SAFE public projects git trees. - safe/jmp/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_count_addresses(struct inet6_dev *idev);
 123
 124 /*
 125  *      Configured unicast address hash table
 126  */
 127 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 128 static DEFINE_RWLOCK(addrconf_hash_lock);
 129
 130 static void addrconf_verify(unsigned long);
 131
 132 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 133 static DEFINE_SPINLOCK(addrconf_verify_lock);
 134
 135 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 136 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 137
 138 static int addrconf_ifdown(struct net_device *dev, int how);
 139
 140 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 141 static void addrconf_dad_timer(unsigned long data);
 142 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 143 static void addrconf_dad_run(struct inet6_dev *idev);
 144 static void addrconf_rs_timer(unsigned long data);
 145 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 146 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 147
 148 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 149                                 struct prefix_info *pinfo);
 150 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 151                               struct net_device *dev);
 152
 153 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 154
 155 struct ipv6_devconf ipv6_devconf __read_mostly = {
 156         .forwarding             = 0,
 157         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 158         .mtu6                   = IPV6_MIN_MTU,
 159         .accept_ra              = 1,
 160         .accept_redirects       = 1,
 161         .autoconf               = 1,
 162         .force_mld_version      = 0,
 163         .dad_transmits          = 1,
 164         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 165         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 166         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 167 #ifdef CONFIG_IPV6_PRIVACY
 168         .use_tempaddr           = 0,
 169         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 170         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 171         .regen_max_retry        = REGEN_MAX_RETRY,
 172         .max_desync_factor      = MAX_DESYNC_FACTOR,
 173 #endif
 174         .max_addresses          = IPV6_MAX_ADDRESSES,
 175         .accept_ra_defrtr       = 1,
 176         .accept_ra_pinfo        = 1,
 177 #ifdef CONFIG_IPV6_ROUTER_PREF
 178         .accept_ra_rtr_pref     = 1,
 179         .rtr_probe_interval     = 60 * HZ,
 180 #ifdef CONFIG_IPV6_ROUTE_INFO
 181         .accept_ra_rt_info_max_plen = 0,
 182 #endif
 183 #endif
 184         .proxy_ndp              = 0,
 185         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 186         .disable_ipv6           = 0,
 187 };
 188
 189 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 190         .forwarding             = 0,
 191         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 192         .mtu6                   = IPV6_MIN_MTU,
 193         .accept_ra              = 1,
 194         .accept_redirects       = 1,
 195         .autoconf               = 1,
 196         .dad_transmits          = 1,
 197         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 198         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 199         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 200 #ifdef CONFIG_IPV6_PRIVACY
 201         .use_tempaddr           = 0,
 202         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 203         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 204         .regen_max_retry        = REGEN_MAX_RETRY,
 205         .max_desync_factor      = MAX_DESYNC_FACTOR,
 206 #endif
 207         .max_addresses          = IPV6_MAX_ADDRESSES,
 208         .accept_ra_defrtr       = 1,
 209         .accept_ra_pinfo        = 1,
 210 #ifdef CONFIG_IPV6_ROUTER_PREF
 211         .accept_ra_rtr_pref     = 1,
 212         .rtr_probe_interval     = 60 * HZ,
 213 #ifdef CONFIG_IPV6_ROUTE_INFO
 214         .accept_ra_rt_info_max_plen = 0,
 215 #endif
 216 #endif
 217         .proxy_ndp              = 0,
 218         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 219         .disable_ipv6           = 0,
 220 };
 221
 222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 225 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 226 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 227
 228 /* Check if a valid qdisc is available */
 229 static inline int addrconf_qdisc_ok(struct net_device *dev)
 230 {
 231         return (dev->qdisc != &noop_qdisc);
 232 }
 233
 234 /* Check if a route is valid prefix route */
 235 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 236 {
 237         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 238 }
 239
 240 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 241 {
 242         if (del_timer(&ifp->timer))
 243                 __in6_ifa_put(ifp);
 244 }
 245
 246 enum addrconf_timer_t
 247 {
 248         AC_NONE,
 249         AC_DAD,
 250         AC_RS,
 251 };
 252
 253 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 254                                enum addrconf_timer_t what,
 255                                unsigned long when)
 256 {
 257         if (!del_timer(&ifp->timer))
 258                 in6_ifa_hold(ifp);
 259
 260         switch (what) {
 261         case AC_DAD:
 262                 ifp->timer.function = addrconf_dad_timer;
 263                 break;
 264         case AC_RS:
 265                 ifp->timer.function = addrconf_rs_timer;
 266                 break;
 267         default:;
 268         }
 269         ifp->timer.expires = jiffies + when;
 270         add_timer(&ifp->timer);
 271 }
 272
 273 static int snmp6_alloc_dev(struct inet6_dev *idev)
 274 {
 275         if (snmp_mib_init((void **)idev->stats.ipv6,
 276                           sizeof(struct ipstats_mib)) < 0)
 277                 goto err_ip;
 278         if (snmp_mib_init((void **)idev->stats.icmpv6,
 279                           sizeof(struct icmpv6_mib)) < 0)
 280                 goto err_icmp;
 281         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 282                           sizeof(struct icmpv6msg_mib)) < 0)
 283                 goto err_icmpmsg;
 284
 285         return 0;
 286
 287 err_icmpmsg:
 288         snmp_mib_free((void **)idev->stats.icmpv6);
 289 err_icmp:
 290         snmp_mib_free((void **)idev->stats.ipv6);
 291 err_ip:
 292         return -ENOMEM;
 293 }
 294
 295 static void snmp6_free_dev(struct inet6_dev *idev)
 296 {
 297         snmp_mib_free((void **)idev->stats.icmpv6msg);
 298         snmp_mib_free((void **)idev->stats.icmpv6);
 299         snmp_mib_free((void **)idev->stats.ipv6);
 300 }
 301
 302 /* Nobody refers to this device, we may destroy it. */
 303
 304 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 305 {
 306         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 307         kfree(idev);
 308 }
 309
 310 void in6_dev_finish_destroy(struct inet6_dev *idev)
 311 {
 312         struct net_device *dev = idev->dev;
 313         BUG_TRAP(idev->addr_list==NULL);
 314         BUG_TRAP(idev->mc_list==NULL);
 315 #ifdef NET_REFCNT_DEBUG
 316         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 317 #endif
 318         dev_put(dev);
 319         if (!idev->dead) {
 320                 printk("Freeing alive inet6 device %p\n", idev);
 321                 return;
 322         }
 323         snmp6_free_dev(idev);
 324         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 325 }
 326
 327 EXPORT_SYMBOL(in6_dev_finish_destroy);
 328
 329 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 330 {
 331         struct inet6_dev *ndev;
 332
 333         ASSERT_RTNL();
 334
 335         if (dev->mtu < IPV6_MIN_MTU)
 336                 return NULL;
 337
 338         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 339
 340         if (ndev == NULL)
 341                 return NULL;
 342
 343         rwlock_init(&ndev->lock);
 344         ndev->dev = dev;
 345         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 346         ndev->cnf.mtu6 = dev->mtu;
 347         ndev->cnf.sysctl = NULL;
 348         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 349         if (ndev->nd_parms == NULL) {
 350                 kfree(ndev);
 351                 return NULL;
 352         }
 353         if (ndev->cnf.forwarding)
 354                 dev_disable_lro(dev);
 355         /* We refer to the device */
 356         dev_hold(dev);
 357
 358         if (snmp6_alloc_dev(ndev) < 0) {
 359                 ADBG((KERN_WARNING
 360                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 361                         __func__, dev->name));
 362                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 363                 ndev->dead = 1;
 364                 in6_dev_finish_destroy(ndev);
 365                 return NULL;
 366         }
 367
 368         if (snmp6_register_dev(ndev) < 0) {
 369                 ADBG((KERN_WARNING
 370                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 371                         __func__, dev->name));
 372                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 373                 ndev->dead = 1;
 374                 in6_dev_finish_destroy(ndev);
 375                 return NULL;
 376         }
 377
 378         /* One reference from device.  We must do this before
 379          * we invoke __ipv6_regen_rndid().
 380          */
 381         in6_dev_hold(ndev);
 382
 383 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 384         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 385                 printk(KERN_INFO
 386                        "%s: Disabled Multicast RS\n",
 387                        dev->name);
 388                 ndev->cnf.rtr_solicits = 0;
 389         }
 390 #endif
 391
 392 #ifdef CONFIG_IPV6_PRIVACY
 393         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 394         if ((dev->flags&IFF_LOOPBACK) ||
 395             dev->type == ARPHRD_TUNNEL ||
 396             dev->type == ARPHRD_TUNNEL6 ||
 397             dev->type == ARPHRD_SIT ||
 398             dev->type == ARPHRD_NONE) {
 399                 printk(KERN_INFO
 400                        "%s: Disabled Privacy Extensions\n",
 401                        dev->name);
 402                 ndev->cnf.use_tempaddr = -1;
 403         } else {
 404                 in6_dev_hold(ndev);
 405                 ipv6_regen_rndid((unsigned long) ndev);
 406         }
 407 #endif
 408
 409         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 410                 ndev->if_flags |= IF_READY;
 411
 412         ipv6_mc_init_dev(ndev);
 413         ndev->tstamp = jiffies;
 414         addrconf_sysctl_register(ndev);
 415         /* protected by rtnl_lock */
 416         rcu_assign_pointer(dev->ip6_ptr, ndev);
 417
 418         /* Join all-node multicast group */
 419         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 420
 421         return ndev;
 422 }
 423
 424 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 425 {
 426         struct inet6_dev *idev;
 427
 428         ASSERT_RTNL();
 429
 430         if ((idev = __in6_dev_get(dev)) == NULL) {
 431                 if ((idev = ipv6_add_dev(dev)) == NULL)
 432                         return NULL;
 433         }
 434
 435         if (dev->flags&IFF_UP)
 436                 ipv6_mc_up(idev);
 437         return idev;
 438 }
 439
 440 #ifdef CONFIG_SYSCTL
 441 static void dev_forward_change(struct inet6_dev *idev)
 442 {
 443         struct net_device *dev;
 444         struct inet6_ifaddr *ifa;
 445
 446         if (!idev)
 447                 return;
 448         dev = idev->dev;
 449         if (idev->cnf.forwarding)
 450                 dev_disable_lro(dev);
 451         if (dev && (dev->flags & IFF_MULTICAST)) {
 452                 if (idev->cnf.forwarding)
 453                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 454                 else
 455                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 456         }
 457         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 458                 if (ifa->flags&IFA_F_TENTATIVE)
 459                         continue;
 460                 if (idev->cnf.forwarding)
 461                         addrconf_join_anycast(ifa);
 462                 else
 463                         addrconf_leave_anycast(ifa);
 464         }
 465 }
 466
 467
 468 static void addrconf_forward_change(struct net *net, __s32 newf)
 469 {
 470         struct net_device *dev;
 471         struct inet6_dev *idev;
 472
 473         read_lock(&dev_base_lock);
 474         for_each_netdev(net, dev) {
 475                 rcu_read_lock();
 476                 idev = __in6_dev_get(dev);
 477                 if (idev) {
 478                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 479                         idev->cnf.forwarding = newf;
 480                         if (changed)
 481                                 dev_forward_change(idev);
 482                 }
 483                 rcu_read_unlock();
 484         }
 485         read_unlock(&dev_base_lock);
 486 }
 487
 488 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 489 {
 490         struct net *net;
 491
 492         net = (struct net *)table->extra2;
 493         if (p == &net->ipv6.devconf_dflt->forwarding)
 494                 return;
 495
 496         rtnl_lock();
 497         if (p == &net->ipv6.devconf_all->forwarding) {
 498                 __s32 newf = net->ipv6.devconf_all->forwarding;
 499                 net->ipv6.devconf_dflt->forwarding = newf;
 500                 addrconf_forward_change(net, newf);
 501         } else if ((!*p) ^ (!old))
 502                 dev_forward_change((struct inet6_dev *)table->extra1);
 503         rtnl_unlock();
 504
 505         if (*p)
 506                 rt6_purge_dflt_routers(net);
 507 }
 508 #endif
 509
 510 /* Nobody refers to this ifaddr, destroy it */
 511
 512 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 513 {
 514         BUG_TRAP(ifp->if_next==NULL);
 515         BUG_TRAP(ifp->lst_next==NULL);
 516 #ifdef NET_REFCNT_DEBUG
 517         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 518 #endif
 519
 520         in6_dev_put(ifp->idev);
 521
 522         if (del_timer(&ifp->timer))
 523                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 524
 525         if (!ifp->dead) {
 526                 printk("Freeing alive inet6 address %p\n", ifp);
 527                 return;
 528         }
 529         dst_release(&ifp->rt->u.dst);
 530
 531         kfree(ifp);
 532 }
 533
 534 static void
 535 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 536 {
 537         struct inet6_ifaddr *ifa, **ifap;
 538         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 539
 540         /*
 541          * Each device address list is sorted in order of scope -
 542          * global before linklocal.
 543          */
 544         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 545              ifap = &ifa->if_next) {
 546                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 547                         break;
 548         }
 549
 550         ifp->if_next = *ifap;
 551         *ifap = ifp;
 552 }
 553
 554 /*
 555  *      Hash function taken from net_alias.c
 556  */
 557 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 558 {
 559         __u32 word;
 560
 561         /*
 562          * We perform the hash function over the last 64 bits of the address
 563          * This will include the IEEE address token on links that support it.
