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