net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 104 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 105 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 106
 107 #ifdef CONFIG_SYSCTL
 108 static void addrconf_sysctl_register(struct inet6_dev *idev);
 109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 110 #else
 111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 112 {
 113 }
 114
 115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 116 {
 117 }
 118 #endif
 119
 120 #ifdef CONFIG_IPV6_PRIVACY
 121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 123 static void ipv6_regen_rndid(unsigned long data);
 124
 125 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 126 #endif
 127
 128 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 129 static int ipv6_count_addresses(struct inet6_dev *idev);
 130
 131 /*
 132  *      Configured unicast address hash table
 133  */
 134 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 135 static DEFINE_SPINLOCK(addrconf_hash_lock);
 136
 137 static void addrconf_verify(unsigned long);
 138
 139 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 140 static DEFINE_SPINLOCK(addrconf_verify_lock);
 141
 142 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 143 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 144
 145 static void addrconf_type_change(struct net_device *dev,
 146                                  unsigned long event);
 147 static int addrconf_ifdown(struct net_device *dev, int how);
 148
 149 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 150 static void addrconf_dad_timer(unsigned long data);
 151 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 152 static void addrconf_dad_run(struct inet6_dev *idev);
 153 static void addrconf_rs_timer(unsigned long data);
 154 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 155 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 156
 157 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 158                                 struct prefix_info *pinfo);
 159 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 160                                struct net_device *dev);
 161
 162 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 163
 164 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 165         .forwarding             = 0,
 166         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 167         .mtu6                   = IPV6_MIN_MTU,
 168         .accept_ra              = 1,
 169         .accept_redirects       = 1,
 170         .autoconf               = 1,
 171         .force_mld_version      = 0,
 172         .dad_transmits          = 1,
 173         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 174         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 175         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 176 #ifdef CONFIG_IPV6_PRIVACY
 177         .use_tempaddr           = 0,
 178         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 179         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 180         .regen_max_retry        = REGEN_MAX_RETRY,
 181         .max_desync_factor      = MAX_DESYNC_FACTOR,
 182 #endif
 183         .max_addresses          = IPV6_MAX_ADDRESSES,
 184         .accept_ra_defrtr       = 1,
 185         .accept_ra_pinfo        = 1,
 186 #ifdef CONFIG_IPV6_ROUTER_PREF
 187         .accept_ra_rtr_pref     = 1,
 188         .rtr_probe_interval     = 60 * HZ,
 189 #ifdef CONFIG_IPV6_ROUTE_INFO
 190         .accept_ra_rt_info_max_plen = 0,
 191 #endif
 192 #endif
 193         .proxy_ndp              = 0,
 194         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 195         .disable_ipv6           = 0,
 196         .accept_dad             = 1,
 197 };
 198
 199 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 200         .forwarding             = 0,
 201         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 202         .mtu6                   = IPV6_MIN_MTU,
 203         .accept_ra              = 1,
 204         .accept_redirects       = 1,
 205         .autoconf               = 1,
 206         .dad_transmits          = 1,
 207         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 208         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 209         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 210 #ifdef CONFIG_IPV6_PRIVACY
 211         .use_tempaddr           = 0,
 212         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 213         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 214         .regen_max_retry        = REGEN_MAX_RETRY,
 215         .max_desync_factor      = MAX_DESYNC_FACTOR,
 216 #endif
 217         .max_addresses          = IPV6_MAX_ADDRESSES,
 218         .accept_ra_defrtr       = 1,
 219         .accept_ra_pinfo        = 1,
 220 #ifdef CONFIG_IPV6_ROUTER_PREF
 221         .accept_ra_rtr_pref     = 1,
 222         .rtr_probe_interval     = 60 * HZ,
 223 #ifdef CONFIG_IPV6_ROUTE_INFO
 224         .accept_ra_rt_info_max_plen = 0,
 225 #endif
 226 #endif
 227         .proxy_ndp              = 0,
 228         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 229         .disable_ipv6           = 0,
 230         .accept_dad             = 1,
 231 };
 232
 233 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 234 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 235 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 236 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 237 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 238
 239 /* Check if a valid qdisc is available */
 240 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 241 {
 242         return !qdisc_tx_is_noop(dev);
 243 }
 244
 245 /* Check if a route is valid prefix route */
 246 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 247 {
 248         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 249 }
 250
 251 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 252 {
 253         if (del_timer(&ifp->timer))
 254                 __in6_ifa_put(ifp);
 255 }
 256
 257 enum addrconf_timer_t {
 258         AC_NONE,
 259         AC_DAD,
 260         AC_RS,
 261 };
 262
 263 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 264                                enum addrconf_timer_t what,
 265                                unsigned long when)
 266 {
 267         if (!del_timer(&ifp->timer))
 268                 in6_ifa_hold(ifp);
 269
 270         switch (what) {
 271         case AC_DAD:
 272                 ifp->timer.function = addrconf_dad_timer;
 273                 break;
 274         case AC_RS:
 275                 ifp->timer.function = addrconf_rs_timer;
 276                 break;
 277         default:
 278                 break;
 279         }
 280         ifp->timer.expires = jiffies + when;
 281         add_timer(&ifp->timer);
 282 }
 283
 284 static int snmp6_alloc_dev(struct inet6_dev *idev)
 285 {
 286         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 287                           sizeof(struct ipstats_mib)) < 0)
 288                 goto err_ip;
 289         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 290                           sizeof(struct icmpv6_mib)) < 0)
 291                 goto err_icmp;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 293                           sizeof(struct icmpv6msg_mib)) < 0)
 294                 goto err_icmpmsg;
 295
 296         return 0;
 297
 298 err_icmpmsg:
 299         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 300 err_icmp:
 301         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 302 err_ip:
 303         return -ENOMEM;
 304 }
 305
 306 static void snmp6_free_dev(struct inet6_dev *idev)
 307 {
 308         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 309         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 311 }
 312
 313 /* Nobody refers to this device, we may destroy it. */
 314
 315 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 316 {
 317         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 318         kfree(idev);
 319 }
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 ndev->dead = 1;
 379                 in6_dev_finish_destroy(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 printk(KERN_INFO
 419                        "%s: Disabled Privacy Extensions\n",
 420                        dev->name);
 421                 ndev->cnf.use_tempaddr = -1;
 422         } else {
 423                 in6_dev_hold(ndev);
 424                 ipv6_regen_rndid((unsigned long) ndev);
 425         }
 426 #endif
 427
 428         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 429                 ndev->if_flags |= IF_READY;
 430
 431         ipv6_mc_init_dev(ndev);
 432         ndev->tstamp = jiffies;
 433         addrconf_sysctl_register(ndev);
 434         /* protected by rtnl_lock */
 435         rcu_assign_pointer(dev->ip6_ptr, ndev);
 436
 437         /* Join all-node multicast group */
 438         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 439
 440         return ndev;
 441 }
 442
 443 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 444 {
 445         struct inet6_dev *idev;
 446
 447         ASSERT_RTNL();
 448
 449         idev = __in6_dev_get(dev);
 450         if (!idev) {
 451                 idev = ipv6_add_dev(dev);
 452                 if (!idev)
 453                         return NULL;
 454         }
 455
 456         if (dev->flags&IFF_UP)
 457                 ipv6_mc_up(idev);
 458         return idev;
 459 }
 460
 461 #ifdef CONFIG_SYSCTL
 462 static void dev_forward_change(struct inet6_dev *idev)
 463 {
 464         struct net_device *dev;
 465         struct inet6_ifaddr *ifa;
 466
 467         if (!idev)
 468                 return;
 469         dev = idev->dev;
 470         if (idev->cnf.forwarding)
 471                 dev_disable_lro(dev);
 472         if (dev && (dev->flags & IFF_MULTICAST)) {
 473                 if (idev->cnf.forwarding)
 474                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 475                 else
 476                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 477         }
 478
 479         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 480                 if (ifa->flags&IFA_F_TENTATIVE)
 481                         continue;
 482                 if (idev->cnf.forwarding)
 483                         addrconf_join_anycast(ifa);
 484                 else
 485                         addrconf_leave_anycast(ifa);
 486         }
 487 }
 488
 489
 490 static void addrconf_forward_change(struct net *net, __s32 newf)
 491 {
 492         struct net_device *dev;
 493         struct inet6_dev *idev;
 494
 495         rcu_read_lock();
 496         for_each_netdev_rcu(net, dev) {
 497                 idev = __in6_dev_get(dev);
 498                 if (idev) {
 499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 500                         idev->cnf.forwarding = newf;
 501                         if (changed)
 502                                 dev_forward_change(idev);
 503                 }
 504         }
 505         rcu_read_unlock();
 506 }
 507
 508 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 509 {
 510         struct net *net;
 511
 512         net = (struct net *)table->extra2;
 513         if (p == &net->ipv6.devconf_dflt->forwarding)
 514                 return 0;
 515
 516         if (!rtnl_trylock()) {
 517                 /* Restore the original values before restarting */
 518                 *p = old;
 519                 return restart_syscall();
 520         }
 521
 522         if (p == &net->ipv6.devconf_all->forwarding) {
 523                 __s32 newf = net->ipv6.devconf_all->forwarding;
 524                 net->ipv6.devconf_dflt->forwarding = newf;
 525                 addrconf_forward_change(net, newf);
 526         } else if ((!*p) ^ (!old))
 527                 dev_forward_change((struct inet6_dev *)table->extra1);
 528         rtnl_unlock();
 529
 530         if (*p)
 531                 rt6_purge_dflt_routers(net);
 532         return 1;
 533 }
 534 #endif
 535
 536 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 537 {
 538         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 539         kfree(ifp);
 540 }
 541
 542 /* Nobody refers to this ifaddr, destroy it */
 543 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 544 {
 545         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 546
 547 #ifdef NET_REFCNT_DEBUG
 548         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 549 #endif
 550
 551         in6_dev_put(ifp->idev);
 552
 553         if (del_timer(&ifp->timer))
 554                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 555
 556         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 557                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 558                 return;
 559         }
 560         dst_release(&ifp->rt->u.dst);
 561
 562         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 563 }
 564
 565 static void
 566 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 567 {
 568         struct list_head *p;
 569         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 570
 571         /*
 572          * Each device address list is sorted in order of scope -
 573          * global before linklocal.
 574          */
 575         list_for_each(p, &idev->addr_list) {
 576                 struct inet6_ifaddr *ifa
 577                         = list_entry(p, struct inet6_ifaddr, if_list);
 578                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 579                         break;
 580         }
 581
 582         list_add_tail(&ifp->if_list, p);
 583 }
 584
 585 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 586 {
 587         /*
 588          * We perform the hash function over the last 64 bits of the address
 589          * This will include the IEEE address token on links that support it.
 590          */
 591         return jhash_2words((__force u32)addr->s6_addr32[2],
 592                             (__force u32)addr->s6_addr32[3], 0)
 593                 & (IN6_ADDR_HSIZE - 1);
 594 }
 595
 596 /* On success it returns ifp with increased reference count */
 597
 598 static struct inet6_ifaddr *
 599 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 600               int scope, u32 flags)
 601 {
 602         struct inet6_ifaddr *ifa = NULL;
 603         struct rt6_info *rt;
 604         unsigned int hash;
 605         int err = 0;
 606         int addr_type = ipv6_addr_type(addr);
 607
 608         if (addr_type == IPV6_ADDR_ANY ||
 609             addr_type & IPV6_ADDR_MULTICAST ||
 610             (!(idev->dev->flags & IFF_LOOPBACK) &&
 611              addr_type & IPV6_ADDR_LOOPBACK))
 612                 return ERR_PTR(-EADDRNOTAVAIL);
 613
 614         rcu_read_lock_bh();
 615         if (idev->dead) {
 616                 err = -ENODEV;                  /*XXX*/
 617                 goto out2;
 618         }
 619
 620         if (idev->cnf.disable_ipv6) {
 621                 err = -EACCES;
 622                 goto out2;
 623         }
 624
 625         spin_lock(&addrconf_hash_lock);
 626
 627         /* Ignore adding duplicate addresses on an interface */
 628         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 629                 ADBG(("ipv6_add_addr: already assigned\n"));
 630                 err = -EEXIST;
 631                 goto out;
 632         }
 633
 634         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 635
 636         if (ifa == NULL) {
 637                 ADBG(("ipv6_add_addr: malloc failed\n"));
 638                 err = -ENOBUFS;
 639                 goto out;
 640         }
 641
 642         rt = addrconf_dst_alloc(idev, addr, 0);
 643         if (IS_ERR(rt)) {
 644                 err = PTR_ERR(rt);
 645                 goto out;
 646         }
 647
 648         ipv6_addr_copy(&ifa->addr, addr);
 649
 650         spin_lock_init(&ifa->lock);
 651         spin_lock_init(&ifa->state_lock);
 652         init_timer(&ifa->timer);
 653         INIT_HLIST_NODE(&ifa->addr_lst);
 654         ifa->timer.data = (unsigned long) ifa;
 655         ifa->scope = scope;
 656         ifa->prefix_len = pfxlen;
 657         ifa->flags = flags | IFA_F_TENTATIVE;
 658         ifa->cstamp = ifa->tstamp = jiffies;
 659
 660         ifa->rt = rt;
 661
 662         /*
 663          * part one of RFC 4429, section 3.3
 664          * We should not configure an address as
 665          * optimistic if we do not yet know the link
 666          * layer address of our nexhop router
 667          */
 668
 669         if (rt->rt6i_nexthop == NULL)
 670                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 671
 672         ifa->idev = idev;
 673         in6_dev_hold(idev);
 674         /* For caller */
 675         in6_ifa_hold(ifa);
 676
 677         /* Add to big hash table */
 678         hash = ipv6_addr_hash(addr);
 679
 680         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 681         spin_unlock(&addrconf_hash_lock);
 682
 683         write_lock(&idev->lock);
 684         /* Add to inet6_dev unicast addr list. */
 685         ipv6_link_dev_addr(idev, ifa);
 686
 687 #ifdef CONFIG_IPV6_PRIVACY
 688         if (ifa->flags&IFA_F_TEMPORARY) {
 689                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 690                 in6_ifa_hold(ifa);
 691         }
 692 #endif
 693
 694         in6_ifa_hold(ifa);
 695         write_unlock(&idev->lock);
 696 out2:
 697         rcu_read_unlock_bh();
 698
 699         if (likely(err == 0))
 700                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 701         else {
 702                 kfree(ifa);
 703                 ifa = ERR_PTR(err);
 704         }
 705
 706         return ifa;
 707 out:
 708         spin_unlock(&addrconf_hash_lock);
 709         goto out2;
 710 }
 711
 712 /* This function wants to get referenced ifp and releases it before return */
 713
 714 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 715 {
 716         struct inet6_ifaddr *ifa, *ifn;
 717         struct inet6_dev *idev = ifp->idev;
 718         int state;
 719         int hash;
 720         int deleted = 0, onlink = 0;
 721         unsigned long expires = jiffies;
 722
 723         hash = ipv6_addr_hash(&ifp->addr);
 724
 725         spin_lock_bh(&ifp->state_lock);
 726         state = ifp->state;
 727         ifp->state = INET6_IFADDR_STATE_DEAD;
 728         spin_unlock_bh(&ifp->state_lock);
 729
 730         if (state == INET6_IFADDR_STATE_DEAD)
 731                 goto out;
 732
 733         spin_lock_bh(&addrconf_hash_lock);
 734         hlist_del_init_rcu(&ifp->addr_lst);
 735         spin_unlock_bh(&addrconf_hash_lock);
 736
 737         write_lock_bh(&idev->lock);
 738 #ifdef CONFIG_IPV6_PRIVACY
 739         if (ifp->flags&IFA_F_TEMPORARY) {
 740                 list_del(&ifp->tmp_list);
 741                 if (ifp->ifpub) {
 742                         in6_ifa_put(ifp->ifpub);
 743                         ifp->ifpub = NULL;
 744                 }
 745                 __in6_ifa_put(ifp);
 746         }
 747 #endif
 748
 749         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 750                 if (ifa == ifp) {
 751                         list_del_init(&ifp->if_list);
 752                         __in6_ifa_put(ifp);
 753
 754                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 755                                 break;
 756                         deleted = 1;
 757                         continue;
 758                 } else if (ifp->flags & IFA_F_PERMANENT) {
 759                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 760                                               ifp->prefix_len)) {
 761                                 if (ifa->flags & IFA_F_PERMANENT) {
 762                                         onlink = 1;
 763                                         if (deleted)
 764                                                 break;
 765                                 } else {
 766                                         unsigned long lifetime;
 767
 768                                         if (!onlink)
 769                                                 onlink = -1;
 770
 771                                         spin_lock(&ifa->lock);
 772
 773                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 774                                         /*
 775                                          * Note: Because this address is
 776                                          * not permanent, lifetime <
 777                                          * LONG_MAX / HZ here.