 564          */
 565
 566         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 567         word ^= (word >> 16);
 568         word ^= (word >> 8);
 569
 570         return ((word ^ (word >> 4)) & 0x0f);
 571 }
 572
 573 /* On success it returns ifp with increased reference count */
 574
 575 static struct inet6_ifaddr *
 576 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 577               int scope, u32 flags)
 578 {
 579         struct inet6_ifaddr *ifa = NULL;
 580         struct rt6_info *rt;
 581         int hash;
 582         int err = 0;
 583         int addr_type = ipv6_addr_type(addr);
 584
 585         if (addr_type == IPV6_ADDR_ANY ||
 586             addr_type & IPV6_ADDR_MULTICAST ||
 587             (!(idev->dev->flags & IFF_LOOPBACK) &&
 588              addr_type & IPV6_ADDR_LOOPBACK))
 589                 return ERR_PTR(-EADDRNOTAVAIL);
 590
 591         rcu_read_lock_bh();
 592         if (idev->dead) {
 593                 err = -ENODEV;                  /*XXX*/
 594                 goto out2;
 595         }
 596
 597         write_lock(&addrconf_hash_lock);
 598
 599         /* Ignore adding duplicate addresses on an interface */
 600         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 601                 ADBG(("ipv6_add_addr: already assigned\n"));
 602                 err = -EEXIST;
 603                 goto out;
 604         }
 605
 606         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 607
 608         if (ifa == NULL) {
 609                 ADBG(("ipv6_add_addr: malloc failed\n"));
 610                 err = -ENOBUFS;
 611                 goto out;
 612         }
 613
 614         rt = addrconf_dst_alloc(idev, addr, 0);
 615         if (IS_ERR(rt)) {
 616                 err = PTR_ERR(rt);
 617                 goto out;
 618         }
 619
 620         ipv6_addr_copy(&ifa->addr, addr);
 621
 622         spin_lock_init(&ifa->lock);
 623         init_timer(&ifa->timer);
 624         ifa->timer.data = (unsigned long) ifa;
 625         ifa->scope = scope;
 626         ifa->prefix_len = pfxlen;
 627         ifa->flags = flags | IFA_F_TENTATIVE;
 628         ifa->cstamp = ifa->tstamp = jiffies;
 629
 630         ifa->rt = rt;
 631
 632         /*
 633          * part one of RFC 4429, section 3.3
 634          * We should not configure an address as
 635          * optimistic if we do not yet know the link
 636          * layer address of our nexhop router
 637          */
 638
 639         if (rt->rt6i_nexthop == NULL)
 640                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 641
 642         ifa->idev = idev;
 643         in6_dev_hold(idev);
 644         /* For caller */
 645         in6_ifa_hold(ifa);
 646
 647         /* Add to big hash table */
 648         hash = ipv6_addr_hash(addr);
 649
 650         ifa->lst_next = inet6_addr_lst[hash];
 651         inet6_addr_lst[hash] = ifa;
 652         in6_ifa_hold(ifa);
 653         write_unlock(&addrconf_hash_lock);
 654
 655         write_lock(&idev->lock);
 656         /* Add to inet6_dev unicast addr list. */
 657         ipv6_link_dev_addr(idev, ifa);
 658
 659 #ifdef CONFIG_IPV6_PRIVACY
 660         if (ifa->flags&IFA_F_TEMPORARY) {
 661                 ifa->tmp_next = idev->tempaddr_list;
 662                 idev->tempaddr_list = ifa;
 663                 in6_ifa_hold(ifa);
 664         }
 665 #endif
 666
 667         in6_ifa_hold(ifa);
 668         write_unlock(&idev->lock);
 669 out2:
 670         rcu_read_unlock_bh();
 671
 672         if (likely(err == 0))
 673                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 674         else {
 675                 kfree(ifa);
 676                 ifa = ERR_PTR(err);
 677         }
 678
 679         return ifa;
 680 out:
 681         write_unlock(&addrconf_hash_lock);
 682         goto out2;
 683 }
 684
 685 /* This function wants to get referenced ifp and releases it before return */
 686
 687 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 688 {
 689         struct inet6_ifaddr *ifa, **ifap;
 690         struct inet6_dev *idev = ifp->idev;
 691         int hash;
 692         int deleted = 0, onlink = 0;
 693         unsigned long expires = jiffies;
 694
 695         hash = ipv6_addr_hash(&ifp->addr);
 696
 697         ifp->dead = 1;
 698
 699         write_lock_bh(&addrconf_hash_lock);
 700         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 701              ifap = &ifa->lst_next) {
 702                 if (ifa == ifp) {
 703                         *ifap = ifa->lst_next;
 704                         __in6_ifa_put(ifp);
 705                         ifa->lst_next = NULL;
 706                         break;
 707                 }
 708         }
 709         write_unlock_bh(&addrconf_hash_lock);
 710
 711         write_lock_bh(&idev->lock);
 712 #ifdef CONFIG_IPV6_PRIVACY
 713         if (ifp->flags&IFA_F_TEMPORARY) {
 714                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 715                      ifap = &ifa->tmp_next) {
 716                         if (ifa == ifp) {
 717                                 *ifap = ifa->tmp_next;
 718                                 if (ifp->ifpub) {
 719                                         in6_ifa_put(ifp->ifpub);
 720                                         ifp->ifpub = NULL;
 721                                 }
 722                                 __in6_ifa_put(ifp);
 723                                 ifa->tmp_next = NULL;
 724                                 break;
 725                         }
 726                 }
 727         }
 728 #endif
 729
 730         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 731                 if (ifa == ifp) {
 732                         *ifap = ifa->if_next;
 733                         __in6_ifa_put(ifp);
 734                         ifa->if_next = NULL;
 735                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 736                                 break;
 737                         deleted = 1;
 738                         continue;
 739                 } else if (ifp->flags & IFA_F_PERMANENT) {
 740                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 741                                               ifp->prefix_len)) {
 742                                 if (ifa->flags & IFA_F_PERMANENT) {
 743                                         onlink = 1;
 744                                         if (deleted)
 745                                                 break;
 746                                 } else {
 747                                         unsigned long lifetime;
 748
 749                                         if (!onlink)
 750                                                 onlink = -1;
 751
 752                                         spin_lock(&ifa->lock);
 753
 754                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 755                                         /*
 756                                          * Note: Because this address is
 757                                          * not permanent, lifetime <
 758                                          * LONG_MAX / HZ here.
 759                                          */
 760                                         if (time_before(expires,
 761                                                         ifa->tstamp + lifetime * HZ))
 762                                                 expires = ifa->tstamp + lifetime * HZ;
 763                                         spin_unlock(&ifa->lock);
 764                                 }
 765                         }
 766                 }
 767                 ifap = &ifa->if_next;
 768         }
 769         write_unlock_bh(&idev->lock);
 770
 771         ipv6_ifa_notify(RTM_DELADDR, ifp);
 772
 773         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 774
 775         addrconf_del_timer(ifp);
 776
 777         /*
 778          * Purge or update corresponding prefix
 779          *
 780          * 1) we don't purge prefix here if address was not permanent.
 781          *    prefix is managed by its own lifetime.
 782          * 2) if there're no addresses, delete prefix.
 783          * 3) if there're still other permanent address(es),
 784          *    corresponding prefix is still permanent.
 785          * 4) otherwise, update prefix lifetime to the
 786          *    longest valid lifetime among the corresponding
 787          *    addresses on the device.
 788          *    Note: subsequent RA will update lifetime.
 789          *
 790          * --yoshfuji
 791          */
 792         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 793                 struct in6_addr prefix;
 794                 struct rt6_info *rt;
 795                 struct net *net = dev_net(ifp->idev->dev);
 796                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 797                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 798
 799                 if (rt && addrconf_is_prefix_route(rt)) {
 800                         if (onlink == 0) {
 801                                 ip6_del_rt(rt);
 802                                 rt = NULL;
 803                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 804                                 rt->rt6i_expires = expires;
 805                                 rt->rt6i_flags |= RTF_EXPIRES;
 806                         }
 807                 }
 808                 dst_release(&rt->u.dst);
 809         }
 810
 811         in6_ifa_put(ifp);
 812 }
 813
 814 #ifdef CONFIG_IPV6_PRIVACY
 815 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 816 {
 817         struct inet6_dev *idev = ifp->idev;
 818         struct in6_addr addr, *tmpaddr;
 819         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 820         unsigned long regen_advance;
 821         int tmp_plen;
 822         int ret = 0;
 823         int max_addresses;
 824         u32 addr_flags;
 825
 826         write_lock(&idev->lock);
 827         if (ift) {
 828                 spin_lock_bh(&ift->lock);
 829                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 830                 spin_unlock_bh(&ift->lock);
 831                 tmpaddr = &addr;
 832         } else {
 833                 tmpaddr = NULL;
 834         }
 835 retry:
 836         in6_dev_hold(idev);
 837         if (idev->cnf.use_tempaddr <= 0) {
 838                 write_unlock(&idev->lock);
 839                 printk(KERN_INFO
 840                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 841                 in6_dev_put(idev);
 842                 ret = -1;
 843                 goto out;
 844         }
 845         spin_lock_bh(&ifp->lock);
 846         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 847                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 848                 spin_unlock_bh(&ifp->lock);
 849                 write_unlock(&idev->lock);
 850                 printk(KERN_WARNING
 851                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 852                 in6_dev_put(idev);
 853                 ret = -1;
 854                 goto out;
 855         }
 856         in6_ifa_hold(ifp);
 857         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 858         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 859                 spin_unlock_bh(&ifp->lock);
 860                 write_unlock(&idev->lock);
 861                 printk(KERN_WARNING
 862                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 863                 in6_ifa_put(ifp);
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 869         tmp_valid_lft = min_t(__u32,
 870                               ifp->valid_lft,
 871                               idev->cnf.temp_valid_lft);
 872         tmp_prefered_lft = min_t(__u32,
 873                                  ifp->prefered_lft,
 874                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 875         tmp_plen = ifp->prefix_len;
 876         max_addresses = idev->cnf.max_addresses;
 877         tmp_cstamp = ifp->cstamp;
 878         tmp_tstamp = ifp->tstamp;
 879         spin_unlock_bh(&ifp->lock);
 880
 881         regen_advance = idev->cnf.regen_max_retry *
 882                         idev->cnf.dad_transmits *
 883                         idev->nd_parms->retrans_time / HZ;
 884         write_unlock(&idev->lock);
 885
 886         /* A temporary address is created only if this calculated Preferred
 887          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 888          * an implementation must not create a temporary address with a zero
 889          * Preferred Lifetime.
 890          */
 891         if (tmp_prefered_lft <= regen_advance) {
 892                 in6_ifa_put(ifp);
 893                 in6_dev_put(idev);
 894                 ret = -1;
 895                 goto out;
 896         }
 897
 898         addr_flags = IFA_F_TEMPORARY;
 899         /* set in addrconf_prefix_rcv() */
 900         if (ifp->flags & IFA_F_OPTIMISTIC)
 901                 addr_flags |= IFA_F_OPTIMISTIC;
 902
 903         ift = !max_addresses ||
 904               ipv6_count_addresses(idev) < max_addresses ?
 905                 ipv6_add_addr(idev, &addr, tmp_plen,
 906                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 907                               addr_flags) : NULL;
 908         if (!ift || IS_ERR(ift)) {
 909                 in6_ifa_put(ifp);
 910                 in6_dev_put(idev);
 911                 printk(KERN_INFO
 912                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 913                 tmpaddr = &addr;
 914                 write_lock(&idev->lock);
 915                 goto retry;
 916         }
 917
 918         spin_lock_bh(&ift->lock);
 919         ift->ifpub = ifp;
 920         ift->valid_lft = tmp_valid_lft;
 921         ift->prefered_lft = tmp_prefered_lft;
 922         ift->cstamp = tmp_cstamp;
 923         ift->tstamp = tmp_tstamp;
 924         spin_unlock_bh(&ift->lock);
 925
 926         addrconf_dad_start(ift, 0);
 927         in6_ifa_put(ift);
 928         in6_dev_put(idev);
 929 out:
 930         return ret;
 931 }
 932 #endif
 933
 934 /*
 935  *      Choose an appropriate source address (RFC3484)
 936  */
 937 enum {
 938         IPV6_SADDR_RULE_INIT = 0,
 939         IPV6_SADDR_RULE_LOCAL,
 940         IPV6_SADDR_RULE_SCOPE,
 941         IPV6_SADDR_RULE_PREFERRED,
 942 #ifdef CONFIG_IPV6_MIP6
 943         IPV6_SADDR_RULE_HOA,
 944 #endif
 945         IPV6_SADDR_RULE_OIF,
 946         IPV6_SADDR_RULE_LABEL,
 947 #ifdef CONFIG_IPV6_PRIVACY
 948         IPV6_SADDR_RULE_PRIVACY,
 949 #endif
 950         IPV6_SADDR_RULE_ORCHID,
 951         IPV6_SADDR_RULE_PREFIX,
 952         IPV6_SADDR_RULE_MAX
 953 };
 954
 955 struct ipv6_saddr_score {
 956         int                     rule;
 957         int                     addr_type;
 958         struct inet6_ifaddr     *ifa;
 959         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 960         int                     scopedist;
 961         int                     matchlen;
 962 };
 963
 964 struct ipv6_saddr_dst {
 965         const struct in6_addr *addr;
 966         int ifindex;
 967         int scope;
 968         int label;
 969         unsigned int prefs;
 970 };
 971
 972 static inline int ipv6_saddr_preferred(int type)
 973 {
 974         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 975                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 976                 return 1;
 977         return 0;
 978 }
 979
 980 static int ipv6_get_saddr_eval(struct net *net,
 981                                struct ipv6_saddr_score *score,
 982                                struct ipv6_saddr_dst *dst,
 983                                int i)
 984 {
 985         int ret;
 986
 987         if (i <= score->rule) {
 988                 switch (i) {
 989                 case IPV6_SADDR_RULE_SCOPE:
 990                         ret = score->scopedist;
 991                         break;
 992                 case IPV6_SADDR_RULE_PREFIX:
 993                         ret = score->matchlen;
 994                         break;
 995                 default:
 996                         ret = !!test_bit(i, score->scorebits);
 997                 }
 998                 goto out;
 999         }
1000
1001         switch (i) {
1002         case IPV6_SADDR_RULE_INIT:
1003                 /* Rule 0: remember if hiscore is not ready yet */
1004                 ret = !!score->ifa;
1005                 break;
1006         case IPV6_SADDR_RULE_LOCAL:
1007                 /* Rule 1: Prefer same address */
1008                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1009                 break;
1010         case IPV6_SADDR_RULE_SCOPE:
1011                 /* Rule 2: Prefer appropriate scope
1012                  *
1013                  *      ret
1014                  *       ^
1015                  *    -1 |  d 15
1016                  *    ---+--+-+---> scope
1017                  *       |
1018                  *       |             d is scope of the destination.
1019                  *  B-d  |  \
1020                  *       |   \      <- smaller scope is better if
1021                  *  B-15 |    \        if scope is enough for destinaion.
1022                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1023                  * d-C-1 | /
1024                  *       |/         <- greater is better
1025                  *   -C  /             if scope is not enough for destination.
1026                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1027                  *
1028                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1029                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1030                  * Assume B = 0 and we get C > 29.
1031                  */
1032                 ret = __ipv6_addr_src_scope(score->addr_type);
1033                 if (ret >= dst->scope)
1034                         ret = -ret;
1035                 else
1036                         ret -= 128;     /* 30 is enough */
1037                 score->scopedist = ret;
1038                 break;
1039         case IPV6_SADDR_RULE_PREFERRED:
1040                 /* Rule 3: Avoid deprecated and optimistic addresses */
1041                 ret = ipv6_saddr_preferred(score->addr_type) ||
1042                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1043                 break;
1044 #ifdef CONFIG_IPV6_MIP6
1045         case IPV6_SADDR_RULE_HOA:
1046             {
1047                 /* Rule 4: Prefer home address */
1048                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1049                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1050                 break;
1051             }
1052 #endif
1053         case IPV6_SADDR_RULE_OIF:
1054                 /* Rule 5: Prefer outgoing interface */
1055                 ret = (!dst->ifindex ||
1056                        dst->ifindex == score->ifa->idev->dev->ifindex);
1057                 break;
1058         case IPV6_SADDR_RULE_LABEL:
1059                 /* Rule 6: Prefer matching label */
1060                 ret = ipv6_addr_label(net,
1061                                       &score->ifa->addr, score->addr_type,
1062                                       score->ifa->idev->dev->ifindex) == dst->label;
1063                 break;
1064 #ifdef CONFIG_IPV6_PRIVACY
1065         case IPV6_SADDR_RULE_PRIVACY:
1066             {
1067                 /* Rule 7: Prefer public address
1068                  * Note: prefer temprary address if use_tempaddr >= 2
1069                  */
1070                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1071                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1072                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1073                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1074                 break;
1075             }
1076 #endif
1077         case IPV6_SADDR_RULE_ORCHID:
1078                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1079                  *          non-ORCHID vs non-ORCHID
1080                  */
1081                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1082                         ipv6_addr_orchid(dst->addr));
1083                 break;
1084         case IPV6_SADDR_RULE_PREFIX:
1085                 /* Rule 8: Use longest matching prefix */
1086                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1087                                                        dst->addr);
1088                 break;
1089         default:
1090                 ret = 0;
1091         }
1092
1093         if (ret)
1094                 __set_bit(i, score->scorebits);
1095         score->rule = i;
1096 out:
1097         return ret;
1098 }
1099
1100 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1101                        const struct in6_addr *daddr, unsigned int prefs,
1102                        struct in6_addr *saddr)
1103 {
1104         struct ipv6_saddr_score scores[2],
1105                                 *score = &scores[0], *hiscore = &scores[1];
1106         struct net *net = dev_net(dst_dev);
1107         struct ipv6_saddr_dst dst;
1108         struct net_device *dev;
1109         int dst_type;
1110
1111         dst_type = __ipv6_addr_type(daddr);
1112         dst.addr = daddr;
1113         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1114         dst.scope = __ipv6_addr_src_scope(dst_type);
1115         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1116         dst.prefs = prefs;
1117
1118         hiscore->rule = -1;
1119         hiscore->ifa = NULL;
1120
1121         read_lock(&dev_base_lock);
1122         rcu_read_lock();
1123
1124         for_each_netdev(net, dev) {
1125                 struct inet6_dev *idev;
1126
1127                 /* Candidate Source Address (section 4)
1128                  *  - multicast and link-local destination address,
1129                  *    the set of candidate source address MUST only
1130                  *    include addresses assigned to interfaces
1131                  *    belonging to the same link as the outgoing
1132                  *    interface.
1133                  * (- For site-local destination addresses, the
1134                  *    set of candidate source addresses MUST only
1135                  *    include addresses assigned to interfaces
1136                  *    belonging to the same site as the outgoing
1137                  *    interface.)
1138                  */
1139                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1140                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1141                     dst.ifindex && dev->ifindex != dst.ifindex)
1142                         continue;
1143
1144                 idev = __in6_dev_get(dev);
1145                 if (!idev)
1146                         continue;
1147
1148                 read_lock_bh(&idev->lock);
1149                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1150                         int i;
1151
1152                         /*
1153                          * - Tentative Address (RFC2462 section 5.4)
1154                          *  - A tentative address is not considered
1155                          *    "assigned to an interface" in the traditional
1156                          *    sense, unless it is also flagged as optimistic.
