 120 #ifdef CONFIG_IPV6_PRIVACY
 121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 123 static void ipv6_regen_rndid(unsigned long data);
 124
 125 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 126 #endif
 127
 128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 129 static int ipv6_count_addresses(struct inet6_dev *idev);
 130
 131 /*
 132  *      Configured unicast address hash table
 133  */
 134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 135 static DEFINE_SPINLOCK(addrconf_hash_lock);
 136
 137 static void addrconf_verify(unsigned long);
 138
 139 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 140 static DEFINE_SPINLOCK(addrconf_verify_lock);
 141
 142 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 143 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 144
 145 static void addrconf_type_change(struct net_device *dev,
 146                                  unsigned long event);
 147 static int addrconf_ifdown(struct net_device *dev, int how);
 148
 149 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 150 static void addrconf_dad_timer(unsigned long data);
 151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 152 static void addrconf_dad_run(struct inet6_dev *idev);
 153 static void addrconf_rs_timer(unsigned long data);
 154 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 155 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 156
 157 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 158                                 struct prefix_info *pinfo);
 159 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 160                                struct net_device *dev);
 161
 162 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 163
 164 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 165         .forwarding             = 0,
 166         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 167         .mtu6                   = IPV6_MIN_MTU,
 168         .accept_ra              = 1,
 169         .accept_redirects       = 1,
 170         .autoconf               = 1,
 171         .force_mld_version      = 0,
 172         .dad_transmits          = 1,
 173         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 174         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 175         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 176 #ifdef CONFIG_IPV6_PRIVACY
 177         .use_tempaddr           = 0,
 178         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 179         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 180         .regen_max_retry        = REGEN_MAX_RETRY,
 181         .max_desync_factor      = MAX_DESYNC_FACTOR,
 182 #endif
 183         .max_addresses          = IPV6_MAX_ADDRESSES,
 184         .accept_ra_defrtr       = 1,
 185         .accept_ra_pinfo        = 1,
 186 #ifdef CONFIG_IPV6_ROUTER_PREF
 187         .accept_ra_rtr_pref     = 1,
 188         .rtr_probe_interval     = 60 * HZ,
 189 #ifdef CONFIG_IPV6_ROUTE_INFO
 190         .accept_ra_rt_info_max_plen = 0,
 191 #endif
 192 #endif
 193         .proxy_ndp              = 0,
 194         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 195         .disable_ipv6           = 0,
 196         .accept_dad             = 1,
 197 };
 198
 199 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 200         .forwarding             = 0,
 201         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 202         .mtu6                   = IPV6_MIN_MTU,
 203         .accept_ra              = 1,
 204         .accept_redirects       = 1,
 205         .autoconf               = 1,
 206         .dad_transmits          = 1,
 207         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 208         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 209         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 210 #ifdef CONFIG_IPV6_PRIVACY
 211         .use_tempaddr           = 0,
 212         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 213         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 214         .regen_max_retry        = REGEN_MAX_RETRY,
 215         .max_desync_factor      = MAX_DESYNC_FACTOR,
 216 #endif
 217         .max_addresses          = IPV6_MAX_ADDRESSES,
 218         .accept_ra_defrtr       = 1,
 219         .accept_ra_pinfo        = 1,
 220 #ifdef CONFIG_IPV6_ROUTER_PREF
 221         .accept_ra_rtr_pref     = 1,
 222         .rtr_probe_interval     = 60 * HZ,
 223 #ifdef CONFIG_IPV6_ROUTE_INFO
 224         .accept_ra_rt_info_max_plen = 0,
 225 #endif
 226 #endif
 227         .proxy_ndp              = 0,
 228         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 229         .disable_ipv6           = 0,
 230         .accept_dad             = 1,
 231 };
 232
 233 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 234 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 235 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 236 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 237 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 238
 239 /* Check if a valid qdisc is available */
 240 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 241 {
 242         return !qdisc_tx_is_noop(dev);
 243 }
 244
 245 /* Check if a route is valid prefix route */
 246 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 247 {
 248         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 249 }
 250
 251 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 252 {
 253         if (del_timer(&ifp->timer))
 254                 __in6_ifa_put(ifp);
 255 }
 256
 257 enum addrconf_timer_t {
 258         AC_NONE,
 259         AC_DAD,
 260         AC_RS,
 261 };
 262
 263 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 264                                enum addrconf_timer_t what,
 265                                unsigned long when)
 266 {
 267         if (!del_timer(&ifp->timer))
 268                 in6_ifa_hold(ifp);
 269
 270         switch (what) {
 271         case AC_DAD:
 272                 ifp->timer.function = addrconf_dad_timer;
 273                 break;
 274         case AC_RS:
 275                 ifp->timer.function = addrconf_rs_timer;
 276                 break;
 277         default:
 278                 break;
 279         }
 280         ifp->timer.expires = jiffies + when;
 281         add_timer(&ifp->timer);
 282 }
 283
 284 static int snmp6_alloc_dev(struct inet6_dev *idev)
 285 {
 286         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                           sizeof(struct ipstats_mib)) < 0)
 288                 goto err_ip;
 289         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 290                           sizeof(struct icmpv6_mib)) < 0)
 291                 goto err_icmp;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 293                           sizeof(struct icmpv6msg_mib)) < 0)
 294                 goto err_icmpmsg;
 295
 296         return 0;
 297
 298 err_icmpmsg:
 299         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 300 err_icmp:
 301         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 302 err_ip:
 303         return -ENOMEM;
 304 }
 305
 306 static void snmp6_free_dev(struct inet6_dev *idev)
 307 {
 308         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 309         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 311 }
 312
 313 /* Nobody refers to this device, we may destroy it. */
 314
 315 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 316 {
 317         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 318         kfree(idev);
 319 }
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 ndev->dead = 1;
 379                 in6_dev_finish_destroy(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 printk(KERN_INFO
 419                        "%s: Disabled Privacy Extensions\n",
 420                        dev->name);
 421                 ndev->cnf.use_tempaddr = -1;
 422         } else {
 423                 in6_dev_hold(ndev);
 424                 ipv6_regen_rndid((unsigned long) ndev);
 425         }
 426 #endif
 427
 428         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 429                 ndev->if_flags |= IF_READY;
 430
 431         ipv6_mc_init_dev(ndev);
 432         ndev->tstamp = jiffies;
 433         addrconf_sysctl_register(ndev);
 434         /* protected by rtnl_lock */
 435         rcu_assign_pointer(dev->ip6_ptr, ndev);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         return ndev;
 441 }
 442
 443 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 444 {
 445         struct inet6_dev *idev;
 446
 447         ASSERT_RTNL();
 448
 449         idev = __in6_dev_get(dev);
 450         if (!idev) {
 451                 idev = ipv6_add_dev(dev);
 452                 if (!idev)
 453                         return NULL;
 454         }
 455
 456         if (dev->flags&IFF_UP)
 457                 ipv6_mc_up(idev);
 458         return idev;
 459 }
 460
 461 #ifdef CONFIG_SYSCTL
 462 static void dev_forward_change(struct inet6_dev *idev)
 463 {
 464         struct net_device *dev;
 465         struct inet6_ifaddr *ifa;
 466
 467         if (!idev)
 468                 return;
 469         dev = idev->dev;
 470         if (idev->cnf.forwarding)
 471                 dev_disable_lro(dev);
 472         if (dev && (dev->flags & IFF_MULTICAST)) {
 473                 if (idev->cnf.forwarding)
 474                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 475                 else
 476                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 477         }
 478
 479         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 480                 if (ifa->flags&IFA_F_TENTATIVE)
 481                         continue;
 482                 if (idev->cnf.forwarding)
 483                         addrconf_join_anycast(ifa);
 484                 else
 485                         addrconf_leave_anycast(ifa);
 486         }
 487 }
 488
 489
 490 static void addrconf_forward_change(struct net *net, __s32 newf)
 491 {
 492         struct net_device *dev;
 493         struct inet6_dev *idev;
 494
 495         rcu_read_lock();
 496         for_each_netdev_rcu(net, dev) {
 497                 idev = __in6_dev_get(dev);
 498                 if (idev) {
 499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 500                         idev->cnf.forwarding = newf;
 501                         if (changed)
 502                                 dev_forward_change(idev);
 503                 }
 504         }
 505         rcu_read_unlock();
 506 }
 507
 508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 509 {
 510         struct net *net;
 511
 512         net = (struct net *)table->extra2;
 513         if (p == &net->ipv6.devconf_dflt->forwarding)
 514                 return 0;
 515
 516         if (!rtnl_trylock()) {
 517                 /* Restore the original values before restarting */
 518                 *p = old;
 519                 return restart_syscall();
 520         }
 521
 522         if (p == &net->ipv6.devconf_all->forwarding) {
 523                 __s32 newf = net->ipv6.devconf_all->forwarding;
 524                 net->ipv6.devconf_dflt->forwarding = newf;
 525                 addrconf_forward_change(net, newf);
 526         } else if ((!*p) ^ (!old))
 527                 dev_forward_change((struct inet6_dev *)table->extra1);
 528         rtnl_unlock();
 529
 530         if (*p)
 531                 rt6_purge_dflt_routers(net);
 532         return 1;
 533 }
 534 #endif
 535
 536 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 537 {
 538         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 539         kfree(ifp);
 540 }
 541
 542 /* Nobody refers to this ifaddr, destroy it */
 543 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 544 {
 545         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 546
 547 #ifdef NET_REFCNT_DEBUG
 548         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 549 #endif
 550
 551         in6_dev_put(ifp->idev);
 552
 553         if (del_timer(&ifp->timer))
 554                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 555
 556         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 557                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 558                 return;
 559         }
 560         dst_release(&ifp->rt->u.dst);
 561
 562         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 563 }
 564
 565 static void
 566 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 567 {
 568         struct list_head *p;
 569         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 570
 571         /*
 572          * Each device address list is sorted in order of scope -
 573          * global before linklocal.
 574          */
 575         list_for_each(p, &idev->addr_list) {
 576                 struct inet6_ifaddr *ifa
 577                         = list_entry(p, struct inet6_ifaddr, if_list);
 578                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 579                         break;
 580         }
 581
 582         list_add_tail(&ifp->if_list, p);
 583 }
 584
 585 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 586 {
 587         /*
 588          * We perform the hash function over the last 64 bits of the address
 589          * This will include the IEEE address token on links that support it.
 590          */
 591         return jhash_2words((__force u32)addr->s6_addr32[2],
 592                             (__force u32)addr->s6_addr32[3], 0)
 593                 & (IN6_ADDR_HSIZE - 1);
 594 }
 595
 596 /* On success it returns ifp with increased reference count */
 597
 598 static struct inet6_ifaddr *
 599 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 600               int scope, u32 flags)
 601 {
 602         struct inet6_ifaddr *ifa = NULL;
 603         struct rt6_info *rt;
 604         unsigned int hash;
 605         int err = 0;
 606         int addr_type = ipv6_addr_type(addr);
 607
 608         if (addr_type == IPV6_ADDR_ANY ||
 609             addr_type & IPV6_ADDR_MULTICAST ||
 610             (!(idev->dev->flags & IFF_LOOPBACK) &&
 611              addr_type & IPV6_ADDR_LOOPBACK))
 612                 return ERR_PTR(-EADDRNOTAVAIL);
 613
 614         rcu_read_lock_bh();
 615         if (idev->dead) {
 616                 err = -ENODEV;                  /*XXX*/
 617                 goto out2;
 618         }
 619
 620         if (idev->cnf.disable_ipv6) {
 621                 err = -EACCES;
 622                 goto out2;
 623         }
 624
 625         spin_lock(&addrconf_hash_lock);
 626
 627         /* Ignore adding duplicate addresses on an interface */
 628         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 629                 ADBG(("ipv6_add_addr: already assigned\n"));
 630                 err = -EEXIST;
 631                 goto out;
 632         }
 633
 634         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 635
 636         if (ifa == NULL) {
 637                 ADBG(("ipv6_add_addr: malloc failed\n"));
 638                 err = -ENOBUFS;
 639                 goto out;
 640         }
 641
 642         rt = addrconf_dst_alloc(idev, addr, 0);
 643         if (IS_ERR(rt)) {
 644                 err = PTR_ERR(rt);
 645                 goto out;
 646         }
 647
 648         ipv6_addr_copy(&ifa->addr, addr);
 649
 650         spin_lock_init(&ifa->lock);
 651         spin_lock_init(&ifa->state_lock);
 652         init_timer(&ifa->timer);
 653         INIT_HLIST_NODE(&ifa->addr_lst);
 654         ifa->timer.data = (unsigned long) ifa;
 655         ifa->scope = scope;
 656         ifa->prefix_len = pfxlen;
 657         ifa->flags = flags | IFA_F_TENTATIVE;
 658         ifa->cstamp = ifa->tstamp = jiffies;
 659
 660         ifa->rt = rt;
 661
 662         /*
 663          * part one of RFC 4429, section 3.3
 664          * We should not configure an address as
 665          * optimistic if we do not yet know the link
 666          * layer address of our nexhop router
 667          */
 668
 669         if (rt->rt6i_nexthop == NULL)
 670                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 671
 672         ifa->idev = idev;
 673         in6_dev_hold(idev);
 674         /* For caller */
 675         in6_ifa_hold(ifa);
 676
 677         /* Add to big hash table */
 678         hash = ipv6_addr_hash(addr);
 679
 680         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 681         spin_unlock(&addrconf_hash_lock);
 682
 683         write_lock(&idev->lock);
 684         /* Add to inet6_dev unicast addr list. */
 685         ipv6_link_dev_addr(idev, ifa);
 686
 687 #ifdef CONFIG_IPV6_PRIVACY
 688         if (ifa->flags&IFA_F_TEMPORARY) {
 689                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 690                 in6_ifa_hold(ifa);
 691         }
 692 #endif
 693
 694         in6_ifa_hold(ifa);
 695         write_unlock(&idev->lock);
 696 out2:
 697         rcu_read_unlock_bh();
 698
 699         if (likely(err == 0))
 700                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 701         else {
 702                 kfree(ifa);
 703                 ifa = ERR_PTR(err);
 704         }
 705
 706         return ifa;
 707 out:
 708         spin_unlock(&addrconf_hash_lock);
 709         goto out2;
 710 }
 711
 712 /* This function wants to get referenced ifp and releases it before return */
 713
 714 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 715 {
 716         struct inet6_ifaddr *ifa, *ifn;
 717         struct inet6_dev *idev = ifp->idev;
 718         int state;
 719         int hash;
 720         int deleted = 0, onlink = 0;
 721         unsigned long expires = jiffies;
 722
 723         hash = ipv6_addr_hash(&ifp->addr);
 724
 725         spin_lock_bh(&ifp->state_lock);
 726         state = ifp->state;
 727         ifp->state = INET6_IFADDR_STATE_DEAD;
 728         spin_unlock_bh(&ifp->state_lock);
 729
 730         if (state == INET6_IFADDR_STATE_DEAD)
 731                 goto out;
 732
 733         spin_lock_bh(&addrconf_hash_lock);
 734         hlist_del_init_rcu(&ifp->addr_lst);
 735         spin_unlock_bh(&addrconf_hash_lock);
 736
 737         write_lock_bh(&idev->lock);
 738 #ifdef CONFIG_IPV6_PRIVACY
 739         if (ifp->flags&IFA_F_TEMPORARY) {
 740                 list_del(&ifp->tmp_list);
 741                 if (ifp->ifpub) {
 742                         in6_ifa_put(ifp->ifpub);
 743                         ifp->ifpub = NULL;
 744                 }
 745                 __in6_ifa_put(ifp);
 746         }
 747 #endif
 748
 749         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 750                 if (ifa == ifp) {
 751                         list_del_init(&ifp->if_list);
 752                         __in6_ifa_put(ifp);
 753
 754                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 755                                 break;
 756                         deleted = 1;
 757                         continue;
 758                 } else if (ifp->flags & IFA_F_PERMANENT) {
 759                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 760                                               ifp->prefix_len)) {
 761                                 if (ifa->flags & IFA_F_PERMANENT) {
 762                                         onlink = 1;
 763                                         if (deleted)
 764                                                 break;
 765                                 } else {
 766                                         unsigned long lifetime;
 767
 768                                         if (!onlink)
 769                                                 onlink = -1;
 770
 771                                         spin_lock(&ifa->lock);
 772
 773                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 774                                         /*
 775                                          * Note: Because this address is
 776                                          * not permanent, lifetime <
 777                                          * LONG_MAX / HZ here.
