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 #endif
 125
 126 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 127 static int ipv6_count_addresses(struct inet6_dev *idev);
 128
 129 /*
 130  *      Configured unicast address hash table
 131  */
 132 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 133 static DEFINE_SPINLOCK(addrconf_hash_lock);
 134
 135 static void addrconf_verify(unsigned long);
 136
 137 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 138 static DEFINE_SPINLOCK(addrconf_verify_lock);
 139
 140 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 141 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 142
 143 static void addrconf_type_change(struct net_device *dev,
 144                                  unsigned long event);
 145 static int addrconf_ifdown(struct net_device *dev, int how);
 146
 147 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 148 static void addrconf_dad_timer(unsigned long data);
 149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 150 static void addrconf_dad_run(struct inet6_dev *idev);
 151 static void addrconf_rs_timer(unsigned long data);
 152 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 153 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 154
 155 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 156                                 struct prefix_info *pinfo);
 157 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 158                                struct net_device *dev);
 159
 160 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 161
 162 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 163         .forwarding             = 0,
 164         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 165         .mtu6                   = IPV6_MIN_MTU,
 166         .accept_ra              = 1,
 167         .accept_redirects       = 1,
 168         .autoconf               = 1,
 169         .force_mld_version      = 0,
 170         .dad_transmits          = 1,
 171         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 172         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 173         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 174 #ifdef CONFIG_IPV6_PRIVACY
 175         .use_tempaddr           = 0,
 176         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 177         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 178         .regen_max_retry        = REGEN_MAX_RETRY,
 179         .max_desync_factor      = MAX_DESYNC_FACTOR,
 180 #endif
 181         .max_addresses          = IPV6_MAX_ADDRESSES,
 182         .accept_ra_defrtr       = 1,
 183         .accept_ra_pinfo        = 1,
 184 #ifdef CONFIG_IPV6_ROUTER_PREF
 185         .accept_ra_rtr_pref     = 1,
 186         .rtr_probe_interval     = 60 * HZ,
 187 #ifdef CONFIG_IPV6_ROUTE_INFO
 188         .accept_ra_rt_info_max_plen = 0,
 189 #endif
 190 #endif
 191         .proxy_ndp              = 0,
 192         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 193         .disable_ipv6           = 0,
 194         .accept_dad             = 1,
 195 };
 196
 197 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 198         .forwarding             = 0,
 199         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 200         .mtu6                   = IPV6_MIN_MTU,
 201         .accept_ra              = 1,
 202         .accept_redirects       = 1,
 203         .autoconf               = 1,
 204         .dad_transmits          = 1,
 205         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 206         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 207         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 208 #ifdef CONFIG_IPV6_PRIVACY
 209         .use_tempaddr           = 0,
 210         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 211         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 212         .regen_max_retry        = REGEN_MAX_RETRY,
 213         .max_desync_factor      = MAX_DESYNC_FACTOR,
 214 #endif
 215         .max_addresses          = IPV6_MAX_ADDRESSES,
 216         .accept_ra_defrtr       = 1,
 217         .accept_ra_pinfo        = 1,
 218 #ifdef CONFIG_IPV6_ROUTER_PREF
 219         .accept_ra_rtr_pref     = 1,
 220         .rtr_probe_interval     = 60 * HZ,
 221 #ifdef CONFIG_IPV6_ROUTE_INFO
 222         .accept_ra_rt_info_max_plen = 0,
 223 #endif
 224 #endif
 225         .proxy_ndp              = 0,
 226         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 227         .disable_ipv6           = 0,
 228         .accept_dad             = 1,
 229 };
 230
 231 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 232 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 233 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 234 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 235 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 236
 237 /* Check if a valid qdisc is available */
 238 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 239 {
 240         return !qdisc_tx_is_noop(dev);
 241 }
 242
 243 /* Check if a route is valid prefix route */
 244 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 245 {
 246         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 247 }
 248
 249 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 250 {
 251         if (del_timer(&ifp->timer))
 252                 __in6_ifa_put(ifp);
 253 }
 254
 255 enum addrconf_timer_t {
 256         AC_NONE,
 257         AC_DAD,
 258         AC_RS,
 259 };
 260
 261 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 262                                enum addrconf_timer_t what,
 263                                unsigned long when)
 264 {
 265         if (!del_timer(&ifp->timer))
 266                 in6_ifa_hold(ifp);
 267
 268         switch (what) {
 269         case AC_DAD:
 270                 ifp->timer.function = addrconf_dad_timer;
 271                 break;
 272         case AC_RS:
 273                 ifp->timer.function = addrconf_rs_timer;
 274                 break;
 275         default:
 276                 break;
 277         }
 278         ifp->timer.expires = jiffies + when;
 279         add_timer(&ifp->timer);
 280 }
 281
 282 static int snmp6_alloc_dev(struct inet6_dev *idev)
 283 {
 284         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 285                           sizeof(struct ipstats_mib),
 286                           __alignof__(struct ipstats_mib)) < 0)
 287                 goto err_ip;
 288         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 289                           sizeof(struct icmpv6_mib),
 290                           __alignof__(struct icmpv6_mib)) < 0)
 291                 goto err_icmp;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 293                           sizeof(struct icmpv6msg_mib),
 294                           __alignof__(struct icmpv6msg_mib)) < 0)
 295                 goto err_icmpmsg;
 296
 297         return 0;
 298
 299 err_icmpmsg:
 300         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 301 err_icmp:
 302         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 303 err_ip:
 304         return -ENOMEM;
 305 }
 306
 307 static void snmp6_free_dev(struct inet6_dev *idev)
 308 {
 309         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 310         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 }
 313
 314 /* Nobody refers to this device, we may destroy it. */
 315
 316 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 317 {
 318         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 319         kfree(idev);
 320 }
 321
 322 void in6_dev_finish_destroy(struct inet6_dev *idev)
 323 {
 324         struct net_device *dev = idev->dev;
 325
 326         WARN_ON(!list_empty(&idev->addr_list));
 327         WARN_ON(idev->mc_list != NULL);
 328
 329 #ifdef NET_REFCNT_DEBUG
 330         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 331 #endif
 332         dev_put(dev);
 333         if (!idev->dead) {
 334                 pr_warning("Freeing alive inet6 device %p\n", idev);
 335                 return;
 336         }
 337         snmp6_free_dev(idev);
 338         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 339 }
 340
 341 EXPORT_SYMBOL(in6_dev_finish_destroy);
 342
 343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 344 {
 345         struct inet6_dev *ndev;
 346
 347         ASSERT_RTNL();
 348
 349         if (dev->mtu < IPV6_MIN_MTU)
 350                 return NULL;
 351
 352         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 353
 354         if (ndev == NULL)
 355                 return NULL;
 356
 357         rwlock_init(&ndev->lock);
 358         ndev->dev = dev;
 359         INIT_LIST_HEAD(&ndev->addr_list);
 360
 361         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 362         ndev->cnf.mtu6 = dev->mtu;
 363         ndev->cnf.sysctl = NULL;
 364         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 365         if (ndev->nd_parms == NULL) {
 366                 kfree(ndev);
 367                 return NULL;
 368         }
 369         if (ndev->cnf.forwarding)
 370                 dev_disable_lro(dev);
 371         /* We refer to the device */
 372         dev_hold(dev);
 373
 374         if (snmp6_alloc_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         if (snmp6_register_dev(ndev) < 0) {
 385                 ADBG((KERN_WARNING
 386                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 387                         __func__, dev->name));
 388                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 389                 ndev->dead = 1;
 390                 in6_dev_finish_destroy(ndev);
 391                 return NULL;
 392         }
 393
 394         /* One reference from device.  We must do this before
 395          * we invoke __ipv6_regen_rndid().
 396          */
 397         in6_dev_hold(ndev);
 398
 399         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 400                 ndev->cnf.accept_dad = -1;
 401
 402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 403         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 404                 printk(KERN_INFO
 405                        "%s: Disabled Multicast RS\n",
 406                        dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 printk(KERN_INFO
 420                        "%s: Disabled Privacy Extensions\n",
 421                        dev->name);
 422                 ndev->cnf.use_tempaddr = -1;
 423         } else {
 424                 in6_dev_hold(ndev);
 425                 ipv6_regen_rndid((unsigned long) ndev);
 426         }
 427 #endif
 428
 429         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 430                 ndev->if_flags |= IF_READY;
 431
 432         ipv6_mc_init_dev(ndev);
 433         ndev->tstamp = jiffies;
 434         addrconf_sysctl_register(ndev);
 435         /* protected by rtnl_lock */
 436         rcu_assign_pointer(dev->ip6_ptr, ndev);
 437
 438         /* Join all-node multicast group */
 439         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         rcu_read_lock();
 497         for_each_netdev_rcu(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506         rcu_read_unlock();
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 510 {
 511         struct net *net;
 512
 513         net = (struct net *)table->extra2;
 514         if (p == &net->ipv6.devconf_dflt->forwarding)
 515                 return 0;
 516
 517         if (!rtnl_trylock()) {
 518                 /* Restore the original values before restarting */
 519                 *p = old;
 520                 return restart_syscall();
 521         }
 522
 523         if (p == &net->ipv6.devconf_all->forwarding) {
 524                 __s32 newf = net->ipv6.devconf_all->forwarding;
 525                 net->ipv6.devconf_dflt->forwarding = newf;
 526                 addrconf_forward_change(net, newf);
 527         } else if ((!*p) ^ (!old))
 528                 dev_forward_change((struct inet6_dev *)table->extra1);
 529         rtnl_unlock();
 530
 531         if (*p)
 532                 rt6_purge_dflt_routers(net);
 533         return 1;
 534 }
 535 #endif
 536
 537 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 538 {
 539         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 540         kfree(ifp);
 541 }
 542
 543 /* Nobody refers to this ifaddr, destroy it */
 544 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 545 {
 546         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 547
 548 #ifdef NET_REFCNT_DEBUG
 549         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 550 #endif
 551
 552         in6_dev_put(ifp->idev);
 553
 554         if (del_timer(&ifp->timer))
 555                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 556
 557         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 558                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 559                 return;
 560         }
 561         dst_release(&ifp->rt->dst);
 562
 563         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 564 }
 565
 566 static void
 567 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 568 {
 569         struct list_head *p;
 570         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 571
 572         /*
 573          * Each device address list is sorted in order of scope -
 574          * global before linklocal.
 575          */
 576         list_for_each(p, &idev->addr_list) {
 577                 struct inet6_ifaddr *ifa
 578                         = list_entry(p, struct inet6_ifaddr, if_list);
 579                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 580                         break;
 581         }
 582
 583         list_add_tail(&ifp->if_list, p);
 584 }
 585
 586 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 587 {
 588         /*
 589          * We perform the hash function over the last 64 bits of the address
 590          * This will include the IEEE address token on links that support it.
 591          */
 592         return jhash_2words((__force u32)addr->s6_addr32[2],
 593                             (__force u32)addr->s6_addr32[3], 0)
 594                 & (IN6_ADDR_HSIZE - 1);
 595 }
 596
 597 /* On success it returns ifp with increased reference count */
 598
 599 static struct inet6_ifaddr *
 600 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 601               int scope, u32 flags)
 602 {
 603         struct inet6_ifaddr *ifa = NULL;
 604         struct rt6_info *rt;
 605         unsigned int hash;
 606         int err = 0;
 607         int addr_type = ipv6_addr_type(addr);
 608
 609         if (addr_type == IPV6_ADDR_ANY ||
 610             addr_type & IPV6_ADDR_MULTICAST ||
 611             (!(idev->dev->flags & IFF_LOOPBACK) &&
 612              addr_type & IPV6_ADDR_LOOPBACK))
 613                 return ERR_PTR(-EADDRNOTAVAIL);
 614
 615         rcu_read_lock_bh();
 616         if (idev->dead) {
 617                 err = -ENODEV;                  /*XXX*/
 618                 goto out2;
 619         }
 620
 621         if (idev->cnf.disable_ipv6) {
 622                 err = -EACCES;
 623                 goto out2;
 624         }
 625
 626         spin_lock(&addrconf_hash_lock);
 627
 628         /* Ignore adding duplicate addresses on an interface */
 629         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 630                 ADBG(("ipv6_add_addr: already assigned\n"));
 631                 err = -EEXIST;
 632                 goto out;
 633         }
 634
 635         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 636
 637         if (ifa == NULL) {
 638                 ADBG(("ipv6_add_addr: malloc failed\n"));
 639                 err = -ENOBUFS;
 640                 goto out;
 641         }
 642
 643         rt = addrconf_dst_alloc(idev, addr, 0);
 644         if (IS_ERR(rt)) {
 645                 err = PTR_ERR(rt);
 646                 goto out;
 647         }
 648
 649         ipv6_addr_copy(&ifa->addr, addr);
 650
 651         spin_lock_init(&ifa->lock);
 652         spin_lock_init(&ifa->state_lock);
 653         init_timer(&ifa->timer);
 654         INIT_HLIST_NODE(&ifa->addr_lst);
 655         ifa->timer.data = (unsigned long) ifa;
 656         ifa->scope = scope;
 657         ifa->prefix_len = pfxlen;
 658         ifa->flags = flags | IFA_F_TENTATIVE;
 659         ifa->cstamp = ifa->tstamp = jiffies;
 660
 661         ifa->rt = rt;
 662
 663         /*
 664          * part one of RFC 4429, section 3.3
 665          * We should not configure an address as
 666          * optimistic if we do not yet know the link
 667          * layer address of our nexhop router
 668          */
 669
 670         if (rt->rt6i_nexthop == NULL)
 671                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 672
 673         ifa->idev = idev;
 674         in6_dev_hold(idev);
 675         /* For caller */
 676         in6_ifa_hold(ifa);
 677
 678         /* Add to big hash table */
 679         hash = ipv6_addr_hash(addr);
 680
 681         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 682         spin_unlock(&addrconf_hash_lock);
 683
 684         write_lock(&idev->lock);
 685         /* Add to inet6_dev unicast addr list. */
 686         ipv6_link_dev_addr(idev, ifa);
 687
 688 #ifdef CONFIG_IPV6_PRIVACY
 689         if (ifa->flags&IFA_F_TEMPORARY) {
 690                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 691                 in6_ifa_hold(ifa);
 692         }
 693 #endif
 694
 695         in6_ifa_hold(ifa);
 696         write_unlock(&idev->lock);
 697 out2:
 698         rcu_read_unlock_bh();
 699
 700         if (likely(err == 0))
 701                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 702         else {
 703                 kfree(ifa);
 704                 ifa = ERR_PTR(err);
 705         }
 706
 707         return ifa;
 708 out:
 709         spin_unlock(&addrconf_hash_lock);
 710         goto out2;
 711 }
 712
 713 /* This function wants to get referenced ifp and releases it before return */
 714
 715 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 716 {
 717         struct inet6_ifaddr *ifa, *ifn;
 718         struct inet6_dev *idev = ifp->idev;
 719         int state;
 720         int hash;
 721         int deleted = 0, onlink = 0;
 722         unsigned long expires = jiffies;
 723
 724         hash = ipv6_addr_hash(&ifp->addr);
 725
 726         spin_lock_bh(&ifp->state_lock);
 727         state = ifp->state;
 728         ifp->state = INET6_IFADDR_STATE_DEAD;
 729         spin_unlock_bh(&ifp->state_lock);
 730
 731         if (state == INET6_IFADDR_STATE_DEAD)
 732                 goto out;
 733
 734         spin_lock_bh(&addrconf_hash_lock);
 735         hlist_del_init_rcu(&ifp->addr_lst);
 736         spin_unlock_bh(&addrconf_hash_lock);
 737
 738         write_lock_bh(&idev->lock);
 739 #ifdef CONFIG_IPV6_PRIVACY
 740         if (ifp->flags&IFA_F_TEMPORARY) {
 741                 list_del(&ifp->tmp_list);
 742                 if (ifp->ifpub) {
 743                         in6_ifa_put(ifp->ifpub);
 744                         ifp->ifpub = NULL;
 745                 }
 746                 __in6_ifa_put(ifp);
 747         }
 748 #endif
 749
 750         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 751                 if (ifa == ifp) {
 752                         list_del_init(&ifp->if_list);
 753                         __in6_ifa_put(ifp);
 754
 755                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 756                                 break;
 757                         deleted = 1;
 758                         continue;
 759                 } else if (ifp->flags & IFA_F_PERMANENT) {
 760                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 761                                               ifp->prefix_len)) {
 762                                 if (ifa->flags & IFA_F_PERMANENT) {
 763                                         onlink = 1;
 764                                         if (deleted)
 765                                                 break;
 766                                 } else {
 767                                         unsigned long lifetime;
 768
 769                                         if (!onlink)
 770                                                 onlink = -1;
 771
 772                                         spin_lock(&ifa->lock);
 773
 774                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 775                                         /*
 776                                          * Note: Because this address is
 777                                          * not permanent, lifetime <
 778                                          * LONG_MAX / HZ here.