1157                          * - Candidate Source Address (section 4)
1158                          *  - In any case, anycast addresses, multicast
1159                          *    addresses, and the unspecified address MUST
1160                          *    NOT be included in a candidate set.
1161                          */
1162                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1163                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1164                                 continue;
1165
1166                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1167
1168                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1169                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1170                                 LIMIT_NETDEBUG(KERN_DEBUG
1171                                                "ADDRCONF: unspecified / multicast address "
1172                                                "assigned as unicast address on %s",
1173                                                dev->name);
1174                                 continue;
1175                         }
1176
1177                         score->rule = -1;
1178                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1179
1180                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1181                                 int minihiscore, miniscore;
1182
1183                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1184                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1185
1186                                 if (minihiscore > miniscore) {
1187                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1188                                             score->scopedist > 0) {
1189                                                 /*
1190                                                  * special case:
1191                                                  * each remaining entry
1192                                                  * has too small (not enough)
1193                                                  * scope, because ifa entries
1194                                                  * are sorted by their scope
1195                                                  * values.
1196                                                  */
1197                                                 goto try_nextdev;
1198                                         }
1199                                         break;
1200                                 } else if (minihiscore < miniscore) {
1201                                         struct ipv6_saddr_score *tmp;
1202
1203                                         if (hiscore->ifa)
1204                                                 in6_ifa_put(hiscore->ifa);
1205
1206                                         in6_ifa_hold(score->ifa);
1207
1208                                         tmp = hiscore;
1209                                         hiscore = score;
1210                                         score = tmp;
1211
1212                                         /* restore our iterator */
1213                                         score->ifa = hiscore->ifa;
1214
1215                                         break;
1216                                 }
1217                         }
1218                 }
1219 try_nextdev:
1220                 read_unlock_bh(&idev->lock);
1221         }
1222         rcu_read_unlock();
1223         read_unlock(&dev_base_lock);
1224
1225         if (!hiscore->ifa)
1226                 return -EADDRNOTAVAIL;
1227
1228         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1229         in6_ifa_put(hiscore->ifa);
1230         return 0;
1231 }
1232
1233 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1234
1235 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1236                     unsigned char banned_flags)
1237 {
1238         struct inet6_dev *idev;
1239         int err = -EADDRNOTAVAIL;
1240
1241         rcu_read_lock();
1242         if ((idev = __in6_dev_get(dev)) != NULL) {
1243                 struct inet6_ifaddr *ifp;
1244
1245                 read_lock_bh(&idev->lock);
1246                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1247                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1248                                 ipv6_addr_copy(addr, &ifp->addr);
1249                                 err = 0;
1250                                 break;
1251                         }
1252                 }
1253                 read_unlock_bh(&idev->lock);
1254         }
1255         rcu_read_unlock();
1256         return err;
1257 }
1258
1259 static int ipv6_count_addresses(struct inet6_dev *idev)
1260 {
1261         int cnt = 0;
1262         struct inet6_ifaddr *ifp;
1263
1264         read_lock_bh(&idev->lock);
1265         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1266                 cnt++;
1267         read_unlock_bh(&idev->lock);
1268         return cnt;
1269 }
1270
1271 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1272                   struct net_device *dev, int strict)
1273 {
1274         struct inet6_ifaddr * ifp;
1275         u8 hash = ipv6_addr_hash(addr);
1276
1277         read_lock_bh(&addrconf_hash_lock);
1278         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1279                 if (!net_eq(dev_net(ifp->idev->dev), net))
1280                         continue;
1281                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1282                     !(ifp->flags&IFA_F_TENTATIVE)) {
1283                         if (dev == NULL || ifp->idev->dev == dev ||
1284                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1285                                 break;
1286                 }
1287         }
1288         read_unlock_bh(&addrconf_hash_lock);
1289         return ifp != NULL;
1290 }
1291 EXPORT_SYMBOL(ipv6_chk_addr);
1292
1293 static
1294 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1295                        struct net_device *dev)
1296 {
1297         struct inet6_ifaddr * ifp;
1298         u8 hash = ipv6_addr_hash(addr);
1299
1300         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1301                 if (!net_eq(dev_net(ifp->idev->dev), net))
1302                         continue;
1303                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1304                         if (dev == NULL || ifp->idev->dev == dev)
1305                                 break;
1306                 }
1307         }
1308         return ifp != NULL;
1309 }
1310
1311 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1312 {
1313         struct inet6_dev *idev;
1314         struct inet6_ifaddr *ifa;
1315         int     onlink;
1316
1317         onlink = 0;
1318         rcu_read_lock();
1319         idev = __in6_dev_get(dev);
1320         if (idev) {
1321                 read_lock_bh(&idev->lock);
1322                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1323                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1324                                                    ifa->prefix_len);
1325                         if (onlink)
1326                                 break;
1327                 }
1328                 read_unlock_bh(&idev->lock);
1329         }
1330         rcu_read_unlock();
1331         return onlink;
1332 }
1333
1334 EXPORT_SYMBOL(ipv6_chk_prefix);
1335
1336 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1337                                      struct net_device *dev, int strict)
1338 {
1339         struct inet6_ifaddr * ifp;
1340         u8 hash = ipv6_addr_hash(addr);
1341
1342         read_lock_bh(&addrconf_hash_lock);
1343         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1344                 if (!net_eq(dev_net(ifp->idev->dev), net))
1345                         continue;
1346                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1347                         if (dev == NULL || ifp->idev->dev == dev ||
1348                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1349                                 in6_ifa_hold(ifp);
1350                                 break;
1351                         }
1352                 }
1353         }
1354         read_unlock_bh(&addrconf_hash_lock);
1355
1356         return ifp;
1357 }
1358
1359 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1360 {
1361         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1362         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1363         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1364         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1365         int sk_ipv6only = ipv6_only_sock(sk);
1366         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1367         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1368         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1369
1370         if (!sk2_rcv_saddr && !sk_ipv6only)
1371                 return 1;
1372
1373         if (addr_type2 == IPV6_ADDR_ANY &&
1374             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1375                 return 1;
1376
1377         if (addr_type == IPV6_ADDR_ANY &&
1378             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1379                 return 1;
1380
1381         if (sk2_rcv_saddr6 &&
1382             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1383                 return 1;
1384
1385         if (addr_type == IPV6_ADDR_MAPPED &&
1386             !sk2_ipv6only &&
1387             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1388                 return 1;
1389
1390         return 0;
1391 }
1392
1393 /* Gets referenced address, destroys ifaddr */
1394
1395 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1396 {
1397         if (ifp->flags&IFA_F_PERMANENT) {
1398                 spin_lock_bh(&ifp->lock);
1399                 addrconf_del_timer(ifp);
1400                 ifp->flags |= IFA_F_TENTATIVE;
1401                 spin_unlock_bh(&ifp->lock);
1402                 in6_ifa_put(ifp);
1403 #ifdef CONFIG_IPV6_PRIVACY
1404         } else if (ifp->flags&IFA_F_TEMPORARY) {
1405                 struct inet6_ifaddr *ifpub;
1406                 spin_lock_bh(&ifp->lock);
1407                 ifpub = ifp->ifpub;
1408                 if (ifpub) {
1409                         in6_ifa_hold(ifpub);
1410                         spin_unlock_bh(&ifp->lock);
1411                         ipv6_create_tempaddr(ifpub, ifp);
1412                         in6_ifa_put(ifpub);
1413                 } else {
1414                         spin_unlock_bh(&ifp->lock);
1415                 }
1416                 ipv6_del_addr(ifp);
1417 #endif
1418         } else
1419                 ipv6_del_addr(ifp);
1420 }
1421
1422 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1423 {
1424         if (net_ratelimit())
1425                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1426         addrconf_dad_stop(ifp);
1427 }
1428
1429 /* Join to solicited addr multicast group. */
1430
1431 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1432 {
1433         struct in6_addr maddr;
1434
1435         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1436                 return;
1437
1438         addrconf_addr_solict_mult(addr, &maddr);
1439         ipv6_dev_mc_inc(dev, &maddr);
1440 }
1441
1442 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1443 {
1444         struct in6_addr maddr;
1445
1446         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1447                 return;
1448
1449         addrconf_addr_solict_mult(addr, &maddr);
1450         __ipv6_dev_mc_dec(idev, &maddr);
1451 }
1452
1453 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1454 {
1455         struct in6_addr addr;
1456         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1457         if (ipv6_addr_any(&addr))
1458                 return;
1459         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1460 }
1461
1462 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1463 {
1464         struct in6_addr addr;
1465         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1466         if (ipv6_addr_any(&addr))
1467                 return;
1468         __ipv6_dev_ac_dec(ifp->idev, &addr);
1469 }
1470
1471 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1472 {
1473         if (dev->addr_len != ETH_ALEN)
1474                 return -1;
1475         memcpy(eui, dev->dev_addr, 3);
1476         memcpy(eui + 5, dev->dev_addr + 3, 3);
1477
1478         /*
1479          * The zSeries OSA network cards can be shared among various
1480          * OS instances, but the OSA cards have only one MAC address.
1481          * This leads to duplicate address conflicts in conjunction
1482          * with IPv6 if more than one instance uses the same card.
1483          *
1484          * The driver for these cards can deliver a unique 16-bit
1485          * identifier for each instance sharing the same card.  It is
1486          * placed instead of 0xFFFE in the interface identifier.  The
1487          * "u" bit of the interface identifier is not inverted in this
1488          * case.  Hence the resulting interface identifier has local
1489          * scope according to RFC2373.
1490          */
1491         if (dev->dev_id) {
1492                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1493                 eui[4] = dev->dev_id & 0xFF;
1494         } else {
1495                 eui[3] = 0xFF;
1496                 eui[4] = 0xFE;
1497                 eui[0] ^= 2;
1498         }
1499         return 0;
1500 }
1501
1502 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1503 {
1504         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1505         if (dev->addr_len != ARCNET_ALEN)
1506                 return -1;
1507         memset(eui, 0, 7);
1508         eui[7] = *(u8*)dev->dev_addr;
1509         return 0;
1510 }
1511
1512 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1513 {
1514         if (dev->addr_len != INFINIBAND_ALEN)
1515                 return -1;
1516         memcpy(eui, dev->dev_addr + 12, 8);
1517         eui[0] |= 2;
1518         return 0;
1519 }
1520
1521 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1522 {
1523         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1524                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1525                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1526                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1527                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1528                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1529         eui[1] = 0;
1530         eui[2] = 0x5E;
1531         eui[3] = 0xFE;
1532         memcpy(eui + 4, &addr, 4);
1533         return 0;
1534 }
1535 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1536
1537 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1538 {
1539         if (dev->priv_flags & IFF_ISATAP)
1540                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1541         return -1;
1542 }
1543
1544 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1545 {
1546         switch (dev->type) {
1547         case ARPHRD_ETHER:
1548         case ARPHRD_FDDI:
1549         case ARPHRD_IEEE802_TR:
1550                 return addrconf_ifid_eui48(eui, dev);
1551         case ARPHRD_ARCNET:
1552                 return addrconf_ifid_arcnet(eui, dev);
1553         case ARPHRD_INFINIBAND:
1554                 return addrconf_ifid_infiniband(eui, dev);
1555         case ARPHRD_SIT:
1556                 return addrconf_ifid_sit(eui, dev);
1557         }
1558         return -1;
1559 }
1560
1561 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1562 {
1563         int err = -1;
1564         struct inet6_ifaddr *ifp;
1565
1566         read_lock_bh(&idev->lock);
1567         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1568                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1569                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1570                         err = 0;
1571                         break;
1572                 }
1573         }
1574         read_unlock_bh(&idev->lock);
1575         return err;
1576 }
1577
1578 #ifdef CONFIG_IPV6_PRIVACY
1579 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1580 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1581 {
1582 regen:
1583         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1584         idev->rndid[0] &= ~0x02;
1585
1586         /*
1587          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1588          * check if generated address is not inappropriate
1589          *
1590          *  - Reserved subnet anycast (RFC 2526)
1591          *      11111101 11....11 1xxxxxxx
1592          *  - ISATAP (RFC4214) 6.1
1593          *      00-00-5E-FE-xx-xx-xx-xx
1594          *  - value 0
1595          *  - XXX: already assigned to an address on the device
1596          */
1597         if (idev->rndid[0] == 0xfd &&
1598             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1599             (idev->rndid[7]&0x80))
1600                 goto regen;
1601         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1602                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1603                         goto regen;
1604                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1605                         goto regen;
1606         }
1607
1608         return 0;
1609 }
1610
1611 static void ipv6_regen_rndid(unsigned long data)
1612 {
1613         struct inet6_dev *idev = (struct inet6_dev *) data;
1614         unsigned long expires;
1615
1616         rcu_read_lock_bh();
1617         write_lock_bh(&idev->lock);
1618
1619         if (idev->dead)
1620                 goto out;
1621
1622         if (__ipv6_regen_rndid(idev) < 0)
1623                 goto out;
1624
1625         expires = jiffies +
1626                 idev->cnf.temp_prefered_lft * HZ -
1627                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1628         if (time_before(expires, jiffies)) {
1629                 printk(KERN_WARNING
1630                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1631                         idev->dev->name);
1632                 goto out;
1633         }
1634
1635         if (!mod_timer(&idev->regen_timer, expires))
1636                 in6_dev_hold(idev);
1637
1638 out:
1639         write_unlock_bh(&idev->lock);
1640         rcu_read_unlock_bh();
1641         in6_dev_put(idev);
1642 }
1643
1644 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1645         int ret = 0;
1646
1647         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1648                 ret = __ipv6_regen_rndid(idev);
1649         return ret;
1650 }
1651 #endif
1652
1653 /*
1654  *      Add prefix route.
1655  */
1656
1657 static void
1658 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1659                       unsigned long expires, u32 flags)
1660 {
1661         struct fib6_config cfg = {
1662                 .fc_table = RT6_TABLE_PREFIX,
1663                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1664                 .fc_ifindex = dev->ifindex,
1665                 .fc_expires = expires,
1666                 .fc_dst_len = plen,
1667                 .fc_flags = RTF_UP | flags,
1668                 .fc_nlinfo.nl_net = dev_net(dev),
1669         };
1670
1671         ipv6_addr_copy(&cfg.fc_dst, pfx);
1672
1673         /* Prevent useless cloning on PtP SIT.
1674            This thing is done here expecting that the whole
1675            class of non-broadcast devices need not cloning.
1676          */
1677 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1678         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1679                 cfg.fc_flags |= RTF_NONEXTHOP;
1680 #endif
1681
1682         ip6_route_add(&cfg);
1683 }
1684
1685 /* Create "default" multicast route to the interface */
1686
1687 static void addrconf_add_mroute(struct net_device *dev)
1688 {
1689         struct fib6_config cfg = {
1690                 .fc_table = RT6_TABLE_LOCAL,
1691                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1692                 .fc_ifindex = dev->ifindex,
1693                 .fc_dst_len = 8,
1694                 .fc_flags = RTF_UP,
1695                 .fc_nlinfo.nl_net = dev_net(dev),
1696         };
1697
1698         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1699
1700         ip6_route_add(&cfg);
1701 }
1702
1703 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1704 static void sit_route_add(struct net_device *dev)
1705 {
1706         struct fib6_config cfg = {
1707                 .fc_table = RT6_TABLE_MAIN,
1708                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1709                 .fc_ifindex = dev->ifindex,
1710                 .fc_dst_len = 96,
1711                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1712                 .fc_nlinfo.nl_net = dev_net(dev),
1713         };
1714
1715         /* prefix length - 96 bits "::d.d.d.d" */
1716         ip6_route_add(&cfg);
1717 }
1718 #endif
1719
1720 static void addrconf_add_lroute(struct net_device *dev)
1721 {
1722         struct in6_addr addr;
1723
1724         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1725         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1726 }
1727
1728 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1729 {
1730         struct inet6_dev *idev;
1731
1732         ASSERT_RTNL();
1733
1734         if ((idev = ipv6_find_idev(dev)) == NULL)
1735                 return NULL;
1736
1737         /* Add default multicast route */
1738         addrconf_add_mroute(dev);
1739
1740         /* Add link local route */
1741         addrconf_add_lroute(dev);
1742         return idev;
1743 }
1744
1745 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1746 {
1747         struct prefix_info *pinfo;
1748         __u32 valid_lft;
1749         __u32 prefered_lft;
1750         int addr_type;
1751         struct inet6_dev *in6_dev;
1752
1753         pinfo = (struct prefix_info *) opt;
1754
1755         if (len < sizeof(struct prefix_info)) {
1756                 ADBG(("addrconf: prefix option too short\n"));
1757                 return;
1758         }
1759
1760         /*
1761          *      Validation checks ([ADDRCONF], page 19)
1762          */
1763
1764         addr_type = ipv6_addr_type(&pinfo->prefix);
1765
1766         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1767                 return;
1768
1769         valid_lft = ntohl(pinfo->valid);
1770         prefered_lft = ntohl(pinfo->prefered);
1771
1772         if (prefered_lft > valid_lft) {
1773                 if (net_ratelimit())
1774                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1775                 return;
1776         }
1777
1778         in6_dev = in6_dev_get(dev);
1779
1780         if (in6_dev == NULL) {
1781                 if (net_ratelimit())
1782                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1783                 return;
1784         }
1785
1786         /*
1787          *      Two things going on here:
1788          *      1) Add routes for on-link prefixes
1789          *      2) Configure prefixes with the auto flag set
1790          */
1791
1792         if (pinfo->onlink) {
1793                 struct rt6_info *rt;
1794                 unsigned long rt_expires;
1795
1796                 /* Avoid arithmetic overflow. Really, we could
1797                  * save rt_expires in seconds, likely valid_lft,
1798                  * but it would require division in fib gc, that it
1799                  * not good.