 778                                          */
 779                                         if (time_before(expires,
 780                                                         ifa->tstamp + lifetime * HZ))
 781                                                 expires = ifa->tstamp + lifetime * HZ;
 782                                         spin_unlock(&ifa->lock);
 783                                 }
 784                         }
 785                 }
 786         }
 787         write_unlock_bh(&idev->lock);
 788
 789         addrconf_del_timer(ifp);
 790
 791         ipv6_ifa_notify(RTM_DELADDR, ifp);
 792
 793         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 794
 795         /*
 796          * Purge or update corresponding prefix
 797          *
 798          * 1) we don't purge prefix here if address was not permanent.
 799          *    prefix is managed by its own lifetime.
 800          * 2) if there're no addresses, delete prefix.
 801          * 3) if there're still other permanent address(es),
 802          *    corresponding prefix is still permanent.
 803          * 4) otherwise, update prefix lifetime to the
 804          *    longest valid lifetime among the corresponding
 805          *    addresses on the device.
 806          *    Note: subsequent RA will update lifetime.
 807          *
 808          * --yoshfuji
 809          */
 810         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 811                 struct in6_addr prefix;
 812                 struct rt6_info *rt;
 813                 struct net *net = dev_net(ifp->idev->dev);
 814                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 815                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 816
 817                 if (rt && addrconf_is_prefix_route(rt)) {
 818                         if (onlink == 0) {
 819                                 ip6_del_rt(rt);
 820                                 rt = NULL;
 821                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 822                                 rt->rt6i_expires = expires;
 823                                 rt->rt6i_flags |= RTF_EXPIRES;
 824                         }
 825                 }
 826                 dst_release(&rt->u.dst);
 827         }
 828
 829 out:
 830         in6_ifa_put(ifp);
 831 }
 832
 833 #ifdef CONFIG_IPV6_PRIVACY
 834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 835 {
 836         struct inet6_dev *idev = ifp->idev;
 837         struct in6_addr addr, *tmpaddr;
 838         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 839         unsigned long regen_advance;
 840         int tmp_plen;
 841         int ret = 0;
 842         int max_addresses;
 843         u32 addr_flags;
 844
 845         write_lock(&idev->lock);
 846         if (ift) {
 847                 spin_lock_bh(&ift->lock);
 848                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 849                 spin_unlock_bh(&ift->lock);
 850                 tmpaddr = &addr;
 851         } else {
 852                 tmpaddr = NULL;
 853         }
 854 retry:
 855         in6_dev_hold(idev);
 856         if (idev->cnf.use_tempaddr <= 0) {
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_INFO
 859                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         spin_lock_bh(&ifp->lock);
 865         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 866                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         in6_ifa_hold(ifp);
 876         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 877         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 878                 spin_unlock_bh(&ifp->lock);
 879                 write_unlock(&idev->lock);
 880                 printk(KERN_WARNING
 881                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 882                 in6_ifa_put(ifp);
 883                 in6_dev_put(idev);
 884                 ret = -1;
 885                 goto out;
 886         }
 887         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 894         tmp_plen = ifp->prefix_len;
 895         max_addresses = idev->cnf.max_addresses;
 896         tmp_cstamp = ifp->cstamp;
 897         tmp_tstamp = ifp->tstamp;
 898         spin_unlock_bh(&ifp->lock);
 899
 900         regen_advance = idev->cnf.regen_max_retry *
 901                         idev->cnf.dad_transmits *
 902                         idev->nd_parms->retrans_time / HZ;
 903         write_unlock(&idev->lock);
 904
 905         /* A temporary address is created only if this calculated Preferred
 906          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 907          * an implementation must not create a temporary address with a zero
 908          * Preferred Lifetime.
 909          */
 910         if (tmp_prefered_lft <= regen_advance) {
 911                 in6_ifa_put(ifp);
 912                 in6_dev_put(idev);
 913                 ret = -1;
 914                 goto out;
 915         }
 916
 917         addr_flags = IFA_F_TEMPORARY;
 918         /* set in addrconf_prefix_rcv() */
 919         if (ifp->flags & IFA_F_OPTIMISTIC)
 920                 addr_flags |= IFA_F_OPTIMISTIC;
 921
 922         ift = !max_addresses ||
 923               ipv6_count_addresses(idev) < max_addresses ?
 924                 ipv6_add_addr(idev, &addr, tmp_plen,
 925                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 926                               addr_flags) : NULL;
 927         if (!ift || IS_ERR(ift)) {
 928                 in6_ifa_put(ifp);
 929                 in6_dev_put(idev);
 930                 printk(KERN_INFO
 931                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 932                 tmpaddr = &addr;
 933                 write_lock(&idev->lock);
 934                 goto retry;
 935         }
 936
 937         spin_lock_bh(&ift->lock);
 938         ift->ifpub = ifp;
 939         ift->valid_lft = tmp_valid_lft;
 940         ift->prefered_lft = tmp_prefered_lft;
 941         ift->cstamp = tmp_cstamp;
 942         ift->tstamp = tmp_tstamp;
 943         spin_unlock_bh(&ift->lock);
 944
 945         addrconf_dad_start(ift, 0);
 946         in6_ifa_put(ift);
 947         in6_dev_put(idev);
 948 out:
 949         return ret;
 950 }
 951 #endif
 952
 953 /*
 954  *      Choose an appropriate source address (RFC3484)
 955  */
 956 enum {
 957         IPV6_SADDR_RULE_INIT = 0,
 958         IPV6_SADDR_RULE_LOCAL,
 959         IPV6_SADDR_RULE_SCOPE,
 960         IPV6_SADDR_RULE_PREFERRED,
 961 #ifdef CONFIG_IPV6_MIP6
 962         IPV6_SADDR_RULE_HOA,
 963 #endif
 964         IPV6_SADDR_RULE_OIF,
 965         IPV6_SADDR_RULE_LABEL,
 966 #ifdef CONFIG_IPV6_PRIVACY
 967         IPV6_SADDR_RULE_PRIVACY,
 968 #endif
 969         IPV6_SADDR_RULE_ORCHID,
 970         IPV6_SADDR_RULE_PREFIX,
 971         IPV6_SADDR_RULE_MAX
 972 };
 973
 974 struct ipv6_saddr_score {
 975         int                     rule;
 976         int                     addr_type;
 977         struct inet6_ifaddr     *ifa;
 978         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 979         int                     scopedist;
 980         int                     matchlen;
 981 };
 982
 983 struct ipv6_saddr_dst {
 984         const struct in6_addr *addr;
 985         int ifindex;
 986         int scope;
 987         int label;
 988         unsigned int prefs;
 989 };
 990
 991 static inline int ipv6_saddr_preferred(int type)
 992 {
 993         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 994                 return 1;
 995         return 0;
 996 }
 997
 998 static int ipv6_get_saddr_eval(struct net *net,
 999                                struct ipv6_saddr_score *score,
1000                                struct ipv6_saddr_dst *dst,
1001                                int i)
1002 {
1003         int ret;
1004
1005         if (i <= score->rule) {
1006                 switch (i) {
1007                 case IPV6_SADDR_RULE_SCOPE:
1008                         ret = score->scopedist;
1009                         break;
1010                 case IPV6_SADDR_RULE_PREFIX:
1011                         ret = score->matchlen;
1012                         break;
1013                 default:
1014                         ret = !!test_bit(i, score->scorebits);
1015                 }
1016                 goto out;
1017         }
1018
1019         switch (i) {
1020         case IPV6_SADDR_RULE_INIT:
1021                 /* Rule 0: remember if hiscore is not ready yet */
1022                 ret = !!score->ifa;
1023                 break;
1024         case IPV6_SADDR_RULE_LOCAL:
1025                 /* Rule 1: Prefer same address */
1026                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027                 break;
1028         case IPV6_SADDR_RULE_SCOPE:
1029                 /* Rule 2: Prefer appropriate scope
1030                  *
1031                  *      ret
1032                  *       ^
1033                  *    -1 |  d 15
1034                  *    ---+--+-+---> scope
1035                  *       |
1036                  *       |             d is scope of the destination.
1037                  *  B-d  |  \
1038                  *       |   \      <- smaller scope is better if
1039                  *  B-15 |    \        if scope is enough for destinaion.
1040                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1041                  * d-C-1 | /
1042                  *       |/         <- greater is better
1043                  *   -C  /             if scope is not enough for destination.
1044                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1045                  *
1046                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048                  * Assume B = 0 and we get C > 29.
1049                  */
1050                 ret = __ipv6_addr_src_scope(score->addr_type);
1051                 if (ret >= dst->scope)
1052                         ret = -ret;
1053                 else
1054                         ret -= 128;     /* 30 is enough */
1055                 score->scopedist = ret;
1056                 break;
1057         case IPV6_SADDR_RULE_PREFERRED:
1058                 /* Rule 3: Avoid deprecated and optimistic addresses */
1059                 ret = ipv6_saddr_preferred(score->addr_type) ||
1060                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061                 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063         case IPV6_SADDR_RULE_HOA:
1064             {
1065                 /* Rule 4: Prefer home address */
1066                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068                 break;
1069             }
1070 #endif
1071         case IPV6_SADDR_RULE_OIF:
1072                 /* Rule 5: Prefer outgoing interface */
1073                 ret = (!dst->ifindex ||
1074                        dst->ifindex == score->ifa->idev->dev->ifindex);
1075                 break;
1076         case IPV6_SADDR_RULE_LABEL:
1077                 /* Rule 6: Prefer matching label */
1078                 ret = ipv6_addr_label(net,
1079                                       &score->ifa->addr, score->addr_type,
1080                                       score->ifa->idev->dev->ifindex) == dst->label;
1081                 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083         case IPV6_SADDR_RULE_PRIVACY:
1084             {
1085                 /* Rule 7: Prefer public address
1086                  * Note: prefer temprary address if use_tempaddr >= 2
1087                  */
1088                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1091                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092                 break;
1093             }
1094 #endif
1095         case IPV6_SADDR_RULE_ORCHID:
1096                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097                  *          non-ORCHID vs non-ORCHID
1098                  */
1099                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100                         ipv6_addr_orchid(dst->addr));
1101                 break;
1102         case IPV6_SADDR_RULE_PREFIX:
1103                 /* Rule 8: Use longest matching prefix */
1104                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105                                                        dst->addr);
1106                 break;
1107         default:
1108                 ret = 0;
1109         }
1110
1111         if (ret)
1112                 __set_bit(i, score->scorebits);
1113         score->rule = i;
1114 out:
1115         return ret;
1116 }
1117
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119                        const struct in6_addr *daddr, unsigned int prefs,
1120                        struct in6_addr *saddr)
1121 {
1122         struct ipv6_saddr_score scores[2],
1123                                 *score = &scores[0], *hiscore = &scores[1];
1124         struct ipv6_saddr_dst dst;
1125         struct net_device *dev;
1126         int dst_type;
1127
1128         dst_type = __ipv6_addr_type(daddr);
1129         dst.addr = daddr;
1130         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131         dst.scope = __ipv6_addr_src_scope(dst_type);
1132         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133         dst.prefs = prefs;
1134
1135         hiscore->rule = -1;
1136         hiscore->ifa = NULL;
1137
1138         rcu_read_lock();
1139
1140         for_each_netdev_rcu(net, dev) {
1141                 struct inet6_dev *idev;
1142
1143                 /* Candidate Source Address (section 4)
1144                  *  - multicast and link-local destination address,
1145                  *    the set of candidate source address MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same link as the outgoing
1148                  *    interface.
1149                  * (- For site-local destination addresses, the
1150                  *    set of candidate source addresses MUST only
1151                  *    include addresses assigned to interfaces
1152                  *    belonging to the same site as the outgoing
1153                  *    interface.)
1154                  */
1155                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1156                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1157                     dst.ifindex && dev->ifindex != dst.ifindex)
1158                         continue;
1159
1160                 idev = __in6_dev_get(dev);
1161                 if (!idev)
1162                         continue;
1163
1164                 read_lock_bh(&idev->lock);
1165                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1166                         int i;
1167
1168                         /*
1169                          * - Tentative Address (RFC2462 section 5.4)
1170                          *  - A tentative address is not considered
1171                          *    "assigned to an interface" in the traditional
1172                          *    sense, unless it is also flagged as optimistic.
1173                          * - Candidate Source Address (section 4)
1174                          *  - In any case, anycast addresses, multicast
1175                          *    addresses, and the unspecified address MUST
1176                          *    NOT be included in a candidate set.
1177                          */
1178                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1179                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1180                                 continue;
1181
1182                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1183
1184                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1185                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1186                                 LIMIT_NETDEBUG(KERN_DEBUG
1187                                                "ADDRCONF: unspecified / multicast address "
1188                                                "assigned as unicast address on %s",
1189                                                dev->name);
1190                                 continue;
1191                         }
1192
1193                         score->rule = -1;
1194                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1195
1196                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1197                                 int minihiscore, miniscore;
1198
1199                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1200                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1201
1202                                 if (minihiscore > miniscore) {
1203                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1204                                             score->scopedist > 0) {
1205                                                 /*
1206                                                  * special case:
1207                                                  * each remaining entry
1208                                                  * has too small (not enough)
1209                                                  * scope, because ifa entries
1210                                                  * are sorted by their scope
1211                                                  * values.