 779                                          */
 780                                         if (time_before(expires,
 781                                                         ifa->tstamp + lifetime * HZ))
 782                                                 expires = ifa->tstamp + lifetime * HZ;
 783                                         spin_unlock(&ifa->lock);
 784                                 }
 785                         }
 786                 }
 787         }
 788         write_unlock_bh(&idev->lock);
 789
 790         addrconf_del_timer(ifp);
 791
 792         ipv6_ifa_notify(RTM_DELADDR, ifp);
 793
 794         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 795
 796         /*
 797          * Purge or update corresponding prefix
 798          *
 799          * 1) we don't purge prefix here if address was not permanent.
 800          *    prefix is managed by its own lifetime.
 801          * 2) if there're no addresses, delete prefix.
 802          * 3) if there're still other permanent address(es),
 803          *    corresponding prefix is still permanent.
 804          * 4) otherwise, update prefix lifetime to the
 805          *    longest valid lifetime among the corresponding
 806          *    addresses on the device.
 807          *    Note: subsequent RA will update lifetime.
 808          *
 809          * --yoshfuji
 810          */
 811         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 812                 struct in6_addr prefix;
 813                 struct rt6_info *rt;
 814                 struct net *net = dev_net(ifp->idev->dev);
 815                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 816                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 817
 818                 if (rt && addrconf_is_prefix_route(rt)) {
 819                         if (onlink == 0) {
 820                                 ip6_del_rt(rt);
 821                                 rt = NULL;
 822                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 823                                 rt->rt6i_expires = expires;
 824                                 rt->rt6i_flags |= RTF_EXPIRES;
 825                         }
 826                 }
 827                 dst_release(&rt->dst);
 828         }
 829
 830 out:
 831         in6_ifa_put(ifp);
 832 }
 833
 834 #ifdef CONFIG_IPV6_PRIVACY
 835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 836 {
 837         struct inet6_dev *idev = ifp->idev;
 838         struct in6_addr addr, *tmpaddr;
 839         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 840         unsigned long regen_advance;
 841         int tmp_plen;
 842         int ret = 0;
 843         int max_addresses;
 844         u32 addr_flags;
 845
 846         write_lock(&idev->lock);
 847         if (ift) {
 848                 spin_lock_bh(&ift->lock);
 849                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 850                 spin_unlock_bh(&ift->lock);
 851                 tmpaddr = &addr;
 852         } else {
 853                 tmpaddr = NULL;
 854         }
 855 retry:
 856         in6_dev_hold(idev);
 857         if (idev->cnf.use_tempaddr <= 0) {
 858                 write_unlock(&idev->lock);
 859                 printk(KERN_INFO
 860                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 861                 in6_dev_put(idev);
 862                 ret = -1;
 863                 goto out;
 864         }
 865         spin_lock_bh(&ifp->lock);
 866         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 867                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 868                 spin_unlock_bh(&ifp->lock);
 869                 write_unlock(&idev->lock);
 870                 printk(KERN_WARNING
 871                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         in6_ifa_hold(ifp);
 877         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 878         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 879                 spin_unlock_bh(&ifp->lock);
 880                 write_unlock(&idev->lock);
 881                 printk(KERN_WARNING
 882                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 883                 in6_ifa_put(ifp);
 884                 in6_dev_put(idev);
 885                 ret = -1;
 886                 goto out;
 887         }
 888         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 889         tmp_valid_lft = min_t(__u32,
 890                               ifp->valid_lft,
 891                               idev->cnf.temp_valid_lft);
 892         tmp_prefered_lft = min_t(__u32,
 893                                  ifp->prefered_lft,
 894                                  idev->cnf.temp_prefered_lft -
 895                                  idev->cnf.max_desync_factor);
 896         tmp_plen = ifp->prefix_len;
 897         max_addresses = idev->cnf.max_addresses;
 898         tmp_cstamp = ifp->cstamp;
 899         tmp_tstamp = ifp->tstamp;
 900         spin_unlock_bh(&ifp->lock);
 901
 902         regen_advance = idev->cnf.regen_max_retry *
 903                         idev->cnf.dad_transmits *
 904                         idev->nd_parms->retrans_time / HZ;
 905         write_unlock(&idev->lock);
 906
 907         /* A temporary address is created only if this calculated Preferred
 908          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 909          * an implementation must not create a temporary address with a zero
 910          * Preferred Lifetime.
 911          */
 912         if (tmp_prefered_lft <= regen_advance) {
 913                 in6_ifa_put(ifp);
 914                 in6_dev_put(idev);
 915                 ret = -1;
 916                 goto out;
 917         }
 918
 919         addr_flags = IFA_F_TEMPORARY;
 920         /* set in addrconf_prefix_rcv() */
 921         if (ifp->flags & IFA_F_OPTIMISTIC)
 922                 addr_flags |= IFA_F_OPTIMISTIC;
 923
 924         ift = !max_addresses ||
 925               ipv6_count_addresses(idev) < max_addresses ?
 926                 ipv6_add_addr(idev, &addr, tmp_plen,
 927                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 928                               addr_flags) : NULL;
 929         if (!ift || IS_ERR(ift)) {
 930                 in6_ifa_put(ifp);
 931                 in6_dev_put(idev);
 932                 printk(KERN_INFO
 933                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 934                 tmpaddr = &addr;
 935                 write_lock(&idev->lock);
 936                 goto retry;
 937         }
 938
 939         spin_lock_bh(&ift->lock);
 940         ift->ifpub = ifp;
 941         ift->valid_lft = tmp_valid_lft;
 942         ift->prefered_lft = tmp_prefered_lft;
 943         ift->cstamp = tmp_cstamp;
 944         ift->tstamp = tmp_tstamp;
 945         spin_unlock_bh(&ift->lock);
 946
 947         addrconf_dad_start(ift, 0);
 948         in6_ifa_put(ift);
 949         in6_dev_put(idev);
 950 out:
 951         return ret;
 952 }
 953 #endif
 954
 955 /*
 956  *      Choose an appropriate source address (RFC3484)
 957  */
 958 enum {
 959         IPV6_SADDR_RULE_INIT = 0,
 960         IPV6_SADDR_RULE_LOCAL,
 961         IPV6_SADDR_RULE_SCOPE,
 962         IPV6_SADDR_RULE_PREFERRED,
 963 #ifdef CONFIG_IPV6_MIP6
 964         IPV6_SADDR_RULE_HOA,
 965 #endif
 966         IPV6_SADDR_RULE_OIF,
 967         IPV6_SADDR_RULE_LABEL,
 968 #ifdef CONFIG_IPV6_PRIVACY
 969         IPV6_SADDR_RULE_PRIVACY,
 970 #endif
 971         IPV6_SADDR_RULE_ORCHID,
 972         IPV6_SADDR_RULE_PREFIX,
 973         IPV6_SADDR_RULE_MAX
 974 };
 975
 976 struct ipv6_saddr_score {
 977         int                     rule;
 978         int                     addr_type;
 979         struct inet6_ifaddr     *ifa;
 980         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 981         int                     scopedist;
 982         int                     matchlen;
 983 };
 984
 985 struct ipv6_saddr_dst {
 986         const struct in6_addr *addr;
 987         int ifindex;
 988         int scope;
 989         int label;
 990         unsigned int prefs;
 991 };
 992
 993 static inline int ipv6_saddr_preferred(int type)
 994 {
 995         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 996                 return 1;
 997         return 0;
 998 }
 999
1000 static int ipv6_get_saddr_eval(struct net *net,
1001                                struct ipv6_saddr_score *score,
1002                                struct ipv6_saddr_dst *dst,
1003                                int i)
1004 {
1005         int ret;
1006
1007         if (i <= score->rule) {
1008                 switch (i) {
1009                 case IPV6_SADDR_RULE_SCOPE:
1010                         ret = score->scopedist;
1011                         break;
1012                 case IPV6_SADDR_RULE_PREFIX:
1013                         ret = score->matchlen;
1014                         break;
1015                 default:
1016                         ret = !!test_bit(i, score->scorebits);
1017                 }
1018                 goto out;
1019         }
1020
1021         switch (i) {
1022         case IPV6_SADDR_RULE_INIT:
1023                 /* Rule 0: remember if hiscore is not ready yet */
1024                 ret = !!score->ifa;
1025                 break;
1026         case IPV6_SADDR_RULE_LOCAL:
1027                 /* Rule 1: Prefer same address */
1028                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029                 break;
1030         case IPV6_SADDR_RULE_SCOPE:
1031                 /* Rule 2: Prefer appropriate scope
1032                  *
1033                  *      ret
1034                  *       ^
1035                  *    -1 |  d 15
1036                  *    ---+--+-+---> scope
1037                  *       |
1038                  *       |             d is scope of the destination.
1039                  *  B-d  |  \
1040                  *       |   \      <- smaller scope is better if
1041                  *  B-15 |    \        if scope is enough for destinaion.
1042                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1043                  * d-C-1 | /
1044                  *       |/         <- greater is better
1045                  *   -C  /             if scope is not enough for destination.
1046                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1047                  *
1048                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050                  * Assume B = 0 and we get C > 29.
1051                  */
1052                 ret = __ipv6_addr_src_scope(score->addr_type);
1053                 if (ret >= dst->scope)
1054                         ret = -ret;
1055                 else
1056                         ret -= 128;     /* 30 is enough */
1057                 score->scopedist = ret;
1058                 break;
1059         case IPV6_SADDR_RULE_PREFERRED:
1060                 /* Rule 3: Avoid deprecated and optimistic addresses */
1061                 ret = ipv6_saddr_preferred(score->addr_type) ||
1062                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063                 break;
1064 #ifdef CONFIG_IPV6_MIP6
1065         case IPV6_SADDR_RULE_HOA:
1066             {
1067                 /* Rule 4: Prefer home address */
1068                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070                 break;
1071             }
1072 #endif
1073         case IPV6_SADDR_RULE_OIF:
1074                 /* Rule 5: Prefer outgoing interface */
1075                 ret = (!dst->ifindex ||
1076                        dst->ifindex == score->ifa->idev->dev->ifindex);
1077                 break;
1078         case IPV6_SADDR_RULE_LABEL:
1079                 /* Rule 6: Prefer matching label */
1080                 ret = ipv6_addr_label(net,
1081                                       &score->ifa->addr, score->addr_type,
1082                                       score->ifa->idev->dev->ifindex) == dst->label;
1083                 break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085         case IPV6_SADDR_RULE_PRIVACY:
1086             {
1087                 /* Rule 7: Prefer public address
1088                  * Note: prefer temprary address if use_tempaddr >= 2
1089                  */
1090                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1093                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094                 break;
1095             }
1096 #endif
1097         case IPV6_SADDR_RULE_ORCHID:
1098                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1099                  *          non-ORCHID vs non-ORCHID
1100                  */
1101                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102                         ipv6_addr_orchid(dst->addr));
1103                 break;
1104         case IPV6_SADDR_RULE_PREFIX:
1105                 /* Rule 8: Use longest matching prefix */
1106                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107                                                        dst->addr);
1108                 break;
1109         default:
1110                 ret = 0;
1111         }
1112
1113         if (ret)
1114                 __set_bit(i, score->scorebits);
1115         score->rule = i;
1116 out:
1117         return ret;
1118 }
1119
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121                        const struct in6_addr *daddr, unsigned int prefs,
1122                        struct in6_addr *saddr)
1123 {
1124         struct ipv6_saddr_score scores[2],
1125                                 *score = &scores[0], *hiscore = &scores[1];
1126         struct ipv6_saddr_dst dst;
1127         struct net_device *dev;
1128         int dst_type;
1129
1130         dst_type = __ipv6_addr_type(daddr);
1131         dst.addr = daddr;
1132         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133         dst.scope = __ipv6_addr_src_scope(dst_type);
1134         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135         dst.prefs = prefs;
1136
1137         hiscore->rule = -1;
1138         hiscore->ifa = NULL;
1139
1140         rcu_read_lock();
1141
1142         for_each_netdev_rcu(net, dev) {
1143                 struct inet6_dev *idev;
1144
1145                 /* Candidate Source Address (section 4)
1146                  *  - multicast and link-local destination address,
1147                  *    the set of candidate source address MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same link as the outgoing
1150                  *    interface.
1151                  * (- For site-local destination addresses, the
1152                  *    set of candidate source addresses MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same site as the outgoing
1155                  *    interface.)
1156                  */
1157                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159                     dst.ifindex && dev->ifindex != dst.ifindex)
1160                         continue;
1161
1162                 idev = __in6_dev_get(dev);
1163                 if (!idev)
1164                         continue;
1165
1166                 read_lock_bh(&idev->lock);
1167                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1168                         int i;
1169
1170                         /*
1171                          * - Tentative Address (RFC2462 section 5.4)
1172                          *  - A tentative address is not considered
1173                          *    "assigned to an interface" in the traditional
1174                          *    sense, unless it is also flagged as optimistic.
1175                          * - Candidate Source Address (section 4)
1176                          *  - In any case, anycast addresses, multicast
1177                          *    addresses, and the unspecified address MUST
1178                          *    NOT be included in a candidate set.
1179                          */
1180                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182                                 continue;
1183
1184                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185
1186                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1188                                 LIMIT_NETDEBUG(KERN_DEBUG
1189                                                "ADDRCONF: unspecified / multicast address "
1190                                                "assigned as unicast address on %s",
1191                                                dev->name);
1192                                 continue;
1193                         }
1194
1195                         score->rule = -1;
1196                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197
1198                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199                                 int minihiscore, miniscore;
1200
1201                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203
1204                                 if (minihiscore > miniscore) {
1205                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1206                                             score->scopedist > 0) {
1207                                                 /*
1208                                                  * special case:
1209                                                  * each remaining entry
1210                                                  * has too small (not enough)
1211                                                  * scope, because ifa entries
1212                                                  * are sorted by their scope
1213                                                  * values.