1800                  */
1801                 if (HZ > USER_HZ)
1802                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1803                 else
1804                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1805
1806                 if (addrconf_finite_timeout(rt_expires))
1807                         rt_expires *= HZ;
1808
1809                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1810                                 dev->ifindex, 1);
1811
1812                 if (rt && addrconf_is_prefix_route(rt)) {
1813                         /* Autoconf prefix route */
1814                         if (valid_lft == 0) {
1815                                 ip6_del_rt(rt);
1816                                 rt = NULL;
1817                         } else if (addrconf_finite_timeout(rt_expires)) {
1818                                 /* not infinity */
1819                                 rt->rt6i_expires = jiffies + rt_expires;
1820                                 rt->rt6i_flags |= RTF_EXPIRES;
1821                         } else {
1822                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1823                                 rt->rt6i_expires = 0;
1824                         }
1825                 } else if (valid_lft) {
1826                         clock_t expires = 0;
1827                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1828                         if (addrconf_finite_timeout(rt_expires)) {
1829                                 /* not infinity */
1830                                 flags |= RTF_EXPIRES;
1831                                 expires = jiffies_to_clock_t(rt_expires);
1832                         }
1833                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1834                                               dev, expires, flags);
1835                 }
1836                 if (rt)
1837                         dst_release(&rt->u.dst);
1838         }
1839
1840         /* Try to figure out our local address for this prefix */
1841
1842         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1843                 struct inet6_ifaddr * ifp;
1844                 struct in6_addr addr;
1845                 int create = 0, update_lft = 0;
1846
1847                 if (pinfo->prefix_len == 64) {
1848                         memcpy(&addr, &pinfo->prefix, 8);
1849                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1850                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1851                                 in6_dev_put(in6_dev);
1852                                 return;
1853                         }
1854                         goto ok;
1855                 }
1856                 if (net_ratelimit())
1857                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1858                                pinfo->prefix_len);
1859                 in6_dev_put(in6_dev);
1860                 return;
1861
1862 ok:
1863
1864                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1865
1866                 if (ifp == NULL && valid_lft) {
1867                         int max_addresses = in6_dev->cnf.max_addresses;
1868                         u32 addr_flags = 0;
1869
1870 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1871                         if (in6_dev->cnf.optimistic_dad &&
1872                             !ipv6_devconf.forwarding)
1873                                 addr_flags = IFA_F_OPTIMISTIC;
1874 #endif
1875
1876                         /* Do not allow to create too much of autoconfigured
1877                          * addresses; this would be too easy way to crash kernel.
1878                          */
1879                         if (!max_addresses ||
1880                             ipv6_count_addresses(in6_dev) < max_addresses)
1881                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1882                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1883                                                     addr_flags);
1884
1885                         if (!ifp || IS_ERR(ifp)) {
1886                                 in6_dev_put(in6_dev);
1887                                 return;
1888                         }
1889
1890                         update_lft = create = 1;
1891                         ifp->cstamp = jiffies;
1892                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1893                 }
1894
1895                 if (ifp) {
1896                         int flags;
1897                         unsigned long now;
1898 #ifdef CONFIG_IPV6_PRIVACY
1899                         struct inet6_ifaddr *ift;
1900 #endif
1901                         u32 stored_lft;
1902
1903                         /* update lifetime (RFC2462 5.5.3 e) */
1904                         spin_lock(&ifp->lock);
1905                         now = jiffies;
1906                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1907                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1908                         else
1909                                 stored_lft = 0;
1910                         if (!update_lft && stored_lft) {
1911                                 if (valid_lft > MIN_VALID_LIFETIME ||
1912                                     valid_lft > stored_lft)
1913                                         update_lft = 1;
1914                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1915                                         /* valid_lft <= stored_lft is always true */
1916                                         /* XXX: IPsec */
1917                                         update_lft = 0;
1918                                 } else {
1919                                         valid_lft = MIN_VALID_LIFETIME;
1920                                         if (valid_lft < prefered_lft)
1921                                                 prefered_lft = valid_lft;
1922                                         update_lft = 1;
1923                                 }
1924                         }
1925
1926                         if (update_lft) {
1927                                 ifp->valid_lft = valid_lft;
1928                                 ifp->prefered_lft = prefered_lft;
1929                                 ifp->tstamp = now;
1930                                 flags = ifp->flags;
1931                                 ifp->flags &= ~IFA_F_DEPRECATED;
1932                                 spin_unlock(&ifp->lock);
1933
1934                                 if (!(flags&IFA_F_TENTATIVE))
1935                                         ipv6_ifa_notify(0, ifp);
1936                         } else
1937                                 spin_unlock(&ifp->lock);
1938
1939 #ifdef CONFIG_IPV6_PRIVACY
1940                         read_lock_bh(&in6_dev->lock);
1941                         /* update all temporary addresses in the list */
1942                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1943                                 /*
1944                                  * When adjusting the lifetimes of an existing
1945                                  * temporary address, only lower the lifetimes.
1946                                  * Implementations must not increase the
1947                                  * lifetimes of an existing temporary address
1948                                  * when processing a Prefix Information Option.
1949                                  */
1950                                 if (ifp != ift->ifpub)
1951                                         continue;
1952
1953                                 spin_lock(&ift->lock);
1954                                 flags = ift->flags;
1955                                 if (ift->valid_lft > valid_lft &&
1956                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1957                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1958                                 if (ift->prefered_lft > prefered_lft &&
1959                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1960                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1961                                 spin_unlock(&ift->lock);
1962                                 if (!(flags&IFA_F_TENTATIVE))
1963                                         ipv6_ifa_notify(0, ift);
1964                         }
1965
1966                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1967                                 /*
1968                                  * When a new public address is created as described in [ADDRCONF],
1969                                  * also create a new temporary address.
1970                                  */
1971                                 read_unlock_bh(&in6_dev->lock);
1972                                 ipv6_create_tempaddr(ifp, NULL);
1973                         } else {
1974                                 read_unlock_bh(&in6_dev->lock);
1975                         }
1976 #endif
1977                         in6_ifa_put(ifp);
1978                         addrconf_verify(0);
1979                 }
1980         }
1981         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1982         in6_dev_put(in6_dev);
1983 }
1984
1985 /*
1986  *      Set destination address.
1987  *      Special case for SIT interfaces where we create a new "virtual"
1988  *      device.
1989  */
1990 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1991 {
1992         struct in6_ifreq ireq;
1993         struct net_device *dev;
1994         int err = -EINVAL;
1995
1996         rtnl_lock();
1997
1998         err = -EFAULT;
1999         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2000                 goto err_exit;
2001
2002         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2003
2004         err = -ENODEV;
2005         if (dev == NULL)
2006                 goto err_exit;
2007
2008 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2009         if (dev->type == ARPHRD_SIT) {
2010                 struct ifreq ifr;
2011                 mm_segment_t    oldfs;
2012                 struct ip_tunnel_parm p;
2013
2014                 err = -EADDRNOTAVAIL;
2015                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2016                         goto err_exit;
2017
2018                 memset(&p, 0, sizeof(p));
2019                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2020                 p.iph.saddr = 0;
2021                 p.iph.version = 4;
2022                 p.iph.ihl = 5;
2023                 p.iph.protocol = IPPROTO_IPV6;
2024                 p.iph.ttl = 64;
2025                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2026
2027                 oldfs = get_fs(); set_fs(KERNEL_DS);
2028                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2029                 set_fs(oldfs);
2030
2031                 if (err == 0) {
2032                         err = -ENOBUFS;
2033                         dev = __dev_get_by_name(net, p.name);
2034                         if (!dev)
2035                                 goto err_exit;
2036                         err = dev_open(dev);
2037                 }
2038         }
2039 #endif
2040
2041 err_exit:
2042         rtnl_unlock();
2043         return err;
2044 }
2045
2046 /*
2047  *      Manual configuration of address on an interface
2048  */
2049 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2050                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2051                           __u32 valid_lft)
2052 {
2053         struct inet6_ifaddr *ifp;
2054         struct inet6_dev *idev;
2055         struct net_device *dev;
2056         int scope;
2057         u32 flags;
2058         clock_t expires;
2059         unsigned long timeout;
2060
2061         ASSERT_RTNL();
2062
2063         if (plen > 128)
2064                 return -EINVAL;
2065
2066         /* check the lifetime */
2067         if (!valid_lft || prefered_lft > valid_lft)
2068                 return -EINVAL;
2069
2070         dev = __dev_get_by_index(net, ifindex);
2071         if (!dev)
2072                 return -ENODEV;
2073
2074         if ((idev = addrconf_add_dev(dev)) == NULL)
2075                 return -ENOBUFS;
2076
2077         scope = ipv6_addr_scope(pfx);
2078
2079         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2080         if (addrconf_finite_timeout(timeout)) {
2081                 expires = jiffies_to_clock_t(timeout * HZ);
2082                 valid_lft = timeout;
2083                 flags = RTF_EXPIRES;
2084         } else {
2085                 expires = 0;
2086                 flags = 0;
2087                 ifa_flags |= IFA_F_PERMANENT;
2088         }
2089
2090         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2091         if (addrconf_finite_timeout(timeout)) {
2092                 if (timeout == 0)
2093                         ifa_flags |= IFA_F_DEPRECATED;
2094                 prefered_lft = timeout;
2095         }
2096
2097         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2098
2099         if (!IS_ERR(ifp)) {
2100                 spin_lock_bh(&ifp->lock);
2101                 ifp->valid_lft = valid_lft;
2102                 ifp->prefered_lft = prefered_lft;
2103                 ifp->tstamp = jiffies;
2104                 spin_unlock_bh(&ifp->lock);
2105
2106                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2107                                       expires, flags);
2108                 /*
2109                  * Note that section 3.1 of RFC 4429 indicates
2110                  * that the Optimistic flag should not be set for
2111                  * manually configured addresses
2112                  */
2113                 addrconf_dad_start(ifp, 0);
2114                 in6_ifa_put(ifp);
2115                 addrconf_verify(0);
2116                 return 0;
2117         }
2118
2119         return PTR_ERR(ifp);
2120 }
2121
2122 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2123                           unsigned int plen)
2124 {
2125         struct inet6_ifaddr *ifp;
2126         struct inet6_dev *idev;
2127         struct net_device *dev;
2128
2129         if (plen > 128)
2130                 return -EINVAL;
2131
2132         dev = __dev_get_by_index(net, ifindex);
2133         if (!dev)
2134                 return -ENODEV;
2135
2136         if ((idev = __in6_dev_get(dev)) == NULL)
2137                 return -ENXIO;
2138
2139         read_lock_bh(&idev->lock);
2140         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2141                 if (ifp->prefix_len == plen &&
2142                     ipv6_addr_equal(pfx, &ifp->addr)) {
2143                         in6_ifa_hold(ifp);
2144                         read_unlock_bh(&idev->lock);
2145
2146                         ipv6_del_addr(ifp);
2147
2148                         /* If the last address is deleted administratively,
2149                            disable IPv6 on this interface.
