net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static void addrconf_type_change(struct net_device *dev,
 141                                  unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         INIT_LIST_HEAD(&ndev->tempaddr_list);
 405         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 406         if ((dev->flags&IFF_LOOPBACK) ||
 407             dev->type == ARPHRD_TUNNEL ||
 408             dev->type == ARPHRD_TUNNEL6 ||
 409             dev->type == ARPHRD_SIT ||
 410             dev->type == ARPHRD_NONE) {
 411                 printk(KERN_INFO
 412                        "%s: Disabled Privacy Extensions\n",
 413                        dev->name);
 414                 ndev->cnf.use_tempaddr = -1;
 415         } else {
 416                 in6_dev_hold(ndev);
 417                 ipv6_regen_rndid((unsigned long) ndev);
 418         }
 419 #endif
 420
 421         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 422                 ndev->if_flags |= IF_READY;
 423
 424         ipv6_mc_init_dev(ndev);
 425         ndev->tstamp = jiffies;
 426         addrconf_sysctl_register(ndev);
 427         /* protected by rtnl_lock */
 428         rcu_assign_pointer(dev->ip6_ptr, ndev);
 429
 430         /* Join all-node multicast group */
 431         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 432
 433         return ndev;
 434 }
 435
 436 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 437 {
 438         struct inet6_dev *idev;
 439
 440         ASSERT_RTNL();
 441
 442         if ((idev = __in6_dev_get(dev)) == NULL) {
 443                 if ((idev = ipv6_add_dev(dev)) == NULL)
 444                         return NULL;
 445         }
 446
 447         if (dev->flags&IFF_UP)
 448                 ipv6_mc_up(idev);
 449         return idev;
 450 }
 451
 452 #ifdef CONFIG_SYSCTL
 453 static void dev_forward_change(struct inet6_dev *idev)
 454 {
 455         struct net_device *dev;
 456         struct inet6_ifaddr *ifa;
 457
 458         if (!idev)
 459                 return;
 460         dev = idev->dev;
 461         if (idev->cnf.forwarding)
 462                 dev_disable_lro(dev);
 463         if (dev && (dev->flags & IFF_MULTICAST)) {
 464                 if (idev->cnf.forwarding)
 465                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 466                 else
 467                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 468         }
 469         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 470                 if (ifa->flags&IFA_F_TENTATIVE)
 471                         continue;
 472                 if (idev->cnf.forwarding)
 473                         addrconf_join_anycast(ifa);
 474                 else
 475                         addrconf_leave_anycast(ifa);
 476         }
 477 }
 478
 479
 480 static void addrconf_forward_change(struct net *net, __s32 newf)
 481 {
 482         struct net_device *dev;
 483         struct inet6_dev *idev;
 484
 485         rcu_read_lock();
 486         for_each_netdev_rcu(net, dev) {
 487                 idev = __in6_dev_get(dev);
 488                 if (idev) {
 489                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 490                         idev->cnf.forwarding = newf;
 491                         if (changed)
 492                                 dev_forward_change(idev);
 493                 }
 494         }
 495         rcu_read_unlock();
 496 }
 497
 498 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 499 {
 500         struct net *net;
 501
 502         net = (struct net *)table->extra2;
 503         if (p == &net->ipv6.devconf_dflt->forwarding)
 504                 return 0;
 505
 506         if (!rtnl_trylock()) {
 507                 /* Restore the original values before restarting */
 508                 *p = old;
 509                 return restart_syscall();
 510         }
 511
 512         if (p == &net->ipv6.devconf_all->forwarding) {
 513                 __s32 newf = net->ipv6.devconf_all->forwarding;
 514                 net->ipv6.devconf_dflt->forwarding = newf;
 515                 addrconf_forward_change(net, newf);
 516         } else if ((!*p) ^ (!old))
 517                 dev_forward_change((struct inet6_dev *)table->extra1);
 518         rtnl_unlock();
 519
 520         if (*p)
 521                 rt6_purge_dflt_routers(net);
 522         return 1;
 523 }
 524 #endif
 525
 526 /* Nobody refers to this ifaddr, destroy it */
 527
 528 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 529 {
 530         WARN_ON(ifp->if_next != NULL);
 531         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 532
 533 #ifdef NET_REFCNT_DEBUG
 534         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 535 #endif
 536
 537         in6_dev_put(ifp->idev);
 538
 539         if (del_timer(&ifp->timer))
 540                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 541
 542         if (!ifp->dead) {
 543                 printk("Freeing alive inet6 address %p\n", ifp);
 544                 return;
 545         }
 546         dst_release(&ifp->rt->u.dst);
 547
 548         kfree(ifp);
 549 }
 550
 551 static void
 552 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 553 {
 554         struct inet6_ifaddr *ifa, **ifap;
 555         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 556
 557         /*
 558          * Each device address list is sorted in order of scope -
 559          * global before linklocal.
 560          */
 561         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 562              ifap = &ifa->if_next) {
 563                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 564                         break;
 565         }
 566
 567         ifp->if_next = *ifap;
 568         *ifap = ifp;
 569 }
 570
 571 /*
 572  *      Hash function taken from net_alias.c
 573  */
 574 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 575 {
 576         __u32 word;
 577
 578         /*
 579          * We perform the hash function over the last 64 bits of the address
 580          * This will include the IEEE address token on links that support it.
 581          */
 582
 583         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 584         word ^= (word >> 16);
 585         word ^= (word >> 8);
 586
 587         return ((word ^ (word >> 4)) & 0x0f);
 588 }
 589
 590 /* On success it returns ifp with increased reference count */
 591
 592 static struct inet6_ifaddr *
 593 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 594               int scope, u32 flags)
 595 {
 596         struct inet6_ifaddr *ifa = NULL;
 597         struct rt6_info *rt;
 598         int hash;
 599         int err = 0;
 600         int addr_type = ipv6_addr_type(addr);
 601
 602         if (addr_type == IPV6_ADDR_ANY ||
 603             addr_type & IPV6_ADDR_MULTICAST ||
 604             (!(idev->dev->flags & IFF_LOOPBACK) &&
 605              addr_type & IPV6_ADDR_LOOPBACK))
 606                 return ERR_PTR(-EADDRNOTAVAIL);
 607
 608         rcu_read_lock_bh();
 609         if (idev->dead) {
 610                 err = -ENODEV;                  /*XXX*/
 611                 goto out2;
 612         }
 613
 614         if (idev->cnf.disable_ipv6) {
 615                 err = -EACCES;
 616                 goto out2;
 617         }
 618
 619         write_lock(&addrconf_hash_lock);
 620
 621         /* Ignore adding duplicate addresses on an interface */
 622         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 623                 ADBG(("ipv6_add_addr: already assigned\n"));
 624                 err = -EEXIST;
 625                 goto out;
 626         }
 627
 628         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 629
 630         if (ifa == NULL) {
 631                 ADBG(("ipv6_add_addr: malloc failed\n"));
 632                 err = -ENOBUFS;
 633                 goto out;
 634         }
 635
 636         rt = addrconf_dst_alloc(idev, addr, 0);
 637         if (IS_ERR(rt)) {
 638                 err = PTR_ERR(rt);
 639                 goto out;
 640         }
 641
 642         ipv6_addr_copy(&ifa->addr, addr);
 643
 644         spin_lock_init(&ifa->lock);
 645         init_timer(&ifa->timer);
 646         INIT_HLIST_NODE(&ifa->addr_lst);
 647         ifa->timer.data = (unsigned long) ifa;
 648         ifa->scope = scope;
 649         ifa->prefix_len = pfxlen;
 650         ifa->flags = flags | IFA_F_TENTATIVE;
 651         ifa->cstamp = ifa->tstamp = jiffies;
 652
 653         ifa->rt = rt;
 654
 655         /*
 656          * part one of RFC 4429, section 3.3
 657          * We should not configure an address as
 658          * optimistic if we do not yet know the link
 659          * layer address of our nexhop router
 660          */
 661
 662         if (rt->rt6i_nexthop == NULL)
 663                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 664
 665         ifa->idev = idev;
 666         in6_dev_hold(idev);
 667         /* For caller */
 668         in6_ifa_hold(ifa);
 669
 670         /* Add to big hash table */
 671         hash = ipv6_addr_hash(addr);
 672
 673         hlist_add_head(&ifa->addr_lst, &inet6_addr_lst[hash]);
 674         in6_ifa_hold(ifa);
 675         write_unlock(&addrconf_hash_lock);
 676
 677         write_lock(&idev->lock);
 678         /* Add to inet6_dev unicast addr list. */
 679         ipv6_link_dev_addr(idev, ifa);
 680
 681 #ifdef CONFIG_IPV6_PRIVACY
 682         if (ifa->flags&IFA_F_TEMPORARY) {
 683                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 684                 in6_ifa_hold(ifa);
 685         }
 686 #endif
 687
 688         in6_ifa_hold(ifa);
 689         write_unlock(&idev->lock);
 690 out2:
 691         rcu_read_unlock_bh();
 692
 693         if (likely(err == 0))
 694                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 695         else {
 696                 kfree(ifa);
 697                 ifa = ERR_PTR(err);
 698         }
 699
 700         return ifa;
 701 out:
 702         write_unlock(&addrconf_hash_lock);
 703         goto out2;
 704 }
 705
 706 /* This function wants to get referenced ifp and releases it before return */
 707
 708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 709 {
 710         struct inet6_ifaddr *ifa, **ifap;
 711         struct inet6_dev *idev = ifp->idev;
 712         int hash;
 713         int deleted = 0, onlink = 0;
 714         unsigned long expires = jiffies;
 715
 716         hash = ipv6_addr_hash(&ifp->addr);
 717
 718         ifp->dead = 1;
 719
 720         write_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init(&ifp->addr_lst);
 722         __in6_ifa_put(ifp);
 723         write_unlock_bh(&addrconf_hash_lock);
 724
 725         write_lock_bh(&idev->lock);
 726 #ifdef CONFIG_IPV6_PRIVACY
 727         if (ifp->flags&IFA_F_TEMPORARY) {
 728                 list_del(&ifp->tmp_list);
 729                 if (ifp->ifpub) {
 730                         in6_ifa_put(ifp->ifpub);
 731                         ifp->ifpub = NULL;
 732                 }
 733                 __in6_ifa_put(ifp);
 734         }
 735 #endif
 736
 737         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 738                 if (ifa == ifp) {
 739                         *ifap = ifa->if_next;
 740                         __in6_ifa_put(ifp);
 741                         ifa->if_next = NULL;
 742                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 743                                 break;
 744                         deleted = 1;
 745                         continue;
 746                 } else if (ifp->flags & IFA_F_PERMANENT) {
 747                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 748                                               ifp->prefix_len)) {
 749                                 if (ifa->flags & IFA_F_PERMANENT) {
 750                                         onlink = 1;
 751                                         if (deleted)
 752                                                 break;
 753                                 } else {
 754                                         unsigned long lifetime;
 755
 756                                         if (!onlink)
 757                                                 onlink = -1;
 758
 759                                         spin_lock(&ifa->lock);
 760
 761                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 762                                         /*
 763                                          * Note: Because this address is
 764                                          * not permanent, lifetime <
 765                                          * LONG_MAX / HZ here.
 766                                          */
 767                                         if (time_before(expires,
 768                                                         ifa->tstamp + lifetime * HZ))
 769                                                 expires = ifa->tstamp + lifetime * HZ;
 770                                         spin_unlock(&ifa->lock);
 771                                 }
 772                         }
 773                 }
 774                 ifap = &ifa->if_next;
 775         }
 776         write_unlock_bh(&idev->lock);
 777
 778         addrconf_del_timer(ifp);
 779
 780         ipv6_ifa_notify(RTM_DELADDR, ifp);
 781
 782         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 783
 784         /*
 785          * Purge or update corresponding prefix
 786          *
 787          * 1) we don't purge prefix here if address was not permanent.
 788          *    prefix is managed by its own lifetime.
 789          * 2) if there're no addresses, delete prefix.
 790          * 3) if there're still other permanent address(es),
 791          *    corresponding prefix is still permanent.
 792          * 4) otherwise, update prefix lifetime to the
 793          *    longest valid lifetime among the corresponding
 794          *    addresses on the device.
 795          *    Note: subsequent RA will update lifetime.
 796          *
 797          * --yoshfuji
 798          */
 799         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 800                 struct in6_addr prefix;
 801                 struct rt6_info *rt;
 802                 struct net *net = dev_net(ifp->idev->dev);
 803                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 804                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 805
 806                 if (rt && addrconf_is_prefix_route(rt)) {
 807                         if (onlink == 0) {
 808                                 ip6_del_rt(rt);
 809                                 rt = NULL;
 810                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 811                                 rt->rt6i_expires = expires;
 812                                 rt->rt6i_flags |= RTF_EXPIRES;
 813                         }
 814                 }
 815                 dst_release(&rt->u.dst);
 816         }
 817
 818         in6_ifa_put(ifp);
 819 }
 820
 821 #ifdef CONFIG_IPV6_PRIVACY
 822 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 823 {
 824         struct inet6_dev *idev = ifp->idev;
 825         struct in6_addr addr, *tmpaddr;
 826         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 827         unsigned long regen_advance;
 828         int tmp_plen;
 829         int ret = 0;
 830         int max_addresses;
 831         u32 addr_flags;
 832
 833         write_lock(&idev->lock);
 834         if (ift) {
 835                 spin_lock_bh(&ift->lock);
 836                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 837                 spin_unlock_bh(&ift->lock);
 838                 tmpaddr = &addr;
 839         } else {
 840                 tmpaddr = NULL;
 841         }
 842 retry:
 843         in6_dev_hold(idev);
 844         if (idev->cnf.use_tempaddr <= 0) {
 845                 write_unlock(&idev->lock);
 846                 printk(KERN_INFO
 847                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 848                 in6_dev_put(idev);
 849                 ret = -1;
 850                 goto out;
 851         }
 852         spin_lock_bh(&ifp->lock);
 853         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 854                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 855                 spin_unlock_bh(&ifp->lock);
 856                 write_unlock(&idev->lock);
 857                 printk(KERN_WARNING
 858                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 859                 in6_dev_put(idev);
 860                 ret = -1;
 861                 goto out;
 862         }
 863         in6_ifa_hold(ifp);
 864         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 865         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 866                 spin_unlock_bh(&ifp->lock);
 867                 write_unlock(&idev->lock);
 868                 printk(KERN_WARNING
 869                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 870                 in6_ifa_put(ifp);
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 876         tmp_valid_lft = min_t(__u32,
 877                               ifp->valid_lft,
 878                               idev->cnf.temp_valid_lft);
 879         tmp_prefered_lft = min_t(__u32,
 880                                  ifp->prefered_lft,
 881                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 882         tmp_plen = ifp->prefix_len;
 883         max_addresses = idev->cnf.max_addresses;
 884         tmp_cstamp = ifp->cstamp;
 885         tmp_tstamp = ifp->tstamp;
 886         spin_unlock_bh(&ifp->lock);
 887
 888         regen_advance = idev->cnf.regen_max_retry *
 889                         idev->cnf.dad_transmits *
 890                         idev->nd_parms->retrans_time / HZ;
 891         write_unlock(&idev->lock);
 892
 893         /* A temporary address is created only if this calculated Preferred
 894          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 895          * an implementation must not create a temporary address with a zero
 896          * Preferred Lifetime.
 897          */
 898         if (tmp_prefered_lft <= regen_advance) {
 899                 in6_ifa_put(ifp);
 900                 in6_dev_put(idev);
 901                 ret = -1;
 902                 goto out;
 903         }
 904
 905         addr_flags = IFA_F_TEMPORARY;
 906         /* set in addrconf_prefix_rcv() */
 907         if (ifp->flags & IFA_F_OPTIMISTIC)
 908                 addr_flags |= IFA_F_OPTIMISTIC;
 909
 910         ift = !max_addresses ||
 911               ipv6_count_addresses(idev) < max_addresses ?
