git.oblomov.eu Git - linux-2.6/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/in6.h>
  47 #include <linux/netdevice.h>
  48 #include <linux/if_addr.h>
  49 #include <linux/if_arp.h>
  50 #include <linux/if_arcnet.h>
  51 #include <linux/if_infiniband.h>
  52 #include <linux/route.h>
  53 #include <linux/inetdevice.h>
  54 #include <linux/init.h>
  55 #ifdef CONFIG_SYSCTL
  56 #include <linux/sysctl.h>
  57 #endif
  58 #include <linux/capability.h>
  59 #include <linux/delay.h>
  60 #include <linux/notifier.h>
  61 #include <linux/string.h>
  62
  63 #include <net/net_namespace.h>
  64 #include <net/sock.h>
  65 #include <net/snmp.h>
  66
  67 #include <net/ipv6.h>
  68 #include <net/protocol.h>
  69 #include <net/ndisc.h>
  70 #include <net/ip6_route.h>
  71 #include <net/addrconf.h>
  72 #include <net/tcp.h>
  73 #include <net/ip.h>
  74 #include <net/netlink.h>
  75 #include <net/pkt_sched.h>
  76 #include <linux/if_tunnel.h>
  77 #include <linux/rtnetlink.h>
  78
  79 #ifdef CONFIG_IPV6_PRIVACY
  80 #include <linux/random.h>
  81 #endif
  82
  83 #include <asm/uaccess.h>
  84 #include <asm/unaligned.h>
  85
  86 #include <linux/proc_fs.h>
  87 #include <linux/seq_file.h>
  88
  89 /* Set to 3 to get tracing... */
  90 #define ACONF_DEBUG 2
  91
  92 #if ACONF_DEBUG >= 3
  93 #define ADBG(x) printk x
  94 #else
  95 #define ADBG(x)
  96 #endif
  97
  98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 100
 101 #ifdef CONFIG_SYSCTL
 102 static void addrconf_sysctl_register(struct inet6_dev *idev);
 103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 104 #else
 105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 106 {
 107 }
 108
 109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 110 {
 111 }
 112 #endif
 113
 114 #ifdef CONFIG_IPV6_PRIVACY
 115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 117 static void ipv6_regen_rndid(unsigned long data);
 118
 119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 120 #endif
 121
 122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 123 static int ipv6_count_addresses(struct inet6_dev *idev);
 124
 125 /*
 126  *      Configured unicast address hash table
 127  */
 128 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 129 static DEFINE_RWLOCK(addrconf_hash_lock);
 130
 131 static void addrconf_verify(unsigned long);
 132
 133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 134 static DEFINE_SPINLOCK(addrconf_verify_lock);
 135
 136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 138
 139 static int addrconf_ifdown(struct net_device *dev, int how);
 140
 141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 142 static void addrconf_dad_timer(unsigned long data);
 143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 144 static void addrconf_dad_run(struct inet6_dev *idev);
 145 static void addrconf_rs_timer(unsigned long data);
 146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148
 149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 150                                 struct prefix_info *pinfo);
 151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 152                               struct net_device *dev);
 153
 154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 155
 156 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 157         .forwarding             = 0,
 158         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 159         .mtu6                   = IPV6_MIN_MTU,
 160         .accept_ra              = 1,
 161         .accept_redirects       = 1,
 162         .autoconf               = 1,
 163         .force_mld_version      = 0,
 164         .dad_transmits          = 1,
 165         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 166         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 167         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 168 #ifdef CONFIG_IPV6_PRIVACY
 169         .use_tempaddr           = 0,
 170         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 171         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 172         .regen_max_retry        = REGEN_MAX_RETRY,
 173         .max_desync_factor      = MAX_DESYNC_FACTOR,
 174 #endif
 175         .max_addresses          = IPV6_MAX_ADDRESSES,
 176         .accept_ra_defrtr       = 1,
 177         .accept_ra_pinfo        = 1,
 178 #ifdef CONFIG_IPV6_ROUTER_PREF
 179         .accept_ra_rtr_pref     = 1,
 180         .rtr_probe_interval     = 60 * HZ,
 181 #ifdef CONFIG_IPV6_ROUTE_INFO
 182         .accept_ra_rt_info_max_plen = 0,
 183 #endif
 184 #endif
 185         .proxy_ndp              = 0,
 186         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 187         .disable_ipv6           = 0,
 188         .accept_dad             = 1,
 189 };
 190
 191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 192         .forwarding             = 0,
 193         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 194         .mtu6                   = IPV6_MIN_MTU,
 195         .accept_ra              = 1,
 196         .accept_redirects       = 1,
 197         .autoconf               = 1,
 198         .dad_transmits          = 1,
 199         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 200         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202 #ifdef CONFIG_IPV6_PRIVACY
 203         .use_tempaddr           = 0,
 204         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 205         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 206         .regen_max_retry        = REGEN_MAX_RETRY,
 207         .max_desync_factor      = MAX_DESYNC_FACTOR,
 208 #endif
 209         .max_addresses          = IPV6_MAX_ADDRESSES,
 210         .accept_ra_defrtr       = 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_max_plen = 0,
 217 #endif
 218 #endif
 219         .proxy_ndp              = 0,
 220         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 221         .disable_ipv6           = 0,
 222         .accept_dad             = 1,
 223 };
 224
 225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 230
 231 /* Check if a valid qdisc is available */
 232 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 233 {
 234         return !qdisc_tx_is_noop(dev);
 235 }
 236
 237 /* Check if a route is valid prefix route */
 238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 239 {
 240         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 241 }
 242
 243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 244 {
 245         if (del_timer(&ifp->timer))
 246                 __in6_ifa_put(ifp);
 247 }
 248
 249 enum addrconf_timer_t
 250 {
 251         AC_NONE,
 252         AC_DAD,
 253         AC_RS,
 254 };
 255
 256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 257                                enum addrconf_timer_t what,
 258                                unsigned long when)
 259 {
 260         if (!del_timer(&ifp->timer))
 261                 in6_ifa_hold(ifp);
 262
 263         switch (what) {
 264         case AC_DAD:
 265                 ifp->timer.function = addrconf_dad_timer;
 266                 break;
 267         case AC_RS:
 268                 ifp->timer.function = addrconf_rs_timer;
 269                 break;
 270         default:;
 271         }
 272         ifp->timer.expires = jiffies + when;
 273         add_timer(&ifp->timer);
 274 }
 275
 276 static int snmp6_alloc_dev(struct inet6_dev *idev)
 277 {
 278         if (snmp_mib_init((void **)idev->stats.ipv6,
 279                           sizeof(struct ipstats_mib)) < 0)
 280                 goto err_ip;
 281         if (snmp_mib_init((void **)idev->stats.icmpv6,
 282                           sizeof(struct icmpv6_mib)) < 0)
 283                 goto err_icmp;
 284         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 285                           sizeof(struct icmpv6msg_mib)) < 0)
 286                 goto err_icmpmsg;
 287
 288         return 0;
 289
 290 err_icmpmsg:
 291         snmp_mib_free((void **)idev->stats.icmpv6);
 292 err_icmp:
 293         snmp_mib_free((void **)idev->stats.ipv6);
 294 err_ip:
 295         return -ENOMEM;
 296 }
 297
 298 static void snmp6_free_dev(struct inet6_dev *idev)
 299 {
 300         snmp_mib_free((void **)idev->stats.icmpv6msg);
 301         snmp_mib_free((void **)idev->stats.icmpv6);
 302         snmp_mib_free((void **)idev->stats.ipv6);
 303 }
 304
 305 /* Nobody refers to this device, we may destroy it. */
 306
 307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 308 {
 309         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 310         kfree(idev);
 311 }
 312
 313 void in6_dev_finish_destroy(struct inet6_dev *idev)
 314 {
 315         struct net_device *dev = idev->dev;
 316
 317         WARN_ON(idev->addr_list != NULL);
 318         WARN_ON(idev->mc_list != NULL);
 319
 320 #ifdef NET_REFCNT_DEBUG
 321         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 322 #endif
 323         dev_put(dev);
 324         if (!idev->dead) {
 325                 printk("Freeing alive inet6 device %p\n", idev);
 326                 return;
 327         }
 328         snmp6_free_dev(idev);
 329         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 330 }
 331
 332 EXPORT_SYMBOL(in6_dev_finish_destroy);
 333
 334 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 335 {
 336         struct inet6_dev *ndev;
 337
 338         ASSERT_RTNL();
 339
 340         if (dev->mtu < IPV6_MIN_MTU)
 341                 return NULL;
 342
 343         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 344
 345         if (ndev == NULL)
 346                 return NULL;
 347
 348         rwlock_init(&ndev->lock);
 349         ndev->dev = dev;
 350         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 351         ndev->cnf.mtu6 = dev->mtu;
 352         ndev->cnf.sysctl = NULL;
 353         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 354         if (ndev->nd_parms == NULL) {
 355                 kfree(ndev);
 356                 return NULL;
 357         }
 358         if (ndev->cnf.forwarding)
 359                 dev_disable_lro(dev);
 360         /* We refer to the device */
 361         dev_hold(dev);
 362
 363         if (snmp6_alloc_dev(ndev) < 0) {
 364                 ADBG((KERN_WARNING
 365                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 366                         __func__, dev->name));
 367                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 368                 ndev->dead = 1;
 369                 in6_dev_finish_destroy(ndev);
 370                 return NULL;
 371         }
 372
 373         if (snmp6_register_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot create /proc/net/dev_snmp6/%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         /* One reference from device.  We must do this before
 384          * we invoke __ipv6_regen_rndid().
 385          */
 386         in6_dev_hold(ndev);
 387
 388         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 389                 ndev->cnf.accept_dad = -1;
 390
 391 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 392         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 393                 printk(KERN_INFO
 394                        "%s: Disabled Multicast RS\n",
 395                        dev->name);
 396                 ndev->cnf.rtr_solicits = 0;
 397         }
 398 #endif
 399
 400 #ifdef CONFIG_IPV6_PRIVACY
 401         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 402         if ((dev->flags&IFF_LOOPBACK) ||
 403             dev->type == ARPHRD_TUNNEL ||
 404             dev->type == ARPHRD_TUNNEL6 ||
 405             dev->type == ARPHRD_SIT ||
 406             dev->type == ARPHRD_NONE) {
 407                 printk(KERN_INFO
 408                        "%s: Disabled Privacy Extensions\n",
 409                        dev->name);
 410                 ndev->cnf.use_tempaddr = -1;
 411         } else {
 412                 in6_dev_hold(ndev);
 413                 ipv6_regen_rndid((unsigned long) ndev);
 414         }
 415 #endif
 416
 417         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 418                 ndev->if_flags |= IF_READY;
 419
 420         ipv6_mc_init_dev(ndev);
 421         ndev->tstamp = jiffies;
 422         addrconf_sysctl_register(ndev);
 423         /* protected by rtnl_lock */
 424         rcu_assign_pointer(dev->ip6_ptr, ndev);
 425
 426         /* Join all-node multicast group */
 427         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 428
 429         return ndev;
 430 }
 431
 432 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 433 {
 434         struct inet6_dev *idev;
 435
 436         ASSERT_RTNL();
 437
 438         if ((idev = __in6_dev_get(dev)) == NULL) {
 439                 if ((idev = ipv6_add_dev(dev)) == NULL)
 440                         return NULL;
 441         }
 442
 443         if (dev->flags&IFF_UP)
 444                 ipv6_mc_up(idev);
 445         return idev;
 446 }
 447
 448 #ifdef CONFIG_SYSCTL
 449 static void dev_forward_change(struct inet6_dev *idev)
 450 {
 451         struct net_device *dev;
 452         struct inet6_ifaddr *ifa;
 453
 454         if (!idev)
 455                 return;
 456         dev = idev->dev;
 457         if (idev->cnf.forwarding)
 458                 dev_disable_lro(dev);
 459         if (dev && (dev->flags & IFF_MULTICAST)) {
 460                 if (idev->cnf.forwarding)
 461                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 462                 else
 463                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 464         }
 465         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 466                 if (ifa->flags&IFA_F_TENTATIVE)
 467                         continue;
 468                 if (idev->cnf.forwarding)
 469                         addrconf_join_anycast(ifa);
 470                 else
 471                         addrconf_leave_anycast(ifa);
 472         }
 473 }
 474
 475
 476 static void addrconf_forward_change(struct net *net, __s32 newf)
 477 {
 478         struct net_device *dev;
 479         struct inet6_dev *idev;
 480
 481         read_lock(&dev_base_lock);
 482         for_each_netdev(net, dev) {
 483                 rcu_read_lock();
 484                 idev = __in6_dev_get(dev);
 485                 if (idev) {
 486                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 487                         idev->cnf.forwarding = newf;
 488                         if (changed)
 489                                 dev_forward_change(idev);
 490                 }
 491                 rcu_read_unlock();
 492         }
 493         read_unlock(&dev_base_lock);
 494 }
 495
 496 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 497 {
 498         struct net *net;
 499
 500         net = (struct net *)table->extra2;
 501         if (p == &net->ipv6.devconf_dflt->forwarding)
 502                 return 0;
 503
 504         if (!rtnl_trylock())
 505                 return -ERESTARTSYS;
 506
 507         if (p == &net->ipv6.devconf_all->forwarding) {
 508                 __s32 newf = net->ipv6.devconf_all->forwarding;
 509                 net->ipv6.devconf_dflt->forwarding = newf;
 510                 addrconf_forward_change(net, newf);
 511         } else if ((!*p) ^ (!old))
 512                 dev_forward_change((struct inet6_dev *)table->extra1);
 513         rtnl_unlock();
 514
 515         if (*p)
 516                 rt6_purge_dflt_routers(net);
 517         return 1;
 518 }
 519 #endif
 520
 521 /* Nobody refers to this ifaddr, destroy it */
 522
 523 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 524 {
 525         WARN_ON(ifp->if_next != NULL);
 526         WARN_ON(ifp->lst_next != NULL);
 527
 528 #ifdef NET_REFCNT_DEBUG
 529         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 530 #endif
 531
 532         in6_dev_put(ifp->idev);
 533
 534         if (del_timer(&ifp->timer))
 535                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 536
 537         if (!ifp->dead) {
 538                 printk("Freeing alive inet6 address %p\n", ifp);
 539                 return;
 540         }
 541         dst_release(&ifp->rt->u.dst);
 542
 543         kfree(ifp);
 544 }
 545
 546 static void
 547 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 548 {
 549         struct inet6_ifaddr *ifa, **ifap;
 550         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 551
 552         /*
 553          * Each device address list is sorted in order of scope -
 554          * global before linklocal.
 555          */
 556         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 557              ifap = &ifa->if_next) {
 558                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 559                         break;
 560         }
 561
 562         ifp->if_next = *ifap;
 563         *ifap = ifp;
 564 }
 565
 566 /*
 567  *      Hash function taken from net_alias.c
 568  */
 569 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 570 {
 571         __u32 word;
 572
 573         /*
 574          * We perform the hash function over the last 64 bits of the address
 575          * This will include the IEEE address token on links that support it.