2150                          */
2151                         if (idev->addr_list == NULL)
2152                                 addrconf_ifdown(idev->dev, 1);
2153                         return 0;
2154                 }
2155         }
2156         read_unlock_bh(&idev->lock);
2157         return -EADDRNOTAVAIL;
2158 }
2159
2160
2161 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2162 {
2163         struct in6_ifreq ireq;
2164         int err;
2165
2166         if (!capable(CAP_NET_ADMIN))
2167                 return -EPERM;
2168
2169         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2170                 return -EFAULT;
2171
2172         rtnl_lock();
2173         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2174                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2175                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2176         rtnl_unlock();
2177         return err;
2178 }
2179
2180 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2181 {
2182         struct in6_ifreq ireq;
2183         int err;
2184
2185         if (!capable(CAP_NET_ADMIN))
2186                 return -EPERM;
2187
2188         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2189                 return -EFAULT;
2190
2191         rtnl_lock();
2192         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2193                              ireq.ifr6_prefixlen);
2194         rtnl_unlock();
2195         return err;
2196 }
2197
2198 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2199 static void sit_add_v4_addrs(struct inet6_dev *idev)
2200 {
2201         struct inet6_ifaddr * ifp;
2202         struct in6_addr addr;
2203         struct net_device *dev;
2204         struct net *net = dev_net(idev->dev);
2205         int scope;
2206
2207         ASSERT_RTNL();
2208
2209         memset(&addr, 0, sizeof(struct in6_addr));
2210         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2211
2212         if (idev->dev->flags&IFF_POINTOPOINT) {
2213                 addr.s6_addr32[0] = htonl(0xfe800000);
2214                 scope = IFA_LINK;
2215         } else {
2216                 scope = IPV6_ADDR_COMPATv4;
2217         }
2218
2219         if (addr.s6_addr32[3]) {
2220                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2221                 if (!IS_ERR(ifp)) {
2222                         spin_lock_bh(&ifp->lock);
2223                         ifp->flags &= ~IFA_F_TENTATIVE;
2224                         spin_unlock_bh(&ifp->lock);
2225                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2226                         in6_ifa_put(ifp);
2227                 }
2228                 return;
2229         }
2230
2231         for_each_netdev(net, dev) {
2232                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2233                 if (in_dev && (dev->flags & IFF_UP)) {
2234                         struct in_ifaddr * ifa;
2235
2236                         int flag = scope;
2237
2238                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2239                                 int plen;
2240
2241                                 addr.s6_addr32[3] = ifa->ifa_local;
2242
2243                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2244                                         continue;
2245                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2246                                         if (idev->dev->flags&IFF_POINTOPOINT)
2247                                                 continue;
2248                                         flag |= IFA_HOST;
2249                                 }
2250                                 if (idev->dev->flags&IFF_POINTOPOINT)
2251                                         plen = 64;
2252                                 else
2253                                         plen = 96;
2254
2255                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2256                                                     IFA_F_PERMANENT);
2257                                 if (!IS_ERR(ifp)) {
2258                                         spin_lock_bh(&ifp->lock);
2259                                         ifp->flags &= ~IFA_F_TENTATIVE;
2260                                         spin_unlock_bh(&ifp->lock);
2261                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2262                                         in6_ifa_put(ifp);
2263                                 }
2264                         }
2265                 }
2266         }
2267 }
2268 #endif
2269
2270 static void init_loopback(struct net_device *dev)
2271 {
2272         struct inet6_dev  *idev;
2273         struct inet6_ifaddr * ifp;
2274
2275         /* ::1 */
2276
2277         ASSERT_RTNL();
2278
2279         if ((idev = ipv6_find_idev(dev)) == NULL) {
2280                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2281                 return;
2282         }
2283
2284         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2285         if (!IS_ERR(ifp)) {
2286                 spin_lock_bh(&ifp->lock);
2287                 ifp->flags &= ~IFA_F_TENTATIVE;
2288                 spin_unlock_bh(&ifp->lock);
2289                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2290                 in6_ifa_put(ifp);
2291         }
2292 }
2293
2294 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2295 {
2296         struct inet6_ifaddr * ifp;
2297         u32 addr_flags = IFA_F_PERMANENT;
2298
2299 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2300         if (idev->cnf.optimistic_dad &&
2301             !ipv6_devconf.forwarding)
2302                 addr_flags |= IFA_F_OPTIMISTIC;
2303 #endif
2304
2305
2306         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2307         if (!IS_ERR(ifp)) {
2308                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2309                 addrconf_dad_start(ifp, 0);
2310                 in6_ifa_put(ifp);
2311         }
2312 }
2313
2314 static void addrconf_dev_config(struct net_device *dev)
2315 {
2316         struct in6_addr addr;
2317         struct inet6_dev    * idev;
2318
2319         ASSERT_RTNL();
2320
2321         if ((dev->type != ARPHRD_ETHER) &&
2322             (dev->type != ARPHRD_FDDI) &&
2323             (dev->type != ARPHRD_IEEE802_TR) &&
2324             (dev->type != ARPHRD_ARCNET) &&
2325             (dev->type != ARPHRD_INFINIBAND)) {
2326                 /* Alas, we support only Ethernet autoconfiguration. */
2327                 return;
2328         }
2329
2330         idev = addrconf_add_dev(dev);
2331         if (idev == NULL)
2332                 return;
2333
2334         memset(&addr, 0, sizeof(struct in6_addr));
2335         addr.s6_addr32[0] = htonl(0xFE800000);
2336
2337         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2338                 addrconf_add_linklocal(idev, &addr);
2339 }
2340
2341 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2342 static void addrconf_sit_config(struct net_device *dev)
2343 {
2344         struct inet6_dev *idev;
2345
2346         ASSERT_RTNL();
2347
2348         /*
2349          * Configure the tunnel with one of our IPv4
2350          * addresses... we should configure all of
2351          * our v4 addrs in the tunnel
2352          */
2353
2354         if ((idev = ipv6_find_idev(dev)) == NULL) {
2355                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2356                 return;
2357         }
2358
2359         if (dev->priv_flags & IFF_ISATAP) {
2360                 struct in6_addr addr;
2361
2362                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2363                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2364                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2365                         addrconf_add_linklocal(idev, &addr);
2366                 return;
2367         }
2368
2369         sit_add_v4_addrs(idev);
2370
2371         if (dev->flags&IFF_POINTOPOINT) {
2372                 addrconf_add_mroute(dev);
2373                 addrconf_add_lroute(dev);
2374         } else
2375                 sit_route_add(dev);
2376 }
2377 #endif
2378
2379 static inline int
2380 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2381 {
2382         struct in6_addr lladdr;
2383
2384         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2385                 addrconf_add_linklocal(idev, &lladdr);
2386                 return 0;
2387         }
2388         return -1;
2389 }
2390
2391 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2392 {
2393         struct net_device *link_dev;
2394         struct net *net = dev_net(idev->dev);
2395
2396         /* first try to inherit the link-local address from the link device */
2397         if (idev->dev->iflink &&
2398             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2399                 if (!ipv6_inherit_linklocal(idev, link_dev))
2400                         return;
2401         }
2402         /* then try to inherit it from any device */
2403         for_each_netdev(net, link_dev) {
2404                 if (!ipv6_inherit_linklocal(idev, link_dev))
2405                         return;
2406         }
2407         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2408 }
2409
2410 /*
2411  * Autoconfigure tunnel with a link-local address so routing protocols,
2412  * DHCPv6, MLD etc. can be run over the virtual link
2413  */
2414
2415 static void addrconf_ip6_tnl_config(struct net_device *dev)
2416 {
2417         struct inet6_dev *idev;
2418
2419         ASSERT_RTNL();
2420
2421         if ((idev = addrconf_add_dev(dev)) == NULL) {
2422                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2423                 return;
2424         }
2425         ip6_tnl_add_linklocal(idev);
2426 }
2427
2428 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2429                            void * data)
2430 {
2431         struct net_device *dev = (struct net_device *) data;
2432         struct inet6_dev *idev = __in6_dev_get(dev);
2433         int run_pending = 0;
2434         int err;
2435
2436         switch(event) {
2437         case NETDEV_REGISTER:
2438                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2439                         idev = ipv6_add_dev(dev);
2440                         if (!idev)
2441                                 return notifier_from_errno(-ENOMEM);
2442                 }
2443                 break;
2444         case NETDEV_UP:
2445         case NETDEV_CHANGE:
2446                 if (dev->flags & IFF_SLAVE)
2447                         break;
2448
2449                 if (event == NETDEV_UP) {
2450                         if (!addrconf_qdisc_ok(dev)) {
2451                                 /* device is not ready yet. */
2452                                 printk(KERN_INFO
2453                                         "ADDRCONF(NETDEV_UP): %s: "
2454                                         "link is not ready\n",
2455                                         dev->name);
2456                                 break;
2457                         }
2458
2459                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2460                                 idev = ipv6_add_dev(dev);
2461
2462                         if (idev)
2463                                 idev->if_flags |= IF_READY;
2464                 } else {
2465                         if (!addrconf_qdisc_ok(dev)) {
2466                                 /* device is still not ready. */
2467                                 break;
2468                         }
2469
2470                         if (idev) {
2471                                 if (idev->if_flags & IF_READY) {
2472                                         /* device is already configured. */
2473                                         break;
2474                                 }
2475                                 idev->if_flags |= IF_READY;
2476                         }
2477
2478                         printk(KERN_INFO
2479                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2480                                         "link becomes ready\n",
2481                                         dev->name);
2482
2483                         run_pending = 1;
2484                 }
2485
2486                 switch(dev->type) {
2487 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2488                 case ARPHRD_SIT:
2489                         addrconf_sit_config(dev);
2490                         break;
2491 #endif
2492                 case ARPHRD_TUNNEL6:
2493                         addrconf_ip6_tnl_config(dev);
2494                         break;
2495                 case ARPHRD_LOOPBACK:
2496                         init_loopback(dev);
2497                         break;
2498
2499                 default:
2500                         addrconf_dev_config(dev);
2501                         break;
2502                 }
2503                 if (idev) {
2504                         if (run_pending)
2505                                 addrconf_dad_run(idev);
2506
2507                         /* If the MTU changed during the interface down, when the
2508                            interface up, the changed MTU must be reflected in the
2509                            idev as well as routers.
2510                          */
2511                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2512                                 rt6_mtu_change(dev, dev->mtu);
2513                                 idev->cnf.mtu6 = dev->mtu;
2514                         }
2515                         idev->tstamp = jiffies;
2516                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2517                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2518                            stop IPv6 on this interface.
2519                          */
2520                         if (dev->mtu < IPV6_MIN_MTU)
2521                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2522                 }
2523                 break;
2524
2525         case NETDEV_CHANGEMTU:
2526                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2527                         rt6_mtu_change(dev, dev->mtu);
2528                         idev->cnf.mtu6 = dev->mtu;
2529                         break;
2530                 }
2531
2532                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2533                         idev = ipv6_add_dev(dev);
2534                         if (idev)
2535                                 break;
2536                 }
2537
2538                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2539
2540         case NETDEV_DOWN:
2541         case NETDEV_UNREGISTER:
2542                 /*
2543                  *      Remove all addresses from this interface.
2544                  */
2545                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2546                 break;
2547
2548         case NETDEV_CHANGENAME:
2549                 if (idev) {
2550                         snmp6_unregister_dev(idev);
2551                         addrconf_sysctl_unregister(idev);
2552                         addrconf_sysctl_register(idev);
2553                         err = snmp6_register_dev(idev);
2554                         if (err)
2555                                 return notifier_from_errno(err);
2556                 }
2557                 break;
2558         }
2559
2560         return NOTIFY_OK;
2561 }
2562
2563 /*
2564  *      addrconf module should be notified of a device going up
2565  */
2566 static struct notifier_block ipv6_dev_notf = {
2567         .notifier_call = addrconf_notify,
2568         .priority = 0
2569 };
2570
2571 static int addrconf_ifdown(struct net_device *dev, int how)
2572 {
2573         struct inet6_dev *idev;
2574         struct inet6_ifaddr *ifa, **bifa;
2575         struct net *net = dev_net(dev);
2576         int i;
2577
2578         ASSERT_RTNL();
2579
2580         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2581                 how = 0;
2582
2583         rt6_ifdown(net, dev);
2584         neigh_ifdown(&nd_tbl, dev);
2585
2586         idev = __in6_dev_get(dev);
2587         if (idev == NULL)
2588                 return -ENODEV;
2589
2590         /* Step 1: remove reference to ipv6 device from parent device.
2591                    Do not dev_put!
2592          */
2593         if (how) {
2594                 idev->dead = 1;
2595
2596                 /* protected by rtnl_lock */
2597                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2598
2599                 /* Step 1.5: remove snmp6 entry */
2600                 snmp6_unregister_dev(idev);
2601
2602         }
2603
2604         /* Step 2: clear hash table */
2605         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2606                 bifa = &inet6_addr_lst[i];
2607
2608                 write_lock_bh(&addrconf_hash_lock);
2609                 while ((ifa = *bifa) != NULL) {
2610                         if (ifa->idev == idev) {
2611                                 *bifa = ifa->lst_next;
2612                                 ifa->lst_next = NULL;
2613                                 addrconf_del_timer(ifa);
2614                                 in6_ifa_put(ifa);
2615                                 continue;
2616                         }
2617                         bifa = &ifa->lst_next;
2618                 }
2619                 write_unlock_bh(&addrconf_hash_lock);
2620         }
2621
2622         write_lock_bh(&idev->lock);
2623
2624         /* Step 3: clear flags for stateless addrconf */
2625         if (!how)
2626                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2627
2628         /* Step 4: clear address list */
2629 #ifdef CONFIG_IPV6_PRIVACY
2630         if (how && del_timer(&idev->regen_timer))
2631                 in6_dev_put(idev);
2632
2633         /* clear tempaddr list */
2634         while ((ifa = idev->tempaddr_list) != NULL) {
2635                 idev->tempaddr_list = ifa->tmp_next;
2636                 ifa->tmp_next = NULL;
2637                 ifa->dead = 1;
2638                 write_unlock_bh(&idev->lock);
2639                 spin_lock_bh(&ifa->lock);
2640
2641                 if (ifa->ifpub) {
2642                         in6_ifa_put(ifa->ifpub);
2643                         ifa->ifpub = NULL;
2644                 }
2645                 spin_unlock_bh(&ifa->lock);
2646                 in6_ifa_put(ifa);
2647                 write_lock_bh(&idev->lock);
2648         }
2649 #endif
2650         while ((ifa = idev->addr_list) != NULL) {
2651                 idev->addr_list = ifa->if_next;
2652                 ifa->if_next = NULL;
2653                 ifa->dead = 1;
2654                 addrconf_del_timer(ifa);
2655                 write_unlock_bh(&idev->lock);
2656
2657                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2658                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2659                 in6_ifa_put(ifa);
2660
2661                 write_lock_bh(&idev->lock);
2662         }
2663         write_unlock_bh(&idev->lock);
2664
2665         /* Step 5: Discard multicast list */
2666
2667         if (how)
2668                 ipv6_mc_destroy_dev(idev);
2669         else
2670                 ipv6_mc_down(idev);
2671
2672         idev->tstamp = jiffies;
2673
2674         /* Shot the device (if unregistered) */
2675
2676         if (how) {
2677                 addrconf_sysctl_unregister(idev);
2678                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2679                 neigh_ifdown(&nd_tbl, dev);
2680                 in6_dev_put(idev);
2681         }
2682         return 0;
2683 }
2684
2685 static void addrconf_rs_timer(unsigned long data)
2686 {
2687         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2688
2689         if (ifp->idev->cnf.forwarding)
2690                 goto out;
2691
2692         if (ifp->idev->if_flags & IF_RA_RCVD) {
2693                 /*
2694                  *      Announcement received after solicitation
2695                  *      was sent
2696                  */
2697                 goto out;
2698         }
2699
2700         spin_lock(&ifp->lock);
2701         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2702                 /* The wait after the last probe can be shorter */
2703                 addrconf_mod_timer(ifp, AC_RS,
2704                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2705                                    ifp->idev->cnf.rtr_solicit_delay :
2706                                    ifp->idev->cnf.rtr_solicit_interval);
2707                 spin_unlock(&ifp->lock);
2708
2709                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2710         } else {
2711                 spin_unlock(&ifp->lock);
2712                 /*
2713                  * Note: we do not support deprecated "all on-link"
2714                  * assumption any longer.
2715                  */
2716                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2717                        ifp->idev->dev->name);
2718         }
2719
2720 out:
2721         in6_ifa_put(ifp);
2722 }
2723
2724 /*
2725  *      Duplicate Address Detection
2726  */
2727 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2728 {
2729         unsigned long rand_num;
2730         struct inet6_dev *idev = ifp->idev;
2731
2732         if (ifp->flags & IFA_F_OPTIMISTIC)
2733                 rand_num = 0;
2734         else
2735                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2736
2737         ifp->probes = idev->cnf.dad_transmits;
2738         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2739 }
2740
2741 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2742 {
2743         struct inet6_dev *idev = ifp->idev;
2744         struct net_device *dev = idev->dev;
2745
2746         addrconf_join_solict(dev, &ifp->addr);
2747
2748         net_srandom(ifp->addr.s6_addr32[3]);
2749
2750         read_lock_bh(&idev->lock);
2751         if (ifp->dead)
2752                 goto out;
2753         spin_lock_bh(&ifp->lock);
2754
2755         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2756             !(ifp->flags&IFA_F_TENTATIVE) ||
2757             ifp->flags & IFA_F_NODAD) {
2758                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2759                 spin_unlock_bh(&ifp->lock);
2760                 read_unlock_bh(&idev->lock);
2761
2762                 addrconf_dad_completed(ifp);
2763                 return;
2764         }
2765
2766         if (!(idev->if_flags & IF_READY)) {
2767                 spin_unlock_bh(&ifp->lock);
2768                 read_unlock_bh(&idev->lock);
2769                 /*
2770                  * If the defice is not ready:
2771                  * - keep it tentative if it is a permanent address.
2772                  * - otherwise, kill it.
2773                  */
2774                 in6_ifa_hold(ifp);
2775                 addrconf_dad_stop(ifp);
2776                 return;
2777         }
2778
2779         /*
2780          * Optimistic nodes can start receiving
2781          * Frames right away
2782          */
2783         if(ifp->flags & IFA_F_OPTIMISTIC)
2784                 ip6_ins_rt(ifp->rt);
2785
2786         addrconf_dad_kick(ifp);
2787         spin_unlock_bh(&ifp->lock);
2788 out:
2789         read_unlock_bh(&idev->lock);
2790 }
2791
2792 static void addrconf_dad_timer(unsigned long data)
2793 {
2794         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2795         struct inet6_dev *idev = ifp->idev;
2796         struct in6_addr mcaddr;
2797
2798         read_lock_bh(&idev->lock);
2799         if (idev->dead) {
2800                 read_unlock_bh(&idev->lock);
2801                 goto out;
2802         }
2803         spin_lock_bh(&ifp->lock);
2804         if (ifp->probes == 0) {
2805                 /*
2806                  * DAD was successful
2807                  */
2808
2809                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2810                 spin_unlock_bh(&ifp->lock);
2811                 read_unlock_bh(&idev->lock);
2812
2813                 addrconf_dad_completed(ifp);
2814
2815                 goto out;
2816         }
2817
2818         ifp->probes--;
2819         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2820         spin_unlock_bh(&ifp->lock);
2821         read_unlock_bh(&idev->lock);
2822
2823         /* send a neighbour solicitation for our addr */
2824         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2825         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2826 out:
2827         in6_ifa_put(ifp);
2828 }
2829
2830 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2831 {
2832         struct net_device *     dev = ifp->idev->dev;
2833
2834         /*
2835          *      Configure the address for reception. Now it is valid.
2836          */
2837
2838         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2839
2840         /* If added prefix is link local and forwarding is off,
2841            start sending router solicitations.