 912                 ipv6_add_addr(idev, &addr, tmp_plen,
 913                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 914                               addr_flags) : NULL;
 915         if (!ift || IS_ERR(ift)) {
 916                 in6_ifa_put(ifp);
 917                 in6_dev_put(idev);
 918                 printk(KERN_INFO
 919                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 920                 tmpaddr = &addr;
 921                 write_lock(&idev->lock);
 922                 goto retry;
 923         }
 924
 925         spin_lock_bh(&ift->lock);
 926         ift->ifpub = ifp;
 927         ift->valid_lft = tmp_valid_lft;
 928         ift->prefered_lft = tmp_prefered_lft;
 929         ift->cstamp = tmp_cstamp;
 930         ift->tstamp = tmp_tstamp;
 931         spin_unlock_bh(&ift->lock);
 932
 933         addrconf_dad_start(ift, 0);
 934         in6_ifa_put(ift);
 935         in6_dev_put(idev);
 936 out:
 937         return ret;
 938 }
 939 #endif
 940
 941 /*
 942  *      Choose an appropriate source address (RFC3484)
 943  */
 944 enum {
 945         IPV6_SADDR_RULE_INIT = 0,
 946         IPV6_SADDR_RULE_LOCAL,
 947         IPV6_SADDR_RULE_SCOPE,
 948         IPV6_SADDR_RULE_PREFERRED,
 949 #ifdef CONFIG_IPV6_MIP6
 950         IPV6_SADDR_RULE_HOA,
 951 #endif
 952         IPV6_SADDR_RULE_OIF,
 953         IPV6_SADDR_RULE_LABEL,
 954 #ifdef CONFIG_IPV6_PRIVACY
 955         IPV6_SADDR_RULE_PRIVACY,
 956 #endif
 957         IPV6_SADDR_RULE_ORCHID,
 958         IPV6_SADDR_RULE_PREFIX,
 959         IPV6_SADDR_RULE_MAX
 960 };
 961
 962 struct ipv6_saddr_score {
 963         int                     rule;
 964         int                     addr_type;
 965         struct inet6_ifaddr     *ifa;
 966         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 967         int                     scopedist;
 968         int                     matchlen;
 969 };
 970
 971 struct ipv6_saddr_dst {
 972         const struct in6_addr *addr;
 973         int ifindex;
 974         int scope;
 975         int label;
 976         unsigned int prefs;
 977 };
 978
 979 static inline int ipv6_saddr_preferred(int type)
 980 {
 981         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 982                 return 1;
 983         return 0;
 984 }
 985
 986 static int ipv6_get_saddr_eval(struct net *net,
 987                                struct ipv6_saddr_score *score,
 988                                struct ipv6_saddr_dst *dst,
 989                                int i)
 990 {
 991         int ret;
 992
 993         if (i <= score->rule) {
 994                 switch (i) {
 995                 case IPV6_SADDR_RULE_SCOPE:
 996                         ret = score->scopedist;
 997                         break;
 998                 case IPV6_SADDR_RULE_PREFIX:
 999                         ret = score->matchlen;
1000                         break;
1001                 default:
1002                         ret = !!test_bit(i, score->scorebits);
1003                 }
1004                 goto out;
1005         }
1006
1007         switch (i) {
1008         case IPV6_SADDR_RULE_INIT:
1009                 /* Rule 0: remember if hiscore is not ready yet */
1010                 ret = !!score->ifa;
1011                 break;
1012         case IPV6_SADDR_RULE_LOCAL:
1013                 /* Rule 1: Prefer same address */
1014                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1015                 break;
1016         case IPV6_SADDR_RULE_SCOPE:
1017                 /* Rule 2: Prefer appropriate scope
1018                  *
1019                  *      ret
1020                  *       ^
1021                  *    -1 |  d 15
1022                  *    ---+--+-+---> scope
1023                  *       |
1024                  *       |             d is scope of the destination.
1025                  *  B-d  |  \
1026                  *       |   \      <- smaller scope is better if
1027                  *  B-15 |    \        if scope is enough for destinaion.
1028                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1029                  * d-C-1 | /
1030                  *       |/         <- greater is better
1031                  *   -C  /             if scope is not enough for destination.
1032                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1033                  *
1034                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1035                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1036                  * Assume B = 0 and we get C > 29.
1037                  */
1038                 ret = __ipv6_addr_src_scope(score->addr_type);
1039                 if (ret >= dst->scope)
1040                         ret = -ret;
1041                 else
1042                         ret -= 128;     /* 30 is enough */
1043                 score->scopedist = ret;
1044                 break;
1045         case IPV6_SADDR_RULE_PREFERRED:
1046                 /* Rule 3: Avoid deprecated and optimistic addresses */
1047                 ret = ipv6_saddr_preferred(score->addr_type) ||
1048                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1049                 break;
1050 #ifdef CONFIG_IPV6_MIP6
1051         case IPV6_SADDR_RULE_HOA:
1052             {
1053                 /* Rule 4: Prefer home address */
1054                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1055                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1056                 break;
1057             }
1058 #endif
1059         case IPV6_SADDR_RULE_OIF:
1060                 /* Rule 5: Prefer outgoing interface */
1061                 ret = (!dst->ifindex ||
1062                        dst->ifindex == score->ifa->idev->dev->ifindex);
1063                 break;
1064         case IPV6_SADDR_RULE_LABEL:
1065                 /* Rule 6: Prefer matching label */
1066                 ret = ipv6_addr_label(net,
1067                                       &score->ifa->addr, score->addr_type,
1068                                       score->ifa->idev->dev->ifindex) == dst->label;
1069                 break;
1070 #ifdef CONFIG_IPV6_PRIVACY
1071         case IPV6_SADDR_RULE_PRIVACY:
1072             {
1073                 /* Rule 7: Prefer public address
1074                  * Note: prefer temprary address if use_tempaddr >= 2
1075                  */
1076                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1077                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1078                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1079                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1080                 break;
1081             }
1082 #endif
1083         case IPV6_SADDR_RULE_ORCHID:
1084                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1085                  *          non-ORCHID vs non-ORCHID
1086                  */
1087                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1088                         ipv6_addr_orchid(dst->addr));
1089                 break;
1090         case IPV6_SADDR_RULE_PREFIX:
1091                 /* Rule 8: Use longest matching prefix */
1092                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1093                                                        dst->addr);
1094                 break;
1095         default:
1096                 ret = 0;
1097         }
1098
1099         if (ret)
1100                 __set_bit(i, score->scorebits);
1101         score->rule = i;
1102 out:
1103         return ret;
1104 }
1105
1106 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1107                        const struct in6_addr *daddr, unsigned int prefs,
1108                        struct in6_addr *saddr)
1109 {
1110         struct ipv6_saddr_score scores[2],
1111                                 *score = &scores[0], *hiscore = &scores[1];
1112         struct ipv6_saddr_dst dst;
1113         struct net_device *dev;
1114         int dst_type;
1115
1116         dst_type = __ipv6_addr_type(daddr);
1117         dst.addr = daddr;
1118         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1119         dst.scope = __ipv6_addr_src_scope(dst_type);
1120         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1121         dst.prefs = prefs;
1122
1123         hiscore->rule = -1;
1124         hiscore->ifa = NULL;
1125
1126         rcu_read_lock();
1127
1128         for_each_netdev_rcu(net, dev) {
1129                 struct inet6_dev *idev;
1130
1131                 /* Candidate Source Address (section 4)
1132                  *  - multicast and link-local destination address,
1133                  *    the set of candidate source address MUST only
1134                  *    include addresses assigned to interfaces
1135                  *    belonging to the same link as the outgoing
1136                  *    interface.
1137                  * (- For site-local destination addresses, the
1138                  *    set of candidate source addresses MUST only
1139                  *    include addresses assigned to interfaces
1140                  *    belonging to the same site as the outgoing
1141                  *    interface.)
1142                  */
1143                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1144                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1145                     dst.ifindex && dev->ifindex != dst.ifindex)
1146                         continue;
1147
1148                 idev = __in6_dev_get(dev);
1149                 if (!idev)
1150                         continue;
1151
1152                 read_lock_bh(&idev->lock);
1153                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1154                         int i;
1155
1156                         /*
1157                          * - Tentative Address (RFC2462 section 5.4)
1158                          *  - A tentative address is not considered
1159                          *    "assigned to an interface" in the traditional
1160                          *    sense, unless it is also flagged as optimistic.
1161                          * - Candidate Source Address (section 4)
1162                          *  - In any case, anycast addresses, multicast
1163                          *    addresses, and the unspecified address MUST
1164                          *    NOT be included in a candidate set.
1165                          */
1166                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1167                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1168                                 continue;
1169
1170                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1171
1172                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1173                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1174                                 LIMIT_NETDEBUG(KERN_DEBUG
1175                                                "ADDRCONF: unspecified / multicast address "
1176                                                "assigned as unicast address on %s",
1177                                                dev->name);
1178                                 continue;
1179                         }
1180
1181                         score->rule = -1;
1182                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1183
1184                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1185                                 int minihiscore, miniscore;
1186
1187                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1188                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1189
1190                                 if (minihiscore > miniscore) {
1191                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1192                                             score->scopedist > 0) {
1193                                                 /*
1194                                                  * special case:
1195                                                  * each remaining entry
1196                                                  * has too small (not enough)
1197                                                  * scope, because ifa entries
1198                                                  * are sorted by their scope
1199                                                  * values.
1200                                                  */
1201                                                 goto try_nextdev;
1202                                         }
1203                                         break;
1204                                 } else if (minihiscore < miniscore) {
1205                                         if (hiscore->ifa)
1206                                                 in6_ifa_put(hiscore->ifa);
1207
1208                                         in6_ifa_hold(score->ifa);
1209
1210                                         swap(hiscore, score);
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
1224         if (!hiscore->ifa)
1225                 return -EADDRNOTAVAIL;
1226
1227         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1228         in6_ifa_put(hiscore->ifa);
1229         return 0;
1230 }
1231
1232 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1233
1234 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1235                     unsigned char banned_flags)
1236 {
1237         struct inet6_dev *idev;
1238         int err = -EADDRNOTAVAIL;
1239
1240         rcu_read_lock();
1241         if ((idev = __in6_dev_get(dev)) != NULL) {
1242                 struct inet6_ifaddr *ifp;
1243
1244                 read_lock_bh(&idev->lock);
1245                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1246                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1247                                 ipv6_addr_copy(addr, &ifp->addr);
1248                                 err = 0;
1249                                 break;
1250                         }
1251                 }
1252                 read_unlock_bh(&idev->lock);
1253         }
1254         rcu_read_unlock();
1255         return err;
1256 }
1257
1258 static int ipv6_count_addresses(struct inet6_dev *idev)
1259 {
1260         int cnt = 0;
1261         struct inet6_ifaddr *ifp;
1262
1263         read_lock_bh(&idev->lock);
1264         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1265                 cnt++;
1266         read_unlock_bh(&idev->lock);
1267         return cnt;
1268 }
1269
1270 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1271                   struct net_device *dev, int strict)
1272 {
1273         struct inet6_ifaddr *ifp = NULL;
1274         struct hlist_node *node;
1275         u8 hash = ipv6_addr_hash(addr);
1276
1277         read_lock_bh(&addrconf_hash_lock);
1278         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
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         struct hlist_node *node;
1299         u8 hash = ipv6_addr_hash(addr);
1300
1301         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1302                 if (!net_eq(dev_net(ifp->idev->dev), net))
1303                         continue;
1304                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1305                         if (dev == NULL || ifp->idev->dev == dev)
1306                                 break;
1307                 }
1308         }
1309         return ifp != NULL;
1310 }
1311
1312 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1313 {
1314         struct inet6_dev *idev;
1315         struct inet6_ifaddr *ifa;
1316         int     onlink;
1317
1318         onlink = 0;
1319         rcu_read_lock();
1320         idev = __in6_dev_get(dev);
1321         if (idev) {
1322                 read_lock_bh(&idev->lock);
1323                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1324                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1325                                                    ifa->prefix_len);
1326                         if (onlink)
1327                                 break;
1328                 }
1329                 read_unlock_bh(&idev->lock);
1330         }
1331         rcu_read_unlock();
1332         return onlink;
1333 }
1334
1335 EXPORT_SYMBOL(ipv6_chk_prefix);
1336
1337 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1338                                      struct net_device *dev, int strict)
1339 {
1340         struct inet6_ifaddr *ifp = NULL;
1341         struct hlist_node *node;
1342         u8 hash = ipv6_addr_hash(addr);
1343
1344         read_lock_bh(&addrconf_hash_lock);
1345         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1346                 if (!net_eq(dev_net(ifp->idev->dev), net))
1347                         continue;
1348                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1349                         if (dev == NULL || ifp->idev->dev == dev ||
1350                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1351                                 in6_ifa_hold(ifp);
1352                                 break;
1353                         }
1354                 }
1355         }
1356         read_unlock_bh(&addrconf_hash_lock);
1357
1358         return ifp;
1359 }
1360
1361 /* Gets referenced address, destroys ifaddr */
1362
1363 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1364 {
1365         if (ifp->flags&IFA_F_PERMANENT) {
1366                 spin_lock_bh(&ifp->lock);
1367                 addrconf_del_timer(ifp);
1368                 ifp->flags |= IFA_F_TENTATIVE;
1369                 if (dad_failed)
1370                         ifp->flags |= IFA_F_DADFAILED;
1371                 spin_unlock_bh(&ifp->lock);
1372                 if (dad_failed)
1373                         ipv6_ifa_notify(0, ifp);
1374                 in6_ifa_put(ifp);
1375 #ifdef CONFIG_IPV6_PRIVACY
1376         } else if (ifp->flags&IFA_F_TEMPORARY) {
1377                 struct inet6_ifaddr *ifpub;
1378                 spin_lock_bh(&ifp->lock);
1379                 ifpub = ifp->ifpub;
1380                 if (ifpub) {
1381                         in6_ifa_hold(ifpub);
1382                         spin_unlock_bh(&ifp->lock);
1383                         ipv6_create_tempaddr(ifpub, ifp);
1384                         in6_ifa_put(ifpub);
1385                 } else {
1386                         spin_unlock_bh(&ifp->lock);
1387                 }
1388                 ipv6_del_addr(ifp);
1389 #endif
1390         } else
1391                 ipv6_del_addr(ifp);
1392 }
1393
1394 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1395 {
1396         struct inet6_dev *idev = ifp->idev;
1397
1398         if (net_ratelimit())
1399                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1400                         ifp->idev->dev->name, &ifp->addr);
1401
1402         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1403                 struct in6_addr addr;
1404
1405                 addr.s6_addr32[0] = htonl(0xfe800000);
1406                 addr.s6_addr32[1] = 0;
1407
1408                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1409                     ipv6_addr_equal(&ifp->addr, &addr)) {
1410                         /* DAD failed for link-local based on MAC address */
1411                         idev->cnf.disable_ipv6 = 1;
1412
1413                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1414                                 ifp->idev->dev->name);
1415                 }
1416         }
1417
1418         addrconf_dad_stop(ifp, 1);
1419 }
1420
1421 /* Join to solicited addr multicast group. */
1422
1423 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1424 {
1425         struct in6_addr maddr;
1426
1427         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1428                 return;
1429
1430         addrconf_addr_solict_mult(addr, &maddr);
1431         ipv6_dev_mc_inc(dev, &maddr);
1432 }
1433
1434 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1435 {
1436         struct in6_addr maddr;
1437
1438         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1439                 return;
1440
1441         addrconf_addr_solict_mult(addr, &maddr);
1442         __ipv6_dev_mc_dec(idev, &maddr);
1443 }
1444
1445 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1446 {
1447         struct in6_addr addr;
1448         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1449         if (ipv6_addr_any(&addr))
1450                 return;
1451         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1452 }
1453
1454 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1455 {
1456         struct in6_addr addr;
1457         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1458         if (ipv6_addr_any(&addr))
1459                 return;
1460         __ipv6_dev_ac_dec(ifp->idev, &addr);
1461 }
1462
1463 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1464 {
1465         if (dev->addr_len != ETH_ALEN)
1466                 return -1;
1467         memcpy(eui, dev->dev_addr, 3);
1468         memcpy(eui + 5, dev->dev_addr + 3, 3);
1469
1470         /*
1471          * The zSeries OSA network cards can be shared among various
1472          * OS instances, but the OSA cards have only one MAC address.