 778                                          */
 779                                         if (time_before(expires,
 780                                                         ifa->tstamp + lifetime * HZ))
 781                                                 expires = ifa->tstamp + lifetime * HZ;
 782                                         spin_unlock(&ifa->lock);
 783                                 }
 784                         }
 785                 }
 786         }
 787         write_unlock_bh(&idev->lock);
 788
 789         addrconf_del_timer(ifp);
 790
 791         ipv6_ifa_notify(RTM_DELADDR, ifp);
 792
 793         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 794
 795         /*
 796          * Purge or update corresponding prefix
 797          *
 798          * 1) we don't purge prefix here if address was not permanent.
 799          *    prefix is managed by its own lifetime.
 800          * 2) if there're no addresses, delete prefix.
 801          * 3) if there're still other permanent address(es),
 802          *    corresponding prefix is still permanent.
 803          * 4) otherwise, update prefix lifetime to the
 804          *    longest valid lifetime among the corresponding
 805          *    addresses on the device.
 806          *    Note: subsequent RA will update lifetime.
 807          *
 808          * --yoshfuji
 809          */
 810         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 811                 struct in6_addr prefix;
 812                 struct rt6_info *rt;
 813                 struct net *net = dev_net(ifp->idev->dev);
 814                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 815                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 816
 817                 if (rt && addrconf_is_prefix_route(rt)) {
 818                         if (onlink == 0) {
 819                                 ip6_del_rt(rt);
 820                                 rt = NULL;
 821                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 822                                 rt->rt6i_expires = expires;
 823                                 rt->rt6i_flags |= RTF_EXPIRES;
 824                         }
 825                 }
 826                 dst_release(&rt->u.dst);
 827         }
 828
 829 out:
 830         in6_ifa_put(ifp);
 831 }
 832
 833 #ifdef CONFIG_IPV6_PRIVACY
 834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 835 {
 836         struct inet6_dev *idev = ifp->idev;
 837         struct in6_addr addr, *tmpaddr;
 838         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 839         unsigned long regen_advance;
 840         int tmp_plen;
 841         int ret = 0;
 842         int max_addresses;
 843         u32 addr_flags;
 844
 845         write_lock(&idev->lock);
 846         if (ift) {
 847                 spin_lock_bh(&ift->lock);
 848                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 849                 spin_unlock_bh(&ift->lock);
 850                 tmpaddr = &addr;
 851         } else {
 852                 tmpaddr = NULL;
 853         }
 854 retry:
 855         in6_dev_hold(idev);
 856         if (idev->cnf.use_tempaddr <= 0) {
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_INFO
 859                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         spin_lock_bh(&ifp->lock);
 865         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 866                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         in6_ifa_hold(ifp);
 876         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 877         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 878                 spin_unlock_bh(&ifp->lock);
 879                 write_unlock(&idev->lock);
 880                 printk(KERN_WARNING
 881                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 882                 in6_ifa_put(ifp);
 883                 in6_dev_put(idev);
 884                 ret = -1;
 885                 goto out;
 886         }
 887         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 894         tmp_plen = ifp->prefix_len;
 895         max_addresses = idev->cnf.max_addresses;
 896         tmp_cstamp = ifp->cstamp;
 897         tmp_tstamp = ifp->tstamp;
 898         spin_unlock_bh(&ifp->lock);
 899
 900         regen_advance = idev->cnf.regen_max_retry *
 901                         idev->cnf.dad_transmits *
 902                         idev->nd_parms->retrans_time / HZ;
 903         write_unlock(&idev->lock);
 904
 905         /* A temporary address is created only if this calculated Preferred
 906          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 907          * an implementation must not create a temporary address with a zero
 908          * Preferred Lifetime.
 909          */
 910         if (tmp_prefered_lft <= regen_advance) {
 911                 in6_ifa_put(ifp);
 912                 in6_dev_put(idev);
 913                 ret = -1;
 914                 goto out;
 915         }
 916
 917         addr_flags = IFA_F_TEMPORARY;
 918         /* set in addrconf_prefix_rcv() */
 919         if (ifp->flags & IFA_F_OPTIMISTIC)
 920                 addr_flags |= IFA_F_OPTIMISTIC;
 921
 922         ift = !max_addresses ||
 923               ipv6_count_addresses(idev) < max_addresses ?
 924                 ipv6_add_addr(idev, &addr, tmp_plen,
 925                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 926                               addr_flags) : NULL;
 927         if (!ift || IS_ERR(ift)) {
 928                 in6_ifa_put(ifp);
 929                 in6_dev_put(idev);
 930                 printk(KERN_INFO
 931                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 932                 tmpaddr = &addr;
 933                 write_lock(&idev->lock);
 934                 goto retry;
 935         }
 936
 937         spin_lock_bh(&ift->lock);
 938         ift->ifpub = ifp;
 939         ift->valid_lft = tmp_valid_lft;
 940         ift->prefered_lft = tmp_prefered_lft;
 941         ift->cstamp = tmp_cstamp;
 942         ift->tstamp = tmp_tstamp;
 943         spin_unlock_bh(&ift->lock);
 944
 945         addrconf_dad_start(ift, 0);
 946         in6_ifa_put(ift);
 947         in6_dev_put(idev);
 948 out:
 949         return ret;
 950 }
 951 #endif
 952
 953 /*
 954  *      Choose an appropriate source address (RFC3484)
 955  */
 956 enum {
 957         IPV6_SADDR_RULE_INIT = 0,
 958         IPV6_SADDR_RULE_LOCAL,
 959         IPV6_SADDR_RULE_SCOPE,
 960         IPV6_SADDR_RULE_PREFERRED,
 961 #ifdef CONFIG_IPV6_MIP6
 962         IPV6_SADDR_RULE_HOA,
 963 #endif
 964         IPV6_SADDR_RULE_OIF,
 965         IPV6_SADDR_RULE_LABEL,
 966 #ifdef CONFIG_IPV6_PRIVACY
 967         IPV6_SADDR_RULE_PRIVACY,
 968 #endif
 969         IPV6_SADDR_RULE_ORCHID,
 970         IPV6_SADDR_RULE_PREFIX,
 971         IPV6_SADDR_RULE_MAX
 972 };
 973
 974 struct ipv6_saddr_score {
 975         int                     rule;
 976         int                     addr_type;
 977         struct inet6_ifaddr     *ifa;
 978         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 979         int                     scopedist;
 980         int                     matchlen;
 981 };
 982
 983 struct ipv6_saddr_dst {
 984         const struct in6_addr *addr;
 985         int ifindex;
 986         int scope;
 987         int label;
 988         unsigned int prefs;
 989 };
 990
 991 static inline int ipv6_saddr_preferred(int type)
 992 {
 993         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 994                 return 1;
 995         return 0;
 996 }
 997
 998 static int ipv6_get_saddr_eval(struct net *net,
 999                                struct ipv6_saddr_score *score,
1000                                struct ipv6_saddr_dst *dst,
1001                                int i)
1002 {
1003         int ret;
1004
1005         if (i <= score->rule) {
1006                 switch (i) {
1007                 case IPV6_SADDR_RULE_SCOPE:
1008                         ret = score->scopedist;
1009                         break;
1010                 case IPV6_SADDR_RULE_PREFIX:
1011                         ret = score->matchlen;
1012                         break;
1013                 default:
1014                         ret = !!test_bit(i, score->scorebits);
1015                 }
1016                 goto out;
1017         }
1018
1019         switch (i) {
1020         case IPV6_SADDR_RULE_INIT:
1021                 /* Rule 0: remember if hiscore is not ready yet */
1022                 ret = !!score->ifa;
1023                 break;
1024         case IPV6_SADDR_RULE_LOCAL:
1025                 /* Rule 1: Prefer same address */
1026                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027                 break;
1028         case IPV6_SADDR_RULE_SCOPE:
1029                 /* Rule 2: Prefer appropriate scope
1030                  *
1031                  *      ret
1032                  *       ^
1033                  *    -1 |  d 15
1034                  *    ---+--+-+---> scope
1035                  *       |
1036                  *       |             d is scope of the destination.
1037                  *  B-d  |  \
1038                  *       |   \      <- smaller scope is better if
1039                  *  B-15 |    \        if scope is enough for destinaion.
1040                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1041                  * d-C-1 | /
1042                  *       |/         <- greater is better
1043                  *   -C  /             if scope is not enough for destination.
1044                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1045                  *
1046                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048                  * Assume B = 0 and we get C > 29.
1049                  */
1050                 ret = __ipv6_addr_src_scope(score->addr_type);
1051                 if (ret >= dst->scope)
1052                         ret = -ret;
1053                 else
1054                         ret -= 128;     /* 30 is enough */
1055                 score->scopedist = ret;
1056                 break;
1057         case IPV6_SADDR_RULE_PREFERRED:
1058                 /* Rule 3: Avoid deprecated and optimistic addresses */
1059                 ret = ipv6_saddr_preferred(score->addr_type) ||
1060                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061                 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063         case IPV6_SADDR_RULE_HOA:
1064             {
1065                 /* Rule 4: Prefer home address */
1066                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068                 break;
1069             }
1070 #endif
1071         case IPV6_SADDR_RULE_OIF:
1072                 /* Rule 5: Prefer outgoing interface */
1073                 ret = (!dst->ifindex ||
1074                        dst->ifindex == score->ifa->idev->dev->ifindex);
1075                 break;
1076         case IPV6_SADDR_RULE_LABEL:
1077                 /* Rule 6: Prefer matching label */
1078                 ret = ipv6_addr_label(net,
1079                                       &score->ifa->addr, score->addr_type,
1080                                       score->ifa->idev->dev->ifindex) == dst->label;
1081                 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083         case IPV6_SADDR_RULE_PRIVACY:
1084             {
1085                 /* Rule 7: Prefer public address
1086                  * Note: prefer temprary address if use_tempaddr >= 2
1087                  */
1088                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1091                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092                 break;
1093             }
1094 #endif
1095         case IPV6_SADDR_RULE_ORCHID:
1096                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097                  *          non-ORCHID vs non-ORCHID
1098                  */
1099                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100                         ipv6_addr_orchid(dst->addr));
1101                 break;
1102         case IPV6_SADDR_RULE_PREFIX:
1103                 /* Rule 8: Use longest matching prefix */
1104                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105                                                        dst->addr);
1106                 break;
1107         default:
1108                 ret = 0;
1109         }
1110
1111         if (ret)
1112                 __set_bit(i, score->scorebits);
1113         score->rule = i;
1114 out:
1115         return ret;
1116 }
1117
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119                        const struct in6_addr *daddr, unsigned int prefs,
1120                        struct in6_addr *saddr)
1121 {
1122         struct ipv6_saddr_score scores[2],
1123                                 *score = &scores[0], *hiscore = &scores[1];
1124         struct ipv6_saddr_dst dst;
1125         struct net_device *dev;
1126         int dst_type;
1127
1128         dst_type = __ipv6_addr_type(daddr);
1129         dst.addr = daddr;
1130         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131         dst.scope = __ipv6_addr_src_scope(dst_type);
1132         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133         dst.prefs = prefs;
1134
1135         hiscore->rule = -1;
1136         hiscore->ifa = NULL;
1137
1138         rcu_read_lock();
1139
1140         for_each_netdev_rcu(net, dev) {
1141                 struct inet6_dev *idev;
1142
1143                 /* Candidate Source Address (section 4)
1144                  *  - multicast and link-local destination address,
1145                  *    the set of candidate source address MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same link as the outgoing
1148                  *    interface.
1149                  * (- For site-local destination addresses, the
1150                  *    set of candidate source addresses MUST only
1151                  *    include addresses assigned to interfaces
1152                  *    belonging to the same site as the outgoing
1153                  *    interface.)
1154                  */
1155                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1156                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1157                     dst.ifindex && dev->ifindex != dst.ifindex)
1158                         continue;
1159
1160                 idev = __in6_dev_get(dev);
1161                 if (!idev)
1162                         continue;
1163
1164                 read_lock_bh(&idev->lock);
1165                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1166                         int i;
1167
1168                         /*
1169                          * - Tentative Address (RFC2462 section 5.4)
1170                          *  - A tentative address is not considered
1171                          *    "assigned to an interface" in the traditional
1172                          *    sense, unless it is also flagged as optimistic.
1173                          * - Candidate Source Address (section 4)
1174                          *  - In any case, anycast addresses, multicast
1175                          *    addresses, and the unspecified address MUST
1176                          *    NOT be included in a candidate set.
1177                          */
1178                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1179                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1180                                 continue;
1181
1182                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1183
1184                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1185                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1186                                 LIMIT_NETDEBUG(KERN_DEBUG
1187                                                "ADDRCONF: unspecified / multicast address "
1188                                                "assigned as unicast address on %s",
1189                                                dev->name);
1190                                 continue;
1191                         }
1192
1193                         score->rule = -1;
1194                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1195
1196                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1197                                 int minihiscore, miniscore;
1198
1199                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1200                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1201
1202                                 if (minihiscore > miniscore) {
1203                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1204                                             score->scopedist > 0) {
1205                                                 /*
1206                                                  * special case:
1207                                                  * each remaining entry
1208                                                  * has too small (not enough)
1209                                                  * scope, because ifa entries
1210                                                  * are sorted by their scope
1211                                                  * values.
