SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/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_SPINLOCK(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 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 527 {
 528         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 529         kfree(ifp);
 530 }
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(ifp->if_next != NULL);
 536         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 537
 538 #ifdef NET_REFCNT_DEBUG
 539         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 540 #endif
 541
 542         in6_dev_put(ifp->idev);
 543
 544         if (del_timer(&ifp->timer))
 545                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 546
 547         if (!ifp->dead) {
 548                 printk("Freeing alive inet6 address %p\n", ifp);
 549                 return;
 550         }
 551         dst_release(&ifp->rt->u.dst);
 552
 553         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 554 }
 555
 556 static void
 557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 558 {
 559         struct inet6_ifaddr *ifa, **ifap;
 560         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 561
 562         /*
 563          * Each device address list is sorted in order of scope -
 564          * global before linklocal.
 565          */
 566         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 567              ifap = &ifa->if_next) {
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         ifp->if_next = *ifap;
 573         *ifap = ifp;
 574 }
 575
 576 /*
 577  *      Hash function taken from net_alias.c
 578  */
 579 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 580 {
 581         __u32 word;
 582
 583         /*
 584          * We perform the hash function over the last 64 bits of the address
 585          * This will include the IEEE address token on links that support it.
 586          */
 587
 588         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 589         word ^= (word >> 16);
 590         word ^= (word >> 8);
 591
 592         return ((word ^ (word >> 4)) & 0x0f);
 593 }
 594
 595 /* On success it returns ifp with increased reference count */
 596
 597 static struct inet6_ifaddr *
 598 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 599               int scope, u32 flags)
 600 {
 601         struct inet6_ifaddr *ifa = NULL;
 602         struct rt6_info *rt;
 603         int hash;
 604         int err = 0;
 605         int addr_type = ipv6_addr_type(addr);
 606
 607         if (addr_type == IPV6_ADDR_ANY ||
 608             addr_type & IPV6_ADDR_MULTICAST ||
 609             (!(idev->dev->flags & IFF_LOOPBACK) &&
 610              addr_type & IPV6_ADDR_LOOPBACK))
 611                 return ERR_PTR(-EADDRNOTAVAIL);
 612
 613         rcu_read_lock_bh();
 614         if (idev->dead) {
 615                 err = -ENODEV;                  /*XXX*/
 616                 goto out2;
 617         }
 618
 619         if (idev->cnf.disable_ipv6) {
 620                 err = -EACCES;
 621                 goto out2;
 622         }
 623
 624         spin_lock(&addrconf_hash_lock);
 625
 626         /* Ignore adding duplicate addresses on an interface */
 627         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 628                 ADBG(("ipv6_add_addr: already assigned\n"));
 629                 err = -EEXIST;
 630                 goto out;
 631         }
 632
 633         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 634
 635         if (ifa == NULL) {
 636                 ADBG(("ipv6_add_addr: malloc failed\n"));
 637                 err = -ENOBUFS;
 638                 goto out;
 639         }
 640
 641         rt = addrconf_dst_alloc(idev, addr, 0);
 642         if (IS_ERR(rt)) {
 643                 err = PTR_ERR(rt);
 644                 goto out;
 645         }
 646
 647         ipv6_addr_copy(&ifa->addr, addr);
 648
 649         spin_lock_init(&ifa->lock);
 650         init_timer(&ifa->timer);
 651         INIT_HLIST_NODE(&ifa->addr_lst);
 652         ifa->timer.data = (unsigned long) ifa;
 653         ifa->scope = scope;
 654         ifa->prefix_len = pfxlen;
 655         ifa->flags = flags | IFA_F_TENTATIVE;
 656         ifa->cstamp = ifa->tstamp = jiffies;
 657
 658         ifa->rt = rt;
 659
 660         /*
 661          * part one of RFC 4429, section 3.3
 662          * We should not configure an address as
 663          * optimistic if we do not yet know the link
 664          * layer address of our nexhop router
 665          */
 666
 667         if (rt->rt6i_nexthop == NULL)
 668                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 669