1473          * This leads to duplicate address conflicts in conjunction
1474          * with IPv6 if more than one instance uses the same card.
1475          *
1476          * The driver for these cards can deliver a unique 16-bit
1477          * identifier for each instance sharing the same card.  It is
1478          * placed instead of 0xFFFE in the interface identifier.  The
1479          * "u" bit of the interface identifier is not inverted in this
1480          * case.  Hence the resulting interface identifier has local
1481          * scope according to RFC2373.
1482          */
1483         if (dev->dev_id) {
1484                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1485                 eui[4] = dev->dev_id & 0xFF;
1486         } else {
1487                 eui[3] = 0xFF;
1488                 eui[4] = 0xFE;
1489                 eui[0] ^= 2;
1490         }
1491         return 0;
1492 }
1493
1494 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1495 {
1496         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1497         if (dev->addr_len != ARCNET_ALEN)
1498                 return -1;
1499         memset(eui, 0, 7);
1500         eui[7] = *(u8*)dev->dev_addr;
1501         return 0;
1502 }
1503
1504 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1505 {
1506         if (dev->addr_len != INFINIBAND_ALEN)
1507                 return -1;
1508         memcpy(eui, dev->dev_addr + 12, 8);
1509         eui[0] |= 2;
1510         return 0;
1511 }
1512
1513 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1514 {
1515         if (addr == 0)
1516                 return -1;
1517         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1518                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1519                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1520                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1521                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1522                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1523         eui[1] = 0;
1524         eui[2] = 0x5E;
1525         eui[3] = 0xFE;
1526         memcpy(eui + 4, &addr, 4);
1527         return 0;
1528 }
1529 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1530
1531 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1532 {
1533         if (dev->priv_flags & IFF_ISATAP)
1534                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1535         return -1;
1536 }
1537
1538 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1539 {
1540         switch (dev->type) {
1541         case ARPHRD_ETHER:
1542         case ARPHRD_FDDI:
1543         case ARPHRD_IEEE802_TR:
1544                 return addrconf_ifid_eui48(eui, dev);
1545         case ARPHRD_ARCNET:
1546                 return addrconf_ifid_arcnet(eui, dev);
1547         case ARPHRD_INFINIBAND:
1548                 return addrconf_ifid_infiniband(eui, dev);
1549         case ARPHRD_SIT:
1550                 return addrconf_ifid_sit(eui, dev);
1551         }
1552         return -1;
1553 }
1554
1555 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1556 {
1557         int err = -1;
1558         struct inet6_ifaddr *ifp;
1559
1560         read_lock_bh(&idev->lock);
1561         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1562                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1563                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1564                         err = 0;
1565                         break;
1566                 }
1567         }
1568         read_unlock_bh(&idev->lock);
1569         return err;
1570 }
1571
1572 #ifdef CONFIG_IPV6_PRIVACY
1573 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1574 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1575 {
1576 regen:
1577         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1578         idev->rndid[0] &= ~0x02;
1579
1580         /*
1581          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1582          * check if generated address is not inappropriate
1583          *
1584          *  - Reserved subnet anycast (RFC 2526)
1585          *      11111101 11....11 1xxxxxxx
1586          *  - ISATAP (RFC4214) 6.1
1587          *      00-00-5E-FE-xx-xx-xx-xx
1588          *  - value 0
1589          *  - XXX: already assigned to an address on the device
1590          */
1591         if (idev->rndid[0] == 0xfd &&
1592             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1593             (idev->rndid[7]&0x80))
1594                 goto regen;
1595         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1596                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1597                         goto regen;
1598                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1599                         goto regen;
1600         }
1601
1602         return 0;
1603 }
1604
1605 static void ipv6_regen_rndid(unsigned long data)
1606 {
1607         struct inet6_dev *idev = (struct inet6_dev *) data;
1608         unsigned long expires;
1609
1610         rcu_read_lock_bh();
1611         write_lock_bh(&idev->lock);
1612
1613         if (idev->dead)
1614                 goto out;
1615
1616         if (__ipv6_regen_rndid(idev) < 0)
1617                 goto out;
1618
1619         expires = jiffies +
1620                 idev->cnf.temp_prefered_lft * HZ -
1621                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1622         if (time_before(expires, jiffies)) {
1623                 printk(KERN_WARNING
1624                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1625                         idev->dev->name);
1626                 goto out;
1627         }
1628
1629         if (!mod_timer(&idev->regen_timer, expires))
1630                 in6_dev_hold(idev);
1631
1632 out:
1633         write_unlock_bh(&idev->lock);
1634         rcu_read_unlock_bh();
1635         in6_dev_put(idev);
1636 }
1637
1638 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1639         int ret = 0;
1640
1641         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1642                 ret = __ipv6_regen_rndid(idev);
1643         return ret;
1644 }
1645 #endif
1646
1647 /*
1648  *      Add prefix route.
1649  */
1650
1651 static void
1652 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1653                       unsigned long expires, u32 flags)
1654 {
1655         struct fib6_config cfg = {
1656                 .fc_table = RT6_TABLE_PREFIX,
1657                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1658                 .fc_ifindex = dev->ifindex,
1659                 .fc_expires = expires,
1660                 .fc_dst_len = plen,
1661                 .fc_flags = RTF_UP | flags,
1662                 .fc_nlinfo.nl_net = dev_net(dev),
1663                 .fc_protocol = RTPROT_KERNEL,
1664         };
1665
1666         ipv6_addr_copy(&cfg.fc_dst, pfx);
1667
1668         /* Prevent useless cloning on PtP SIT.
1669            This thing is done here expecting that the whole
1670            class of non-broadcast devices need not cloning.
1671          */
1672 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1673         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1674                 cfg.fc_flags |= RTF_NONEXTHOP;
1675 #endif
1676
1677         ip6_route_add(&cfg);
1678 }
1679
1680 /* Create "default" multicast route to the interface */
1681
1682 static void addrconf_add_mroute(struct net_device *dev)
1683 {
1684         struct fib6_config cfg = {
1685                 .fc_table = RT6_TABLE_LOCAL,
1686                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1687                 .fc_ifindex = dev->ifindex,
1688                 .fc_dst_len = 8,
1689                 .fc_flags = RTF_UP,
1690                 .fc_nlinfo.nl_net = dev_net(dev),
1691         };
1692
1693         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1694
1695         ip6_route_add(&cfg);
1696 }
1697
1698 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1699 static void sit_route_add(struct net_device *dev)
1700 {
1701         struct fib6_config cfg = {
1702                 .fc_table = RT6_TABLE_MAIN,
1703                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1704                 .fc_ifindex = dev->ifindex,
1705                 .fc_dst_len = 96,
1706                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1707                 .fc_nlinfo.nl_net = dev_net(dev),
1708         };
1709
1710         /* prefix length - 96 bits "::d.d.d.d" */
1711         ip6_route_add(&cfg);
1712 }
1713 #endif
1714
1715 static void addrconf_add_lroute(struct net_device *dev)
1716 {
1717         struct in6_addr addr;
1718
1719         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1720         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1721 }
1722
1723 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1724 {
1725         struct inet6_dev *idev;
1726
1727         ASSERT_RTNL();
1728
1729         if ((idev = ipv6_find_idev(dev)) == NULL)
1730                 return NULL;
1731
1732         /* Add default multicast route */
1733         addrconf_add_mroute(dev);
1734
1735         /* Add link local route */
1736         addrconf_add_lroute(dev);
1737         return idev;
1738 }
1739
1740 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1741 {
1742         struct prefix_info *pinfo;
1743         __u32 valid_lft;
1744         __u32 prefered_lft;
1745         int addr_type;
1746         struct inet6_dev *in6_dev;
1747         struct net *net = dev_net(dev);
1748
1749         pinfo = (struct prefix_info *) opt;
1750
1751         if (len < sizeof(struct prefix_info)) {
1752                 ADBG(("addrconf: prefix option too short\n"));
1753                 return;
1754         }
1755
1756         /*
1757          *      Validation checks ([ADDRCONF], page 19)
1758          */
1759
1760         addr_type = ipv6_addr_type(&pinfo->prefix);
1761
1762         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1763                 return;
1764
1765         valid_lft = ntohl(pinfo->valid);
1766         prefered_lft = ntohl(pinfo->prefered);
1767
1768         if (prefered_lft > valid_lft) {
1769                 if (net_ratelimit())
1770                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1771                 return;
1772         }
1773
1774         in6_dev = in6_dev_get(dev);
1775
1776         if (in6_dev == NULL) {
1777                 if (net_ratelimit())
1778                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1779                 return;
1780         }
1781
1782         /*
1783          *      Two things going on here:
1784          *      1) Add routes for on-link prefixes
1785          *      2) Configure prefixes with the auto flag set
1786          */
1787
1788         if (pinfo->onlink) {
1789                 struct rt6_info *rt;
1790                 unsigned long rt_expires;
1791
1792                 /* Avoid arithmetic overflow. Really, we could
1793                  * save rt_expires in seconds, likely valid_lft,
1794                  * but it would require division in fib gc, that it
1795                  * not good.
1796                  */
1797                 if (HZ > USER_HZ)
1798                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1799                 else
1800                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1801
1802                 if (addrconf_finite_timeout(rt_expires))
1803                         rt_expires *= HZ;
1804
1805                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1806                                 dev->ifindex, 1);
1807
1808                 if (rt && addrconf_is_prefix_route(rt)) {
1809                         /* Autoconf prefix route */
1810                         if (valid_lft == 0) {
1811                                 ip6_del_rt(rt);
1812                                 rt = NULL;
1813                         } else if (addrconf_finite_timeout(rt_expires)) {
1814                                 /* not infinity */
1815                                 rt->rt6i_expires = jiffies + rt_expires;
1816                                 rt->rt6i_flags |= RTF_EXPIRES;
1817                         } else {
1818                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1819                                 rt->rt6i_expires = 0;
1820                         }
1821                 } else if (valid_lft) {
1822                         clock_t expires = 0;
1823                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1824                         if (addrconf_finite_timeout(rt_expires)) {
1825                                 /* not infinity */
1826                                 flags |= RTF_EXPIRES;
1827                                 expires = jiffies_to_clock_t(rt_expires);
1828                         }
1829                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1830                                               dev, expires, flags);
1831                 }
1832                 if (rt)
1833                         dst_release(&rt->u.dst);
1834         }
1835
1836         /* Try to figure out our local address for this prefix */
1837
1838         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1839                 struct inet6_ifaddr * ifp;
1840                 struct in6_addr addr;
1841                 int create = 0, update_lft = 0;
1842
1843                 if (pinfo->prefix_len == 64) {
1844                         memcpy(&addr, &pinfo->prefix, 8);
1845                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1846                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1847                                 in6_dev_put(in6_dev);
1848                                 return;
1849                         }
1850                         goto ok;
1851                 }
1852                 if (net_ratelimit())
1853                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1854                                pinfo->prefix_len);
1855                 in6_dev_put(in6_dev);
1856                 return;
1857
1858 ok:
1859
1860                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1861
1862                 if (ifp == NULL && valid_lft) {
1863                         int max_addresses = in6_dev->cnf.max_addresses;
1864                         u32 addr_flags = 0;
1865
1866 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1867                         if (in6_dev->cnf.optimistic_dad &&
1868                             !net->ipv6.devconf_all->forwarding)
1869                                 addr_flags = IFA_F_OPTIMISTIC;
1870 #endif
1871
1872                         /* Do not allow to create too much of autoconfigured
1873                          * addresses; this would be too easy way to crash kernel.
1874                          */
1875                         if (!max_addresses ||
1876                             ipv6_count_addresses(in6_dev) < max_addresses)
1877                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1878                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1879                                                     addr_flags);
1880
1881                         if (!ifp || IS_ERR(ifp)) {
1882                                 in6_dev_put(in6_dev);
1883                                 return;
1884                         }
1885
1886                         update_lft = create = 1;
1887                         ifp->cstamp = jiffies;
1888                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1889                 }
1890
1891                 if (ifp) {
1892                         int flags;
1893                         unsigned long now;
1894 #ifdef CONFIG_IPV6_PRIVACY
1895                         struct inet6_ifaddr *ift;
1896 #endif
1897                         u32 stored_lft;
1898
1899                         /* update lifetime (RFC2462 5.5.3 e) */
1900                         spin_lock(&ifp->lock);
1901                         now = jiffies;
1902                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1903                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1904                         else
1905                                 stored_lft = 0;
1906                         if (!update_lft && stored_lft) {
1907                                 if (valid_lft > MIN_VALID_LIFETIME ||
1908                                     valid_lft > stored_lft)
1909                                         update_lft = 1;
1910                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1911                                         /* valid_lft <= stored_lft is always true */
1912                                         /*
1913                                          * RFC 4862 Section 5.5.3e:
1914                                          * "Note that the preferred lifetime of
1915                                          *  the corresponding address is always
1916                                          *  reset to the Preferred Lifetime in
1917                                          *  the received Prefix Information
1918                                          *  option, regardless of whether the
1919                                          *  valid lifetime is also reset or
1920                                          *  ignored."
1921                                          *
1922                                          *  So if the preferred lifetime in
1923                                          *  this advertisement is different
1924                                          *  than what we have stored, but the
1925                                          *  valid lifetime is invalid, just
1926                                          *  reset prefered_lft.
1927                                          *
1928                                          *  We must set the valid lifetime
1929                                          *  to the stored lifetime since we'll
1930                                          *  be updating the timestamp below,
1931                                          *  else we'll set it back to the
1932                                          *  minumum.
1933                                          */
1934                                         if (prefered_lft != ifp->prefered_lft) {
1935                                                 valid_lft = stored_lft;
1936                                                 update_lft = 1;
1937                                         }
1938                                 } else {
1939                                         valid_lft = MIN_VALID_LIFETIME;
1940                                         if (valid_lft < prefered_lft)
1941                                                 prefered_lft = valid_lft;
1942                                         update_lft = 1;
1943                                 }
1944                         }
1945
1946                         if (update_lft) {
1947                                 ifp->valid_lft = valid_lft;
1948                                 ifp->prefered_lft = prefered_lft;
1949                                 ifp->tstamp = now;
1950                                 flags = ifp->flags;
1951                                 ifp->flags &= ~IFA_F_DEPRECATED;
1952                                 spin_unlock(&ifp->lock);
1953
1954                                 if (!(flags&IFA_F_TENTATIVE))
1955                                         ipv6_ifa_notify(0, ifp);
1956                         } else
1957                                 spin_unlock(&ifp->lock);
1958
1959 #ifdef CONFIG_IPV6_PRIVACY
1960                         read_lock_bh(&in6_dev->lock);
1961                         /* update all temporary addresses in the list */
1962                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1963                                 /*
1964                                  * When adjusting the lifetimes of an existing
1965                                  * temporary address, only lower the lifetimes.