 576          */
 577
 578         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 579         word ^= (word >> 16);
 580         word ^= (word >> 8);
 581
 582         return ((word ^ (word >> 4)) & 0x0f);
 583 }
 584
 585 /* On success it returns ifp with increased reference count */
 586
 587 static struct inet6_ifaddr *
 588 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 589               int scope, u32 flags)
 590 {
 591         struct inet6_ifaddr *ifa = NULL;
 592         struct rt6_info *rt;
 593         int hash;
 594         int err = 0;
 595         int addr_type = ipv6_addr_type(addr);
 596
 597         if (addr_type == IPV6_ADDR_ANY ||
 598             addr_type & IPV6_ADDR_MULTICAST ||
 599             (!(idev->dev->flags & IFF_LOOPBACK) &&
 600              addr_type & IPV6_ADDR_LOOPBACK))
 601                 return ERR_PTR(-EADDRNOTAVAIL);
 602
 603         rcu_read_lock_bh();
 604         if (idev->dead) {
 605                 err = -ENODEV;                  /*XXX*/
 606                 goto out2;
 607         }
 608
 609         write_lock(&addrconf_hash_lock);
 610
 611         /* Ignore adding duplicate addresses on an interface */
 612         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 613                 ADBG(("ipv6_add_addr: already assigned\n"));
 614                 err = -EEXIST;
 615                 goto out;
 616         }
 617
 618         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 619
 620         if (ifa == NULL) {
 621                 ADBG(("ipv6_add_addr: malloc failed\n"));
 622                 err = -ENOBUFS;
 623                 goto out;
 624         }
 625
 626         rt = addrconf_dst_alloc(idev, addr, 0);
 627         if (IS_ERR(rt)) {
 628                 err = PTR_ERR(rt);
 629                 goto out;
 630         }
 631
 632         ipv6_addr_copy(&ifa->addr, addr);
 633
 634         spin_lock_init(&ifa->lock);
 635         init_timer(&ifa->timer);
 636         ifa->timer.data = (unsigned long) ifa;
 637         ifa->scope = scope;
 638         ifa->prefix_len = pfxlen;
 639         ifa->flags = flags | IFA_F_TENTATIVE;
 640         ifa->cstamp = ifa->tstamp = jiffies;
 641
 642         ifa->rt = rt;
 643
 644         /*
 645          * part one of RFC 4429, section 3.3
 646          * We should not configure an address as
 647          * optimistic if we do not yet know the link
 648          * layer address of our nexhop router
 649          */
 650
 651         if (rt->rt6i_nexthop == NULL)
 652                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 653
 654         ifa->idev = idev;
 655         in6_dev_hold(idev);
 656         /* For caller */
 657         in6_ifa_hold(ifa);
 658
 659         /* Add to big hash table */
 660         hash = ipv6_addr_hash(addr);
 661
 662         ifa->lst_next = inet6_addr_lst[hash];
 663         inet6_addr_lst[hash] = ifa;
 664         in6_ifa_hold(ifa);
 665         write_unlock(&addrconf_hash_lock);
 666
 667         write_lock(&idev->lock);
 668         /* Add to inet6_dev unicast addr list. */
 669         ipv6_link_dev_addr(idev, ifa);
 670
 671 #ifdef CONFIG_IPV6_PRIVACY
 672         if (ifa->flags&IFA_F_TEMPORARY) {
 673                 ifa->tmp_next = idev->tempaddr_list;
 674                 idev->tempaddr_list = ifa;
 675                 in6_ifa_hold(ifa);
 676         }
 677 #endif
 678
 679         in6_ifa_hold(ifa);
 680         write_unlock(&idev->lock);
 681 out2:
 682         rcu_read_unlock_bh();
 683
 684         if (likely(err == 0))
 685                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 686         else {
 687                 kfree(ifa);
 688                 ifa = ERR_PTR(err);
 689         }
 690
 691         return ifa;
 692 out:
 693         write_unlock(&addrconf_hash_lock);
 694         goto out2;
 695 }
 696
 697 /* This function wants to get referenced ifp and releases it before return */
 698
 699 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 700 {
 701         struct inet6_ifaddr *ifa, **ifap;
 702         struct inet6_dev *idev = ifp->idev;
 703         int hash;
 704         int deleted = 0, onlink = 0;
 705         unsigned long expires = jiffies;
 706
 707         hash = ipv6_addr_hash(&ifp->addr);
 708
 709         ifp->dead = 1;
 710
 711         write_lock_bh(&addrconf_hash_lock);
 712         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 713              ifap = &ifa->lst_next) {
 714                 if (ifa == ifp) {
 715                         *ifap = ifa->lst_next;
 716                         __in6_ifa_put(ifp);
 717                         ifa->lst_next = NULL;
 718                         break;
 719                 }
 720         }
 721         write_unlock_bh(&addrconf_hash_lock);
 722
 723         write_lock_bh(&idev->lock);
 724 #ifdef CONFIG_IPV6_PRIVACY
 725         if (ifp->flags&IFA_F_TEMPORARY) {
 726                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 727                      ifap = &ifa->tmp_next) {
 728                         if (ifa == ifp) {
 729                                 *ifap = ifa->tmp_next;
 730                                 if (ifp->ifpub) {
 731                                         in6_ifa_put(ifp->ifpub);
 732                                         ifp->ifpub = NULL;
 733                                 }
 734                                 __in6_ifa_put(ifp);
 735                                 ifa->tmp_next = NULL;
 736                                 break;
 737                         }
 738                 }
 739         }
 740 #endif
 741
 742         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 743                 if (ifa == ifp) {
 744                         *ifap = ifa->if_next;
 745                         __in6_ifa_put(ifp);
 746                         ifa->if_next = NULL;
 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                 ifap = &ifa->if_next;
 780         }
 781         write_unlock_bh(&idev->lock);
 782
 783         addrconf_del_timer(ifp);
 784
 785         ipv6_ifa_notify(RTM_DELADDR, ifp);
 786
 787         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 788
 789         /*
 790          * Purge or update corresponding prefix
 791          *
 792          * 1) we don't purge prefix here if address was not permanent.
 793          *    prefix is managed by its own lifetime.
 794          * 2) if there're no addresses, delete prefix.
 795          * 3) if there're still other permanent address(es),
 796          *    corresponding prefix is still permanent.
 797          * 4) otherwise, update prefix lifetime to the
 798          *    longest valid lifetime among the corresponding
 799          *    addresses on the device.
 800          *    Note: subsequent RA will update lifetime.
 801          *
 802          * --yoshfuji
 803          */
 804         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 805                 struct in6_addr prefix;
 806                 struct rt6_info *rt;
 807                 struct net *net = dev_net(ifp->idev->dev);
 808                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 809                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 810
 811                 if (rt && addrconf_is_prefix_route(rt)) {
 812                         if (onlink == 0) {
 813                                 ip6_del_rt(rt);
 814                                 rt = NULL;
 815                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 816                                 rt->rt6i_expires = expires;
 817                                 rt->rt6i_flags |= RTF_EXPIRES;
 818                         }
 819                 }
 820                 dst_release(&rt->u.dst);
 821         }
 822
 823         in6_ifa_put(ifp);
 824 }
 825
 826 #ifdef CONFIG_IPV6_PRIVACY
 827 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 828 {
 829         struct inet6_dev *idev = ifp->idev;
 830         struct in6_addr addr, *tmpaddr;
 831         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 832         unsigned long regen_advance;
 833         int tmp_plen;
 834         int ret = 0;
 835         int max_addresses;
 836         u32 addr_flags;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         tmp_valid_lft = min_t(__u32,
 882                               ifp->valid_lft,
 883                               idev->cnf.temp_valid_lft);
 884         tmp_prefered_lft = min_t(__u32,
 885                                  ifp->prefered_lft,
 886                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 887         tmp_plen = ifp->prefix_len;
 888         max_addresses = idev->cnf.max_addresses;
 889         tmp_cstamp = ifp->cstamp;
 890         tmp_tstamp = ifp->tstamp;
 891         spin_unlock_bh(&ifp->lock);
 892
 893         regen_advance = idev->cnf.regen_max_retry *
 894                         idev->cnf.dad_transmits *
 895                         idev->nd_parms->retrans_time / HZ;
 896         write_unlock(&idev->lock);
 897
 898         /* A temporary address is created only if this calculated Preferred
 899          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 900          * an implementation must not create a temporary address with a zero
 901          * Preferred Lifetime.
 902          */
 903         if (tmp_prefered_lft <= regen_advance) {
 904                 in6_ifa_put(ifp);
 905                 in6_dev_put(idev);
 906                 ret = -1;
 907                 goto out;
 908         }
 909
 910         addr_flags = IFA_F_TEMPORARY;
 911         /* set in addrconf_prefix_rcv() */
 912         if (ifp->flags & IFA_F_OPTIMISTIC)
 913                 addr_flags |= IFA_F_OPTIMISTIC;
 914
 915         ift = !max_addresses ||
 916               ipv6_count_addresses(idev) < max_addresses ?
 917                 ipv6_add_addr(idev, &addr, tmp_plen,
 918                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 919                               addr_flags) : NULL;
 920         if (!ift || IS_ERR(ift)) {
 921                 in6_ifa_put(ifp);
 922                 in6_dev_put(idev);
 923                 printk(KERN_INFO
 924                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 925                 tmpaddr = &addr;
 926                 write_lock(&idev->lock);
 927                 goto retry;
 928         }
 929
 930         spin_lock_bh(&ift->lock);
 931         ift->ifpub = ifp;
 932         ift->valid_lft = tmp_valid_lft;
 933         ift->prefered_lft = tmp_prefered_lft;
 934         ift->cstamp = tmp_cstamp;
 935         ift->tstamp = tmp_tstamp;
 936         spin_unlock_bh(&ift->lock);
 937
 938         addrconf_dad_start(ift, 0);
 939         in6_ifa_put(ift);
 940         in6_dev_put(idev);
 941 out:
 942         return ret;
 943 }
 944 #endif
 945
 946 /*
 947  *      Choose an appropriate source address (RFC3484)
 948  */
 949 enum {
 950         IPV6_SADDR_RULE_INIT = 0,
 951         IPV6_SADDR_RULE_LOCAL,
 952         IPV6_SADDR_RULE_SCOPE,
 953         IPV6_SADDR_RULE_PREFERRED,
 954 #ifdef CONFIG_IPV6_MIP6
 955         IPV6_SADDR_RULE_HOA,
 956 #endif
 957         IPV6_SADDR_RULE_OIF,
 958         IPV6_SADDR_RULE_LABEL,
 959 #ifdef CONFIG_IPV6_PRIVACY
 960         IPV6_SADDR_RULE_PRIVACY,
 961 #endif
 962         IPV6_SADDR_RULE_ORCHID,
 963         IPV6_SADDR_RULE_PREFIX,
 964         IPV6_SADDR_RULE_MAX
 965 };
 966
 967 struct ipv6_saddr_score {
 968         int                     rule;
 969         int                     addr_type;
 970         struct inet6_ifaddr     *ifa;
 971         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 972         int                     scopedist;
 973         int                     matchlen;
 974 };
 975
 976 struct ipv6_saddr_dst {
 977         const struct in6_addr *addr;
 978         int ifindex;
 979         int scope;
 980         int label;
 981         unsigned int prefs;
 982 };
 983
 984 static inline int ipv6_saddr_preferred(int type)
 985 {
 986         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 987                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 988                 return 1;
 989         return 0;
 990 }
 991
 992 static int ipv6_get_saddr_eval(struct net *net,
 993                                struct ipv6_saddr_score *score,
 994                                struct ipv6_saddr_dst *dst,
 995                                int i)
 996 {
 997         int ret;
 998
 999         if (i <= score->rule) {
1000                 switch (i) {
1001                 case IPV6_SADDR_RULE_SCOPE:
1002                         ret = score->scopedist;
1003                         break;
1004                 case IPV6_SADDR_RULE_PREFIX:
1005                         ret = score->matchlen;
1006                         break;
1007                 default:
1008                         ret = !!test_bit(i, score->scorebits);
1009                 }
1010                 goto out;
1011         }
1012
1013         switch (i) {
1014         case IPV6_SADDR_RULE_INIT:
1015                 /* Rule 0: remember if hiscore is not ready yet */
1016                 ret = !!score->ifa;
1017                 break;
1018         case IPV6_SADDR_RULE_LOCAL:
1019                 /* Rule 1: Prefer same address */
1020                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1021                 break;
1022         case IPV6_SADDR_RULE_SCOPE:
1023                 /* Rule 2: Prefer appropriate scope
1024                  *
1025                  *      ret
1026                  *       ^
1027                  *    -1 |  d 15
1028                  *    ---+--+-+---> scope
1029                  *       |
1030                  *       |             d is scope of the destination.
1031                  *  B-d  |  \
1032                  *       |   \      <- smaller scope is better if
1033                  *  B-15 |    \        if scope is enough for destinaion.
1034                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1035                  * d-C-1 | /
1036                  *       |/         <- greater is better
1037                  *   -C  /             if scope is not enough for destination.
1038                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1039                  *
1040                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1041                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1042                  * Assume B = 0 and we get C > 29.
1043                  */
1044                 ret = __ipv6_addr_src_scope(score->addr_type);
1045                 if (ret >= dst->scope)
1046                         ret = -ret;
1047                 else
1048                         ret -= 128;     /* 30 is enough */
1049                 score->scopedist = ret;
1050                 break;
1051         case IPV6_SADDR_RULE_PREFERRED:
1052                 /* Rule 3: Avoid deprecated and optimistic addresses */
1053                 ret = ipv6_saddr_preferred(score->addr_type) ||
1054                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1055                 break;
1056 #ifdef CONFIG_IPV6_MIP6
1057         case IPV6_SADDR_RULE_HOA:
1058             {
1059                 /* Rule 4: Prefer home address */
1060                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1061                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1062                 break;
1063             }
1064 #endif
1065         case IPV6_SADDR_RULE_OIF:
1066                 /* Rule 5: Prefer outgoing interface */
1067                 ret = (!dst->ifindex ||
1068                        dst->ifindex == score->ifa->idev->dev->ifindex);
1069                 break;
1070         case IPV6_SADDR_RULE_LABEL:
1071                 /* Rule 6: Prefer matching label */
1072                 ret = ipv6_addr_label(net,
1073                                       &score->ifa->addr, score->addr_type,
1074                                       score->ifa->idev->dev->ifindex) == dst->label;
1075                 break;
1076 #ifdef CONFIG_IPV6_PRIVACY
1077         case IPV6_SADDR_RULE_PRIVACY:
1078             {
1079                 /* Rule 7: Prefer public address
1080                  * Note: prefer temprary address if use_tempaddr >= 2
1081                  */
1082                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1083                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1084                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1085                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1086                 break;
1087             }
1088 #endif
1089         case IPV6_SADDR_RULE_ORCHID:
1090                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1091                  *          non-ORCHID vs non-ORCHID
1092                  */
1093                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1094                         ipv6_addr_orchid(dst->addr));
1095                 break;
1096         case IPV6_SADDR_RULE_PREFIX:
1097                 /* Rule 8: Use longest matching prefix */
1098                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1099                                                        dst->addr);
1100                 break;
1101         default:
1102                 ret = 0;
1103         }
1104
1105         if (ret)
1106                 __set_bit(i, score->scorebits);
1107         score->rule = i;
1108 out:
1109         return ret;
1110 }
1111
1112 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1113                        const struct in6_addr *daddr, unsigned int prefs,
1114                        struct in6_addr *saddr)
1115 {
1116         struct ipv6_saddr_score scores[2],
1117                                 *score = &scores[0], *hiscore = &scores[1];
1118         struct ipv6_saddr_dst dst;
1119         struct net_device *dev;
1120         int dst_type;
1121
1122         dst_type = __ipv6_addr_type(daddr);
1123         dst.addr = daddr;
1124         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1125         dst.scope = __ipv6_addr_src_scope(dst_type);
1126         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1127         dst.prefs = prefs;
1128
1129         hiscore->rule = -1;
1130         hiscore->ifa = NULL;
1131
1132         read_lock(&dev_base_lock);
1133         rcu_read_lock();
1134
1135         for_each_netdev(net, dev) {
1136                 struct inet6_dev *idev;
1137
1138                 /* Candidate Source Address (section 4)
1139                  *  - multicast and link-local destination address,
1140                  *    the set of candidate source address MUST only
1141                  *    include addresses assigned to interfaces
1142                  *    belonging to the same link as the outgoing
1143                  *    interface.