 670         ifa->idev = idev;
 671         in6_dev_hold(idev);
 672         /* For caller */
 673         in6_ifa_hold(ifa);
 674
 675         /* Add to big hash table */
 676         hash = ipv6_addr_hash(addr);
 677
 678         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 679         in6_ifa_hold(ifa);
 680         spin_unlock(&addrconf_hash_lock);
 681
 682         write_lock(&idev->lock);
 683         /* Add to inet6_dev unicast addr list. */
 684         ipv6_link_dev_addr(idev, ifa);
 685
 686 #ifdef CONFIG_IPV6_PRIVACY
 687         if (ifa->flags&IFA_F_TEMPORARY) {
 688                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 689                 in6_ifa_hold(ifa);
 690         }
 691 #endif
 692
 693         in6_ifa_hold(ifa);
 694         write_unlock(&idev->lock);
 695 out2:
 696         rcu_read_unlock_bh();
 697
 698         if (likely(err == 0))
 699                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 700         else {
 701                 kfree(ifa);
 702                 ifa = ERR_PTR(err);
 703         }
 704
 705         return ifa;
 706 out:
 707         spin_unlock(&addrconf_hash_lock);
 708         goto out2;
 709 }
 710
 711 /* This function wants to get referenced ifp and releases it before return */
 712
 713 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 714 {
 715         struct inet6_ifaddr *ifa, **ifap;
 716         struct inet6_dev *idev = ifp->idev;
 717         int hash;
 718         int deleted = 0, onlink = 0;
 719         unsigned long expires = jiffies;
 720
 721         hash = ipv6_addr_hash(&ifp->addr);
 722
 723         ifp->dead = 1;
 724
 725         spin_lock_bh(&addrconf_hash_lock);
 726         hlist_del_init_rcu(&ifp->addr_lst);
 727         __in6_ifa_put(ifp);
 728         spin_unlock_bh(&addrconf_hash_lock);
 729
 730         write_lock_bh(&idev->lock);
 731 #ifdef CONFIG_IPV6_PRIVACY
 732         if (ifp->flags&IFA_F_TEMPORARY) {
 733                 list_del(&ifp->tmp_list);
 734                 if (ifp->ifpub) {
 735                         in6_ifa_put(ifp->ifpub);
 736                         ifp->ifpub = NULL;
 737                 }
 738                 __in6_ifa_put(ifp);
 739         }
 740 #endif
 741
 742         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 743                 if (ifa == ifp) {
 744                         *ifap = ifa->if_next;
 745                         __in6_ifa_put(ifp);
 746                         ifa->if_next = NULL;
 747                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 748                                 break;
 749                         deleted = 1;
 750                         continue;
 751                 } else if (ifp->flags & IFA_F_PERMANENT) {
 752                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 753                                               ifp->prefix_len)) {
 754                                 if (ifa->flags & IFA_F_PERMANENT) {
 755                                         onlink = 1;
 756                                         if (deleted)
 757                                                 break;
 758                                 } else {
 759                                         unsigned long lifetime;
 760
 761                                         if (!onlink)
 762                                                 onlink = -1;
 763
 764                                         spin_lock(&ifa->lock);
 765
 766                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 767                                         /*
 768                                          * Note: Because this address is
 769                                          * not permanent, lifetime <
 770                                          * LONG_MAX / HZ here.
 771                                          */
 772                                         if (time_before(expires,
 773                                                         ifa->tstamp + lifetime * HZ))
 774                                                 expires = ifa->tstamp + lifetime * HZ;
 775                                         spin_unlock(&ifa->lock);
 776                                 }
 777                         }
 778                 }
 779                 ifap = &ifa->if_next;
 780         }
 781         write_unlock_bh(&idev->lock);
 782
 783         addrconf_del_timer(ifp);
 784
 785         ipv6_ifa_notify(RTM_DELADDR, ifp);
 786
 787         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 788
 789         /*
 790          * Purge or update corresponding prefix
 791          *
 792          * 1) we don't purge prefix here if address was not permanent.