1212                                                  */
1213                                                 goto try_nextdev;
1214                                         }
1215                                         break;
1216                                 } else if (minihiscore < miniscore) {
1217                                         if (hiscore->ifa)
1218                                                 in6_ifa_put(hiscore->ifa);
1219
1220                                         in6_ifa_hold(score->ifa);
1221
1222                                         swap(hiscore, score);
1223
1224                                         /* restore our iterator */
1225                                         score->ifa = hiscore->ifa;
1226
1227                                         break;
1228                                 }
1229                         }
1230                 }
1231 try_nextdev:
1232                 read_unlock_bh(&idev->lock);
1233         }
1234         rcu_read_unlock();
1235
1236         if (!hiscore->ifa)
1237                 return -EADDRNOTAVAIL;
1238
1239         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240         in6_ifa_put(hiscore->ifa);
1241         return 0;
1242 }
1243 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1244
1245 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1246                     unsigned char banned_flags)
1247 {
1248         struct inet6_dev *idev;
1249         int err = -EADDRNOTAVAIL;
1250
1251         rcu_read_lock();
1252         idev = __in6_dev_get(dev);
1253         if (idev) {
1254                 struct inet6_ifaddr *ifp;
1255
1256                 read_lock_bh(&idev->lock);
1257                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1258                         if (ifp->scope == IFA_LINK &&
1259                             !(ifp->flags & banned_flags)) {
1260                                 ipv6_addr_copy(addr, &ifp->addr);
1261                                 err = 0;
1262                                 break;
1263                         }
1264                 }
1265                 read_unlock_bh(&idev->lock);
1266         }
1267         rcu_read_unlock();
1268         return err;
1269 }
1270
1271 static int ipv6_count_addresses(struct inet6_dev *idev)
1272 {
1273         int cnt = 0;
1274         struct inet6_ifaddr *ifp;
1275
1276         read_lock_bh(&idev->lock);
1277         list_for_each_entry(ifp, &idev->addr_list, if_list)
1278                 cnt++;
1279         read_unlock_bh(&idev->lock);
1280         return cnt;
1281 }
1282
1283 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1284                   struct net_device *dev, int strict)
1285 {
1286         struct inet6_ifaddr *ifp;
1287         struct hlist_node *node;
1288         unsigned int hash = ipv6_addr_hash(addr);
1289
1290         rcu_read_lock_bh();
1291         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1292                 if (!net_eq(dev_net(ifp->idev->dev), net))
1293                         continue;
1294                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295                     !(ifp->flags&IFA_F_TENTATIVE) &&
1296                     (dev == NULL || ifp->idev->dev == dev ||
1297                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1298                         rcu_read_unlock_bh();
1299                         return 1;
1300                 }
1301         }
1302
1303         rcu_read_unlock_bh();
1304         return 0;
1305 }
1306 EXPORT_SYMBOL(ipv6_chk_addr);
1307
1308 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                                struct net_device *dev)
1310 {
1311         unsigned int hash = ipv6_addr_hash(addr);
1312         struct inet6_ifaddr *ifp;
1313         struct hlist_node *node;
1314
1315         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1316                 if (!net_eq(dev_net(ifp->idev->dev), net))
1317                         continue;
1318                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1319                         if (dev == NULL || ifp->idev->dev == dev)
1320                                 return true;
1321                 }
1322         }
1323         return false;
1324 }
1325
1326 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1327 {
1328         struct inet6_dev *idev;
1329         struct inet6_ifaddr *ifa;
1330         int     onlink;
1331
1332         onlink = 0;
1333         rcu_read_lock();
1334         idev = __in6_dev_get(dev);
1335         if (idev) {
1336                 read_lock_bh(&idev->lock);
1337                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1338                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1339                                                    ifa->prefix_len);
1340                         if (onlink)
1341                                 break;
1342                 }
1343                 read_unlock_bh(&idev->lock);
1344         }
1345         rcu_read_unlock();
1346         return onlink;
1347 }
1348
1349 EXPORT_SYMBOL(ipv6_chk_prefix);
1350
1351 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1352                                      struct net_device *dev, int strict)
1353 {
1354         struct inet6_ifaddr *ifp, *result = NULL;
1355         unsigned int hash = ipv6_addr_hash(addr);
1356         struct hlist_node *node;
1357
1358         rcu_read_lock_bh();
1359         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1360                 if (!net_eq(dev_net(ifp->idev->dev), net))
1361                         continue;
1362                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1363                         if (dev == NULL || ifp->idev->dev == dev ||
1364                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1365                                 result = ifp;
1366                                 in6_ifa_hold(ifp);
1367                                 break;
1368                         }
1369                 }
1370         }
1371         rcu_read_unlock_bh();
1372
1373         return result;
1374 }
1375
1376 /* Gets referenced address, destroys ifaddr */
1377
1378 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1379 {
1380         if (ifp->flags&IFA_F_PERMANENT) {
1381                 spin_lock_bh(&ifp->lock);
1382                 addrconf_del_timer(ifp);
1383                 ifp->flags |= IFA_F_TENTATIVE;
1384                 if (dad_failed)
1385                         ifp->flags |= IFA_F_DADFAILED;
1386                 spin_unlock_bh(&ifp->lock);
1387                 if (dad_failed)
1388                         ipv6_ifa_notify(0, ifp);
1389                 in6_ifa_put(ifp);
1390 #ifdef CONFIG_IPV6_PRIVACY
1391         } else if (ifp->flags&IFA_F_TEMPORARY) {
1392                 struct inet6_ifaddr *ifpub;
1393                 spin_lock_bh(&ifp->lock);
1394                 ifpub = ifp->ifpub;
1395                 if (ifpub) {
1396                         in6_ifa_hold(ifpub);
1397                         spin_unlock_bh(&ifp->lock);
1398                         ipv6_create_tempaddr(ifpub, ifp);
1399                         in6_ifa_put(ifpub);
1400                 } else {
1401                         spin_unlock_bh(&ifp->lock);
1402                 }
1403                 ipv6_del_addr(ifp);
1404 #endif
1405         } else
1406                 ipv6_del_addr(ifp);
1407 }
1408
1409 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1410 {
1411         int err = -ENOENT;
1412
1413         spin_lock(&ifp->state_lock);
1414         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1415                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1416                 err = 0;
1417         }
1418         spin_unlock(&ifp->state_lock);
1419
1420         return err;
1421 }
1422
1423 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1424 {
1425         struct inet6_dev *idev = ifp->idev;
1426
1427         if (addrconf_dad_end(ifp))
1428                 return;
1429
1430         if (net_ratelimit())
1431                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1432                         ifp->idev->dev->name, &ifp->addr);
1433
1434         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1435                 struct in6_addr addr;
1436
1437                 addr.s6_addr32[0] = htonl(0xfe800000);
1438                 addr.s6_addr32[1] = 0;
1439
1440                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1441                     ipv6_addr_equal(&ifp->addr, &addr)) {
1442                         /* DAD failed for link-local based on MAC address */
1443                         idev->cnf.disable_ipv6 = 1;
1444
1445                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1446                                 ifp->idev->dev->name);
1447                 }
1448         }
1449
1450         addrconf_dad_stop(ifp, 1);
1451 }
1452
1453 /* Join to solicited addr multicast group. */
1454
1455 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1456 {
1457         struct in6_addr maddr;
1458
1459         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1460                 return;
1461
1462         addrconf_addr_solict_mult(addr, &maddr);
1463         ipv6_dev_mc_inc(dev, &maddr);
1464 }
1465
1466 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1467 {
1468         struct in6_addr maddr;
1469
1470         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1471                 return;
1472
1473         addrconf_addr_solict_mult(addr, &maddr);
1474         __ipv6_dev_mc_dec(idev, &maddr);
1475 }
1476
1477 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1478 {
1479         struct in6_addr addr;
1480         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481         if (ipv6_addr_any(&addr))
1482                 return;
1483         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1484 }
1485
1486 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1487 {
1488         struct in6_addr addr;
1489         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490         if (ipv6_addr_any(&addr))
1491                 return;
1492         __ipv6_dev_ac_dec(ifp->idev, &addr);
1493 }
1494
1495 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1496 {
1497         if (dev->addr_len != ETH_ALEN)
1498                 return -1;
1499         memcpy(eui, dev->dev_addr, 3);
1500         memcpy(eui + 5, dev->dev_addr + 3, 3);
1501
1502         /*
1503          * The zSeries OSA network cards can be shared among various
1504          * OS instances, but the OSA cards have only one MAC address.
1505          * This leads to duplicate address conflicts in conjunction
1506          * with IPv6 if more than one instance uses the same card.
1507          *
1508          * The driver for these cards can deliver a unique 16-bit
1509          * identifier for each instance sharing the same card.  It is
1510          * placed instead of 0xFFFE in the interface identifier.  The
1511          * "u" bit of the interface identifier is not inverted in this
1512          * case.  Hence the resulting interface identifier has local
1513          * scope according to RFC2373.
1514          */
1515         if (dev->dev_id) {
1516                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1517                 eui[4] = dev->dev_id & 0xFF;
1518         } else {
1519                 eui[3] = 0xFF;
1520                 eui[4] = 0xFE;
1521                 eui[0] ^= 2;
1522         }
1523         return 0;
1524 }
1525
1526 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1527 {
1528         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1529         if (dev->addr_len != ARCNET_ALEN)
1530                 return -1;
1531         memset(eui, 0, 7);
1532         eui[7] = *(u8*)dev->dev_addr;
1533         return 0;
1534 }
1535
1536 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1537 {
1538         if (dev->addr_len != INFINIBAND_ALEN)
1539                 return -1;
1540         memcpy(eui, dev->dev_addr + 12, 8);
1541         eui[0] |= 2;
1542         return 0;
1543 }
1544
1545 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1546 {
1547         if (addr == 0)
1548                 return -1;
1549         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1550                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1551                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1552                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1553                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1554                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1555         eui[1] = 0;
1556         eui[2] = 0x5E;
1557         eui[3] = 0xFE;
1558         memcpy(eui + 4, &addr, 4);
1559         return 0;
1560 }
1561 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1562
1563 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1564 {
1565         if (dev->priv_flags & IFF_ISATAP)
1566                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1567         return -1;
1568 }
1569
1570 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1571 {
1572         switch (dev->type) {
1573         case ARPHRD_ETHER:
1574         case ARPHRD_FDDI:
1575         case ARPHRD_IEEE802_TR:
1576                 return addrconf_ifid_eui48(eui, dev);
1577         case ARPHRD_ARCNET:
1578                 return addrconf_ifid_arcnet(eui, dev);
1579         case ARPHRD_INFINIBAND:
1580                 return addrconf_ifid_infiniband(eui, dev);
1581         case ARPHRD_SIT:
1582                 return addrconf_ifid_sit(eui, dev);
1583         }
1584         return -1;
1585 }
1586
1587 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1588 {
1589         int err = -1;
1590         struct inet6_ifaddr *ifp;
1591
1592         read_lock_bh(&idev->lock);
1593         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1594                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1595                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1596                         err = 0;
1597                         break;
1598                 }
1599         }
1600         read_unlock_bh(&idev->lock);
1601         return err;
1602 }
1603
1604 #ifdef CONFIG_IPV6_PRIVACY
1605 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1606 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1607 {
1608 regen:
1609         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1610         idev->rndid[0] &= ~0x02;
1611
1612         /*
1613          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1614          * check if generated address is not inappropriate
1615          *
1616          *  - Reserved subnet anycast (RFC 2526)
1617          *      11111101 11....11 1xxxxxxx
1618          *  - ISATAP (RFC4214) 6.1
1619          *      00-00-5E-FE-xx-xx-xx-xx
1620          *  - value 0
1621          *  - XXX: already assigned to an address on the device
1622          */
1623         if (idev->rndid[0] == 0xfd &&
1624             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1625             (idev->rndid[7]&0x80))
1626                 goto regen;
1627         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1628                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1629                         goto regen;
1630                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1631                         goto regen;
1632         }
1633
1634         return 0;
1635 }
1636
1637 static void ipv6_regen_rndid(unsigned long data)
1638 {
1639         struct inet6_dev *idev = (struct inet6_dev *) data;
1640         unsigned long expires;
1641
1642         rcu_read_lock_bh();
1643         write_lock_bh(&idev->lock);
1644
1645         if (idev->dead)
1646                 goto out;
1647
1648         if (__ipv6_regen_rndid(idev) < 0)
1649                 goto out;
1650
1651         expires = jiffies +
1652                 idev->cnf.temp_prefered_lft * HZ -
1653                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1654         if (time_before(expires, jiffies)) {
1655                 printk(KERN_WARNING
1656                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1657                         idev->dev->name);
1658                 goto out;
1659         }
1660
1661         if (!mod_timer(&idev->regen_timer, expires))
1662                 in6_dev_hold(idev);
1663
1664 out:
1665         write_unlock_bh(&idev->lock);
1666         rcu_read_unlock_bh();
1667         in6_dev_put(idev);
1668 }
1669
1670 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1671         int ret = 0;
1672
1673         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674                 ret = __ipv6_regen_rndid(idev);
1675         return ret;
1676 }
1677 #endif
1678
1679 /*
1680  *      Add prefix route.
1681  */
1682
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685                       unsigned long expires, u32 flags)
1686 {
1687         struct fib6_config cfg = {
1688                 .fc_table = RT6_TABLE_PREFIX,
1689                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1690                 .fc_ifindex = dev->ifindex,
1691                 .fc_expires = expires,
1692                 .fc_dst_len = plen,
1693                 .fc_flags = RTF_UP | flags,
1694                 .fc_nlinfo.nl_net = dev_net(dev),
1695                 .fc_protocol = RTPROT_KERNEL,
1696         };
1697
1698         ipv6_addr_copy(&cfg.fc_dst, pfx);
1699
1700         /* Prevent useless cloning on PtP SIT.
1701            This thing is done here expecting that the whole
1702            class of non-broadcast devices need not cloning.
