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