1144                  * (- For site-local destination addresses, the
1145                  *    set of candidate source addresses MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same site as the outgoing
1148                  *    interface.)
1149                  */
1150                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1151                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1152                     dst.ifindex && dev->ifindex != dst.ifindex)
1153                         continue;
1154
1155                 idev = __in6_dev_get(dev);
1156                 if (!idev)
1157                         continue;
1158
1159                 read_lock_bh(&idev->lock);
1160                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1161                         int i;
1162
1163                         /*
1164                          * - Tentative Address (RFC2462 section 5.4)
1165                          *  - A tentative address is not considered
1166                          *    "assigned to an interface" in the traditional
1167                          *    sense, unless it is also flagged as optimistic.
1168                          * - Candidate Source Address (section 4)
1169                          *  - In any case, anycast addresses, multicast
1170                          *    addresses, and the unspecified address MUST
1171                          *    NOT be included in a candidate set.
1172                          */
1173                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1174                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1175                                 continue;
1176
1177                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1178
1179                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1180                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1181                                 LIMIT_NETDEBUG(KERN_DEBUG
1182                                                "ADDRCONF: unspecified / multicast address "
1183                                                "assigned as unicast address on %s",
1184                                                dev->name);
1185                                 continue;
1186                         }
1187
1188                         score->rule = -1;
1189                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1190
1191                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1192                                 int minihiscore, miniscore;
1193
1194                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1195                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1196
1197                                 if (minihiscore > miniscore) {
1198                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1199                                             score->scopedist > 0) {
1200                                                 /*
1201                                                  * special case:
1202                                                  * each remaining entry
1203                                                  * has too small (not enough)
1204                                                  * scope, because ifa entries
1205                                                  * are sorted by their scope
1206                                                  * values.
1207                                                  */
1208                                                 goto try_nextdev;
1209                                         }
1210                                         break;
1211                                 } else if (minihiscore < miniscore) {
1212                                         struct ipv6_saddr_score *tmp;
1213
1214                                         if (hiscore->ifa)
1215                                                 in6_ifa_put(hiscore->ifa);
1216
1217                                         in6_ifa_hold(score->ifa);
1218
1219                                         tmp = hiscore;
1220                                         hiscore = score;
1221                                         score = tmp;
1222
1223                                         /* restore our iterator */
1224                                         score->ifa = hiscore->ifa;
1225
1226                                         break;
1227                                 }
1228                         }
1229                 }
1230 try_nextdev:
1231                 read_unlock_bh(&idev->lock);
1232         }
1233         rcu_read_unlock();
1234         read_unlock(&dev_base_lock);
1235
1236         if (!hiscore->ifa)
1237                 return -EADDRNOTAVAIL;
1238
1239         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240         in6_ifa_put(hiscore->ifa);
1241         return 0;
1242 }
1243
1244 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1245
1246 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1247                     unsigned char banned_flags)
1248 {
1249         struct inet6_dev *idev;
1250         int err = -EADDRNOTAVAIL;
1251
1252         rcu_read_lock();
1253         if ((idev = __in6_dev_get(dev)) != NULL) {
1254                 struct inet6_ifaddr *ifp;
1255
1256                 read_lock_bh(&idev->lock);
1257                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1258                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1259                                 ipv6_addr_copy(addr, &ifp->addr);
1260                                 err = 0;
1261                                 break;
1262                         }
1263                 }
1264                 read_unlock_bh(&idev->lock);
1265         }
1266         rcu_read_unlock();
1267         return err;
1268 }
1269
1270 static int ipv6_count_addresses(struct inet6_dev *idev)
1271 {
1272         int cnt = 0;
1273         struct inet6_ifaddr *ifp;
1274
1275         read_lock_bh(&idev->lock);
1276         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1277                 cnt++;
1278         read_unlock_bh(&idev->lock);
1279         return cnt;
1280 }
1281
1282 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1283                   struct net_device *dev, int strict)
1284 {
1285         struct inet6_ifaddr * ifp;
1286         u8 hash = ipv6_addr_hash(addr);
1287
1288         read_lock_bh(&addrconf_hash_lock);
1289         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1290                 if (!net_eq(dev_net(ifp->idev->dev), net))
1291                         continue;
1292                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1293                     !(ifp->flags&IFA_F_TENTATIVE)) {
1294                         if (dev == NULL || ifp->idev->dev == dev ||
1295                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1296                                 break;
1297                 }
1298         }
1299         read_unlock_bh(&addrconf_hash_lock);
1300         return ifp != NULL;
1301 }
1302 EXPORT_SYMBOL(ipv6_chk_addr);
1303
1304 static
1305 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1306                        struct net_device *dev)
1307 {
1308         struct inet6_ifaddr * ifp;
1309         u8 hash = ipv6_addr_hash(addr);
1310
1311         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1312                 if (!net_eq(dev_net(ifp->idev->dev), net))
1313                         continue;
1314                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1315                         if (dev == NULL || ifp->idev->dev == dev)
1316                                 break;
1317                 }
1318         }
1319         return ifp != NULL;
1320 }
1321
1322 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1323 {
1324         struct inet6_dev *idev;
1325         struct inet6_ifaddr *ifa;
1326         int     onlink;
1327
1328         onlink = 0;
1329         rcu_read_lock();
1330         idev = __in6_dev_get(dev);
1331         if (idev) {
1332                 read_lock_bh(&idev->lock);
1333                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1334                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1335                                                    ifa->prefix_len);
1336                         if (onlink)
1337                                 break;
1338                 }
1339                 read_unlock_bh(&idev->lock);
1340         }
1341         rcu_read_unlock();
1342         return onlink;
1343 }
1344
1345 EXPORT_SYMBOL(ipv6_chk_prefix);
1346
1347 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1348                                      struct net_device *dev, int strict)
1349 {
1350         struct inet6_ifaddr * ifp;
1351         u8 hash = ipv6_addr_hash(addr);
1352
1353         read_lock_bh(&addrconf_hash_lock);
1354         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1355                 if (!net_eq(dev_net(ifp->idev->dev), net))
1356                         continue;
1357                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1358                         if (dev == NULL || ifp->idev->dev == dev ||
1359                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1360                                 in6_ifa_hold(ifp);
1361                                 break;
1362                         }
1363                 }
1364         }
1365         read_unlock_bh(&addrconf_hash_lock);
1366
1367         return ifp;
1368 }
1369
1370 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1371 {
1372         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1373         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1374         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1375         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1376         int sk_ipv6only = ipv6_only_sock(sk);
1377         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1378         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1379         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1380
1381         if (!sk2_rcv_saddr && !sk_ipv6only)
1382                 return 1;
1383
1384         if (addr_type2 == IPV6_ADDR_ANY &&
1385             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1386                 return 1;
1387
1388         if (addr_type == IPV6_ADDR_ANY &&
1389             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1390                 return 1;
1391
1392         if (sk2_rcv_saddr6 &&
1393             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1394                 return 1;
1395
1396         if (addr_type == IPV6_ADDR_MAPPED &&
1397             !sk2_ipv6only &&
1398             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1399                 return 1;
1400
1401         return 0;
1402 }
1403
1404 /* Gets referenced address, destroys ifaddr */
1405
1406 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1407 {
1408         if (ifp->flags&IFA_F_PERMANENT) {
1409                 spin_lock_bh(&ifp->lock);
1410                 addrconf_del_timer(ifp);
1411                 ifp->flags |= IFA_F_TENTATIVE;
1412                 spin_unlock_bh(&ifp->lock);
1413                 in6_ifa_put(ifp);
1414 #ifdef CONFIG_IPV6_PRIVACY
1415         } else if (ifp->flags&IFA_F_TEMPORARY) {
1416                 struct inet6_ifaddr *ifpub;
1417                 spin_lock_bh(&ifp->lock);
1418                 ifpub = ifp->ifpub;
1419                 if (ifpub) {
1420                         in6_ifa_hold(ifpub);
1421                         spin_unlock_bh(&ifp->lock);
1422                         ipv6_create_tempaddr(ifpub, ifp);
1423                         in6_ifa_put(ifpub);
1424                 } else {
1425                         spin_unlock_bh(&ifp->lock);
1426                 }
1427                 ipv6_del_addr(ifp);
1428 #endif
1429         } else
1430                 ipv6_del_addr(ifp);
1431 }
1432
1433 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1434 {
1435         struct inet6_dev *idev = ifp->idev;
1436         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1437                 struct in6_addr addr;
1438
1439                 addr.s6_addr32[0] = htonl(0xfe800000);
1440                 addr.s6_addr32[1] = 0;
1441
1442                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1443                     ipv6_addr_equal(&ifp->addr, &addr)) {
1444                         /* DAD failed for link-local based on MAC address */
1445                         idev->cnf.disable_ipv6 = 1;
1446                 }
1447         }
1448
1449         if (net_ratelimit())
1450                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1451         addrconf_dad_stop(ifp);
1452 }
1453
1454 /* Join to solicited addr multicast group. */
1455
1456 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1457 {
1458         struct in6_addr maddr;
1459
1460         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1461                 return;
1462
1463         addrconf_addr_solict_mult(addr, &maddr);
1464         ipv6_dev_mc_inc(dev, &maddr);
1465 }
1466
1467 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1468 {
1469         struct in6_addr maddr;
1470
1471         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1472                 return;
1473
1474         addrconf_addr_solict_mult(addr, &maddr);
1475         __ipv6_dev_mc_dec(idev, &maddr);
1476 }
1477
1478 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1479 {
1480         struct in6_addr addr;
1481         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1482         if (ipv6_addr_any(&addr))
1483                 return;
1484         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1485 }
1486
1487 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1488 {
1489         struct in6_addr addr;
1490         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1491         if (ipv6_addr_any(&addr))
1492                 return;
1493         __ipv6_dev_ac_dec(ifp->idev, &addr);
1494 }
1495
1496 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1497 {
1498         if (dev->addr_len != ETH_ALEN)
1499                 return -1;
1500         memcpy(eui, dev->dev_addr, 3);
1501         memcpy(eui + 5, dev->dev_addr + 3, 3);
1502
1503         /*
1504          * The zSeries OSA network cards can be shared among various
1505          * OS instances, but the OSA cards have only one MAC address.
1506          * This leads to duplicate address conflicts in conjunction
1507          * with IPv6 if more than one instance uses the same card.
1508          *
1509          * The driver for these cards can deliver a unique 16-bit
1510          * identifier for each instance sharing the same card.  It is
1511          * placed instead of 0xFFFE in the interface identifier.  The
1512          * "u" bit of the interface identifier is not inverted in this
1513          * case.  Hence the resulting interface identifier has local
1514          * scope according to RFC2373.
1515          */
1516         if (dev->dev_id) {
1517                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1518                 eui[4] = dev->dev_id & 0xFF;
1519         } else {
1520                 eui[3] = 0xFF;
1521                 eui[4] = 0xFE;
1522                 eui[0] ^= 2;
1523         }
1524         return 0;
1525 }
1526
1527 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1528 {
1529         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1530         if (dev->addr_len != ARCNET_ALEN)
1531                 return -1;
1532         memset(eui, 0, 7);
1533         eui[7] = *(u8*)dev->dev_addr;
1534         return 0;
1535 }
1536
1537 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1538 {
1539         if (dev->addr_len != INFINIBAND_ALEN)
1540                 return -1;
1541         memcpy(eui, dev->dev_addr + 12, 8);
1542         eui[0] |= 2;
1543         return 0;
1544 }
1545
1546 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1547 {
1548         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1549                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1550                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1551                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1552                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1553                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1554         eui[1] = 0;
1555         eui[2] = 0x5E;
1556         eui[3] = 0xFE;
1557         memcpy(eui + 4, &addr, 4);
1558         return 0;
1559 }
1560 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1561
1562 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1563 {
1564         if (dev->priv_flags & IFF_ISATAP)
1565                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1566         return -1;
1567 }
1568
1569 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1570 {
1571         switch (dev->type) {
1572         case ARPHRD_ETHER:
1573         case ARPHRD_FDDI:
1574         case ARPHRD_IEEE802_TR:
1575                 return addrconf_ifid_eui48(eui, dev);
1576         case ARPHRD_ARCNET:
1577                 return addrconf_ifid_arcnet(eui, dev);
1578         case ARPHRD_INFINIBAND:
1579                 return addrconf_ifid_infiniband(eui, dev);
1580         case ARPHRD_SIT:
1581                 return addrconf_ifid_sit(eui, dev);
1582         }
1583         return -1;
1584 }
1585
1586 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 {
1588         int err = -1;
1589         struct inet6_ifaddr *ifp;
1590
1591         read_lock_bh(&idev->lock);
1592         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1593                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1594                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1595                         err = 0;
1596                         break;
1597                 }
1598         }
1599         read_unlock_bh(&idev->lock);
1600         return err;
1601 }
1602
1603 #ifdef CONFIG_IPV6_PRIVACY
1604 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1605 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 {
1607 regen:
1608         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1609         idev->rndid[0] &= ~0x02;
1610
1611         /*
1612          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1613          * check if generated address is not inappropriate
1614          *
1615          *  - Reserved subnet anycast (RFC 2526)
1616          *      11111101 11....11 1xxxxxxx
1617          *  - ISATAP (RFC4214) 6.1
1618          *      00-00-5E-FE-xx-xx-xx-xx
1619          *  - value 0
1620          *  - XXX: already assigned to an address on the device
1621          */
1622         if (idev->rndid[0] == 0xfd &&
1623             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1624             (idev->rndid[7]&0x80))
1625                 goto regen;
1626         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1627                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1628                         goto regen;
1629                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1630                         goto regen;
1631         }
1632
1633         return 0;
1634 }
1635
1636 static void ipv6_regen_rndid(unsigned long data)
1637 {
1638         struct inet6_dev *idev = (struct inet6_dev *) data;
1639         unsigned long expires;
1640
1641         rcu_read_lock_bh();
1642         write_lock_bh(&idev->lock);
1643
1644         if (idev->dead)
1645                 goto out;
1646
1647         if (__ipv6_regen_rndid(idev) < 0)
1648                 goto out;
1649
1650         expires = jiffies +
1651                 idev->cnf.temp_prefered_lft * HZ -
1652                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1653         if (time_before(expires, jiffies)) {
1654                 printk(KERN_WARNING
1655                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1656                         idev->dev->name);
1657                 goto out;
1658         }
1659
1660         if (!mod_timer(&idev->regen_timer, expires))
1661                 in6_dev_hold(idev);
1662
1663 out:
1664         write_unlock_bh(&idev->lock);
1665         rcu_read_unlock_bh();
1666         in6_dev_put(idev);
1667 }
1668
1669 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1670         int ret = 0;
1671
1672         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1673                 ret = __ipv6_regen_rndid(idev);
1674         return ret;
1675 }
1676 #endif
1677
1678 /*
1679  *      Add prefix route.