 793          *    prefix is managed by its own lifetime.
 794          * 2) if there're no addresses, delete prefix.
 795          * 3) if there're still other permanent address(es),
 796          *    corresponding prefix is still permanent.
 797          * 4) otherwise, update prefix lifetime to the
 798          *    longest valid lifetime among the corresponding
 799          *    addresses on the device.
 800          *    Note: subsequent RA will update lifetime.
 801          *
 802          * --yoshfuji
 803          */
 804         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 805                 struct in6_addr prefix;
 806                 struct rt6_info *rt;
 807                 struct net *net = dev_net(ifp->idev->dev);
 808                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 809                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 810
 811                 if (rt && addrconf_is_prefix_route(rt)) {
 812                         if (onlink == 0) {
 813                                 ip6_del_rt(rt);
 814                                 rt = NULL;
 815                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 816                                 rt->rt6i_expires = expires;
 817                                 rt->rt6i_flags |= RTF_EXPIRES;
 818                         }
 819                 }
 820                 dst_release(&rt->u.dst);
 821         }
 822
 823         in6_ifa_put(ifp);
 824 }
 825
 826 #ifdef CONFIG_IPV6_PRIVACY
 827 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 828 {
 829         struct inet6_dev *idev = ifp->idev;
 830         struct in6_addr addr, *tmpaddr;
 831         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 832         unsigned long regen_advance;
 833         int tmp_plen;
 834         int ret = 0;
 835         int max_addresses;
 836         u32 addr_flags;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         tmp_valid_lft = min_t(__u32,
 882                               ifp->valid_lft,
 883                               idev->cnf.temp_valid_lft);
 884         tmp_prefered_lft = min_t(__u32,
 885                                  ifp->prefered_lft,
 886                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 887         tmp_plen = ifp->prefix_len;
 888         max_addresses = idev->cnf.max_addresses;
 889         tmp_cstamp = ifp->cstamp;
 890         tmp_tstamp = ifp->tstamp;
 891         spin_unlock_bh(&ifp->lock);
 892
 893         regen_advance = idev->cnf.regen_max_retry *
 894                         idev->cnf.dad_transmits *
 895                         idev->nd_parms->retrans_time / HZ;
 896         write_unlock(&idev->lock);
 897
 898         /* A temporary address is created only if this calculated Preferred
 899          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 900          * an implementation must not create a temporary address with a zero
 901          * Preferred Lifetime.
 902          */
 903         if (tmp_prefered_lft <= regen_advance) {
 904                 in6_ifa_put(ifp);
 905                 in6_dev_put(idev);
 906                 ret = -1;
 907                 goto out;
 908         }
 909
 910         addr_flags = IFA_F_TEMPORARY;
 911         /* set in addrconf_prefix_rcv() */
 912         if (ifp->flags & IFA_F_OPTIMISTIC)
 913                 addr_flags |= IFA_F_OPTIMISTIC;
 914
 915         ift = !max_addresses ||
 916               ipv6_count_addresses(idev) < max_addresses ?