1703          */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706                 cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708
1709         ip6_route_add(&cfg);
1710 }
1711
1712 /* Create "default" multicast route to the interface */
1713
1714 static void addrconf_add_mroute(struct net_device *dev)
1715 {
1716         struct fib6_config cfg = {
1717                 .fc_table = RT6_TABLE_LOCAL,
1718                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1719                 .fc_ifindex = dev->ifindex,
1720                 .fc_dst_len = 8,
1721                 .fc_flags = RTF_UP,
1722                 .fc_nlinfo.nl_net = dev_net(dev),
1723         };
1724
1725         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1726
1727         ip6_route_add(&cfg);
1728 }
1729
1730 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1731 static void sit_route_add(struct net_device *dev)
1732 {
1733         struct fib6_config cfg = {
1734                 .fc_table = RT6_TABLE_MAIN,
1735                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1736                 .fc_ifindex = dev->ifindex,
1737                 .fc_dst_len = 96,
1738                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1739                 .fc_nlinfo.nl_net = dev_net(dev),
1740         };
1741
1742         /* prefix length - 96 bits "::d.d.d.d" */
1743         ip6_route_add(&cfg);
1744 }
1745 #endif
1746
1747 static void addrconf_add_lroute(struct net_device *dev)
1748 {
1749         struct in6_addr addr;
1750
1751         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1752         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1753 }
1754
1755 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1756 {
1757         struct inet6_dev *idev;
1758
1759         ASSERT_RTNL();
1760
1761         idev = ipv6_find_idev(dev);
1762         if (!idev)
1763                 return ERR_PTR(-ENOBUFS);
1764
1765         if (idev->cnf.disable_ipv6)
1766                 return ERR_PTR(-EACCES);
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->u.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         idev = addrconf_add_dev(dev);
2136         if (IS_ERR(idev))
2137                 return PTR_ERR(idev);
2138
2139         scope = ipv6_addr_scope(pfx);
2140
2141         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2142         if (addrconf_finite_timeout(timeout)) {
2143                 expires = jiffies_to_clock_t(timeout * HZ);
2144                 valid_lft = timeout;
2145                 flags = RTF_EXPIRES;
2146         } else {
2147                 expires = 0;
2148                 flags = 0;
2149                 ifa_flags |= IFA_F_PERMANENT;
2150         }
2151
2152         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2153         if (addrconf_finite_timeout(timeout)) {
2154                 if (timeout == 0)
2155                         ifa_flags |= IFA_F_DEPRECATED;
2156                 prefered_lft = timeout;
2157         }
2158
2159         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2160
2161         if (!IS_ERR(ifp)) {
2162                 spin_lock_bh(&ifp->lock);
2163                 ifp->valid_lft = valid_lft;
2164                 ifp->prefered_lft = prefered_lft;
2165                 ifp->tstamp = jiffies;
2166                 spin_unlock_bh(&ifp->lock);
2167
2168                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2169                                       expires, flags);
2170                 /*
2171                  * Note that section 3.1 of RFC 4429 indicates
2172                  * that the Optimistic flag should not be set for
2173                  * manually configured addresses
2174                  */
2175                 addrconf_dad_start(ifp, 0);
2176                 in6_ifa_put(ifp);
2177                 addrconf_verify(0);
2178                 return 0;
2179         }
2180
2181         return PTR_ERR(ifp);
2182 }
2183
2184 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2185                           unsigned int plen)
2186 {
2187         struct inet6_ifaddr *ifp;
2188         struct inet6_dev *idev;
2189         struct net_device *dev;
2190
2191         if (plen > 128)
2192                 return -EINVAL;
2193
2194         dev = __dev_get_by_index(net, ifindex);
2195         if (!dev)
2196                 return -ENODEV;
2197
2198         if ((idev = __in6_dev_get(dev)) == NULL)
2199                 return -ENXIO;
2200
2201         read_lock_bh(&idev->lock);
2202         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2203                 if (ifp->prefix_len == plen &&
2204                     ipv6_addr_equal(pfx, &ifp->addr)) {
2205                         in6_ifa_hold(ifp);
2206                         read_unlock_bh(&idev->lock);
2207
2208                         ipv6_del_addr(ifp);
2209
2210                         /* If the last address is deleted administratively,
2211                            disable IPv6 on this interface.
2212                          */
2213                         if (list_empty(&idev->addr_list))
2214                                 addrconf_ifdown(idev->dev, 1);
2215                         return 0;
2216                 }
2217         }
2218         read_unlock_bh(&idev->lock);
2219         return -EADDRNOTAVAIL;
2220 }
2221
2222
2223 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2224 {
2225         struct in6_ifreq ireq;
2226         int err;
2227
2228         if (!capable(CAP_NET_ADMIN))
2229                 return -EPERM;
2230
2231         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2232                 return -EFAULT;
2233
2234         rtnl_lock();
2235         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2236                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2237                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2238         rtnl_unlock();
2239         return err;
2240 }
2241
2242 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2243 {
2244         struct in6_ifreq ireq;
2245         int err;
2246
2247         if (!capable(CAP_NET_ADMIN))
2248                 return -EPERM;
2249
2250         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2251                 return -EFAULT;
2252
2253         rtnl_lock();
2254         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2255                              ireq.ifr6_prefixlen);
2256         rtnl_unlock();
2257         return err;
2258 }
2259
2260 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2261                      int plen, int scope)
2262 {
2263         struct inet6_ifaddr *ifp;
2264
2265         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2266         if (!IS_ERR(ifp)) {
2267                 spin_lock_bh(&ifp->lock);
2268                 ifp->flags &= ~IFA_F_TENTATIVE;
2269                 spin_unlock_bh(&ifp->lock);
2270                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2271                 in6_ifa_put(ifp);
2272         }
2273 }
2274
2275 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2276 static void sit_add_v4_addrs(struct inet6_dev *idev)
2277 {
2278         struct in6_addr addr;
2279         struct net_device *dev;
2280         struct net *net = dev_net(idev->dev);
2281         int scope;
2282
2283         ASSERT_RTNL();
2284
2285         memset(&addr, 0, sizeof(struct in6_addr));
2286         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2287
2288         if (idev->dev->flags&IFF_POINTOPOINT) {
2289                 addr.s6_addr32[0] = htonl(0xfe800000);
2290                 scope = IFA_LINK;
2291         } else {
2292                 scope = IPV6_ADDR_COMPATv4;
2293         }
2294
2295         if (addr.s6_addr32[3]) {
2296                 add_addr(idev, &addr, 128, scope);
2297                 return;
2298         }
2299
2300         for_each_netdev(net, dev) {
2301                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2302                 if (in_dev && (dev->flags & IFF_UP)) {
2303                         struct in_ifaddr * ifa;
2304
2305                         int flag = scope;
2306
2307                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2308                                 int plen;
2309
2310                                 addr.s6_addr32[3] = ifa->ifa_local;
2311
2312                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2313                                         continue;
2314                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2315                                         if (idev->dev->flags&IFF_POINTOPOINT)
2316                                                 continue;
2317                                         flag |= IFA_HOST;
2318                                 }
2319                                 if (idev->dev->flags&IFF_POINTOPOINT)
2320                                         plen = 64;
2321                                 else
2322                                         plen = 96;
2323
2324                                 add_addr(idev, &addr, plen, flag);
2325                         }
2326                 }
2327         }
2328 }
2329 #endif
2330
2331 static void init_loopback(struct net_device *dev)
2332 {
2333         struct inet6_dev  *idev;
2334
2335         /* ::1 */
2336
2337         ASSERT_RTNL();
2338
2339         if ((idev = ipv6_find_idev(dev)) == NULL) {
2340                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2341                 return;
2342         }
2343
2344         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2345 }
2346
2347 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2348 {
2349         struct inet6_ifaddr * ifp;
2350         u32 addr_flags = IFA_F_PERMANENT;
2351
2352 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2353         if (idev->cnf.optimistic_dad &&
2354             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2355                 addr_flags |= IFA_F_OPTIMISTIC;
2356 #endif
2357
2358
2359         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2360         if (!IS_ERR(ifp)) {
2361                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2362                 addrconf_dad_start(ifp, 0);
2363                 in6_ifa_put(ifp);
2364         }
2365 }
2366
2367 static void addrconf_dev_config(struct net_device *dev)
2368 {
2369         struct in6_addr addr;
2370         struct inet6_dev    * idev;
2371
2372         ASSERT_RTNL();
2373
2374         if ((dev->type != ARPHRD_ETHER) &&
2375             (dev->type != ARPHRD_FDDI) &&
2376             (dev->type != ARPHRD_IEEE802_TR) &&
2377             (dev->type != ARPHRD_ARCNET) &&
2378             (dev->type != ARPHRD_INFINIBAND)) {
2379                 /* Alas, we support only Ethernet autoconfiguration. */
2380                 return;
2381         }
2382
2383         idev = addrconf_add_dev(dev);
2384         if (IS_ERR(idev))
2385                 return;
2386
2387         memset(&addr, 0, sizeof(struct in6_addr));
2388         addr.s6_addr32[0] = htonl(0xFE800000);
2389
2390         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2391                 addrconf_add_linklocal(idev, &addr);
2392 }
2393
2394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2395 static void addrconf_sit_config(struct net_device *dev)
2396 {
2397         struct inet6_dev *idev;
2398
2399         ASSERT_RTNL();
2400
2401         /*
2402          * Configure the tunnel with one of our IPv4
2403          * addresses... we should configure all of
2404          * our v4 addrs in the tunnel
2405          */
2406
2407         if ((idev = ipv6_find_idev(dev)) == NULL) {
2408                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2409                 return;
2410         }
2411
2412         if (dev->priv_flags & IFF_ISATAP) {
2413                 struct in6_addr addr;
2414
2415                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2416                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2417                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2418                         addrconf_add_linklocal(idev, &addr);
2419                 return;
2420         }
2421
2422         sit_add_v4_addrs(idev);
2423
2424         if (dev->flags&IFF_POINTOPOINT) {
2425                 addrconf_add_mroute(dev);
2426                 addrconf_add_lroute(dev);
2427         } else
2428                 sit_route_add(dev);
2429 }
2430 #endif
2431
2432 static inline int
2433 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2434 {
2435         struct in6_addr lladdr;
2436
2437         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2438                 addrconf_add_linklocal(idev, &lladdr);
2439                 return 0;
2440         }
2441         return -1;
2442 }
2443
2444 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2445 {
2446         struct net_device *link_dev;
2447         struct net *net = dev_net(idev->dev);
2448
2449         /* first try to inherit the link-local address from the link device */
2450         if (idev->dev->iflink &&
2451             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2452                 if (!ipv6_inherit_linklocal(idev, link_dev))
2453                         return;
2454         }
2455         /* then try to inherit it from any device */
2456         for_each_netdev(net, link_dev) {
2457                 if (!ipv6_inherit_linklocal(idev, link_dev))
2458                         return;
2459         }
2460         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2461 }
2462
2463 /*
2464  * Autoconfigure tunnel with a link-local address so routing protocols,
2465  * DHCPv6, MLD etc. can be run over the virtual link
2466  */
2467
2468 static void addrconf_ip6_tnl_config(struct net_device *dev)
2469 {
2470         struct inet6_dev *idev;
2471
2472         ASSERT_RTNL();
2473
2474         idev = addrconf_add_dev(dev);
2475         if (IS_ERR(idev)) {
2476                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2477                 return;
2478         }
2479         ip6_tnl_add_linklocal(idev);
2480 }
2481
2482 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2483                            void * data)
2484 {
2485         struct net_device *dev = (struct net_device *) data;
2486         struct inet6_dev *idev = __in6_dev_get(dev);
2487         int run_pending = 0;
2488         int err;
2489
2490         switch (event) {
2491         case NETDEV_REGISTER:
2492                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2493                         idev = ipv6_add_dev(dev);
2494                         if (!idev)
2495                                 return notifier_from_errno(-ENOMEM);
2496                 }
2497                 break;
2498
2499         case NETDEV_UP:
2500         case NETDEV_CHANGE:
2501                 if (dev->flags & IFF_SLAVE)
2502                         break;
2503
2504                 if (event == NETDEV_UP) {
2505                         if (!addrconf_qdisc_ok(dev)) {
2506                                 /* device is not ready yet. */
2507                                 printk(KERN_INFO
2508                                         "ADDRCONF(NETDEV_UP): %s: "
2509                                         "link is not ready\n",
2510                                         dev->name);
2511                                 break;
2512                         }
2513
2514                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2515                                 idev = ipv6_add_dev(dev);
2516
2517                         if (idev) {
2518                                 idev->if_flags |= IF_READY;
2519                                 run_pending = 1;
2520                         }
2521                 } else {
2522                         if (!addrconf_qdisc_ok(dev)) {
2523                                 /* device is still not ready. */
2524                                 break;
2525                         }
2526
2527                         if (idev) {
2528                                 if (idev->if_flags & IF_READY)
2529                                         /* device is already configured. */
2530                                         break;
2531                                 idev->if_flags |= IF_READY;
2532                         }
2533
2534                         printk(KERN_INFO
2535                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2536                                         "link becomes ready\n",
2537                                         dev->name);
2538
2539                         run_pending = 1;
2540                 }
2541
2542                 switch (dev->type) {
2543 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2544                 case ARPHRD_SIT:
2545                         addrconf_sit_config(dev);
2546                         break;
2547 #endif
2548                 case ARPHRD_TUNNEL6:
2549                         addrconf_ip6_tnl_config(dev);
2550                         break;
2551                 case ARPHRD_LOOPBACK:
2552                         init_loopback(dev);
2553                         break;
2554
2555                 default:
2556                         addrconf_dev_config(dev);
2557                         break;
2558                 }
2559
2560                 if (idev) {
2561                         if (run_pending)
2562                                 addrconf_dad_run(idev);
2563
2564                         /*
2565                          * If the MTU changed during the interface down,
2566                          * when the interface up, the changed MTU must be
2567                          * reflected in the idev as well as routers.
2568                          */
2569                         if (idev->cnf.mtu6 != dev->mtu &&
2570                             dev->mtu >= IPV6_MIN_MTU) {
2571                                 rt6_mtu_change(dev, dev->mtu);
2572                                 idev->cnf.mtu6 = dev->mtu;
2573                         }
2574                         idev->tstamp = jiffies;
2575                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2576
2577                         /*
2578                          * If the changed mtu during down is lower than
2579                          * IPV6_MIN_MTU stop IPv6 on this interface.
2580                          */
2581                         if (dev->mtu < IPV6_MIN_MTU)
2582                                 addrconf_ifdown(dev, 1);
2583                 }
2584                 break;
2585
2586         case NETDEV_CHANGEMTU:
2587                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2588                         rt6_mtu_change(dev, dev->mtu);
2589                         idev->cnf.mtu6 = dev->mtu;
2590                         break;
2591                 }
2592
2593                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2594                         idev = ipv6_add_dev(dev);
2595                         if (idev)
2596                                 break;
2597                 }
2598
2599                 /*
2600                  * MTU falled under IPV6_MIN_MTU.
2601                  * Stop IPv6 on this interface.
2602                  */
2603
2604         case NETDEV_DOWN:
2605         case NETDEV_UNREGISTER:
2606                 /*
2607                  *      Remove all addresses from this interface.
2608                  */
2609                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2610                 break;
2611
2612         case NETDEV_CHANGENAME:
2613                 if (idev) {
2614                         snmp6_unregister_dev(idev);
2615                         addrconf_sysctl_unregister(idev);
2616                         addrconf_sysctl_register(idev);
2617                         err = snmp6_register_dev(idev);
2618                         if (err)
2619                                 return notifier_from_errno(err);
2620                 }
2621                 break;
2622
2623         case NETDEV_PRE_TYPE_CHANGE:
2624         case NETDEV_POST_TYPE_CHANGE:
2625                 addrconf_type_change(dev, event);
2626                 break;
2627         }
2628
2629         return NOTIFY_OK;
2630 }
2631
2632 /*
2633  *      addrconf module should be notified of a device going up
2634  */
2635 static struct notifier_block ipv6_dev_notf = {
2636         .notifier_call = addrconf_notify,
2637 };
2638
2639 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2640 {
2641         struct inet6_dev *idev;
2642         ASSERT_RTNL();
2643
2644         idev = __in6_dev_get(dev);
2645
2646         if (event == NETDEV_POST_TYPE_CHANGE)
2647                 ipv6_mc_remap(idev);
2648         else if (event == NETDEV_PRE_TYPE_CHANGE)
2649                 ipv6_mc_unmap(idev);
2650 }
2651
2652 static int addrconf_ifdown(struct net_device *dev, int how)
2653 {
2654         struct net *net = dev_net(dev);
2655         struct inet6_dev *idev;
2656         struct inet6_ifaddr *ifa;
2657         LIST_HEAD(keep_list);
2658         int state;
2659
2660         ASSERT_RTNL();
2661
2662         rt6_ifdown(net, dev);
2663         neigh_ifdown(&nd_tbl, dev);
2664
2665         idev = __in6_dev_get(dev);
2666         if (idev == NULL)
2667                 return -ENODEV;
2668
2669         /*
2670          * Step 1: remove reference to ipv6 device from parent device.