1212                                                  */
1213                                                 goto try_nextdev;
1214                                         }
1215                                         break;
1216                                 } else if (minihiscore < miniscore) {
1217                                         if (hiscore->ifa)
1218                                                 in6_ifa_put(hiscore->ifa);
1219
1220                                         in6_ifa_hold(score->ifa);
1221
1222                                         swap(hiscore, score);
1223
1224                                         /* restore our iterator */
1225                                         score->ifa = hiscore->ifa;
1226
1227                                         break;
1228                                 }
1229                         }
1230                 }
1231 try_nextdev:
1232                 read_unlock_bh(&idev->lock);
1233         }
1234         rcu_read_unlock();
1235
1236         if (!hiscore->ifa)
1237                 return -EADDRNOTAVAIL;
1238
1239         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240         in6_ifa_put(hiscore->ifa);
1241         return 0;
1242 }
1243 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1244
1245 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1246                     unsigned char banned_flags)
1247 {
1248         struct inet6_dev *idev;
1249         int err = -EADDRNOTAVAIL;
1250
1251         rcu_read_lock();
1252         idev = __in6_dev_get(dev);
1253         if (idev) {
1254                 struct inet6_ifaddr *ifp;
1255
1256                 read_lock_bh(&idev->lock);
1257                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1258                         if (ifp->scope == IFA_LINK &&
1259                             !(ifp->flags & banned_flags)) {
1260                                 ipv6_addr_copy(addr, &ifp->addr);
1261                                 err = 0;
1262                                 break;
1263                         }
1264                 }
1265                 read_unlock_bh(&idev->lock);
1266         }
1267         rcu_read_unlock();
1268         return err;
1269 }
1270
1271 static int ipv6_count_addresses(struct inet6_dev *idev)
1272 {
1273         int cnt = 0;
1274         struct inet6_ifaddr *ifp;
1275
1276         read_lock_bh(&idev->lock);
1277         list_for_each_entry(ifp, &idev->addr_list, if_list)
1278                 cnt++;
1279         read_unlock_bh(&idev->lock);
1280         return cnt;
1281 }
1282
1283 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1284                   struct net_device *dev, int strict)
1285 {
1286         struct inet6_ifaddr *ifp;
1287         struct hlist_node *node;
1288         unsigned int hash = ipv6_addr_hash(addr);
1289
1290         rcu_read_lock_bh();
1291         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1292                 if (!net_eq(dev_net(ifp->idev->dev), net))
1293                         continue;
1294                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295                     !(ifp->flags&IFA_F_TENTATIVE) &&
1296                     (dev == NULL || ifp->idev->dev == dev ||
1297                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1298                         rcu_read_unlock_bh();
1299                         return 1;
1300                 }
1301         }
1302
1303         rcu_read_unlock_bh();
1304         return 0;
1305 }
1306 EXPORT_SYMBOL(ipv6_chk_addr);
1307
1308 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                                struct net_device *dev)
1310 {
1311         unsigned int hash = ipv6_addr_hash(addr);
1312         struct inet6_ifaddr *ifp;
1313         struct hlist_node *node;
1314
1315         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1316                 if (!net_eq(dev_net(ifp->idev->dev), net))
1317                         continue;
1318                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1319                         if (dev == NULL || ifp->idev->dev == dev)
1320                                 return true;
1321                 }
1322         }
1323         return false;
1324 }
1325
1326 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1327 {
1328         struct inet6_dev *idev;
1329         struct inet6_ifaddr *ifa;
1330         int     onlink;
1331
1332         onlink = 0;
1333         rcu_read_lock();
1334         idev = __in6_dev_get(dev);
1335         if (idev) {
1336                 read_lock_bh(&idev->lock);
1337                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1338                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1339                                                    ifa->prefix_len);
1340                         if (onlink)
1341                                 break;
1342                 }
1343                 read_unlock_bh(&idev->lock);
1344         }
1345         rcu_read_unlock();
1346         return onlink;
1347 }
1348
1349 EXPORT_SYMBOL(ipv6_chk_prefix);
1350
1351 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1352                                      struct net_device *dev, int strict)
1353 {
1354         struct inet6_ifaddr *ifp, *result = NULL;
1355         unsigned int hash = ipv6_addr_hash(addr);
1356         struct hlist_node *node;
1357
1358         rcu_read_lock_bh();
1359         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1360                 if (!net_eq(dev_net(ifp->idev->dev), net))
1361                         continue;
1362                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1363                         if (dev == NULL || ifp->idev->dev == dev ||
1364                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1365                                 result = ifp;
1366                                 in6_ifa_hold(ifp);
1367                                 break;
1368                         }
1369                 }
1370         }
1371         rcu_read_unlock_bh();
1372
1373         return result;
1374 }
1375
1376 /* Gets referenced address, destroys ifaddr */
1377
1378 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1379 {
1380         if (ifp->flags&IFA_F_PERMANENT) {
1381                 spin_lock_bh(&ifp->lock);
1382                 addrconf_del_timer(ifp);
1383                 ifp->flags |= IFA_F_TENTATIVE;
1384                 if (dad_failed)
1385                         ifp->flags |= IFA_F_DADFAILED;
1386                 spin_unlock_bh(&ifp->lock);
1387                 if (dad_failed)
1388                         ipv6_ifa_notify(0, ifp);
1389                 in6_ifa_put(ifp);
1390 #ifdef CONFIG_IPV6_PRIVACY
1391         } else if (ifp->flags&IFA_F_TEMPORARY) {
1392                 struct inet6_ifaddr *ifpub;
1393                 spin_lock_bh(&ifp->lock);
1394                 ifpub = ifp->ifpub;
1395                 if (ifpub) {
1396                         in6_ifa_hold(ifpub);
1397                         spin_unlock_bh(&ifp->lock);
1398                         ipv6_create_tempaddr(ifpub, ifp);
1399                         in6_ifa_put(ifpub);
1400                 } else {
1401                         spin_unlock_bh(&ifp->lock);
1402                 }
1403                 ipv6_del_addr(ifp);
1404 #endif
1405         } else
1406                 ipv6_del_addr(ifp);
1407 }
1408
1409 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1410 {
1411         struct inet6_dev *idev = ifp->idev;
1412
1413         if (net_ratelimit())
1414                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1415                         ifp->idev->dev->name, &ifp->addr);
1416
1417         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1418                 struct in6_addr addr;
1419
1420                 addr.s6_addr32[0] = htonl(0xfe800000);
1421                 addr.s6_addr32[1] = 0;
1422
1423                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1424                     ipv6_addr_equal(&ifp->addr, &addr)) {
1425                         /* DAD failed for link-local based on MAC address */
1426                         idev->cnf.disable_ipv6 = 1;
1427
1428                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1429                                 ifp->idev->dev->name);
1430                 }
1431         }
1432
1433         addrconf_dad_stop(ifp, 1);
1434 }
1435
1436 /* Join to solicited addr multicast group. */
1437
1438 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1439 {
1440         struct in6_addr maddr;
1441
1442         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1443                 return;
1444
1445         addrconf_addr_solict_mult(addr, &maddr);
1446         ipv6_dev_mc_inc(dev, &maddr);
1447 }
1448
1449 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1450 {
1451         struct in6_addr maddr;
1452
1453         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1454                 return;
1455
1456         addrconf_addr_solict_mult(addr, &maddr);
1457         __ipv6_dev_mc_dec(idev, &maddr);
1458 }
1459
1460 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1461 {
1462         struct in6_addr addr;
1463         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1464         if (ipv6_addr_any(&addr))
1465                 return;
1466         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1467 }
1468
1469 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1470 {
1471         struct in6_addr addr;
1472         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1473         if (ipv6_addr_any(&addr))
1474                 return;
1475         __ipv6_dev_ac_dec(ifp->idev, &addr);
1476 }
1477
1478 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1479 {
1480         if (dev->addr_len != ETH_ALEN)
1481                 return -1;
1482         memcpy(eui, dev->dev_addr, 3);
1483         memcpy(eui + 5, dev->dev_addr + 3, 3);
1484
1485         /*
1486          * The zSeries OSA network cards can be shared among various
1487          * OS instances, but the OSA cards have only one MAC address.
1488          * This leads to duplicate address conflicts in conjunction
1489          * with IPv6 if more than one instance uses the same card.
1490          *
1491          * The driver for these cards can deliver a unique 16-bit
1492          * identifier for each instance sharing the same card.  It is
1493          * placed instead of 0xFFFE in the interface identifier.  The
1494          * "u" bit of the interface identifier is not inverted in this
1495          * case.  Hence the resulting interface identifier has local
1496          * scope according to RFC2373.
1497          */
1498         if (dev->dev_id) {
1499                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1500                 eui[4] = dev->dev_id & 0xFF;
1501         } else {
1502                 eui[3] = 0xFF;
1503                 eui[4] = 0xFE;
1504                 eui[0] ^= 2;
1505         }
1506         return 0;
1507 }
1508
1509 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1510 {
1511         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1512         if (dev->addr_len != ARCNET_ALEN)
1513                 return -1;
1514         memset(eui, 0, 7);
1515         eui[7] = *(u8*)dev->dev_addr;
1516         return 0;
1517 }
1518
1519 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1520 {
1521         if (dev->addr_len != INFINIBAND_ALEN)
1522                 return -1;
1523         memcpy(eui, dev->dev_addr + 12, 8);
1524         eui[0] |= 2;
1525         return 0;
1526 }
1527
1528 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1529 {
1530         if (addr == 0)
1531                 return -1;
1532         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1533                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1534                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1535                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1536                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1537                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1538         eui[1] = 0;
1539         eui[2] = 0x5E;
1540         eui[3] = 0xFE;
1541         memcpy(eui + 4, &addr, 4);
1542         return 0;
1543 }
1544 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1545
1546 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1547 {
1548         if (dev->priv_flags & IFF_ISATAP)
1549                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1550         return -1;
1551 }
1552
1553 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1554 {
1555         switch (dev->type) {
1556         case ARPHRD_ETHER:
1557         case ARPHRD_FDDI:
1558         case ARPHRD_IEEE802_TR:
1559                 return addrconf_ifid_eui48(eui, dev);
1560         case ARPHRD_ARCNET:
1561                 return addrconf_ifid_arcnet(eui, dev);
1562         case ARPHRD_INFINIBAND:
1563                 return addrconf_ifid_infiniband(eui, dev);
1564         case ARPHRD_SIT:
1565                 return addrconf_ifid_sit(eui, dev);
1566         }
1567         return -1;
1568 }
1569
1570 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1571 {
1572         int err = -1;
1573         struct inet6_ifaddr *ifp;
1574
1575         read_lock_bh(&idev->lock);
1576         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1577                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1578                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1579                         err = 0;
1580                         break;
1581                 }
1582         }
1583         read_unlock_bh(&idev->lock);
1584         return err;
1585 }
1586
1587 #ifdef CONFIG_IPV6_PRIVACY
1588 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1589 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1590 {
1591 regen:
1592         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1593         idev->rndid[0] &= ~0x02;
1594
1595         /*
1596          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1597          * check if generated address is not inappropriate
1598          *
1599          *  - Reserved subnet anycast (RFC 2526)
1600          *      11111101 11....11 1xxxxxxx
1601          *  - ISATAP (RFC4214) 6.1
1602          *      00-00-5E-FE-xx-xx-xx-xx
1603          *  - value 0
1604          *  - XXX: already assigned to an address on the device
1605          */
1606         if (idev->rndid[0] == 0xfd &&
1607             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1608             (idev->rndid[7]&0x80))
1609                 goto regen;
1610         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1611                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1612                         goto regen;
1613                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1614                         goto regen;
1615         }
1616
1617         return 0;
1618 }
1619
1620 static void ipv6_regen_rndid(unsigned long data)
1621 {
1622         struct inet6_dev *idev = (struct inet6_dev *) data;
1623         unsigned long expires;
1624
1625         rcu_read_lock_bh();
1626         write_lock_bh(&idev->lock);
1627
1628         if (idev->dead)
1629                 goto out;
1630
1631         if (__ipv6_regen_rndid(idev) < 0)
1632                 goto out;
1633
1634         expires = jiffies +
1635                 idev->cnf.temp_prefered_lft * HZ -
1636                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1637         if (time_before(expires, jiffies)) {
1638                 printk(KERN_WARNING
1639                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1640                         idev->dev->name);
1641                 goto out;
1642         }
1643
1644         if (!mod_timer(&idev->regen_timer, expires))
1645                 in6_dev_hold(idev);
1646
1647 out:
1648         write_unlock_bh(&idev->lock);
1649         rcu_read_unlock_bh();
1650         in6_dev_put(idev);
1651 }
1652
1653 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1654         int ret = 0;
1655
1656         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1657                 ret = __ipv6_regen_rndid(idev);
1658         return ret;
1659 }
1660 #endif
1661
1662 /*
1663  *      Add prefix route.
1664  */
1665
1666 static void
1667 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1668                       unsigned long expires, u32 flags)
1669 {
1670         struct fib6_config cfg = {
1671                 .fc_table = RT6_TABLE_PREFIX,
1672                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1673                 .fc_ifindex = dev->ifindex,
1674                 .fc_expires = expires,
1675                 .fc_dst_len = plen,
1676                 .fc_flags = RTF_UP | flags,
1677                 .fc_nlinfo.nl_net = dev_net(dev),
1678                 .fc_protocol = RTPROT_KERNEL,
1679         };
1680
1681         ipv6_addr_copy(&cfg.fc_dst, pfx);
1682
1683         /* Prevent useless cloning on PtP SIT.
1684            This thing is done here expecting that the whole
1685            class of non-broadcast devices need not cloning.
1686          */
1687 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1688         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1689                 cfg.fc_flags |= RTF_NONEXTHOP;
1690 #endif
1691
1692         ip6_route_add(&cfg);
1693 }
1694
1695 /* Create "default" multicast route to the interface */
1696
1697 static void addrconf_add_mroute(struct net_device *dev)
1698 {
1699         struct fib6_config cfg = {
1700                 .fc_table = RT6_TABLE_LOCAL,
1701                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1702                 .fc_ifindex = dev->ifindex,
1703                 .fc_dst_len = 8,
1704                 .fc_flags = RTF_UP,
1705                 .fc_nlinfo.nl_net = dev_net(dev),
1706         };
1707
1708         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1709
1710         ip6_route_add(&cfg);
1711 }
1712
1713 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1714 static void sit_route_add(struct net_device *dev)
1715 {
1716         struct fib6_config cfg = {
1717                 .fc_table = RT6_TABLE_MAIN,
1718                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1719                 .fc_ifindex = dev->ifindex,
1720                 .fc_dst_len = 96,
1721                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1722                 .fc_nlinfo.nl_net = dev_net(dev),
1723         };
1724
1725         /* prefix length - 96 bits "::d.d.d.d" */
1726         ip6_route_add(&cfg);
1727 }
1728 #endif
1729
1730 static void addrconf_add_lroute(struct net_device *dev)
1731 {
1732         struct in6_addr addr;
1733
1734         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1735         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1736 }
1737
1738 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1739 {
1740         struct inet6_dev *idev;
1741
1742         ASSERT_RTNL();
1743
1744         idev = ipv6_find_idev(dev);
1745         if (!idev)
1746                 return NULL;
1747
1748         /* Add default multicast route */
1749         addrconf_add_mroute(dev);
1750
1751         /* Add link local route */
1752         addrconf_add_lroute(dev);
1753         return idev;
1754 }
1755
1756 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1757 {
1758         struct prefix_info *pinfo;
1759         __u32 valid_lft;
1760         __u32 prefered_lft;
1761         int addr_type;
1762         struct inet6_dev *in6_dev;
1763         struct net *net = dev_net(dev);
1764
1765         pinfo = (struct prefix_info *) opt;
1766
1767         if (len < sizeof(struct prefix_info)) {
1768                 ADBG(("addrconf: prefix option too short\n"));
1769                 return;
1770         }
1771
1772         /*
1773          *      Validation checks ([ADDRCONF], page 19)
1774          */
1775
1776         addr_type = ipv6_addr_type(&pinfo->prefix);
1777
1778         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1779                 return;
1780
1781         valid_lft = ntohl(pinfo->valid);
1782         prefered_lft = ntohl(pinfo->prefered);
1783
1784         if (prefered_lft > valid_lft) {
1785                 if (net_ratelimit())
1786                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1787                 return;
1788         }
1789
1790         in6_dev = in6_dev_get(dev);
1791
1792         if (in6_dev == NULL) {
1793                 if (net_ratelimit())
1794                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1795                 return;
1796         }
1797
1798         /*
1799          *      Two things going on here:
1800          *      1) Add routes for on-link prefixes
1801          *      2) Configure prefixes with the auto flag set
1802          */
1803
1804         if (pinfo->onlink) {
1805                 struct rt6_info *rt;
1806                 unsigned long rt_expires;
1807
1808                 /* Avoid arithmetic overflow. Really, we could
1809                  * save rt_expires in seconds, likely valid_lft,
1810                  * but it would require division in fib gc, that it
1811                  * not good.
1812                  */
1813                 if (HZ > USER_HZ)
1814                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1815                 else
1816                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1817
1818                 if (addrconf_finite_timeout(rt_expires))
1819                         rt_expires *= HZ;
1820
1821                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1822                                 dev->ifindex, 1);
1823
1824                 if (rt && addrconf_is_prefix_route(rt)) {
1825                         /* Autoconf prefix route */
1826                         if (valid_lft == 0) {
1827                                 ip6_del_rt(rt);
1828                                 rt = NULL;
1829                         } else if (addrconf_finite_timeout(rt_expires)) {
1830                                 /* not infinity */
1831                                 rt->rt6i_expires = jiffies + rt_expires;
1832                                 rt->rt6i_flags |= RTF_EXPIRES;
1833                         } else {
1834                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1835                                 rt->rt6i_expires = 0;
1836                         }
1837                 } else if (valid_lft) {
1838                         clock_t expires = 0;
1839                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1840                         if (addrconf_finite_timeout(rt_expires)) {
1841                                 /* not infinity */
1842                                 flags |= RTF_EXPIRES;
1843                                 expires = jiffies_to_clock_t(rt_expires);
1844                         }
1845                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1846                                               dev, expires, flags);
1847                 }
1848                 if (rt)
1849                         dst_release(&rt->u.dst);
1850         }
1851
1852         /* Try to figure out our local address for this prefix */
1853
1854         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1855                 struct inet6_ifaddr * ifp;
1856                 struct in6_addr addr;
1857                 int create = 0, update_lft = 0;
1858
1859                 if (pinfo->prefix_len == 64) {
1860                         memcpy(&addr, &pinfo->prefix, 8);
1861                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1862                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1863                                 in6_dev_put(in6_dev);
1864                                 return;
1865                         }
1866                         goto ok;
1867                 }
1868                 if (net_ratelimit())
1869                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1870                                pinfo->prefix_len);
1871                 in6_dev_put(in6_dev);
1872                 return;
1873
1874 ok:
1875
1876                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1877
1878                 if (ifp == NULL && valid_lft) {
1879                         int max_addresses = in6_dev->cnf.max_addresses;
1880                         u32 addr_flags = 0;
1881
1882 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1883                         if (in6_dev->cnf.optimistic_dad &&
1884                             !net->ipv6.devconf_all->forwarding)
1885                                 addr_flags = IFA_F_OPTIMISTIC;
1886 #endif
1887
1888                         /* Do not allow to create too much of autoconfigured
1889                          * addresses; this would be too easy way to crash kernel.