1680  */
1681
1682 static void
1683 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1684                       unsigned long expires, u32 flags)
1685 {
1686         struct fib6_config cfg = {
1687                 .fc_table = RT6_TABLE_PREFIX,
1688                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1689                 .fc_ifindex = dev->ifindex,
1690                 .fc_expires = expires,
1691                 .fc_dst_len = plen,
1692                 .fc_flags = RTF_UP | flags,
1693                 .fc_nlinfo.nl_net = dev_net(dev),
1694                 .fc_protocol = RTPROT_KERNEL,
1695         };
1696
1697         ipv6_addr_copy(&cfg.fc_dst, pfx);
1698
1699         /* Prevent useless cloning on PtP SIT.
1700            This thing is done here expecting that the whole
1701            class of non-broadcast devices need not cloning.
1702          */
1703 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1704         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1705                 cfg.fc_flags |= RTF_NONEXTHOP;
1706 #endif
1707
1708         ip6_route_add(&cfg);
1709 }
1710
1711 /* Create "default" multicast route to the interface */
1712
1713 static void addrconf_add_mroute(struct net_device *dev)
1714 {
1715         struct fib6_config cfg = {
1716                 .fc_table = RT6_TABLE_LOCAL,
1717                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1718                 .fc_ifindex = dev->ifindex,
1719                 .fc_dst_len = 8,
1720                 .fc_flags = RTF_UP,
1721                 .fc_nlinfo.nl_net = dev_net(dev),
1722         };
1723
1724         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1725
1726         ip6_route_add(&cfg);
1727 }
1728
1729 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1730 static void sit_route_add(struct net_device *dev)
1731 {
1732         struct fib6_config cfg = {
1733                 .fc_table = RT6_TABLE_MAIN,
1734                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1735                 .fc_ifindex = dev->ifindex,
1736                 .fc_dst_len = 96,
1737                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1738                 .fc_nlinfo.nl_net = dev_net(dev),
1739         };
1740
1741         /* prefix length - 96 bits "::d.d.d.d" */
1742         ip6_route_add(&cfg);
1743 }
1744 #endif
1745
1746 static void addrconf_add_lroute(struct net_device *dev)
1747 {
1748         struct in6_addr addr;
1749
1750         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1751         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1752 }
1753
1754 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1755 {
1756         struct inet6_dev *idev;
1757
1758         ASSERT_RTNL();
1759
1760         if ((idev = ipv6_find_idev(dev)) == NULL)
1761                 return NULL;
1762
1763         /* Add default multicast route */
1764         addrconf_add_mroute(dev);
1765
1766         /* Add link local route */
1767         addrconf_add_lroute(dev);
1768         return idev;
1769 }
1770
1771 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1772 {
1773         struct prefix_info *pinfo;
1774         __u32 valid_lft;
1775         __u32 prefered_lft;
1776         int addr_type;
1777         struct inet6_dev *in6_dev;
1778
1779         pinfo = (struct prefix_info *) opt;
1780
1781         if (len < sizeof(struct prefix_info)) {
1782                 ADBG(("addrconf: prefix option too short\n"));
1783                 return;
1784         }
1785
1786         /*
1787          *      Validation checks ([ADDRCONF], page 19)
1788          */
1789
1790         addr_type = ipv6_addr_type(&pinfo->prefix);
1791
1792         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1793                 return;
1794
1795         valid_lft = ntohl(pinfo->valid);
1796         prefered_lft = ntohl(pinfo->prefered);
1797
1798         if (prefered_lft > valid_lft) {
1799                 if (net_ratelimit())
1800                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1801                 return;
1802         }
1803
1804         in6_dev = in6_dev_get(dev);
1805
1806         if (in6_dev == NULL) {
1807                 if (net_ratelimit())
1808                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1809                 return;
1810         }
1811
1812         /*
1813          *      Two things going on here:
1814          *      1) Add routes for on-link prefixes
1815          *      2) Configure prefixes with the auto flag set
1816          */
1817
1818         if (pinfo->onlink) {
1819                 struct rt6_info *rt;
1820                 unsigned long rt_expires;
1821
1822                 /* Avoid arithmetic overflow. Really, we could
1823                  * save rt_expires in seconds, likely valid_lft,
1824                  * but it would require division in fib gc, that it
1825                  * not good.
1826                  */
1827                 if (HZ > USER_HZ)
1828                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1829                 else
1830                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1831
1832                 if (addrconf_finite_timeout(rt_expires))
1833                         rt_expires *= HZ;
1834
1835                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1836                                 dev->ifindex, 1);
1837
1838                 if (rt && addrconf_is_prefix_route(rt)) {
1839                         /* Autoconf prefix route */
1840                         if (valid_lft == 0) {
1841                                 ip6_del_rt(rt);
1842                                 rt = NULL;
1843                         } else if (addrconf_finite_timeout(rt_expires)) {
1844                                 /* not infinity */
1845                                 rt->rt6i_expires = jiffies + rt_expires;
1846                                 rt->rt6i_flags |= RTF_EXPIRES;
1847                         } else {
1848                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1849                                 rt->rt6i_expires = 0;
1850                         }
1851                 } else if (valid_lft) {
1852                         clock_t expires = 0;
1853                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1854                         if (addrconf_finite_timeout(rt_expires)) {
1855                                 /* not infinity */
1856                                 flags |= RTF_EXPIRES;
1857                                 expires = jiffies_to_clock_t(rt_expires);
1858                         }
1859                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1860                                               dev, expires, flags);
1861                 }
1862                 if (rt)
1863                         dst_release(&rt->u.dst);
1864         }
1865
1866         /* Try to figure out our local address for this prefix */
1867
1868         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1869                 struct inet6_ifaddr * ifp;
1870                 struct in6_addr addr;
1871                 int create = 0, update_lft = 0;
1872                 struct net *net = dev_net(dev);
1873
1874                 if (pinfo->prefix_len == 64) {
1875                         memcpy(&addr, &pinfo->prefix, 8);
1876                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1877                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1878                                 in6_dev_put(in6_dev);
1879                                 return;
1880                         }
1881                         goto ok;
1882                 }
1883                 if (net_ratelimit())
1884                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1885                                pinfo->prefix_len);
1886                 in6_dev_put(in6_dev);
1887                 return;
1888
1889 ok:
1890
1891                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1892
1893                 if (ifp == NULL && valid_lft) {
1894                         int max_addresses = in6_dev->cnf.max_addresses;
1895                         u32 addr_flags = 0;
1896
1897 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1898                         if (in6_dev->cnf.optimistic_dad &&
1899                             !net->ipv6.devconf_all->forwarding)
1900                                 addr_flags = IFA_F_OPTIMISTIC;
1901 #endif
1902
1903                         /* Do not allow to create too much of autoconfigured
1904                          * addresses; this would be too easy way to crash kernel.
1905                          */
1906                         if (!max_addresses ||
1907                             ipv6_count_addresses(in6_dev) < max_addresses)
1908                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1909                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1910                                                     addr_flags);
1911
1912                         if (!ifp || IS_ERR(ifp)) {
1913                                 in6_dev_put(in6_dev);
1914                                 return;
1915                         }
1916
1917                         update_lft = create = 1;
1918                         ifp->cstamp = jiffies;
1919                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1920                 }
1921
1922                 if (ifp) {
1923                         int flags;
1924                         unsigned long now;
1925 #ifdef CONFIG_IPV6_PRIVACY
1926                         struct inet6_ifaddr *ift;
1927 #endif
1928                         u32 stored_lft;
1929
1930                         /* update lifetime (RFC2462 5.5.3 e) */
1931                         spin_lock(&ifp->lock);
1932                         now = jiffies;
1933                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1934                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1935                         else
1936                                 stored_lft = 0;
1937                         if (!update_lft && stored_lft) {
1938                                 if (valid_lft > MIN_VALID_LIFETIME ||
1939                                     valid_lft > stored_lft)
1940                                         update_lft = 1;
1941                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1942                                         /* valid_lft <= stored_lft is always true */
1943                                         /* XXX: IPsec */
1944                                         update_lft = 0;
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                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
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         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
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 (idev->addr_list == NULL)
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 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2231 static void sit_add_v4_addrs(struct inet6_dev *idev)
2232 {
2233         struct inet6_ifaddr * ifp;
2234         struct in6_addr addr;
2235         struct net_device *dev;
2236         struct net *net = dev_net(idev->dev);
2237         int scope;
2238
2239         ASSERT_RTNL();
2240
2241         memset(&addr, 0, sizeof(struct in6_addr));
2242         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2243
2244         if (idev->dev->flags&IFF_POINTOPOINT) {
2245                 addr.s6_addr32[0] = htonl(0xfe800000);
2246                 scope = IFA_LINK;
2247         } else {
2248                 scope = IPV6_ADDR_COMPATv4;
2249         }
2250
2251         if (addr.s6_addr32[3]) {
2252                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2253                 if (!IS_ERR(ifp)) {
2254                         spin_lock_bh(&ifp->lock);
2255                         ifp->flags &= ~IFA_F_TENTATIVE;
2256                         spin_unlock_bh(&ifp->lock);
2257                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2258                         in6_ifa_put(ifp);
2259                 }
2260                 return;
2261         }
2262
2263         for_each_netdev(net, dev) {
2264                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2265                 if (in_dev && (dev->flags & IFF_UP)) {
2266                         struct in_ifaddr * ifa;
2267
2268                         int flag = scope;
2269
2270                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2271                                 int plen;
2272
2273                                 addr.s6_addr32[3] = ifa->ifa_local;
2274
2275                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2276                                         continue;
2277                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2278                                         if (idev->dev->flags&IFF_POINTOPOINT)
2279                                                 continue;
2280                                         flag |= IFA_HOST;
2281                                 }
2282                                 if (idev->dev->flags&IFF_POINTOPOINT)
2283                                         plen = 64;
2284                                 else
2285                                         plen = 96;
2286
2287                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2288                                                     IFA_F_PERMANENT);
2289                                 if (!IS_ERR(ifp)) {
2290                                         spin_lock_bh(&ifp->lock);
2291                                         ifp->flags &= ~IFA_F_TENTATIVE;
2292                                         spin_unlock_bh(&ifp->lock);
2293                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2294                                         in6_ifa_put(ifp);
2295                                 }
2296                         }
2297                 }
2298         }
2299 }
2300 #endif
2301
2302 static void init_loopback(struct net_device *dev)
2303 {
2304         struct inet6_dev  *idev;
2305         struct inet6_ifaddr * ifp;
2306
2307         /* ::1 */
2308
2309         ASSERT_RTNL();
2310
2311         if ((idev = ipv6_find_idev(dev)) == NULL) {
2312                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2313                 return;
2314         }
2315
2316         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2317         if (!IS_ERR(ifp)) {
2318                 spin_lock_bh(&ifp->lock);
2319                 ifp->flags &= ~IFA_F_TENTATIVE;
2320                 spin_unlock_bh(&ifp->lock);
2321                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2322                 in6_ifa_put(ifp);
2323         }
2324 }
2325
2326 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2327 {
2328         struct inet6_ifaddr * ifp;
2329         u32 addr_flags = IFA_F_PERMANENT;
2330
2331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2332         if (idev->cnf.optimistic_dad &&
2333             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2334                 addr_flags |= IFA_F_OPTIMISTIC;
2335 #endif
2336
2337
2338         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2339         if (!IS_ERR(ifp)) {
2340                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2341                 addrconf_dad_start(ifp, 0);
2342                 in6_ifa_put(ifp);
2343         }
2344 }
2345
2346 static void addrconf_dev_config(struct net_device *dev)
2347 {
2348         struct in6_addr addr;
2349         struct inet6_dev    * idev;
2350
2351         ASSERT_RTNL();
2352
2353         if ((dev->type != ARPHRD_ETHER) &&
2354             (dev->type != ARPHRD_FDDI) &&
2355             (dev->type != ARPHRD_IEEE802_TR) &&
2356             (dev->type != ARPHRD_ARCNET) &&
2357             (dev->type != ARPHRD_INFINIBAND)) {
2358                 /* Alas, we support only Ethernet autoconfiguration. */
2359                 return;
2360         }
2361
2362         idev = addrconf_add_dev(dev);
2363         if (idev == NULL)
2364                 return;
2365
2366         memset(&addr, 0, sizeof(struct in6_addr));
2367         addr.s6_addr32[0] = htonl(0xFE800000);
2368
2369         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2370                 addrconf_add_linklocal(idev, &addr);
2371 }
2372
2373 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2374 static void addrconf_sit_config(struct net_device *dev)
2375 {
2376         struct inet6_dev *idev;
2377
2378         ASSERT_RTNL();
2379
2380         /*
2381          * Configure the tunnel with one of our IPv4
2382          * addresses... we should configure all of
2383          * our v4 addrs in the tunnel
2384          */
2385
2386         if ((idev = ipv6_find_idev(dev)) == NULL) {
2387                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2388                 return;
2389         }
2390
2391         if (dev->priv_flags & IFF_ISATAP) {
2392                 struct in6_addr addr;
2393
2394                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2395                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2396                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2397                         addrconf_add_linklocal(idev, &addr);
2398                 return;
2399         }
2400
2401         sit_add_v4_addrs(idev);
2402
2403         if (dev->flags&IFF_POINTOPOINT) {
2404                 addrconf_add_mroute(dev);
2405                 addrconf_add_lroute(dev);
2406         } else
2407                 sit_route_add(dev);
2408 }
2409 #endif
2410
2411 static inline int
2412 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2413 {
2414         struct in6_addr lladdr;
2415
2416         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2417                 addrconf_add_linklocal(idev, &lladdr);
2418                 return 0;
2419         }
2420         return -1;
2421 }
2422
2423 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2424 {
2425         struct net_device *link_dev;
2426         struct net *net = dev_net(idev->dev);
2427
2428         /* first try to inherit the link-local address from the link device */
2429         if (idev->dev->iflink &&
2430             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2431                 if (!ipv6_inherit_linklocal(idev, link_dev))
2432                         return;
2433         }
2434         /* then try to inherit it from any device */
2435         for_each_netdev(net, link_dev) {
2436                 if (!ipv6_inherit_linklocal(idev, link_dev))
2437                         return;
2438         }
2439         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2440 }
2441
2442 /*
2443  * Autoconfigure tunnel with a link-local address so routing protocols,
2444  * DHCPv6, MLD etc. can be run over the virtual link
2445  */
2446
2447 static void addrconf_ip6_tnl_config(struct net_device *dev)
2448 {
2449         struct inet6_dev *idev;
2450
2451         ASSERT_RTNL();
2452
2453         if ((idev = addrconf_add_dev(dev)) == NULL) {
2454                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2455                 return;
2456         }
2457         ip6_tnl_add_linklocal(idev);
2458 }
2459
2460 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2461                            void * data)
2462 {
2463         struct net_device *dev = (struct net_device *) data;
2464         struct inet6_dev *idev = __in6_dev_get(dev);
2465         int run_pending = 0;
2466         int err;
2467
2468         switch(event) {
2469         case NETDEV_REGISTER:
2470                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2471                         idev = ipv6_add_dev(dev);
2472                         if (!idev)
2473                                 return notifier_from_errno(-ENOMEM);
2474                 }
2475                 break;
2476         case NETDEV_UP:
2477         case NETDEV_CHANGE:
2478                 if (dev->flags & IFF_SLAVE)
2479                         break;
2480
2481                 if (event == NETDEV_UP) {
2482                         if (!addrconf_qdisc_ok(dev)) {
2483                                 /* device is not ready yet. */
2484                                 printk(KERN_INFO
2485                                         "ADDRCONF(NETDEV_UP): %s: "
2486                                         "link is not ready\n",
2487                                         dev->name);
2488                                 break;
2489                         }
2490
2491                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2492                                 idev = ipv6_add_dev(dev);
2493
2494                         if (idev) {
2495                                 idev->if_flags |= IF_READY;
2496                                 run_pending = 1;
2497                         }
2498                 } else {
2499                         if (!addrconf_qdisc_ok(dev)) {
2500                                 /* device is still not ready. */
2501                                 break;
2502                         }
2503
2504                         if (idev) {
2505                                 if (idev->if_flags & IF_READY) {
2506                                         /* device is already configured. */
2507                                         break;
2508                                 }
2509                                 idev->if_flags |= IF_READY;
2510                         }
2511
2512                         printk(KERN_INFO
2513                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2514                                         "link becomes ready\n",
2515                                         dev->name);
2516
2517                         run_pending = 1;
2518                 }
2519
2520                 switch(dev->type) {
2521 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2522                 case ARPHRD_SIT:
2523                         addrconf_sit_config(dev);
2524                         break;
2525 #endif
2526                 case ARPHRD_TUNNEL6:
2527                         addrconf_ip6_tnl_config(dev);
2528                         break;
2529                 case ARPHRD_LOOPBACK:
2530                         init_loopback(dev);
2531                         break;
2532
2533                 default:
2534                         addrconf_dev_config(dev);
2535                         break;
2536                 }
2537                 if (idev) {
2538                         if (run_pending)
2539                                 addrconf_dad_run(idev);
2540
2541                         /* If the MTU changed during the interface down, when the
2542                            interface up, the changed MTU must be reflected in the
2543                            idev as well as routers.