1214                                                  */
1215                                                 goto try_nextdev;
1216                                         }
1217                                         break;
1218                                 } else if (minihiscore < miniscore) {
1219                                         if (hiscore->ifa)
1220                                                 in6_ifa_put(hiscore->ifa);
1221
1222                                         in6_ifa_hold(score->ifa);
1223
1224                                         swap(hiscore, score);
1225
1226                                         /* restore our iterator */
1227                                         score->ifa = hiscore->ifa;
1228
1229                                         break;
1230                                 }
1231                         }
1232                 }
1233 try_nextdev:
1234                 read_unlock_bh(&idev->lock);
1235         }
1236         rcu_read_unlock();
1237
1238         if (!hiscore->ifa)
1239                 return -EADDRNOTAVAIL;
1240
1241         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242         in6_ifa_put(hiscore->ifa);
1243         return 0;
1244 }
1245 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1246
1247 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1248                     unsigned char banned_flags)
1249 {
1250         struct inet6_dev *idev;
1251         int err = -EADDRNOTAVAIL;
1252
1253         rcu_read_lock();
1254         idev = __in6_dev_get(dev);
1255         if (idev) {
1256                 struct inet6_ifaddr *ifp;
1257
1258                 read_lock_bh(&idev->lock);
1259                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1260                         if (ifp->scope == IFA_LINK &&
1261                             !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         list_for_each_entry(ifp, &idev->addr_list, if_list)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr *ifp;
1289         struct hlist_node *node;
1290         unsigned int hash = ipv6_addr_hash(addr);
1291
1292         rcu_read_lock_bh();
1293         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1294                 if (!net_eq(dev_net(ifp->idev->dev), net))
1295                         continue;
1296                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1297                     !(ifp->flags&IFA_F_TENTATIVE) &&
1298                     (dev == NULL || ifp->idev->dev == dev ||
1299                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1300                         rcu_read_unlock_bh();
1301                         return 1;
1302                 }
1303         }
1304
1305         rcu_read_unlock_bh();
1306         return 0;
1307 }
1308 EXPORT_SYMBOL(ipv6_chk_addr);
1309
1310 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1311                                struct net_device *dev)
1312 {
1313         unsigned int hash = ipv6_addr_hash(addr);
1314         struct inet6_ifaddr *ifp;
1315         struct hlist_node *node;
1316
1317         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1318                 if (!net_eq(dev_net(ifp->idev->dev), net))
1319                         continue;
1320                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1321                         if (dev == NULL || ifp->idev->dev == dev)
1322                                 return true;
1323                 }
1324         }
1325         return false;
1326 }
1327
1328 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1329 {
1330         struct inet6_dev *idev;
1331         struct inet6_ifaddr *ifa;
1332         int     onlink;
1333
1334         onlink = 0;
1335         rcu_read_lock();
1336         idev = __in6_dev_get(dev);
1337         if (idev) {
1338                 read_lock_bh(&idev->lock);
1339                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1340                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1341                                                    ifa->prefix_len);
1342                         if (onlink)
1343                                 break;
1344                 }
1345                 read_unlock_bh(&idev->lock);
1346         }
1347         rcu_read_unlock();
1348         return onlink;
1349 }
1350
1351 EXPORT_SYMBOL(ipv6_chk_prefix);
1352
1353 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1354                                      struct net_device *dev, int strict)
1355 {
1356         struct inet6_ifaddr *ifp, *result = NULL;
1357         unsigned int hash = ipv6_addr_hash(addr);
1358         struct hlist_node *node;
1359
1360         rcu_read_lock_bh();
1361         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1362                 if (!net_eq(dev_net(ifp->idev->dev), net))
1363                         continue;
1364                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1365                         if (dev == NULL || ifp->idev->dev == dev ||
1366                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1367                                 result = ifp;
1368                                 in6_ifa_hold(ifp);
1369                                 break;
1370                         }
1371                 }
1372         }
1373         rcu_read_unlock_bh();
1374
1375         return result;
1376 }
1377
1378 /* Gets referenced address, destroys ifaddr */
1379
1380 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1381 {
1382         if (ifp->flags&IFA_F_PERMANENT) {
1383                 spin_lock_bh(&ifp->lock);
1384                 addrconf_del_timer(ifp);
1385                 ifp->flags |= IFA_F_TENTATIVE;
1386                 if (dad_failed)
1387                         ifp->flags |= IFA_F_DADFAILED;
1388                 spin_unlock_bh(&ifp->lock);
1389                 if (dad_failed)
1390                         ipv6_ifa_notify(0, ifp);
1391                 in6_ifa_put(ifp);
1392 #ifdef CONFIG_IPV6_PRIVACY
1393         } else if (ifp->flags&IFA_F_TEMPORARY) {
1394                 struct inet6_ifaddr *ifpub;
1395                 spin_lock_bh(&ifp->lock);
1396                 ifpub = ifp->ifpub;
1397                 if (ifpub) {
1398                         in6_ifa_hold(ifpub);
1399                         spin_unlock_bh(&ifp->lock);
1400                         ipv6_create_tempaddr(ifpub, ifp);
1401                         in6_ifa_put(ifpub);
1402                 } else {
1403                         spin_unlock_bh(&ifp->lock);
1404                 }
1405                 ipv6_del_addr(ifp);
1406 #endif
1407         } else
1408                 ipv6_del_addr(ifp);
1409 }
1410
1411 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1412 {
1413         int err = -ENOENT;
1414
1415         spin_lock(&ifp->state_lock);
1416         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1417                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1418                 err = 0;
1419         }
1420         spin_unlock(&ifp->state_lock);
1421
1422         return err;
1423 }
1424
1425 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1426 {
1427         struct inet6_dev *idev = ifp->idev;
1428
1429         if (addrconf_dad_end(ifp))
1430                 return;
1431
1432         if (net_ratelimit())
1433                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1434                         ifp->idev->dev->name, &ifp->addr);
1435
1436         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1437                 struct in6_addr addr;
1438
1439                 addr.s6_addr32[0] = htonl(0xfe800000);
1440                 addr.s6_addr32[1] = 0;
1441
1442                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1443                     ipv6_addr_equal(&ifp->addr, &addr)) {
1444                         /* DAD failed for link-local based on MAC address */
1445                         idev->cnf.disable_ipv6 = 1;
1446
1447                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1448                                 ifp->idev->dev->name);
1449                 }
1450         }
1451
1452         addrconf_dad_stop(ifp, 1);
1453 }
1454
1455 /* Join to solicited addr multicast group. */
1456
1457 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1458 {
1459         struct in6_addr maddr;
1460
1461         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1462                 return;
1463
1464         addrconf_addr_solict_mult(addr, &maddr);
1465         ipv6_dev_mc_inc(dev, &maddr);
1466 }
1467
1468 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1469 {
1470         struct in6_addr maddr;
1471
1472         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1473                 return;
1474
1475         addrconf_addr_solict_mult(addr, &maddr);
1476         __ipv6_dev_mc_dec(idev, &maddr);
1477 }
1478
1479 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1480 {
1481         struct in6_addr addr;
1482         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483         if (ipv6_addr_any(&addr))
1484                 return;
1485         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1486 }
1487
1488 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1489 {
1490         struct in6_addr addr;
1491         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1492         if (ipv6_addr_any(&addr))
1493                 return;
1494         __ipv6_dev_ac_dec(ifp->idev, &addr);
1495 }
1496
1497 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1498 {
1499         if (dev->addr_len != ETH_ALEN)
1500                 return -1;
1501         memcpy(eui, dev->dev_addr, 3);
1502         memcpy(eui + 5, dev->dev_addr + 3, 3);
1503
1504         /*
1505          * The zSeries OSA network cards can be shared among various
1506          * OS instances, but the OSA cards have only one MAC address.
1507          * This leads to duplicate address conflicts in conjunction
1508          * with IPv6 if more than one instance uses the same card.
1509          *
1510          * The driver for these cards can deliver a unique 16-bit
1511          * identifier for each instance sharing the same card.  It is
1512          * placed instead of 0xFFFE in the interface identifier.  The
1513          * "u" bit of the interface identifier is not inverted in this
1514          * case.  Hence the resulting interface identifier has local
1515          * scope according to RFC2373.
1516          */
1517         if (dev->dev_id) {
1518                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1519                 eui[4] = dev->dev_id & 0xFF;
1520         } else {
1521                 eui[3] = 0xFF;
1522                 eui[4] = 0xFE;
1523                 eui[0] ^= 2;
1524         }
1525         return 0;
1526 }
1527
1528 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1529 {
1530         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1531         if (dev->addr_len != ARCNET_ALEN)
1532                 return -1;
1533         memset(eui, 0, 7);
1534         eui[7] = *(u8*)dev->dev_addr;
1535         return 0;
1536 }
1537
1538 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1539 {
1540         if (dev->addr_len != INFINIBAND_ALEN)
1541                 return -1;
1542         memcpy(eui, dev->dev_addr + 12, 8);
1543         eui[0] |= 2;
1544         return 0;
1545 }
1546
1547 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1548 {
1549         if (addr == 0)
1550                 return -1;
1551         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1552                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1553                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1554                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1555                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1556                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1557         eui[1] = 0;
1558         eui[2] = 0x5E;
1559         eui[3] = 0xFE;
1560         memcpy(eui + 4, &addr, 4);
1561         return 0;
1562 }
1563 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1564
1565 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1566 {
1567         if (dev->priv_flags & IFF_ISATAP)
1568                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569         return -1;
1570 }
1571
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574         switch (dev->type) {
1575         case ARPHRD_ETHER:
1576         case ARPHRD_FDDI:
1577         case ARPHRD_IEEE802_TR:
1578                 return addrconf_ifid_eui48(eui, dev);
1579         case ARPHRD_ARCNET:
1580                 return addrconf_ifid_arcnet(eui, dev);
1581         case ARPHRD_INFINIBAND:
1582                 return addrconf_ifid_infiniband(eui, dev);
1583         case ARPHRD_SIT:
1584                 return addrconf_ifid_sit(eui, dev);
1585         }
1586         return -1;
1587 }
1588
1589 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1590 {
1591         int err = -1;
1592         struct inet6_ifaddr *ifp;
1593
1594         read_lock_bh(&idev->lock);
1595         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1596                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1597                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1598                         err = 0;
1599                         break;
1600                 }
1601         }
1602         read_unlock_bh(&idev->lock);
1603         return err;
1604 }
1605
1606 #ifdef CONFIG_IPV6_PRIVACY
1607 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1608 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1609 {
1610 regen:
1611         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1612         idev->rndid[0] &= ~0x02;
1613
1614         /*
1615          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1616          * check if generated address is not inappropriate
1617          *
1618          *  - Reserved subnet anycast (RFC 2526)
1619          *      11111101 11....11 1xxxxxxx
1620          *  - ISATAP (RFC4214) 6.1
1621          *      00-00-5E-FE-xx-xx-xx-xx
1622          *  - value 0
1623          *  - XXX: already assigned to an address on the device
1624          */
1625         if (idev->rndid[0] == 0xfd &&
1626             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1627             (idev->rndid[7]&0x80))
1628                 goto regen;
1629         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1630                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1631                         goto regen;
1632                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1633                         goto regen;
1634         }
1635
1636         return 0;
1637 }
1638
1639 static void ipv6_regen_rndid(unsigned long data)
1640 {
1641         struct inet6_dev *idev = (struct inet6_dev *) data;
1642         unsigned long expires;
1643
1644         rcu_read_lock_bh();
1645         write_lock_bh(&idev->lock);
1646
1647         if (idev->dead)
1648                 goto out;
1649
1650         if (__ipv6_regen_rndid(idev) < 0)
1651                 goto out;
1652
1653         expires = jiffies +
1654                 idev->cnf.temp_prefered_lft * HZ -
1655                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1656                 idev->cnf.max_desync_factor * HZ;
1657         if (time_before(expires, jiffies)) {
1658                 printk(KERN_WARNING
1659                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1660                         idev->dev->name);
1661                 goto out;
1662         }
1663
1664         if (!mod_timer(&idev->regen_timer, expires))
1665                 in6_dev_hold(idev);
1666
1667 out:
1668         write_unlock_bh(&idev->lock);
1669         rcu_read_unlock_bh();
1670         in6_dev_put(idev);
1671 }
1672
1673 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1674         int ret = 0;
1675
1676         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1677                 ret = __ipv6_regen_rndid(idev);
1678         return ret;
1679 }
1680 #endif
1681
1682 /*
1683  *      Add prefix route.
1684  */
1685
1686 static void
1687 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1688                       unsigned long expires, u32 flags)
1689 {
1690         struct fib6_config cfg = {
1691                 .fc_table = RT6_TABLE_PREFIX,
1692                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1693                 .fc_ifindex = dev->ifindex,
1694                 .fc_expires = expires,
1695                 .fc_dst_len = plen,
1696                 .fc_flags = RTF_UP | flags,
1697                 .fc_nlinfo.nl_net = dev_net(dev),
1698                 .fc_protocol = RTPROT_KERNEL,
1699         };
1700
1701         ipv6_addr_copy(&cfg.fc_dst, pfx);
1702
1703         /* Prevent useless cloning on PtP SIT.
1704            This thing is done here expecting that the whole
1705            class of non-broadcast devices need not cloning.
1706          */
1707 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1708         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1709                 cfg.fc_flags |= RTF_NONEXTHOP;
1710 #endif
1711
1712         ip6_route_add(&cfg);
1713 }
1714
1715 /* Create "default" multicast route to the interface */
1716
1717 static void addrconf_add_mroute(struct net_device *dev)
1718 {
1719         struct fib6_config cfg = {
1720                 .fc_table = RT6_TABLE_LOCAL,
1721                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1722                 .fc_ifindex = dev->ifindex,
1723                 .fc_dst_len = 8,
1724                 .fc_flags = RTF_UP,
1725                 .fc_nlinfo.nl_net = dev_net(dev),
1726         };
1727
1728         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1729
1730         ip6_route_add(&cfg);
1731 }
1732
1733 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1734 static void sit_route_add(struct net_device *dev)
1735 {
1736         struct fib6_config cfg = {
1737                 .fc_table = RT6_TABLE_MAIN,
1738                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1739                 .fc_ifindex = dev->ifindex,
1740                 .fc_dst_len = 96,
1741                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1742                 .fc_nlinfo.nl_net = dev_net(dev),
1743         };
1744
1745         /* prefix length - 96 bits "::d.d.d.d" */
1746         ip6_route_add(&cfg);
1747 }
1748 #endif
1749
1750 static void addrconf_add_lroute(struct net_device *dev)
1751 {
1752         struct in6_addr addr;
1753
1754         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1755         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1756 }
1757
1758 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1759 {
1760         struct inet6_dev *idev;
1761
1762         ASSERT_RTNL();
1763
1764         idev = ipv6_find_idev(dev);
1765         if (!idev)
1766                 return NULL;
1767
1768         /* Add default multicast route */
1769         addrconf_add_mroute(dev);
1770
1771         /* Add link local route */
1772         addrconf_add_lroute(dev);
1773         return idev;
1774 }
1775
1776 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1777 {
1778         struct prefix_info *pinfo;
1779         __u32 valid_lft;
1780         __u32 prefered_lft;
1781         int addr_type;
1782         struct inet6_dev *in6_dev;
1783         struct net *net = dev_net(dev);
1784
1785         pinfo = (struct prefix_info *) opt;
1786
1787         if (len < sizeof(struct prefix_info)) {
1788                 ADBG(("addrconf: prefix option too short\n"));
1789                 return;
1790         }
1791
1792         /*
1793          *      Validation checks ([ADDRCONF], page 19)
1794          */
1795
1796         addr_type = ipv6_addr_type(&pinfo->prefix);
1797
1798         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1799                 return;
1800
1801         valid_lft = ntohl(pinfo->valid);
1802         prefered_lft = ntohl(pinfo->prefered);
1803
1804         if (prefered_lft > valid_lft) {
1805                 if (net_ratelimit())
1806                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1807                 return;
1808         }
1809
1810         in6_dev = in6_dev_get(dev);
1811
1812         if (in6_dev == NULL) {
1813                 if (net_ratelimit())
1814                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1815                 return;
1816         }
1817
1818         /*
1819          *      Two things going on here:
1820          *      1) Add routes for on-link prefixes
1821          *      2) Configure prefixes with the auto flag set
1822          */
1823
1824         if (pinfo->onlink) {
1825                 struct rt6_info *rt;
1826                 unsigned long rt_expires;
1827
1828                 /* Avoid arithmetic overflow. Really, we could
1829                  * save rt_expires in seconds, likely valid_lft,
1830                  * but it would require division in fib gc, that it
1831                  * not good.
1832                  */
1833                 if (HZ > USER_HZ)
1834                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1835                 else
1836                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1837
1838                 if (addrconf_finite_timeout(rt_expires))
1839                         rt_expires *= HZ;
1840
1841                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1842                                 dev->ifindex, 1);
1843
1844                 if (rt && addrconf_is_prefix_route(rt)) {
1845                         /* Autoconf prefix route */
1846                         if (valid_lft == 0) {
1847                                 ip6_del_rt(rt);
1848                                 rt = NULL;
1849                         } else if (addrconf_finite_timeout(rt_expires)) {
1850                                 /* not infinity */
1851                                 rt->rt6i_expires = jiffies + rt_expires;
1852                                 rt->rt6i_flags |= RTF_EXPIRES;
1853                         } else {
1854                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1855                                 rt->rt6i_expires = 0;
1856                         }
1857                 } else if (valid_lft) {
1858                         clock_t expires = 0;
1859                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1860                         if (addrconf_finite_timeout(rt_expires)) {
1861                                 /* not infinity */
1862                                 flags |= RTF_EXPIRES;
1863                                 expires = jiffies_to_clock_t(rt_expires);
1864                         }
1865                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1866                                               dev, expires, flags);
1867                 }
1868                 if (rt)
1869                         dst_release(&rt->dst);
1870         }
1871
1872         /* Try to figure out our local address for this prefix */
1873
1874         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1875                 struct inet6_ifaddr * ifp;
1876                 struct in6_addr addr;
1877                 int create = 0, update_lft = 0;
1878
1879                 if (pinfo->prefix_len == 64) {
1880                         memcpy(&addr, &pinfo->prefix, 8);
1881                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1882                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1883                                 in6_dev_put(in6_dev);
1884                                 return;
1885                         }
1886                         goto ok;
1887                 }
1888                 if (net_ratelimit())
1889                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1890                                pinfo->prefix_len);
1891                 in6_dev_put(in6_dev);
1892                 return;
1893
1894 ok:
1895
1896                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1897
1898                 if (ifp == NULL && valid_lft) {
1899                         int max_addresses = in6_dev->cnf.max_addresses;
1900                         u32 addr_flags = 0;
1901
1902 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1903                         if (in6_dev->cnf.optimistic_dad &&
1904                             !net->ipv6.devconf_all->forwarding)
1905                                 addr_flags = IFA_F_OPTIMISTIC;
1906 #endif
1907
1908                         /* Do not allow to create too much of autoconfigured
1909                          * addresses; this would be too easy way to crash kernel.