1890                          */
1891                         if (!max_addresses ||
1892                             ipv6_count_addresses(in6_dev) < max_addresses)
1893                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1894                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1895                                                     addr_flags);
1896
1897                         if (!ifp || IS_ERR(ifp)) {
1898                                 in6_dev_put(in6_dev);
1899                                 return;
1900                         }
1901
1902                         update_lft = create = 1;
1903                         ifp->cstamp = jiffies;
1904                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1905                 }
1906
1907                 if (ifp) {
1908                         int flags;
1909                         unsigned long now;
1910 #ifdef CONFIG_IPV6_PRIVACY
1911                         struct inet6_ifaddr *ift;
1912 #endif
1913                         u32 stored_lft;
1914
1915                         /* update lifetime (RFC2462 5.5.3 e) */
1916                         spin_lock(&ifp->lock);
1917                         now = jiffies;
1918                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1919                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1920                         else
1921                                 stored_lft = 0;
1922                         if (!update_lft && stored_lft) {
1923                                 if (valid_lft > MIN_VALID_LIFETIME ||
1924                                     valid_lft > stored_lft)
1925                                         update_lft = 1;
1926                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1927                                         /* valid_lft <= stored_lft is always true */
1928                                         /*
1929                                          * RFC 4862 Section 5.5.3e:
1930                                          * "Note that the preferred lifetime of
1931                                          *  the corresponding address is always
1932                                          *  reset to the Preferred Lifetime in
1933                                          *  the received Prefix Information
1934                                          *  option, regardless of whether the
1935                                          *  valid lifetime is also reset or
1936                                          *  ignored."
1937                                          *
1938                                          *  So if the preferred lifetime in
1939                                          *  this advertisement is different
1940                                          *  than what we have stored, but the
1941                                          *  valid lifetime is invalid, just
1942                                          *  reset prefered_lft.
1943                                          *
1944                                          *  We must set the valid lifetime
1945                                          *  to the stored lifetime since we'll
1946                                          *  be updating the timestamp below,
1947                                          *  else we'll set it back to the
1948                                          *  minumum.
1949                                          */
1950                                         if (prefered_lft != ifp->prefered_lft) {
1951                                                 valid_lft = stored_lft;
1952                                                 update_lft = 1;
1953                                         }
1954                                 } else {
1955                                         valid_lft = MIN_VALID_LIFETIME;
1956                                         if (valid_lft < prefered_lft)
1957                                                 prefered_lft = valid_lft;
1958                                         update_lft = 1;
1959                                 }
1960                         }
1961
1962                         if (update_lft) {
1963                                 ifp->valid_lft = valid_lft;
1964                                 ifp->prefered_lft = prefered_lft;
1965                                 ifp->tstamp = now;
1966                                 flags = ifp->flags;
1967                                 ifp->flags &= ~IFA_F_DEPRECATED;
1968                                 spin_unlock(&ifp->lock);
1969
1970                                 if (!(flags&IFA_F_TENTATIVE))
1971                                         ipv6_ifa_notify(0, ifp);
1972                         } else
1973                                 spin_unlock(&ifp->lock);
1974
1975 #ifdef CONFIG_IPV6_PRIVACY
1976                         read_lock_bh(&in6_dev->lock);
1977                         /* update all temporary addresses in the list */
1978                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1979                                 /*
1980                                  * When adjusting the lifetimes of an existing
1981                                  * temporary address, only lower the lifetimes.
1982                                  * Implementations must not increase the
1983                                  * lifetimes of an existing temporary address
1984                                  * when processing a Prefix Information Option.
1985                                  */
1986                                 if (ifp != ift->ifpub)
1987                                         continue;
1988
1989                                 spin_lock(&ift->lock);
1990                                 flags = ift->flags;
1991                                 if (ift->valid_lft > valid_lft &&
1992                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1993                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1994                                 if (ift->prefered_lft > prefered_lft &&
1995                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1996                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1997                                 spin_unlock(&ift->lock);
1998                                 if (!(flags&IFA_F_TENTATIVE))
1999                                         ipv6_ifa_notify(0, ift);
2000                         }
2001
2002                         if (create && in6_dev->cnf.use_tempaddr > 0) {
2003                                 /*
2004                                  * When a new public address is created as described in [ADDRCONF],
2005                                  * also create a new temporary address.
2006                                  */
2007                                 read_unlock_bh(&in6_dev->lock);
2008                                 ipv6_create_tempaddr(ifp, NULL);
2009                         } else {
2010                                 read_unlock_bh(&in6_dev->lock);
2011                         }
2012 #endif
2013                         in6_ifa_put(ifp);
2014                         addrconf_verify(0);
2015                 }
2016         }
2017         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2018         in6_dev_put(in6_dev);
2019 }
2020
2021 /*
2022  *      Set destination address.
2023  *      Special case for SIT interfaces where we create a new "virtual"
2024  *      device.
2025  */
2026 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2027 {
2028         struct in6_ifreq ireq;
2029         struct net_device *dev;
2030         int err = -EINVAL;
2031
2032         rtnl_lock();
2033
2034         err = -EFAULT;
2035         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2036                 goto err_exit;
2037
2038         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2039
2040         err = -ENODEV;
2041         if (dev == NULL)
2042                 goto err_exit;
2043
2044 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2045         if (dev->type == ARPHRD_SIT) {
2046                 const struct net_device_ops *ops = dev->netdev_ops;
2047                 struct ifreq ifr;
2048                 struct ip_tunnel_parm p;
2049
2050                 err = -EADDRNOTAVAIL;
2051                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2052                         goto err_exit;
2053
2054                 memset(&p, 0, sizeof(p));
2055                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2056                 p.iph.saddr = 0;
2057                 p.iph.version = 4;
2058                 p.iph.ihl = 5;
2059                 p.iph.protocol = IPPROTO_IPV6;
2060                 p.iph.ttl = 64;
2061                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2062
2063                 if (ops->ndo_do_ioctl) {
2064                         mm_segment_t oldfs = get_fs();
2065
2066                         set_fs(KERNEL_DS);
2067                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2068                         set_fs(oldfs);
2069                 } else
2070                         err = -EOPNOTSUPP;
2071
2072                 if (err == 0) {
2073                         err = -ENOBUFS;
2074                         dev = __dev_get_by_name(net, p.name);
2075                         if (!dev)
2076                                 goto err_exit;
2077                         err = dev_open(dev);
2078                 }
2079         }
2080 #endif
2081
2082 err_exit:
2083         rtnl_unlock();
2084         return err;
2085 }
2086
2087 /*
2088  *      Manual configuration of address on an interface
2089  */
2090 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2091                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2092                           __u32 valid_lft)
2093 {
2094         struct inet6_ifaddr *ifp;
2095         struct inet6_dev *idev;
2096         struct net_device *dev;
2097         int scope;
2098         u32 flags;
2099         clock_t expires;
2100         unsigned long timeout;
2101
2102         ASSERT_RTNL();
2103
2104         if (plen > 128)
2105                 return -EINVAL;
2106
2107         /* check the lifetime */
2108         if (!valid_lft || prefered_lft > valid_lft)
2109                 return -EINVAL;
2110
2111         dev = __dev_get_by_index(net, ifindex);
2112         if (!dev)
2113                 return -ENODEV;
2114
2115         if ((idev = addrconf_add_dev(dev)) == NULL)
2116                 return -ENOBUFS;
2117
2118         scope = ipv6_addr_scope(pfx);
2119
2120         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2121         if (addrconf_finite_timeout(timeout)) {
2122                 expires = jiffies_to_clock_t(timeout * HZ);
2123                 valid_lft = timeout;
2124                 flags = RTF_EXPIRES;
2125         } else {
2126                 expires = 0;
2127                 flags = 0;
2128                 ifa_flags |= IFA_F_PERMANENT;
2129         }
2130
2131         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2132         if (addrconf_finite_timeout(timeout)) {
2133                 if (timeout == 0)
2134                         ifa_flags |= IFA_F_DEPRECATED;
2135                 prefered_lft = timeout;
2136         }
2137
2138         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2139
2140         if (!IS_ERR(ifp)) {
2141                 spin_lock_bh(&ifp->lock);
2142                 ifp->valid_lft = valid_lft;
2143                 ifp->prefered_lft = prefered_lft;
2144                 ifp->tstamp = jiffies;
2145                 spin_unlock_bh(&ifp->lock);
2146
2147                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2148                                       expires, flags);
2149                 /*
2150                  * Note that section 3.1 of RFC 4429 indicates
2151                  * that the Optimistic flag should not be set for
2152                  * manually configured addresses
2153                  */
2154                 addrconf_dad_start(ifp, 0);
2155                 in6_ifa_put(ifp);
2156                 addrconf_verify(0);
2157                 return 0;
2158         }
2159
2160         return PTR_ERR(ifp);
2161 }
2162
2163 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2164                           unsigned int plen)
2165 {
2166         struct inet6_ifaddr *ifp;
2167         struct inet6_dev *idev;
2168         struct net_device *dev;
2169
2170         if (plen > 128)
2171                 return -EINVAL;
2172
2173         dev = __dev_get_by_index(net, ifindex);
2174         if (!dev)
2175                 return -ENODEV;
2176
2177         if ((idev = __in6_dev_get(dev)) == NULL)
2178                 return -ENXIO;
2179
2180         read_lock_bh(&idev->lock);
2181         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2182                 if (ifp->prefix_len == plen &&
2183                     ipv6_addr_equal(pfx, &ifp->addr)) {
2184                         in6_ifa_hold(ifp);
2185                         read_unlock_bh(&idev->lock);
2186
2187                         ipv6_del_addr(ifp);
2188
2189                         /* If the last address is deleted administratively,
2190                            disable IPv6 on this interface.
2191                          */
2192                         if (list_empty(&idev->addr_list))
2193                                 addrconf_ifdown(idev->dev, 1);
2194                         return 0;
2195                 }
2196         }
2197         read_unlock_bh(&idev->lock);
2198         return -EADDRNOTAVAIL;
2199 }
2200
2201
2202 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2203 {
2204         struct in6_ifreq ireq;
2205         int err;
2206
2207         if (!capable(CAP_NET_ADMIN))
2208                 return -EPERM;
2209
2210         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2211                 return -EFAULT;
2212
2213         rtnl_lock();
2214         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2215                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2216                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2217         rtnl_unlock();
2218         return err;
2219 }
2220
2221 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2222 {
2223         struct in6_ifreq ireq;
2224         int err;
2225
2226         if (!capable(CAP_NET_ADMIN))
2227                 return -EPERM;
2228
2229         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2230                 return -EFAULT;
2231
2232         rtnl_lock();
2233         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2234                              ireq.ifr6_prefixlen);
2235         rtnl_unlock();
2236         return err;
2237 }
2238
2239 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2240                      int plen, int scope)
2241 {
2242         struct inet6_ifaddr *ifp;
2243
2244         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2245         if (!IS_ERR(ifp)) {
2246                 spin_lock_bh(&ifp->lock);
2247                 ifp->flags &= ~IFA_F_TENTATIVE;
2248                 spin_unlock_bh(&ifp->lock);
2249                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2250                 in6_ifa_put(ifp);
2251         }
2252 }
2253
2254 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2255 static void sit_add_v4_addrs(struct inet6_dev *idev)
2256 {
2257         struct in6_addr addr;
2258         struct net_device *dev;
2259         struct net *net = dev_net(idev->dev);
2260         int scope;
2261
2262         ASSERT_RTNL();
2263
2264         memset(&addr, 0, sizeof(struct in6_addr));
2265         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2266
2267         if (idev->dev->flags&IFF_POINTOPOINT) {
2268                 addr.s6_addr32[0] = htonl(0xfe800000);
2269                 scope = IFA_LINK;
2270         } else {
2271                 scope = IPV6_ADDR_COMPATv4;
2272         }
2273
2274         if (addr.s6_addr32[3]) {
2275                 add_addr(idev, &addr, 128, scope);
2276                 return;
2277         }
2278
2279         for_each_netdev(net, dev) {
2280                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2281                 if (in_dev && (dev->flags & IFF_UP)) {
2282                         struct in_ifaddr * ifa;
2283
2284                         int flag = scope;
2285
2286                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2287                                 int plen;
2288
2289                                 addr.s6_addr32[3] = ifa->ifa_local;
2290
2291                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2292                                         continue;
2293                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2294                                         if (idev->dev->flags&IFF_POINTOPOINT)
2295                                                 continue;
2296                                         flag |= IFA_HOST;
2297                                 }
2298                                 if (idev->dev->flags&IFF_POINTOPOINT)
2299                                         plen = 64;
2300                                 else
2301                                         plen = 96;
2302
2303                                 add_addr(idev, &addr, plen, flag);
2304                         }
2305                 }
2306         }
2307 }
2308 #endif
2309
2310 static void init_loopback(struct net_device *dev)
2311 {
2312         struct inet6_dev  *idev;
2313
2314         /* ::1 */
2315
2316         ASSERT_RTNL();
2317
2318         if ((idev = ipv6_find_idev(dev)) == NULL) {
2319                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2320                 return;
2321         }
2322
2323         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2324 }
2325
2326 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2327 {
2328         struct inet6_ifaddr * ifp;
2329         u32 addr_flags = IFA_F_PERMANENT;
2330
2331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2332         if (idev->cnf.optimistic_dad &&
2333             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2334                 addr_flags |= IFA_F_OPTIMISTIC;
2335 #endif
2336
2337
2338         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2339         if (!IS_ERR(ifp)) {
2340                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2341                 addrconf_dad_start(ifp, 0);
2342                 in6_ifa_put(ifp);
2343         }
2344 }
2345
2346 static void addrconf_dev_config(struct net_device *dev)
2347 {
2348         struct in6_addr addr;
2349         struct inet6_dev    * idev;
2350
2351         ASSERT_RTNL();
2352
2353         if ((dev->type != ARPHRD_ETHER) &&
2354             (dev->type != ARPHRD_FDDI) &&
2355             (dev->type != ARPHRD_IEEE802_TR) &&
2356             (dev->type != ARPHRD_ARCNET) &&
2357             (dev->type != ARPHRD_INFINIBAND)) {
2358                 /* Alas, we support only Ethernet autoconfiguration. */
2359                 return;
2360         }
2361
2362         idev = addrconf_add_dev(dev);
2363         if (idev == NULL)
2364                 return;
2365
2366         memset(&addr, 0, sizeof(struct in6_addr));
2367         addr.s6_addr32[0] = htonl(0xFE800000);
2368
2369         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2370                 addrconf_add_linklocal(idev, &addr);
2371 }
2372
2373 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2374 static void addrconf_sit_config(struct net_device *dev)
2375 {
2376         struct inet6_dev *idev;
2377
2378         ASSERT_RTNL();
2379
2380         /*
2381          * Configure the tunnel with one of our IPv4
2382          * addresses... we should configure all of
2383          * our v4 addrs in the tunnel
2384          */
2385
2386         if ((idev = ipv6_find_idev(dev)) == NULL) {
2387                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2388                 return;
2389         }
2390
2391         if (dev->priv_flags & IFF_ISATAP) {
2392                 struct in6_addr addr;
2393
2394                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2395                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2396                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2397                         addrconf_add_linklocal(idev, &addr);
2398                 return;
2399         }
2400
2401         sit_add_v4_addrs(idev);
2402
2403         if (dev->flags&IFF_POINTOPOINT) {
2404                 addrconf_add_mroute(dev);
2405                 addrconf_add_lroute(dev);
2406         } else
2407                 sit_route_add(dev);
2408 }
2409 #endif
2410
2411 static inline int
2412 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2413 {
2414         struct in6_addr lladdr;
2415
2416         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2417                 addrconf_add_linklocal(idev, &lladdr);
2418                 return 0;
2419         }
2420         return -1;
2421 }
2422
2423 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2424 {
2425         struct net_device *link_dev;
2426         struct net *net = dev_net(idev->dev);
2427
2428         /* first try to inherit the link-local address from the link device */
2429         if (idev->dev->iflink &&
2430             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2431                 if (!ipv6_inherit_linklocal(idev, link_dev))
2432                         return;
2433         }
2434         /* then try to inherit it from any device */
2435         for_each_netdev(net, link_dev) {
2436                 if (!ipv6_inherit_linklocal(idev, link_dev))
2437                         return;
2438         }
2439         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2440 }
2441
2442 /*
2443  * Autoconfigure tunnel with a link-local address so routing protocols,
2444  * DHCPv6, MLD etc. can be run over the virtual link
2445  */
2446
2447 static void addrconf_ip6_tnl_config(struct net_device *dev)
2448 {
2449         struct inet6_dev *idev;
2450
2451         ASSERT_RTNL();
2452
2453         idev = addrconf_add_dev(dev);
2454         if (!idev) {
2455                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2456                 return;
2457         }
2458         ip6_tnl_add_linklocal(idev);
2459 }
2460
2461 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2462                            void * data)
2463 {
2464         struct net_device *dev = (struct net_device *) data;
2465         struct inet6_dev *idev = __in6_dev_get(dev);
2466         int run_pending = 0;
2467         int err;
2468
2469         switch (event) {
2470         case NETDEV_REGISTER:
2471                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2472                         idev = ipv6_add_dev(dev);
2473                         if (!idev)
2474                                 return notifier_from_errno(-ENOMEM);
2475                 }
2476                 break;
2477
2478         case NETDEV_UP:
2479         case NETDEV_CHANGE:
2480                 if (dev->flags & IFF_SLAVE)
2481                         break;
2482
2483                 if (event == NETDEV_UP) {
2484                         if (!addrconf_qdisc_ok(dev)) {
2485                                 /* device is not ready yet. */
2486                                 printk(KERN_INFO
2487                                         "ADDRCONF(NETDEV_UP): %s: "
2488                                         "link is not ready\n",
2489                                         dev->name);
2490                                 break;
2491                         }
2492
2493                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2494                                 idev = ipv6_add_dev(dev);
2495
2496                         if (idev) {
2497                                 idev->if_flags |= IF_READY;
2498                                 run_pending = 1;
2499                         }
2500                 } else {
2501                         if (!addrconf_qdisc_ok(dev)) {
2502                                 /* device is still not ready. */
2503                                 break;
2504                         }
2505
2506                         if (idev) {
2507                                 if (idev->if_flags & IF_READY)
2508                                         /* device is already configured. */
2509                                         break;
2510                                 idev->if_flags |= IF_READY;
2511                         }
2512
2513                         printk(KERN_INFO
2514                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2515                                         "link becomes ready\n",
2516                                         dev->name);
2517
2518                         run_pending = 1;
2519                 }
2520
2521                 switch (dev->type) {
2522 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2523                 case ARPHRD_SIT:
2524                         addrconf_sit_config(dev);
2525                         break;
2526 #endif
2527                 case ARPHRD_TUNNEL6:
2528                         addrconf_ip6_tnl_config(dev);
2529                         break;
2530                 case ARPHRD_LOOPBACK:
2531                         init_loopback(dev);
2532                         break;
2533
2534                 default:
2535                         addrconf_dev_config(dev);
2536                         break;
2537                 }
2538
2539                 if (idev) {
2540                         if (run_pending)
2541                                 addrconf_dad_run(idev);
2542
2543                         /*
2544                          * If the MTU changed during the interface down,
2545                          * when the interface up, the changed MTU must be
2546                          * reflected in the idev as well as routers.