2544                          */
2545                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2546                                 rt6_mtu_change(dev, dev->mtu);
2547                                 idev->cnf.mtu6 = dev->mtu;
2548                         }
2549                         idev->tstamp = jiffies;
2550                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2551                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2552                            stop IPv6 on this interface.
2553                          */
2554                         if (dev->mtu < IPV6_MIN_MTU)
2555                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2556                 }
2557                 break;
2558
2559         case NETDEV_CHANGEMTU:
2560                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2561                         rt6_mtu_change(dev, dev->mtu);
2562                         idev->cnf.mtu6 = dev->mtu;
2563                         break;
2564                 }
2565
2566                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2567                         idev = ipv6_add_dev(dev);
2568                         if (idev)
2569                                 break;
2570                 }
2571
2572                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
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
2594         return NOTIFY_OK;
2595 }
2596
2597 /*
2598  *      addrconf module should be notified of a device going up
2599  */
2600 static struct notifier_block ipv6_dev_notf = {
2601         .notifier_call = addrconf_notify,
2602         .priority = 0
2603 };
2604
2605 static int addrconf_ifdown(struct net_device *dev, int how)
2606 {
2607         struct inet6_dev *idev;
2608         struct inet6_ifaddr *ifa, **bifa;
2609         struct net *net = dev_net(dev);
2610         int i;
2611
2612         ASSERT_RTNL();
2613
2614         rt6_ifdown(net, dev);
2615         neigh_ifdown(&nd_tbl, dev);
2616
2617         idev = __in6_dev_get(dev);
2618         if (idev == NULL)
2619                 return -ENODEV;
2620
2621         /* Step 1: remove reference to ipv6 device from parent device.
2622                    Do not dev_put!
2623          */
2624         if (how) {
2625                 idev->dead = 1;
2626
2627                 /* protected by rtnl_lock */
2628                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2629
2630                 /* Step 1.5: remove snmp6 entry */
2631                 snmp6_unregister_dev(idev);
2632
2633         }
2634
2635         /* Step 2: clear hash table */
2636         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2637                 bifa = &inet6_addr_lst[i];
2638
2639                 write_lock_bh(&addrconf_hash_lock);
2640                 while ((ifa = *bifa) != NULL) {
2641                         if (ifa->idev == idev) {
2642                                 *bifa = ifa->lst_next;
2643                                 ifa->lst_next = NULL;
2644                                 addrconf_del_timer(ifa);
2645                                 in6_ifa_put(ifa);
2646                                 continue;
2647                         }
2648                         bifa = &ifa->lst_next;
2649                 }
2650                 write_unlock_bh(&addrconf_hash_lock);
2651         }
2652
2653         write_lock_bh(&idev->lock);
2654
2655         /* Step 3: clear flags for stateless addrconf */
2656         if (!how)
2657                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2658
2659         /* Step 4: clear address list */
2660 #ifdef CONFIG_IPV6_PRIVACY
2661         if (how && del_timer(&idev->regen_timer))
2662                 in6_dev_put(idev);
2663
2664         /* clear tempaddr list */
2665         while ((ifa = idev->tempaddr_list) != NULL) {
2666                 idev->tempaddr_list = ifa->tmp_next;
2667                 ifa->tmp_next = NULL;
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         while ((ifa = idev->addr_list) != NULL) {
2682                 idev->addr_list = ifa->if_next;
2683                 ifa->if_next = NULL;
2684                 ifa->dead = 1;
2685                 addrconf_del_timer(ifa);
2686                 write_unlock_bh(&idev->lock);
2687
2688                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2689                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2690                 in6_ifa_put(ifa);
2691
2692                 write_lock_bh(&idev->lock);
2693         }
2694         write_unlock_bh(&idev->lock);
2695
2696         /* Step 5: Discard multicast list */
2697
2698         if (how)
2699                 ipv6_mc_destroy_dev(idev);
2700         else
2701                 ipv6_mc_down(idev);
2702
2703         idev->tstamp = jiffies;
2704
2705         /* Shot the device (if unregistered) */
2706
2707         if (how) {
2708                 addrconf_sysctl_unregister(idev);
2709                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2710                 neigh_ifdown(&nd_tbl, dev);
2711                 in6_dev_put(idev);
2712         }
2713         return 0;
2714 }
2715
2716 static void addrconf_rs_timer(unsigned long data)
2717 {
2718         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2719
2720         if (ifp->idev->cnf.forwarding)
2721                 goto out;
2722
2723         if (ifp->idev->if_flags & IF_RA_RCVD) {
2724                 /*
2725                  *      Announcement received after solicitation
2726                  *      was sent
2727                  */
2728                 goto out;
2729         }
2730
2731         spin_lock(&ifp->lock);
2732         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2733                 /* The wait after the last probe can be shorter */
2734                 addrconf_mod_timer(ifp, AC_RS,
2735                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2736                                    ifp->idev->cnf.rtr_solicit_delay :
2737                                    ifp->idev->cnf.rtr_solicit_interval);
2738                 spin_unlock(&ifp->lock);
2739
2740                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2741         } else {
2742                 spin_unlock(&ifp->lock);
2743                 /*
2744                  * Note: we do not support deprecated "all on-link"
2745                  * assumption any longer.
2746                  */
2747                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2748                        ifp->idev->dev->name);
2749         }
2750
2751 out:
2752         in6_ifa_put(ifp);
2753 }
2754
2755 /*
2756  *      Duplicate Address Detection
2757  */
2758 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2759 {
2760         unsigned long rand_num;
2761         struct inet6_dev *idev = ifp->idev;
2762
2763         if (ifp->flags & IFA_F_OPTIMISTIC)
2764                 rand_num = 0;
2765         else
2766                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2767
2768         ifp->probes = idev->cnf.dad_transmits;
2769         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2770 }
2771
2772 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2773 {
2774         struct inet6_dev *idev = ifp->idev;
2775         struct net_device *dev = idev->dev;
2776
2777         addrconf_join_solict(dev, &ifp->addr);
2778
2779         net_srandom(ifp->addr.s6_addr32[3]);
2780
2781         read_lock_bh(&idev->lock);
2782         if (ifp->dead)
2783                 goto out;
2784         spin_lock_bh(&ifp->lock);
2785
2786         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2787             idev->cnf.accept_dad < 1 ||
2788             !(ifp->flags&IFA_F_TENTATIVE) ||
2789             ifp->flags & IFA_F_NODAD) {
2790                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2791                 spin_unlock_bh(&ifp->lock);
2792                 read_unlock_bh(&idev->lock);
2793
2794                 addrconf_dad_completed(ifp);
2795                 return;
2796         }
2797
2798         if (!(idev->if_flags & IF_READY)) {
2799                 spin_unlock_bh(&ifp->lock);
2800                 read_unlock_bh(&idev->lock);
2801                 /*
2802                  * If the defice is not ready:
2803                  * - keep it tentative if it is a permanent address.
2804                  * - otherwise, kill it.
2805                  */
2806                 in6_ifa_hold(ifp);
2807                 addrconf_dad_stop(ifp);
2808                 return;
2809         }
2810
2811         /*
2812          * Optimistic nodes can start receiving
2813          * Frames right away
2814          */
2815         if(ifp->flags & IFA_F_OPTIMISTIC)
2816                 ip6_ins_rt(ifp->rt);
2817
2818         addrconf_dad_kick(ifp);
2819         spin_unlock_bh(&ifp->lock);
2820 out:
2821         read_unlock_bh(&idev->lock);
2822 }
2823
2824 static void addrconf_dad_timer(unsigned long data)
2825 {
2826         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2827         struct inet6_dev *idev = ifp->idev;
2828         struct in6_addr mcaddr;
2829
2830         read_lock_bh(&idev->lock);
2831         if (idev->dead) {
2832                 read_unlock_bh(&idev->lock);
2833                 goto out;
2834         }
2835         if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2836                 read_unlock_bh(&idev->lock);
2837                 addrconf_dad_failure(ifp);
2838                 return;
2839         }
2840         spin_lock_bh(&ifp->lock);
2841         if (ifp->probes == 0) {
2842                 /*
2843                  * DAD was successful
2844                  */
2845
2846                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2847                 spin_unlock_bh(&ifp->lock);
2848                 read_unlock_bh(&idev->lock);
2849
2850                 addrconf_dad_completed(ifp);
2851
2852                 goto out;
2853         }
2854
2855         ifp->probes--;
2856         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2857         spin_unlock_bh(&ifp->lock);
2858         read_unlock_bh(&idev->lock);
2859
2860         /* send a neighbour solicitation for our addr */
2861         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2862         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2863 out:
2864         in6_ifa_put(ifp);
2865 }
2866
2867 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2868 {
2869         struct net_device *     dev = ifp->idev->dev;
2870
2871         /*
2872          *      Configure the address for reception. Now it is valid.
2873          */
2874
2875         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2876
2877         /* If added prefix is link local and forwarding is off,
2878            start sending router solicitations.