1966                                  * Implementations must not increase the
1967                                  * lifetimes of an existing temporary address
1968                                  * when processing a Prefix Information Option.
1969                                  */
1970                                 if (ifp != ift->ifpub)
1971                                         continue;
1972
1973                                 spin_lock(&ift->lock);
1974                                 flags = ift->flags;
1975                                 if (ift->valid_lft > valid_lft &&
1976                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1977                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1978                                 if (ift->prefered_lft > prefered_lft &&
1979                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1980                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1981                                 spin_unlock(&ift->lock);
1982                                 if (!(flags&IFA_F_TENTATIVE))
1983                                         ipv6_ifa_notify(0, ift);
1984                         }
1985
1986                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1987                                 /*
1988                                  * When a new public address is created as described in [ADDRCONF],
1989                                  * also create a new temporary address.
1990                                  */
1991                                 read_unlock_bh(&in6_dev->lock);
1992                                 ipv6_create_tempaddr(ifp, NULL);
1993                         } else {
1994                                 read_unlock_bh(&in6_dev->lock);
1995                         }
1996 #endif
1997                         in6_ifa_put(ifp);
1998                         addrconf_verify(0);
1999                 }
2000         }
2001         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2002         in6_dev_put(in6_dev);
2003 }
2004
2005 /*
2006  *      Set destination address.
2007  *      Special case for SIT interfaces where we create a new "virtual"
2008  *      device.
2009  */
2010 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2011 {
2012         struct in6_ifreq ireq;
2013         struct net_device *dev;
2014         int err = -EINVAL;
2015
2016         rtnl_lock();
2017
2018         err = -EFAULT;
2019         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2020                 goto err_exit;
2021
2022         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2023
2024         err = -ENODEV;
2025         if (dev == NULL)
2026                 goto err_exit;
2027
2028 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2029         if (dev->type == ARPHRD_SIT) {
2030                 const struct net_device_ops *ops = dev->netdev_ops;
2031                 struct ifreq ifr;
2032                 struct ip_tunnel_parm p;
2033
2034                 err = -EADDRNOTAVAIL;
2035                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2036                         goto err_exit;
2037
2038                 memset(&p, 0, sizeof(p));
2039                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2040                 p.iph.saddr = 0;
2041                 p.iph.version = 4;
2042                 p.iph.ihl = 5;
2043                 p.iph.protocol = IPPROTO_IPV6;
2044                 p.iph.ttl = 64;
2045                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2046
2047                 if (ops->ndo_do_ioctl) {
2048                         mm_segment_t oldfs = get_fs();
2049
2050                         set_fs(KERNEL_DS);
2051                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2052                         set_fs(oldfs);
2053                 } else
2054                         err = -EOPNOTSUPP;
2055
2056                 if (err == 0) {
2057                         err = -ENOBUFS;
2058                         dev = __dev_get_by_name(net, p.name);
2059                         if (!dev)
2060                                 goto err_exit;
2061                         err = dev_open(dev);
2062                 }
2063         }
2064 #endif
2065
2066 err_exit:
2067         rtnl_unlock();
2068         return err;
2069 }
2070
2071 /*
2072  *      Manual configuration of address on an interface
2073  */
2074 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2075                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2076                           __u32 valid_lft)
2077 {
2078         struct inet6_ifaddr *ifp;
2079         struct inet6_dev *idev;
2080         struct net_device *dev;
2081         int scope;
2082         u32 flags;
2083         clock_t expires;
2084         unsigned long timeout;
2085
2086         ASSERT_RTNL();
2087
2088         if (plen > 128)
2089                 return -EINVAL;
2090
2091         /* check the lifetime */
2092         if (!valid_lft || prefered_lft > valid_lft)
2093                 return -EINVAL;
2094
2095         dev = __dev_get_by_index(net, ifindex);
2096         if (!dev)
2097                 return -ENODEV;
2098
2099         if ((idev = addrconf_add_dev(dev)) == NULL)
2100                 return -ENOBUFS;
2101
2102         scope = ipv6_addr_scope(pfx);
2103
2104         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2105         if (addrconf_finite_timeout(timeout)) {
2106                 expires = jiffies_to_clock_t(timeout * HZ);
2107                 valid_lft = timeout;
2108                 flags = RTF_EXPIRES;
2109         } else {
2110                 expires = 0;
2111                 flags = 0;
2112                 ifa_flags |= IFA_F_PERMANENT;
2113         }
2114
2115         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2116         if (addrconf_finite_timeout(timeout)) {
2117                 if (timeout == 0)
2118                         ifa_flags |= IFA_F_DEPRECATED;
2119                 prefered_lft = timeout;
2120         }
2121
2122         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2123
2124         if (!IS_ERR(ifp)) {
2125                 spin_lock_bh(&ifp->lock);
2126                 ifp->valid_lft = valid_lft;
2127                 ifp->prefered_lft = prefered_lft;
2128                 ifp->tstamp = jiffies;
2129                 spin_unlock_bh(&ifp->lock);
2130
2131                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2132                                       expires, flags);
2133                 /*
2134                  * Note that section 3.1 of RFC 4429 indicates
2135                  * that the Optimistic flag should not be set for
2136                  * manually configured addresses
2137                  */
2138                 addrconf_dad_start(ifp, 0);
2139                 in6_ifa_put(ifp);
2140                 addrconf_verify(0);
2141                 return 0;
2142         }
2143
2144         return PTR_ERR(ifp);
2145 }
2146
2147 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2148                           unsigned int plen)
2149 {
2150         struct inet6_ifaddr *ifp;
2151         struct inet6_dev *idev;
2152         struct net_device *dev;
2153
2154         if (plen > 128)
2155                 return -EINVAL;
2156
2157         dev = __dev_get_by_index(net, ifindex);
2158         if (!dev)
2159                 return -ENODEV;
2160
2161         if ((idev = __in6_dev_get(dev)) == NULL)
2162                 return -ENXIO;
2163
2164         read_lock_bh(&idev->lock);
2165         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2166                 if (ifp->prefix_len == plen &&
2167                     ipv6_addr_equal(pfx, &ifp->addr)) {
2168                         in6_ifa_hold(ifp);
2169                         read_unlock_bh(&idev->lock);
2170
2171                         ipv6_del_addr(ifp);
2172
2173                         /* If the last address is deleted administratively,
2174                            disable IPv6 on this interface.
2175                          */
2176                         if (idev->addr_list == NULL)
2177                                 addrconf_ifdown(idev->dev, 1);
2178                         return 0;
2179                 }
2180         }
2181         read_unlock_bh(&idev->lock);
2182         return -EADDRNOTAVAIL;
2183 }
2184
2185
2186 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2187 {
2188         struct in6_ifreq ireq;
2189         int err;
2190
2191         if (!capable(CAP_NET_ADMIN))
2192                 return -EPERM;
2193
2194         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2195                 return -EFAULT;
2196
2197         rtnl_lock();
2198         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2199                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2200                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2201         rtnl_unlock();
2202         return err;
2203 }
2204
2205 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2206 {
2207         struct in6_ifreq ireq;
2208         int err;
2209
2210         if (!capable(CAP_NET_ADMIN))
2211                 return -EPERM;
2212
2213         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2214                 return -EFAULT;
2215
2216         rtnl_lock();
2217         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2218                              ireq.ifr6_prefixlen);
2219         rtnl_unlock();
2220         return err;
2221 }
2222
2223 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2224                      int plen, int scope)
2225 {
2226         struct inet6_ifaddr *ifp;
2227
2228         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2229         if (!IS_ERR(ifp)) {
2230                 spin_lock_bh(&ifp->lock);
2231                 ifp->flags &= ~IFA_F_TENTATIVE;
2232                 spin_unlock_bh(&ifp->lock);
2233                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2234                 in6_ifa_put(ifp);
2235         }
2236 }
2237
2238 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2239 static void sit_add_v4_addrs(struct inet6_dev *idev)
2240 {
2241         struct in6_addr addr;
2242         struct net_device *dev;
2243         struct net *net = dev_net(idev->dev);
2244         int scope;
2245
2246         ASSERT_RTNL();
2247
2248         memset(&addr, 0, sizeof(struct in6_addr));
2249         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2250
2251         if (idev->dev->flags&IFF_POINTOPOINT) {
2252                 addr.s6_addr32[0] = htonl(0xfe800000);
2253                 scope = IFA_LINK;
2254         } else {
2255                 scope = IPV6_ADDR_COMPATv4;
2256         }
2257
2258         if (addr.s6_addr32[3]) {
2259                 add_addr(idev, &addr, 128, scope);
2260                 return;
2261         }
2262
2263         for_each_netdev(net, dev) {
2264                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2265                 if (in_dev && (dev->flags & IFF_UP)) {
2266                         struct in_ifaddr * ifa;
2267
2268                         int flag = scope;
2269
2270                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2271                                 int plen;
2272
2273                                 addr.s6_addr32[3] = ifa->ifa_local;
2274
2275                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2276                                         continue;
2277                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2278                                         if (idev->dev->flags&IFF_POINTOPOINT)
2279                                                 continue;
2280                                         flag |= IFA_HOST;
2281                                 }
2282                                 if (idev->dev->flags&IFF_POINTOPOINT)
2283                                         plen = 64;
2284                                 else
2285                                         plen = 96;
2286
2287                                 add_addr(idev, &addr, plen, flag);
2288                         }
2289                 }
2290         }
2291 }
2292 #endif
2293
2294 static void init_loopback(struct net_device *dev)
2295 {
2296         struct inet6_dev  *idev;
2297
2298         /* ::1 */
2299
2300         ASSERT_RTNL();
2301
2302         if ((idev = ipv6_find_idev(dev)) == NULL) {
2303                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2304                 return;
2305         }
2306
2307         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2308 }
2309
2310 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2311 {
2312         struct inet6_ifaddr * ifp;
2313         u32 addr_flags = IFA_F_PERMANENT;
2314
2315 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2316         if (idev->cnf.optimistic_dad &&
2317             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2318                 addr_flags |= IFA_F_OPTIMISTIC;
2319 #endif
2320
2321
2322         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2323         if (!IS_ERR(ifp)) {
2324                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2325                 addrconf_dad_start(ifp, 0);
2326                 in6_ifa_put(ifp);
2327         }
2328 }
2329
2330 static void addrconf_dev_config(struct net_device *dev)
2331 {
2332         struct in6_addr addr;
2333         struct inet6_dev    * idev;
2334
2335         ASSERT_RTNL();
2336
2337         if ((dev->type != ARPHRD_ETHER) &&
2338             (dev->type != ARPHRD_FDDI) &&
2339             (dev->type != ARPHRD_IEEE802_TR) &&
2340             (dev->type != ARPHRD_ARCNET) &&
2341             (dev->type != ARPHRD_INFINIBAND)) {
2342                 /* Alas, we support only Ethernet autoconfiguration. */
2343                 return;
2344         }
2345
2346         idev = addrconf_add_dev(dev);
2347         if (idev == NULL)
2348                 return;
2349
2350         memset(&addr, 0, sizeof(struct in6_addr));
2351         addr.s6_addr32[0] = htonl(0xFE800000);
2352
2353         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2354                 addrconf_add_linklocal(idev, &addr);
2355 }
2356
2357 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2358 static void addrconf_sit_config(struct net_device *dev)
2359 {
2360         struct inet6_dev *idev;
2361
2362         ASSERT_RTNL();
2363
2364         /*
2365          * Configure the tunnel with one of our IPv4
2366          * addresses... we should configure all of
2367          * our v4 addrs in the tunnel
2368          */
2369
2370         if ((idev = ipv6_find_idev(dev)) == NULL) {
2371                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2372                 return;
2373         }
2374
2375         if (dev->priv_flags & IFF_ISATAP) {
2376                 struct in6_addr addr;
2377
2378                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2379                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2380                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2381                         addrconf_add_linklocal(idev, &addr);
2382                 return;
2383         }
2384
2385         sit_add_v4_addrs(idev);
2386
2387         if (dev->flags&IFF_POINTOPOINT) {
2388                 addrconf_add_mroute(dev);
2389                 addrconf_add_lroute(dev);
2390         } else
2391                 sit_route_add(dev);
2392 }
2393 #endif
2394
2395 static inline int
2396 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2397 {
2398         struct in6_addr lladdr;
2399
2400         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2401                 addrconf_add_linklocal(idev, &lladdr);
2402                 return 0;
2403         }
2404         return -1;
2405 }
2406
2407 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2408 {
2409         struct net_device *link_dev;
2410         struct net *net = dev_net(idev->dev);
2411
2412         /* first try to inherit the link-local address from the link device */
2413         if (idev->dev->iflink &&
2414             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2415                 if (!ipv6_inherit_linklocal(idev, link_dev))
2416                         return;
2417         }
2418         /* then try to inherit it from any device */
2419         for_each_netdev(net, link_dev) {
2420                 if (!ipv6_inherit_linklocal(idev, link_dev))
2421                         return;
2422         }
2423         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2424 }
2425
2426 /*
2427  * Autoconfigure tunnel with a link-local address so routing protocols,
2428  * DHCPv6, MLD etc. can be run over the virtual link
2429  */
2430
2431 static void addrconf_ip6_tnl_config(struct net_device *dev)
2432 {
2433         struct inet6_dev *idev;
2434
2435         ASSERT_RTNL();
2436
2437         if ((idev = addrconf_add_dev(dev)) == NULL) {
2438                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2439                 return;
2440         }
2441         ip6_tnl_add_linklocal(idev);
2442 }
2443
2444 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2445                            void * data)
2446 {
2447         struct net_device *dev = (struct net_device *) data;
2448         struct inet6_dev *idev = __in6_dev_get(dev);
2449         int run_pending = 0;
2450         int err;
2451
2452         switch(event) {
2453         case NETDEV_REGISTER:
2454                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2455                         idev = ipv6_add_dev(dev);
2456                         if (!idev)
2457                                 return notifier_from_errno(-ENOMEM);
2458                 }
2459                 break;
2460         case NETDEV_UP:
2461         case NETDEV_CHANGE:
2462                 if (dev->flags & IFF_SLAVE)
2463                         break;
2464
2465                 if (event == NETDEV_UP) {
2466                         if (!addrconf_qdisc_ok(dev)) {
2467                                 /* device is not ready yet. */
2468                                 printk(KERN_INFO
2469                                         "ADDRCONF(NETDEV_UP): %s: "
2470                                         "link is not ready\n",
2471                                         dev->name);
2472                                 break;
2473                         }
2474
2475                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2476                                 idev = ipv6_add_dev(dev);
2477
2478                         if (idev) {
2479                                 idev->if_flags |= IF_READY;
2480                                 run_pending = 1;
2481                         }
2482                 } else {
2483                         if (!addrconf_qdisc_ok(dev)) {
2484                                 /* device is still not ready. */
2485                                 break;
2486                         }
2487
2488                         if (idev) {
2489                                 if (idev->if_flags & IF_READY) {
2490                                         /* device is already configured. */
2491                                         break;
2492                                 }
2493                                 idev->if_flags |= IF_READY;
2494                         }
2495
2496                         printk(KERN_INFO
2497                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2498                                         "link becomes ready\n",
2499                                         dev->name);
2500
2501                         run_pending = 1;
2502                 }
2503
2504                 switch(dev->type) {
2505 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2506                 case ARPHRD_SIT:
2507                         addrconf_sit_config(dev);
2508                         break;
2509 #endif
2510                 case ARPHRD_TUNNEL6:
2511                         addrconf_ip6_tnl_config(dev);
2512                         break;
2513                 case ARPHRD_LOOPBACK:
2514                         init_loopback(dev);
2515                         break;
2516
2517                 default:
2518                         addrconf_dev_config(dev);
2519                         break;
2520                 }
2521                 if (idev) {
2522                         if (run_pending)
2523                                 addrconf_dad_run(idev);
2524
2525                         /* If the MTU changed during the interface down, when the
2526                            interface up, the changed MTU must be reflected in the
2527                            idev as well as routers.