 917                 ipv6_add_addr(idev, &addr, tmp_plen,
 918                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 919                               addr_flags) : NULL;
 920         if (!ift || IS_ERR(ift)) {
 921                 in6_ifa_put(ifp);
 922                 in6_dev_put(idev);
 923                 printk(KERN_INFO
 924                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 925                 tmpaddr = &addr;
 926                 write_lock(&idev->lock);
 927                 goto retry;
 928         }
 929
 930         spin_lock_bh(&ift->lock);
 931         ift->ifpub = ifp;
 932         ift->valid_lft = tmp_valid_lft;
 933         ift->prefered_lft = tmp_prefered_lft;
 934         ift->cstamp = tmp_cstamp;
 935         ift->tstamp = tmp_tstamp;
 936         spin_unlock_bh(&ift->lock);
 937
 938         addrconf_dad_start(ift, 0);
 939         in6_ifa_put(ift);
 940         in6_dev_put(idev);
 941 out:
 942         return ret;
 943 }
 944 #endif
 945
 946 /*
 947  *      Choose an appropriate source address (RFC3484)
 948  */
 949 enum {
 950         IPV6_SADDR_RULE_INIT = 0,
 951         IPV6_SADDR_RULE_LOCAL,
 952         IPV6_SADDR_RULE_SCOPE,
 953         IPV6_SADDR_RULE_PREFERRED,
 954 #ifdef CONFIG_IPV6_MIP6
 955         IPV6_SADDR_RULE_HOA,
 956 #endif
 957         IPV6_SADDR_RULE_OIF,
 958         IPV6_SADDR_RULE_LABEL,
 959 #ifdef CONFIG_IPV6_PRIVACY
 960         IPV6_SADDR_RULE_PRIVACY,
 961 #endif
 962         IPV6_SADDR_RULE_ORCHID,
 963         IPV6_SADDR_RULE_PREFIX,
 964         IPV6_SADDR_RULE_MAX
 965 };
 966
 967 struct ipv6_saddr_score {
 968         int                     rule;
 969         int                     addr_type;
 970         struct inet6_ifaddr     *ifa;
 971         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 972         int                     scopedist;
 973         int                     matchlen;
 974 };
 975
 976 struct ipv6_saddr_dst {
 977         const struct in6_addr *addr;
 978         int ifindex;
 979         int scope;
 980         int label;
 981         unsigned int prefs;
 982 };
 983
 984 static inline int ipv6_saddr_preferred(int type)
 985 {
 986         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 987                 return 1;
 988         return 0;
 989 }
 990
 991 static int ipv6_get_saddr_eval(struct net *net,
 992                                struct ipv6_saddr_score *score,
 993                                struct ipv6_saddr_dst *dst,
 994                                int i)
 995 {
 996         int ret;
 997
 998         if (i <= score->rule) {
 999                 switch (i) {
1000                 case IPV6_SADDR_RULE_SCOPE:
1001                         ret = score->scopedist;
1002                         break;
1003                 case IPV6_SADDR_RULE_PREFIX:
1004                         ret = score->matchlen;
1005                         break;
1006                 default:
1007                         ret = !!test_bit(i, score->scorebits);
1008                 }
1009                 goto out;
1010         }
1011
1012         switch (i) {
1013         case IPV6_SADDR_RULE_INIT:
1014                 /* Rule 0: remember if hiscore is not ready yet */
1015                 ret = !!score->ifa;
1016                 break;
1017         case IPV6_SADDR_RULE_LOCAL:
1018                 /* Rule 1: Prefer same address */
1019                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1020                 break;
1021         case IPV6_SADDR_RULE_SCOPE:
1022                 /* Rule 2: Prefer appropriate scope
1023                  *
1024                  *      ret
1025                  *       ^
1026                  *    -1 |  d 15
1027                  *    ---+--+-+---> scope
1028                  *       |
1029                  *       |             d is scope of the destination.
1030                  *  B-d  |  \
1031                  *       |   \      <- smaller scope is better if
1032                  *  B-15 |    \        if scope is enough for destinaion.
1033                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1034                  * d-C-1 | /
1035                  *       |/         <- greater is better
1036                  *   -C  /             if scope is not enough for destination.
1037                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1038                  *
1039                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1040                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1041                  * Assume B = 0 and we get C > 29.