2879          */
2880
2881         if (ifp->idev->cnf.forwarding == 0 &&
2882             ifp->idev->cnf.rtr_solicits > 0 &&
2883             (dev->flags&IFF_LOOPBACK) == 0 &&
2884             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2885                 /*
2886                  *      If a host as already performed a random delay
2887                  *      [...] as part of DAD [...] there is no need
2888                  *      to delay again before sending the first RS
2889                  */
2890                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2891
2892                 spin_lock_bh(&ifp->lock);
2893                 ifp->probes = 1;
2894                 ifp->idev->if_flags |= IF_RS_SENT;
2895                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2896                 spin_unlock_bh(&ifp->lock);
2897         }
2898 }
2899
2900 static void addrconf_dad_run(struct inet6_dev *idev) {
2901         struct inet6_ifaddr *ifp;
2902
2903         read_lock_bh(&idev->lock);
2904         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2905                 spin_lock_bh(&ifp->lock);
2906                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2907                         spin_unlock_bh(&ifp->lock);
2908                         continue;
2909                 }
2910                 spin_unlock_bh(&ifp->lock);
2911                 addrconf_dad_kick(ifp);
2912         }
2913         read_unlock_bh(&idev->lock);
2914 }
2915
2916 #ifdef CONFIG_PROC_FS
2917 struct if6_iter_state {
2918         struct seq_net_private p;
2919         int bucket;
2920 };
2921
2922 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2923 {
2924         struct inet6_ifaddr *ifa = NULL;
2925         struct if6_iter_state *state = seq->private;
2926         struct net *net = seq_file_net(seq);
2927
2928         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2929                 ifa = inet6_addr_lst[state->bucket];
2930
2931                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2932                         ifa = ifa->lst_next;
2933                 if (ifa)
2934                         break;
2935         }
2936         return ifa;
2937 }
2938
2939 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2940 {
2941         struct if6_iter_state *state = seq->private;
2942         struct net *net = seq_file_net(seq);
2943
2944         ifa = ifa->lst_next;
2945 try_again:
2946         if (ifa) {
2947                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2948                         ifa = ifa->lst_next;
2949                         goto try_again;
2950                 }
2951         }
2952
2953         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2954                 ifa = inet6_addr_lst[state->bucket];
2955                 goto try_again;
2956         }
2957
2958         return ifa;
2959 }
2960
2961 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2962 {
2963         struct inet6_ifaddr *ifa = if6_get_first(seq);
2964
2965         if (ifa)
2966                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2967                         --pos;
2968         return pos ? NULL : ifa;
2969 }
2970
2971 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2972         __acquires(addrconf_hash_lock)
2973 {
2974         read_lock_bh(&addrconf_hash_lock);
2975         return if6_get_idx(seq, *pos);
2976 }
2977
2978 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2979 {
2980         struct inet6_ifaddr *ifa;
2981
2982         ifa = if6_get_next(seq, v);
2983         ++*pos;
2984         return ifa;
2985 }
2986
2987 static void if6_seq_stop(struct seq_file *seq, void *v)
2988         __releases(addrconf_hash_lock)
2989 {
2990         read_unlock_bh(&addrconf_hash_lock);
2991 }
2992
2993 static int if6_seq_show(struct seq_file *seq, void *v)
2994 {
2995         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2996         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2997                    &ifp->addr,
2998                    ifp->idev->dev->ifindex,
2999                    ifp->prefix_len,
3000                    ifp->scope,
3001                    ifp->flags,
3002                    ifp->idev->dev->name);
3003         return 0;
3004 }
3005
3006 static const struct seq_operations if6_seq_ops = {
3007         .start  = if6_seq_start,
3008         .next   = if6_seq_next,
3009         .show   = if6_seq_show,
3010         .stop   = if6_seq_stop,
3011 };
3012
3013 static int if6_seq_open(struct inode *inode, struct file *file)
3014 {
3015         return seq_open_net(inode, file, &if6_seq_ops,
3016                             sizeof(struct if6_iter_state));
3017 }
3018
3019 static const struct file_operations if6_fops = {
3020         .owner          = THIS_MODULE,
3021         .open           = if6_seq_open,
3022         .read           = seq_read,
3023         .llseek         = seq_lseek,
3024         .release        = seq_release_net,
3025 };
3026
3027 static int if6_proc_net_init(struct net *net)
3028 {
3029         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3030                 return -ENOMEM;
3031         return 0;
3032 }
3033
3034 static void if6_proc_net_exit(struct net *net)
3035 {
3036        proc_net_remove(net, "if_inet6");
3037 }
3038
3039 static struct pernet_operations if6_proc_net_ops = {
3040        .init = if6_proc_net_init,
3041        .exit = if6_proc_net_exit,
3042 };
3043
3044 int __init if6_proc_init(void)
3045 {
3046         return register_pernet_subsys(&if6_proc_net_ops);
3047 }
3048
3049 void if6_proc_exit(void)
3050 {
3051         unregister_pernet_subsys(&if6_proc_net_ops);
3052 }
3053 #endif  /* CONFIG_PROC_FS */
3054
3055 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3056 /* Check if address is a home address configured on any interface. */
3057 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3058 {
3059         int ret = 0;
3060         struct inet6_ifaddr * ifp;
3061         u8 hash = ipv6_addr_hash(addr);
3062         read_lock_bh(&addrconf_hash_lock);
3063         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3064                 if (!net_eq(dev_net(ifp->idev->dev), net))
3065                         continue;
3066                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3067                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3068                         ret = 1;
3069                         break;
3070                 }
3071         }
3072         read_unlock_bh(&addrconf_hash_lock);
3073         return ret;
3074 }
3075 #endif
3076
3077 /*
3078  *      Periodic address status verification
3079  */
3080
3081 static void addrconf_verify(unsigned long foo)
3082 {
3083         struct inet6_ifaddr *ifp;
3084         unsigned long now, next;
3085         int i;
3086
3087         spin_lock_bh(&addrconf_verify_lock);
3088         now = jiffies;
3089         next = now + ADDR_CHECK_FREQUENCY;
3090
3091         del_timer(&addr_chk_timer);
3092
3093         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3094
3095 restart:
3096                 read_lock(&addrconf_hash_lock);
3097                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3098                         unsigned long age;
3099 #ifdef CONFIG_IPV6_PRIVACY
3100                         unsigned long regen_advance;
3101 #endif
3102
3103                         if (ifp->flags & IFA_F_PERMANENT)
3104                                 continue;
3105
3106                         spin_lock(&ifp->lock);
3107                         age = (now - ifp->tstamp) / HZ;
3108
3109 #ifdef CONFIG_IPV6_PRIVACY
3110                         regen_advance = ifp->idev->cnf.regen_max_retry *
3111                                         ifp->idev->cnf.dad_transmits *
3112                                         ifp->idev->nd_parms->retrans_time / HZ;
3113 #endif
3114
3115                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3116                             age >= ifp->valid_lft) {
3117                                 spin_unlock(&ifp->lock);
3118                                 in6_ifa_hold(ifp);
3119                                 read_unlock(&addrconf_hash_lock);
3120                                 ipv6_del_addr(ifp);
3121                                 goto restart;
3122                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3123                                 spin_unlock(&ifp->lock);
3124                                 continue;
3125                         } else if (age >= ifp->prefered_lft) {
3126                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3127                                 int deprecate = 0;
3128
3129                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3130                                         deprecate = 1;
3131                                         ifp->flags |= IFA_F_DEPRECATED;
3132                                 }
3133
3134                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3135                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3136
3137                                 spin_unlock(&ifp->lock);
3138
3139                                 if (deprecate) {
3140                                         in6_ifa_hold(ifp);
3141                                         read_unlock(&addrconf_hash_lock);
3142
3143                                         ipv6_ifa_notify(0, ifp);
3144                                         in6_ifa_put(ifp);
3145                                         goto restart;
3146                                 }
3147 #ifdef CONFIG_IPV6_PRIVACY
3148                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3149                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3150                                 if (age >= ifp->prefered_lft - regen_advance) {
3151                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3152                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3153                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3154                                         if (!ifp->regen_count && ifpub) {
3155                                                 ifp->regen_count++;
3156                                                 in6_ifa_hold(ifp);
3157                                                 in6_ifa_hold(ifpub);
3158                                                 spin_unlock(&ifp->lock);
3159                                                 read_unlock(&addrconf_hash_lock);
3160                                                 spin_lock(&ifpub->lock);
3161                                                 ifpub->regen_count = 0;
3162                                                 spin_unlock(&ifpub->lock);
3163                                                 ipv6_create_tempaddr(ifpub, ifp);
3164                                                 in6_ifa_put(ifpub);
3165                                                 in6_ifa_put(ifp);
3166                                                 goto restart;
3167                                         }
3168                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3169                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3170                                 spin_unlock(&ifp->lock);
3171 #endif
3172                         } else {
3173                                 /* ifp->prefered_lft <= ifp->valid_lft */
3174                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3175                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3176                                 spin_unlock(&ifp->lock);
3177                         }
3178                 }
3179                 read_unlock(&addrconf_hash_lock);
3180         }
3181
3182         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3183         add_timer(&addr_chk_timer);
3184         spin_unlock_bh(&addrconf_verify_lock);
3185 }
3186
3187 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3188 {
3189         struct in6_addr *pfx = NULL;
3190
3191         if (addr)
3192                 pfx = nla_data(addr);
3193
3194         if (local) {
3195                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3196                         pfx = NULL;
3197                 else
3198                         pfx = nla_data(local);
3199         }
3200
3201         return pfx;
3202 }
3203
3204 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3205         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3206         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3207         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3208 };
3209
3210 static int
3211 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3212 {
3213         struct net *net = sock_net(skb->sk);
3214         struct ifaddrmsg *ifm;
3215         struct nlattr *tb[IFA_MAX+1];
3216         struct in6_addr *pfx;
3217         int err;
3218
3219         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3220         if (err < 0)
3221                 return err;
3222
3223         ifm = nlmsg_data(nlh);
3224         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3225         if (pfx == NULL)
3226                 return -EINVAL;
3227
3228         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3229 }
3230
3231 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3232                              u32 prefered_lft, u32 valid_lft)
3233 {
3234         u32 flags;
3235         clock_t expires;
3236         unsigned long timeout;
3237
3238         if (!valid_lft || (prefered_lft > valid_lft))
3239                 return -EINVAL;
3240
3241         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3242         if (addrconf_finite_timeout(timeout)) {
3243                 expires = jiffies_to_clock_t(timeout * HZ);
3244                 valid_lft = timeout;
3245                 flags = RTF_EXPIRES;
3246         } else {
3247                 expires = 0;
3248                 flags = 0;
3249                 ifa_flags |= IFA_F_PERMANENT;
3250         }
3251
3252         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3253         if (addrconf_finite_timeout(timeout)) {
3254                 if (timeout == 0)
3255                         ifa_flags |= IFA_F_DEPRECATED;
3256                 prefered_lft = timeout;
3257         }
3258
3259         spin_lock_bh(&ifp->lock);
3260         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3261         ifp->tstamp = jiffies;
3262         ifp->valid_lft = valid_lft;
3263         ifp->prefered_lft = prefered_lft;
3264
3265         spin_unlock_bh(&ifp->lock);
3266         if (!(ifp->flags&IFA_F_TENTATIVE))
3267                 ipv6_ifa_notify(0, ifp);
3268
3269         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3270                               expires, flags);
3271         addrconf_verify(0);
3272
3273         return 0;
3274 }
3275
3276 static int
3277 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3278 {
3279         struct net *net = sock_net(skb->sk);
3280         struct ifaddrmsg *ifm;
3281         struct nlattr *tb[IFA_MAX+1];
3282         struct in6_addr *pfx;
3283         struct inet6_ifaddr *ifa;
3284         struct net_device *dev;
3285         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3286         u8 ifa_flags;
3287         int err;
3288
3289         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3290         if (err < 0)
3291                 return err;
3292
3293         ifm = nlmsg_data(nlh);
3294         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3295         if (pfx == NULL)
3296                 return -EINVAL;
3297
3298         if (tb[IFA_CACHEINFO]) {
3299                 struct ifa_cacheinfo *ci;
3300
3301                 ci = nla_data(tb[IFA_CACHEINFO]);
3302                 valid_lft = ci->ifa_valid;
3303                 preferred_lft = ci->ifa_prefered;
3304         } else {
3305                 preferred_lft = INFINITY_LIFE_TIME;
3306                 valid_lft = INFINITY_LIFE_TIME;
3307         }
3308
3309         dev =  __dev_get_by_index(net, ifm->ifa_index);
3310         if (dev == NULL)
3311                 return -ENODEV;
3312
3313         /* We ignore other flags so far. */
3314         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3315
3316         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3317         if (ifa == NULL) {
3318                 /*
3319                  * It would be best to check for !NLM_F_CREATE here but
3320                  * userspace alreay relies on not having to provide this.
3321                  */
3322                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3323                                       ifm->ifa_prefixlen, ifa_flags,
3324                                       preferred_lft, valid_lft);
3325         }
3326
3327         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3328             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3329                 err = -EEXIST;
3330         else
3331                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3332
3333         in6_ifa_put(ifa);
3334
3335         return err;
3336 }
3337
3338 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3339                           u8 scope, int ifindex)
3340 {
3341         struct ifaddrmsg *ifm;
3342
3343         ifm = nlmsg_data(nlh);
3344         ifm->ifa_family = AF_INET6;
3345         ifm->ifa_prefixlen = prefixlen;
3346         ifm->ifa_flags = flags;
3347         ifm->ifa_scope = scope;
3348         ifm->ifa_index = ifindex;
3349 }
3350
3351 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3352                          unsigned long tstamp, u32 preferred, u32 valid)
3353 {
3354         struct ifa_cacheinfo ci;
3355
3356         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3357                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3358         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3359                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3360         ci.ifa_prefered = preferred;
3361         ci.ifa_valid = valid;
3362
3363         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3364 }
3365
3366 static inline int rt_scope(int ifa_scope)
3367 {
3368         if (ifa_scope & IFA_HOST)
3369                 return RT_SCOPE_HOST;
3370         else if (ifa_scope & IFA_LINK)
3371                 return RT_SCOPE_LINK;
3372         else if (ifa_scope & IFA_SITE)
3373                 return RT_SCOPE_SITE;
3374         else
3375                 return RT_SCOPE_UNIVERSE;
3376 }
3377
3378 static inline int inet6_ifaddr_msgsize(void)
3379 {
3380         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3381                + nla_total_size(16) /* IFA_ADDRESS */
3382                + nla_total_size(sizeof(struct ifa_cacheinfo));
3383 }
3384
3385 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3386                              u32 pid, u32 seq, int event, unsigned int flags)
3387 {
3388         struct nlmsghdr  *nlh;
3389         u32 preferred, valid;
3390
3391         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3392         if (nlh == NULL)
3393                 return -EMSGSIZE;
3394
3395         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3396                       ifa->idev->dev->ifindex);
3397
3398         if (!(ifa->flags&IFA_F_PERMANENT)) {
3399                 preferred = ifa->prefered_lft;
3400                 valid = ifa->valid_lft;
3401                 if (preferred != INFINITY_LIFE_TIME) {
3402                         long tval = (jiffies - ifa->tstamp)/HZ;
3403                         preferred -= tval;
3404                         if (valid != INFINITY_LIFE_TIME)
3405                                 valid -= tval;
3406                 }
3407         } else {
3408                 preferred = INFINITY_LIFE_TIME;
3409                 valid = INFINITY_LIFE_TIME;
3410         }
3411
3412         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3413             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3414                 nlmsg_cancel(skb, nlh);
3415                 return -EMSGSIZE;
3416         }
3417
3418         return nlmsg_end(skb, nlh);
3419 }
3420
3421 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3422                                 u32 pid, u32 seq, int event, u16 flags)
3423 {
3424         struct nlmsghdr  *nlh;
3425         u8 scope = RT_SCOPE_UNIVERSE;
3426         int ifindex = ifmca->idev->dev->ifindex;
3427
3428         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3429                 scope = RT_SCOPE_SITE;
3430
3431         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3432         if (nlh == NULL)
3433                 return -EMSGSIZE;
3434
3435         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3436         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3437             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3438                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3439                 nlmsg_cancel(skb, nlh);
3440                 return -EMSGSIZE;
3441         }
3442
3443         return nlmsg_end(skb, nlh);
3444 }
3445
3446 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3447                                 u32 pid, u32 seq, int event, unsigned int flags)
3448 {
3449         struct nlmsghdr  *nlh;
3450         u8 scope = RT_SCOPE_UNIVERSE;
3451         int ifindex = ifaca->aca_idev->dev->ifindex;
3452
3453         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3454                 scope = RT_SCOPE_SITE;
3455
3456         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3457         if (nlh == NULL)
3458                 return -EMSGSIZE;
3459
3460         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3461         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3462             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3463                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3464                 nlmsg_cancel(skb, nlh);
3465                 return -EMSGSIZE;
3466         }
3467
3468         return nlmsg_end(skb, nlh);
3469 }
3470
3471 enum addr_type_t
3472 {
3473         UNICAST_ADDR,
3474         MULTICAST_ADDR,