2528                          */
2529                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2530                                 rt6_mtu_change(dev, dev->mtu);
2531                                 idev->cnf.mtu6 = dev->mtu;
2532                         }
2533                         idev->tstamp = jiffies;
2534                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2535                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2536                            stop IPv6 on this interface.
2537                          */
2538                         if (dev->mtu < IPV6_MIN_MTU)
2539                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2540                 }
2541                 break;
2542
2543         case NETDEV_CHANGEMTU:
2544                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2545                         rt6_mtu_change(dev, dev->mtu);
2546                         idev->cnf.mtu6 = dev->mtu;
2547                         break;
2548                 }
2549
2550                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2551                         idev = ipv6_add_dev(dev);
2552                         if (idev)
2553                                 break;
2554                 }
2555
2556                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2557
2558         case NETDEV_DOWN:
2559         case NETDEV_UNREGISTER:
2560                 /*
2561                  *      Remove all addresses from this interface.
2562                  */
2563                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2564                 break;
2565
2566         case NETDEV_CHANGENAME:
2567                 if (idev) {
2568                         snmp6_unregister_dev(idev);
2569                         addrconf_sysctl_unregister(idev);
2570                         addrconf_sysctl_register(idev);
2571                         err = snmp6_register_dev(idev);
2572                         if (err)
2573                                 return notifier_from_errno(err);
2574                 }
2575                 break;
2576         case NETDEV_PRE_TYPE_CHANGE:
2577         case NETDEV_POST_TYPE_CHANGE:
2578                 addrconf_type_change(dev, event);
2579                 break;
2580         }
2581
2582         return NOTIFY_OK;
2583 }
2584
2585 /*
2586  *      addrconf module should be notified of a device going up
2587  */
2588 static struct notifier_block ipv6_dev_notf = {
2589         .notifier_call = addrconf_notify,
2590         .priority = 0
2591 };
2592
2593 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2594 {
2595         struct inet6_dev *idev;
2596         ASSERT_RTNL();
2597
2598         idev = __in6_dev_get(dev);
2599
2600         if (event == NETDEV_POST_TYPE_CHANGE)
2601                 ipv6_mc_remap(idev);
2602         else if (event == NETDEV_PRE_TYPE_CHANGE)
2603                 ipv6_mc_unmap(idev);
2604 }
2605
2606 static int addrconf_ifdown(struct net_device *dev, int how)
2607 {
2608         struct inet6_dev *idev;
2609         struct inet6_ifaddr *ifa, *keep_list, **bifa;
2610         struct net *net = dev_net(dev);
2611
2612         ASSERT_RTNL();
2613
2614         rt6_ifdown(net, dev);
2615         neigh_ifdown(&nd_tbl, dev);
2616
2617         idev = __in6_dev_get(dev);
2618         if (idev == NULL)
2619                 return -ENODEV;
2620
2621         /* Step 1: remove reference to ipv6 device from parent device.
2622                    Do not dev_put!
2623          */
2624         if (how) {
2625                 idev->dead = 1;
2626
2627                 /* protected by rtnl_lock */
2628                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2629
2630                 /* Step 1.5: remove snmp6 entry */
2631                 snmp6_unregister_dev(idev);
2632
2633         }
2634
2635         write_lock_bh(&idev->lock);
2636
2637         /* Step 3: clear flags for stateless addrconf */
2638         if (!how)
2639                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2640
2641         /* Step 4: clear address list */
2642 #ifdef CONFIG_IPV6_PRIVACY
2643         if (how && del_timer(&idev->regen_timer))
2644                 in6_dev_put(idev);
2645
2646         /* clear tempaddr list */
2647         while (!list_empty(&idev->tempaddr_list)) {
2648                 ifa = list_first_entry(&idev->tempaddr_list,
2649                                        struct inet6_ifaddr, tmp_list);
2650                 list_del(&ifa->tmp_list);
2651                 ifa->dead = 1;
2652                 write_unlock_bh(&idev->lock);
2653                 spin_lock_bh(&ifa->lock);
2654
2655                 if (ifa->ifpub) {
2656                         in6_ifa_put(ifa->ifpub);
2657                         ifa->ifpub = NULL;
2658                 }
2659                 spin_unlock_bh(&ifa->lock);
2660                 in6_ifa_put(ifa);
2661                 write_lock_bh(&idev->lock);
2662         }
2663 #endif
2664         keep_list = NULL;
2665         bifa = &keep_list;
2666         while ((ifa = idev->addr_list) != NULL) {
2667                 idev->addr_list = ifa->if_next;
2668                 ifa->if_next = NULL;
2669
2670                 addrconf_del_timer(ifa);
2671
2672                 /* If just doing link down, and address is permanent
2673                    and not link-local, then retain it. */
2674                 if (how == 0 &&
2675                     (ifa->flags&IFA_F_PERMANENT) &&
2676                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2677
2678                         /* Move to holding list */
2679                         *bifa = ifa;
2680                         bifa = &ifa->if_next;
2681
2682                         /* If not doing DAD on this address, just keep it. */
2683                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2684                             idev->cnf.accept_dad <= 0 ||
2685                             (ifa->flags & IFA_F_NODAD))
2686                                 continue;
2687
2688                         /* If it was tentative already, no need to notify */
2689                         if (ifa->flags & IFA_F_TENTATIVE)
2690                                 continue;
2691
2692                         /* Flag it for later restoration when link comes up */
2693                         ifa->flags |= IFA_F_TENTATIVE;
2694                         in6_ifa_hold(ifa);
2695                 } else {
2696                         ifa->dead = 1;
2697                 }
2698                 write_unlock_bh(&idev->lock);
2699
2700                 /* clear hash table */
2701                 write_lock_bh(&addrconf_hash_lock);
2702                 hlist_del_init(&ifa->addr_lst);
2703                 __in6_ifa_put(ifa);
2704                 write_unlock_bh(&addrconf_hash_lock);
2705
2706                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2707                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2708                 in6_ifa_put(ifa);
2709
2710                 write_lock_bh(&idev->lock);
2711         }
2712
2713         idev->addr_list = keep_list;
2714
2715         write_unlock_bh(&idev->lock);
2716
2717         /* Step 5: Discard multicast list */
2718
2719         if (how)
2720                 ipv6_mc_destroy_dev(idev);
2721         else
2722                 ipv6_mc_down(idev);
2723
2724         idev->tstamp = jiffies;
2725
2726         /* Shot the device (if unregistered) */
2727
2728         if (how) {
2729                 addrconf_sysctl_unregister(idev);
2730                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2731                 neigh_ifdown(&nd_tbl, dev);
2732                 in6_dev_put(idev);
2733         }
2734         return 0;
2735 }
2736
2737 static void addrconf_rs_timer(unsigned long data)
2738 {
2739         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2740         struct inet6_dev *idev = ifp->idev;
2741
2742         read_lock(&idev->lock);
2743         if (idev->dead || !(idev->if_flags & IF_READY))
2744                 goto out;
2745
2746         if (idev->cnf.forwarding)
2747                 goto out;
2748
2749         /* Announcement received after solicitation was sent */
2750         if (idev->if_flags & IF_RA_RCVD)
2751                 goto out;
2752
2753         spin_lock(&ifp->lock);
2754         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2755                 /* The wait after the last probe can be shorter */
2756                 addrconf_mod_timer(ifp, AC_RS,
2757                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2758                                    idev->cnf.rtr_solicit_delay :
2759                                    idev->cnf.rtr_solicit_interval);
2760                 spin_unlock(&ifp->lock);
2761
2762                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2763         } else {
2764                 spin_unlock(&ifp->lock);
2765                 /*
2766                  * Note: we do not support deprecated "all on-link"
2767                  * assumption any longer.
2768                  */
2769                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2770                        idev->dev->name);
2771         }
2772
2773 out:
2774         read_unlock(&idev->lock);
2775         in6_ifa_put(ifp);
2776 }
2777
2778 /*
2779  *      Duplicate Address Detection
2780  */
2781 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2782 {
2783         unsigned long rand_num;
2784         struct inet6_dev *idev = ifp->idev;
2785
2786         if (ifp->flags & IFA_F_OPTIMISTIC)
2787                 rand_num = 0;
2788         else
2789                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2790
2791         ifp->probes = idev->cnf.dad_transmits;
2792         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2793 }
2794
2795 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2796 {
2797         struct inet6_dev *idev = ifp->idev;
2798         struct net_device *dev = idev->dev;
2799
2800         addrconf_join_solict(dev, &ifp->addr);
2801
2802         net_srandom(ifp->addr.s6_addr32[3]);
2803
2804         read_lock_bh(&idev->lock);
2805         if (ifp->dead)
2806                 goto out;
2807
2808         spin_lock(&ifp->lock);
2809         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2810             idev->cnf.accept_dad < 1 ||
2811             !(ifp->flags&IFA_F_TENTATIVE) ||
2812             ifp->flags & IFA_F_NODAD) {
2813                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2814                 spin_unlock(&ifp->lock);
2815                 read_unlock_bh(&idev->lock);
2816
2817                 addrconf_dad_completed(ifp);
2818                 return;
2819         }
2820
2821         if (!(idev->if_flags & IF_READY)) {
2822                 spin_unlock(&ifp->lock);
2823                 read_unlock_bh(&idev->lock);
2824                 /*
2825                  * If the device is not ready:
2826                  * - keep it tentative if it is a permanent address.
2827                  * - otherwise, kill it.
2828                  */
2829                 in6_ifa_hold(ifp);
2830                 addrconf_dad_stop(ifp, 0);
2831                 return;
2832         }
2833
2834         /*
2835          * Optimistic nodes can start receiving
2836          * Frames right away
2837          */
2838         if(ifp->flags & IFA_F_OPTIMISTIC)
2839                 ip6_ins_rt(ifp->rt);
2840
2841         addrconf_dad_kick(ifp);
2842         spin_unlock(&ifp->lock);
2843 out:
2844         read_unlock_bh(&idev->lock);
2845 }
2846
2847 static void addrconf_dad_timer(unsigned long data)
2848 {
2849         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2850         struct inet6_dev *idev = ifp->idev;
2851         struct in6_addr mcaddr;
2852
2853         read_lock(&idev->lock);
2854         if (idev->dead || !(idev->if_flags & IF_READY)) {
2855                 read_unlock(&idev->lock);
2856                 goto out;
2857         }
2858
2859         spin_lock(&ifp->lock);
2860         if (ifp->probes == 0) {
2861                 /*
2862                  * DAD was successful
2863                  */
2864
2865                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2866                 spin_unlock(&ifp->lock);
2867                 read_unlock(&idev->lock);
2868
2869                 addrconf_dad_completed(ifp);
2870
2871                 goto out;
2872         }
2873
2874         ifp->probes--;
2875         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2876         spin_unlock(&ifp->lock);
2877         read_unlock(&idev->lock);
2878
2879         /* send a neighbour solicitation for our addr */
2880         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2881         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2882 out:
2883         in6_ifa_put(ifp);
2884 }
2885
2886 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2887 {
2888         struct net_device *     dev = ifp->idev->dev;
2889
2890         /*
2891          *      Configure the address for reception. Now it is valid.
2892          */
2893
2894         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2895
2896         /* If added prefix is link local and forwarding is off,
2897            start sending router solicitations.