1042                  */
1043                 ret = __ipv6_addr_src_scope(score->addr_type);
1044                 if (ret >= dst->scope)
1045                         ret = -ret;
1046                 else
1047                         ret -= 128;     /* 30 is enough */
1048                 score->scopedist = ret;
1049                 break;
1050         case IPV6_SADDR_RULE_PREFERRED:
1051                 /* Rule 3: Avoid deprecated and optimistic addresses */
1052                 ret = ipv6_saddr_preferred(score->addr_type) ||
1053                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1054                 break;
1055 #ifdef CONFIG_IPV6_MIP6
1056         case IPV6_SADDR_RULE_HOA:
1057             {
1058                 /* Rule 4: Prefer home address */
1059                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1060                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1061                 break;
1062             }
1063 #endif
1064         case IPV6_SADDR_RULE_OIF:
1065                 /* Rule 5: Prefer outgoing interface */
1066                 ret = (!dst->ifindex ||
1067                        dst->ifindex == score->ifa->idev->dev->ifindex);
1068                 break;
1069         case IPV6_SADDR_RULE_LABEL:
1070                 /* Rule 6: Prefer matching label */
1071                 ret = ipv6_addr_label(net,
1072                                       &score->ifa->addr, score->addr_type,
1073                                       score->ifa->idev->dev->ifindex) == dst->label;
1074                 break;
1075 #ifdef CONFIG_IPV6_PRIVACY
1076         case IPV6_SADDR_RULE_PRIVACY:
1077             {
1078                 /* Rule 7: Prefer public address
1079                  * Note: prefer temprary address if use_tempaddr >= 2
1080                  */
1081                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1082                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1083                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1084                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1085                 break;
1086             }
1087 #endif
1088         case IPV6_SADDR_RULE_ORCHID:
1089                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1090                  *          non-ORCHID vs non-ORCHID
1091                  */
1092                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1093                         ipv6_addr_orchid(dst->addr));
1094                 break;
1095         case IPV6_SADDR_RULE_PREFIX:
1096                 /* Rule 8: Use longest matching prefix */
1097                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1098                                                        dst->addr);
1099                 break;
1100         default:
1101                 ret = 0;
1102         }
1103
1104         if (ret)
1105                 __set_bit(i, score->scorebits);
1106         score->rule = i;
1107 out:
1108         return ret;
1109 }
1110
1111 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1112                        const struct in6_addr *daddr, unsigned int prefs,
1113                        struct in6_addr *saddr)
1114 {
1115         struct ipv6_saddr_score scores[2],
1116                                 *score = &scores[0], *hiscore = &scores[1];
1117         struct ipv6_saddr_dst dst;
1118         struct net_device *dev;
1119         int dst_type;
1120
1121         dst_type = __ipv6_addr_type(daddr);
1122         dst.addr = daddr;
1123         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1124         dst.scope = __ipv6_addr_src_scope(dst_type);
1125         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1126         dst.prefs = prefs;
1127
1128         hiscore->rule = -1;
1129         hiscore->ifa = NULL;
1130
1131         rcu_read_lock();
1132
1133         for_each_netdev_rcu(net, dev) {
1134                 struct inet6_dev *idev;
1135
1136                 /* Candidate Source Address (section 4)
1137                  *  - multicast and link-local destination address,
1138                  *    the set of candidate source address MUST only
1139                  *    include addresses assigned to interfaces
1140                  *    belonging to the same link as the outgoing
1141                  *    interface.
1142                  * (- For site-local destination addresses, the
1143                  *    set of candidate source addresses MUST only
1144                  *    include addresses assigned to interfaces
1145                  *    belonging to the same site as the outgoing
1146                  *    interface.)
1147                  */
1148                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1149                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1150                     dst.ifindex && dev->ifindex != dst.ifindex)
1151                         continue;
1152
1153                 idev = __in6_dev_get(dev);
1154                 if (!idev)
1155                         continue;
1156
1157                 read_lock_bh(&idev->lock);
1158                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1159                         int i;
1160
1161                         /*
1162                          * - Tentative Address (RFC2462 section 5.4)
1163                          *  - A tentative address is not considered
1164                          *    "assigned to an interface" in the traditional
1165                          *    sense, unless it is also flagged as optimistic.