2671          *         Do not dev_put!
2672          */
2673         if (how) {
2674                 idev->dead = 1;
2675
2676                 /* protected by rtnl_lock */
2677                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2678
2679                 /* Step 1.5: remove snmp6 entry */
2680                 snmp6_unregister_dev(idev);
2681
2682         }
2683
2684         write_lock_bh(&idev->lock);
2685
2686         /* Step 2: clear flags for stateless addrconf */
2687         if (!how)
2688                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2689
2690 #ifdef CONFIG_IPV6_PRIVACY
2691         if (how && del_timer(&idev->regen_timer))
2692                 in6_dev_put(idev);
2693
2694         /* Step 3: clear tempaddr list */
2695         while (!list_empty(&idev->tempaddr_list)) {
2696                 ifa = list_first_entry(&idev->tempaddr_list,
2697                                        struct inet6_ifaddr, tmp_list);
2698                 list_del(&ifa->tmp_list);
2699                 write_unlock_bh(&idev->lock);
2700                 spin_lock_bh(&ifa->lock);
2701
2702                 if (ifa->ifpub) {
2703                         in6_ifa_put(ifa->ifpub);
2704                         ifa->ifpub = NULL;
2705                 }
2706                 spin_unlock_bh(&ifa->lock);
2707                 in6_ifa_put(ifa);
2708                 write_lock_bh(&idev->lock);
2709         }
2710 #endif
2711
2712         while (!list_empty(&idev->addr_list)) {
2713                 ifa = list_first_entry(&idev->addr_list,
2714                                        struct inet6_ifaddr, if_list);
2715                 addrconf_del_timer(ifa);
2716
2717                 /* If just doing link down, and address is permanent
2718                    and not link-local, then retain it. */
2719                 if (!how &&
2720                     (ifa->flags&IFA_F_PERMANENT) &&
2721                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2722                         list_move_tail(&ifa->if_list, &keep_list);
2723
2724                         /* If not doing DAD on this address, just keep it. */
2725                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2726                             idev->cnf.accept_dad <= 0 ||
2727                             (ifa->flags & IFA_F_NODAD))
2728                                 continue;
2729
2730                         /* If it was tentative already, no need to notify */
2731                         if (ifa->flags & IFA_F_TENTATIVE)
2732                                 continue;
2733
2734                         /* Flag it for later restoration when link comes up */
2735                         ifa->flags |= IFA_F_TENTATIVE;
2736                         ifa->state = INET6_IFADDR_STATE_DAD;
2737
2738                         write_unlock_bh(&idev->lock);
2739
2740                         in6_ifa_hold(ifa);
2741                 } else {
2742                         list_del(&ifa->if_list);
2743
2744                         /* clear hash table */
2745                         spin_lock_bh(&addrconf_hash_lock);
2746                         hlist_del_init_rcu(&ifa->addr_lst);
2747                         spin_unlock_bh(&addrconf_hash_lock);
2748
2749                         write_unlock_bh(&idev->lock);
2750                         spin_lock_bh(&ifa->state_lock);
2751                         state = ifa->state;
2752                         ifa->state = INET6_IFADDR_STATE_DEAD;
2753                         spin_unlock_bh(&ifa->state_lock);
2754
2755                         if (state == INET6_IFADDR_STATE_DEAD)
2756                                 goto put_ifa;
2757                 }
2758
2759                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2760                 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2761                         atomic_notifier_call_chain(&inet6addr_chain,
2762                                                    NETDEV_DOWN, ifa);
2763
2764 put_ifa:
2765                 in6_ifa_put(ifa);
2766
2767                 write_lock_bh(&idev->lock);
2768         }
2769
2770         list_splice(&keep_list, &idev->addr_list);
2771
2772         write_unlock_bh(&idev->lock);
2773
2774         /* Step 5: Discard multicast list */
2775         if (how)
2776                 ipv6_mc_destroy_dev(idev);
2777         else
2778                 ipv6_mc_down(idev);
2779
2780         idev->tstamp = jiffies;
2781
2782         /* Last: Shot the device (if unregistered) */
2783         if (how) {
2784                 addrconf_sysctl_unregister(idev);
2785                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2786                 neigh_ifdown(&nd_tbl, dev);
2787                 in6_dev_put(idev);
2788         }
2789         return 0;
2790 }
2791
2792 static void addrconf_rs_timer(unsigned long data)
2793 {
2794         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2795         struct inet6_dev *idev = ifp->idev;
2796
2797         read_lock(&idev->lock);
2798         if (idev->dead || !(idev->if_flags & IF_READY))
2799                 goto out;
2800
2801         if (idev->cnf.forwarding)
2802                 goto out;
2803
2804         /* Announcement received after solicitation was sent */
2805         if (idev->if_flags & IF_RA_RCVD)
2806                 goto out;
2807
2808         spin_lock(&ifp->lock);
2809         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2810                 /* The wait after the last probe can be shorter */
2811                 addrconf_mod_timer(ifp, AC_RS,
2812                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2813                                    idev->cnf.rtr_solicit_delay :
2814                                    idev->cnf.rtr_solicit_interval);
2815                 spin_unlock(&ifp->lock);
2816
2817                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2818         } else {
2819                 spin_unlock(&ifp->lock);
2820                 /*
2821                  * Note: we do not support deprecated "all on-link"
2822                  * assumption any longer.
2823                  */
2824                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2825                        idev->dev->name);
2826         }
2827
2828 out:
2829         read_unlock(&idev->lock);
2830         in6_ifa_put(ifp);
2831 }
2832
2833 /*
2834  *      Duplicate Address Detection
2835  */
2836 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2837 {
2838         unsigned long rand_num;
2839         struct inet6_dev *idev = ifp->idev;
2840
2841         if (ifp->flags & IFA_F_OPTIMISTIC)
2842                 rand_num = 0;
2843         else
2844                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2845
2846         ifp->probes = idev->cnf.dad_transmits;
2847         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2848 }
2849
2850 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2851 {
2852         struct inet6_dev *idev = ifp->idev;
2853         struct net_device *dev = idev->dev;
2854
2855         addrconf_join_solict(dev, &ifp->addr);
2856
2857         net_srandom(ifp->addr.s6_addr32[3]);
2858
2859         read_lock_bh(&idev->lock);
2860         spin_lock(&ifp->lock);
2861         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2862                 goto out;
2863
2864         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2865             idev->cnf.accept_dad < 1 ||
2866             !(ifp->flags&IFA_F_TENTATIVE) ||
2867             ifp->flags & IFA_F_NODAD) {
2868                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2869                 spin_unlock(&ifp->lock);
2870                 read_unlock_bh(&idev->lock);
2871
2872                 addrconf_dad_completed(ifp);
2873                 return;
2874         }
2875
2876         if (!(idev->if_flags & IF_READY)) {
2877                 spin_unlock(&ifp->lock);
2878                 read_unlock_bh(&idev->lock);
2879                 /*
2880                  * If the device is not ready:
2881                  * - keep it tentative if it is a permanent address.
2882                  * - otherwise, kill it.
2883                  */
2884                 in6_ifa_hold(ifp);
2885                 addrconf_dad_stop(ifp, 0);
2886                 return;
2887         }
2888
2889         /*
2890          * Optimistic nodes can start receiving
2891          * Frames right away
2892          */
2893         if (ifp->flags & IFA_F_OPTIMISTIC)
2894                 ip6_ins_rt(ifp->rt);
2895
2896         addrconf_dad_kick(ifp);
2897 out:
2898         spin_unlock(&ifp->lock);
2899         read_unlock_bh(&idev->lock);
2900 }
2901
2902 static void addrconf_dad_timer(unsigned long data)
2903 {
2904         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2905         struct inet6_dev *idev = ifp->idev;
2906         struct in6_addr mcaddr;
2907
2908         if (!ifp->probes && addrconf_dad_end(ifp))
2909                 goto out;
2910
2911         read_lock(&idev->lock);
2912         if (idev->dead || !(idev->if_flags & IF_READY)) {
2913                 read_unlock(&idev->lock);
2914                 goto out;
2915         }
2916
2917         spin_lock(&ifp->lock);
2918         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2919                 spin_unlock(&ifp->lock);
2920                 read_unlock(&idev->lock);
2921                 goto out;
2922         }
2923
2924         if (ifp->probes == 0) {
2925                 /*
2926                  * DAD was successful
2927                  */
2928
2929                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2930                 spin_unlock(&ifp->lock);
2931                 read_unlock(&idev->lock);
2932
2933                 addrconf_dad_completed(ifp);
2934
2935                 goto out;
2936         }
2937
2938         ifp->probes--;
2939         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2940         spin_unlock(&ifp->lock);
2941         read_unlock(&idev->lock);
2942
2943         /* send a neighbour solicitation for our addr */
2944         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2945         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2946 out:
2947         in6_ifa_put(ifp);
2948 }
2949
2950 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2951 {
2952         struct net_device *dev = ifp->idev->dev;
2953
2954         /*
2955          *      Configure the address for reception. Now it is valid.
2956          */
2957
2958         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2959
2960         /* If added prefix is link local and forwarding is off,
2961            start sending router solicitations.
2962          */
2963
2964         if (ifp->idev->cnf.forwarding == 0 &&
2965             ifp->idev->cnf.rtr_solicits > 0 &&
2966             (dev->flags&IFF_LOOPBACK) == 0 &&
2967             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2968                 /*
2969                  *      If a host as already performed a random delay
2970                  *      [...] as part of DAD [...] there is no need
2971                  *      to delay again before sending the first RS
2972                  */
2973                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2974
2975                 spin_lock_bh(&ifp->lock);
2976                 ifp->probes = 1;
2977                 ifp->idev->if_flags |= IF_RS_SENT;
2978                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2979                 spin_unlock_bh(&ifp->lock);
2980         }
2981 }
2982
2983 static void addrconf_dad_run(struct inet6_dev *idev)
2984 {
2985         struct inet6_ifaddr *ifp;
2986
2987         read_lock_bh(&idev->lock);
2988         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2989                 spin_lock(&ifp->lock);
2990                 if (ifp->flags & IFA_F_TENTATIVE &&
2991                     ifp->state == INET6_IFADDR_STATE_DAD)
2992                         addrconf_dad_kick(ifp);
2993                 spin_unlock(&ifp->lock);
2994         }
2995         read_unlock_bh(&idev->lock);
2996 }
2997
2998 #ifdef CONFIG_PROC_FS
2999 struct if6_iter_state {
3000         struct seq_net_private p;
3001         int bucket;
3002 };
3003
3004 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3005 {
3006         struct inet6_ifaddr *ifa = NULL;
3007         struct if6_iter_state *state = seq->private;
3008         struct net *net = seq_file_net(seq);
3009
3010         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3011                 struct hlist_node *n;
3012                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3013                                          addr_lst)
3014                         if (net_eq(dev_net(ifa->idev->dev), net))
3015                                 return ifa;
3016         }
3017         return NULL;
3018 }
3019
3020 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3021                                          struct inet6_ifaddr *ifa)
3022 {
3023         struct if6_iter_state *state = seq->private;
3024         struct net *net = seq_file_net(seq);
3025         struct hlist_node *n = &ifa->addr_lst;
3026
3027         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3028                 if (net_eq(dev_net(ifa->idev->dev), net))
3029                         return ifa;
3030
3031         while (++state->bucket < IN6_ADDR_HSIZE) {
3032                 hlist_for_each_entry_rcu_bh(ifa, n,
3033                                      &inet6_addr_lst[state->bucket], addr_lst) {
3034                         if (net_eq(dev_net(ifa->idev->dev), net))
3035                                 return ifa;
3036                 }
3037         }
3038
3039         return NULL;
3040 }
3041
3042 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3043 {
3044         struct inet6_ifaddr *ifa = if6_get_first(seq);
3045
3046         if (ifa)
3047                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3048                         --pos;
3049         return pos ? NULL : ifa;
3050 }
3051
3052 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3053         __acquires(rcu_bh)
3054 {
3055         rcu_read_lock_bh();
3056         return if6_get_idx(seq, *pos);
3057 }
3058
3059 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3060 {
3061         struct inet6_ifaddr *ifa;
3062
3063         ifa = if6_get_next(seq, v);
3064         ++*pos;
3065         return ifa;
3066 }
3067
3068 static void if6_seq_stop(struct seq_file *seq, void *v)
3069         __releases(rcu_bh)
3070 {
3071         rcu_read_unlock_bh();
3072 }
3073
3074 static int if6_seq_show(struct seq_file *seq, void *v)
3075 {
3076         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3077         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3078                    &ifp->addr,
3079                    ifp->idev->dev->ifindex,
3080                    ifp->prefix_len,
3081                    ifp->scope,
3082                    ifp->flags,
3083                    ifp->idev->dev->name);
3084         return 0;
3085 }
3086
3087 static const struct seq_operations if6_seq_ops = {
3088         .start  = if6_seq_start,
3089         .next   = if6_seq_next,
3090         .show   = if6_seq_show,
3091         .stop   = if6_seq_stop,
3092 };
3093
3094 static int if6_seq_open(struct inode *inode, struct file *file)
3095 {
3096         return seq_open_net(inode, file, &if6_seq_ops,
3097                             sizeof(struct if6_iter_state));
3098 }
3099
3100 static const struct file_operations if6_fops = {
3101         .owner          = THIS_MODULE,
3102         .open           = if6_seq_open,
3103         .read           = seq_read,
3104         .llseek         = seq_lseek,
3105         .release        = seq_release_net,
3106 };
3107
3108 static int __net_init if6_proc_net_init(struct net *net)
3109 {
3110         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3111                 return -ENOMEM;
3112         return 0;
3113 }
3114
3115 static void __net_exit if6_proc_net_exit(struct net *net)
3116 {
3117        proc_net_remove(net, "if_inet6");
3118 }
3119
3120 static struct pernet_operations if6_proc_net_ops = {
3121        .init = if6_proc_net_init,
3122        .exit = if6_proc_net_exit,
3123 };
3124
3125 int __init if6_proc_init(void)
3126 {
3127         return register_pernet_subsys(&if6_proc_net_ops);
3128 }
3129
3130 void if6_proc_exit(void)
3131 {
3132         unregister_pernet_subsys(&if6_proc_net_ops);
3133 }
3134 #endif  /* CONFIG_PROC_FS */
3135
3136 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3137 /* Check if address is a home address configured on any interface. */
3138 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3139 {
3140         int ret = 0;
3141         struct inet6_ifaddr *ifp = NULL;
3142         struct hlist_node *n;
3143         unsigned int hash = ipv6_addr_hash(addr);
3144
3145         rcu_read_lock_bh();
3146         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3147                 if (!net_eq(dev_net(ifp->idev->dev), net))
3148                         continue;
3149                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3150                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3151                         ret = 1;
3152                         break;
3153                 }
3154         }
3155         rcu_read_unlock_bh();
3156         return ret;
3157 }
3158 #endif
3159
3160 /*
3161  *      Periodic address status verification
3162  */
3163
3164 static void addrconf_verify(unsigned long foo)
3165 {
3166         unsigned long now, next, next_sec, next_sched;
3167         struct inet6_ifaddr *ifp;
3168         struct hlist_node *node;
3169         int i;
3170
3171         rcu_read_lock_bh();
3172         spin_lock(&addrconf_verify_lock);
3173         now = jiffies;
3174         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3175
3176         del_timer(&addr_chk_timer);
3177
3178         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3179 restart:
3180                 hlist_for_each_entry_rcu_bh(ifp, node,
3181                                          &inet6_addr_lst[i], addr_lst) {
3182                         unsigned long age;
3183
3184                         if (ifp->flags & IFA_F_PERMANENT)
3185                                 continue;
3186
3187                         spin_lock(&ifp->lock);
3188                         /* We try to batch several events at once. */
3189                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3190
3191                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3192                             age >= ifp->valid_lft) {
3193                                 spin_unlock(&ifp->lock);
3194                                 in6_ifa_hold(ifp);
3195                                 ipv6_del_addr(ifp);
3196                                 goto restart;
3197                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3198                                 spin_unlock(&ifp->lock);
3199                                 continue;
3200                         } else if (age >= ifp->prefered_lft) {
3201                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3202                                 int deprecate = 0;
3203
3204                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3205                                         deprecate = 1;
3206                                         ifp->flags |= IFA_F_DEPRECATED;