2898          */
2899
2900         if (ifp->idev->cnf.forwarding == 0 &&
2901             ifp->idev->cnf.rtr_solicits > 0 &&
2902             (dev->flags&IFF_LOOPBACK) == 0 &&
2903             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2904                 /*
2905                  *      If a host as already performed a random delay
2906                  *      [...] as part of DAD [...] there is no need
2907                  *      to delay again before sending the first RS
2908                  */
2909                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2910
2911                 spin_lock_bh(&ifp->lock);
2912                 ifp->probes = 1;
2913                 ifp->idev->if_flags |= IF_RS_SENT;
2914                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2915                 spin_unlock_bh(&ifp->lock);
2916         }
2917 }
2918
2919 static void addrconf_dad_run(struct inet6_dev *idev) {
2920         struct inet6_ifaddr *ifp;
2921
2922         read_lock_bh(&idev->lock);
2923         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2924                 spin_lock(&ifp->lock);
2925                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2926                         spin_unlock(&ifp->lock);
2927                         continue;
2928                 }
2929                 spin_unlock(&ifp->lock);
2930                 addrconf_dad_kick(ifp);
2931         }
2932         read_unlock_bh(&idev->lock);
2933 }
2934
2935 #ifdef CONFIG_PROC_FS
2936 struct if6_iter_state {
2937         struct seq_net_private p;
2938         int bucket;
2939 };
2940
2941 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2942 {
2943         struct inet6_ifaddr *ifa = NULL;
2944         struct if6_iter_state *state = seq->private;
2945         struct net *net = seq_file_net(seq);
2946
2947         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2948                 struct hlist_node *n;
2949                 hlist_for_each_entry(ifa, n,
2950                                      &inet6_addr_lst[state->bucket], addr_lst) {
2951                         if (net_eq(dev_net(ifa->idev->dev), net))
2952                                 return ifa;
2953                 }
2954         }
2955         return NULL;
2956 }
2957
2958 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2959                                          struct inet6_ifaddr *ifa)
2960 {
2961         struct if6_iter_state *state = seq->private;
2962         struct net *net = seq_file_net(seq);
2963         struct hlist_node *n = &ifa->addr_lst;
2964
2965         hlist_for_each_entry_continue(ifa, n, addr_lst) {
2966                 if (net_eq(dev_net(ifa->idev->dev), net))
2967                         return ifa;
2968         }
2969
2970         while (++state->bucket < IN6_ADDR_HSIZE) {
2971                 hlist_for_each_entry(ifa, n,
2972                                      &inet6_addr_lst[state->bucket], addr_lst) {
2973                         if (net_eq(dev_net(ifa->idev->dev), net))
2974                                 return ifa;
2975                 }
2976         }
2977
2978         return NULL;
2979 }
2980
2981 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2982 {
2983         struct inet6_ifaddr *ifa = if6_get_first(seq);
2984
2985         if (ifa)
2986                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2987                         --pos;
2988         return pos ? NULL : ifa;
2989 }
2990
2991 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2992         __acquires(addrconf_hash_lock)
2993 {
2994         read_lock_bh(&addrconf_hash_lock);
2995         return if6_get_idx(seq, *pos);
2996 }
2997
2998 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2999 {
3000         struct inet6_ifaddr *ifa;
3001
3002         ifa = if6_get_next(seq, v);
3003         ++*pos;
3004         return ifa;
3005 }
3006
3007 static void if6_seq_stop(struct seq_file *seq, void *v)
3008         __releases(addrconf_hash_lock)
3009 {
3010         read_unlock_bh(&addrconf_hash_lock);
3011 }
3012
3013 static int if6_seq_show(struct seq_file *seq, void *v)
3014 {
3015         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3016         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3017                    &ifp->addr,
3018                    ifp->idev->dev->ifindex,
3019                    ifp->prefix_len,
3020                    ifp->scope,
3021                    ifp->flags,
3022                    ifp->idev->dev->name);
3023         return 0;
3024 }
3025
3026 static const struct seq_operations if6_seq_ops = {
3027         .start  = if6_seq_start,
3028         .next   = if6_seq_next,
3029         .show   = if6_seq_show,
3030         .stop   = if6_seq_stop,
3031 };
3032
3033 static int if6_seq_open(struct inode *inode, struct file *file)
3034 {
3035         return seq_open_net(inode, file, &if6_seq_ops,
3036                             sizeof(struct if6_iter_state));
3037 }
3038
3039 static const struct file_operations if6_fops = {
3040         .owner          = THIS_MODULE,
3041         .open           = if6_seq_open,
3042         .read           = seq_read,
3043         .llseek         = seq_lseek,
3044         .release        = seq_release_net,
3045 };
3046
3047 static int __net_init if6_proc_net_init(struct net *net)
3048 {
3049         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3050                 return -ENOMEM;
3051         return 0;
3052 }
3053
3054 static void __net_exit if6_proc_net_exit(struct net *net)
3055 {
3056        proc_net_remove(net, "if_inet6");
3057 }
3058
3059 static struct pernet_operations if6_proc_net_ops = {
3060        .init = if6_proc_net_init,
3061        .exit = if6_proc_net_exit,
3062 };
3063
3064 int __init if6_proc_init(void)
3065 {
3066         return register_pernet_subsys(&if6_proc_net_ops);
3067 }
3068
3069 void if6_proc_exit(void)
3070 {
3071         unregister_pernet_subsys(&if6_proc_net_ops);
3072 }
3073 #endif  /* CONFIG_PROC_FS */
3074
3075 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3076 /* Check if address is a home address configured on any interface. */
3077 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3078 {
3079         int ret = 0;
3080         struct inet6_ifaddr *ifp = NULL;
3081         struct hlist_node *n;
3082         u8 hash = ipv6_addr_hash(addr);
3083
3084         read_lock_bh(&addrconf_hash_lock);
3085         hlist_for_each_entry(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3086                 if (!net_eq(dev_net(ifp->idev->dev), net))
3087                         continue;
3088                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3089                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3090                         ret = 1;
3091                         break;
3092                 }
3093         }
3094         read_unlock_bh(&addrconf_hash_lock);
3095         return ret;
3096 }
3097 #endif
3098
3099 /*
3100  *      Periodic address status verification
3101  */
3102
3103 static void addrconf_verify(unsigned long foo)
3104 {
3105         struct inet6_ifaddr *ifp;
3106         struct hlist_node *node;
3107         unsigned long now, next;
3108         int i;
3109
3110         spin_lock_bh(&addrconf_verify_lock);
3111         now = jiffies;
3112         next = now + ADDR_CHECK_FREQUENCY;
3113
3114         del_timer(&addr_chk_timer);
3115
3116         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3117
3118 restart:
3119                 read_lock(&addrconf_hash_lock);
3120                 hlist_for_each_entry(ifp, node, &inet6_addr_lst[i], addr_lst) {
3121                         unsigned long age;
3122 #ifdef CONFIG_IPV6_PRIVACY
3123                         unsigned long regen_advance;
3124 #endif
3125
3126                         if (ifp->flags & IFA_F_PERMANENT)
3127                                 continue;
3128
3129                         spin_lock(&ifp->lock);
3130                         age = (now - ifp->tstamp) / HZ;
3131
3132 #ifdef CONFIG_IPV6_PRIVACY
3133                         regen_advance = ifp->idev->cnf.regen_max_retry *
3134                                         ifp->idev->cnf.dad_transmits *
3135                                         ifp->idev->nd_parms->retrans_time / HZ;
3136 #endif
3137
3138                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3139                             age >= ifp->valid_lft) {
3140                                 spin_unlock(&ifp->lock);
3141                                 in6_ifa_hold(ifp);
3142                                 read_unlock(&addrconf_hash_lock);
3143                                 ipv6_del_addr(ifp);
3144                                 goto restart;
3145                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3146                                 spin_unlock(&ifp->lock);
3147                                 continue;
3148                         } else if (age >= ifp->prefered_lft) {
3149                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3150                                 int deprecate = 0;
3151
3152                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3153                                         deprecate = 1;
3154                                         ifp->flags |= IFA_F_DEPRECATED;
3155                                 }
3156
3157                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3158                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3159
3160                                 spin_unlock(&ifp->lock);
3161
3162                                 if (deprecate) {
3163                                         in6_ifa_hold(ifp);
3164                                         read_unlock(&addrconf_hash_lock);
3165
3166                                         ipv6_ifa_notify(0, ifp);
3167                                         in6_ifa_put(ifp);
3168                                         goto restart;
3169                                 }
3170 #ifdef CONFIG_IPV6_PRIVACY
3171                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3172                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3173                                 if (age >= ifp->prefered_lft - regen_advance) {
3174                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3175                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3176                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3177                                         if (!ifp->regen_count && ifpub) {
3178                                                 ifp->regen_count++;
3179                                                 in6_ifa_hold(ifp);
3180                                                 in6_ifa_hold(ifpub);
3181                                                 spin_unlock(&ifp->lock);
3182                                                 read_unlock(&addrconf_hash_lock);
3183                                                 spin_lock(&ifpub->lock);
3184                                                 ifpub->regen_count = 0;
3185                                                 spin_unlock(&ifpub->lock);
3186                                                 ipv6_create_tempaddr(ifpub, ifp);
3187                                                 in6_ifa_put(ifpub);
3188                                                 in6_ifa_put(ifp);
3189                                                 goto restart;
3190                                         }
3191                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3192                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3193                                 spin_unlock(&ifp->lock);
3194 #endif
3195                         } else {
3196                                 /* ifp->prefered_lft <= ifp->valid_lft */
3197                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3198                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3199                                 spin_unlock(&ifp->lock);
3200                         }
3201                 }
3202                 read_unlock(&addrconf_hash_lock);
3203         }
3204
3205         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3206         add_timer(&addr_chk_timer);
3207         spin_unlock_bh(&addrconf_verify_lock);
3208 }
3209
3210 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3211 {
3212         struct in6_addr *pfx = NULL;
3213
3214         if (addr)
3215                 pfx = nla_data(addr);
3216
3217         if (local) {
3218                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3219                         pfx = NULL;
3220                 else
3221                         pfx = nla_data(local);
3222         }
3223
3224         return pfx;
3225 }
3226
3227 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3228         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3229         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3230         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3231 };
3232
3233 static int
3234 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3235 {
3236         struct net *net = sock_net(skb->sk);
3237         struct ifaddrmsg *ifm;
3238         struct nlattr *tb[IFA_MAX+1];
3239         struct in6_addr *pfx;
3240         int err;
3241
3242         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3243         if (err < 0)
3244                 return err;
3245
3246         ifm = nlmsg_data(nlh);
3247         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3248         if (pfx == NULL)
3249                 return -EINVAL;
3250
3251         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3252 }
3253
3254 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3255                              u32 prefered_lft, u32 valid_lft)
3256 {
3257         u32 flags;
3258         clock_t expires;
3259         unsigned long timeout;
3260
3261         if (!valid_lft || (prefered_lft > valid_lft))
3262                 return -EINVAL;
3263
3264         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3265         if (addrconf_finite_timeout(timeout)) {
3266                 expires = jiffies_to_clock_t(timeout * HZ);
3267                 valid_lft = timeout;
3268                 flags = RTF_EXPIRES;
3269         } else {
3270                 expires = 0;
3271                 flags = 0;
3272                 ifa_flags |= IFA_F_PERMANENT;
3273         }
3274
3275         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3276         if (addrconf_finite_timeout(timeout)) {
3277                 if (timeout == 0)
3278                         ifa_flags |= IFA_F_DEPRECATED;
3279                 prefered_lft = timeout;
3280         }
3281
3282         spin_lock_bh(&ifp->lock);
3283         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3284         ifp->tstamp = jiffies;
3285         ifp->valid_lft = valid_lft;
3286         ifp->prefered_lft = prefered_lft;
3287
3288         spin_unlock_bh(&ifp->lock);
3289         if (!(ifp->flags&IFA_F_TENTATIVE))
3290                 ipv6_ifa_notify(0, ifp);
3291
3292         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3293                               expires, flags);
3294         addrconf_verify(0);
3295
3296         return 0;
3297 }
3298
3299 static int
3300 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3301 {
3302         struct net *net = sock_net(skb->sk);
3303         struct ifaddrmsg *ifm;
3304         struct nlattr *tb[IFA_MAX+1];
3305         struct in6_addr *pfx;
3306         struct inet6_ifaddr *ifa;
3307         struct net_device *dev;
3308         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3309         u8 ifa_flags;
3310         int err;
3311
3312         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3313         if (err < 0)
3314                 return err;
3315
3316         ifm = nlmsg_data(nlh);
3317         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3318         if (pfx == NULL)
3319                 return -EINVAL;
3320
3321         if (tb[IFA_CACHEINFO]) {
3322                 struct ifa_cacheinfo *ci;
3323
3324                 ci = nla_data(tb[IFA_CACHEINFO]);
3325                 valid_lft = ci->ifa_valid;
3326                 preferred_lft = ci->ifa_prefered;
3327         } else {
3328                 preferred_lft = INFINITY_LIFE_TIME;
3329                 valid_lft = INFINITY_LIFE_TIME;
3330         }
3331
3332         dev =  __dev_get_by_index(net, ifm->ifa_index);
3333         if (dev == NULL)
3334                 return -ENODEV;
3335
3336         /* We ignore other flags so far. */
3337         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3338
3339         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3340         if (ifa == NULL) {
3341                 /*
3342                  * It would be best to check for !NLM_F_CREATE here but
3343                  * userspace alreay relies on not having to provide this.
3344                  */
3345                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3346                                       ifm->ifa_prefixlen, ifa_flags,
3347                                       preferred_lft, valid_lft);
3348         }
3349
3350         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3351             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3352                 err = -EEXIST;
3353         else
3354                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3355
3356         in6_ifa_put(ifa);
3357
3358         return err;
3359 }
3360
3361 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3362                           u8 scope, int ifindex)
3363 {
3364         struct ifaddrmsg *ifm;
3365
3366         ifm = nlmsg_data(nlh);
3367         ifm->ifa_family = AF_INET6;
3368         ifm->ifa_prefixlen = prefixlen;
3369         ifm->ifa_flags = flags;
3370         ifm->ifa_scope = scope;
3371         ifm->ifa_index = ifindex;
3372 }
3373
3374 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3375                          unsigned long tstamp, u32 preferred, u32 valid)
3376 {
3377         struct ifa_cacheinfo ci;
3378
3379         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3380                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3381         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3382                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3383         ci.ifa_prefered = preferred;
3384         ci.ifa_valid = valid;
3385
3386         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3387 }
3388
3389 static inline int rt_scope(int ifa_scope)
3390 {
3391         if (ifa_scope & IFA_HOST)
3392                 return RT_SCOPE_HOST;
3393         else if (ifa_scope & IFA_LINK)
3394                 return RT_SCOPE_LINK;
3395         else if (ifa_scope & IFA_SITE)
3396                 return RT_SCOPE_SITE;
3397         else
3398                 return RT_SCOPE_UNIVERSE;
3399 }
3400
3401 static inline int inet6_ifaddr_msgsize(void)
3402 {
3403         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3404                + nla_total_size(16) /* IFA_ADDRESS */
3405                + nla_total_size(sizeof(struct ifa_cacheinfo));
3406 }
3407
3408 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3409                              u32 pid, u32 seq, int event, unsigned int flags)
3410 {
3411         struct nlmsghdr  *nlh;
3412         u32 preferred, valid;
3413
3414         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3415         if (nlh == NULL)
3416                 return -EMSGSIZE;
3417
3418         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3419                       ifa->idev->dev->ifindex);
3420
3421         if (!(ifa->flags&IFA_F_PERMANENT)) {
3422                 preferred = ifa->prefered_lft;
3423                 valid = ifa->valid_lft;
3424                 if (preferred != INFINITY_LIFE_TIME) {
3425                         long tval = (jiffies - ifa->tstamp)/HZ;
3426                         if (preferred > tval)
3427                                 preferred -= tval;
3428                         else
3429                                 preferred = 0;
3430                         if (valid != INFINITY_LIFE_TIME)
3431                                 valid -= tval;
3432                 }
3433         } else {
3434                 preferred = INFINITY_LIFE_TIME;
3435                 valid = INFINITY_LIFE_TIME;
3436         }
3437
3438         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3439             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3440                 nlmsg_cancel(skb, nlh);
3441                 return -EMSGSIZE;
3442         }
3443
3444         return nlmsg_end(skb, nlh);
3445 }
3446
3447 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3448                                 u32 pid, u32 seq, int event, u16 flags)
3449 {
3450         struct nlmsghdr  *nlh;
3451         u8 scope = RT_SCOPE_UNIVERSE;
3452         int ifindex = ifmca->idev->dev->ifindex;
3453
3454         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3455                 scope = RT_SCOPE_SITE;
3456
3457         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3458         if (nlh == NULL)
3459                 return -EMSGSIZE;
3460
3461         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3462         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3463             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3464                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3465                 nlmsg_cancel(skb, nlh);
3466                 return -EMSGSIZE;
3467         }
3468
3469         return nlmsg_end(skb, nlh);
3470 }
3471
3472 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3473                                 u32 pid, u32 seq, int event, unsigned int flags)
3474 {
3475         struct nlmsghdr  *nlh;
3476         u8 scope = RT_SCOPE_UNIVERSE;
3477         int ifindex = ifaca->aca_idev->dev->ifindex;
3478
3479         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3480                 scope = RT_SCOPE_SITE;
3481
3482         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3483         if (nlh == NULL)
3484                 return -EMSGSIZE;
3485
3486         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3487         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3488             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3489                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3490                 nlmsg_cancel(skb, nlh);