2547                          */
2548                         if (idev->cnf.mtu6 != dev->mtu &&
2549                             dev->mtu >= IPV6_MIN_MTU) {
2550                                 rt6_mtu_change(dev, dev->mtu);
2551                                 idev->cnf.mtu6 = dev->mtu;
2552                         }
2553                         idev->tstamp = jiffies;
2554                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2555
2556                         /*
2557                          * If the changed mtu during down is lower than
2558                          * IPV6_MIN_MTU stop IPv6 on this interface.
2559                          */
2560                         if (dev->mtu < IPV6_MIN_MTU)
2561                                 addrconf_ifdown(dev, 1);
2562                 }
2563                 break;
2564
2565         case NETDEV_CHANGEMTU:
2566                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2567                         rt6_mtu_change(dev, dev->mtu);
2568                         idev->cnf.mtu6 = dev->mtu;
2569                         break;
2570                 }
2571
2572                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2573                         idev = ipv6_add_dev(dev);
2574                         if (idev)
2575                                 break;
2576                 }
2577
2578                 /*
2579                  * MTU falled under IPV6_MIN_MTU.
2580                  * Stop IPv6 on this interface.
2581                  */
2582
2583         case NETDEV_DOWN:
2584         case NETDEV_UNREGISTER:
2585                 /*
2586                  *      Remove all addresses from this interface.
2587                  */
2588                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2589                 break;
2590
2591         case NETDEV_CHANGENAME:
2592                 if (idev) {
2593                         snmp6_unregister_dev(idev);
2594                         addrconf_sysctl_unregister(idev);
2595                         addrconf_sysctl_register(idev);
2596                         err = snmp6_register_dev(idev);
2597                         if (err)
2598                                 return notifier_from_errno(err);
2599                 }
2600                 break;
2601
2602         case NETDEV_PRE_TYPE_CHANGE:
2603         case NETDEV_POST_TYPE_CHANGE:
2604                 addrconf_type_change(dev, event);
2605                 break;
2606         }
2607
2608         return NOTIFY_OK;
2609 }
2610
2611 /*
2612  *      addrconf module should be notified of a device going up
2613  */
2614 static struct notifier_block ipv6_dev_notf = {
2615         .notifier_call = addrconf_notify,
2616 };
2617
2618 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2619 {
2620         struct inet6_dev *idev;
2621         ASSERT_RTNL();
2622
2623         idev = __in6_dev_get(dev);
2624
2625         if (event == NETDEV_POST_TYPE_CHANGE)
2626                 ipv6_mc_remap(idev);
2627         else if (event == NETDEV_PRE_TYPE_CHANGE)
2628                 ipv6_mc_unmap(idev);
2629 }
2630
2631 static int addrconf_ifdown(struct net_device *dev, int how)
2632 {
2633         struct net *net = dev_net(dev);
2634         struct inet6_dev *idev;
2635         struct inet6_ifaddr *ifa;
2636         LIST_HEAD(keep_list);
2637         int state;
2638
2639         ASSERT_RTNL();
2640
2641         rt6_ifdown(net, dev);
2642         neigh_ifdown(&nd_tbl, dev);
2643
2644         idev = __in6_dev_get(dev);
2645         if (idev == NULL)
2646                 return -ENODEV;
2647
2648         /*
2649          * Step 1: remove reference to ipv6 device from parent device.
2650          *         Do not dev_put!
2651          */
2652         if (how) {
2653                 idev->dead = 1;
2654
2655                 /* protected by rtnl_lock */
2656                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2657
2658                 /* Step 1.5: remove snmp6 entry */
2659                 snmp6_unregister_dev(idev);
2660
2661         }
2662
2663         write_lock_bh(&idev->lock);
2664
2665         /* Step 2: clear flags for stateless addrconf */
2666         if (!how)
2667                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2668
2669 #ifdef CONFIG_IPV6_PRIVACY
2670         if (how && del_timer(&idev->regen_timer))
2671                 in6_dev_put(idev);
2672
2673         /* Step 3: clear tempaddr list */
2674         while (!list_empty(&idev->tempaddr_list)) {
2675                 ifa = list_first_entry(&idev->tempaddr_list,
2676                                        struct inet6_ifaddr, tmp_list);
2677                 list_del(&ifa->tmp_list);
2678                 write_unlock_bh(&idev->lock);
2679                 spin_lock_bh(&ifa->lock);
2680
2681                 if (ifa->ifpub) {
2682                         in6_ifa_put(ifa->ifpub);
2683                         ifa->ifpub = NULL;
2684                 }
2685                 spin_unlock_bh(&ifa->lock);
2686                 in6_ifa_put(ifa);
2687                 write_lock_bh(&idev->lock);
2688         }
2689 #endif
2690
2691         while (!list_empty(&idev->addr_list)) {
2692                 ifa = list_first_entry(&idev->addr_list,
2693                                        struct inet6_ifaddr, if_list);
2694                 addrconf_del_timer(ifa);
2695
2696                 /* If just doing link down, and address is permanent
2697                    and not link-local, then retain it. */
2698                 if (!how &&
2699                     (ifa->flags&IFA_F_PERMANENT) &&
2700                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2701                         list_move_tail(&ifa->if_list, &keep_list);
2702
2703                         /* If not doing DAD on this address, just keep it. */
2704                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2705                             idev->cnf.accept_dad <= 0 ||
2706                             (ifa->flags & IFA_F_NODAD))
2707                                 continue;
2708
2709                         /* If it was tentative already, no need to notify */
2710                         if (ifa->flags & IFA_F_TENTATIVE)
2711                                 continue;
2712
2713                         /* Flag it for later restoration when link comes up */
2714                         ifa->flags |= IFA_F_TENTATIVE;
2715
2716                         write_unlock_bh(&idev->lock);
2717
2718                         in6_ifa_hold(ifa);
2719                 } else {
2720                         list_del(&ifa->if_list);
2721
2722                         /* clear hash table */
2723                         spin_lock_bh(&addrconf_hash_lock);
2724                         hlist_del_init_rcu(&ifa->addr_lst);
2725                         spin_unlock_bh(&addrconf_hash_lock);
2726
2727                         write_unlock_bh(&idev->lock);
2728                         spin_lock_bh(&ifa->state_lock);
2729                         state = ifa->state;
2730                         ifa->state = INET6_IFADDR_STATE_DEAD;
2731                         spin_unlock_bh(&ifa->state_lock);
2732
2733                         if (state == INET6_IFADDR_STATE_DEAD)
2734                                 goto put_ifa;
2735                 }
2736
2737                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2738                 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2739                         atomic_notifier_call_chain(&inet6addr_chain,
2740                                                    NETDEV_DOWN, ifa);
2741
2742 put_ifa:
2743                 in6_ifa_put(ifa);
2744
2745                 write_lock_bh(&idev->lock);
2746         }
2747
2748         list_splice(&keep_list, &idev->addr_list);
2749
2750         write_unlock_bh(&idev->lock);
2751
2752         /* Step 5: Discard multicast list */
2753         if (how)
2754                 ipv6_mc_destroy_dev(idev);
2755         else
2756                 ipv6_mc_down(idev);
2757
2758         idev->tstamp = jiffies;
2759
2760         /* Last: Shot the device (if unregistered) */
2761         if (how) {
2762                 addrconf_sysctl_unregister(idev);
2763                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2764                 neigh_ifdown(&nd_tbl, dev);
2765                 in6_dev_put(idev);
2766         }
2767         return 0;
2768 }
2769
2770 static void addrconf_rs_timer(unsigned long data)
2771 {
2772         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2773         struct inet6_dev *idev = ifp->idev;
2774
2775         read_lock(&idev->lock);
2776         if (idev->dead || !(idev->if_flags & IF_READY))
2777                 goto out;
2778
2779         if (idev->cnf.forwarding)
2780                 goto out;
2781
2782         /* Announcement received after solicitation was sent */
2783         if (idev->if_flags & IF_RA_RCVD)
2784                 goto out;
2785
2786         spin_lock(&ifp->lock);
2787         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2788                 /* The wait after the last probe can be shorter */
2789                 addrconf_mod_timer(ifp, AC_RS,
2790                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2791                                    idev->cnf.rtr_solicit_delay :
2792                                    idev->cnf.rtr_solicit_interval);
2793                 spin_unlock(&ifp->lock);
2794
2795                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2796         } else {
2797                 spin_unlock(&ifp->lock);
2798                 /*
2799                  * Note: we do not support deprecated "all on-link"
2800                  * assumption any longer.
2801                  */
2802                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2803                        idev->dev->name);
2804         }
2805
2806 out:
2807         read_unlock(&idev->lock);
2808         in6_ifa_put(ifp);
2809 }
2810
2811 /*
2812  *      Duplicate Address Detection
2813  */
2814 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2815 {
2816         unsigned long rand_num;
2817         struct inet6_dev *idev = ifp->idev;
2818
2819         if (ifp->flags & IFA_F_OPTIMISTIC)
2820                 rand_num = 0;
2821         else
2822                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2823
2824         ifp->probes = idev->cnf.dad_transmits;
2825         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2826 }
2827
2828 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2829 {
2830         struct inet6_dev *idev = ifp->idev;
2831         struct net_device *dev = idev->dev;
2832
2833         addrconf_join_solict(dev, &ifp->addr);
2834
2835         net_srandom(ifp->addr.s6_addr32[3]);
2836
2837         read_lock_bh(&idev->lock);
2838         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2839                 goto out;
2840
2841         spin_lock(&ifp->lock);
2842         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2843             idev->cnf.accept_dad < 1 ||
2844             !(ifp->flags&IFA_F_TENTATIVE) ||
2845             ifp->flags & IFA_F_NODAD) {
2846                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2847                 spin_unlock(&ifp->lock);
2848                 read_unlock_bh(&idev->lock);
2849
2850                 addrconf_dad_completed(ifp);
2851                 return;
2852         }
2853
2854         if (!(idev->if_flags & IF_READY)) {
2855                 spin_unlock(&ifp->lock);
2856                 read_unlock_bh(&idev->lock);
2857                 /*
2858                  * If the device is not ready:
2859                  * - keep it tentative if it is a permanent address.
2860                  * - otherwise, kill it.
2861                  */
2862                 in6_ifa_hold(ifp);
2863                 addrconf_dad_stop(ifp, 0);
2864                 return;
2865         }
2866
2867         /*
2868          * Optimistic nodes can start receiving
2869          * Frames right away
2870          */
2871         if (ifp->flags & IFA_F_OPTIMISTIC)
2872                 ip6_ins_rt(ifp->rt);
2873
2874         addrconf_dad_kick(ifp);
2875         spin_unlock(&ifp->lock);
2876 out:
2877         read_unlock_bh(&idev->lock);
2878 }
2879
2880 static void addrconf_dad_timer(unsigned long data)
2881 {
2882         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2883         struct inet6_dev *idev = ifp->idev;
2884         struct in6_addr mcaddr;
2885
2886         read_lock(&idev->lock);
2887         if (idev->dead || !(idev->if_flags & IF_READY)) {
2888                 read_unlock(&idev->lock);
2889                 goto out;
2890         }
2891
2892         spin_lock(&ifp->lock);
2893         if (ifp->probes == 0) {
2894                 /*
2895                  * DAD was successful
2896                  */
2897
2898                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2899                 spin_unlock(&ifp->lock);
2900                 read_unlock(&idev->lock);
2901
2902                 addrconf_dad_completed(ifp);
2903
2904                 goto out;
2905         }
2906
2907         ifp->probes--;
2908         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2909         spin_unlock(&ifp->lock);
2910         read_unlock(&idev->lock);
2911
2912         /* send a neighbour solicitation for our addr */
2913         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2914         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2915 out:
2916         in6_ifa_put(ifp);
2917 }
2918
2919 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2920 {
2921         struct net_device *dev = ifp->idev->dev;
2922
2923         /*
2924          *      Configure the address for reception. Now it is valid.
2925          */
2926
2927         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2928
2929         /* If added prefix is link local and forwarding is off,
2930            start sending router solicitations.