1910                          */
1911                         if (!max_addresses ||
1912                             ipv6_count_addresses(in6_dev) < max_addresses)
1913                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1914                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1915                                                     addr_flags);
1916
1917                         if (!ifp || IS_ERR(ifp)) {
1918                                 in6_dev_put(in6_dev);
1919                                 return;
1920                         }
1921
1922                         update_lft = create = 1;
1923                         ifp->cstamp = jiffies;
1924                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1925                 }
1926
1927                 if (ifp) {
1928                         int flags;
1929                         unsigned long now;
1930 #ifdef CONFIG_IPV6_PRIVACY
1931                         struct inet6_ifaddr *ift;
1932 #endif
1933                         u32 stored_lft;
1934
1935                         /* update lifetime (RFC2462 5.5.3 e) */
1936                         spin_lock(&ifp->lock);
1937                         now = jiffies;
1938                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1939                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1940                         else
1941                                 stored_lft = 0;
1942                         if (!update_lft && stored_lft) {
1943                                 if (valid_lft > MIN_VALID_LIFETIME ||
1944                                     valid_lft > stored_lft)
1945                                         update_lft = 1;
1946                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1947                                         /* valid_lft <= stored_lft is always true */
1948                                         /*
1949                                          * RFC 4862 Section 5.5.3e:
1950                                          * "Note that the preferred lifetime of
1951                                          *  the corresponding address is always
1952                                          *  reset to the Preferred Lifetime in
1953                                          *  the received Prefix Information
1954                                          *  option, regardless of whether the
1955                                          *  valid lifetime is also reset or
1956                                          *  ignored."
1957                                          *
1958                                          *  So if the preferred lifetime in
1959                                          *  this advertisement is different
1960                                          *  than what we have stored, but the
1961                                          *  valid lifetime is invalid, just
1962                                          *  reset prefered_lft.
1963                                          *
1964                                          *  We must set the valid lifetime
1965                                          *  to the stored lifetime since we'll
1966                                          *  be updating the timestamp below,
1967                                          *  else we'll set it back to the
1968                                          *  minumum.
1969                                          */
1970                                         if (prefered_lft != ifp->prefered_lft) {
1971                                                 valid_lft = stored_lft;
1972                                                 update_lft = 1;
1973                                         }
1974                                 } else {
1975                                         valid_lft = MIN_VALID_LIFETIME;
1976                                         if (valid_lft < prefered_lft)
1977                                                 prefered_lft = valid_lft;
1978                                         update_lft = 1;
1979                                 }
1980                         }
1981
1982                         if (update_lft) {
1983                                 ifp->valid_lft = valid_lft;
1984                                 ifp->prefered_lft = prefered_lft;
1985                                 ifp->tstamp = now;
1986                                 flags = ifp->flags;
1987                                 ifp->flags &= ~IFA_F_DEPRECATED;
1988                                 spin_unlock(&ifp->lock);
1989
1990                                 if (!(flags&IFA_F_TENTATIVE))
1991                                         ipv6_ifa_notify(0, ifp);
1992                         } else
1993                                 spin_unlock(&ifp->lock);
1994
1995 #ifdef CONFIG_IPV6_PRIVACY
1996                         read_lock_bh(&in6_dev->lock);
1997                         /* update all temporary addresses in the list */
1998                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1999                                 /*
2000                                  * When adjusting the lifetimes of an existing
2001                                  * temporary address, only lower the lifetimes.
2002                                  * Implementations must not increase the
2003                                  * lifetimes of an existing temporary address
2004                                  * when processing a Prefix Information Option.
2005                                  */
2006                                 if (ifp != ift->ifpub)
2007                                         continue;
2008
2009                                 spin_lock(&ift->lock);
2010                                 flags = ift->flags;
2011                                 if (ift->valid_lft > valid_lft &&
2012                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2013                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2014                                 if (ift->prefered_lft > prefered_lft &&
2015                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2016                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2017                                 spin_unlock(&ift->lock);
2018                                 if (!(flags&IFA_F_TENTATIVE))
2019                                         ipv6_ifa_notify(0, ift);
2020                         }
2021
2022                         if (create && in6_dev->cnf.use_tempaddr > 0) {
2023                                 /*
2024                                  * When a new public address is created as described in [ADDRCONF],
2025                                  * also create a new temporary address.
2026                                  */
2027                                 read_unlock_bh(&in6_dev->lock);
2028                                 ipv6_create_tempaddr(ifp, NULL);
2029                         } else {
2030                                 read_unlock_bh(&in6_dev->lock);
2031                         }
2032 #endif
2033                         in6_ifa_put(ifp);
2034                         addrconf_verify(0);
2035                 }
2036         }
2037         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2038         in6_dev_put(in6_dev);
2039 }
2040
2041 /*
2042  *      Set destination address.
2043  *      Special case for SIT interfaces where we create a new "virtual"
2044  *      device.
2045  */
2046 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2047 {
2048         struct in6_ifreq ireq;
2049         struct net_device *dev;
2050         int err = -EINVAL;
2051
2052         rtnl_lock();
2053
2054         err = -EFAULT;
2055         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2056                 goto err_exit;
2057
2058         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2059
2060         err = -ENODEV;
2061         if (dev == NULL)
2062                 goto err_exit;
2063
2064 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2065         if (dev->type == ARPHRD_SIT) {
2066                 const struct net_device_ops *ops = dev->netdev_ops;
2067                 struct ifreq ifr;
2068                 struct ip_tunnel_parm p;
2069
2070                 err = -EADDRNOTAVAIL;
2071                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2072                         goto err_exit;
2073
2074                 memset(&p, 0, sizeof(p));
2075                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2076                 p.iph.saddr = 0;
2077                 p.iph.version = 4;
2078                 p.iph.ihl = 5;
2079                 p.iph.protocol = IPPROTO_IPV6;
2080                 p.iph.ttl = 64;
2081                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2082
2083                 if (ops->ndo_do_ioctl) {
2084                         mm_segment_t oldfs = get_fs();
2085
2086                         set_fs(KERNEL_DS);
2087                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2088                         set_fs(oldfs);
2089                 } else
2090                         err = -EOPNOTSUPP;
2091
2092                 if (err == 0) {
2093                         err = -ENOBUFS;
2094                         dev = __dev_get_by_name(net, p.name);
2095                         if (!dev)
2096                                 goto err_exit;
2097                         err = dev_open(dev);
2098                 }
2099         }
2100 #endif
2101
2102 err_exit:
2103         rtnl_unlock();
2104         return err;
2105 }
2106
2107 /*
2108  *      Manual configuration of address on an interface
2109  */
2110 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2111                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2112                           __u32 valid_lft)
2113 {
2114         struct inet6_ifaddr *ifp;
2115         struct inet6_dev *idev;
2116         struct net_device *dev;
2117         int scope;
2118         u32 flags;
2119         clock_t expires;
2120         unsigned long timeout;
2121
2122         ASSERT_RTNL();
2123
2124         if (plen > 128)
2125                 return -EINVAL;
2126
2127         /* check the lifetime */
2128         if (!valid_lft || prefered_lft > valid_lft)
2129                 return -EINVAL;
2130
2131         dev = __dev_get_by_index(net, ifindex);
2132         if (!dev)
2133                 return -ENODEV;
2134
2135         if ((idev = addrconf_add_dev(dev)) == NULL)
2136                 return -ENOBUFS;
2137
2138         scope = ipv6_addr_scope(pfx);
2139
2140         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2141         if (addrconf_finite_timeout(timeout)) {
2142                 expires = jiffies_to_clock_t(timeout * HZ);
2143                 valid_lft = timeout;
2144                 flags = RTF_EXPIRES;
2145         } else {
2146                 expires = 0;
2147                 flags = 0;
2148                 ifa_flags |= IFA_F_PERMANENT;
2149         }
2150
2151         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2152         if (addrconf_finite_timeout(timeout)) {
2153                 if (timeout == 0)
2154                         ifa_flags |= IFA_F_DEPRECATED;
2155                 prefered_lft = timeout;
2156         }
2157
2158         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2159
2160         if (!IS_ERR(ifp)) {
2161                 spin_lock_bh(&ifp->lock);
2162                 ifp->valid_lft = valid_lft;
2163                 ifp->prefered_lft = prefered_lft;
2164                 ifp->tstamp = jiffies;
2165                 spin_unlock_bh(&ifp->lock);
2166
2167                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2168                                       expires, flags);
2169                 /*
2170                  * Note that section 3.1 of RFC 4429 indicates
2171                  * that the Optimistic flag should not be set for
2172                  * manually configured addresses
2173                  */
2174                 addrconf_dad_start(ifp, 0);
2175                 in6_ifa_put(ifp);
2176                 addrconf_verify(0);
2177                 return 0;
2178         }
2179
2180         return PTR_ERR(ifp);
2181 }
2182
2183 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2184                           unsigned int plen)
2185 {
2186         struct inet6_ifaddr *ifp;
2187         struct inet6_dev *idev;
2188         struct net_device *dev;
2189
2190         if (plen > 128)
2191                 return -EINVAL;
2192
2193         dev = __dev_get_by_index(net, ifindex);
2194         if (!dev)
2195                 return -ENODEV;
2196
2197         if ((idev = __in6_dev_get(dev)) == NULL)
2198                 return -ENXIO;
2199
2200         read_lock_bh(&idev->lock);
2201         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2202                 if (ifp->prefix_len == plen &&
2203                     ipv6_addr_equal(pfx, &ifp->addr)) {
2204                         in6_ifa_hold(ifp);
2205                         read_unlock_bh(&idev->lock);
2206
2207                         ipv6_del_addr(ifp);
2208
2209                         /* If the last address is deleted administratively,
2210                            disable IPv6 on this interface.
2211                          */
2212                         if (list_empty(&idev->addr_list))
2213                                 addrconf_ifdown(idev->dev, 1);
2214                         return 0;
2215                 }
2216         }
2217         read_unlock_bh(&idev->lock);
2218         return -EADDRNOTAVAIL;
2219 }
2220
2221
2222 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2223 {
2224         struct in6_ifreq ireq;
2225         int err;
2226
2227         if (!capable(CAP_NET_ADMIN))
2228                 return -EPERM;
2229
2230         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2231                 return -EFAULT;
2232
2233         rtnl_lock();
2234         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2235                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2236                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2237         rtnl_unlock();
2238         return err;
2239 }
2240
2241 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2242 {
2243         struct in6_ifreq ireq;
2244         int err;
2245
2246         if (!capable(CAP_NET_ADMIN))
2247                 return -EPERM;
2248
2249         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2250                 return -EFAULT;
2251
2252         rtnl_lock();
2253         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2254                              ireq.ifr6_prefixlen);
2255         rtnl_unlock();
2256         return err;
2257 }
2258
2259 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2260                      int plen, int scope)
2261 {
2262         struct inet6_ifaddr *ifp;
2263
2264         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2265         if (!IS_ERR(ifp)) {
2266                 spin_lock_bh(&ifp->lock);
2267                 ifp->flags &= ~IFA_F_TENTATIVE;
2268                 spin_unlock_bh(&ifp->lock);
2269                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2270                 in6_ifa_put(ifp);
2271         }
2272 }
2273
2274 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2275 static void sit_add_v4_addrs(struct inet6_dev *idev)
2276 {
2277         struct in6_addr addr;
2278         struct net_device *dev;
2279         struct net *net = dev_net(idev->dev);
2280         int scope;
2281
2282         ASSERT_RTNL();
2283
2284         memset(&addr, 0, sizeof(struct in6_addr));
2285         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2286
2287         if (idev->dev->flags&IFF_POINTOPOINT) {
2288                 addr.s6_addr32[0] = htonl(0xfe800000);
2289                 scope = IFA_LINK;
2290         } else {
2291                 scope = IPV6_ADDR_COMPATv4;
2292         }
2293
2294         if (addr.s6_addr32[3]) {
2295                 add_addr(idev, &addr, 128, scope);
2296                 return;
2297         }
2298
2299         for_each_netdev(net, dev) {
2300                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2301                 if (in_dev && (dev->flags & IFF_UP)) {
2302                         struct in_ifaddr * ifa;
2303
2304                         int flag = scope;
2305
2306                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2307                                 int plen;
2308
2309                                 addr.s6_addr32[3] = ifa->ifa_local;
2310
2311                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2312                                         continue;
2313                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2314                                         if (idev->dev->flags&IFF_POINTOPOINT)
2315                                                 continue;
2316                                         flag |= IFA_HOST;
2317                                 }
2318                                 if (idev->dev->flags&IFF_POINTOPOINT)
2319                                         plen = 64;
2320                                 else
2321                                         plen = 96;
2322
2323                                 add_addr(idev, &addr, plen, flag);
2324                         }
2325                 }
2326         }
2327 }
2328 #endif
2329
2330 static void init_loopback(struct net_device *dev)
2331 {
2332         struct inet6_dev  *idev;
2333
2334         /* ::1 */
2335
2336         ASSERT_RTNL();
2337
2338         if ((idev = ipv6_find_idev(dev)) == NULL) {
2339                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2340                 return;
2341         }
2342
2343         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2344 }
2345
2346 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2347 {
2348         struct inet6_ifaddr * ifp;
2349         u32 addr_flags = IFA_F_PERMANENT;
2350
2351 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2352         if (idev->cnf.optimistic_dad &&
2353             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2354                 addr_flags |= IFA_F_OPTIMISTIC;
2355 #endif
2356
2357
2358         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2359         if (!IS_ERR(ifp)) {
2360                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2361                 addrconf_dad_start(ifp, 0);
2362                 in6_ifa_put(ifp);
2363         }
2364 }
2365
2366 static void addrconf_dev_config(struct net_device *dev)
2367 {
2368         struct in6_addr addr;
2369         struct inet6_dev    * idev;
2370
2371         ASSERT_RTNL();
2372
2373         if ((dev->type != ARPHRD_ETHER) &&
2374             (dev->type != ARPHRD_FDDI) &&
2375             (dev->type != ARPHRD_IEEE802_TR) &&
2376             (dev->type != ARPHRD_ARCNET) &&
2377             (dev->type != ARPHRD_INFINIBAND)) {
2378                 /* Alas, we support only Ethernet autoconfiguration. */
2379                 return;
2380         }
2381
2382         idev = addrconf_add_dev(dev);
2383         if (idev == NULL)
2384                 return;
2385
2386         memset(&addr, 0, sizeof(struct in6_addr));
2387         addr.s6_addr32[0] = htonl(0xFE800000);
2388
2389         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2390                 addrconf_add_linklocal(idev, &addr);
2391 }
2392
2393 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2394 static void addrconf_sit_config(struct net_device *dev)
2395 {
2396         struct inet6_dev *idev;
2397
2398         ASSERT_RTNL();
2399
2400         /*
2401          * Configure the tunnel with one of our IPv4
2402          * addresses... we should configure all of
2403          * our v4 addrs in the tunnel
2404          */
2405
2406         if ((idev = ipv6_find_idev(dev)) == NULL) {
2407                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2408                 return;
2409         }
2410
2411         if (dev->priv_flags & IFF_ISATAP) {
2412                 struct in6_addr addr;
2413
2414                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2415                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2416                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2417                         addrconf_add_linklocal(idev, &addr);
2418                 return;
2419         }
2420
2421         sit_add_v4_addrs(idev);
2422
2423         if (dev->flags&IFF_POINTOPOINT) {
2424                 addrconf_add_mroute(dev);
2425                 addrconf_add_lroute(dev);
2426         } else
2427                 sit_route_add(dev);
2428 }
2429 #endif
2430
2431 static inline int
2432 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2433 {
2434         struct in6_addr lladdr;
2435
2436         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2437                 addrconf_add_linklocal(idev, &lladdr);
2438                 return 0;
2439         }
2440         return -1;
2441 }
2442
2443 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2444 {
2445         struct net_device *link_dev;
2446         struct net *net = dev_net(idev->dev);
2447
2448         /* first try to inherit the link-local address from the link device */
2449         if (idev->dev->iflink &&
2450             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2451                 if (!ipv6_inherit_linklocal(idev, link_dev))
2452                         return;
2453         }
2454         /* then try to inherit it from any device */
2455         for_each_netdev(net, link_dev) {
2456                 if (!ipv6_inherit_linklocal(idev, link_dev))
2457                         return;
2458         }
2459         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2460 }
2461
2462 /*
2463  * Autoconfigure tunnel with a link-local address so routing protocols,
2464  * DHCPv6, MLD etc. can be run over the virtual link
2465  */
2466
2467 static void addrconf_ip6_tnl_config(struct net_device *dev)
2468 {
2469         struct inet6_dev *idev;
2470
2471         ASSERT_RTNL();
2472
2473         idev = addrconf_add_dev(dev);
2474         if (!idev) {
2475                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2476                 return;
2477         }
2478         ip6_tnl_add_linklocal(idev);
2479 }
2480
2481 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2482                            void * data)
2483 {
2484         struct net_device *dev = (struct net_device *) data;
2485         struct inet6_dev *idev = __in6_dev_get(dev);
2486         int run_pending = 0;
2487         int err;
2488
2489         switch (event) {
2490         case NETDEV_REGISTER:
2491                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2492                         idev = ipv6_add_dev(dev);
2493                         if (!idev)
2494                                 return notifier_from_errno(-ENOMEM);
2495                 }
2496                 break;
2497
2498         case NETDEV_UP:
2499         case NETDEV_CHANGE:
2500                 if (dev->flags & IFF_SLAVE)
2501                         break;
2502
2503                 if (event == NETDEV_UP) {
2504                         if (!addrconf_qdisc_ok(dev)) {
2505                                 /* device is not ready yet. */
2506                                 printk(KERN_INFO
2507                                         "ADDRCONF(NETDEV_UP): %s: "
2508                                         "link is not ready\n",
2509                                         dev->name);
2510                                 break;
2511                         }
2512
2513                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2514                                 idev = ipv6_add_dev(dev);
2515
2516                         if (idev) {
2517                                 idev->if_flags |= IF_READY;
2518                                 run_pending = 1;
2519                         }
2520                 } else {
2521                         if (!addrconf_qdisc_ok(dev)) {
2522                                 /* device is still not ready. */
2523                                 break;
2524                         }
2525
2526                         if (idev) {
2527                                 if (idev->if_flags & IF_READY)
2528                                         /* device is already configured. */
2529                                         break;
2530                                 idev->if_flags |= IF_READY;
2531                         }
2532
2533                         printk(KERN_INFO
2534                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2535                                         "link becomes ready\n",
2536                                         dev->name);
2537
2538                         run_pending = 1;
2539                 }
2540
2541                 switch (dev->type) {
2542 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2543                 case ARPHRD_SIT:
2544                         addrconf_sit_config(dev);
2545                         break;
2546 #endif
2547                 case ARPHRD_TUNNEL6:
2548                         addrconf_ip6_tnl_config(dev);
2549                         break;
2550                 case ARPHRD_LOOPBACK:
2551                         init_loopback(dev);
2552                         break;
2553
2554                 default:
2555                         addrconf_dev_config(dev);
2556                         break;
2557                 }
2558
2559                 if (idev) {
2560                         if (run_pending)
2561                                 addrconf_dad_run(idev);
2562
2563                         /*
2564                          * If the MTU changed during the interface down,
2565                          * when the interface up, the changed MTU must be
2566                          * reflected in the idev as well as routers.
