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

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