3491                 return -EMSGSIZE;
3492         }
3493
3494         return nlmsg_end(skb, nlh);
3495 }
3496
3497 enum addr_type_t
3498 {
3499         UNICAST_ADDR,
3500         MULTICAST_ADDR,
3501         ANYCAST_ADDR,
3502 };
3503
3504 /* called with rcu_read_lock() */
3505 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3506                           struct netlink_callback *cb, enum addr_type_t type,
3507                           int s_ip_idx, int *p_ip_idx)
3508 {
3509         struct inet6_ifaddr *ifa;
3510         struct ifmcaddr6 *ifmca;
3511         struct ifacaddr6 *ifaca;
3512         int err = 1;
3513         int ip_idx = *p_ip_idx;
3514
3515         read_lock_bh(&idev->lock);
3516         switch (type) {
3517         case UNICAST_ADDR:
3518                 /* unicast address incl. temp addr */
3519                 for (ifa = idev->addr_list; ifa;
3520                      ifa = ifa->if_next, ip_idx++) {
3521                         if (ip_idx < s_ip_idx)
3522                                 continue;
3523                         err = inet6_fill_ifaddr(skb, ifa,
3524                                                 NETLINK_CB(cb->skb).pid,
3525                                                 cb->nlh->nlmsg_seq,
3526                                                 RTM_NEWADDR,
3527                                                 NLM_F_MULTI);
3528                         if (err <= 0)
3529                                 break;
3530                 }
3531                 break;
3532         case MULTICAST_ADDR:
3533                 /* multicast address */
3534                 for (ifmca = idev->mc_list; ifmca;
3535                      ifmca = ifmca->next, ip_idx++) {
3536                         if (ip_idx < s_ip_idx)
3537                                 continue;
3538                         err = inet6_fill_ifmcaddr(skb, ifmca,
3539                                                   NETLINK_CB(cb->skb).pid,
3540                                                   cb->nlh->nlmsg_seq,
3541                                                   RTM_GETMULTICAST,
3542                                                   NLM_F_MULTI);
3543                         if (err <= 0)
3544                                 break;
3545                 }
3546                 break;
3547         case ANYCAST_ADDR:
3548                 /* anycast address */
3549                 for (ifaca = idev->ac_list; ifaca;
3550                      ifaca = ifaca->aca_next, ip_idx++) {
3551                         if (ip_idx < s_ip_idx)
3552                                 continue;
3553                         err = inet6_fill_ifacaddr(skb, ifaca,
3554                                                   NETLINK_CB(cb->skb).pid,
3555                                                   cb->nlh->nlmsg_seq,
3556                                                   RTM_GETANYCAST,
3557                                                   NLM_F_MULTI);
3558                         if (err <= 0)
3559                                 break;
3560                 }
3561                 break;
3562         default:
3563                 break;
3564         }
3565         read_unlock_bh(&idev->lock);
3566         *p_ip_idx = ip_idx;
3567         return err;
3568 }
3569
3570 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3571                            enum addr_type_t type)
3572 {
3573         struct net *net = sock_net(skb->sk);
3574         int h, s_h;
3575         int idx, ip_idx;
3576         int s_idx, s_ip_idx;
3577         struct net_device *dev;
3578         struct inet6_dev *idev;
3579         struct hlist_head *head;
3580         struct hlist_node *node;
3581
3582         s_h = cb->args[0];
3583         s_idx = idx = cb->args[1];
3584         s_ip_idx = ip_idx = cb->args[2];
3585
3586         rcu_read_lock();
3587         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3588                 idx = 0;
3589                 head = &net->dev_index_head[h];
3590                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3591                         if (idx < s_idx)
3592                                 goto cont;
3593                         if (idx > s_idx)
3594                                 s_ip_idx = 0;
3595                         ip_idx = 0;
3596                         if ((idev = __in6_dev_get(dev)) == NULL)
3597                                 goto cont;
3598
3599                         if (in6_dump_addrs(idev, skb, cb, type,
3600                                            s_ip_idx, &ip_idx) <= 0)
3601                                 goto done;
3602 cont:
3603                         idx++;
3604                 }
3605         }
3606 done:
3607         rcu_read_unlock();
3608         cb->args[0] = h;
3609         cb->args[1] = idx;
3610         cb->args[2] = ip_idx;
3611
3612         return skb->len;
3613 }
3614
3615 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3616 {
3617         enum addr_type_t type = UNICAST_ADDR;
3618
3619         return inet6_dump_addr(skb, cb, type);
3620 }
3621
3622 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3623 {
3624         enum addr_type_t type = MULTICAST_ADDR;
3625
3626         return inet6_dump_addr(skb, cb, type);
3627 }
3628
3629
3630 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3631 {
3632         enum addr_type_t type = ANYCAST_ADDR;
3633
3634         return inet6_dump_addr(skb, cb, type);
3635 }
3636
3637 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3638                              void *arg)
3639 {
3640         struct net *net = sock_net(in_skb->sk);
3641         struct ifaddrmsg *ifm;
3642         struct nlattr *tb[IFA_MAX+1];
3643         struct in6_addr *addr = NULL;
3644         struct net_device *dev = NULL;
3645         struct inet6_ifaddr *ifa;
3646         struct sk_buff *skb;
3647         int err;
3648
3649         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3650         if (err < 0)
3651                 goto errout;
3652
3653         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3654         if (addr == NULL) {
3655                 err = -EINVAL;
3656                 goto errout;
3657         }
3658
3659         ifm = nlmsg_data(nlh);
3660         if (ifm->ifa_index)
3661                 dev = __dev_get_by_index(net, ifm->ifa_index);
3662
3663         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3664                 err = -EADDRNOTAVAIL;
3665                 goto errout;
3666         }
3667
3668         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3669                 err = -ENOBUFS;
3670                 goto errout_ifa;
3671         }
3672
3673         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3674                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3675         if (err < 0) {
3676                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3677                 WARN_ON(err == -EMSGSIZE);
3678                 kfree_skb(skb);
3679                 goto errout_ifa;
3680         }
3681         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3682 errout_ifa:
3683         in6_ifa_put(ifa);
3684 errout:
3685         return err;
3686 }
3687
3688 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3689 {
3690         struct sk_buff *skb;
3691         struct net *net = dev_net(ifa->idev->dev);
3692         int err = -ENOBUFS;
3693
3694         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3695         if (skb == NULL)
3696                 goto errout;
3697
3698         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3699         if (err < 0) {
3700                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3701                 WARN_ON(err == -EMSGSIZE);
3702                 kfree_skb(skb);
3703                 goto errout;
3704         }
3705         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3706         return;
3707 errout:
3708         if (err < 0)
3709                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3710 }
3711
3712 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3713                                 __s32 *array, int bytes)
3714 {
3715         BUG_ON(bytes < (DEVCONF_MAX * 4));
3716
3717         memset(array, 0, bytes);
3718         array[DEVCONF_FORWARDING] = cnf->forwarding;
3719         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3720         array[DEVCONF_MTU6] = cnf->mtu6;
3721         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3722         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3723         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3724         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3725         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3726         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3727         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3728         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3729 #ifdef CONFIG_IPV6_PRIVACY
3730         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3731         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3732         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3733         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3734         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3735 #endif
3736         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3737         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3738         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3739 #ifdef CONFIG_IPV6_ROUTER_PREF
3740         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3741         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3742 #ifdef CONFIG_IPV6_ROUTE_INFO
3743         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3744 #endif
3745 #endif
3746         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3747         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3748 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3749         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3750 #endif
3751 #ifdef CONFIG_IPV6_MROUTE
3752         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3753 #endif
3754         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3755         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3756         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3757 }
3758
3759 static inline size_t inet6_if_nlmsg_size(void)
3760 {
3761         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3762                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3763                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3764                + nla_total_size(4) /* IFLA_MTU */
3765                + nla_total_size(4) /* IFLA_LINK */
3766                + nla_total_size( /* IFLA_PROTINFO */
3767                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3768                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3769                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3770                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3771                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3772                  );
3773 }
3774
3775 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3776                                       int items, int bytes)
3777 {
3778         int i;
3779         int pad = bytes - sizeof(u64) * items;
3780         BUG_ON(pad < 0);
3781
3782         /* Use put_unaligned() because stats may not be aligned for u64. */
3783         put_unaligned(items, &stats[0]);
3784         for (i = 1; i < items; i++)
3785                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3786
3787         memset(&stats[items], 0, pad);
3788 }
3789
3790 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3791                              int bytes)
3792 {
3793         switch(attrtype) {
3794         case IFLA_INET6_STATS:
3795                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3796                 break;
3797         case IFLA_INET6_ICMP6STATS:
3798                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3799                 break;
3800         }
3801 }
3802
3803 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3804                              u32 pid, u32 seq, int event, unsigned int flags)
3805 {
3806         struct net_device *dev = idev->dev;
3807         struct nlattr *nla;
3808         struct ifinfomsg *hdr;
3809         struct nlmsghdr *nlh;
3810         void *protoinfo;
3811         struct ifla_cacheinfo ci;
3812
3813         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3814         if (nlh == NULL)
3815                 return -EMSGSIZE;
3816
3817         hdr = nlmsg_data(nlh);
3818         hdr->ifi_family = AF_INET6;
3819         hdr->__ifi_pad = 0;
3820         hdr->ifi_type = dev->type;
3821         hdr->ifi_index = dev->ifindex;
3822         hdr->ifi_flags = dev_get_flags(dev);
3823         hdr->ifi_change = 0;
3824
3825         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3826
3827         if (dev->addr_len)
3828                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3829
3830         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3831         if (dev->ifindex != dev->iflink)
3832                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3833
3834         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3835         if (protoinfo == NULL)
3836                 goto nla_put_failure;
3837
3838         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3839
3840         ci.max_reasm_len = IPV6_MAXPLEN;
3841         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3842                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3843         ci.reachable_time = idev->nd_parms->reachable_time;
3844         ci.retrans_time = idev->nd_parms->retrans_time;
3845         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3846
3847         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3848         if (nla == NULL)
3849                 goto nla_put_failure;
3850         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3851
3852         /* XXX - MC not implemented */
3853
3854         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3855         if (nla == NULL)
3856                 goto nla_put_failure;
3857         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3858
3859         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3860         if (nla == NULL)
3861                 goto nla_put_failure;
3862         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3863
3864         nla_nest_end(skb, protoinfo);
3865         return nlmsg_end(skb, nlh);
3866
3867 nla_put_failure:
3868         nlmsg_cancel(skb, nlh);
3869         return -EMSGSIZE;
3870 }
3871
3872 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3873 {
3874         struct net *net = sock_net(skb->sk);
3875         int h, s_h;
3876         int idx = 0, s_idx;
3877         struct net_device *dev;
3878         struct inet6_dev *idev;
3879         struct hlist_head *head;
3880         struct hlist_node *node;
3881
3882         s_h = cb->args[0];
3883         s_idx = cb->args[1];
3884
3885         rcu_read_lock();
3886         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3887                 idx = 0;
3888                 head = &net->dev_index_head[h];
3889                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3890                         if (idx < s_idx)
3891                                 goto cont;
3892                         idev = __in6_dev_get(dev);
3893                         if (!idev)
3894                                 goto cont;
3895                         if (inet6_fill_ifinfo(skb, idev,
3896                                               NETLINK_CB(cb->skb).pid,
3897                                               cb->nlh->nlmsg_seq,
3898                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3899                                 goto out;
3900 cont:
3901                         idx++;
3902                 }
3903         }
3904 out:
3905         rcu_read_unlock();
3906         cb->args[1] = idx;
3907         cb->args[0] = h;
3908
3909         return skb->len;
3910 }
3911
3912 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3913 {
3914         struct sk_buff *skb;
3915         struct net *net = dev_net(idev->dev);
3916         int err = -ENOBUFS;
3917
3918         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3919         if (skb == NULL)
3920                 goto errout;
3921
3922         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3923         if (err < 0) {
3924                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3925                 WARN_ON(err == -EMSGSIZE);
3926                 kfree_skb(skb);
3927                 goto errout;
3928         }
3929         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3930         return;
3931 errout:
3932         if (err < 0)
3933                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3934 }
3935
3936 static inline size_t inet6_prefix_nlmsg_size(void)
3937 {
3938         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3939                + nla_total_size(sizeof(struct in6_addr))
3940                + nla_total_size(sizeof(struct prefix_cacheinfo));
3941 }
3942
3943 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3944                              struct prefix_info *pinfo, u32 pid, u32 seq,
3945                              int event, unsigned int flags)
3946 {
3947         struct prefixmsg *pmsg;
3948         struct nlmsghdr *nlh;
3949         struct prefix_cacheinfo ci;
3950
3951         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3952         if (nlh == NULL)
3953                 return -EMSGSIZE;
3954
3955         pmsg = nlmsg_data(nlh);
3956         pmsg->prefix_family = AF_INET6;
3957         pmsg->prefix_pad1 = 0;
3958         pmsg->prefix_pad2 = 0;
3959         pmsg->prefix_ifindex = idev->dev->ifindex;
3960         pmsg->prefix_len = pinfo->prefix_len;
3961         pmsg->prefix_type = pinfo->type;
3962         pmsg->prefix_pad3 = 0;
3963         pmsg->prefix_flags = 0;
3964         if (pinfo->onlink)
3965                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3966         if (pinfo->autoconf)
3967                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3968
3969         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3970
3971         ci.preferred_time = ntohl(pinfo->prefered);
3972         ci.valid_time = ntohl(pinfo->valid);
3973         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3974
3975         return nlmsg_end(skb, nlh);
3976
3977 nla_put_failure:
3978         nlmsg_cancel(skb, nlh);
3979         return -EMSGSIZE;
3980 }
3981
3982 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3983                          struct prefix_info *pinfo)
3984 {
3985         struct sk_buff *skb;
3986         struct net *net = dev_net(idev->dev);
3987         int err = -ENOBUFS;
3988
3989         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3990         if (skb == NULL)
3991                 goto errout;
3992
3993         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3994         if (err < 0) {
3995                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3996                 WARN_ON(err == -EMSGSIZE);
3997                 kfree_skb(skb);
3998                 goto errout;
3999         }
4000         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4001         return;
4002 errout:
4003         if (err < 0)
4004                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4005 }
4006
4007 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4008 {
4009         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4010
4011         switch (event) {
4012         case RTM_NEWADDR:
4013                 /*
4014                  * If the address was optimistic
4015                  * we inserted the route at the start of
4016                  * our DAD process, so we don't need
4017                  * to do it again
4018                  */
4019                 if (!(ifp->rt->rt6i_node))
4020                         ip6_ins_rt(ifp->rt);
4021                 if (ifp->idev->cnf.forwarding)
4022                         addrconf_join_anycast(ifp);
4023                 break;
4024         case RTM_DELADDR:
4025                 if (ifp->idev->cnf.forwarding)
4026                         addrconf_leave_anycast(ifp);
4027                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4028                 dst_hold(&ifp->rt->u.dst);
4029                 if (ip6_del_rt(ifp->rt))
4030                         dst_free(&ifp->rt->u.dst);
4031                 break;
4032         }
4033 }
4034
4035 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4036 {
4037         rcu_read_lock_bh();
4038         if (likely(ifp->idev->dead == 0))
4039                 __ipv6_ifa_notify(event, ifp);
4040         rcu_read_unlock_bh();
4041 }
4042
4043 #ifdef CONFIG_SYSCTL
4044
4045 static
4046 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4047                            void __user *buffer, size_t *lenp, loff_t *ppos)
4048 {
4049         int *valp = ctl->data;
4050         int val = *valp;
4051         loff_t pos = *ppos;
4052         int ret;
4053
4054         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4055
4056         if (write)
4057                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4058         if (ret)
4059                 *ppos = pos;
4060         return ret;
4061 }
4062
4063 static void dev_disable_change(struct inet6_dev *idev)
4064 {
4065         if (!idev || !idev->dev)
4066                 return;
4067
4068         if (idev->cnf.disable_ipv6)
4069                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4070         else
4071                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4072 }
4073
4074 static void addrconf_disable_change(struct net *net, __s32 newf)
4075 {
4076         struct net_device *dev;
4077         struct inet6_dev *idev;
4078
4079         rcu_read_lock();
4080         for_each_netdev_rcu(net, dev) {
4081                 idev = __in6_dev_get(dev);
4082                 if (idev) {
4083                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4084                         idev->cnf.disable_ipv6 = newf;