2567                          */
2568                         if (idev->cnf.mtu6 != dev->mtu &&
2569                             dev->mtu >= IPV6_MIN_MTU) {
2570                                 rt6_mtu_change(dev, dev->mtu);
2571                                 idev->cnf.mtu6 = dev->mtu;
2572                         }
2573                         idev->tstamp = jiffies;
2574                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2575
2576                         /*
2577                          * If the changed mtu during down is lower than
2578                          * IPV6_MIN_MTU stop IPv6 on this interface.
2579                          */
2580                         if (dev->mtu < IPV6_MIN_MTU)
2581                                 addrconf_ifdown(dev, 1);
2582                 }
2583                 break;
2584
2585         case NETDEV_CHANGEMTU:
2586                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2587                         rt6_mtu_change(dev, dev->mtu);
2588                         idev->cnf.mtu6 = dev->mtu;
2589                         break;
2590                 }
2591
2592                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2593                         idev = ipv6_add_dev(dev);
2594                         if (idev)
2595                                 break;
2596                 }
2597
2598                 /*
2599                  * MTU falled under IPV6_MIN_MTU.
2600                  * Stop IPv6 on this interface.
2601                  */
2602
2603         case NETDEV_DOWN:
2604         case NETDEV_UNREGISTER:
2605                 /*
2606                  *      Remove all addresses from this interface.
2607                  */
2608                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2609                 break;
2610
2611         case NETDEV_CHANGENAME:
2612                 if (idev) {
2613                         snmp6_unregister_dev(idev);
2614                         addrconf_sysctl_unregister(idev);
2615                         addrconf_sysctl_register(idev);
2616                         err = snmp6_register_dev(idev);
2617                         if (err)
2618                                 return notifier_from_errno(err);
2619                 }
2620                 break;
2621
2622         case NETDEV_PRE_TYPE_CHANGE:
2623         case NETDEV_POST_TYPE_CHANGE:
2624                 addrconf_type_change(dev, event);
2625                 break;
2626         }
2627
2628         return NOTIFY_OK;
2629 }
2630
2631 /*
2632  *      addrconf module should be notified of a device going up
2633  */
2634 static struct notifier_block ipv6_dev_notf = {
2635         .notifier_call = addrconf_notify,
2636 };
2637
2638 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2639 {
2640         struct inet6_dev *idev;
2641         ASSERT_RTNL();
2642
2643         idev = __in6_dev_get(dev);
2644
2645         if (event == NETDEV_POST_TYPE_CHANGE)
2646                 ipv6_mc_remap(idev);
2647         else if (event == NETDEV_PRE_TYPE_CHANGE)
2648                 ipv6_mc_unmap(idev);
2649 }
2650
2651 static int addrconf_ifdown(struct net_device *dev, int how)
2652 {
2653         struct net *net = dev_net(dev);
2654         struct inet6_dev *idev;
2655         struct inet6_ifaddr *ifa;
2656         LIST_HEAD(keep_list);
2657         int state;
2658
2659         ASSERT_RTNL();
2660
2661         rt6_ifdown(net, dev);
2662         neigh_ifdown(&nd_tbl, dev);
2663
2664         idev = __in6_dev_get(dev);
2665         if (idev == NULL)
2666                 return -ENODEV;
2667
2668         /*
2669          * Step 1: remove reference to ipv6 device from parent device.
2670          *         Do not dev_put!
2671          */
2672         if (how) {
2673                 idev->dead = 1;
2674
2675                 /* protected by rtnl_lock */
2676                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2677
2678                 /* Step 1.5: remove snmp6 entry */
2679                 snmp6_unregister_dev(idev);
2680
2681         }
2682
2683         write_lock_bh(&idev->lock);
2684
2685         /* Step 2: clear flags for stateless addrconf */
2686         if (!how)
2687                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2688
2689 #ifdef CONFIG_IPV6_PRIVACY
2690         if (how && del_timer(&idev->regen_timer))
2691                 in6_dev_put(idev);
2692
2693         /* Step 3: clear tempaddr list */
2694         while (!list_empty(&idev->tempaddr_list)) {
2695                 ifa = list_first_entry(&idev->tempaddr_list,
2696                                        struct inet6_ifaddr, tmp_list);
2697                 list_del(&ifa->tmp_list);
2698                 write_unlock_bh(&idev->lock);
2699                 spin_lock_bh(&ifa->lock);
2700
2701                 if (ifa->ifpub) {
2702                         in6_ifa_put(ifa->ifpub);
2703                         ifa->ifpub = NULL;
2704                 }
2705                 spin_unlock_bh(&ifa->lock);
2706                 in6_ifa_put(ifa);
2707                 write_lock_bh(&idev->lock);
2708         }
2709 #endif
2710
2711         while (!list_empty(&idev->addr_list)) {
2712                 ifa = list_first_entry(&idev->addr_list,
2713                                        struct inet6_ifaddr, if_list);
2714                 addrconf_del_timer(ifa);
2715
2716                 /* If just doing link down, and address is permanent
2717                    and not link-local, then retain it. */
2718                 if (!how &&
2719                     (ifa->flags&IFA_F_PERMANENT) &&
2720                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2721                         list_move_tail(&ifa->if_list, &keep_list);
2722
2723                         /* If not doing DAD on this address, just keep it. */
2724                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2725                             idev->cnf.accept_dad <= 0 ||
2726                             (ifa->flags & IFA_F_NODAD))
2727                                 continue;
2728
2729                         /* If it was tentative already, no need to notify */
2730                         if (ifa->flags & IFA_F_TENTATIVE)
2731                                 continue;
2732
2733                         /* Flag it for later restoration when link comes up */
2734                         ifa->flags |= IFA_F_TENTATIVE;
2735                         ifa->state = INET6_IFADDR_STATE_DAD;
2736
2737                         write_unlock_bh(&idev->lock);
2738
2739                         in6_ifa_hold(ifa);
2740                 } else {
2741                         list_del(&ifa->if_list);
2742
2743                         /* clear hash table */
2744                         spin_lock_bh(&addrconf_hash_lock);
2745                         hlist_del_init_rcu(&ifa->addr_lst);
2746                         spin_unlock_bh(&addrconf_hash_lock);
2747
2748                         write_unlock_bh(&idev->lock);
2749                         spin_lock_bh(&ifa->state_lock);
2750                         state = ifa->state;
2751                         ifa->state = INET6_IFADDR_STATE_DEAD;
2752                         spin_unlock_bh(&ifa->state_lock);
2753
2754                         if (state == INET6_IFADDR_STATE_DEAD)
2755                                 goto put_ifa;
2756                 }
2757
2758                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2759                 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2760                         atomic_notifier_call_chain(&inet6addr_chain,
2761                                                    NETDEV_DOWN, ifa);
2762
2763 put_ifa:
2764                 in6_ifa_put(ifa);
2765
2766                 write_lock_bh(&idev->lock);
2767         }
2768
2769         list_splice(&keep_list, &idev->addr_list);
2770
2771         write_unlock_bh(&idev->lock);
2772
2773         /* Step 5: Discard multicast list */
2774         if (how)
2775                 ipv6_mc_destroy_dev(idev);
2776         else
2777                 ipv6_mc_down(idev);
2778
2779         idev->tstamp = jiffies;
2780
2781         /* Last: Shot the device (if unregistered) */
2782         if (how) {
2783                 addrconf_sysctl_unregister(idev);
2784                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2785                 neigh_ifdown(&nd_tbl, dev);
2786                 in6_dev_put(idev);
2787         }
2788         return 0;
2789 }
2790
2791 static void addrconf_rs_timer(unsigned long data)
2792 {
2793         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2794         struct inet6_dev *idev = ifp->idev;
2795
2796         read_lock(&idev->lock);
2797         if (idev->dead || !(idev->if_flags & IF_READY))
2798                 goto out;
2799
2800         if (idev->cnf.forwarding)
2801                 goto out;
2802
2803         /* Announcement received after solicitation was sent */
2804         if (idev->if_flags & IF_RA_RCVD)
2805                 goto out;
2806
2807         spin_lock(&ifp->lock);
2808         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2809                 /* The wait after the last probe can be shorter */
2810                 addrconf_mod_timer(ifp, AC_RS,
2811                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2812                                    idev->cnf.rtr_solicit_delay :
2813                                    idev->cnf.rtr_solicit_interval);
2814                 spin_unlock(&ifp->lock);
2815
2816                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2817         } else {
2818                 spin_unlock(&ifp->lock);
2819                 /*
2820                  * Note: we do not support deprecated "all on-link"
2821                  * assumption any longer.
2822                  */
2823                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2824                        idev->dev->name);
2825         }
2826
2827 out:
2828         read_unlock(&idev->lock);
2829         in6_ifa_put(ifp);
2830 }
2831
2832 /*
2833  *      Duplicate Address Detection
2834  */
2835 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2836 {
2837         unsigned long rand_num;
2838         struct inet6_dev *idev = ifp->idev;
2839
2840         if (ifp->flags & IFA_F_OPTIMISTIC)
2841                 rand_num = 0;
2842         else
2843                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2844
2845         ifp->probes = idev->cnf.dad_transmits;
2846         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2847 }
2848
2849 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2850 {
2851         struct inet6_dev *idev = ifp->idev;
2852         struct net_device *dev = idev->dev;
2853
2854         addrconf_join_solict(dev, &ifp->addr);
2855
2856         net_srandom(ifp->addr.s6_addr32[3]);
2857
2858         read_lock_bh(&idev->lock);
2859         spin_lock(&ifp->lock);
2860         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2861                 goto out;
2862
2863         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2864             idev->cnf.accept_dad < 1 ||
2865             !(ifp->flags&IFA_F_TENTATIVE) ||
2866             ifp->flags & IFA_F_NODAD) {
2867                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2868                 spin_unlock(&ifp->lock);
2869                 read_unlock_bh(&idev->lock);
2870
2871                 addrconf_dad_completed(ifp);
2872                 return;
2873         }
2874
2875         if (!(idev->if_flags & IF_READY)) {
2876                 spin_unlock(&ifp->lock);
2877                 read_unlock_bh(&idev->lock);
2878                 /*
2879                  * If the device is not ready:
2880                  * - keep it tentative if it is a permanent address.
2881                  * - otherwise, kill it.
2882                  */
2883                 in6_ifa_hold(ifp);
2884                 addrconf_dad_stop(ifp, 0);
2885                 return;
2886         }
2887
2888         /*
2889          * Optimistic nodes can start receiving
2890          * Frames right away
2891          */
2892         if (ifp->flags & IFA_F_OPTIMISTIC)
2893                 ip6_ins_rt(ifp->rt);
2894
2895         addrconf_dad_kick(ifp);
2896 out:
2897         spin_unlock(&ifp->lock);
2898         read_unlock_bh(&idev->lock);
2899 }
2900
2901 static void addrconf_dad_timer(unsigned long data)
2902 {
2903         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2904         struct inet6_dev *idev = ifp->idev;
2905         struct in6_addr mcaddr;
2906
2907         if (!ifp->probes && addrconf_dad_end(ifp))
2908                 goto out;
2909
2910         read_lock(&idev->lock);
2911         if (idev->dead || !(idev->if_flags & IF_READY)) {
2912                 read_unlock(&idev->lock);
2913                 goto out;
2914         }
2915
2916         spin_lock(&ifp->lock);
2917         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2918                 spin_unlock(&ifp->lock);
2919                 read_unlock(&idev->lock);
2920                 goto out;
2921         }
2922
2923         if (ifp->probes == 0) {
2924                 /*
2925                  * DAD was successful
2926                  */
2927
2928                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2929                 spin_unlock(&ifp->lock);
2930                 read_unlock(&idev->lock);
2931
2932                 addrconf_dad_completed(ifp);
2933
2934                 goto out;
2935         }
2936
2937         ifp->probes--;
2938         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2939         spin_unlock(&ifp->lock);
2940         read_unlock(&idev->lock);
2941
2942         /* send a neighbour solicitation for our addr */
2943         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2944         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2945 out:
2946         in6_ifa_put(ifp);
2947 }
2948
2949 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2950 {
2951         struct net_device *dev = ifp->idev->dev;
2952
2953         /*
2954          *      Configure the address for reception. Now it is valid.
2955          */
2956
2957         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2958
2959         /* If added prefix is link local and forwarding is off,
2960            start sending router solicitations.