1166                          * - Candidate Source Address (section 4)
1167                          *  - In any case, anycast addresses, multicast
1168                          *    addresses, and the unspecified address MUST
1169                          *    NOT be included in a candidate set.
1170                          */
1171                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1172                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1173                                 continue;
1174
1175                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1176
1177                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1178                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1179                                 LIMIT_NETDEBUG(KERN_DEBUG
1180                                                "ADDRCONF: unspecified / multicast address "
1181                                                "assigned as unicast address on %s",
1182                                                dev->name);
1183                                 continue;
1184                         }
1185
1186                         score->rule = -1;
1187                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1188
1189                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1190                                 int minihiscore, miniscore;
1191
1192                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1193                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1194
1195                                 if (minihiscore > miniscore) {
1196                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1197                                             score->scopedist > 0) {
1198                                                 /*
1199                                                  * special case:
1200                                                  * each remaining entry
1201                                                  * has too small (not enough)
1202                                                  * scope, because ifa entries
1203                                                  * are sorted by their scope
1204                                                  * values.
1205                                                  */
1206                                                 goto try_nextdev;
1207                                         }
1208                                         break;
1209                                 } else if (minihiscore < miniscore) {
1210                                         if (hiscore->ifa)
1211                                                 in6_ifa_put(hiscore->ifa);
1212
1213                                         in6_ifa_hold(score->ifa);
1214
1215                                         swap(hiscore, score);
1216
1217                                         /* restore our iterator */
1218                                         score->ifa = hiscore->ifa;
1219
1220                                         break;
1221                                 }
1222                         }
1223                 }
1224 try_nextdev:
1225                 read_unlock_bh(&idev->lock);
1226         }
1227         rcu_read_unlock();
1228
1229         if (!hiscore->ifa)
1230                 return -EADDRNOTAVAIL;
1231
1232         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1233         in6_ifa_put(hiscore->ifa);
1234         return 0;
1235 }
1236
1237 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1238
1239 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1240                     unsigned char banned_flags)
1241 {
1242         struct inet6_dev *idev;
1243         int err = -EADDRNOTAVAIL;
1244
1245         rcu_read_lock();
1246         if ((idev = __in6_dev_get(dev)) != NULL) {
1247                 struct inet6_ifaddr *ifp;
1248
1249                 read_lock_bh(&idev->lock);
1250                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1251                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1252                                 ipv6_addr_copy(addr, &ifp->addr);
1253                                 err = 0;
1254                                 break;
1255                         }
1256                 }
1257                 read_unlock_bh(&idev->lock);
1258         }
1259         rcu_read_unlock();
1260         return err;
1261 }
1262
1263 static int ipv6_count_addresses(struct inet6_dev *idev)
1264 {
1265         int cnt = 0;
1266         struct inet6_ifaddr *ifp;
1267
1268         read_lock_bh(&idev->lock);
1269         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1270                 cnt++;
1271         read_unlock_bh(&idev->lock);
1272         return cnt;
1273 }
1274
1275 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1276                   struct net_device *dev, int strict)
1277 {
1278         struct inet6_ifaddr *ifp = NULL;
1279         struct hlist_node *node;
1280         u8 hash = ipv6_addr_hash(addr);
1281
1282         rcu_read_lock_bh();
1283         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1284                 if (!net_eq(dev_net(ifp->idev->dev), net))
1285                         continue;
1286                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1287                     !(ifp->flags&IFA_F_TENTATIVE)) {
1288                         if (dev == NULL || ifp->idev->dev == dev ||
1289                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1290                                 break;
1291                 }
1292         }