3207                                 }
3208
3209                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3210                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3211
3212                                 spin_unlock(&ifp->lock);
3213
3214                                 if (deprecate) {
3215                                         in6_ifa_hold(ifp);
3216
3217                                         ipv6_ifa_notify(0, ifp);
3218                                         in6_ifa_put(ifp);
3219                                         goto restart;
3220                                 }
3221 #ifdef CONFIG_IPV6_PRIVACY
3222                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3223                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3224                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3225                                         ifp->idev->cnf.dad_transmits *
3226                                         ifp->idev->nd_parms->retrans_time / HZ;
3227
3228                                 if (age >= ifp->prefered_lft - regen_advance) {
3229                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3230                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3231                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3232                                         if (!ifp->regen_count && ifpub) {
3233                                                 ifp->regen_count++;
3234                                                 in6_ifa_hold(ifp);
3235                                                 in6_ifa_hold(ifpub);
3236                                                 spin_unlock(&ifp->lock);
3237
3238                                                 spin_lock(&ifpub->lock);
3239                                                 ifpub->regen_count = 0;
3240                                                 spin_unlock(&ifpub->lock);
3241                                                 ipv6_create_tempaddr(ifpub, ifp);
3242                                                 in6_ifa_put(ifpub);
3243                                                 in6_ifa_put(ifp);
3244                                                 goto restart;
3245                                         }
3246                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3247                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3248                                 spin_unlock(&ifp->lock);
3249 #endif
3250                         } else {
3251                                 /* ifp->prefered_lft <= ifp->valid_lft */
3252                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3253                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3254                                 spin_unlock(&ifp->lock);
3255                         }
3256                 }
3257         }
3258
3259         next_sec = round_jiffies_up(next);
3260         next_sched = next;
3261
3262         /* If rounded timeout is accurate enough, accept it. */
3263         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3264                 next_sched = next_sec;
3265
3266         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3267         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3268                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3269
3270         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3271               now, next, next_sec, next_sched));
3272
3273         addr_chk_timer.expires = next_sched;
3274         add_timer(&addr_chk_timer);
3275         spin_unlock(&addrconf_verify_lock);
3276         rcu_read_unlock_bh();
3277 }
3278
3279 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3280 {
3281         struct in6_addr *pfx = NULL;
3282
3283         if (addr)
3284                 pfx = nla_data(addr);
3285
3286         if (local) {
3287                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3288                         pfx = NULL;
3289                 else
3290                         pfx = nla_data(local);
3291         }
3292
3293         return pfx;
3294 }
3295
3296 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3297         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3298         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3299         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3300 };
3301
3302 static int
3303 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3304 {
3305         struct net *net = sock_net(skb->sk);
3306         struct ifaddrmsg *ifm;
3307         struct nlattr *tb[IFA_MAX+1];
3308         struct in6_addr *pfx;
3309         int err;
3310
3311         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3312         if (err < 0)
3313                 return err;
3314
3315         ifm = nlmsg_data(nlh);
3316         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3317         if (pfx == NULL)
3318                 return -EINVAL;
3319
3320         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3321 }
3322
3323 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3324                              u32 prefered_lft, u32 valid_lft)
3325 {
3326         u32 flags;
3327         clock_t expires;
3328         unsigned long timeout;
3329
3330         if (!valid_lft || (prefered_lft > valid_lft))
3331                 return -EINVAL;
3332
3333         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3334         if (addrconf_finite_timeout(timeout)) {
3335                 expires = jiffies_to_clock_t(timeout * HZ);
3336                 valid_lft = timeout;
3337                 flags = RTF_EXPIRES;
3338         } else {
3339                 expires = 0;
3340                 flags = 0;
3341                 ifa_flags |= IFA_F_PERMANENT;
3342         }
3343
3344         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3345         if (addrconf_finite_timeout(timeout)) {
3346                 if (timeout == 0)
3347                         ifa_flags |= IFA_F_DEPRECATED;
3348                 prefered_lft = timeout;
3349         }
3350
3351         spin_lock_bh(&ifp->lock);
3352         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3353         ifp->tstamp = jiffies;
3354         ifp->valid_lft = valid_lft;
3355         ifp->prefered_lft = prefered_lft;
3356
3357         spin_unlock_bh(&ifp->lock);
3358         if (!(ifp->flags&IFA_F_TENTATIVE))
3359                 ipv6_ifa_notify(0, ifp);
3360
3361         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3362                               expires, flags);
3363         addrconf_verify(0);
3364
3365         return 0;
3366 }
3367
3368 static int
3369 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3370 {
3371         struct net *net = sock_net(skb->sk);
3372         struct ifaddrmsg *ifm;
3373         struct nlattr *tb[IFA_MAX+1];
3374         struct in6_addr *pfx;
3375         struct inet6_ifaddr *ifa;
3376         struct net_device *dev;
3377         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3378         u8 ifa_flags;
3379         int err;
3380
3381         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3382         if (err < 0)
3383                 return err;
3384
3385         ifm = nlmsg_data(nlh);
3386         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3387         if (pfx == NULL)
3388                 return -EINVAL;
3389
3390         if (tb[IFA_CACHEINFO]) {
3391                 struct ifa_cacheinfo *ci;
3392
3393                 ci = nla_data(tb[IFA_CACHEINFO]);
3394                 valid_lft = ci->ifa_valid;
3395                 preferred_lft = ci->ifa_prefered;
3396         } else {
3397                 preferred_lft = INFINITY_LIFE_TIME;
3398                 valid_lft = INFINITY_LIFE_TIME;
3399         }
3400
3401         dev =  __dev_get_by_index(net, ifm->ifa_index);
3402         if (dev == NULL)
3403                 return -ENODEV;
3404
3405         /* We ignore other flags so far. */
3406         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3407
3408         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3409         if (ifa == NULL) {
3410                 /*
3411                  * It would be best to check for !NLM_F_CREATE here but
3412                  * userspace alreay relies on not having to provide this.
3413                  */
3414                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3415                                       ifm->ifa_prefixlen, ifa_flags,
3416                                       preferred_lft, valid_lft);
3417         }
3418
3419         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3420             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3421                 err = -EEXIST;
3422         else
3423                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3424
3425         in6_ifa_put(ifa);
3426
3427         return err;
3428 }
3429
3430 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3431                           u8 scope, int ifindex)
3432 {
3433         struct ifaddrmsg *ifm;
3434
3435         ifm = nlmsg_data(nlh);
3436         ifm->ifa_family = AF_INET6;
3437         ifm->ifa_prefixlen = prefixlen;
3438         ifm->ifa_flags = flags;
3439         ifm->ifa_scope = scope;
3440         ifm->ifa_index = ifindex;
3441 }
3442
3443 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3444                          unsigned long tstamp, u32 preferred, u32 valid)
3445 {
3446         struct ifa_cacheinfo ci;
3447
3448         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3449                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3450         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3451                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3452         ci.ifa_prefered = preferred;
3453         ci.ifa_valid = valid;
3454
3455         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3456 }
3457
3458 static inline int rt_scope(int ifa_scope)
3459 {
3460         if (ifa_scope & IFA_HOST)
3461                 return RT_SCOPE_HOST;
3462         else if (ifa_scope & IFA_LINK)
3463                 return RT_SCOPE_LINK;
3464         else if (ifa_scope & IFA_SITE)
3465                 return RT_SCOPE_SITE;
3466         else
3467                 return RT_SCOPE_UNIVERSE;
3468 }
3469
3470 static inline int inet6_ifaddr_msgsize(void)
3471 {
3472         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3473                + nla_total_size(16) /* IFA_ADDRESS */
3474                + nla_total_size(sizeof(struct ifa_cacheinfo));
3475 }
3476
3477 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3478                              u32 pid, u32 seq, int event, unsigned int flags)
3479 {
3480         struct nlmsghdr  *nlh;
3481         u32 preferred, valid;
3482
3483         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3484         if (nlh == NULL)
3485                 return -EMSGSIZE;
3486
3487         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3488                       ifa->idev->dev->ifindex);
3489
3490         if (!(ifa->flags&IFA_F_PERMANENT)) {
3491                 preferred = ifa->prefered_lft;
3492                 valid = ifa->valid_lft;
3493                 if (preferred != INFINITY_LIFE_TIME) {
3494                         long tval = (jiffies - ifa->tstamp)/HZ;
3495                         if (preferred > tval)
3496                                 preferred -= tval;
3497                         else
3498                                 preferred = 0;
3499                         if (valid != INFINITY_LIFE_TIME)
3500                                 valid -= tval;
3501                 }
3502         } else {
3503                 preferred = INFINITY_LIFE_TIME;
3504                 valid = INFINITY_LIFE_TIME;
3505         }
3506
3507         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3508             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3509                 nlmsg_cancel(skb, nlh);
3510                 return -EMSGSIZE;
3511         }
3512
3513         return nlmsg_end(skb, nlh);
3514 }
3515
3516 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3517                                 u32 pid, u32 seq, int event, u16 flags)
3518 {
3519         struct nlmsghdr  *nlh;
3520         u8 scope = RT_SCOPE_UNIVERSE;
3521         int ifindex = ifmca->idev->dev->ifindex;
3522
3523         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3524                 scope = RT_SCOPE_SITE;
3525
3526         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3527         if (nlh == NULL)
3528                 return -EMSGSIZE;
3529
3530         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3531         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3532             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3533                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3534                 nlmsg_cancel(skb, nlh);
3535                 return -EMSGSIZE;
3536         }
3537
3538         return nlmsg_end(skb, nlh);
3539 }
3540
3541 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3542                                 u32 pid, u32 seq, int event, unsigned int flags)
3543 {
3544         struct nlmsghdr  *nlh;
3545         u8 scope = RT_SCOPE_UNIVERSE;
3546         int ifindex = ifaca->aca_idev->dev->ifindex;
3547
3548         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3549                 scope = RT_SCOPE_SITE;
3550
3551         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3552         if (nlh == NULL)
3553                 return -EMSGSIZE;
3554
3555         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3556         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3557             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3558                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3559                 nlmsg_cancel(skb, nlh);
3560                 return -EMSGSIZE;
3561         }
3562
3563         return nlmsg_end(skb, nlh);
3564 }
3565
3566 enum addr_type_t {
3567         UNICAST_ADDR,
3568         MULTICAST_ADDR,
3569         ANYCAST_ADDR,
3570 };
3571
3572 /* called with rcu_read_lock() */
3573 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3574                           struct netlink_callback *cb, enum addr_type_t type,
3575                           int s_ip_idx, int *p_ip_idx)
3576 {
3577         struct ifmcaddr6 *ifmca;
3578         struct ifacaddr6 *ifaca;
3579         int err = 1;
3580         int ip_idx = *p_ip_idx;
3581
3582         read_lock_bh(&idev->lock);
3583         switch (type) {
3584         case UNICAST_ADDR: {
3585                 struct inet6_ifaddr *ifa;
3586
3587                 /* unicast address incl. temp addr */
3588                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3589                         if (++ip_idx < s_ip_idx)
3590                                 continue;
3591                         err = inet6_fill_ifaddr(skb, ifa,
3592                                                 NETLINK_CB(cb->skb).pid,
3593                                                 cb->nlh->nlmsg_seq,
3594                                                 RTM_NEWADDR,
3595                                                 NLM_F_MULTI);
3596                         if (err <= 0)
3597                                 break;
3598                 }
3599                 break;
3600         }
3601         case MULTICAST_ADDR:
3602                 /* multicast address */
3603                 for (ifmca = idev->mc_list; ifmca;
3604                      ifmca = ifmca->next, ip_idx++) {
3605                         if (ip_idx < s_ip_idx)
3606                                 continue;
3607                         err = inet6_fill_ifmcaddr(skb, ifmca,
3608                                                   NETLINK_CB(cb->skb).pid,
3609                                                   cb->nlh->nlmsg_seq,
3610                                                   RTM_GETMULTICAST,
3611                                                   NLM_F_MULTI);
3612                         if (err <= 0)
3613                                 break;
3614                 }
3615                 break;
3616         case ANYCAST_ADDR:
3617                 /* anycast address */
3618                 for (ifaca = idev->ac_list; ifaca;
3619                      ifaca = ifaca->aca_next, ip_idx++) {
3620                         if (ip_idx < s_ip_idx)
3621                                 continue;
3622                         err = inet6_fill_ifacaddr(skb, ifaca,
3623                                                   NETLINK_CB(cb->skb).pid,
3624                                                   cb->nlh->nlmsg_seq,
3625                                                   RTM_GETANYCAST,
3626                                                   NLM_F_MULTI);
3627                         if (err <= 0)
3628                                 break;
3629                 }
3630                 break;
3631         default:
3632                 break;
3633         }
3634         read_unlock_bh(&idev->lock);
3635         *p_ip_idx = ip_idx;
3636         return err;
3637 }
3638
3639 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3640                            enum addr_type_t type)
3641 {
3642         struct net *net = sock_net(skb->sk);
3643         int h, s_h;
3644         int idx, ip_idx;
3645         int s_idx, s_ip_idx;
3646         struct net_device *dev;
3647         struct inet6_dev *idev;
3648         struct hlist_head *head;
3649         struct hlist_node *node;
3650
3651         s_h = cb->args[0];
3652         s_idx = idx = cb->args[1];
3653         s_ip_idx = ip_idx = cb->args[2];
3654
3655         rcu_read_lock();
3656         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3657                 idx = 0;
3658                 head = &net->dev_index_head[h];
3659                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3660                         if (idx < s_idx)
3661                                 goto cont;
3662                         if (h > s_h || idx > s_idx)
3663                                 s_ip_idx = 0;
3664                         ip_idx = 0;
3665                         idev = __in6_dev_get(dev);
3666                         if (!idev)
3667                                 goto cont;
3668
3669                         if (in6_dump_addrs(idev, skb, cb, type,
3670                                            s_ip_idx, &ip_idx) <= 0)
3671                                 goto done;
3672 cont:
3673                         idx++;
3674                 }
3675         }
3676 done:
3677         rcu_read_unlock();
3678         cb->args[0] = h;
3679         cb->args[1] = idx;
3680         cb->args[2] = ip_idx;
3681
3682         return skb->len;
3683 }
3684
3685 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3686 {
3687         enum addr_type_t type = UNICAST_ADDR;
3688
3689         return inet6_dump_addr(skb, cb, type);
3690 }
3691
3692 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3693 {
3694         enum addr_type_t type = MULTICAST_ADDR;
3695
3696         return inet6_dump_addr(skb, cb, type);
3697 }
3698
3699
3700 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3701 {
3702         enum addr_type_t type = ANYCAST_ADDR;
3703
3704         return inet6_dump_addr(skb, cb, type);
3705 }
3706