2931          */
2932
2933         if (ifp->idev->cnf.forwarding == 0 &&
2934             ifp->idev->cnf.rtr_solicits > 0 &&
2935             (dev->flags&IFF_LOOPBACK) == 0 &&
2936             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2937                 /*
2938                  *      If a host as already performed a random delay
2939                  *      [...] as part of DAD [...] there is no need
2940                  *      to delay again before sending the first RS
2941                  */
2942                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2943
2944                 spin_lock_bh(&ifp->lock);
2945                 ifp->probes = 1;
2946                 ifp->idev->if_flags |= IF_RS_SENT;
2947                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2948                 spin_unlock_bh(&ifp->lock);
2949         }
2950 }
2951
2952 static void addrconf_dad_run(struct inet6_dev *idev)
2953 {
2954         struct inet6_ifaddr *ifp;
2955
2956         read_lock_bh(&idev->lock);
2957         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2958                 spin_lock(&ifp->lock);
2959                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2960                         spin_unlock(&ifp->lock);
2961                         continue;
2962                 }
2963                 spin_unlock(&ifp->lock);
2964                 addrconf_dad_kick(ifp);
2965         }
2966         read_unlock_bh(&idev->lock);
2967 }
2968
2969 #ifdef CONFIG_PROC_FS
2970 struct if6_iter_state {
2971         struct seq_net_private p;
2972         int bucket;
2973 };
2974
2975 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2976 {
2977         struct inet6_ifaddr *ifa = NULL;
2978         struct if6_iter_state *state = seq->private;
2979         struct net *net = seq_file_net(seq);
2980
2981         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2982                 struct hlist_node *n;
2983                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2984                                          addr_lst)
2985                         if (net_eq(dev_net(ifa->idev->dev), net))
2986                                 return ifa;
2987         }
2988         return NULL;
2989 }
2990
2991 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2992                                          struct inet6_ifaddr *ifa)
2993 {
2994         struct if6_iter_state *state = seq->private;
2995         struct net *net = seq_file_net(seq);
2996         struct hlist_node *n = &ifa->addr_lst;
2997
2998         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
2999                 if (net_eq(dev_net(ifa->idev->dev), net))
3000                         return ifa;
3001
3002         while (++state->bucket < IN6_ADDR_HSIZE) {
3003                 hlist_for_each_entry_rcu_bh(ifa, n,
3004                                      &inet6_addr_lst[state->bucket], addr_lst) {
3005                         if (net_eq(dev_net(ifa->idev->dev), net))
3006                                 return ifa;
3007                 }
3008         }
3009
3010         return NULL;
3011 }
3012
3013 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3014 {
3015         struct inet6_ifaddr *ifa = if6_get_first(seq);
3016
3017         if (ifa)
3018                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3019                         --pos;
3020         return pos ? NULL : ifa;
3021 }
3022
3023 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3024         __acquires(rcu_bh)
3025 {
3026         rcu_read_lock_bh();
3027         return if6_get_idx(seq, *pos);
3028 }
3029
3030 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3031 {
3032         struct inet6_ifaddr *ifa;
3033
3034         ifa = if6_get_next(seq, v);
3035         ++*pos;
3036         return ifa;
3037 }
3038
3039 static void if6_seq_stop(struct seq_file *seq, void *v)
3040         __releases(rcu_bh)
3041 {
3042         rcu_read_unlock_bh();
3043 }
3044
3045 static int if6_seq_show(struct seq_file *seq, void *v)
3046 {
3047         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3048         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3049                    &ifp->addr,
3050                    ifp->idev->dev->ifindex,
3051                    ifp->prefix_len,
3052                    ifp->scope,
3053                    ifp->flags,
3054                    ifp->idev->dev->name);
3055         return 0;
3056 }
3057
3058 static const struct seq_operations if6_seq_ops = {
3059         .start  = if6_seq_start,
3060         .next   = if6_seq_next,
3061         .show   = if6_seq_show,
3062         .stop   = if6_seq_stop,
3063 };
3064
3065 static int if6_seq_open(struct inode *inode, struct file *file)
3066 {
3067         return seq_open_net(inode, file, &if6_seq_ops,
3068                             sizeof(struct if6_iter_state));
3069 }
3070
3071 static const struct file_operations if6_fops = {
3072         .owner          = THIS_MODULE,
3073         .open           = if6_seq_open,
3074         .read           = seq_read,
3075         .llseek         = seq_lseek,
3076         .release        = seq_release_net,
3077 };
3078
3079 static int __net_init if6_proc_net_init(struct net *net)
3080 {
3081         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3082                 return -ENOMEM;
3083         return 0;
3084 }
3085
3086 static void __net_exit if6_proc_net_exit(struct net *net)
3087 {
3088        proc_net_remove(net, "if_inet6");
3089 }
3090
3091 static struct pernet_operations if6_proc_net_ops = {
3092        .init = if6_proc_net_init,
3093        .exit = if6_proc_net_exit,
3094 };
3095
3096 int __init if6_proc_init(void)
3097 {
3098         return register_pernet_subsys(&if6_proc_net_ops);
3099 }
3100
3101 void if6_proc_exit(void)
3102 {
3103         unregister_pernet_subsys(&if6_proc_net_ops);
3104 }
3105 #endif  /* CONFIG_PROC_FS */
3106
3107 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3108 /* Check if address is a home address configured on any interface. */
3109 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3110 {
3111         int ret = 0;
3112         struct inet6_ifaddr *ifp = NULL;
3113         struct hlist_node *n;
3114         unsigned int hash = ipv6_addr_hash(addr);
3115
3116         rcu_read_lock_bh();
3117         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3118                 if (!net_eq(dev_net(ifp->idev->dev), net))
3119                         continue;
3120                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3121                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3122                         ret = 1;
3123                         break;
3124                 }
3125         }
3126         rcu_read_unlock_bh();
3127         return ret;
3128 }
3129 #endif
3130
3131 /*
3132  *      Periodic address status verification
3133  */
3134
3135 static void addrconf_verify(unsigned long foo)
3136 {
3137         unsigned long now, next, next_sec, next_sched;
3138         struct inet6_ifaddr *ifp;
3139         struct hlist_node *node;
3140         int i;
3141
3142         rcu_read_lock_bh();
3143         spin_lock(&addrconf_verify_lock);
3144         now = jiffies;
3145         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3146
3147         del_timer(&addr_chk_timer);
3148
3149         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3150 restart:
3151                 hlist_for_each_entry_rcu_bh(ifp, node,
3152                                          &inet6_addr_lst[i], addr_lst) {
3153                         unsigned long age;
3154
3155                         if (ifp->flags & IFA_F_PERMANENT)
3156                                 continue;
3157
3158                         spin_lock(&ifp->lock);
3159                         /* We try to batch several events at once. */
3160                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3161
3162                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3163                             age >= ifp->valid_lft) {
3164                                 spin_unlock(&ifp->lock);
3165                                 in6_ifa_hold(ifp);
3166                                 ipv6_del_addr(ifp);
3167                                 goto restart;
3168                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3169                                 spin_unlock(&ifp->lock);
3170                                 continue;
3171                         } else if (age >= ifp->prefered_lft) {
3172                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3173                                 int deprecate = 0;
3174
3175                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3176                                         deprecate = 1;
3177                                         ifp->flags |= IFA_F_DEPRECATED;
3178                                 }
3179
3180                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3181                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3182
3183                                 spin_unlock(&ifp->lock);
3184
3185                                 if (deprecate) {
3186                                         in6_ifa_hold(ifp);
3187
3188                                         ipv6_ifa_notify(0, ifp);
3189                                         in6_ifa_put(ifp);
3190                                         goto restart;
3191                                 }
3192 #ifdef CONFIG_IPV6_PRIVACY
3193                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3194                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3195                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3196                                         ifp->idev->cnf.dad_transmits *
3197                                         ifp->idev->nd_parms->retrans_time / HZ;
3198
3199                                 if (age >= ifp->prefered_lft - regen_advance) {
3200                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3201                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3202                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3203                                         if (!ifp->regen_count && ifpub) {
3204                                                 ifp->regen_count++;
3205                                                 in6_ifa_hold(ifp);
3206                                                 in6_ifa_hold(ifpub);
3207                                                 spin_unlock(&ifp->lock);
3208
3209                                                 spin_lock(&ifpub->lock);
3210                                                 ifpub->regen_count = 0;
3211                                                 spin_unlock(&ifpub->lock);
3212                                                 ipv6_create_tempaddr(ifpub, ifp);
3213                                                 in6_ifa_put(ifpub);
3214                                                 in6_ifa_put(ifp);
3215                                                 goto restart;
3216                                         }
3217                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3218                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3219                                 spin_unlock(&ifp->lock);
3220 #endif
3221                         } else {
3222                                 /* ifp->prefered_lft <= ifp->valid_lft */
3223                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3224                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3225                                 spin_unlock(&ifp->lock);
3226                         }
3227                 }
3228         }
3229
3230         next_sec = round_jiffies_up(next);
3231         next_sched = next;
3232
3233         /* If rounded timeout is accurate enough, accept it. */
3234         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3235                 next_sched = next_sec;
3236
3237         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3238         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3239                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3240
3241         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3242               now, next, next_sec, next_sched));
3243
3244         addr_chk_timer.expires = next_sched;
3245         add_timer(&addr_chk_timer);
3246         spin_unlock(&addrconf_verify_lock);
3247         rcu_read_unlock_bh();
3248 }
3249
3250 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3251 {
3252         struct in6_addr *pfx = NULL;
3253
3254         if (addr)
3255                 pfx = nla_data(addr);
3256
3257         if (local) {
3258                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3259                         pfx = NULL;
3260                 else
3261                         pfx = nla_data(local);
3262         }
3263
3264         return pfx;
3265 }
3266
3267 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3268         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3269         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3270         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3271 };
3272
3273 static int
3274 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3275 {
3276         struct net *net = sock_net(skb->sk);
3277         struct ifaddrmsg *ifm;
3278         struct nlattr *tb[IFA_MAX+1];
3279         struct in6_addr *pfx;
3280         int err;
3281
3282         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3283         if (err < 0)
3284                 return err;
3285
3286         ifm = nlmsg_data(nlh);
3287         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3288         if (pfx == NULL)
3289                 return -EINVAL;
3290
3291         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3292 }
3293
3294 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3295                              u32 prefered_lft, u32 valid_lft)
3296 {
3297         u32 flags;
3298         clock_t expires;
3299         unsigned long timeout;
3300
3301         if (!valid_lft || (prefered_lft > valid_lft))
3302                 return -EINVAL;
3303
3304         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3305         if (addrconf_finite_timeout(timeout)) {
3306                 expires = jiffies_to_clock_t(timeout * HZ);
3307                 valid_lft = timeout;
3308                 flags = RTF_EXPIRES;
3309         } else {
3310                 expires = 0;
3311                 flags = 0;
3312                 ifa_flags |= IFA_F_PERMANENT;
3313         }
3314
3315         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3316         if (addrconf_finite_timeout(timeout)) {
3317                 if (timeout == 0)
3318                         ifa_flags |= IFA_F_DEPRECATED;
3319                 prefered_lft = timeout;
3320         }
3321
3322         spin_lock_bh(&ifp->lock);
3323         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3324         ifp->tstamp = jiffies;
3325         ifp->valid_lft = valid_lft;
3326         ifp->prefered_lft = prefered_lft;
3327
3328         spin_unlock_bh(&ifp->lock);
3329         if (!(ifp->flags&IFA_F_TENTATIVE))
3330                 ipv6_ifa_notify(0, ifp);
3331
3332         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3333                               expires, flags);
3334         addrconf_verify(0);
3335
3336         return 0;
3337 }
3338
3339 static int
3340 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3341 {
3342         struct net *net = sock_net(skb->sk);
3343         struct ifaddrmsg *ifm;
3344         struct nlattr *tb[IFA_MAX+1];
3345         struct in6_addr *pfx;
3346         struct inet6_ifaddr *ifa;
3347         struct net_device *dev;
3348         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3349         u8 ifa_flags;
3350         int err;
3351
3352         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3353         if (err < 0)
3354                 return err;
3355
3356         ifm = nlmsg_data(nlh);
3357         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3358         if (pfx == NULL)
3359                 return -EINVAL;
3360
3361         if (tb[IFA_CACHEINFO]) {
3362                 struct ifa_cacheinfo *ci;
3363
3364                 ci = nla_data(tb[IFA_CACHEINFO]);
3365                 valid_lft = ci->ifa_valid;
3366                 preferred_lft = ci->ifa_prefered;
3367         } else {
3368                 preferred_lft = INFINITY_LIFE_TIME;
3369                 valid_lft = INFINITY_LIFE_TIME;
3370         }
3371
3372         dev =  __dev_get_by_index(net, ifm->ifa_index);
3373         if (dev == NULL)
3374                 return -ENODEV;
3375
3376         /* We ignore other flags so far. */
3377         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3378
3379         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3380         if (ifa == NULL) {
3381                 /*
3382                  * It would be best to check for !NLM_F_CREATE here but
3383                  * userspace alreay relies on not having to provide this.
3384                  */
3385                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3386                                       ifm->ifa_prefixlen, ifa_flags,
3387                                       preferred_lft, valid_lft);
3388         }
3389
3390         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3391             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3392                 err = -EEXIST;
3393         else
3394                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3395
3396         in6_ifa_put(ifa);
3397
3398         return err;
3399 }
3400
3401 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3402                           u8 scope, int ifindex)
3403 {
3404         struct ifaddrmsg *ifm;
3405
3406         ifm = nlmsg_data(nlh);
3407         ifm->ifa_family = AF_INET6;
3408         ifm->ifa_prefixlen = prefixlen;
3409         ifm->ifa_flags = flags;
3410         ifm->ifa_scope = scope;
3411         ifm->ifa_index = ifindex;
3412 }
3413
3414 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3415                          unsigned long tstamp, u32 preferred, u32 valid)
3416 {
3417         struct ifa_cacheinfo ci;
3418
3419         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3420                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3421         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3422                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3423         ci.ifa_prefered = preferred;
3424         ci.ifa_valid = valid;
3425
3426         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3427 }
3428
3429 static inline int rt_scope(int ifa_scope)
3430 {
3431         if (ifa_scope & IFA_HOST)
3432                 return RT_SCOPE_HOST;
3433         else if (ifa_scope & IFA_LINK)
3434                 return RT_SCOPE_LINK;
3435         else if (ifa_scope & IFA_SITE)
3436                 return RT_SCOPE_SITE;
3437         else
3438                 return RT_SCOPE_UNIVERSE;
3439 }
3440
3441 static inline int inet6_ifaddr_msgsize(void)
3442 {
3443         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3444                + nla_total_size(16) /* IFA_ADDRESS */
3445                + nla_total_size(sizeof(struct ifa_cacheinfo));
3446 }
3447
3448 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3449                              u32 pid, u32 seq, int event, unsigned int flags)
3450 {
3451         struct nlmsghdr  *nlh;
3452         u32 preferred, valid;
3453
3454         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3455         if (nlh == NULL)
3456                 return -EMSGSIZE;
3457
3458         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3459                       ifa->idev->dev->ifindex);
3460
3461         if (!(ifa->flags&IFA_F_PERMANENT)) {
3462                 preferred = ifa->prefered_lft;
3463                 valid = ifa->valid_lft;
3464                 if (preferred != INFINITY_LIFE_TIME) {
3465                         long tval = (jiffies - ifa->tstamp)/HZ;
3466                         if (preferred > tval)
3467                                 preferred -= tval;
3468                         else
3469                                 preferred = 0;
3470                         if (valid != INFINITY_LIFE_TIME)
3471                                 valid -= tval;
3472                 }
3473         } else {
3474                 preferred = INFINITY_LIFE_TIME;
3475                 valid = INFINITY_LIFE_TIME;
3476         }
3477
3478         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3479             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3480                 nlmsg_cancel(skb, nlh);
3481                 return -EMSGSIZE;
3482         }
3483
3484         return nlmsg_end(skb, nlh);
3485 }
3486
3487 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3488                                 u32 pid, u32 seq, int event, u16 flags)
3489 {
3490         struct nlmsghdr  *nlh;
3491         u8 scope = RT_SCOPE_UNIVERSE;
3492         int ifindex = ifmca->idev->dev->ifindex;
3493
3494         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3495                 scope = RT_SCOPE_SITE;
3496
3497         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3498         if (nlh == NULL)
3499                 return -EMSGSIZE;
3500
3501         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3502         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3503             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3504                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3505                 nlmsg_cancel(skb, nlh);
3506                 return -EMSGSIZE;
3507         }
3508
3509         return nlmsg_end(skb, nlh);
3510 }
3511
3512 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3513                                 u32 pid, u32 seq, int event, unsigned int flags)
3514 {
3515         struct nlmsghdr  *nlh;
3516         u8 scope = RT_SCOPE_UNIVERSE;
3517         int ifindex = ifaca->aca_idev->dev->ifindex;
3518
3519         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3520                 scope = RT_SCOPE_SITE;
