net/ipv6/addrconf.c

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