2842          */
2843
2844         if (ifp->idev->cnf.forwarding == 0 &&
2845             ifp->idev->cnf.rtr_solicits > 0 &&
2846             (dev->flags&IFF_LOOPBACK) == 0 &&
2847             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2848                 /*
2849                  *      If a host as already performed a random delay
2850                  *      [...] as part of DAD [...] there is no need
2851                  *      to delay again before sending the first RS
2852                  */
2853                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2854
2855                 spin_lock_bh(&ifp->lock);
2856                 ifp->probes = 1;
2857                 ifp->idev->if_flags |= IF_RS_SENT;
2858                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2859                 spin_unlock_bh(&ifp->lock);
2860         }
2861 }
2862
2863 static void addrconf_dad_run(struct inet6_dev *idev) {
2864         struct inet6_ifaddr *ifp;
2865
2866         read_lock_bh(&idev->lock);
2867         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2868                 spin_lock_bh(&ifp->lock);
2869                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2870                         spin_unlock_bh(&ifp->lock);
2871                         continue;
2872                 }
2873                 spin_unlock_bh(&ifp->lock);
2874                 addrconf_dad_kick(ifp);
2875         }
2876         read_unlock_bh(&idev->lock);
2877 }
2878
2879 #ifdef CONFIG_PROC_FS
2880 struct if6_iter_state {
2881         struct seq_net_private p;
2882         int bucket;
2883 };
2884
2885 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2886 {
2887         struct inet6_ifaddr *ifa = NULL;
2888         struct if6_iter_state *state = seq->private;
2889         struct net *net = seq_file_net(seq);
2890
2891         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2892                 ifa = inet6_addr_lst[state->bucket];
2893
2894                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2895                         ifa = ifa->lst_next;
2896                 if (ifa)
2897                         break;
2898         }
2899         return ifa;
2900 }
2901
2902 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2903 {
2904         struct if6_iter_state *state = seq->private;
2905         struct net *net = seq_file_net(seq);
2906
2907         ifa = ifa->lst_next;
2908 try_again:
2909         if (ifa) {
2910                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2911                         ifa = ifa->lst_next;
2912                         goto try_again;
2913                 }
2914         }
2915
2916         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2917                 ifa = inet6_addr_lst[state->bucket];
2918                 goto try_again;
2919         }
2920
2921         return ifa;
2922 }
2923
2924 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2925 {
2926         struct inet6_ifaddr *ifa = if6_get_first(seq);
2927
2928         if (ifa)
2929                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2930                         --pos;
2931         return pos ? NULL : ifa;
2932 }
2933
2934 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2935         __acquires(addrconf_hash_lock)
2936 {
2937         read_lock_bh(&addrconf_hash_lock);
2938         return if6_get_idx(seq, *pos);
2939 }
2940
2941 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2942 {
2943         struct inet6_ifaddr *ifa;
2944
2945         ifa = if6_get_next(seq, v);
2946         ++*pos;
2947         return ifa;
2948 }
2949
2950 static void if6_seq_stop(struct seq_file *seq, void *v)
2951         __releases(addrconf_hash_lock)
2952 {
2953         read_unlock_bh(&addrconf_hash_lock);
2954 }
2955
2956 static int if6_seq_show(struct seq_file *seq, void *v)
2957 {
2958         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2959         seq_printf(seq,
2960                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2961                    NIP6(ifp->addr),
2962                    ifp->idev->dev->ifindex,
2963                    ifp->prefix_len,
2964                    ifp->scope,
2965                    ifp->flags,
2966                    ifp->idev->dev->name);
2967         return 0;
2968 }
2969
2970 static const struct seq_operations if6_seq_ops = {
2971         .start  = if6_seq_start,
2972         .next   = if6_seq_next,
2973         .show   = if6_seq_show,
2974         .stop   = if6_seq_stop,
2975 };
2976
2977 static int if6_seq_open(struct inode *inode, struct file *file)
2978 {
2979         return seq_open_net(inode, file, &if6_seq_ops,
2980                             sizeof(struct if6_iter_state));
2981 }
2982
2983 static const struct file_operations if6_fops = {
2984         .owner          = THIS_MODULE,
2985         .open           = if6_seq_open,
2986         .read           = seq_read,
2987         .llseek         = seq_lseek,
2988         .release        = seq_release_net,
2989 };
2990
2991 static int if6_proc_net_init(struct net *net)
2992 {
2993         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2994                 return -ENOMEM;
2995         return 0;
2996 }
2997
2998 static void if6_proc_net_exit(struct net *net)
2999 {
3000        proc_net_remove(net, "if_inet6");
3001 }
3002
3003 static struct pernet_operations if6_proc_net_ops = {
3004        .init = if6_proc_net_init,
3005        .exit = if6_proc_net_exit,
3006 };
3007
3008 int __init if6_proc_init(void)
3009 {
3010         return register_pernet_subsys(&if6_proc_net_ops);
3011 }
3012
3013 void if6_proc_exit(void)
3014 {
3015         unregister_pernet_subsys(&if6_proc_net_ops);
3016 }
3017 #endif  /* CONFIG_PROC_FS */
3018
3019 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3020 /* Check if address is a home address configured on any interface. */
3021 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3022 {
3023         int ret = 0;
3024         struct inet6_ifaddr * ifp;
3025         u8 hash = ipv6_addr_hash(addr);
3026         read_lock_bh(&addrconf_hash_lock);
3027         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3028                 if (!net_eq(dev_net(ifp->idev->dev), net))
3029                         continue;
3030                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3031                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3032                         ret = 1;
3033                         break;
3034                 }
3035         }
3036         read_unlock_bh(&addrconf_hash_lock);
3037         return ret;
3038 }
3039 #endif
3040
3041 /*
3042  *      Periodic address status verification
3043  */
3044
3045 static void addrconf_verify(unsigned long foo)
3046 {
3047         struct inet6_ifaddr *ifp;
3048         unsigned long now, next;
3049         int i;
3050
3051         spin_lock_bh(&addrconf_verify_lock);
3052         now = jiffies;
3053         next = now + ADDR_CHECK_FREQUENCY;
3054
3055         del_timer(&addr_chk_timer);
3056
3057         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3058
3059 restart:
3060                 read_lock(&addrconf_hash_lock);
3061                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3062                         unsigned long age;
3063 #ifdef CONFIG_IPV6_PRIVACY
3064                         unsigned long regen_advance;
3065 #endif
3066
3067                         if (ifp->flags & IFA_F_PERMANENT)
3068                                 continue;
3069
3070                         spin_lock(&ifp->lock);
3071                         age = (now - ifp->tstamp) / HZ;
3072
3073 #ifdef CONFIG_IPV6_PRIVACY
3074                         regen_advance = ifp->idev->cnf.regen_max_retry *
3075                                         ifp->idev->cnf.dad_transmits *
3076                                         ifp->idev->nd_parms->retrans_time / HZ;
3077 #endif
3078
3079                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3080                             age >= ifp->valid_lft) {
3081                                 spin_unlock(&ifp->lock);
3082                                 in6_ifa_hold(ifp);
3083                                 read_unlock(&addrconf_hash_lock);
3084                                 ipv6_del_addr(ifp);
3085                                 goto restart;
3086                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3087                                 spin_unlock(&ifp->lock);
3088                                 continue;
3089                         } else if (age >= ifp->prefered_lft) {
3090                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3091                                 int deprecate = 0;
3092
3093                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3094                                         deprecate = 1;
3095                                         ifp->flags |= IFA_F_DEPRECATED;
3096                                 }
3097
3098                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3099                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3100
3101                                 spin_unlock(&ifp->lock);
3102
3103                                 if (deprecate) {
3104                                         in6_ifa_hold(ifp);
3105                                         read_unlock(&addrconf_hash_lock);
3106
3107                                         ipv6_ifa_notify(0, ifp);
3108                                         in6_ifa_put(ifp);
3109                                         goto restart;
3110                                 }
3111 #ifdef CONFIG_IPV6_PRIVACY
3112                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3113                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3114                                 if (age >= ifp->prefered_lft - regen_advance) {
3115                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3116                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3117                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3118                                         if (!ifp->regen_count && ifpub) {
3119                                                 ifp->regen_count++;
3120                                                 in6_ifa_hold(ifp);
3121                                                 in6_ifa_hold(ifpub);
3122                                                 spin_unlock(&ifp->lock);
3123                                                 read_unlock(&addrconf_hash_lock);
3124                                                 spin_lock(&ifpub->lock);
3125                                                 ifpub->regen_count = 0;
3126                                                 spin_unlock(&ifpub->lock);
3127                                                 ipv6_create_tempaddr(ifpub, ifp);
3128                                                 in6_ifa_put(ifpub);
3129                                                 in6_ifa_put(ifp);
3130                                                 goto restart;
3131                                         }
3132                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3133                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3134                                 spin_unlock(&ifp->lock);
3135 #endif
3136                         } else {
3137                                 /* ifp->prefered_lft <= ifp->valid_lft */
3138                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3139                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3140                                 spin_unlock(&ifp->lock);
3141                         }
3142                 }
3143                 read_unlock(&addrconf_hash_lock);
3144         }
3145
3146         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3147         add_timer(&addr_chk_timer);
3148         spin_unlock_bh(&addrconf_verify_lock);
3149 }
3150
3151 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3152 {
3153         struct in6_addr *pfx = NULL;
3154
3155         if (addr)
3156                 pfx = nla_data(addr);
3157
3158         if (local) {
3159                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3160                         pfx = NULL;
3161                 else
3162                         pfx = nla_data(local);
3163         }
3164
3165         return pfx;
3166 }
3167
3168 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3169         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3170         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3171         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3172 };
3173
3174 static int
3175 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3176 {
3177         struct net *net = sock_net(skb->sk);
3178         struct ifaddrmsg *ifm;
3179         struct nlattr *tb[IFA_MAX+1];
3180         struct in6_addr *pfx;
3181         int err;
3182
3183         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3184         if (err < 0)
3185                 return err;
3186
3187         ifm = nlmsg_data(nlh);
3188         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3189         if (pfx == NULL)
3190                 return -EINVAL;
3191
3192         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3193 }
3194
3195 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3196                              u32 prefered_lft, u32 valid_lft)
3197 {
3198         u32 flags;
3199         clock_t expires;
3200         unsigned long timeout;
3201
3202         if (!valid_lft || (prefered_lft > valid_lft))
3203                 return -EINVAL;
3204
3205         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3206         if (addrconf_finite_timeout(timeout)) {
3207                 expires = jiffies_to_clock_t(timeout * HZ);
3208                 valid_lft = timeout;
3209                 flags = RTF_EXPIRES;
3210         } else {
3211                 expires = 0;
3212                 flags = 0;
3213                 ifa_flags |= IFA_F_PERMANENT;
3214         }
3215
3216         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3217         if (addrconf_finite_timeout(timeout)) {
3218                 if (timeout == 0)
3219                         ifa_flags |= IFA_F_DEPRECATED;
3220                 prefered_lft = timeout;
3221         }
3222
3223         spin_lock_bh(&ifp->lock);
3224         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3225         ifp->tstamp = jiffies;
3226         ifp->valid_lft = valid_lft;
3227         ifp->prefered_lft = prefered_lft;
3228
3229         spin_unlock_bh(&ifp->lock);
3230         if (!(ifp->flags&IFA_F_TENTATIVE))
3231                 ipv6_ifa_notify(0, ifp);
3232
3233         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3234                               expires, flags);
3235         addrconf_verify(0);
3236
3237         return 0;
3238 }
3239
3240 static int
3241 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3242 {
3243         struct net *net = sock_net(skb->sk);
3244         struct ifaddrmsg *ifm;
3245         struct nlattr *tb[IFA_MAX+1];
3246         struct in6_addr *pfx;
3247         struct inet6_ifaddr *ifa;
3248         struct net_device *dev;
3249         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3250         u8 ifa_flags;
3251         int err;
3252
3253         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3254         if (err < 0)
3255                 return err;
3256
3257         ifm = nlmsg_data(nlh);
3258         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3259         if (pfx == NULL)
3260                 return -EINVAL;
3261
3262         if (tb[IFA_CACHEINFO]) {
3263                 struct ifa_cacheinfo *ci;
3264
3265                 ci = nla_data(tb[IFA_CACHEINFO]);
3266                 valid_lft = ci->ifa_valid;
3267                 preferred_lft = ci->ifa_prefered;
3268         } else {
3269                 preferred_lft = INFINITY_LIFE_TIME;
3270                 valid_lft = INFINITY_LIFE_TIME;
3271         }
3272
3273         dev =  __dev_get_by_index(net, ifm->ifa_index);
3274         if (dev == NULL)
3275                 return -ENODEV;
3276
3277         /* We ignore other flags so far. */
3278         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3279
3280         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3281         if (ifa == NULL) {
3282                 /*
3283                  * It would be best to check for !NLM_F_CREATE here but
3284                  * userspace alreay relies on not having to provide this.