2961          */
2962
2963         if (ifp->idev->cnf.forwarding == 0 &&
2964             ifp->idev->cnf.rtr_solicits > 0 &&
2965             (dev->flags&IFF_LOOPBACK) == 0 &&
2966             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2967                 /*
2968                  *      If a host as already performed a random delay
2969                  *      [...] as part of DAD [...] there is no need
2970                  *      to delay again before sending the first RS
2971                  */
2972                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2973
2974                 spin_lock_bh(&ifp->lock);
2975                 ifp->probes = 1;
2976                 ifp->idev->if_flags |= IF_RS_SENT;
2977                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2978                 spin_unlock_bh(&ifp->lock);
2979         }
2980 }
2981
2982 static void addrconf_dad_run(struct inet6_dev *idev)
2983 {
2984         struct inet6_ifaddr *ifp;
2985
2986         read_lock_bh(&idev->lock);
2987         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2988                 spin_lock(&ifp->lock);
2989                 if (ifp->flags & IFA_F_TENTATIVE &&
2990                     ifp->state == INET6_IFADDR_STATE_DAD)
2991                         addrconf_dad_kick(ifp);
2992                 spin_unlock(&ifp->lock);
2993         }
2994         read_unlock_bh(&idev->lock);
2995 }
2996
2997 #ifdef CONFIG_PROC_FS
2998 struct if6_iter_state {
2999         struct seq_net_private p;
3000         int bucket;
3001 };
3002
3003 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3004 {
3005         struct inet6_ifaddr *ifa = NULL;
3006         struct if6_iter_state *state = seq->private;
3007         struct net *net = seq_file_net(seq);
3008
3009         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3010                 struct hlist_node *n;
3011                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3012                                          addr_lst)
3013                         if (net_eq(dev_net(ifa->idev->dev), net))
3014                                 return ifa;
3015         }
3016         return NULL;
3017 }
3018
3019 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3020                                          struct inet6_ifaddr *ifa)
3021 {
3022         struct if6_iter_state *state = seq->private;
3023         struct net *net = seq_file_net(seq);
3024         struct hlist_node *n = &ifa->addr_lst;
3025
3026         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3027                 if (net_eq(dev_net(ifa->idev->dev), net))
3028                         return ifa;
3029
3030         while (++state->bucket < IN6_ADDR_HSIZE) {
3031                 hlist_for_each_entry_rcu_bh(ifa, n,
3032                                      &inet6_addr_lst[state->bucket], addr_lst) {
3033                         if (net_eq(dev_net(ifa->idev->dev), net))
3034                                 return ifa;
3035                 }
3036         }
3037
3038         return NULL;
3039 }
3040
3041 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3042 {
3043         struct inet6_ifaddr *ifa = if6_get_first(seq);
3044
3045         if (ifa)
3046                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3047                         --pos;
3048         return pos ? NULL : ifa;
3049 }
3050
3051 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3052         __acquires(rcu_bh)
3053 {
3054         rcu_read_lock_bh();
3055         return if6_get_idx(seq, *pos);
3056 }
3057
3058 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3059 {
3060         struct inet6_ifaddr *ifa;
3061
3062         ifa = if6_get_next(seq, v);
3063         ++*pos;
3064         return ifa;
3065 }
3066
3067 static void if6_seq_stop(struct seq_file *seq, void *v)
3068         __releases(rcu_bh)
3069 {
3070         rcu_read_unlock_bh();
3071 }
3072
3073 static int if6_seq_show(struct seq_file *seq, void *v)
3074 {
3075         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3076         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3077                    &ifp->addr,
3078                    ifp->idev->dev->ifindex,
3079                    ifp->prefix_len,
3080                    ifp->scope,
3081                    ifp->flags,
3082                    ifp->idev->dev->name);
3083         return 0;
3084 }
3085
3086 static const struct seq_operations if6_seq_ops = {
3087         .start  = if6_seq_start,
3088         .next   = if6_seq_next,
3089         .show   = if6_seq_show,
3090         .stop   = if6_seq_stop,
3091 };
3092
3093 static int if6_seq_open(struct inode *inode, struct file *file)
3094 {
3095         return seq_open_net(inode, file, &if6_seq_ops,
3096                             sizeof(struct if6_iter_state));
3097 }
3098
3099 static const struct file_operations if6_fops = {
3100         .owner          = THIS_MODULE,
3101         .open           = if6_seq_open,
3102         .read           = seq_read,
3103         .llseek         = seq_lseek,
3104         .release        = seq_release_net,
3105 };
3106
3107 static int __net_init if6_proc_net_init(struct net *net)
3108 {
3109         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3110                 return -ENOMEM;
3111         return 0;
3112 }
3113
3114 static void __net_exit if6_proc_net_exit(struct net *net)
3115 {
3116        proc_net_remove(net, "if_inet6");
3117 }
3118
3119 static struct pernet_operations if6_proc_net_ops = {
3120        .init = if6_proc_net_init,
3121        .exit = if6_proc_net_exit,
3122 };
3123
3124 int __init if6_proc_init(void)
3125 {
3126         return register_pernet_subsys(&if6_proc_net_ops);
3127 }
3128
3129 void if6_proc_exit(void)
3130 {
3131         unregister_pernet_subsys(&if6_proc_net_ops);
3132 }
3133 #endif  /* CONFIG_PROC_FS */
3134
3135 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3136 /* Check if address is a home address configured on any interface. */
3137 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3138 {
3139         int ret = 0;
3140         struct inet6_ifaddr *ifp = NULL;
3141         struct hlist_node *n;
3142         unsigned int hash = ipv6_addr_hash(addr);
3143
3144         rcu_read_lock_bh();
3145         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3146                 if (!net_eq(dev_net(ifp->idev->dev), net))
3147                         continue;
3148                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3149                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3150                         ret = 1;
3151                         break;
3152                 }
3153         }
3154         rcu_read_unlock_bh();
3155         return ret;
3156 }
3157 #endif
3158
3159 /*
3160  *      Periodic address status verification
3161  */
3162
3163 static void addrconf_verify(unsigned long foo)
3164 {
3165         unsigned long now, next, next_sec, next_sched;
3166         struct inet6_ifaddr *ifp;
3167         struct hlist_node *node;
3168         int i;
3169
3170         rcu_read_lock_bh();
3171         spin_lock(&addrconf_verify_lock);
3172         now = jiffies;
3173         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3174
3175         del_timer(&addr_chk_timer);
3176
3177         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3178 restart:
3179                 hlist_for_each_entry_rcu_bh(ifp, node,
3180                                          &inet6_addr_lst[i], addr_lst) {
3181                         unsigned long age;
3182
3183                         if (ifp->flags & IFA_F_PERMANENT)
3184                                 continue;
3185
3186                         spin_lock(&ifp->lock);
3187                         /* We try to batch several events at once. */
3188                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3189
3190                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3191                             age >= ifp->valid_lft) {
3192                                 spin_unlock(&ifp->lock);
3193                                 in6_ifa_hold(ifp);
3194                                 ipv6_del_addr(ifp);
3195                                 goto restart;
3196                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3197                                 spin_unlock(&ifp->lock);
3198                                 continue;
3199                         } else if (age >= ifp->prefered_lft) {
3200                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3201                                 int deprecate = 0;
3202
3203                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3204                                         deprecate = 1;
3205                                         ifp->flags |= IFA_F_DEPRECATED;
3206                                 }
3207
3208                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3209                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3210
3211                                 spin_unlock(&ifp->lock);
3212
3213                                 if (deprecate) {
3214                                         in6_ifa_hold(ifp);
3215
3216                                         ipv6_ifa_notify(0, ifp);
3217                                         in6_ifa_put(ifp);
3218                                         goto restart;
3219                                 }
3220 #ifdef CONFIG_IPV6_PRIVACY
3221                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3222                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3223                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3224                                         ifp->idev->cnf.dad_transmits *
3225                                         ifp->idev->nd_parms->retrans_time / HZ;
3226
3227                                 if (age >= ifp->prefered_lft - regen_advance) {
3228                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3229                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3230                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3231                                         if (!ifp->regen_count && ifpub) {
3232                                                 ifp->regen_count++;
3233                                                 in6_ifa_hold(ifp);
3234                                                 in6_ifa_hold(ifpub);
3235                                                 spin_unlock(&ifp->lock);
3236
3237                                                 spin_lock(&ifpub->lock);
3238                                                 ifpub->regen_count = 0;
3239                                                 spin_unlock(&ifpub->lock);
3240                                                 ipv6_create_tempaddr(ifpub, ifp);
3241                                                 in6_ifa_put(ifpub);
3242                                                 in6_ifa_put(ifp);
3243                                                 goto restart;
3244                                         }
3245                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3246                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3247                                 spin_unlock(&ifp->lock);
3248 #endif
3249                         } else {
3250                                 /* ifp->prefered_lft <= ifp->valid_lft */
3251                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3252                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3253                                 spin_unlock(&ifp->lock);
3254                         }
3255                 }
3256         }
3257
3258         next_sec = round_jiffies_up(next);
3259         next_sched = next;
3260
3261         /* If rounded timeout is accurate enough, accept it. */
3262         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3263                 next_sched = next_sec;
3264
3265         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3266         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3267                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3268
3269         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3270               now, next, next_sec, next_sched));
3271
3272         addr_chk_timer.expires = next_sched;
3273         add_timer(&addr_chk_timer);
3274         spin_unlock(&addrconf_verify_lock);
3275         rcu_read_unlock_bh();
3276 }
3277
3278 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3279 {
3280         struct in6_addr *pfx = NULL;
3281
3282         if (addr)
3283                 pfx = nla_data(addr);
3284
3285         if (local) {
3286                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3287                         pfx = NULL;
3288                 else
3289                         pfx = nla_data(local);
3290         }
3291
3292         return pfx;
3293 }
3294
3295 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3296         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3297         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3298         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3299 };
3300
3301 static int
3302 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3303 {
3304         struct net *net = sock_net(skb->sk);
3305         struct ifaddrmsg *ifm;
3306         struct nlattr *tb[IFA_MAX+1];
3307         struct in6_addr *pfx;
3308         int err;
3309
3310         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3311         if (err < 0)
3312                 return err;
3313
3314         ifm = nlmsg_data(nlh);
3315         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3316         if (pfx == NULL)
3317                 return -EINVAL;
3318
3319         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3320 }
3321
3322 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3323                              u32 prefered_lft, u32 valid_lft)
3324 {
3325         u32 flags;
3326         clock_t expires;
3327         unsigned long timeout;
3328
3329         if (!valid_lft || (prefered_lft > valid_lft))
3330                 return -EINVAL;
3331
3332         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3333         if (addrconf_finite_timeout(timeout)) {
3334                 expires = jiffies_to_clock_t(timeout * HZ);
3335                 valid_lft = timeout;
3336                 flags = RTF_EXPIRES;
3337         } else {
3338                 expires = 0;
3339                 flags = 0;
3340                 ifa_flags |= IFA_F_PERMANENT;
3341         }
3342
3343         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3344         if (addrconf_finite_timeout(timeout)) {
3345                 if (timeout == 0)
3346                         ifa_flags |= IFA_F_DEPRECATED;
3347                 prefered_lft = timeout;
3348         }
3349
3350         spin_lock_bh(&ifp->lock);
3351         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3352         ifp->tstamp = jiffies;
3353         ifp->valid_lft = valid_lft;
3354         ifp->prefered_lft = prefered_lft;
3355
3356         spin_unlock_bh(&ifp->lock);
3357         if (!(ifp->flags&IFA_F_TENTATIVE))
3358                 ipv6_ifa_notify(0, ifp);
3359
3360         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3361                               expires, flags);
3362         addrconf_verify(0);
3363
3364         return 0;
3365 }
3366
3367 static int
3368 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3369 {
3370         struct net *net = sock_net(skb->sk);
3371         struct ifaddrmsg *ifm;
3372         struct nlattr *tb[IFA_MAX+1];
3373         struct in6_addr *pfx;
3374         struct inet6_ifaddr *ifa;
3375         struct net_device *dev;
3376         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3377         u8 ifa_flags;
3378         int err;
3379
3380         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3381         if (err < 0)
3382                 return err;
3383
3384         ifm = nlmsg_data(nlh);
3385         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3386         if (pfx == NULL)
3387                 return -EINVAL;
3388
3389         if (tb[IFA_CACHEINFO]) {
3390                 struct ifa_cacheinfo *ci;
3391
3392                 ci = nla_data(tb[IFA_CACHEINFO]);
3393                 valid_lft = ci->ifa_valid;
3394                 preferred_lft = ci->ifa_prefered;
3395         } else {
3396                 preferred_lft = INFINITY_LIFE_TIME;
3397                 valid_lft = INFINITY_LIFE_TIME;
3398         }
3399
3400         dev =  __dev_get_by_index(net, ifm->ifa_index);
3401         if (dev == NULL)
3402                 return -ENODEV;
3403
3404         /* We ignore other flags so far. */
3405         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3406
3407         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3408         if (ifa == NULL) {
3409                 /*
3410                  * It would be best to check for !NLM_F_CREATE here but
3411                  * userspace alreay relies on not having to provide this.