3475         ANYCAST_ADDR,
3476 };
3477
3478 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3479                            enum addr_type_t type)
3480 {
3481         int idx, ip_idx;
3482         int s_idx, s_ip_idx;
3483         int err = 1;
3484         struct net_device *dev;
3485         struct inet6_dev *idev = NULL;
3486         struct inet6_ifaddr *ifa;
3487         struct ifmcaddr6 *ifmca;
3488         struct ifacaddr6 *ifaca;
3489         struct net *net = sock_net(skb->sk);
3490
3491         s_idx = cb->args[0];
3492         s_ip_idx = ip_idx = cb->args[1];
3493
3494         idx = 0;
3495         for_each_netdev(net, dev) {
3496                 if (idx < s_idx)
3497                         goto cont;
3498                 if (idx > s_idx)
3499                         s_ip_idx = 0;
3500                 ip_idx = 0;
3501                 if ((idev = in6_dev_get(dev)) == NULL)
3502                         goto cont;
3503                 read_lock_bh(&idev->lock);
3504                 switch (type) {
3505                 case UNICAST_ADDR:
3506                         /* unicast address incl. temp addr */
3507                         for (ifa = idev->addr_list; ifa;
3508                              ifa = ifa->if_next, ip_idx++) {
3509                                 if (ip_idx < s_ip_idx)
3510                                         continue;
3511                                 err = inet6_fill_ifaddr(skb, ifa,
3512                                                         NETLINK_CB(cb->skb).pid,
3513                                                         cb->nlh->nlmsg_seq,
3514                                                         RTM_NEWADDR,
3515                                                         NLM_F_MULTI);
3516                         }
3517                         break;
3518                 case MULTICAST_ADDR:
3519                         /* multicast address */
3520                         for (ifmca = idev->mc_list; ifmca;
3521                              ifmca = ifmca->next, ip_idx++) {
3522                                 if (ip_idx < s_ip_idx)
3523                                         continue;
3524                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3525                                                           NETLINK_CB(cb->skb).pid,
3526                                                           cb->nlh->nlmsg_seq,
3527                                                           RTM_GETMULTICAST,
3528                                                           NLM_F_MULTI);
3529                         }
3530                         break;
3531                 case ANYCAST_ADDR:
3532                         /* anycast address */
3533                         for (ifaca = idev->ac_list; ifaca;
3534                              ifaca = ifaca->aca_next, ip_idx++) {
3535                                 if (ip_idx < s_ip_idx)
3536                                         continue;
3537                                 err = inet6_fill_ifacaddr(skb, ifaca,
3538                                                           NETLINK_CB(cb->skb).pid,
3539                                                           cb->nlh->nlmsg_seq,
3540                                                           RTM_GETANYCAST,
3541                                                           NLM_F_MULTI);
3542                         }
3543                         break;
3544                 default:
3545                         break;
3546                 }
3547                 read_unlock_bh(&idev->lock);
3548                 in6_dev_put(idev);
3549
3550                 if (err <= 0)
3551                         break;
3552 cont:
3553                 idx++;
3554         }
3555         cb->args[0] = idx;
3556         cb->args[1] = ip_idx;
3557         return skb->len;
3558 }
3559
3560 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3561 {
3562         enum addr_type_t type = UNICAST_ADDR;
3563
3564         return inet6_dump_addr(skb, cb, type);
3565 }
3566
3567 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3568 {
3569         enum addr_type_t type = MULTICAST_ADDR;
3570
3571         return inet6_dump_addr(skb, cb, type);
3572 }
3573
3574
3575 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3576 {
3577         enum addr_type_t type = ANYCAST_ADDR;
3578
3579         return inet6_dump_addr(skb, cb, type);
3580 }
3581
3582 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3583                              void *arg)
3584 {
3585         struct net *net = sock_net(in_skb->sk);
3586         struct ifaddrmsg *ifm;
3587         struct nlattr *tb[IFA_MAX+1];
3588         struct in6_addr *addr = NULL;
3589         struct net_device *dev = NULL;
3590         struct inet6_ifaddr *ifa;
3591         struct sk_buff *skb;
3592         int err;
3593
3594         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3595         if (err < 0)
3596                 goto errout;
3597
3598         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3599         if (addr == NULL) {
3600                 err = -EINVAL;
3601                 goto errout;
3602         }
3603
3604         ifm = nlmsg_data(nlh);
3605         if (ifm->ifa_index)
3606                 dev = __dev_get_by_index(net, ifm->ifa_index);
3607
3608         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3609                 err = -EADDRNOTAVAIL;
3610                 goto errout;
3611         }
3612
3613         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3614                 err = -ENOBUFS;
3615                 goto errout_ifa;
3616         }
3617
3618         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3619                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3620         if (err < 0) {
3621                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3622                 WARN_ON(err == -EMSGSIZE);
3623                 kfree_skb(skb);
3624                 goto errout_ifa;
3625         }
3626         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3627 errout_ifa:
3628         in6_ifa_put(ifa);
3629 errout:
3630         return err;
3631 }
3632
3633 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3634 {
3635         struct sk_buff *skb;
3636         struct net *net = dev_net(ifa->idev->dev);
3637         int err = -ENOBUFS;
3638
3639         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3640         if (skb == NULL)
3641                 goto errout;
3642
3643         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3644         if (err < 0) {
3645                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3646                 WARN_ON(err == -EMSGSIZE);
3647                 kfree_skb(skb);
3648                 goto errout;
3649         }
3650         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3651 errout:
3652         if (err < 0)
3653                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3654 }
3655
3656 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3657                                 __s32 *array, int bytes)
3658 {
3659         BUG_ON(bytes < (DEVCONF_MAX * 4));
3660
3661         memset(array, 0, bytes);
3662         array[DEVCONF_FORWARDING] = cnf->forwarding;
3663         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3664         array[DEVCONF_MTU6] = cnf->mtu6;
3665         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3666         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3667         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3668         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3669         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3670         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3671         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3672         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3673 #ifdef CONFIG_IPV6_PRIVACY
3674         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3675         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3676         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3677         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3678         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3679 #endif
3680         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3681         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3682         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3683 #ifdef CONFIG_IPV6_ROUTER_PREF
3684         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3685         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3686 #ifdef CONFIG_IPV6_ROUTE_INFO
3687         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3688 #endif
3689 #endif
3690         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3691         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3692 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3693         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3694 #endif
3695 #ifdef CONFIG_IPV6_MROUTE
3696         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3697 #endif
3698         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3699         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3700 }
3701
3702 static inline size_t inet6_if_nlmsg_size(void)
3703 {
3704         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3705                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3706                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3707                + nla_total_size(4) /* IFLA_MTU */
3708                + nla_total_size(4) /* IFLA_LINK */
3709                + nla_total_size( /* IFLA_PROTINFO */
3710                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3711                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3712                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3713                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3714                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3715                  );
3716 }
3717
3718 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3719                                       int bytes)
3720 {
3721         int i;
3722         int pad = bytes - sizeof(u64) * items;
3723         BUG_ON(pad < 0);
3724
3725         /* Use put_unaligned() because stats may not be aligned for u64. */
3726         put_unaligned(items, &stats[0]);
3727         for (i = 1; i < items; i++)
3728                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3729
3730         memset(&stats[items], 0, pad);
3731 }
3732
3733 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3734                              int bytes)
3735 {
3736         switch(attrtype) {
3737         case IFLA_INET6_STATS:
3738                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3739                 break;
3740         case IFLA_INET6_ICMP6STATS:
3741                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3742                 break;
3743         }
3744 }
3745
3746 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3747                              u32 pid, u32 seq, int event, unsigned int flags)
3748 {
3749         struct net_device *dev = idev->dev;
3750         struct nlattr *nla;
3751         struct ifinfomsg *hdr;
3752         struct nlmsghdr *nlh;
3753         void *protoinfo;
3754         struct ifla_cacheinfo ci;
3755
3756         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3757         if (nlh == NULL)
3758                 return -EMSGSIZE;
3759
3760         hdr = nlmsg_data(nlh);
3761         hdr->ifi_family = AF_INET6;
3762         hdr->__ifi_pad = 0;
3763         hdr->ifi_type = dev->type;
3764         hdr->ifi_index = dev->ifindex;
3765         hdr->ifi_flags = dev_get_flags(dev);
3766         hdr->ifi_change = 0;
3767
3768         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3769
3770         if (dev->addr_len)
3771                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3772
3773         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3774         if (dev->ifindex != dev->iflink)
3775                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3776
3777         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3778         if (protoinfo == NULL)
3779                 goto nla_put_failure;
3780
3781         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3782
3783         ci.max_reasm_len = IPV6_MAXPLEN;
3784         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3785                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3786         ci.reachable_time = idev->nd_parms->reachable_time;
3787         ci.retrans_time = idev->nd_parms->retrans_time;
3788         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3789
3790         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3791         if (nla == NULL)
3792                 goto nla_put_failure;
3793         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3794
3795         /* XXX - MC not implemented */
3796
3797         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3798         if (nla == NULL)
3799                 goto nla_put_failure;
3800         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3801
3802         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3803         if (nla == NULL)
3804                 goto nla_put_failure;
3805         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3806
3807         nla_nest_end(skb, protoinfo);
3808         return nlmsg_end(skb, nlh);
3809
3810 nla_put_failure:
3811         nlmsg_cancel(skb, nlh);
3812         return -EMSGSIZE;
3813 }
3814
3815 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3816 {
3817         struct net *net = sock_net(skb->sk);
3818         int idx, err;
3819         int s_idx = cb->args[0];
3820         struct net_device *dev;
3821         struct inet6_dev *idev;
3822
3823         read_lock(&dev_base_lock);
3824         idx = 0;
3825         for_each_netdev(net, dev) {
3826                 if (idx < s_idx)
3827                         goto cont;
3828                 if ((idev = in6_dev_get(dev)) == NULL)
3829                         goto cont;
3830                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3831                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3832                 in6_dev_put(idev);
3833                 if (err <= 0)
3834                         break;
3835 cont:
3836                 idx++;
3837         }
3838         read_unlock(&dev_base_lock);
3839         cb->args[0] = idx;
3840
3841         return skb->len;
3842 }
3843
3844 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3845 {
3846         struct sk_buff *skb;
3847         struct net *net = dev_net(idev->dev);
3848         int err = -ENOBUFS;
3849
3850         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3851         if (skb == NULL)
3852                 goto errout;
3853
3854         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3855         if (err < 0) {
3856                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3857                 WARN_ON(err == -EMSGSIZE);
3858                 kfree_skb(skb);
3859                 goto errout;
3860         }
3861         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3862 errout:
3863         if (err < 0)
3864                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3865 }
3866
3867 static inline size_t inet6_prefix_nlmsg_size(void)
3868 {
3869         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3870                + nla_total_size(sizeof(struct in6_addr))
3871                + nla_total_size(sizeof(struct prefix_cacheinfo));
3872 }
3873
3874 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3875                              struct prefix_info *pinfo, u32 pid, u32 seq,
3876                              int event, unsigned int flags)
3877 {
3878         struct prefixmsg *pmsg;
3879         struct nlmsghdr *nlh;
3880         struct prefix_cacheinfo ci;
3881
3882         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3883         if (nlh == NULL)
3884                 return -EMSGSIZE;
3885
3886         pmsg = nlmsg_data(nlh);
3887         pmsg->prefix_family = AF_INET6;
3888         pmsg->prefix_pad1 = 0;
3889         pmsg->prefix_pad2 = 0;
3890         pmsg->prefix_ifindex = idev->dev->ifindex;
3891         pmsg->prefix_len = pinfo->prefix_len;
3892         pmsg->prefix_type = pinfo->type;
3893         pmsg->prefix_pad3 = 0;
3894         pmsg->prefix_flags = 0;
3895         if (pinfo->onlink)
3896                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3897         if (pinfo->autoconf)
3898                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3899
3900         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3901
3902         ci.preferred_time = ntohl(pinfo->prefered);
3903         ci.valid_time = ntohl(pinfo->valid);
3904         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3905
3906         return nlmsg_end(skb, nlh);
3907
3908 nla_put_failure:
3909         nlmsg_cancel(skb, nlh);
3910         return -EMSGSIZE;
3911 }
3912
3913 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3914                          struct prefix_info *pinfo)
3915 {
3916         struct sk_buff *skb;
3917         struct net *net = dev_net(idev->dev);
3918         int err = -ENOBUFS;
3919
3920         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3921         if (skb == NULL)
3922                 goto errout;
3923
3924         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3925         if (err < 0) {
3926                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3927                 WARN_ON(err == -EMSGSIZE);
3928                 kfree_skb(skb);
3929                 goto errout;
3930         }
3931         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3932 errout:
3933         if (err < 0)
3934                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3935 }
3936
3937 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3938 {
3939         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3940
3941         switch (event) {
3942         case RTM_NEWADDR:
3943                 /*
3944                  * If the address was optimistic
3945                  * we inserted the route at the start of
3946                  * our DAD process, so we don't need
3947                  * to do it again
3948                  */
3949                 if (!(ifp->rt->rt6i_node))
3950                         ip6_ins_rt(ifp->rt);
3951                 if (ifp->idev->cnf.forwarding)
3952                         addrconf_join_anycast(ifp);
3953                 break;
3954         case RTM_DELADDR:
3955                 if (ifp->idev->cnf.forwarding)
3956                         addrconf_leave_anycast(ifp);
3957                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3958                 dst_hold(&ifp->rt->u.dst);
3959                 if (ip6_del_rt(ifp->rt))
3960                         dst_free(&ifp->rt->u.dst);
3961                 break;
3962         }
3963 }
3964
3965 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3966 {
3967         rcu_read_lock_bh();
3968         if (likely(ifp->idev->dead == 0))
3969                 __ipv6_ifa_notify(event, ifp);
3970         rcu_read_unlock_bh();
3971 }
3972
3973 #ifdef CONFIG_SYSCTL
3974
3975 static
3976 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3977                            void __user *buffer, size_t *lenp, loff_t *ppos)
3978 {
3979         int *valp = ctl->data;
3980         int val = *valp;
3981         int ret;
3982
3983         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3984
3985         if (write)
3986                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3987         return ret;
3988 }
3989
3990 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3991                                             void __user *oldval,
3992                                             size_t __user *oldlenp,
3993                                             void __user *newval, size_t newlen)
3994 {
3995         int *valp = table->data;
3996         int val = *valp;
3997         int new;
3998
3999         if (!newval || !newlen)
4000                 return 0;
4001         if (newlen != sizeof(int))
4002                 return -EINVAL;
4003         if (get_user(new, (int __user *)newval))
4004                 return -EFAULT;
4005         if (new == *valp)
4006                 return 0;
4007         if (oldval && oldlenp) {
4008                 size_t len;
4009                 if (get_user(len, oldlenp))
4010                         return -EFAULT;
4011                 if (len) {
4012                         if (len > table->maxlen)
4013                                 len = table->maxlen;
4014                         if (copy_to_user(oldval, valp, len))
4015                                 return -EFAULT;
4016                         if (put_user(len, oldlenp))
4017                                 return -EFAULT;
4018                 }
4019         }
4020
4021         *valp = new;
4022         return addrconf_fixup_forwarding(table, valp, val);
4023 }
4024
4025 static struct addrconf_sysctl_table
4026 {
4027         struct ctl_table_header *sysctl_header;
4028         ctl_table addrconf_vars[DEVCONF_MAX+1];
4029         char *dev_name;
4030 } addrconf_sysctl __read_mostly = {
4031         .sysctl_header = NULL,
4032         .addrconf_vars = {
4033                 {
4034                         .ctl_name       =       NET_IPV6_FORWARDING,
4035                         .procname       =       "forwarding",
4036                         .data           =       &ipv6_devconf.forwarding,
4037                         .maxlen         =       sizeof(int),
4038                         .mode           =       0644,
4039                         .proc_handler   =       addrconf_sysctl_forward,
4040                         .strategy       =       addrconf_sysctl_forward_strategy,
4041                 },
4042                 {
4043                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4044                         .procname       =       "hop_limit",