1293         rcu_read_unlock_bh();
1294
1295         return ifp != NULL;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static
1300 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301                        struct net_device *dev)
1302 {
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305         u8 hash = ipv6_addr_hash(addr);
1306
1307         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1308                 if (!net_eq(dev_net(ifp->idev->dev), net))
1309                         continue;
1310                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1311                         if (dev == NULL || ifp->idev->dev == dev)
1312                                 break;
1313                 }
1314         }
1315         return ifp != NULL;
1316 }
1317
1318 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1319 {
1320         struct inet6_dev *idev;
1321         struct inet6_ifaddr *ifa;
1322         int     onlink;
1323
1324         onlink = 0;
1325         rcu_read_lock();
1326         idev = __in6_dev_get(dev);
1327         if (idev) {
1328                 read_lock_bh(&idev->lock);
1329                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1330                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1331                                                    ifa->prefix_len);
1332                         if (onlink)
1333                                 break;
1334                 }
1335                 read_unlock_bh(&idev->lock);
1336         }
1337         rcu_read_unlock();
1338         return onlink;
1339 }
1340
1341 EXPORT_SYMBOL(ipv6_chk_prefix);
1342
1343 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1344                                      struct net_device *dev, int strict)
1345 {
1346         struct inet6_ifaddr *ifp = NULL;
1347         struct hlist_node *node;
1348         u8 hash = ipv6_addr_hash(addr);
1349
1350         rcu_read_lock_bh();
1351         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1352                 if (!net_eq(dev_net(ifp->idev->dev), net))
1353                         continue;
1354                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1355                         if (dev == NULL || ifp->idev->dev == dev ||
1356                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return ifp;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1401 {
1402         struct inet6_dev *idev = ifp->idev;
1403
1404         if (net_ratelimit())
1405                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1406                         ifp->idev->dev->name, &ifp->addr);
1407
1408         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1409                 struct in6_addr addr;
1410
1411                 addr.s6_addr32[0] = htonl(0xfe800000);
1412                 addr.s6_addr32[1] = 0;
1413
1414                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1415                     ipv6_addr_equal(&ifp->addr, &addr)) {
1416                         /* DAD failed for link-local based on MAC address */
1417                         idev->cnf.disable_ipv6 = 1;
1418
1419                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1420                                 ifp->idev->dev->name);
1421                 }
1422         }
1423
1424         addrconf_dad_stop(ifp, 1);
1425 }
1426
1427 /* Join to solicited addr multicast group. */
1428
1429 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1430 {
1431         struct in6_addr maddr;
1432
1433         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1434                 return;
1435
1436         addrconf_addr_solict_mult(addr, &maddr);
1437         ipv6_dev_mc_inc(dev, &maddr);
1438 }
1439
1440 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1441 {
1442         struct in6_addr maddr;
1443
1444         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1445                 return;
1446
1447         addrconf_addr_solict_mult(addr, &maddr);
1448         __ipv6_dev_mc_dec(idev, &maddr);
1449 }
1450
1451 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1452 {
1453         struct in6_addr addr;
1454         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1455         if (ipv6_addr_any(&addr))
1456                 return;
1457         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1458 }
1459
1460 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1461 {
1462         struct in6_addr addr;
1463         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1464         if (ipv6_addr_any(&addr))
1465                 return;
1466         __ipv6_dev_ac_dec(ifp->idev, &addr);
1467 }
1468
1469 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1470 {
1471         if (dev->addr_len != ETH_ALEN)
1472                 return -1;
1473         memcpy(eui, dev->dev_addr, 3);
1474         memcpy(eui + 5, dev->dev_addr + 3, 3);
1475
1476         /*
1477          * The zSeries OSA network cards can be shared among various
1478          * OS instances, but the OSA cards have only one MAC address.
1479          * This leads to duplicate address conflicts in conjunction
1480          * with IPv6 if more than one instance uses the same card.
1481          *
1482          * The driver for these cards can deliver a unique 16-bit
1483          * identifier for each instance sharing the same card.  It is
1484          * placed instead of 0xFFFE in the interface identifier.  The
1485          * "u" bit of the interface identifier is not inverted in this
1486          * case.  Hence the resulting interface identifier has local
1487          * scope according to RFC2373.