3707 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3708                              void *arg)
3709 {
3710         struct net *net = sock_net(in_skb->sk);
3711         struct ifaddrmsg *ifm;
3712         struct nlattr *tb[IFA_MAX+1];
3713         struct in6_addr *addr = NULL;
3714         struct net_device *dev = NULL;
3715         struct inet6_ifaddr *ifa;
3716         struct sk_buff *skb;
3717         int err;
3718
3719         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3720         if (err < 0)
3721                 goto errout;
3722
3723         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3724         if (addr == NULL) {
3725                 err = -EINVAL;
3726                 goto errout;
3727         }
3728
3729         ifm = nlmsg_data(nlh);
3730         if (ifm->ifa_index)
3731                 dev = __dev_get_by_index(net, ifm->ifa_index);
3732
3733         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3734         if (!ifa) {
3735                 err = -EADDRNOTAVAIL;
3736                 goto errout;
3737         }
3738
3739         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3740         if (!skb) {
3741                 err = -ENOBUFS;
3742                 goto errout_ifa;
3743         }
3744
3745         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3746                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3747         if (err < 0) {
3748                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3749                 WARN_ON(err == -EMSGSIZE);
3750                 kfree_skb(skb);
3751                 goto errout_ifa;
3752         }
3753         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3754 errout_ifa:
3755         in6_ifa_put(ifa);
3756 errout:
3757         return err;
3758 }
3759
3760 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3761 {
3762         struct sk_buff *skb;
3763         struct net *net = dev_net(ifa->idev->dev);
3764         int err = -ENOBUFS;
3765
3766         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3767         if (skb == NULL)
3768                 goto errout;
3769
3770         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3771         if (err < 0) {
3772                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3773                 WARN_ON(err == -EMSGSIZE);
3774                 kfree_skb(skb);
3775                 goto errout;
3776         }
3777         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3778         return;
3779 errout:
3780         if (err < 0)
3781                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3782 }
3783
3784 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3785                                 __s32 *array, int bytes)
3786 {
3787         BUG_ON(bytes < (DEVCONF_MAX * 4));
3788
3789         memset(array, 0, bytes);
3790         array[DEVCONF_FORWARDING] = cnf->forwarding;
3791         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3792         array[DEVCONF_MTU6] = cnf->mtu6;
3793         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3794         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3795         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3796         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3797         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3798         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3799         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3800         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3801 #ifdef CONFIG_IPV6_PRIVACY
3802         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3803         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3804         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3805         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3806         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3807 #endif
3808         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3809         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3810         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3811 #ifdef CONFIG_IPV6_ROUTER_PREF
3812         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3813         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3814 #ifdef CONFIG_IPV6_ROUTE_INFO
3815         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3816 #endif
3817 #endif
3818         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3819         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3820 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3821         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3822 #endif
3823 #ifdef CONFIG_IPV6_MROUTE
3824         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3825 #endif
3826         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3827         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3828         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3829 }
3830
3831 static inline size_t inet6_if_nlmsg_size(void)
3832 {
3833         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3834                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3835                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3836                + nla_total_size(4) /* IFLA_MTU */
3837                + nla_total_size(4) /* IFLA_LINK */
3838                + nla_total_size( /* IFLA_PROTINFO */
3839                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3840                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3841                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3842                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3843                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3844                  );
3845 }
3846
3847 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3848                                       int items, int bytes)
3849 {
3850         int i;
3851         int pad = bytes - sizeof(u64) * items;
3852         BUG_ON(pad < 0);
3853
3854         /* Use put_unaligned() because stats may not be aligned for u64. */
3855         put_unaligned(items, &stats[0]);
3856         for (i = 1; i < items; i++)
3857                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3858
3859         memset(&stats[items], 0, pad);
3860 }
3861
3862 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3863                              int bytes)
3864 {
3865         switch (attrtype) {
3866         case IFLA_INET6_STATS:
3867                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3868                 break;
3869         case IFLA_INET6_ICMP6STATS:
3870                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3871                 break;
3872         }
3873 }
3874
3875 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3876                              u32 pid, u32 seq, int event, unsigned int flags)
3877 {
3878         struct net_device *dev = idev->dev;
3879         struct nlattr *nla;
3880         struct ifinfomsg *hdr;
3881         struct nlmsghdr *nlh;
3882         void *protoinfo;
3883         struct ifla_cacheinfo ci;
3884
3885         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3886         if (nlh == NULL)
3887                 return -EMSGSIZE;
3888
3889         hdr = nlmsg_data(nlh);
3890         hdr->ifi_family = AF_INET6;
3891         hdr->__ifi_pad = 0;
3892         hdr->ifi_type = dev->type;
3893         hdr->ifi_index = dev->ifindex;
3894         hdr->ifi_flags = dev_get_flags(dev);
3895         hdr->ifi_change = 0;
3896
3897         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3898
3899         if (dev->addr_len)
3900                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3901
3902         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3903         if (dev->ifindex != dev->iflink)
3904                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3905
3906         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3907         if (protoinfo == NULL)
3908                 goto nla_put_failure;
3909
3910         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3911
3912         ci.max_reasm_len = IPV6_MAXPLEN;
3913         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3914                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3915         ci.reachable_time = idev->nd_parms->reachable_time;
3916         ci.retrans_time = idev->nd_parms->retrans_time;
3917         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3918
3919         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3920         if (nla == NULL)
3921                 goto nla_put_failure;
3922         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3923
3924         /* XXX - MC not implemented */
3925
3926         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3927         if (nla == NULL)
3928                 goto nla_put_failure;
3929         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3930
3931         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3932         if (nla == NULL)
3933                 goto nla_put_failure;
3934         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3935
3936         nla_nest_end(skb, protoinfo);
3937         return nlmsg_end(skb, nlh);
3938
3939 nla_put_failure:
3940         nlmsg_cancel(skb, nlh);
3941         return -EMSGSIZE;
3942 }
3943
3944 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3945 {
3946         struct net *net = sock_net(skb->sk);
3947         int h, s_h;
3948         int idx = 0, s_idx;
3949         struct net_device *dev;
3950         struct inet6_dev *idev;
3951         struct hlist_head *head;
3952         struct hlist_node *node;
3953
3954         s_h = cb->args[0];
3955         s_idx = cb->args[1];
3956
3957         rcu_read_lock();
3958         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3959                 idx = 0;
3960                 head = &net->dev_index_head[h];
3961                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3962                         if (idx < s_idx)
3963                                 goto cont;
3964                         idev = __in6_dev_get(dev);
3965                         if (!idev)
3966                                 goto cont;
3967                         if (inet6_fill_ifinfo(skb, idev,
3968                                               NETLINK_CB(cb->skb).pid,
3969                                               cb->nlh->nlmsg_seq,
3970                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3971                                 goto out;
3972 cont:
3973                         idx++;
3974                 }
3975         }
3976 out:
3977         rcu_read_unlock();
3978         cb->args[1] = idx;
3979         cb->args[0] = h;
3980
3981         return skb->len;
3982 }
3983
3984 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3985 {
3986         struct sk_buff *skb;
3987         struct net *net = dev_net(idev->dev);
3988         int err = -ENOBUFS;
3989
3990         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3991         if (skb == NULL)
3992                 goto errout;
3993
3994         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3995         if (err < 0) {
3996                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3997                 WARN_ON(err == -EMSGSIZE);
3998                 kfree_skb(skb);
3999                 goto errout;
4000         }
4001         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
4002         return;
4003 errout:
4004         if (err < 0)
4005                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
4006 }
4007
4008 static inline size_t inet6_prefix_nlmsg_size(void)
4009 {
4010         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4011                + nla_total_size(sizeof(struct in6_addr))
4012                + nla_total_size(sizeof(struct prefix_cacheinfo));
4013 }
4014
4015 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4016                              struct prefix_info *pinfo, u32 pid, u32 seq,
4017                              int event, unsigned int flags)
4018 {
4019         struct prefixmsg *pmsg;
4020         struct nlmsghdr *nlh;
4021         struct prefix_cacheinfo ci;
4022
4023         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4024         if (nlh == NULL)
4025                 return -EMSGSIZE;
4026
4027         pmsg = nlmsg_data(nlh);
4028         pmsg->prefix_family = AF_INET6;
4029         pmsg->prefix_pad1 = 0;
4030         pmsg->prefix_pad2 = 0;
4031         pmsg->prefix_ifindex = idev->dev->ifindex;
4032         pmsg->prefix_len = pinfo->prefix_len;
4033         pmsg->prefix_type = pinfo->type;
4034         pmsg->prefix_pad3 = 0;
4035         pmsg->prefix_flags = 0;
4036         if (pinfo->onlink)
4037                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4038         if (pinfo->autoconf)
4039                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4040
4041         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4042
4043         ci.preferred_time = ntohl(pinfo->prefered);
4044         ci.valid_time = ntohl(pinfo->valid);
4045         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4046
4047         return nlmsg_end(skb, nlh);
4048
4049 nla_put_failure:
4050         nlmsg_cancel(skb, nlh);
4051         return -EMSGSIZE;
4052 }
4053
4054 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4055                          struct prefix_info *pinfo)
4056 {
4057         struct sk_buff *skb;
4058         struct net *net = dev_net(idev->dev);
4059         int err = -ENOBUFS;
4060
4061         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4062         if (skb == NULL)
4063                 goto errout;
4064
4065         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4066         if (err < 0) {
4067                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4068                 WARN_ON(err == -EMSGSIZE);
4069                 kfree_skb(skb);
4070                 goto errout;
4071         }
4072         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4073         return;
4074 errout:
4075         if (err < 0)
4076                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4077 }
4078
4079 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4080 {
4081         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4082
4083         switch (event) {
4084         case RTM_NEWADDR:
4085                 /*
4086                  * If the address was optimistic
4087                  * we inserted the route at the start of
4088                  * our DAD process, so we don't need
4089                  * to do it again
4090                  */
4091                 if (!(ifp->rt->rt6i_node))
4092                         ip6_ins_rt(ifp->rt);
4093                 if (ifp->idev->cnf.forwarding)
4094                         addrconf_join_anycast(ifp);
4095                 break;
4096         case RTM_DELADDR:
4097                 if (ifp->idev->cnf.forwarding)
4098                         addrconf_leave_anycast(ifp);
4099                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4100                 dst_hold(&ifp->rt->u.dst);
4101
4102                 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4103                     ip6_del_rt(ifp->rt))
4104                         dst_free(&ifp->rt->u.dst);
4105                 break;
4106         }
4107 }
4108
4109 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4110 {
4111         rcu_read_lock_bh();
4112         if (likely(ifp->idev->dead == 0))
4113                 __ipv6_ifa_notify(event, ifp);
4114         rcu_read_unlock_bh();
4115 }
4116
4117 #ifdef CONFIG_SYSCTL
4118
4119 static
4120 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4121                            void __user *buffer, size_t *lenp, loff_t *ppos)
4122 {
4123         int *valp = ctl->data;
4124         int val = *valp;
4125         loff_t pos = *ppos;
4126         int ret;
4127
4128         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4129
4130         if (write)
4131                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4132         if (ret)
4133                 *ppos = pos;
4134         return ret;
4135 }
4136
4137 static void dev_disable_change(struct inet6_dev *idev)
4138 {
4139         if (!idev || !idev->dev)
4140                 return;
4141
4142         if (idev->cnf.disable_ipv6)
4143                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4144         else
4145                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4146 }
4147
4148 static void addrconf_disable_change(struct net *net, __s32 newf)
4149 {
4150         struct net_device *dev;
4151         struct inet6_dev *idev;
4152
4153         rcu_read_lock();
4154         for_each_netdev_rcu(net, dev) {
4155                 idev = __in6_dev_get(dev);
4156                 if (idev) {
4157                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4158                         idev->cnf.disable_ipv6 = newf;
4159                         if (changed)
4160                                 dev_disable_change(idev);
4161                 }
4162         }
4163         rcu_read_unlock();
4164 }
4165
4166 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4167 {
4168         struct net *net;
4169
4170         net = (struct net *)table->extra2;
4171
4172         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4173                 return 0;
4174
4175         if (!rtnl_trylock()) {
4176                 /* Restore the original values before restarting */
4177                 *p = old;
4178                 return restart_syscall();
4179         }
4180
4181         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4182                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4183                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4184                 addrconf_disable_change(net, newf);
4185         } else if ((!*p) ^ (!old))
4186                 dev_disable_change((struct inet6_dev *)table->extra1);
4187
4188         rtnl_unlock();
4189         return 0;
4190 }
4191
4192 static
4193 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4194                             void __user *buffer, size_t *lenp, loff_t *ppos)
4195 {
4196         int *valp = ctl->data;
4197         int val = *valp;
4198         loff_t pos = *ppos;
4199         int ret;
4200
4201         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4202
4203         if (write)
4204                 ret = addrconf_disable_ipv6(ctl, valp, val);
4205         if (ret)
4206                 *ppos = pos;
4207         return ret;
4208 }
4209
4210 static struct addrconf_sysctl_table
4211 {
4212         struct ctl_table_header *sysctl_header;
4213         ctl_table addrconf_vars[DEVCONF_MAX+1];
4214         char *dev_name;
4215 } addrconf_sysctl __read_mostly = {
4216         .sysctl_header = NULL,
4217         .addrconf_vars = {
4218                 {
4219                         .procname       = "forwarding",
4220                         .data           = &ipv6_devconf.forwarding,
4221                         .maxlen         = sizeof(int),
4222                         .mode           = 0644,
4223                         .proc_handler   = addrconf_sysctl_forward,
4224                 },
4225                 {