4045                         .data           =       &ipv6_devconf.hop_limit,
4046                         .maxlen         =       sizeof(int),
4047                         .mode           =       0644,
4048                         .proc_handler   =       proc_dointvec,
4049                 },
4050                 {
4051                         .ctl_name       =       NET_IPV6_MTU,
4052                         .procname       =       "mtu",
4053                         .data           =       &ipv6_devconf.mtu6,
4054                         .maxlen         =       sizeof(int),
4055                         .mode           =       0644,
4056                         .proc_handler   =       proc_dointvec,
4057                 },
4058                 {
4059                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4060                         .procname       =       "accept_ra",
4061                         .data           =       &ipv6_devconf.accept_ra,
4062                         .maxlen         =       sizeof(int),
4063                         .mode           =       0644,
4064                         .proc_handler   =       proc_dointvec,
4065                 },
4066                 {
4067                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4068                         .procname       =       "accept_redirects",
4069                         .data           =       &ipv6_devconf.accept_redirects,
4070                         .maxlen         =       sizeof(int),
4071                         .mode           =       0644,
4072                         .proc_handler   =       proc_dointvec,
4073                 },
4074                 {
4075                         .ctl_name       =       NET_IPV6_AUTOCONF,
4076                         .procname       =       "autoconf",
4077                         .data           =       &ipv6_devconf.autoconf,
4078                         .maxlen         =       sizeof(int),
4079                         .mode           =       0644,
4080                         .proc_handler   =       proc_dointvec,
4081                 },
4082                 {
4083                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4084                         .procname       =       "dad_transmits",
4085                         .data           =       &ipv6_devconf.dad_transmits,
4086                         .maxlen         =       sizeof(int),
4087                         .mode           =       0644,
4088                         .proc_handler   =       proc_dointvec,
4089                 },
4090                 {
4091                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4092                         .procname       =       "router_solicitations",
4093                         .data           =       &ipv6_devconf.rtr_solicits,
4094                         .maxlen         =       sizeof(int),
4095                         .mode           =       0644,
4096                         .proc_handler   =       proc_dointvec,
4097                 },
4098                 {
4099                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4100                         .procname       =       "router_solicitation_interval",
4101                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4102                         .maxlen         =       sizeof(int),
4103                         .mode           =       0644,
4104                         .proc_handler   =       proc_dointvec_jiffies,
4105                         .strategy       =       sysctl_jiffies,
4106                 },
4107                 {
4108                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4109                         .procname       =       "router_solicitation_delay",
4110                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4111                         .maxlen         =       sizeof(int),
4112                         .mode           =       0644,
4113                         .proc_handler   =       proc_dointvec_jiffies,
4114                         .strategy       =       sysctl_jiffies,
4115                 },
4116                 {
4117                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4118                         .procname       =       "force_mld_version",
4119                         .data           =       &ipv6_devconf.force_mld_version,
4120                         .maxlen         =       sizeof(int),
4121                         .mode           =       0644,
4122                         .proc_handler   =       proc_dointvec,
4123                 },
4124 #ifdef CONFIG_IPV6_PRIVACY
4125                 {
4126                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4127                         .procname       =       "use_tempaddr",
4128                         .data           =       &ipv6_devconf.use_tempaddr,
4129                         .maxlen         =       sizeof(int),
4130                         .mode           =       0644,
4131                         .proc_handler   =       proc_dointvec,
4132                 },
4133                 {
4134                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4135                         .procname       =       "temp_valid_lft",
4136                         .data           =       &ipv6_devconf.temp_valid_lft,
4137                         .maxlen         =       sizeof(int),
4138                         .mode           =       0644,
4139                         .proc_handler   =       proc_dointvec,
4140                 },
4141                 {
4142                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4143                         .procname       =       "temp_prefered_lft",
4144                         .data           =       &ipv6_devconf.temp_prefered_lft,
4145                         .maxlen         =       sizeof(int),
4146                         .mode           =       0644,
4147                         .proc_handler   =       proc_dointvec,
4148                 },
4149                 {
4150                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4151                         .procname       =       "regen_max_retry",
4152                         .data           =       &ipv6_devconf.regen_max_retry,
4153                         .maxlen         =       sizeof(int),
4154                         .mode           =       0644,
4155                         .proc_handler   =       proc_dointvec,
4156                 },
4157                 {
4158                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4159                         .procname       =       "max_desync_factor",
4160                         .data           =       &ipv6_devconf.max_desync_factor,
4161                         .maxlen         =       sizeof(int),
4162                         .mode           =       0644,
4163                         .proc_handler   =       proc_dointvec,
4164                 },
4165 #endif
4166                 {
4167                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4168                         .procname       =       "max_addresses",
4169                         .data           =       &ipv6_devconf.max_addresses,
4170                         .maxlen         =       sizeof(int),
4171                         .mode           =       0644,
4172                         .proc_handler   =       proc_dointvec,
4173                 },
4174                 {
4175                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4176                         .procname       =       "accept_ra_defrtr",
4177                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4178                         .maxlen         =       sizeof(int),
4179                         .mode           =       0644,
4180                         .proc_handler   =       proc_dointvec,
4181                 },
4182                 {
4183                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4184                         .procname       =       "accept_ra_pinfo",
4185                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4186                         .maxlen         =       sizeof(int),
4187                         .mode           =       0644,
4188                         .proc_handler   =       proc_dointvec,
4189                 },
4190 #ifdef CONFIG_IPV6_ROUTER_PREF
4191                 {
4192                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4193                         .procname       =       "accept_ra_rtr_pref",
4194                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4195                         .maxlen         =       sizeof(int),
4196                         .mode           =       0644,
4197                         .proc_handler   =       proc_dointvec,
4198                 },
4199                 {
4200                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4201                         .procname       =       "router_probe_interval",
4202                         .data           =       &ipv6_devconf.rtr_probe_interval,
4203                         .maxlen         =       sizeof(int),
4204                         .mode           =       0644,
4205                         .proc_handler   =       proc_dointvec_jiffies,
4206                         .strategy       =       sysctl_jiffies,
4207                 },
4208 #ifdef CONFIG_IPV6_ROUTE_INFO
4209                 {
4210                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4211                         .procname       =       "accept_ra_rt_info_max_plen",
4212                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4213                         .maxlen         =       sizeof(int),
4214                         .mode           =       0644,
4215                         .proc_handler   =       proc_dointvec,
4216                 },
4217 #endif
4218 #endif
4219                 {
4220                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4221                         .procname       =       "proxy_ndp",
4222                         .data           =       &ipv6_devconf.proxy_ndp,
4223                         .maxlen         =       sizeof(int),
4224                         .mode           =       0644,
4225                         .proc_handler   =       proc_dointvec,
4226                 },
4227                 {
4228                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4229                         .procname       =       "accept_source_route",
4230                         .data           =       &ipv6_devconf.accept_source_route,
4231                         .maxlen         =       sizeof(int),
4232                         .mode           =       0644,
4233                         .proc_handler   =       proc_dointvec,
4234                 },
4235 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4236                 {
4237                         .ctl_name       =       CTL_UNNUMBERED,
4238                         .procname       =       "optimistic_dad",
4239                         .data           =       &ipv6_devconf.optimistic_dad,
4240                         .maxlen         =       sizeof(int),
4241                         .mode           =       0644,
4242                         .proc_handler   =       proc_dointvec,
4243
4244                 },
4245 #endif
4246 #ifdef CONFIG_IPV6_MROUTE
4247                 {
4248                         .ctl_name       =       CTL_UNNUMBERED,
4249                         .procname       =       "mc_forwarding",
4250                         .data           =       &ipv6_devconf.mc_forwarding,
4251                         .maxlen         =       sizeof(int),
4252                         .mode           =       0444,
4253                         .proc_handler   =       proc_dointvec,
4254                 },
4255 #endif
4256                 {
4257                         .ctl_name       =       CTL_UNNUMBERED,
4258                         .procname       =       "disable_ipv6",
4259                         .data           =       &ipv6_devconf.disable_ipv6,
4260                         .maxlen         =       sizeof(int),
4261                         .mode           =       0644,
4262                         .proc_handler   =       proc_dointvec,
4263                 },
4264                 {
4265                         .ctl_name       =       CTL_UNNUMBERED,
4266                         .procname       =       "accept_dad",
4267                         .data           =       &ipv6_devconf.accept_dad,
4268                         .maxlen         =       sizeof(int),
4269                         .mode           =       0644,
4270                         .proc_handler   =       proc_dointvec,
4271                 },
4272                 {
4273                         .ctl_name       =       0,      /* sentinel */
4274                 }
4275         },
4276 };
4277
4278 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4279                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4280 {
4281         int i;
4282         struct addrconf_sysctl_table *t;
4283
4284 #define ADDRCONF_CTL_PATH_DEV   3
4285
4286         struct ctl_path addrconf_ctl_path[] = {
4287                 { .procname = "net", .ctl_name = CTL_NET, },
4288                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4289                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4290                 { /* to be set */ },
4291                 { },
4292         };
4293
4294
4295         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4296         if (t == NULL)
4297                 goto out;
4298
4299         for (i=0; t->addrconf_vars[i].data; i++) {
4300                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4301                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4302                 t->addrconf_vars[i].extra2 = net;
4303         }
4304
4305         /*
4306          * Make a copy of dev_name, because '.procname' is regarded as const
4307          * by sysctl and we wouldn't want anyone to change it under our feet
4308          * (see SIOCSIFNAME).
4309          */
4310         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4311         if (!t->dev_name)
4312                 goto free;
4313
4314         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4315         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4316
4317         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4318                         t->addrconf_vars);
4319         if (t->sysctl_header == NULL)
4320                 goto free_procname;
4321
4322         p->sysctl = t;
4323         return 0;
4324
4325 free_procname:
4326         kfree(t->dev_name);
4327 free:
4328         kfree(t);
4329 out:
4330         return -ENOBUFS;
4331 }
4332
4333 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4334 {
4335         struct addrconf_sysctl_table *t;
4336
4337         if (p->sysctl == NULL)
4338                 return;
4339
4340         t = p->sysctl;
4341         p->sysctl = NULL;
4342         unregister_sysctl_table(t->sysctl_header);
4343         kfree(t->dev_name);
4344         kfree(t);
4345 }
4346
4347 static void addrconf_sysctl_register(struct inet6_dev *idev)
4348 {
4349         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4350                               NET_IPV6_NEIGH, "ipv6",
4351                               &ndisc_ifinfo_sysctl_change,
4352                               ndisc_ifinfo_sysctl_strategy);
4353         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4354                         idev->dev->ifindex, idev, &idev->cnf);
4355 }
4356
4357 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4358 {
4359         __addrconf_sysctl_unregister(&idev->cnf);
4360         neigh_sysctl_unregister(idev->nd_parms);
4361 }
4362
4363
4364 #endif
4365
4366 static int addrconf_init_net(struct net *net)
4367 {
4368         int err;
4369         struct ipv6_devconf *all, *dflt;
4370
4371         err = -ENOMEM;
4372         all = &ipv6_devconf;
4373         dflt = &ipv6_devconf_dflt;
4374
4375         if (net != &init_net) {
4376                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4377                 if (all == NULL)
4378                         goto err_alloc_all;
4379
4380                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4381                 if (dflt == NULL)
4382                         goto err_alloc_dflt;
4383         }
4384
4385         net->ipv6.devconf_all = all;
4386         net->ipv6.devconf_dflt = dflt;
4387
4388 #ifdef CONFIG_SYSCTL
4389         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4390                         NULL, all);
4391         if (err < 0)
4392                 goto err_reg_all;
4393
4394         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4395                         NULL, dflt);
4396         if (err < 0)
4397                 goto err_reg_dflt;
4398 #endif
4399         return 0;
4400
4401 #ifdef CONFIG_SYSCTL
4402 err_reg_dflt:
4403         __addrconf_sysctl_unregister(all);
4404 err_reg_all:
4405         kfree(dflt);
4406 #endif
4407 err_alloc_dflt:
4408         kfree(all);
4409 err_alloc_all:
4410         return err;
4411 }
4412
4413 static void addrconf_exit_net(struct net *net)
4414 {
4415 #ifdef CONFIG_SYSCTL
4416         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4417         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4418 #endif
4419         if (net != &init_net) {
4420                 kfree(net->ipv6.devconf_dflt);
4421                 kfree(net->ipv6.devconf_all);
4422         }
4423 }
4424
4425 static struct pernet_operations addrconf_ops = {
4426         .init = addrconf_init_net,
4427         .exit = addrconf_exit_net,
4428 };
4429
4430 /*
4431  *      Device notifier
4432  */
4433
4434 int register_inet6addr_notifier(struct notifier_block *nb)
4435 {
4436         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4437 }
4438
4439 EXPORT_SYMBOL(register_inet6addr_notifier);
4440
4441 int unregister_inet6addr_notifier(struct notifier_block *nb)
4442 {
4443         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4444 }
4445
4446 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4447
4448 /*
4449  *      Init / cleanup code
4450  */
4451
4452 int __init addrconf_init(void)
4453 {
4454         int err;
4455
4456         if ((err = ipv6_addr_label_init()) < 0) {
4457                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4458                         err);
4459                 return err;
4460         }
4461
4462         register_pernet_subsys(&addrconf_ops);
4463
4464         /* The addrconf netdev notifier requires that loopback_dev
4465          * has it's ipv6 private information allocated and setup
4466          * before it can bring up and give link-local addresses
4467          * to other devices which are up.
4468          *
4469          * Unfortunately, loopback_dev is not necessarily the first
4470          * entry in the global dev_base list of net devices.  In fact,
4471          * it is likely to be the very last entry on that list.
4472          * So this causes the notifier registry below to try and
4473          * give link-local addresses to all devices besides loopback_dev
4474          * first, then loopback_dev, which cases all the non-loopback_dev
4475          * devices to fail to get a link-local address.
4476          *
4477          * So, as a temporary fix, allocate the ipv6 structure for
4478          * loopback_dev first by hand.
4479          * Longer term, all of the dependencies ipv6 has upon the loopback
4480          * device and it being up should be removed.
4481          */
4482         rtnl_lock();
4483         if (!ipv6_add_dev(init_net.loopback_dev))
4484                 err = -ENOMEM;
4485         rtnl_unlock();
4486         if (err)
4487                 goto errlo;
4488
4489         register_netdevice_notifier(&ipv6_dev_notf);
4490
4491         addrconf_verify(0);
4492
4493         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4494         if (err < 0)
4495                 goto errout;
4496
4497         /* Only the first call to __rtnl_register can fail */
4498         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4499         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4500         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4501         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4502         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4503
4504         ipv6_addr_label_rtnl_register();
4505
4506         return 0;
4507 errout:
4508         unregister_netdevice_notifier(&ipv6_dev_notf);
4509 errlo:
4510         unregister_pernet_subsys(&addrconf_ops);
4511
4512         return err;
4513 }
4514
4515 void addrconf_cleanup(void)
4516 {
4517         struct inet6_ifaddr *ifa;
4518         struct net_device *dev;
4519         int i;
4520
4521         unregister_netdevice_notifier(&ipv6_dev_notf);
4522         unregister_pernet_subsys(&addrconf_ops);
4523
4524         rtnl_lock();
4525
4526         /* clean dev list */
4527         for_each_netdev(&init_net, dev) {
4528                 if (__in6_dev_get(dev) == NULL)
4529                         continue;
4530                 addrconf_ifdown(dev, 1);
4531         }
4532         addrconf_ifdown(init_net.loopback_dev, 2);
4533
4534         /*
4535          *      Check hash table.
4536          */
4537         write_lock_bh(&addrconf_hash_lock);
4538         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4539                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4540                         struct inet6_ifaddr *bifa;
4541
4542                         bifa = ifa;
4543                         ifa = ifa->lst_next;
4544                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4545                         /* Do not free it; something is wrong.
4546                            Now we can investigate it with debugger.
4547                          */
4548                 }
4549         }
4550         write_unlock_bh(&addrconf_hash_lock);
4551
4552         del_timer(&addr_chk_timer);
4553         rtnl_unlock();
4554 }