3285                  */
3286                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3287                                       ifm->ifa_prefixlen, ifa_flags,
3288                                       preferred_lft, valid_lft);
3289         }
3290
3291         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3292             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3293                 err = -EEXIST;
3294         else
3295                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3296
3297         in6_ifa_put(ifa);
3298
3299         return err;
3300 }
3301
3302 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3303                           u8 scope, int ifindex)
3304 {
3305         struct ifaddrmsg *ifm;
3306
3307         ifm = nlmsg_data(nlh);
3308         ifm->ifa_family = AF_INET6;
3309         ifm->ifa_prefixlen = prefixlen;
3310         ifm->ifa_flags = flags;
3311         ifm->ifa_scope = scope;
3312         ifm->ifa_index = ifindex;
3313 }
3314
3315 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3316                          unsigned long tstamp, u32 preferred, u32 valid)
3317 {
3318         struct ifa_cacheinfo ci;
3319
3320         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3321                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3322         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3323                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3324         ci.ifa_prefered = preferred;
3325         ci.ifa_valid = valid;
3326
3327         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3328 }
3329
3330 static inline int rt_scope(int ifa_scope)
3331 {
3332         if (ifa_scope & IFA_HOST)
3333                 return RT_SCOPE_HOST;
3334         else if (ifa_scope & IFA_LINK)
3335                 return RT_SCOPE_LINK;
3336         else if (ifa_scope & IFA_SITE)
3337                 return RT_SCOPE_SITE;
3338         else
3339                 return RT_SCOPE_UNIVERSE;
3340 }
3341
3342 static inline int inet6_ifaddr_msgsize(void)
3343 {
3344         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3345                + nla_total_size(16) /* IFA_ADDRESS */
3346                + nla_total_size(sizeof(struct ifa_cacheinfo));
3347 }
3348
3349 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3350                              u32 pid, u32 seq, int event, unsigned int flags)
3351 {
3352         struct nlmsghdr  *nlh;
3353         u32 preferred, valid;
3354
3355         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3356         if (nlh == NULL)
3357                 return -EMSGSIZE;
3358
3359         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3360                       ifa->idev->dev->ifindex);
3361
3362         if (!(ifa->flags&IFA_F_PERMANENT)) {
3363                 preferred = ifa->prefered_lft;
3364                 valid = ifa->valid_lft;
3365                 if (preferred != INFINITY_LIFE_TIME) {
3366                         long tval = (jiffies - ifa->tstamp)/HZ;
3367                         preferred -= tval;
3368                         if (valid != INFINITY_LIFE_TIME)
3369                                 valid -= tval;
3370                 }
3371         } else {
3372                 preferred = INFINITY_LIFE_TIME;
3373                 valid = INFINITY_LIFE_TIME;
3374         }
3375
3376         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3377             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3378                 nlmsg_cancel(skb, nlh);
3379                 return -EMSGSIZE;
3380         }
3381
3382         return nlmsg_end(skb, nlh);
3383 }
3384
3385 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3386                                 u32 pid, u32 seq, int event, u16 flags)
3387 {
3388         struct nlmsghdr  *nlh;
3389         u8 scope = RT_SCOPE_UNIVERSE;
3390         int ifindex = ifmca->idev->dev->ifindex;
3391
3392         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3393                 scope = RT_SCOPE_SITE;
3394
3395         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3396         if (nlh == NULL)
3397                 return -EMSGSIZE;
3398
3399         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3400         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3401             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3402                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3403                 nlmsg_cancel(skb, nlh);
3404                 return -EMSGSIZE;
3405         }
3406
3407         return nlmsg_end(skb, nlh);
3408 }
3409
3410 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3411                                 u32 pid, u32 seq, int event, unsigned int flags)
3412 {
3413         struct nlmsghdr  *nlh;
3414         u8 scope = RT_SCOPE_UNIVERSE;
3415         int ifindex = ifaca->aca_idev->dev->ifindex;
3416
3417         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3418                 scope = RT_SCOPE_SITE;
3419
3420         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3421         if (nlh == NULL)
3422                 return -EMSGSIZE;
3423
3424         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3425         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3426             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3427                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3428                 nlmsg_cancel(skb, nlh);
3429                 return -EMSGSIZE;
3430         }
3431
3432         return nlmsg_end(skb, nlh);
3433 }
3434
3435 enum addr_type_t
3436 {
3437         UNICAST_ADDR,
3438         MULTICAST_ADDR,
3439         ANYCAST_ADDR,
3440 };
3441
3442 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3443                            enum addr_type_t type)
3444 {
3445         int idx, ip_idx;
3446         int s_idx, s_ip_idx;
3447         int err = 1;
3448         struct net_device *dev;
3449         struct inet6_dev *idev = NULL;
3450         struct inet6_ifaddr *ifa;
3451         struct ifmcaddr6 *ifmca;
3452         struct ifacaddr6 *ifaca;
3453         struct net *net = sock_net(skb->sk);
3454
3455         s_idx = cb->args[0];
3456         s_ip_idx = ip_idx = cb->args[1];
3457
3458         idx = 0;
3459         for_each_netdev(net, dev) {
3460                 if (idx < s_idx)
3461                         goto cont;
3462                 if (idx > s_idx)
3463                         s_ip_idx = 0;
3464                 ip_idx = 0;
3465                 if ((idev = in6_dev_get(dev)) == NULL)
3466                         goto cont;
3467                 read_lock_bh(&idev->lock);
3468                 switch (type) {
3469                 case UNICAST_ADDR:
3470                         /* unicast address incl. temp addr */
3471                         for (ifa = idev->addr_list; ifa;
3472                              ifa = ifa->if_next, ip_idx++) {
3473                                 if (ip_idx < s_ip_idx)
3474                                         continue;
3475                                 err = inet6_fill_ifaddr(skb, ifa,
3476                                                         NETLINK_CB(cb->skb).pid,
3477                                                         cb->nlh->nlmsg_seq,
3478                                                         RTM_NEWADDR,
3479                                                         NLM_F_MULTI);
3480                         }
3481                         break;
3482                 case MULTICAST_ADDR:
3483                         /* multicast address */
3484                         for (ifmca = idev->mc_list; ifmca;
3485                              ifmca = ifmca->next, ip_idx++) {
3486                                 if (ip_idx < s_ip_idx)
3487                                         continue;
3488                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3489                                                           NETLINK_CB(cb->skb).pid,
3490                                                           cb->nlh->nlmsg_seq,
3491                                                           RTM_GETMULTICAST,
3492                                                           NLM_F_MULTI);
3493                         }
3494                         break;
3495                 case ANYCAST_ADDR:
3496                         /* anycast address */
3497                         for (ifaca = idev->ac_list; ifaca;
3498                              ifaca = ifaca->aca_next, ip_idx++) {
3499                                 if (ip_idx < s_ip_idx)
3500                                         continue;
3501                                 err = inet6_fill_ifacaddr(skb, ifaca,
3502                                                           NETLINK_CB(cb->skb).pid,
3503                                                           cb->nlh->nlmsg_seq,
3504                                                           RTM_GETANYCAST,
3505                                                           NLM_F_MULTI);
3506                         }
3507                         break;
3508                 default:
3509                         break;
3510                 }
3511                 read_unlock_bh(&idev->lock);
3512                 in6_dev_put(idev);
3513
3514                 if (err <= 0)
3515                         break;
3516 cont:
3517                 idx++;
3518         }
3519         cb->args[0] = idx;
3520         cb->args[1] = ip_idx;
3521         return skb->len;
3522 }
3523
3524 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3525 {
3526         enum addr_type_t type = UNICAST_ADDR;
3527
3528         return inet6_dump_addr(skb, cb, type);
3529 }
3530
3531 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3532 {
3533         enum addr_type_t type = MULTICAST_ADDR;
3534
3535         return inet6_dump_addr(skb, cb, type);
3536 }
3537
3538
3539 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3540 {
3541         enum addr_type_t type = ANYCAST_ADDR;
3542
3543         return inet6_dump_addr(skb, cb, type);
3544 }
3545
3546 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3547                              void *arg)
3548 {
3549         struct net *net = sock_net(in_skb->sk);
3550         struct ifaddrmsg *ifm;
3551         struct nlattr *tb[IFA_MAX+1];
3552         struct in6_addr *addr = NULL;
3553         struct net_device *dev = NULL;
3554         struct inet6_ifaddr *ifa;
3555         struct sk_buff *skb;
3556         int err;
3557
3558         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3559         if (err < 0)
3560                 goto errout;
3561
3562         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3563         if (addr == NULL) {
3564                 err = -EINVAL;
3565                 goto errout;
3566         }
3567
3568         ifm = nlmsg_data(nlh);
3569         if (ifm->ifa_index)
3570                 dev = __dev_get_by_index(net, ifm->ifa_index);
3571
3572         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3573                 err = -EADDRNOTAVAIL;
3574                 goto errout;
3575         }
3576
3577         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3578                 err = -ENOBUFS;
3579                 goto errout_ifa;
3580         }
3581
3582         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3583                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3584         if (err < 0) {
3585                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3586                 WARN_ON(err == -EMSGSIZE);
3587                 kfree_skb(skb);
3588                 goto errout_ifa;
3589         }
3590         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3591 errout_ifa:
3592         in6_ifa_put(ifa);
3593 errout:
3594         return err;
3595 }
3596
3597 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3598 {
3599         struct sk_buff *skb;
3600         struct net *net = dev_net(ifa->idev->dev);
3601         int err = -ENOBUFS;
3602
3603         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3604         if (skb == NULL)
3605                 goto errout;
3606
3607         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3608         if (err < 0) {
3609                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3610                 WARN_ON(err == -EMSGSIZE);
3611                 kfree_skb(skb);
3612                 goto errout;
3613         }
3614         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3615 errout:
3616         if (err < 0)
3617                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3618 }
3619
3620 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3621                                 __s32 *array, int bytes)
3622 {
3623         BUG_ON(bytes < (DEVCONF_MAX * 4));
3624
3625         memset(array, 0, bytes);
3626         array[DEVCONF_FORWARDING] = cnf->forwarding;
3627         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3628         array[DEVCONF_MTU6] = cnf->mtu6;
3629         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3630         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3631         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3632         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3633         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3634         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3635         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3636         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3637 #ifdef CONFIG_IPV6_PRIVACY
3638         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3639         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3640         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3641         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3642         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3643 #endif
3644         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3645         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3646         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3647 #ifdef CONFIG_IPV6_ROUTER_PREF
3648         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3649         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3650 #ifdef CONFIG_IPV6_ROUTE_INFO
3651         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3652 #endif
3653 #endif
3654         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3655         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3656 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3657         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3658 #endif
3659 #ifdef CONFIG_IPV6_MROUTE
3660         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3661 #endif
3662         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3663 }
3664
3665 static inline size_t inet6_if_nlmsg_size(void)
3666 {
3667         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3668                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3669                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3670                + nla_total_size(4) /* IFLA_MTU */
3671                + nla_total_size(4) /* IFLA_LINK */
3672                + nla_total_size( /* IFLA_PROTINFO */
3673                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3674                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3675                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3676                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3677                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3678                  );
3679 }
3680
3681 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3682                                       int bytes)
3683 {
3684         int i;
3685         int pad = bytes - sizeof(u64) * items;
3686         BUG_ON(pad < 0);
3687
3688         /* Use put_unaligned() because stats may not be aligned for u64. */
3689         put_unaligned(items, &stats[0]);
3690         for (i = 1; i < items; i++)
3691                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3692
3693         memset(&stats[items], 0, pad);
3694 }
3695
3696 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3697                              int bytes)
3698 {
3699         switch(attrtype) {
3700         case IFLA_INET6_STATS:
3701                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3702                 break;
3703         case IFLA_INET6_ICMP6STATS:
3704                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3705                 break;
3706         }
3707 }
3708
3709 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3710                              u32 pid, u32 seq, int event, unsigned int flags)
3711 {
3712         struct net_device *dev = idev->dev;
3713         struct nlattr *nla;
3714         struct ifinfomsg *hdr;
3715         struct nlmsghdr *nlh;
3716         void *protoinfo;
3717         struct ifla_cacheinfo ci;
3718
3719         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3720         if (nlh == NULL)
3721                 return -EMSGSIZE;
3722
3723         hdr = nlmsg_data(nlh);
3724         hdr->ifi_family = AF_INET6;
3725         hdr->__ifi_pad = 0;
3726         hdr->ifi_type = dev->type;
3727         hdr->ifi_index = dev->ifindex;
3728         hdr->ifi_flags = dev_get_flags(dev);
3729         hdr->ifi_change = 0;
3730
3731         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3732
3733         if (dev->addr_len)
3734                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3735
3736         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3737         if (dev->ifindex != dev->iflink)
3738                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3739
3740         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3741         if (protoinfo == NULL)
3742                 goto nla_put_failure;
3743
3744         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3745
3746         ci.max_reasm_len = IPV6_MAXPLEN;
3747         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3748                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3749         ci.reachable_time = idev->nd_parms->reachable_time;
3750         ci.retrans_time = idev->nd_parms->retrans_time;
3751         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3752
3753         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3754         if (nla == NULL)
3755                 goto nla_put_failure;
3756         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3757
3758         /* XXX - MC not implemented */
3759
3760         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3761         if (nla == NULL)
3762                 goto nla_put_failure;
3763         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3764
3765         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3766         if (nla == NULL)
3767                 goto nla_put_failure;
3768         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3769
3770         nla_nest_end(skb, protoinfo);
3771         return nlmsg_end(skb, nlh);
3772
3773 nla_put_failure:
3774         nlmsg_cancel(skb, nlh);
3775         return -EMSGSIZE;
3776 }
3777
3778 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3779 {
3780         struct net *net = sock_net(skb->sk);
3781         int idx, err;
3782         int s_idx = cb->args[0];
3783         struct net_device *dev;
3784         struct inet6_dev *idev;
3785
3786         read_lock(&dev_base_lock);
3787         idx = 0;
3788         for_each_netdev(net, dev) {
3789                 if (idx < s_idx)
3790                         goto cont;
3791                 if ((idev = in6_dev_get(dev)) == NULL)
3792                         goto cont;
3793                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3794                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3795                 in6_dev_put(idev);
3796                 if (err <= 0)
3797                         break;
3798 cont:
3799                 idx++;
3800         }
3801         read_unlock(&dev_base_lock);
3802         cb->args[0] = idx;
3803
3804         return skb->len;
3805 }
3806
3807 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3808 {
3809         struct sk_buff *skb;
3810         struct net *net = dev_net(idev->dev);
3811         int err = -ENOBUFS;
3812
3813         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3814         if (skb == NULL)
3815                 goto errout;
3816
3817         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3818         if (err < 0) {
3819                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3820                 WARN_ON(err == -EMSGSIZE);
3821                 kfree_skb(skb);
3822                 goto errout;
3823         }
3824         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3825 errout:
3826         if (err < 0)
3827                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3828 }
3829
3830 static inline size_t inet6_prefix_nlmsg_size(void)
3831 {
3832         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3833                + nla_total_size(sizeof(struct in6_addr))
3834                + nla_total_size(sizeof(struct prefix_cacheinfo));
3835 }
3836
3837 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3838                              struct prefix_info *pinfo, u32 pid, u32 seq,
3839                              int event, unsigned int flags)
3840 {
3841         struct prefixmsg *pmsg;
3842         struct nlmsghdr *nlh;
3843         struct prefix_cacheinfo ci;
3844
3845         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3846         if (nlh == NULL)
3847                 return -EMSGSIZE;
3848
3849         pmsg = nlmsg_data(nlh);
3850         pmsg->prefix_family = AF_INET6;
3851         pmsg->prefix_pad1 = 0;
3852         pmsg->prefix_pad2 = 0;
3853         pmsg->prefix_ifindex = idev->dev->ifindex;
3854         pmsg->prefix_len = pinfo->prefix_len;
3855         pmsg->prefix_type = pinfo->type;
3856         pmsg->prefix_pad3 = 0;
3857         pmsg->prefix_flags = 0;
3858         if (pinfo->onlink)
3859                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3860         if (pinfo->autoconf)
3861                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3862
3863         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3864
3865         ci.preferred_time = ntohl(pinfo->prefered);
3866         ci.valid_time = ntohl(pinfo->valid);
3867         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3868
3869         return nlmsg_end(skb, nlh);
3870
3871 nla_put_failure:
3872         nlmsg_cancel(skb, nlh);
3873         return -EMSGSIZE;
3874 }
3875
3876 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3877                          struct prefix_info *pinfo)
3878 {
3879         struct sk_buff *skb;
3880         struct net *net = dev_net(idev->dev);
3881         int err = -ENOBUFS;
3882
3883         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3884         if (skb == NULL)
3885                 goto errout;
3886
3887         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3888         if (err < 0) {
3889                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3890                 WARN_ON(err == -EMSGSIZE);
3891                 kfree_skb(skb);
3892                 goto errout;
3893         }
3894         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3895 errout:
3896         if (err < 0)
3897                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3898 }
3899
3900 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3901 {
3902         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3903
3904         switch (event) {
3905         case RTM_NEWADDR:
3906                 /*
3907                  * If the address was optimistic
3908                  * we inserted the route at the start of
3909                  * our DAD process, so we don't need
3910                  * to do it again
3911                  */
3912                 if (!(ifp->rt->rt6i_node))
3913                         ip6_ins_rt(ifp->rt);
3914                 if (ifp->idev->cnf.forwarding)
3915                         addrconf_join_anycast(ifp);
3916                 break;
3917         case RTM_DELADDR:
3918                 if (ifp->idev->cnf.forwarding)
3919                         addrconf_leave_anycast(ifp);
3920                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3921                 dst_hold(&ifp->rt->u.dst);
3922                 if (ip6_del_rt(ifp->rt))
3923                         dst_free(&ifp->rt->u.dst);
3924                 break;
3925         }
3926 }
3927
3928 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3929 {
3930         rcu_read_lock_bh();
3931         if (likely(ifp->idev->dead == 0))
3932                 __ipv6_ifa_notify(event, ifp);
3933         rcu_read_unlock_bh();
3934 }
3935
3936 #ifdef CONFIG_SYSCTL
3937
3938 static