3412                  */
3413                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3414                                       ifm->ifa_prefixlen, ifa_flags,
3415                                       preferred_lft, valid_lft);
3416         }
3417
3418         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3419             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3420                 err = -EEXIST;
3421         else
3422                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3423
3424         in6_ifa_put(ifa);
3425
3426         return err;
3427 }
3428
3429 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3430                           u8 scope, int ifindex)
3431 {
3432         struct ifaddrmsg *ifm;
3433
3434         ifm = nlmsg_data(nlh);
3435         ifm->ifa_family = AF_INET6;
3436         ifm->ifa_prefixlen = prefixlen;
3437         ifm->ifa_flags = flags;
3438         ifm->ifa_scope = scope;
3439         ifm->ifa_index = ifindex;
3440 }
3441
3442 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3443                          unsigned long tstamp, u32 preferred, u32 valid)
3444 {
3445         struct ifa_cacheinfo ci;
3446
3447         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3448                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3449         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3450                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3451         ci.ifa_prefered = preferred;
3452         ci.ifa_valid = valid;
3453
3454         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3455 }
3456
3457 static inline int rt_scope(int ifa_scope)
3458 {
3459         if (ifa_scope & IFA_HOST)
3460                 return RT_SCOPE_HOST;
3461         else if (ifa_scope & IFA_LINK)
3462                 return RT_SCOPE_LINK;
3463         else if (ifa_scope & IFA_SITE)
3464                 return RT_SCOPE_SITE;
3465         else
3466                 return RT_SCOPE_UNIVERSE;
3467 }
3468
3469 static inline int inet6_ifaddr_msgsize(void)
3470 {
3471         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3472                + nla_total_size(16) /* IFA_ADDRESS */
3473                + nla_total_size(sizeof(struct ifa_cacheinfo));
3474 }
3475
3476 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3477                              u32 pid, u32 seq, int event, unsigned int flags)
3478 {
3479         struct nlmsghdr  *nlh;
3480         u32 preferred, valid;
3481
3482         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3483         if (nlh == NULL)
3484                 return -EMSGSIZE;
3485
3486         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3487                       ifa->idev->dev->ifindex);
3488
3489         if (!(ifa->flags&IFA_F_PERMANENT)) {
3490                 preferred = ifa->prefered_lft;
3491                 valid = ifa->valid_lft;
3492                 if (preferred != INFINITY_LIFE_TIME) {
3493                         long tval = (jiffies - ifa->tstamp)/HZ;
3494                         if (preferred > tval)
3495                                 preferred -= tval;
3496                         else
3497                                 preferred = 0;
3498                         if (valid != INFINITY_LIFE_TIME) {
3499                                 if (valid > tval)
3500                                         valid -= tval;
3501                                 else
3502                                         valid = 0;
3503                         }
3504                 }
3505         } else {
3506                 preferred = INFINITY_LIFE_TIME;
3507                 valid = INFINITY_LIFE_TIME;
3508         }
3509
3510         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3511             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3512                 nlmsg_cancel(skb, nlh);
3513                 return -EMSGSIZE;
3514         }
3515
3516         return nlmsg_end(skb, nlh);
3517 }
3518
3519 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3520                                 u32 pid, u32 seq, int event, u16 flags)
3521 {
3522         struct nlmsghdr  *nlh;
3523         u8 scope = RT_SCOPE_UNIVERSE;
3524         int ifindex = ifmca->idev->dev->ifindex;
3525
3526         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3527                 scope = RT_SCOPE_SITE;
3528
3529         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3530         if (nlh == NULL)
3531                 return -EMSGSIZE;
3532
3533         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3534         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3535             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3536                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3537                 nlmsg_cancel(skb, nlh);
3538                 return -EMSGSIZE;
3539         }
3540
3541         return nlmsg_end(skb, nlh);
3542 }
3543
3544 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3545                                 u32 pid, u32 seq, int event, unsigned int flags)
3546 {
3547         struct nlmsghdr  *nlh;
3548         u8 scope = RT_SCOPE_UNIVERSE;
3549         int ifindex = ifaca->aca_idev->dev->ifindex;
3550
3551         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3552                 scope = RT_SCOPE_SITE;
3553
3554         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3555         if (nlh == NULL)
3556                 return -EMSGSIZE;
3557
3558         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3559         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3560             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3561                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3562                 nlmsg_cancel(skb, nlh);
3563                 return -EMSGSIZE;
3564         }
3565
3566         return nlmsg_end(skb, nlh);
3567 }
3568
3569 enum addr_type_t {
3570         UNICAST_ADDR,
3571         MULTICAST_ADDR,
3572         ANYCAST_ADDR,
3573 };
3574
3575 /* called with rcu_read_lock() */
3576 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3577                           struct netlink_callback *cb, enum addr_type_t type,
3578                           int s_ip_idx, int *p_ip_idx)
3579 {
3580         struct ifmcaddr6 *ifmca;
3581         struct ifacaddr6 *ifaca;
3582         int err = 1;
3583         int ip_idx = *p_ip_idx;
3584
3585         read_lock_bh(&idev->lock);
3586         switch (type) {
3587         case UNICAST_ADDR: {
3588                 struct inet6_ifaddr *ifa;
3589
3590                 /* unicast address incl. temp addr */
3591                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3592                         if (++ip_idx < s_ip_idx)
3593                                 continue;
3594                         err = inet6_fill_ifaddr(skb, ifa,
3595                                                 NETLINK_CB(cb->skb).pid,
3596                                                 cb->nlh->nlmsg_seq,
3597                                                 RTM_NEWADDR,
3598                                                 NLM_F_MULTI);
3599                         if (err <= 0)
3600                                 break;
3601                 }
3602                 break;
3603         }
3604         case MULTICAST_ADDR:
3605                 /* multicast address */
3606                 for (ifmca = idev->mc_list; ifmca;
3607                      ifmca = ifmca->next, ip_idx++) {
3608                         if (ip_idx < s_ip_idx)
3609                                 continue;
3610                         err = inet6_fill_ifmcaddr(skb, ifmca,
3611                                                   NETLINK_CB(cb->skb).pid,
3612                                                   cb->nlh->nlmsg_seq,
3613                                                   RTM_GETMULTICAST,
3614                                                   NLM_F_MULTI);
3615                         if (err <= 0)
3616                                 break;
3617                 }
3618                 break;
3619         case ANYCAST_ADDR:
3620                 /* anycast address */
3621                 for (ifaca = idev->ac_list; ifaca;
3622                      ifaca = ifaca->aca_next, ip_idx++) {
3623                         if (ip_idx < s_ip_idx)
3624                                 continue;
3625                         err = inet6_fill_ifacaddr(skb, ifaca,
3626                                                   NETLINK_CB(cb->skb).pid,
3627                                                   cb->nlh->nlmsg_seq,
3628                                                   RTM_GETANYCAST,
3629                                                   NLM_F_MULTI);
3630                         if (err <= 0)
3631                                 break;
3632                 }
3633                 break;
3634         default:
3635                 break;
3636         }
3637         read_unlock_bh(&idev->lock);
3638         *p_ip_idx = ip_idx;
3639         return err;
3640 }
3641
3642 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3643                            enum addr_type_t type)
3644 {
3645         struct net *net = sock_net(skb->sk);
3646         int h, s_h;
3647         int idx, ip_idx;
3648         int s_idx, s_ip_idx;
3649         struct net_device *dev;
3650         struct inet6_dev *idev;
3651         struct hlist_head *head;
3652         struct hlist_node *node;
3653
3654         s_h = cb->args[0];
3655         s_idx = idx = cb->args[1];
3656         s_ip_idx = ip_idx = cb->args[2];
3657
3658         rcu_read_lock();
3659         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3660                 idx = 0;
3661                 head = &net->dev_index_head[h];
3662                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3663                         if (idx < s_idx)
3664                                 goto cont;
3665                         if (h > s_h || idx > s_idx)
3666                                 s_ip_idx = 0;
3667                         ip_idx = 0;
3668                         idev = __in6_dev_get(dev);
3669                         if (!idev)
3670                                 goto cont;
3671
3672                         if (in6_dump_addrs(idev, skb, cb, type,
3673                                            s_ip_idx, &ip_idx) <= 0)
3674                                 goto done;
3675 cont:
3676                         idx++;
3677                 }
3678         }
3679 done:
3680         rcu_read_unlock();
3681         cb->args[0] = h;
3682         cb->args[1] = idx;
3683         cb->args[2] = ip_idx;
3684
3685         return skb->len;
3686 }
3687
3688 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3689 {
3690         enum addr_type_t type = UNICAST_ADDR;
3691
3692         return inet6_dump_addr(skb, cb, type);
3693 }
3694
3695 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3696 {
3697         enum addr_type_t type = MULTICAST_ADDR;
3698
3699         return inet6_dump_addr(skb, cb, type);
3700 }
3701
3702
3703 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3704 {
3705         enum addr_type_t type = ANYCAST_ADDR;
3706
3707         return inet6_dump_addr(skb, cb, type);
3708 }
3709
3710 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3711                              void *arg)
3712 {
3713         struct net *net = sock_net(in_skb->sk);
3714         struct ifaddrmsg *ifm;
3715         struct nlattr *tb[IFA_MAX+1];
3716         struct in6_addr *addr = NULL;
3717         struct net_device *dev = NULL;
3718         struct inet6_ifaddr *ifa;
3719         struct sk_buff *skb;
3720         int err;
3721
3722         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3723         if (err < 0)
3724                 goto errout;
3725
3726         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3727         if (addr == NULL) {
3728                 err = -EINVAL;
3729                 goto errout;
3730         }
3731
3732         ifm = nlmsg_data(nlh);
3733         if (ifm->ifa_index)
3734                 dev = __dev_get_by_index(net, ifm->ifa_index);
3735
3736         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3737         if (!ifa) {
3738                 err = -EADDRNOTAVAIL;
3739                 goto errout;
3740         }
3741
3742         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3743         if (!skb) {
3744                 err = -ENOBUFS;
3745                 goto errout_ifa;
3746         }
3747
3748         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3749                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3750         if (err < 0) {
3751                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3752                 WARN_ON(err == -EMSGSIZE);
3753                 kfree_skb(skb);
3754                 goto errout_ifa;
3755         }
3756         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3757 errout_ifa:
3758         in6_ifa_put(ifa);
3759 errout:
3760         return err;
3761 }
3762
3763 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3764 {
3765         struct sk_buff *skb;
3766         struct net *net = dev_net(ifa->idev->dev);
3767         int err = -ENOBUFS;
3768
3769         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3770         if (skb == NULL)
3771                 goto errout;
3772
3773         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3774         if (err < 0) {
3775                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3776                 WARN_ON(err == -EMSGSIZE);
3777                 kfree_skb(skb);
3778                 goto errout;
3779         }
3780         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3781         return;
3782 errout:
3783         if (err < 0)
3784                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3785 }
3786
3787 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3788                                 __s32 *array, int bytes)
3789 {
3790         BUG_ON(bytes < (DEVCONF_MAX * 4));
3791
3792         memset(array, 0, bytes);
3793         array[DEVCONF_FORWARDING] = cnf->forwarding;
3794         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3795         array[DEVCONF_MTU6] = cnf->mtu6;
3796         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3797         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3798         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3799         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3800         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3801         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3802         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3803         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3804 #ifdef CONFIG_IPV6_PRIVACY
3805         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3806         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3807         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3808         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3809         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3810 #endif
3811         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3812         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3813         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3814 #ifdef CONFIG_IPV6_ROUTER_PREF
3815         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3816         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3817 #ifdef CONFIG_IPV6_ROUTE_INFO
3818         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3819 #endif
3820 #endif
3821         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3822         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3823 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3824         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3825 #endif
3826 #ifdef CONFIG_IPV6_MROUTE
3827         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3828 #endif
3829         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3830         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3831         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3832 }
3833
3834 static inline size_t inet6_if_nlmsg_size(void)
3835 {
3836         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3837                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3838                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3839                + nla_total_size(4) /* IFLA_MTU */
3840                + nla_total_size(4) /* IFLA_LINK */
3841                + nla_total_size( /* IFLA_PROTINFO */
3842                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3843                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3844                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3845                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3846                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3847                  );
3848 }
3849
3850 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3851                                       int items, int bytes)
3852 {
3853         int i;
3854         int pad = bytes - sizeof(u64) * items;
3855         BUG_ON(pad < 0);
3856
3857         /* Use put_unaligned() because stats may not be aligned for u64. */
3858         put_unaligned(items, &stats[0]);
3859         for (i = 1; i < items; i++)
3860                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3861
3862         memset(&stats[items], 0, pad);
3863 }
3864
3865 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3866                              int bytes)
3867 {
3868         switch (attrtype) {
3869         case IFLA_INET6_STATS:
3870                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3871                 break;
3872         case IFLA_INET6_ICMP6STATS:
3873                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3874                 break;
3875         }
3876 }
3877
3878 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3879                              u32 pid, u32 seq, int event, unsigned int flags)
3880 {
3881         struct net_device *dev = idev->dev;
3882         struct nlattr *nla;
3883         struct ifinfomsg *hdr;
3884         struct nlmsghdr *nlh;
3885         void *protoinfo;
3886         struct ifla_cacheinfo ci;
3887
3888         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3889         if (nlh == NULL)
3890                 return -EMSGSIZE;
3891
3892         hdr = nlmsg_data(nlh);
3893         hdr->ifi_family = AF_INET6;
3894         hdr->__ifi_pad = 0;
3895         hdr->ifi_type = dev->type;
3896         hdr->ifi_index = dev->ifindex;
3897         hdr->ifi_flags = dev_get_flags(dev);
3898         hdr->ifi_change = 0;
3899
3900         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3901
3902         if (dev->addr_len)
3903                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3904
3905         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3906         if (dev->ifindex != dev->iflink)
3907                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3908
3909         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3910         if (protoinfo == NULL)
3911                 goto nla_put_failure;
3912
3913         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3914
3915         ci.max_reasm_len = IPV6_MAXPLEN;
3916         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3917                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3918         ci.reachable_time = idev->nd_parms->reachable_time;
3919         ci.retrans_time = idev->nd_parms->retrans_time;
3920         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3921
3922         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3923         if (nla == NULL)
3924                 goto nla_put_failure;
3925         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3926
3927         /* XXX - MC not implemented */
3928
3929         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3930         if (nla == NULL)
3931                 goto nla_put_failure;
3932         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3933
3934         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3935         if (nla == NULL)
3936                 goto nla_put_failure;
3937         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3938
3939         nla_nest_end(skb, protoinfo);
3940         return nlmsg_end(skb, nlh);
3941
3942 nla_put_failure:
3943         nlmsg_cancel(skb, nlh);
3944         return -EMSGSIZE;
3945 }
3946
3947 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3948 {
3949         struct net *net = sock_net(skb->sk);
3950         int h, s_h;
3951         int idx = 0, s_idx;
3952         struct net_device *dev;
3953         struct inet6_dev *idev;
3954         struct hlist_head *head;
3955         struct hlist_node *node;
3956
3957         s_h = cb->args[0];
3958         s_idx = cb->args[1];
3959
3960         rcu_read_lock();
3961         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3962                 idx = 0;
3963                 head = &net->dev_index_head[h];
3964                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3965                         if (idx < s_idx)
3966                                 goto cont;
3967                         idev = __in6_dev_get(dev);
3968                         if (!idev)
3969                                 goto cont;
3970                         if (inet6_fill_ifinfo(skb, idev,
3971                                               NETLINK_CB(cb->skb).pid,
3972                                               cb->nlh->nlmsg_seq,
3973                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3974                                 goto out;
3975 cont:
3976                         idx++;
3977                 }
3978         }
3979 out:
3980         rcu_read_unlock();
3981         cb->args[1] = idx;
3982         cb->args[0] = h;
3983
3984         return skb->len;
3985 }
3986
3987 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3988 {
3989         struct sk_buff *skb;
3990         struct net *net = dev_net(idev->dev);
3991         int err = -ENOBUFS;
3992
3993         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3994         if (skb == NULL)
3995                 goto errout;
3996
3997         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3998         if (err < 0) {
3999                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4000                 WARN_ON(err == -EMSGSIZE);
4001                 kfree_skb(skb);
4002                 goto errout;
4003         }
4004         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4005         return;
4006 errout:
4007         if (err < 0)
4008                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4009 }
4010
4011 static inline size_t inet6_prefix_nlmsg_size(void)
4012 {
4013         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4014                + nla_total_size(sizeof(struct in6_addr))
4015                + nla_total_size(sizeof(struct prefix_cacheinfo));
4016 }
4017
4018 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4019                              struct prefix_info *pinfo, u32 pid, u32 seq,
4020                              int event, unsigned int flags)
4021 {
4022         struct prefixmsg *pmsg;
4023         struct nlmsghdr *nlh;
4024         struct prefix_cacheinfo ci;
4025
4026         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4027         if (nlh == NULL)
4028                 return -EMSGSIZE;
4029
4030         pmsg = nlmsg_data(nlh);
4031         pmsg->prefix_family = AF_INET6;
4032         pmsg->prefix_pad1 = 0;
4033         pmsg->prefix_pad2 = 0;
4034         pmsg->prefix_ifindex = idev->dev->ifindex;
4035         pmsg->prefix_len = pinfo->prefix_len;
4036         pmsg->prefix_type = pinfo->type;
4037         pmsg->prefix_pad3 = 0;
4038         pmsg->prefix_flags = 0;
4039         if (pinfo->onlink)
4040                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4041         if (pinfo->autoconf)
4042                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4043
4044         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4045
4046         ci.preferred_time = ntohl(pinfo->prefered);
4047         ci.valid_time = ntohl(pinfo->valid);
4048         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4049
4050         return nlmsg_end(skb, nlh);
4051
4052 nla_put_failure:
4053         nlmsg_cancel(skb, nlh);
4054         return -EMSGSIZE;
4055 }
4056
4057 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4058                          struct prefix_info *pinfo)
4059 {
4060         struct sk_buff *skb;
4061         struct net *net = dev_net(idev->dev);
4062         int err = -ENOBUFS;
4063
4064         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4065         if (skb == NULL)