3521
3522         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3523         if (nlh == NULL)
3524                 return -EMSGSIZE;
3525
3526         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3527         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3528             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3529                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3530                 nlmsg_cancel(skb, nlh);
3531                 return -EMSGSIZE;
3532         }
3533
3534         return nlmsg_end(skb, nlh);
3535 }
3536
3537 enum addr_type_t {
3538         UNICAST_ADDR,
3539         MULTICAST_ADDR,
3540         ANYCAST_ADDR,
3541 };
3542
3543 /* called with rcu_read_lock() */
3544 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3545                           struct netlink_callback *cb, enum addr_type_t type,
3546                           int s_ip_idx, int *p_ip_idx)
3547 {
3548         struct ifmcaddr6 *ifmca;
3549         struct ifacaddr6 *ifaca;
3550         int err = 1;
3551         int ip_idx = *p_ip_idx;
3552
3553         read_lock_bh(&idev->lock);
3554         switch (type) {
3555         case UNICAST_ADDR: {
3556                 struct inet6_ifaddr *ifa;
3557
3558                 /* unicast address incl. temp addr */
3559                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3560                         if (++ip_idx < s_ip_idx)
3561                                 continue;
3562                         err = inet6_fill_ifaddr(skb, ifa,
3563                                                 NETLINK_CB(cb->skb).pid,
3564                                                 cb->nlh->nlmsg_seq,
3565                                                 RTM_NEWADDR,
3566                                                 NLM_F_MULTI);
3567                         if (err <= 0)
3568                                 break;
3569                 }
3570                 break;
3571         }
3572         case MULTICAST_ADDR:
3573                 /* multicast address */
3574                 for (ifmca = idev->mc_list; ifmca;
3575                      ifmca = ifmca->next, ip_idx++) {
3576                         if (ip_idx < s_ip_idx)
3577                                 continue;
3578                         err = inet6_fill_ifmcaddr(skb, ifmca,
3579                                                   NETLINK_CB(cb->skb).pid,
3580                                                   cb->nlh->nlmsg_seq,
3581                                                   RTM_GETMULTICAST,
3582                                                   NLM_F_MULTI);
3583                         if (err <= 0)
3584                                 break;
3585                 }
3586                 break;
3587         case ANYCAST_ADDR:
3588                 /* anycast address */
3589                 for (ifaca = idev->ac_list; ifaca;
3590                      ifaca = ifaca->aca_next, ip_idx++) {
3591                         if (ip_idx < s_ip_idx)
3592                                 continue;
3593                         err = inet6_fill_ifacaddr(skb, ifaca,
3594                                                   NETLINK_CB(cb->skb).pid,
3595                                                   cb->nlh->nlmsg_seq,
3596                                                   RTM_GETANYCAST,
3597                                                   NLM_F_MULTI);
3598                         if (err <= 0)
3599                                 break;
3600                 }
3601                 break;
3602         default:
3603                 break;
3604         }
3605         read_unlock_bh(&idev->lock);
3606         *p_ip_idx = ip_idx;
3607         return err;
3608 }
3609
3610 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3611                            enum addr_type_t type)
3612 {
3613         struct net *net = sock_net(skb->sk);
3614         int h, s_h;
3615         int idx, ip_idx;
3616         int s_idx, s_ip_idx;
3617         struct net_device *dev;
3618         struct inet6_dev *idev;
3619         struct hlist_head *head;
3620         struct hlist_node *node;
3621
3622         s_h = cb->args[0];
3623         s_idx = idx = cb->args[1];
3624         s_ip_idx = ip_idx = cb->args[2];
3625
3626         rcu_read_lock();
3627         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3628                 idx = 0;
3629                 head = &net->dev_index_head[h];
3630                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3631                         if (idx < s_idx)
3632                                 goto cont;
3633                         if (h > s_h || idx > s_idx)
3634                                 s_ip_idx = 0;
3635                         ip_idx = 0;
3636                         idev = __in6_dev_get(dev);
3637                         if (!idev)
3638                                 goto cont;
3639
3640                         if (in6_dump_addrs(idev, skb, cb, type,
3641                                            s_ip_idx, &ip_idx) <= 0)
3642                                 goto done;
3643 cont:
3644                         idx++;
3645                 }
3646         }
3647 done:
3648         rcu_read_unlock();
3649         cb->args[0] = h;
3650         cb->args[1] = idx;
3651         cb->args[2] = ip_idx;
3652
3653         return skb->len;
3654 }
3655
3656 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3657 {
3658         enum addr_type_t type = UNICAST_ADDR;
3659
3660         return inet6_dump_addr(skb, cb, type);
3661 }
3662
3663 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3664 {
3665         enum addr_type_t type = MULTICAST_ADDR;
3666
3667         return inet6_dump_addr(skb, cb, type);
3668 }
3669
3670
3671 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3672 {
3673         enum addr_type_t type = ANYCAST_ADDR;
3674
3675         return inet6_dump_addr(skb, cb, type);
3676 }
3677
3678 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3679                              void *arg)
3680 {
3681         struct net *net = sock_net(in_skb->sk);
3682         struct ifaddrmsg *ifm;
3683         struct nlattr *tb[IFA_MAX+1];
3684         struct in6_addr *addr = NULL;
3685         struct net_device *dev = NULL;
3686         struct inet6_ifaddr *ifa;
3687         struct sk_buff *skb;
3688         int err;
3689
3690         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3691         if (err < 0)
3692                 goto errout;
3693
3694         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3695         if (addr == NULL) {
3696                 err = -EINVAL;
3697                 goto errout;
3698         }
3699
3700         ifm = nlmsg_data(nlh);
3701         if (ifm->ifa_index)
3702                 dev = __dev_get_by_index(net, ifm->ifa_index);
3703
3704         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3705         if (!ifa) {
3706                 err = -EADDRNOTAVAIL;
3707                 goto errout;
3708         }
3709
3710         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3711         if (!skb) {
3712                 err = -ENOBUFS;
3713                 goto errout_ifa;
3714         }
3715
3716         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3717                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3718         if (err < 0) {
3719                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3720                 WARN_ON(err == -EMSGSIZE);
3721                 kfree_skb(skb);
3722                 goto errout_ifa;
3723         }
3724         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3725 errout_ifa:
3726         in6_ifa_put(ifa);
3727 errout:
3728         return err;
3729 }
3730
3731 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3732 {
3733         struct sk_buff *skb;
3734         struct net *net = dev_net(ifa->idev->dev);
3735         int err = -ENOBUFS;
3736
3737         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3738         if (skb == NULL)
3739                 goto errout;
3740
3741         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3742         if (err < 0) {
3743                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3744                 WARN_ON(err == -EMSGSIZE);
3745                 kfree_skb(skb);
3746                 goto errout;
3747         }
3748         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3749         return;
3750 errout:
3751         if (err < 0)
3752                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3753 }
3754
3755 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3756                                 __s32 *array, int bytes)
3757 {
3758         BUG_ON(bytes < (DEVCONF_MAX * 4));
3759
3760         memset(array, 0, bytes);
3761         array[DEVCONF_FORWARDING] = cnf->forwarding;
3762         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3763         array[DEVCONF_MTU6] = cnf->mtu6;
3764         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3765         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3766         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3767         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3768         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3769         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3770         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3771         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3772 #ifdef CONFIG_IPV6_PRIVACY
3773         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3774         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3775         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3776         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3777         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3778 #endif
3779         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3780         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3781         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3782 #ifdef CONFIG_IPV6_ROUTER_PREF
3783         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3784         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3785 #ifdef CONFIG_IPV6_ROUTE_INFO
3786         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3787 #endif
3788 #endif
3789         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3790         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3791 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3792         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3793 #endif
3794 #ifdef CONFIG_IPV6_MROUTE
3795         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3796 #endif
3797         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3798         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3799         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3800 }
3801
3802 static inline size_t inet6_if_nlmsg_size(void)
3803 {
3804         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3805                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3806                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3807                + nla_total_size(4) /* IFLA_MTU */
3808                + nla_total_size(4) /* IFLA_LINK */
3809                + nla_total_size( /* IFLA_PROTINFO */
3810                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3811                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3812                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3813                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3814                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3815                  );
3816 }
3817
3818 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3819                                       int items, int bytes)
3820 {
3821         int i;
3822         int pad = bytes - sizeof(u64) * items;
3823         BUG_ON(pad < 0);
3824
3825         /* Use put_unaligned() because stats may not be aligned for u64. */
3826         put_unaligned(items, &stats[0]);
3827         for (i = 1; i < items; i++)
3828                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3829
3830         memset(&stats[items], 0, pad);
3831 }
3832
3833 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3834                              int bytes)
3835 {
3836         switch (attrtype) {
3837         case IFLA_INET6_STATS:
3838                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3839                 break;
3840         case IFLA_INET6_ICMP6STATS:
3841                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3842                 break;
3843         }
3844 }
3845
3846 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3847                              u32 pid, u32 seq, int event, unsigned int flags)
3848 {
3849         struct net_device *dev = idev->dev;
3850         struct nlattr *nla;
3851         struct ifinfomsg *hdr;
3852         struct nlmsghdr *nlh;
3853         void *protoinfo;
3854         struct ifla_cacheinfo ci;
3855
3856         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3857         if (nlh == NULL)
3858                 return -EMSGSIZE;
3859
3860         hdr = nlmsg_data(nlh);
3861         hdr->ifi_family = AF_INET6;
3862         hdr->__ifi_pad = 0;
3863         hdr->ifi_type = dev->type;
3864         hdr->ifi_index = dev->ifindex;
3865         hdr->ifi_flags = dev_get_flags(dev);
3866         hdr->ifi_change = 0;
3867
3868         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3869
3870         if (dev->addr_len)
3871                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3872
3873         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3874         if (dev->ifindex != dev->iflink)
3875                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3876
3877         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3878         if (protoinfo == NULL)
3879                 goto nla_put_failure;
3880
3881         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3882
3883         ci.max_reasm_len = IPV6_MAXPLEN;
3884         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3885                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3886         ci.reachable_time = idev->nd_parms->reachable_time;
3887         ci.retrans_time = idev->nd_parms->retrans_time;
3888         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3889
3890         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3891         if (nla == NULL)
3892                 goto nla_put_failure;
3893         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3894
3895         /* XXX - MC not implemented */
3896
3897         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3898         if (nla == NULL)
3899                 goto nla_put_failure;
3900         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3901
3902         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3903         if (nla == NULL)
3904                 goto nla_put_failure;
3905         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3906
3907         nla_nest_end(skb, protoinfo);
3908         return nlmsg_end(skb, nlh);
3909
3910 nla_put_failure:
3911         nlmsg_cancel(skb, nlh);
3912         return -EMSGSIZE;
3913 }
3914
3915 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3916 {
3917         struct net *net = sock_net(skb->sk);
3918         int h, s_h;
3919         int idx = 0, s_idx;
3920         struct net_device *dev;
3921         struct inet6_dev *idev;
3922         struct hlist_head *head;
3923         struct hlist_node *node;
3924
3925         s_h = cb->args[0];
3926         s_idx = cb->args[1];
3927
3928         rcu_read_lock();
3929         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3930                 idx = 0;
3931                 head = &net->dev_index_head[h];
3932                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3933                         if (idx < s_idx)
3934                                 goto cont;
3935                         idev = __in6_dev_get(dev);
3936                         if (!idev)
3937                                 goto cont;
3938                         if (inet6_fill_ifinfo(skb, idev,
3939                                               NETLINK_CB(cb->skb).pid,
3940                                               cb->nlh->nlmsg_seq,
3941                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3942                                 goto out;
3943 cont:
3944                         idx++;
3945                 }
3946         }
3947 out:
3948         rcu_read_unlock();
3949         cb->args[1] = idx;
3950         cb->args[0] = h;
3951
3952         return skb->len;
3953 }
3954
3955 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3956 {
3957         struct sk_buff *skb;
3958         struct net *net = dev_net(idev->dev);
3959         int err = -ENOBUFS;
3960
3961         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3962         if (skb == NULL)
3963                 goto errout;
3964
3965         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3966         if (err < 0) {
3967                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3968                 WARN_ON(err == -EMSGSIZE);
3969                 kfree_skb(skb);
3970                 goto errout;
3971         }
3972         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3973         return;
3974 errout:
3975         if (err < 0)
3976                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3977 }
3978
3979 static inline size_t inet6_prefix_nlmsg_size(void)
3980 {
3981         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3982                + nla_total_size(sizeof(struct in6_addr))
3983                + nla_total_size(sizeof(struct prefix_cacheinfo));
3984 }
3985
3986 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3987                              struct prefix_info *pinfo, u32 pid, u32 seq,
3988                              int event, unsigned int flags)
3989 {
3990         struct prefixmsg *pmsg;
3991         struct nlmsghdr *nlh;
3992         struct prefix_cacheinfo ci;
3993
3994         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3995         if (nlh == NULL)
3996                 return -EMSGSIZE;
3997
3998         pmsg = nlmsg_data(nlh);
3999         pmsg->prefix_family = AF_INET6;
4000         pmsg->prefix_pad1 = 0;
4001         pmsg->prefix_pad2 = 0;
4002         pmsg->prefix_ifindex = idev->dev->ifindex;
4003         pmsg->prefix_len = pinfo->prefix_len;
4004         pmsg->prefix_type = pinfo->type;
4005         pmsg->prefix_pad3 = 0;
4006         pmsg->prefix_flags = 0;
4007         if (pinfo->onlink)
4008                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4009         if (pinfo->autoconf)
4010                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4011
4012         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4013
4014         ci.preferred_time = ntohl(pinfo->prefered);
4015         ci.valid_time = ntohl(pinfo->valid);
4016         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4017
4018         return nlmsg_end(skb, nlh);
4019
4020 nla_put_failure:
4021         nlmsg_cancel(skb, nlh);
4022         return -EMSGSIZE;
4023 }
4024
4025 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4026                          struct prefix_info *pinfo)
4027 {
4028         struct sk_buff *skb;
4029         struct net *net = dev_net(idev->dev);
4030         int err = -ENOBUFS;
4031
4032         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4033         if (skb == NULL)
4034                 goto errout;
4035
4036         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4037         if (err < 0) {
4038                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4039                 WARN_ON(err == -EMSGSIZE);
4040                 kfree_skb(skb);
4041                 goto errout;
4042         }
4043         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4044         return;
4045 errout:
4046         if (err < 0)
4047                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4048 }
4049
4050 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4051 {
4052         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4053
4054         switch (event) {
4055         case RTM_NEWADDR:
4056                 /*
4057                  * If the address was optimistic
4058                  * we inserted the route at the start of
4059                  * our DAD process, so we don't need
4060                  * to do it again
4061                  */
4062                 if (!(ifp->rt->rt6i_node))
4063                         ip6_ins_rt(ifp->rt);
4064                 if (ifp->idev->cnf.forwarding)
4065                         addrconf_join_anycast(ifp);
4066                 break;
4067         case RTM_DELADDR:
4068                 if (ifp->idev->cnf.forwarding)
4069                         addrconf_leave_anycast(ifp);
4070                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4071                 dst_hold(&ifp->rt->u.dst);
4072
4073                 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4074                     ip6_del_rt(ifp->rt))
4075                         dst_free(&ifp->rt->u.dst);
4076                 break;
4077         }
4078 }
4079
4080 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4081 {
4082         rcu_read_lock_bh();
4083         if (likely(ifp->idev->dead == 0))
4084                 __ipv6_ifa_notify(event, ifp);
4085         rcu_read_unlock_bh();
4086 }
4087
4088 #ifdef CONFIG_SYSCTL
4089
4090 static
4091 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4092                            void __user *buffer, size_t *lenp, loff_t *ppos)
4093 {
4094         int *valp = ctl->data;
4095         int val = *valp;
4096         loff_t pos = *ppos;
4097         int ret;
4098
4099         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4100
4101         if (write)
4102                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4103         if (ret)
4104                 *ppos = pos;
4105         return ret;
4106 }
4107
4108 static void dev_disable_change(struct inet6_dev *idev)
4109 {
4110         if (!idev || !idev->dev)