3939 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3940                            void __user *buffer, size_t *lenp, loff_t *ppos)
3941 {
3942         int *valp = ctl->data;
3943         int val = *valp;
3944         int ret;
3945
3946         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3947
3948         if (write)
3949                 addrconf_fixup_forwarding(ctl, valp, val);
3950         return ret;
3951 }
3952
3953 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3954                                             int __user *name, int nlen,
3955                                             void __user *oldval,
3956                                             size_t __user *oldlenp,
3957                                             void __user *newval, size_t newlen)
3958 {
3959         int *valp = table->data;
3960         int val = *valp;
3961         int new;
3962
3963         if (!newval || !newlen)
3964                 return 0;
3965         if (newlen != sizeof(int))
3966                 return -EINVAL;
3967         if (get_user(new, (int __user *)newval))
3968                 return -EFAULT;
3969         if (new == *valp)
3970                 return 0;
3971         if (oldval && oldlenp) {
3972                 size_t len;
3973                 if (get_user(len, oldlenp))
3974                         return -EFAULT;
3975                 if (len) {
3976                         if (len > table->maxlen)
3977                                 len = table->maxlen;
3978                         if (copy_to_user(oldval, valp, len))
3979                                 return -EFAULT;
3980                         if (put_user(len, oldlenp))
3981                                 return -EFAULT;
3982                 }
3983         }
3984
3985         *valp = new;
3986         addrconf_fixup_forwarding(table, valp, val);
3987         return 1;
3988 }
3989
3990 static struct addrconf_sysctl_table
3991 {
3992         struct ctl_table_header *sysctl_header;
3993         ctl_table addrconf_vars[DEVCONF_MAX+1];
3994         char *dev_name;
3995 } addrconf_sysctl __read_mostly = {
3996         .sysctl_header = NULL,
3997         .addrconf_vars = {
3998                 {
3999                         .ctl_name       =       NET_IPV6_FORWARDING,
4000                         .procname       =       "forwarding",
4001                         .data           =       &ipv6_devconf.forwarding,
4002                         .maxlen         =       sizeof(int),
4003                         .mode           =       0644,
4004                         .proc_handler   =       &addrconf_sysctl_forward,
4005                         .strategy       =       &addrconf_sysctl_forward_strategy,
4006                 },
4007                 {
4008                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4009                         .procname       =       "hop_limit",
4010                         .data           =       &ipv6_devconf.hop_limit,
4011                         .maxlen         =       sizeof(int),
4012                         .mode           =       0644,
4013                         .proc_handler   =       proc_dointvec,
4014                 },
4015                 {
4016                         .ctl_name       =       NET_IPV6_MTU,
4017                         .procname       =       "mtu",
4018                         .data           =       &ipv6_devconf.mtu6,
4019                         .maxlen         =       sizeof(int),
4020                         .mode           =       0644,
4021                         .proc_handler   =       &proc_dointvec,
4022                 },
4023                 {
4024                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4025                         .procname       =       "accept_ra",
4026                         .data           =       &ipv6_devconf.accept_ra,
4027                         .maxlen         =       sizeof(int),
4028                         .mode           =       0644,
4029                         .proc_handler   =       &proc_dointvec,
4030                 },
4031                 {
4032                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4033                         .procname       =       "accept_redirects",
4034                         .data           =       &ipv6_devconf.accept_redirects,
4035                         .maxlen         =       sizeof(int),
4036                         .mode           =       0644,
4037                         .proc_handler   =       &proc_dointvec,
4038                 },
4039                 {
4040                         .ctl_name       =       NET_IPV6_AUTOCONF,
4041                         .procname       =       "autoconf",
4042                         .data           =       &ipv6_devconf.autoconf,
4043                         .maxlen         =       sizeof(int),
4044                         .mode           =       0644,
4045                         .proc_handler   =       &proc_dointvec,
4046                 },
4047                 {
4048                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4049                         .procname       =       "dad_transmits",
4050                         .data           =       &ipv6_devconf.dad_transmits,
4051                         .maxlen         =       sizeof(int),
4052                         .mode           =       0644,
4053                         .proc_handler   =       &proc_dointvec,
4054                 },
4055                 {
4056                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4057                         .procname       =       "router_solicitations",
4058                         .data           =       &ipv6_devconf.rtr_solicits,
4059                         .maxlen         =       sizeof(int),
4060                         .mode           =       0644,
4061                         .proc_handler   =       &proc_dointvec,
4062                 },
4063                 {
4064                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4065                         .procname       =       "router_solicitation_interval",
4066                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4067                         .maxlen         =       sizeof(int),
4068                         .mode           =       0644,
4069                         .proc_handler   =       &proc_dointvec_jiffies,
4070                         .strategy       =       &sysctl_jiffies,
4071                 },
4072                 {
4073                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4074                         .procname       =       "router_solicitation_delay",
4075                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4076                         .maxlen         =       sizeof(int),
4077                         .mode           =       0644,
4078                         .proc_handler   =       &proc_dointvec_jiffies,
4079                         .strategy       =       &sysctl_jiffies,
4080                 },
4081                 {
4082                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4083                         .procname       =       "force_mld_version",
4084                         .data           =       &ipv6_devconf.force_mld_version,
4085                         .maxlen         =       sizeof(int),
4086                         .mode           =       0644,
4087                         .proc_handler   =       &proc_dointvec,
4088                 },
4089 #ifdef CONFIG_IPV6_PRIVACY
4090                 {
4091                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4092                         .procname       =       "use_tempaddr",
4093                         .data           =       &ipv6_devconf.use_tempaddr,
4094                         .maxlen         =       sizeof(int),
4095                         .mode           =       0644,
4096                         .proc_handler   =       &proc_dointvec,
4097                 },
4098                 {
4099                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4100                         .procname       =       "temp_valid_lft",
4101                         .data           =       &ipv6_devconf.temp_valid_lft,
4102                         .maxlen         =       sizeof(int),
4103                         .mode           =       0644,
4104                         .proc_handler   =       &proc_dointvec,
4105                 },
4106                 {
4107                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4108                         .procname       =       "temp_prefered_lft",
4109                         .data           =       &ipv6_devconf.temp_prefered_lft,
4110                         .maxlen         =       sizeof(int),
4111                         .mode           =       0644,
4112                         .proc_handler   =       &proc_dointvec,
4113                 },
4114                 {
4115                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4116                         .procname       =       "regen_max_retry",
4117                         .data           =       &ipv6_devconf.regen_max_retry,
4118                         .maxlen         =       sizeof(int),
4119                         .mode           =       0644,
4120                         .proc_handler   =       &proc_dointvec,
4121                 },
4122                 {
4123                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4124                         .procname       =       "max_desync_factor",
4125                         .data           =       &ipv6_devconf.max_desync_factor,
4126                         .maxlen         =       sizeof(int),
4127                         .mode           =       0644,
4128                         .proc_handler   =       &proc_dointvec,
4129                 },
4130 #endif
4131                 {
4132                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4133                         .procname       =       "max_addresses",
4134                         .data           =       &ipv6_devconf.max_addresses,
4135                         .maxlen         =       sizeof(int),
4136                         .mode           =       0644,
4137                         .proc_handler   =       &proc_dointvec,
4138                 },
4139                 {
4140                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4141                         .procname       =       "accept_ra_defrtr",
4142                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4143                         .maxlen         =       sizeof(int),
4144                         .mode           =       0644,
4145                         .proc_handler   =       &proc_dointvec,
4146                 },
4147                 {
4148                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4149                         .procname       =       "accept_ra_pinfo",
4150                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4151                         .maxlen         =       sizeof(int),
4152                         .mode           =       0644,
4153                         .proc_handler   =       &proc_dointvec,
4154                 },
4155 #ifdef CONFIG_IPV6_ROUTER_PREF
4156                 {
4157                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4158                         .procname       =       "accept_ra_rtr_pref",
4159                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4160                         .maxlen         =       sizeof(int),
4161                         .mode           =       0644,
4162                         .proc_handler   =       &proc_dointvec,
4163                 },
4164                 {
4165                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4166                         .procname       =       "router_probe_interval",
4167                         .data           =       &ipv6_devconf.rtr_probe_interval,
4168                         .maxlen         =       sizeof(int),
4169                         .mode           =       0644,
4170                         .proc_handler   =       &proc_dointvec_jiffies,
4171                         .strategy       =       &sysctl_jiffies,
4172                 },
4173 #ifdef CONFIG_IPV6_ROUTE_INFO
4174                 {
4175                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4176                         .procname       =       "accept_ra_rt_info_max_plen",
4177                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4178                         .maxlen         =       sizeof(int),
4179                         .mode           =       0644,
4180                         .proc_handler   =       &proc_dointvec,
4181                 },
4182 #endif
4183 #endif
4184                 {
4185                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4186                         .procname       =       "proxy_ndp",
4187                         .data           =       &ipv6_devconf.proxy_ndp,
4188                         .maxlen         =       sizeof(int),
4189                         .mode           =       0644,
4190                         .proc_handler   =       &proc_dointvec,
4191                 },
4192                 {
4193                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4194                         .procname       =       "accept_source_route",
4195                         .data           =       &ipv6_devconf.accept_source_route,
4196                         .maxlen         =       sizeof(int),
4197                         .mode           =       0644,
4198                         .proc_handler   =       &proc_dointvec,
4199                 },
4200 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4201                 {
4202                         .ctl_name       =       CTL_UNNUMBERED,
4203                         .procname       =       "optimistic_dad",
4204                         .data           =       &ipv6_devconf.optimistic_dad,
4205                         .maxlen         =       sizeof(int),
4206                         .mode           =       0644,
4207                         .proc_handler   =       &proc_dointvec,
4208
4209                 },
4210 #endif
4211 #ifdef CONFIG_IPV6_MROUTE
4212                 {
4213                         .ctl_name       =       CTL_UNNUMBERED,
4214                         .procname       =       "mc_forwarding",
4215                         .data           =       &ipv6_devconf.mc_forwarding,
4216                         .maxlen         =       sizeof(int),
4217                         .mode           =       0644,
4218                         .proc_handler   =       &proc_dointvec,
4219                 },
4220 #endif
4221                 {
4222                         .ctl_name       =       CTL_UNNUMBERED,
4223                         .procname       =       "disable_ipv6",
4224                         .data           =       &ipv6_devconf.disable_ipv6,
4225                         .maxlen         =       sizeof(int),
4226                         .mode           =       0644,
4227                         .proc_handler   =       &proc_dointvec,
4228                 },
4229                 {
4230                         .ctl_name       =       0,      /* sentinel */
4231                 }
4232         },
4233 };
4234
4235 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4236                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4237 {
4238         int i;
4239         struct addrconf_sysctl_table *t;
4240
4241 #define ADDRCONF_CTL_PATH_DEV   3
4242
4243         struct ctl_path addrconf_ctl_path[] = {
4244                 { .procname = "net", .ctl_name = CTL_NET, },
4245                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4246                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4247                 { /* to be set */ },
4248                 { },
4249         };
4250
4251
4252         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4253         if (t == NULL)
4254                 goto out;
4255
4256         for (i=0; t->addrconf_vars[i].data; i++) {
4257                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4258                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4259                 t->addrconf_vars[i].extra2 = net;
4260         }
4261
4262         /*
4263          * Make a copy of dev_name, because '.procname' is regarded as const
4264          * by sysctl and we wouldn't want anyone to change it under our feet
4265          * (see SIOCSIFNAME).
4266          */
4267         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4268         if (!t->dev_name)
4269                 goto free;
4270
4271         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4272         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4273
4274         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4275                         t->addrconf_vars);
4276         if (t->sysctl_header == NULL)
4277                 goto free_procname;
4278
4279         p->sysctl = t;
4280         return 0;
4281
4282 free_procname:
4283         kfree(t->dev_name);
4284 free:
4285         kfree(t);
4286 out:
4287         return -ENOBUFS;
4288 }
4289
4290 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4291 {
4292         struct addrconf_sysctl_table *t;
4293
4294         if (p->sysctl == NULL)
4295                 return;
4296
4297         t = p->sysctl;
4298         p->sysctl = NULL;
4299         unregister_sysctl_table(t->sysctl_header);
4300         kfree(t->dev_name);
4301         kfree(t);
4302 }
4303
4304 static void addrconf_sysctl_register(struct inet6_dev *idev)
4305 {
4306         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4307                               NET_IPV6_NEIGH, "ipv6",
4308                               &ndisc_ifinfo_sysctl_change,
4309                               ndisc_ifinfo_sysctl_strategy);
4310         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4311                         idev->dev->ifindex, idev, &idev->cnf);
4312 }
4313
4314 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4315 {
4316         __addrconf_sysctl_unregister(&idev->cnf);
4317         neigh_sysctl_unregister(idev->nd_parms);
4318 }
4319
4320
4321 #endif
4322
4323 static int addrconf_init_net(struct net *net)
4324 {
4325         int err;
4326         struct ipv6_devconf *all, *dflt;
4327
4328         err = -ENOMEM;
4329         all = &ipv6_devconf;
4330         dflt = &ipv6_devconf_dflt;
4331
4332         if (net != &init_net) {
4333                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4334                 if (all == NULL)
4335                         goto err_alloc_all;
4336
4337                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4338                 if (dflt == NULL)
4339                         goto err_alloc_dflt;
4340         }
4341
4342         net->ipv6.devconf_all = all;
4343         net->ipv6.devconf_dflt = dflt;
4344
4345 #ifdef CONFIG_SYSCTL
4346         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4347                         NULL, all);
4348         if (err < 0)
4349                 goto err_reg_all;
4350
4351         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4352                         NULL, dflt);
4353         if (err < 0)
4354                 goto err_reg_dflt;
4355 #endif
4356         return 0;
4357
4358 #ifdef CONFIG_SYSCTL
4359 err_reg_dflt:
4360         __addrconf_sysctl_unregister(all);
4361 err_reg_all:
4362         kfree(dflt);
4363 #endif
4364 err_alloc_dflt:
4365         kfree(all);
4366 err_alloc_all:
4367         return err;
4368 }
4369
4370 static void addrconf_exit_net(struct net *net)
4371 {
4372 #ifdef CONFIG_SYSCTL
4373         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4374         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4375 #endif
4376         if (net != &init_net) {
4377                 kfree(net->ipv6.devconf_dflt);
4378                 kfree(net->ipv6.devconf_all);
4379         }
4380 }
4381
4382 static struct pernet_operations addrconf_ops = {
4383         .init = addrconf_init_net,
4384         .exit = addrconf_exit_net,
4385 };
4386
4387 /*
4388  *      Device notifier
4389  */
4390
4391 int register_inet6addr_notifier(struct notifier_block *nb)
4392 {
4393         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4394 }
4395
4396 EXPORT_SYMBOL(register_inet6addr_notifier);
4397
4398 int unregister_inet6addr_notifier(struct notifier_block *nb)
4399 {
4400         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4401 }
4402
4403 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4404
4405 static void addrconf_net_exit(struct net *net)
4406 {
4407         struct net_device *dev;
4408
4409         rtnl_lock();
4410         /* clean dev list */
4411         for_each_netdev(net, dev) {
4412                 if (__in6_dev_get(dev) == NULL)
4413                         continue;
4414                 addrconf_ifdown(dev, 1);
4415         }
4416         addrconf_ifdown(net->loopback_dev, 2);
4417         rtnl_unlock();
4418 }
4419
4420 static struct pernet_operations addrconf_net_ops = {
4421         .exit = addrconf_net_exit,
4422 };
4423
4424 /*
4425  *      Init / cleanup code
4426  */
4427
4428 int __init addrconf_init(void)
4429 {
4430         int err;
4431
4432         if ((err = ipv6_addr_label_init()) < 0) {
4433                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4434                         err);
4435                 return err;
4436         }
4437
4438         register_pernet_subsys(&addrconf_ops);
4439
4440         /* The addrconf netdev notifier requires that loopback_dev
4441          * has it's ipv6 private information allocated and setup
4442          * before it can bring up and give link-local addresses
4443          * to other devices which are up.
4444          *
4445          * Unfortunately, loopback_dev is not necessarily the first
4446          * entry in the global dev_base list of net devices.  In fact,
4447          * it is likely to be the very last entry on that list.
4448          * So this causes the notifier registry below to try and
4449          * give link-local addresses to all devices besides loopback_dev
4450          * first, then loopback_dev, which cases all the non-loopback_dev
4451          * devices to fail to get a link-local address.
4452          *
4453          * So, as a temporary fix, allocate the ipv6 structure for
4454          * loopback_dev first by hand.
4455          * Longer term, all of the dependencies ipv6 has upon the loopback
4456          * device and it being up should be removed.
4457          */
4458         rtnl_lock();
4459         if (!ipv6_add_dev(init_net.loopback_dev))
4460                 err = -ENOMEM;
4461         rtnl_unlock();
4462         if (err)
4463                 goto errlo;
4464
4465         err = register_pernet_device(&addrconf_net_ops);
4466         if (err)
4467                 return err;
4468
4469         register_netdevice_notifier(&ipv6_dev_notf);
4470
4471         addrconf_verify(0);
4472
4473         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4474         if (err < 0)
4475                 goto errout;
4476
4477         /* Only the first call to __rtnl_register can fail */
4478         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4479         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4480         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4481         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4482         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4483
4484         ipv6_addr_label_rtnl_register();
4485
4486         return 0;
4487 errout:
4488         unregister_netdevice_notifier(&ipv6_dev_notf);
4489 errlo:
4490         unregister_pernet_subsys(&addrconf_ops);
4491
4492         return err;
4493 }
4494
4495 void addrconf_cleanup(void)
4496 {
4497         struct inet6_ifaddr *ifa;
4498         int i;
4499
4500         unregister_netdevice_notifier(&ipv6_dev_notf);
4501         unregister_pernet_device(&addrconf_net_ops);
4502
4503         unregister_pernet_subsys(&addrconf_ops);
4504
4505         rtnl_lock();
4506
4507         /*
4508          *      Check hash table.
4509          */
4510         write_lock_bh(&addrconf_hash_lock);
4511         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4512                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4513                         struct inet6_ifaddr *bifa;
4514
4515                         bifa = ifa;
4516                         ifa = ifa->lst_next;
4517                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4518                         /* Do not free it; something is wrong.
4519                            Now we can investigate it with debugger.
4520                          */
4521                 }
4522         }
4523         write_unlock_bh(&addrconf_hash_lock);
4524
4525         del_timer(&addr_chk_timer);
4526         rtnl_unlock();
4527
4528         unregister_pernet_subsys(&addrconf_net_ops);
4529 }