4226                         .procname       = "hop_limit",
4227                         .data           = &ipv6_devconf.hop_limit,
4228                         .maxlen         = sizeof(int),
4229                         .mode           = 0644,
4230                         .proc_handler   = proc_dointvec,
4231                 },
4232                 {
4233                         .procname       = "mtu",
4234                         .data           = &ipv6_devconf.mtu6,
4235                         .maxlen         = sizeof(int),
4236                         .mode           = 0644,
4237                         .proc_handler   = proc_dointvec,
4238                 },
4239                 {
4240                         .procname       = "accept_ra",
4241                         .data           = &ipv6_devconf.accept_ra,
4242                         .maxlen         = sizeof(int),
4243                         .mode           = 0644,
4244                         .proc_handler   = proc_dointvec,
4245                 },
4246                 {
4247                         .procname       = "accept_redirects",
4248                         .data           = &ipv6_devconf.accept_redirects,
4249                         .maxlen         = sizeof(int),
4250                         .mode           = 0644,
4251                         .proc_handler   = proc_dointvec,
4252                 },
4253                 {
4254                         .procname       = "autoconf",
4255                         .data           = &ipv6_devconf.autoconf,
4256                         .maxlen         = sizeof(int),
4257                         .mode           = 0644,
4258                         .proc_handler   = proc_dointvec,
4259                 },
4260                 {
4261                         .procname       = "dad_transmits",
4262                         .data           = &ipv6_devconf.dad_transmits,
4263                         .maxlen         = sizeof(int),
4264                         .mode           = 0644,
4265                         .proc_handler   = proc_dointvec,
4266                 },
4267                 {
4268                         .procname       = "router_solicitations",
4269                         .data           = &ipv6_devconf.rtr_solicits,
4270                         .maxlen         = sizeof(int),
4271                         .mode           = 0644,
4272                         .proc_handler   = proc_dointvec,
4273                 },
4274                 {
4275                         .procname       = "router_solicitation_interval",
4276                         .data           = &ipv6_devconf.rtr_solicit_interval,
4277                         .maxlen         = sizeof(int),
4278                         .mode           = 0644,
4279                         .proc_handler   = proc_dointvec_jiffies,
4280                 },
4281                 {
4282                         .procname       = "router_solicitation_delay",
4283                         .data           = &ipv6_devconf.rtr_solicit_delay,
4284                         .maxlen         = sizeof(int),
4285                         .mode           = 0644,
4286                         .proc_handler   = proc_dointvec_jiffies,
4287                 },
4288                 {
4289                         .procname       = "force_mld_version",
4290                         .data           = &ipv6_devconf.force_mld_version,
4291                         .maxlen         = sizeof(int),
4292                         .mode           = 0644,
4293                         .proc_handler   = proc_dointvec,
4294                 },
4295 #ifdef CONFIG_IPV6_PRIVACY
4296                 {
4297                         .procname       = "use_tempaddr",
4298                         .data           = &ipv6_devconf.use_tempaddr,
4299                         .maxlen         = sizeof(int),
4300                         .mode           = 0644,
4301                         .proc_handler   = proc_dointvec,
4302                 },
4303                 {
4304                         .procname       = "temp_valid_lft",
4305                         .data           = &ipv6_devconf.temp_valid_lft,
4306                         .maxlen         = sizeof(int),
4307                         .mode           = 0644,
4308                         .proc_handler   = proc_dointvec,
4309                 },
4310                 {
4311                         .procname       = "temp_prefered_lft",
4312                         .data           = &ipv6_devconf.temp_prefered_lft,
4313                         .maxlen         = sizeof(int),
4314                         .mode           = 0644,
4315                         .proc_handler   = proc_dointvec,
4316                 },
4317                 {
4318                         .procname       = "regen_max_retry",
4319                         .data           = &ipv6_devconf.regen_max_retry,
4320                         .maxlen         = sizeof(int),
4321                         .mode           = 0644,
4322                         .proc_handler   = proc_dointvec,
4323                 },
4324                 {
4325                         .procname       = "max_desync_factor",
4326                         .data           = &ipv6_devconf.max_desync_factor,
4327                         .maxlen         = sizeof(int),
4328                         .mode           = 0644,
4329                         .proc_handler   = proc_dointvec,
4330                 },
4331 #endif
4332                 {
4333                         .procname       = "max_addresses",
4334                         .data           = &ipv6_devconf.max_addresses,
4335                         .maxlen         = sizeof(int),
4336                         .mode           = 0644,
4337                         .proc_handler   = proc_dointvec,
4338                 },
4339                 {
4340                         .procname       = "accept_ra_defrtr",
4341                         .data           = &ipv6_devconf.accept_ra_defrtr,
4342                         .maxlen         = sizeof(int),
4343                         .mode           = 0644,
4344                         .proc_handler   = proc_dointvec,
4345                 },
4346                 {
4347                         .procname       = "accept_ra_pinfo",
4348                         .data           = &ipv6_devconf.accept_ra_pinfo,
4349                         .maxlen         = sizeof(int),
4350                         .mode           = 0644,
4351                         .proc_handler   = proc_dointvec,
4352                 },
4353 #ifdef CONFIG_IPV6_ROUTER_PREF
4354                 {
4355                         .procname       = "accept_ra_rtr_pref",
4356                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4357                         .maxlen         = sizeof(int),
4358                         .mode           = 0644,
4359                         .proc_handler   = proc_dointvec,
4360                 },
4361                 {
4362                         .procname       = "router_probe_interval",
4363                         .data           = &ipv6_devconf.rtr_probe_interval,
4364                         .maxlen         = sizeof(int),
4365                         .mode           = 0644,
4366                         .proc_handler   = proc_dointvec_jiffies,
4367                 },
4368 #ifdef CONFIG_IPV6_ROUTE_INFO
4369                 {
4370                         .procname       = "accept_ra_rt_info_max_plen",
4371                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4372                         .maxlen         = sizeof(int),
4373                         .mode           = 0644,
4374                         .proc_handler   = proc_dointvec,
4375                 },
4376 #endif
4377 #endif
4378                 {
4379                         .procname       = "proxy_ndp",
4380                         .data           = &ipv6_devconf.proxy_ndp,
4381                         .maxlen         = sizeof(int),
4382                         .mode           = 0644,
4383                         .proc_handler   = proc_dointvec,
4384                 },
4385                 {
4386                         .procname       = "accept_source_route",
4387                         .data           = &ipv6_devconf.accept_source_route,
4388                         .maxlen         = sizeof(int),
4389                         .mode           = 0644,
4390                         .proc_handler   = proc_dointvec,
4391                 },
4392 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4393                 {
4394                         .procname       = "optimistic_dad",
4395                         .data           = &ipv6_devconf.optimistic_dad,
4396                         .maxlen         = sizeof(int),
4397                         .mode           = 0644,
4398                         .proc_handler   = proc_dointvec,
4399
4400                 },
4401 #endif
4402 #ifdef CONFIG_IPV6_MROUTE
4403                 {
4404                         .procname       = "mc_forwarding",
4405                         .data           = &ipv6_devconf.mc_forwarding,
4406                         .maxlen         = sizeof(int),
4407                         .mode           = 0444,
4408                         .proc_handler   = proc_dointvec,
4409                 },
4410 #endif
4411                 {
4412                         .procname       = "disable_ipv6",
4413                         .data           = &ipv6_devconf.disable_ipv6,
4414                         .maxlen         = sizeof(int),
4415                         .mode           = 0644,
4416                         .proc_handler   = addrconf_sysctl_disable,
4417                 },
4418                 {
4419                         .procname       = "accept_dad",
4420                         .data           = &ipv6_devconf.accept_dad,
4421                         .maxlen         = sizeof(int),
4422                         .mode           = 0644,
4423                         .proc_handler   = proc_dointvec,
4424                 },
4425                 {
4426                         .procname       = "force_tllao",
4427                         .data           = &ipv6_devconf.force_tllao,
4428                         .maxlen         = sizeof(int),
4429                         .mode           = 0644,
4430                         .proc_handler   = proc_dointvec
4431                 },
4432                 {
4433                         /* sentinel */
4434                 }
4435         },
4436 };
4437
4438 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4439                 struct inet6_dev *idev, struct ipv6_devconf *p)
4440 {
4441         int i;
4442         struct addrconf_sysctl_table *t;
4443
4444 #define ADDRCONF_CTL_PATH_DEV   3
4445
4446         struct ctl_path addrconf_ctl_path[] = {
4447                 { .procname = "net", },
4448                 { .procname = "ipv6", },
4449                 { .procname = "conf", },
4450                 { /* to be set */ },
4451                 { },
4452         };
4453
4454
4455         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4456         if (t == NULL)
4457                 goto out;
4458
4459         for (i = 0; t->addrconf_vars[i].data; i++) {
4460                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4461                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4462                 t->addrconf_vars[i].extra2 = net;
4463         }
4464
4465         /*
4466          * Make a copy of dev_name, because '.procname' is regarded as const
4467          * by sysctl and we wouldn't want anyone to change it under our feet
4468          * (see SIOCSIFNAME).
4469          */
4470         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4471         if (!t->dev_name)
4472                 goto free;
4473
4474         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4475
4476         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4477                         t->addrconf_vars);
4478         if (t->sysctl_header == NULL)
4479                 goto free_procname;
4480
4481         p->sysctl = t;
4482         return 0;
4483
4484 free_procname:
4485         kfree(t->dev_name);
4486 free:
4487         kfree(t);
4488 out:
4489         return -ENOBUFS;
4490 }
4491
4492 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4493 {
4494         struct addrconf_sysctl_table *t;
4495
4496         if (p->sysctl == NULL)
4497                 return;
4498
4499         t = p->sysctl;
4500         p->sysctl = NULL;
4501         unregister_sysctl_table(t->sysctl_header);
4502         kfree(t->dev_name);
4503         kfree(t);
4504 }
4505
4506 static void addrconf_sysctl_register(struct inet6_dev *idev)
4507 {
4508         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4509                               &ndisc_ifinfo_sysctl_change);
4510         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4511                                         idev, &idev->cnf);
4512 }
4513
4514 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4515 {
4516         __addrconf_sysctl_unregister(&idev->cnf);
4517         neigh_sysctl_unregister(idev->nd_parms);
4518 }
4519
4520
4521 #endif
4522
4523 static int __net_init addrconf_init_net(struct net *net)
4524 {
4525         int err;
4526         struct ipv6_devconf *all, *dflt;
4527
4528         err = -ENOMEM;
4529         all = &ipv6_devconf;
4530         dflt = &ipv6_devconf_dflt;
4531
4532         if (!net_eq(net, &init_net)) {
4533                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4534                 if (all == NULL)
4535                         goto err_alloc_all;
4536
4537                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4538                 if (dflt == NULL)
4539                         goto err_alloc_dflt;
4540         } else {
4541                 /* these will be inherited by all namespaces */
4542                 dflt->autoconf = ipv6_defaults.autoconf;
4543                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4544         }
4545
4546         net->ipv6.devconf_all = all;
4547         net->ipv6.devconf_dflt = dflt;
4548
4549 #ifdef CONFIG_SYSCTL
4550         err = __addrconf_sysctl_register(net, "all", NULL, all);
4551         if (err < 0)
4552                 goto err_reg_all;
4553
4554         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4555         if (err < 0)
4556                 goto err_reg_dflt;
4557 #endif
4558         return 0;
4559
4560 #ifdef CONFIG_SYSCTL
4561 err_reg_dflt:
4562         __addrconf_sysctl_unregister(all);
4563 err_reg_all:
4564         kfree(dflt);
4565 #endif
4566 err_alloc_dflt:
4567         kfree(all);
4568 err_alloc_all:
4569         return err;
4570 }
4571
4572 static void __net_exit addrconf_exit_net(struct net *net)
4573 {
4574 #ifdef CONFIG_SYSCTL
4575         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4576         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4577 #endif
4578         if (!net_eq(net, &init_net)) {
4579                 kfree(net->ipv6.devconf_dflt);
4580                 kfree(net->ipv6.devconf_all);
4581         }
4582 }
4583
4584 static struct pernet_operations addrconf_ops = {
4585         .init = addrconf_init_net,
4586         .exit = addrconf_exit_net,
4587 };
4588
4589 /*
4590  *      Device notifier
4591  */
4592
4593 int register_inet6addr_notifier(struct notifier_block *nb)
4594 {
4595         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4596 }
4597 EXPORT_SYMBOL(register_inet6addr_notifier);
4598
4599 int unregister_inet6addr_notifier(struct notifier_block *nb)
4600 {
4601         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4602 }
4603 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4604
4605 /*
4606  *      Init / cleanup code
4607  */
4608
4609 int __init addrconf_init(void)
4610 {
4611         int i, err;
4612
4613         err = ipv6_addr_label_init();
4614         if (err < 0) {
4615                 printk(KERN_CRIT "IPv6 Addrconf:"
4616                        " cannot initialize default policy table: %d.\n", err);
4617                 return err;
4618         }
4619
4620         register_pernet_subsys(&addrconf_ops);
4621
4622         /* The addrconf netdev notifier requires that loopback_dev
4623          * has it's ipv6 private information allocated and setup
4624          * before it can bring up and give link-local addresses
4625          * to other devices which are up.
4626          *
4627          * Unfortunately, loopback_dev is not necessarily the first
4628          * entry in the global dev_base list of net devices.  In fact,
4629          * it is likely to be the very last entry on that list.
4630          * So this causes the notifier registry below to try and
4631          * give link-local addresses to all devices besides loopback_dev
4632          * first, then loopback_dev, which cases all the non-loopback_dev
4633          * devices to fail to get a link-local address.
4634          *
4635          * So, as a temporary fix, allocate the ipv6 structure for
4636          * loopback_dev first by hand.
4637          * Longer term, all of the dependencies ipv6 has upon the loopback
4638          * device and it being up should be removed.
4639          */
4640         rtnl_lock();
4641         if (!ipv6_add_dev(init_net.loopback_dev))
4642                 err = -ENOMEM;
4643         rtnl_unlock();
4644         if (err)
4645                 goto errlo;
4646
4647         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4648                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4649
4650         register_netdevice_notifier(&ipv6_dev_notf);
4651
4652         addrconf_verify(0);
4653
4654         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4655         if (err < 0)
4656                 goto errout;
4657
4658         /* Only the first call to __rtnl_register can fail */
4659         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4660         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4661         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4662         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4663         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4664
4665         ipv6_addr_label_rtnl_register();
4666
4667         return 0;
4668 errout:
4669         unregister_netdevice_notifier(&ipv6_dev_notf);
4670 errlo:
4671         unregister_pernet_subsys(&addrconf_ops);
4672
4673         return err;
4674 }
4675
4676 void addrconf_cleanup(void)
4677 {
4678         struct net_device *dev;
4679         int i;
4680
4681         unregister_netdevice_notifier(&ipv6_dev_notf);
4682         unregister_pernet_subsys(&addrconf_ops);
4683
4684         rtnl_lock();
4685
4686         /* clean dev list */
4687         for_each_netdev(&init_net, dev) {
4688                 if (__in6_dev_get(dev) == NULL)
4689                         continue;
4690                 addrconf_ifdown(dev, 1);
4691         }
4692         addrconf_ifdown(init_net.loopback_dev, 2);
4693
4694         /*
4695          *      Check hash table.
4696          */
4697         spin_lock_bh(&addrconf_hash_lock);
4698         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4699                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4700         spin_unlock_bh(&addrconf_hash_lock);
4701
4702         del_timer(&addr_chk_timer);
4703         rtnl_unlock();
4704 }