4111                 return;
4112
4113         if (idev->cnf.disable_ipv6)
4114                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4115         else
4116                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4117 }
4118
4119 static void addrconf_disable_change(struct net *net, __s32 newf)
4120 {
4121         struct net_device *dev;
4122         struct inet6_dev *idev;
4123
4124         rcu_read_lock();
4125         for_each_netdev_rcu(net, dev) {
4126                 idev = __in6_dev_get(dev);
4127                 if (idev) {
4128                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4129                         idev->cnf.disable_ipv6 = newf;
4130                         if (changed)
4131                                 dev_disable_change(idev);
4132                 }
4133         }
4134         rcu_read_unlock();
4135 }
4136
4137 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4138 {
4139         struct net *net;
4140
4141         net = (struct net *)table->extra2;
4142
4143         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4144                 return 0;
4145
4146         if (!rtnl_trylock()) {
4147                 /* Restore the original values before restarting */
4148                 *p = old;
4149                 return restart_syscall();
4150         }
4151
4152         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4153                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4154                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4155                 addrconf_disable_change(net, newf);
4156         } else if ((!*p) ^ (!old))
4157                 dev_disable_change((struct inet6_dev *)table->extra1);
4158
4159         rtnl_unlock();
4160         return 0;
4161 }
4162
4163 static
4164 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4165                             void __user *buffer, size_t *lenp, loff_t *ppos)
4166 {
4167         int *valp = ctl->data;
4168         int val = *valp;
4169         loff_t pos = *ppos;
4170         int ret;
4171
4172         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4173
4174         if (write)
4175                 ret = addrconf_disable_ipv6(ctl, valp, val);
4176         if (ret)
4177                 *ppos = pos;
4178         return ret;
4179 }
4180
4181 static struct addrconf_sysctl_table
4182 {
4183         struct ctl_table_header *sysctl_header;
4184         ctl_table addrconf_vars[DEVCONF_MAX+1];
4185         char *dev_name;
4186 } addrconf_sysctl __read_mostly = {
4187         .sysctl_header = NULL,
4188         .addrconf_vars = {
4189                 {
4190                         .procname       = "forwarding",
4191                         .data           = &ipv6_devconf.forwarding,
4192                         .maxlen         = sizeof(int),
4193                         .mode           = 0644,
4194                         .proc_handler   = addrconf_sysctl_forward,
4195                 },
4196                 {
4197                         .procname       = "hop_limit",
4198                         .data           = &ipv6_devconf.hop_limit,
4199                         .maxlen         = sizeof(int),
4200                         .mode           = 0644,
4201                         .proc_handler   = proc_dointvec,
4202                 },
4203                 {
4204                         .procname       = "mtu",
4205                         .data           = &ipv6_devconf.mtu6,
4206                         .maxlen         = sizeof(int),
4207                         .mode           = 0644,
4208                         .proc_handler   = proc_dointvec,
4209                 },
4210                 {
4211                         .procname       = "accept_ra",
4212                         .data           = &ipv6_devconf.accept_ra,
4213                         .maxlen         = sizeof(int),
4214                         .mode           = 0644,
4215                         .proc_handler   = proc_dointvec,
4216                 },
4217                 {
4218                         .procname       = "accept_redirects",
4219                         .data           = &ipv6_devconf.accept_redirects,
4220                         .maxlen         = sizeof(int),
4221                         .mode           = 0644,
4222                         .proc_handler   = proc_dointvec,
4223                 },
4224                 {
4225                         .procname       = "autoconf",
4226                         .data           = &ipv6_devconf.autoconf,
4227                         .maxlen         = sizeof(int),
4228                         .mode           = 0644,
4229                         .proc_handler   = proc_dointvec,
4230                 },
4231                 {
4232                         .procname       = "dad_transmits",
4233                         .data           = &ipv6_devconf.dad_transmits,
4234                         .maxlen         = sizeof(int),
4235                         .mode           = 0644,
4236                         .proc_handler   = proc_dointvec,
4237                 },
4238                 {
4239                         .procname       = "router_solicitations",
4240                         .data           = &ipv6_devconf.rtr_solicits,
4241                         .maxlen         = sizeof(int),
4242                         .mode           = 0644,
4243                         .proc_handler   = proc_dointvec,
4244                 },
4245                 {
4246                         .procname       = "router_solicitation_interval",
4247                         .data           = &ipv6_devconf.rtr_solicit_interval,
4248                         .maxlen         = sizeof(int),
4249                         .mode           = 0644,
4250                         .proc_handler   = proc_dointvec_jiffies,
4251                 },
4252                 {
4253                         .procname       = "router_solicitation_delay",
4254                         .data           = &ipv6_devconf.rtr_solicit_delay,
4255                         .maxlen         = sizeof(int),
4256                         .mode           = 0644,
4257                         .proc_handler   = proc_dointvec_jiffies,
4258                 },
4259                 {
4260                         .procname       = "force_mld_version",
4261                         .data           = &ipv6_devconf.force_mld_version,
4262                         .maxlen         = sizeof(int),
4263                         .mode           = 0644,
4264                         .proc_handler   = proc_dointvec,
4265                 },
4266 #ifdef CONFIG_IPV6_PRIVACY
4267                 {
4268                         .procname       = "use_tempaddr",
4269                         .data           = &ipv6_devconf.use_tempaddr,
4270                         .maxlen         = sizeof(int),
4271                         .mode           = 0644,
4272                         .proc_handler   = proc_dointvec,
4273                 },
4274                 {
4275                         .procname       = "temp_valid_lft",
4276                         .data           = &ipv6_devconf.temp_valid_lft,
4277                         .maxlen         = sizeof(int),
4278                         .mode           = 0644,
4279                         .proc_handler   = proc_dointvec,
4280                 },
4281                 {
4282                         .procname       = "temp_prefered_lft",
4283                         .data           = &ipv6_devconf.temp_prefered_lft,
4284                         .maxlen         = sizeof(int),
4285                         .mode           = 0644,
4286                         .proc_handler   = proc_dointvec,
4287                 },
4288                 {
4289                         .procname       = "regen_max_retry",
4290                         .data           = &ipv6_devconf.regen_max_retry,
4291                         .maxlen         = sizeof(int),
4292                         .mode           = 0644,
4293                         .proc_handler   = proc_dointvec,
4294                 },
4295                 {
4296                         .procname       = "max_desync_factor",
4297                         .data           = &ipv6_devconf.max_desync_factor,
4298                         .maxlen         = sizeof(int),
4299                         .mode           = 0644,
4300                         .proc_handler   = proc_dointvec,
4301                 },
4302 #endif
4303                 {
4304                         .procname       = "max_addresses",
4305                         .data           = &ipv6_devconf.max_addresses,
4306                         .maxlen         = sizeof(int),
4307                         .mode           = 0644,
4308                         .proc_handler   = proc_dointvec,
4309                 },
4310                 {
4311                         .procname       = "accept_ra_defrtr",
4312                         .data           = &ipv6_devconf.accept_ra_defrtr,
4313                         .maxlen         = sizeof(int),
4314                         .mode           = 0644,
4315                         .proc_handler   = proc_dointvec,
4316                 },
4317                 {
4318                         .procname       = "accept_ra_pinfo",
4319                         .data           = &ipv6_devconf.accept_ra_pinfo,
4320                         .maxlen         = sizeof(int),
4321                         .mode           = 0644,
4322                         .proc_handler   = proc_dointvec,
4323                 },
4324 #ifdef CONFIG_IPV6_ROUTER_PREF
4325                 {
4326                         .procname       = "accept_ra_rtr_pref",
4327                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4328                         .maxlen         = sizeof(int),
4329                         .mode           = 0644,
4330                         .proc_handler   = proc_dointvec,
4331                 },
4332                 {
4333                         .procname       = "router_probe_interval",
4334                         .data           = &ipv6_devconf.rtr_probe_interval,
4335                         .maxlen         = sizeof(int),
4336                         .mode           = 0644,
4337                         .proc_handler   = proc_dointvec_jiffies,
4338                 },
4339 #ifdef CONFIG_IPV6_ROUTE_INFO
4340                 {
4341                         .procname       = "accept_ra_rt_info_max_plen",
4342                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4343                         .maxlen         = sizeof(int),
4344                         .mode           = 0644,
4345                         .proc_handler   = proc_dointvec,
4346                 },
4347 #endif
4348 #endif
4349                 {
4350                         .procname       = "proxy_ndp",
4351                         .data           = &ipv6_devconf.proxy_ndp,
4352                         .maxlen         = sizeof(int),
4353                         .mode           = 0644,
4354                         .proc_handler   = proc_dointvec,
4355                 },
4356                 {
4357                         .procname       = "accept_source_route",
4358                         .data           = &ipv6_devconf.accept_source_route,
4359                         .maxlen         = sizeof(int),
4360                         .mode           = 0644,
4361                         .proc_handler   = proc_dointvec,
4362                 },
4363 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4364                 {
4365                         .procname       = "optimistic_dad",
4366                         .data           = &ipv6_devconf.optimistic_dad,
4367                         .maxlen         = sizeof(int),
4368                         .mode           = 0644,
4369                         .proc_handler   = proc_dointvec,
4370
4371                 },
4372 #endif
4373 #ifdef CONFIG_IPV6_MROUTE
4374                 {
4375                         .procname       = "mc_forwarding",
4376                         .data           = &ipv6_devconf.mc_forwarding,
4377                         .maxlen         = sizeof(int),
4378                         .mode           = 0444,
4379                         .proc_handler   = proc_dointvec,
4380                 },
4381 #endif
4382                 {
4383                         .procname       = "disable_ipv6",
4384                         .data           = &ipv6_devconf.disable_ipv6,
4385                         .maxlen         = sizeof(int),
4386                         .mode           = 0644,
4387                         .proc_handler   = addrconf_sysctl_disable,
4388                 },
4389                 {
4390                         .procname       = "accept_dad",
4391                         .data           = &ipv6_devconf.accept_dad,
4392                         .maxlen         = sizeof(int),
4393                         .mode           = 0644,
4394                         .proc_handler   = proc_dointvec,
4395                 },
4396                 {
4397                         .procname       = "force_tllao",
4398                         .data           = &ipv6_devconf.force_tllao,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec
4402                 },
4403                 {
4404                         /* sentinel */
4405                 }
4406         },
4407 };
4408
4409 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4410                 struct inet6_dev *idev, struct ipv6_devconf *p)
4411 {
4412         int i;
4413         struct addrconf_sysctl_table *t;
4414
4415 #define ADDRCONF_CTL_PATH_DEV   3
4416
4417         struct ctl_path addrconf_ctl_path[] = {
4418                 { .procname = "net", },
4419                 { .procname = "ipv6", },
4420                 { .procname = "conf", },
4421                 { /* to be set */ },
4422                 { },
4423         };
4424
4425
4426         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4427         if (t == NULL)
4428                 goto out;
4429
4430         for (i = 0; t->addrconf_vars[i].data; i++) {
4431                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4432                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4433                 t->addrconf_vars[i].extra2 = net;
4434         }
4435
4436         /*
4437          * Make a copy of dev_name, because '.procname' is regarded as const
4438          * by sysctl and we wouldn't want anyone to change it under our feet
4439          * (see SIOCSIFNAME).
4440          */
4441         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4442         if (!t->dev_name)
4443                 goto free;
4444
4445         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4446
4447         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4448                         t->addrconf_vars);
4449         if (t->sysctl_header == NULL)
4450                 goto free_procname;
4451
4452         p->sysctl = t;
4453         return 0;
4454
4455 free_procname:
4456         kfree(t->dev_name);
4457 free:
4458         kfree(t);
4459 out:
4460         return -ENOBUFS;
4461 }
4462
4463 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4464 {
4465         struct addrconf_sysctl_table *t;
4466
4467         if (p->sysctl == NULL)
4468                 return;
4469
4470         t = p->sysctl;
4471         p->sysctl = NULL;
4472         unregister_sysctl_table(t->sysctl_header);
4473         kfree(t->dev_name);
4474         kfree(t);
4475 }
4476
4477 static void addrconf_sysctl_register(struct inet6_dev *idev)
4478 {
4479         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4480                               &ndisc_ifinfo_sysctl_change);
4481         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4482                                         idev, &idev->cnf);
4483 }
4484
4485 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4486 {
4487         __addrconf_sysctl_unregister(&idev->cnf);
4488         neigh_sysctl_unregister(idev->nd_parms);
4489 }
4490
4491
4492 #endif
4493
4494 static int __net_init addrconf_init_net(struct net *net)
4495 {
4496         int err;
4497         struct ipv6_devconf *all, *dflt;
4498
4499         err = -ENOMEM;
4500         all = &ipv6_devconf;
4501         dflt = &ipv6_devconf_dflt;
4502
4503         if (!net_eq(net, &init_net)) {
4504                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4505                 if (all == NULL)
4506                         goto err_alloc_all;
4507
4508                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4509                 if (dflt == NULL)
4510                         goto err_alloc_dflt;
4511         } else {
4512                 /* these will be inherited by all namespaces */
4513                 dflt->autoconf = ipv6_defaults.autoconf;
4514                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4515         }
4516
4517         net->ipv6.devconf_all = all;
4518         net->ipv6.devconf_dflt = dflt;
4519
4520 #ifdef CONFIG_SYSCTL
4521         err = __addrconf_sysctl_register(net, "all", NULL, all);
4522         if (err < 0)
4523                 goto err_reg_all;
4524
4525         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4526         if (err < 0)
4527                 goto err_reg_dflt;
4528 #endif
4529         return 0;
4530
4531 #ifdef CONFIG_SYSCTL
4532 err_reg_dflt:
4533         __addrconf_sysctl_unregister(all);
4534 err_reg_all:
4535         kfree(dflt);
4536 #endif
4537 err_alloc_dflt:
4538         kfree(all);
4539 err_alloc_all:
4540         return err;
4541 }
4542
4543 static void __net_exit addrconf_exit_net(struct net *net)
4544 {
4545 #ifdef CONFIG_SYSCTL
4546         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4547         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4548 #endif
4549         if (!net_eq(net, &init_net)) {
4550                 kfree(net->ipv6.devconf_dflt);
4551                 kfree(net->ipv6.devconf_all);
4552         }
4553 }
4554
4555 static struct pernet_operations addrconf_ops = {
4556         .init = addrconf_init_net,
4557         .exit = addrconf_exit_net,
4558 };
4559
4560 /*
4561  *      Device notifier
4562  */
4563
4564 int register_inet6addr_notifier(struct notifier_block *nb)
4565 {
4566         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4567 }
4568 EXPORT_SYMBOL(register_inet6addr_notifier);
4569
4570 int unregister_inet6addr_notifier(struct notifier_block *nb)
4571 {
4572         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4573 }
4574 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4575
4576 /*
4577  *      Init / cleanup code
4578  */
4579
4580 int __init addrconf_init(void)
4581 {
4582         int i, err;
4583
4584         err = ipv6_addr_label_init();
4585         if (err < 0) {
4586                 printk(KERN_CRIT "IPv6 Addrconf:"
4587                        " cannot initialize default policy table: %d.\n", err);
4588                 return err;
4589         }
4590
4591         register_pernet_subsys(&addrconf_ops);
4592
4593         /* The addrconf netdev notifier requires that loopback_dev
4594          * has it's ipv6 private information allocated and setup
4595          * before it can bring up and give link-local addresses
4596          * to other devices which are up.
4597          *
4598          * Unfortunately, loopback_dev is not necessarily the first
4599          * entry in the global dev_base list of net devices.  In fact,
4600          * it is likely to be the very last entry on that list.
4601          * So this causes the notifier registry below to try and
4602          * give link-local addresses to all devices besides loopback_dev
4603          * first, then loopback_dev, which cases all the non-loopback_dev
4604          * devices to fail to get a link-local address.
4605          *
4606          * So, as a temporary fix, allocate the ipv6 structure for
4607          * loopback_dev first by hand.
4608          * Longer term, all of the dependencies ipv6 has upon the loopback
4609          * device and it being up should be removed.
4610          */
4611         rtnl_lock();
4612         if (!ipv6_add_dev(init_net.loopback_dev))
4613                 err = -ENOMEM;
4614         rtnl_unlock();
4615         if (err)
4616                 goto errlo;
4617
4618         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4619                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4620
4621         register_netdevice_notifier(&ipv6_dev_notf);
4622
4623         addrconf_verify(0);
4624
4625         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4626         if (err < 0)
4627                 goto errout;
4628
4629         /* Only the first call to __rtnl_register can fail */
4630         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4631         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4632         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4633         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4634         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4635
4636         ipv6_addr_label_rtnl_register();
4637
4638         return 0;
4639 errout:
4640         unregister_netdevice_notifier(&ipv6_dev_notf);
4641 errlo:
4642         unregister_pernet_subsys(&addrconf_ops);
4643
4644         return err;
4645 }
4646
4647 void addrconf_cleanup(void)
4648 {
4649         struct net_device *dev;
4650         int i;
4651
4652         unregister_netdevice_notifier(&ipv6_dev_notf);
4653         unregister_pernet_subsys(&addrconf_ops);
4654
4655         rtnl_lock();
4656
4657         /* clean dev list */
4658         for_each_netdev(&init_net, dev) {
4659                 if (__in6_dev_get(dev) == NULL)
4660                         continue;
4661                 addrconf_ifdown(dev, 1);
4662         }
4663         addrconf_ifdown(init_net.loopback_dev, 2);
4664
4665         /*
4666          *      Check hash table.
4667          */
4668         spin_lock_bh(&addrconf_hash_lock);
4669         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4670                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4671         spin_unlock_bh(&addrconf_hash_lock);
4672
4673         del_timer(&addr_chk_timer);
4674         rtnl_unlock();
4675 }