4066                 goto errout;
4067
4068         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4069         if (err < 0) {
4070                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4071                 WARN_ON(err == -EMSGSIZE);
4072                 kfree_skb(skb);
4073                 goto errout;
4074         }
4075         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4076         return;
4077 errout:
4078         if (err < 0)
4079                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4080 }
4081
4082 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4083 {
4084         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4085
4086         switch (event) {
4087         case RTM_NEWADDR:
4088                 /*
4089                  * If the address was optimistic
4090                  * we inserted the route at the start of
4091                  * our DAD process, so we don't need
4092                  * to do it again
4093                  */
4094                 if (!(ifp->rt->rt6i_node))
4095                         ip6_ins_rt(ifp->rt);
4096                 if (ifp->idev->cnf.forwarding)
4097                         addrconf_join_anycast(ifp);
4098                 break;
4099         case RTM_DELADDR:
4100                 if (ifp->idev->cnf.forwarding)
4101                         addrconf_leave_anycast(ifp);
4102                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4103                 dst_hold(&ifp->rt->dst);
4104
4105                 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4106                     ip6_del_rt(ifp->rt))
4107                         dst_free(&ifp->rt->dst);
4108                 break;
4109         }
4110 }
4111
4112 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4113 {
4114         rcu_read_lock_bh();
4115         if (likely(ifp->idev->dead == 0))
4116                 __ipv6_ifa_notify(event, ifp);
4117         rcu_read_unlock_bh();
4118 }
4119
4120 #ifdef CONFIG_SYSCTL
4121
4122 static
4123 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4124                            void __user *buffer, size_t *lenp, loff_t *ppos)
4125 {
4126         int *valp = ctl->data;
4127         int val = *valp;
4128         loff_t pos = *ppos;
4129         int ret;
4130
4131         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4132
4133         if (write)
4134                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4135         if (ret)
4136                 *ppos = pos;
4137         return ret;
4138 }
4139
4140 static void dev_disable_change(struct inet6_dev *idev)
4141 {
4142         if (!idev || !idev->dev)
4143                 return;
4144
4145         if (idev->cnf.disable_ipv6)
4146                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4147         else
4148                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4149 }
4150
4151 static void addrconf_disable_change(struct net *net, __s32 newf)
4152 {
4153         struct net_device *dev;
4154         struct inet6_dev *idev;
4155
4156         rcu_read_lock();
4157         for_each_netdev_rcu(net, dev) {
4158                 idev = __in6_dev_get(dev);
4159                 if (idev) {
4160                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4161                         idev->cnf.disable_ipv6 = newf;
4162                         if (changed)
4163                                 dev_disable_change(idev);
4164                 }
4165         }
4166         rcu_read_unlock();
4167 }
4168
4169 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4170 {
4171         struct net *net;
4172
4173         net = (struct net *)table->extra2;
4174
4175         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4176                 return 0;
4177
4178         if (!rtnl_trylock()) {
4179                 /* Restore the original values before restarting */
4180                 *p = old;
4181                 return restart_syscall();
4182         }
4183
4184         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4185                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4186                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4187                 addrconf_disable_change(net, newf);
4188         } else if ((!*p) ^ (!old))
4189                 dev_disable_change((struct inet6_dev *)table->extra1);
4190
4191         rtnl_unlock();
4192         return 0;
4193 }
4194
4195 static
4196 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4197                             void __user *buffer, size_t *lenp, loff_t *ppos)
4198 {
4199         int *valp = ctl->data;
4200         int val = *valp;
4201         loff_t pos = *ppos;
4202         int ret;
4203
4204         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4205
4206         if (write)
4207                 ret = addrconf_disable_ipv6(ctl, valp, val);
4208         if (ret)
4209                 *ppos = pos;
4210         return ret;
4211 }
4212
4213 static struct addrconf_sysctl_table
4214 {
4215         struct ctl_table_header *sysctl_header;
4216         ctl_table addrconf_vars[DEVCONF_MAX+1];
4217         char *dev_name;
4218 } addrconf_sysctl __read_mostly = {
4219         .sysctl_header = NULL,
4220         .addrconf_vars = {
4221                 {
4222                         .procname       = "forwarding",
4223                         .data           = &ipv6_devconf.forwarding,
4224                         .maxlen         = sizeof(int),
4225                         .mode           = 0644,
4226                         .proc_handler   = addrconf_sysctl_forward,
4227                 },
4228                 {
4229                         .procname       = "hop_limit",
4230                         .data           = &ipv6_devconf.hop_limit,
4231                         .maxlen         = sizeof(int),
4232                         .mode           = 0644,
4233                         .proc_handler   = proc_dointvec,
4234                 },
4235                 {
4236                         .procname       = "mtu",
4237                         .data           = &ipv6_devconf.mtu6,
4238                         .maxlen         = sizeof(int),
4239                         .mode           = 0644,
4240                         .proc_handler   = proc_dointvec,
4241                 },
4242                 {
4243                         .procname       = "accept_ra",
4244                         .data           = &ipv6_devconf.accept_ra,
4245                         .maxlen         = sizeof(int),
4246                         .mode           = 0644,
4247                         .proc_handler   = proc_dointvec,
4248                 },
4249                 {
4250                         .procname       = "accept_redirects",
4251                         .data           = &ipv6_devconf.accept_redirects,
4252                         .maxlen         = sizeof(int),
4253                         .mode           = 0644,
4254                         .proc_handler   = proc_dointvec,
4255                 },
4256                 {
4257                         .procname       = "autoconf",
4258                         .data           = &ipv6_devconf.autoconf,
4259                         .maxlen         = sizeof(int),
4260                         .mode           = 0644,
4261                         .proc_handler   = proc_dointvec,
4262                 },
4263                 {
4264                         .procname       = "dad_transmits",
4265                         .data           = &ipv6_devconf.dad_transmits,
4266                         .maxlen         = sizeof(int),
4267                         .mode           = 0644,
4268                         .proc_handler   = proc_dointvec,
4269                 },
4270                 {
4271                         .procname       = "router_solicitations",
4272                         .data           = &ipv6_devconf.rtr_solicits,
4273                         .maxlen         = sizeof(int),
4274                         .mode           = 0644,
4275                         .proc_handler   = proc_dointvec,
4276                 },
4277                 {
4278                         .procname       = "router_solicitation_interval",
4279                         .data           = &ipv6_devconf.rtr_solicit_interval,
4280                         .maxlen         = sizeof(int),
4281                         .mode           = 0644,
4282                         .proc_handler   = proc_dointvec_jiffies,
4283                 },
4284                 {
4285                         .procname       = "router_solicitation_delay",
4286                         .data           = &ipv6_devconf.rtr_solicit_delay,
4287                         .maxlen         = sizeof(int),
4288                         .mode           = 0644,
4289                         .proc_handler   = proc_dointvec_jiffies,
4290                 },
4291                 {
4292                         .procname       = "force_mld_version",
4293                         .data           = &ipv6_devconf.force_mld_version,
4294                         .maxlen         = sizeof(int),
4295                         .mode           = 0644,
4296                         .proc_handler   = proc_dointvec,
4297                 },
4298 #ifdef CONFIG_IPV6_PRIVACY
4299                 {
4300                         .procname       = "use_tempaddr",
4301                         .data           = &ipv6_devconf.use_tempaddr,
4302                         .maxlen         = sizeof(int),
4303                         .mode           = 0644,
4304                         .proc_handler   = proc_dointvec,
4305                 },
4306                 {
4307                         .procname       = "temp_valid_lft",
4308                         .data           = &ipv6_devconf.temp_valid_lft,
4309                         .maxlen         = sizeof(int),
4310                         .mode           = 0644,
4311                         .proc_handler   = proc_dointvec,
4312                 },
4313                 {
4314                         .procname       = "temp_prefered_lft",
4315                         .data           = &ipv6_devconf.temp_prefered_lft,
4316                         .maxlen         = sizeof(int),
4317                         .mode           = 0644,
4318                         .proc_handler   = proc_dointvec,
4319                 },
4320                 {
4321                         .procname       = "regen_max_retry",
4322                         .data           = &ipv6_devconf.regen_max_retry,
4323                         .maxlen         = sizeof(int),
4324                         .mode           = 0644,
4325                         .proc_handler   = proc_dointvec,
4326                 },
4327                 {
4328                         .procname       = "max_desync_factor",
4329                         .data           = &ipv6_devconf.max_desync_factor,
4330                         .maxlen         = sizeof(int),
4331                         .mode           = 0644,
4332                         .proc_handler   = proc_dointvec,
4333                 },
4334 #endif
4335                 {
4336                         .procname       = "max_addresses",
4337                         .data           = &ipv6_devconf.max_addresses,
4338                         .maxlen         = sizeof(int),
4339                         .mode           = 0644,
4340                         .proc_handler   = proc_dointvec,
4341                 },
4342                 {
4343                         .procname       = "accept_ra_defrtr",
4344                         .data           = &ipv6_devconf.accept_ra_defrtr,
4345                         .maxlen         = sizeof(int),
4346                         .mode           = 0644,
4347                         .proc_handler   = proc_dointvec,
4348                 },
4349                 {
4350                         .procname       = "accept_ra_pinfo",
4351                         .data           = &ipv6_devconf.accept_ra_pinfo,
4352                         .maxlen         = sizeof(int),
4353                         .mode           = 0644,
4354                         .proc_handler   = proc_dointvec,
4355                 },
4356 #ifdef CONFIG_IPV6_ROUTER_PREF
4357                 {
4358                         .procname       = "accept_ra_rtr_pref",
4359                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4360                         .maxlen         = sizeof(int),
4361                         .mode           = 0644,
4362                         .proc_handler   = proc_dointvec,
4363                 },
4364                 {
4365                         .procname       = "router_probe_interval",
4366                         .data           = &ipv6_devconf.rtr_probe_interval,
4367                         .maxlen         = sizeof(int),
4368                         .mode           = 0644,
4369                         .proc_handler   = proc_dointvec_jiffies,
4370                 },
4371 #ifdef CONFIG_IPV6_ROUTE_INFO
4372                 {
4373                         .procname       = "accept_ra_rt_info_max_plen",
4374                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4375                         .maxlen         = sizeof(int),
4376                         .mode           = 0644,
4377                         .proc_handler   = proc_dointvec,
4378                 },
4379 #endif
4380 #endif
4381                 {
4382                         .procname       = "proxy_ndp",
4383                         .data           = &ipv6_devconf.proxy_ndp,
4384                         .maxlen         = sizeof(int),
4385                         .mode           = 0644,
4386                         .proc_handler   = proc_dointvec,
4387                 },
4388                 {
4389                         .procname       = "accept_source_route",
4390                         .data           = &ipv6_devconf.accept_source_route,
4391                         .maxlen         = sizeof(int),
4392                         .mode           = 0644,
4393                         .proc_handler   = proc_dointvec,
4394                 },
4395 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4396                 {
4397                         .procname       = "optimistic_dad",
4398                         .data           = &ipv6_devconf.optimistic_dad,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec,
4402
4403                 },
4404 #endif
4405 #ifdef CONFIG_IPV6_MROUTE
4406                 {
4407                         .procname       = "mc_forwarding",
4408                         .data           = &ipv6_devconf.mc_forwarding,
4409                         .maxlen         = sizeof(int),
4410                         .mode           = 0444,
4411                         .proc_handler   = proc_dointvec,
4412                 },
4413 #endif
4414                 {
4415                         .procname       = "disable_ipv6",
4416                         .data           = &ipv6_devconf.disable_ipv6,
4417                         .maxlen         = sizeof(int),
4418                         .mode           = 0644,
4419                         .proc_handler   = addrconf_sysctl_disable,
4420                 },
4421                 {
4422                         .procname       = "accept_dad",
4423                         .data           = &ipv6_devconf.accept_dad,
4424                         .maxlen         = sizeof(int),
4425                         .mode           = 0644,
4426                         .proc_handler   = proc_dointvec,
4427                 },
4428                 {
4429                         .procname       = "force_tllao",
4430                         .data           = &ipv6_devconf.force_tllao,
4431                         .maxlen         = sizeof(int),
4432                         .mode           = 0644,
4433                         .proc_handler   = proc_dointvec
4434                 },
4435                 {
4436                         /* sentinel */
4437                 }
4438         },
4439 };
4440
4441 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4442                 struct inet6_dev *idev, struct ipv6_devconf *p)
4443 {
4444         int i;
4445         struct addrconf_sysctl_table *t;
4446
4447 #define ADDRCONF_CTL_PATH_DEV   3
4448
4449         struct ctl_path addrconf_ctl_path[] = {
4450                 { .procname = "net", },
4451                 { .procname = "ipv6", },
4452                 { .procname = "conf", },
4453                 { /* to be set */ },
4454                 { },
4455         };
4456
4457
4458         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4459         if (t == NULL)
4460                 goto out;
4461
4462         for (i = 0; t->addrconf_vars[i].data; i++) {
4463                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4464                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4465                 t->addrconf_vars[i].extra2 = net;
4466         }
4467
4468         /*
4469          * Make a copy of dev_name, because '.procname' is regarded as const
4470          * by sysctl and we wouldn't want anyone to change it under our feet
4471          * (see SIOCSIFNAME).
4472          */
4473         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4474         if (!t->dev_name)
4475                 goto free;
4476
4477         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4478
4479         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4480                         t->addrconf_vars);
4481         if (t->sysctl_header == NULL)
4482                 goto free_procname;
4483
4484         p->sysctl = t;
4485         return 0;
4486
4487 free_procname:
4488         kfree(t->dev_name);
4489 free:
4490         kfree(t);
4491 out:
4492         return -ENOBUFS;
4493 }
4494
4495 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4496 {
4497         struct addrconf_sysctl_table *t;
4498
4499         if (p->sysctl == NULL)
4500                 return;
4501
4502         t = p->sysctl;
4503         p->sysctl = NULL;
4504         unregister_sysctl_table(t->sysctl_header);
4505         kfree(t->dev_name);
4506         kfree(t);
4507 }
4508
4509 static void addrconf_sysctl_register(struct inet6_dev *idev)
4510 {
4511         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4512                               &ndisc_ifinfo_sysctl_change);
4513         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4514                                         idev, &idev->cnf);
4515 }
4516
4517 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4518 {
4519         __addrconf_sysctl_unregister(&idev->cnf);
4520         neigh_sysctl_unregister(idev->nd_parms);
4521 }
4522
4523
4524 #endif
4525
4526 static int __net_init addrconf_init_net(struct net *net)
4527 {
4528         int err;
4529         struct ipv6_devconf *all, *dflt;
4530
4531         err = -ENOMEM;
4532         all = &ipv6_devconf;
4533         dflt = &ipv6_devconf_dflt;
4534
4535         if (!net_eq(net, &init_net)) {
4536                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4537                 if (all == NULL)
4538                         goto err_alloc_all;
4539
4540                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4541                 if (dflt == NULL)
4542                         goto err_alloc_dflt;
4543         } else {
4544                 /* these will be inherited by all namespaces */
4545                 dflt->autoconf = ipv6_defaults.autoconf;
4546                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4547         }
4548
4549         net->ipv6.devconf_all = all;
4550         net->ipv6.devconf_dflt = dflt;
4551
4552 #ifdef CONFIG_SYSCTL
4553         err = __addrconf_sysctl_register(net, "all", NULL, all);
4554         if (err < 0)
4555                 goto err_reg_all;
4556
4557         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4558         if (err < 0)
4559                 goto err_reg_dflt;
4560 #endif
4561         return 0;
4562
4563 #ifdef CONFIG_SYSCTL
4564 err_reg_dflt:
4565         __addrconf_sysctl_unregister(all);
4566 err_reg_all:
4567         kfree(dflt);
4568 #endif
4569 err_alloc_dflt:
4570         kfree(all);
4571 err_alloc_all:
4572         return err;
4573 }
4574
4575 static void __net_exit addrconf_exit_net(struct net *net)
4576 {
4577 #ifdef CONFIG_SYSCTL
4578         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4579         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4580 #endif
4581         if (!net_eq(net, &init_net)) {
4582                 kfree(net->ipv6.devconf_dflt);
4583                 kfree(net->ipv6.devconf_all);
4584         }
4585 }
4586
4587 static struct pernet_operations addrconf_ops = {
4588         .init = addrconf_init_net,
4589         .exit = addrconf_exit_net,
4590 };
4591
4592 /*
4593  *      Device notifier
4594  */
4595
4596 int register_inet6addr_notifier(struct notifier_block *nb)
4597 {
4598         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4599 }
4600 EXPORT_SYMBOL(register_inet6addr_notifier);
4601
4602 int unregister_inet6addr_notifier(struct notifier_block *nb)
4603 {
4604         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4605 }
4606 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4607
4608 /*
4609  *      Init / cleanup code
4610  */
4611
4612 int __init addrconf_init(void)
4613 {
4614         int i, err;
4615
4616         err = ipv6_addr_label_init();
4617         if (err < 0) {
4618                 printk(KERN_CRIT "IPv6 Addrconf:"
4619                        " cannot initialize default policy table: %d.\n", err);
4620                 return err;
4621         }
4622
4623         register_pernet_subsys(&addrconf_ops);
4624
4625         /* The addrconf netdev notifier requires that loopback_dev
4626          * has it's ipv6 private information allocated and setup
4627          * before it can bring up and give link-local addresses
4628          * to other devices which are up.
4629          *
4630          * Unfortunately, loopback_dev is not necessarily the first
4631          * entry in the global dev_base list of net devices.  In fact,
4632          * it is likely to be the very last entry on that list.
4633          * So this causes the notifier registry below to try and
4634          * give link-local addresses to all devices besides loopback_dev
4635          * first, then loopback_dev, which cases all the non-loopback_dev
4636          * devices to fail to get a link-local address.
4637          *
4638          * So, as a temporary fix, allocate the ipv6 structure for
4639          * loopback_dev first by hand.
4640          * Longer term, all of the dependencies ipv6 has upon the loopback
4641          * device and it being up should be removed.
4642          */
4643         rtnl_lock();
4644         if (!ipv6_add_dev(init_net.loopback_dev))
4645                 err = -ENOMEM;
4646         rtnl_unlock();
4647         if (err)
4648                 goto errlo;
4649
4650         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4651                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4652
4653         register_netdevice_notifier(&ipv6_dev_notf);
4654
4655         addrconf_verify(0);
4656
4657         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4658         if (err < 0)
4659                 goto errout;
4660
4661         /* Only the first call to __rtnl_register can fail */
4662         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4663         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4664         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4665         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4666         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4667
4668         ipv6_addr_label_rtnl_register();
4669
4670         return 0;
4671 errout:
4672         unregister_netdevice_notifier(&ipv6_dev_notf);
4673 errlo:
4674         unregister_pernet_subsys(&addrconf_ops);
4675
4676         return err;
4677 }
4678
4679 void addrconf_cleanup(void)
4680 {
4681         struct net_device *dev;
4682         int i;
4683
4684         unregister_netdevice_notifier(&ipv6_dev_notf);
4685         unregister_pernet_subsys(&addrconf_ops);
4686
4687         rtnl_lock();
4688
4689         /* clean dev list */
4690         for_each_netdev(&init_net, dev) {
4691                 if (__in6_dev_get(dev) == NULL)
4692                         continue;
4693                 addrconf_ifdown(dev, 1);
4694         }
4695         addrconf_ifdown(init_net.loopback_dev, 2);
4696
4697         /*
4698          *      Check hash table.
4699          */
4700         spin_lock_bh(&addrconf_hash_lock);
4701         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4702                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4703         spin_unlock_bh(&addrconf_hash_lock);
4704
4705         del_timer(&addr_chk_timer);
4706         rtnl_unlock();
4707 }