4085                         if (changed)
4086                                 dev_disable_change(idev);
4087                 }
4088         }
4089         rcu_read_unlock();
4090 }
4091
4092 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4093 {
4094         struct net *net;
4095
4096         net = (struct net *)table->extra2;
4097
4098         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4099                 return 0;
4100
4101         if (!rtnl_trylock()) {
4102                 /* Restore the original values before restarting */
4103                 *p = old;
4104                 return restart_syscall();
4105         }
4106
4107         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4108                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4109                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4110                 addrconf_disable_change(net, newf);
4111         } else if ((!*p) ^ (!old))
4112                 dev_disable_change((struct inet6_dev *)table->extra1);
4113
4114         rtnl_unlock();
4115         return 0;
4116 }
4117
4118 static
4119 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4120                             void __user *buffer, size_t *lenp, loff_t *ppos)
4121 {
4122         int *valp = ctl->data;
4123         int val = *valp;
4124         loff_t pos = *ppos;
4125         int ret;
4126
4127         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4128
4129         if (write)
4130                 ret = addrconf_disable_ipv6(ctl, valp, val);
4131         if (ret)
4132                 *ppos = pos;
4133         return ret;
4134 }
4135
4136 static struct addrconf_sysctl_table
4137 {
4138         struct ctl_table_header *sysctl_header;
4139         ctl_table addrconf_vars[DEVCONF_MAX+1];
4140         char *dev_name;
4141 } addrconf_sysctl __read_mostly = {
4142         .sysctl_header = NULL,
4143         .addrconf_vars = {
4144                 {
4145                         .procname       =       "forwarding",
4146                         .data           =       &ipv6_devconf.forwarding,
4147                         .maxlen         =       sizeof(int),
4148                         .mode           =       0644,
4149                         .proc_handler   =       addrconf_sysctl_forward,
4150                 },
4151                 {
4152                         .procname       =       "hop_limit",
4153                         .data           =       &ipv6_devconf.hop_limit,
4154                         .maxlen         =       sizeof(int),
4155                         .mode           =       0644,
4156                         .proc_handler   =       proc_dointvec,
4157                 },
4158                 {
4159                         .procname       =       "mtu",
4160                         .data           =       &ipv6_devconf.mtu6,
4161                         .maxlen         =       sizeof(int),
4162                         .mode           =       0644,
4163                         .proc_handler   =       proc_dointvec,
4164                 },
4165                 {
4166                         .procname       =       "accept_ra",
4167                         .data           =       &ipv6_devconf.accept_ra,
4168                         .maxlen         =       sizeof(int),
4169                         .mode           =       0644,
4170                         .proc_handler   =       proc_dointvec,
4171                 },
4172                 {
4173                         .procname       =       "accept_redirects",
4174                         .data           =       &ipv6_devconf.accept_redirects,
4175                         .maxlen         =       sizeof(int),
4176                         .mode           =       0644,
4177                         .proc_handler   =       proc_dointvec,
4178                 },
4179                 {
4180                         .procname       =       "autoconf",
4181                         .data           =       &ipv6_devconf.autoconf,
4182                         .maxlen         =       sizeof(int),
4183                         .mode           =       0644,
4184                         .proc_handler   =       proc_dointvec,
4185                 },
4186                 {
4187                         .procname       =       "dad_transmits",
4188                         .data           =       &ipv6_devconf.dad_transmits,
4189                         .maxlen         =       sizeof(int),
4190                         .mode           =       0644,
4191                         .proc_handler   =       proc_dointvec,
4192                 },
4193                 {
4194                         .procname       =       "router_solicitations",
4195                         .data           =       &ipv6_devconf.rtr_solicits,
4196                         .maxlen         =       sizeof(int),
4197                         .mode           =       0644,
4198                         .proc_handler   =       proc_dointvec,
4199                 },
4200                 {
4201                         .procname       =       "router_solicitation_interval",
4202                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4203                         .maxlen         =       sizeof(int),
4204                         .mode           =       0644,
4205                         .proc_handler   =       proc_dointvec_jiffies,
4206                 },
4207                 {
4208                         .procname       =       "router_solicitation_delay",
4209                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4210                         .maxlen         =       sizeof(int),
4211                         .mode           =       0644,
4212                         .proc_handler   =       proc_dointvec_jiffies,
4213                 },
4214                 {
4215                         .procname       =       "force_mld_version",
4216                         .data           =       &ipv6_devconf.force_mld_version,
4217                         .maxlen         =       sizeof(int),
4218                         .mode           =       0644,
4219                         .proc_handler   =       proc_dointvec,
4220                 },
4221 #ifdef CONFIG_IPV6_PRIVACY
4222                 {
4223                         .procname       =       "use_tempaddr",
4224                         .data           =       &ipv6_devconf.use_tempaddr,
4225                         .maxlen         =       sizeof(int),
4226                         .mode           =       0644,
4227                         .proc_handler   =       proc_dointvec,
4228                 },
4229                 {
4230                         .procname       =       "temp_valid_lft",
4231                         .data           =       &ipv6_devconf.temp_valid_lft,
4232                         .maxlen         =       sizeof(int),
4233                         .mode           =       0644,
4234                         .proc_handler   =       proc_dointvec,
4235                 },
4236                 {
4237                         .procname       =       "temp_prefered_lft",
4238                         .data           =       &ipv6_devconf.temp_prefered_lft,
4239                         .maxlen         =       sizeof(int),
4240                         .mode           =       0644,
4241                         .proc_handler   =       proc_dointvec,
4242                 },
4243                 {
4244                         .procname       =       "regen_max_retry",
4245                         .data           =       &ipv6_devconf.regen_max_retry,
4246                         .maxlen         =       sizeof(int),
4247                         .mode           =       0644,
4248                         .proc_handler   =       proc_dointvec,
4249                 },
4250                 {
4251                         .procname       =       "max_desync_factor",
4252                         .data           =       &ipv6_devconf.max_desync_factor,
4253                         .maxlen         =       sizeof(int),
4254                         .mode           =       0644,
4255                         .proc_handler   =       proc_dointvec,
4256                 },
4257 #endif
4258                 {
4259                         .procname       =       "max_addresses",
4260                         .data           =       &ipv6_devconf.max_addresses,
4261                         .maxlen         =       sizeof(int),
4262                         .mode           =       0644,
4263                         .proc_handler   =       proc_dointvec,
4264                 },
4265                 {
4266                         .procname       =       "accept_ra_defrtr",
4267                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4268                         .maxlen         =       sizeof(int),
4269                         .mode           =       0644,
4270                         .proc_handler   =       proc_dointvec,
4271                 },
4272                 {
4273                         .procname       =       "accept_ra_pinfo",
4274                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4275                         .maxlen         =       sizeof(int),
4276                         .mode           =       0644,
4277                         .proc_handler   =       proc_dointvec,
4278                 },
4279 #ifdef CONFIG_IPV6_ROUTER_PREF
4280                 {
4281                         .procname       =       "accept_ra_rtr_pref",
4282                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4283                         .maxlen         =       sizeof(int),
4284                         .mode           =       0644,
4285                         .proc_handler   =       proc_dointvec,
4286                 },
4287                 {
4288                         .procname       =       "router_probe_interval",
4289                         .data           =       &ipv6_devconf.rtr_probe_interval,
4290                         .maxlen         =       sizeof(int),
4291                         .mode           =       0644,
4292                         .proc_handler   =       proc_dointvec_jiffies,
4293                 },
4294 #ifdef CONFIG_IPV6_ROUTE_INFO
4295                 {
4296                         .procname       =       "accept_ra_rt_info_max_plen",
4297                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4298                         .maxlen         =       sizeof(int),
4299                         .mode           =       0644,
4300                         .proc_handler   =       proc_dointvec,
4301                 },
4302 #endif
4303 #endif
4304                 {
4305                         .procname       =       "proxy_ndp",
4306                         .data           =       &ipv6_devconf.proxy_ndp,
4307                         .maxlen         =       sizeof(int),
4308                         .mode           =       0644,
4309                         .proc_handler   =       proc_dointvec,
4310                 },
4311                 {
4312                         .procname       =       "accept_source_route",
4313                         .data           =       &ipv6_devconf.accept_source_route,
4314                         .maxlen         =       sizeof(int),
4315                         .mode           =       0644,
4316                         .proc_handler   =       proc_dointvec,
4317                 },
4318 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4319                 {
4320                         .procname       =       "optimistic_dad",
4321                         .data           =       &ipv6_devconf.optimistic_dad,
4322                         .maxlen         =       sizeof(int),
4323                         .mode           =       0644,
4324                         .proc_handler   =       proc_dointvec,
4325
4326                 },
4327 #endif
4328 #ifdef CONFIG_IPV6_MROUTE
4329                 {
4330                         .procname       =       "mc_forwarding",
4331                         .data           =       &ipv6_devconf.mc_forwarding,
4332                         .maxlen         =       sizeof(int),
4333                         .mode           =       0444,
4334                         .proc_handler   =       proc_dointvec,
4335                 },
4336 #endif
4337                 {
4338                         .procname       =       "disable_ipv6",
4339                         .data           =       &ipv6_devconf.disable_ipv6,
4340                         .maxlen         =       sizeof(int),
4341                         .mode           =       0644,
4342                         .proc_handler   =       addrconf_sysctl_disable,
4343                 },
4344                 {
4345                         .procname       =       "accept_dad",
4346                         .data           =       &ipv6_devconf.accept_dad,
4347                         .maxlen         =       sizeof(int),
4348                         .mode           =       0644,
4349                         .proc_handler   =       proc_dointvec,
4350                 },
4351                 {
4352                         .procname       = "force_tllao",
4353                         .data           = &ipv6_devconf.force_tllao,
4354                         .maxlen         = sizeof(int),
4355                         .mode           = 0644,
4356                         .proc_handler   = proc_dointvec
4357                 },
4358                 {
4359                         /* sentinel */
4360                 }
4361         },
4362 };
4363
4364 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4365                 struct inet6_dev *idev, struct ipv6_devconf *p)
4366 {
4367         int i;
4368         struct addrconf_sysctl_table *t;
4369
4370 #define ADDRCONF_CTL_PATH_DEV   3
4371
4372         struct ctl_path addrconf_ctl_path[] = {
4373                 { .procname = "net", },
4374                 { .procname = "ipv6", },
4375                 { .procname = "conf", },
4376                 { /* to be set */ },
4377                 { },
4378         };
4379
4380
4381         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4382         if (t == NULL)
4383                 goto out;
4384
4385         for (i=0; t->addrconf_vars[i].data; i++) {
4386                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4387                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4388                 t->addrconf_vars[i].extra2 = net;
4389         }
4390
4391         /*
4392          * Make a copy of dev_name, because '.procname' is regarded as const
4393          * by sysctl and we wouldn't want anyone to change it under our feet
4394          * (see SIOCSIFNAME).
4395          */
4396         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4397         if (!t->dev_name)
4398                 goto free;
4399
4400         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4401
4402         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4403                         t->addrconf_vars);
4404         if (t->sysctl_header == NULL)
4405                 goto free_procname;
4406
4407         p->sysctl = t;
4408         return 0;
4409
4410 free_procname:
4411         kfree(t->dev_name);
4412 free:
4413         kfree(t);
4414 out:
4415         return -ENOBUFS;
4416 }
4417
4418 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4419 {
4420         struct addrconf_sysctl_table *t;
4421
4422         if (p->sysctl == NULL)
4423                 return;
4424
4425         t = p->sysctl;
4426         p->sysctl = NULL;
4427         unregister_sysctl_table(t->sysctl_header);
4428         kfree(t->dev_name);
4429         kfree(t);
4430 }
4431
4432 static void addrconf_sysctl_register(struct inet6_dev *idev)
4433 {
4434         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4435                               &ndisc_ifinfo_sysctl_change);
4436         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4437                                         idev, &idev->cnf);
4438 }
4439
4440 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4441 {
4442         __addrconf_sysctl_unregister(&idev->cnf);
4443         neigh_sysctl_unregister(idev->nd_parms);
4444 }
4445
4446
4447 #endif
4448
4449 static int __net_init addrconf_init_net(struct net *net)
4450 {
4451         int err;
4452         struct ipv6_devconf *all, *dflt;
4453
4454         err = -ENOMEM;
4455         all = &ipv6_devconf;
4456         dflt = &ipv6_devconf_dflt;
4457
4458         if (!net_eq(net, &init_net)) {
4459                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4460                 if (all == NULL)
4461                         goto err_alloc_all;
4462
4463                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4464                 if (dflt == NULL)
4465                         goto err_alloc_dflt;
4466         } else {
4467                 /* these will be inherited by all namespaces */
4468                 dflt->autoconf = ipv6_defaults.autoconf;
4469                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4470         }
4471
4472         net->ipv6.devconf_all = all;
4473         net->ipv6.devconf_dflt = dflt;
4474
4475 #ifdef CONFIG_SYSCTL
4476         err = __addrconf_sysctl_register(net, "all", NULL, all);
4477         if (err < 0)
4478                 goto err_reg_all;
4479
4480         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4481         if (err < 0)
4482                 goto err_reg_dflt;
4483 #endif
4484         return 0;
4485
4486 #ifdef CONFIG_SYSCTL
4487 err_reg_dflt:
4488         __addrconf_sysctl_unregister(all);
4489 err_reg_all:
4490         kfree(dflt);
4491 #endif
4492 err_alloc_dflt:
4493         kfree(all);
4494 err_alloc_all:
4495         return err;
4496 }
4497
4498 static void __net_exit addrconf_exit_net(struct net *net)
4499 {
4500 #ifdef CONFIG_SYSCTL
4501         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4502         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4503 #endif
4504         if (!net_eq(net, &init_net)) {
4505                 kfree(net->ipv6.devconf_dflt);
4506                 kfree(net->ipv6.devconf_all);
4507         }
4508 }
4509
4510 static struct pernet_operations addrconf_ops = {
4511         .init = addrconf_init_net,
4512         .exit = addrconf_exit_net,
4513 };
4514
4515 /*
4516  *      Device notifier
4517  */
4518
4519 int register_inet6addr_notifier(struct notifier_block *nb)
4520 {
4521         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4522 }
4523
4524 EXPORT_SYMBOL(register_inet6addr_notifier);
4525
4526 int unregister_inet6addr_notifier(struct notifier_block *nb)
4527 {
4528         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4529 }
4530
4531 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4532
4533 /*
4534  *      Init / cleanup code
4535  */
4536
4537 int __init addrconf_init(void)
4538 {
4539         int i, err;
4540
4541         if ((err = ipv6_addr_label_init()) < 0) {
4542                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4543                         err);
4544                 return err;
4545         }
4546
4547         register_pernet_subsys(&addrconf_ops);
4548
4549         /* The addrconf netdev notifier requires that loopback_dev
4550          * has it's ipv6 private information allocated and setup
4551          * before it can bring up and give link-local addresses
4552          * to other devices which are up.
4553          *
4554          * Unfortunately, loopback_dev is not necessarily the first
4555          * entry in the global dev_base list of net devices.  In fact,
4556          * it is likely to be the very last entry on that list.
4557          * So this causes the notifier registry below to try and
4558          * give link-local addresses to all devices besides loopback_dev
4559          * first, then loopback_dev, which cases all the non-loopback_dev
4560          * devices to fail to get a link-local address.
4561          *
4562          * So, as a temporary fix, allocate the ipv6 structure for
4563          * loopback_dev first by hand.
4564          * Longer term, all of the dependencies ipv6 has upon the loopback
4565          * device and it being up should be removed.
4566          */
4567         rtnl_lock();
4568         if (!ipv6_add_dev(init_net.loopback_dev))
4569                 err = -ENOMEM;
4570         rtnl_unlock();
4571         if (err)
4572                 goto errlo;
4573
4574         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4575                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4576
4577         register_netdevice_notifier(&ipv6_dev_notf);
4578
4579         addrconf_verify(0);
4580
4581         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4582         if (err < 0)
4583                 goto errout;
4584
4585         /* Only the first call to __rtnl_register can fail */
4586         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4587         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4588         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4589         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4590         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4591
4592         ipv6_addr_label_rtnl_register();
4593
4594         return 0;
4595 errout:
4596         unregister_netdevice_notifier(&ipv6_dev_notf);
4597 errlo:
4598         unregister_pernet_subsys(&addrconf_ops);
4599
4600         return err;
4601 }
4602
4603 void addrconf_cleanup(void)
4604 {
4605         struct net_device *dev;
4606         int i;
4607
4608         unregister_netdevice_notifier(&ipv6_dev_notf);
4609         unregister_pernet_subsys(&addrconf_ops);
4610
4611         rtnl_lock();
4612
4613         /* clean dev list */
4614         for_each_netdev(&init_net, dev) {
4615                 if (__in6_dev_get(dev) == NULL)
4616                         continue;
4617                 addrconf_ifdown(dev, 1);
4618         }
4619         addrconf_ifdown(init_net.loopback_dev, 2);
4620
4621         /*
4622          *      Check hash table.
4623          */
4624         write_lock_bh(&addrconf_hash_lock);
4625         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4626                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4627         write_unlock_bh(&addrconf_hash_lock);
4628
4629         del_timer(&addr_chk_timer);
4630         rtnl_unlock();
4631 }