1488          */
1489         if (dev->dev_id) {
1490                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1491                 eui[4] = dev->dev_id & 0xFF;
1492         } else {
1493                 eui[3] = 0xFF;
1494                 eui[4] = 0xFE;
1495                 eui[0] ^= 2;
1496         }
1497         return 0;
1498 }
1499
1500 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1501 {
1502         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1503         if (dev->addr_len != ARCNET_ALEN)
1504                 return -1;
1505         memset(eui, 0, 7);
1506         eui[7] = *(u8*)dev->dev_addr;
1507         return 0;
1508 }
1509
1510 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1511 {
1512         if (dev->addr_len != INFINIBAND_ALEN)
1513                 return -1;
1514         memcpy(eui, dev->dev_addr + 12, 8);
1515         eui[0] |= 2;
1516         return 0;
1517 }
1518
1519 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1520 {
1521         if (addr == 0)
1522                 return -1;
1523         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1524                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1525                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1526                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1527                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1528                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1529         eui[1] = 0;
1530         eui[2] = 0x5E;
1531         eui[3] = 0xFE;
1532         memcpy(eui + 4, &addr, 4);
1533         return 0;
1534 }
1535 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1536
1537 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1538 {
1539         if (dev->priv_flags & IFF_ISATAP)
1540                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1541         return -1;
1542 }
1543
1544 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1545 {
1546         switch (dev->type) {
1547         case ARPHRD_ETHER:
1548         case ARPHRD_FDDI:
1549         case ARPHRD_IEEE802_TR:
1550                 return addrconf_ifid_eui48(eui, dev);
1551         case ARPHRD_ARCNET:
1552                 return addrconf_ifid_arcnet(eui, dev);
1553         case ARPHRD_INFINIBAND:
1554                 return addrconf_ifid_infiniband(eui, dev);
1555         case ARPHRD_SIT:
1556                 return addrconf_ifid_sit(eui, dev);
1557         }
1558         return -1;
1559 }
1560
1561 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1562 {
1563         int err = -1;
1564         struct inet6_ifaddr *ifp;
1565
1566         read_lock_bh(&idev->lock);
1567         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1568                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1569                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1570                         err = 0;
1571                         break;
1572                 }
1573         }
1574         read_unlock_bh(&idev->lock);
1575         return err;
1576 }
1577
1578 #ifdef CONFIG_IPV6_PRIVACY
1579 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1580 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1581 {
1582 regen:
1583         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1584         idev->rndid[0] &= ~0x02;
1585
1586         /*
1587          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1588          * check if generated address is not inappropriate
1589          *
1590          *  - Reserved subnet anycast (RFC 2526)
1591          *      11111101 11....11 1xxxxxxx
1592          *  - ISATAP (RFC4214) 6.1
1593          *      00-00-5E-FE-xx-xx-xx-xx
1594          *  - value 0
1595          *  - XXX: already assigned to an address on the device
1596          */
1597         if (idev->rndid[0] == 0xfd &&
1598             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1599             (idev->rndid[7]&0x80))
1600                 goto regen;
1601         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1602                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1603                         goto regen;
1604                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1605                         goto regen;
1606         }
1607
1608         return 0;
1609 }
1610
1611 static void ipv6_regen_rndid(unsigned long data)
1612 {
1613         struct inet6_dev *idev = (struct inet6_dev *) data;
1614         unsigned long expires;
1615
1616         rcu_read_lock_bh();
1617         write_lock_bh(&idev->lock);
1618
1619         if (idev->dead)
1620                 goto out;
1621
1622         if (__ipv6_regen_rndid(idev) < 0)
1623                 goto out;
1624
1625         expires = jiffies +
1626                 idev->cnf.temp_prefered_lft * HZ -
1627                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1628         if (time_before(expires, jiffies)) {
1629                 printk(KERN_WARNING
1630                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1631                         idev->dev->name);
1632                 goto out;
1633         }
1634
1635         if (!mod_timer(&idev->regen_timer, expires))
1636                 in6_dev_hold(idev);
1637
1638 out:
1639         write_unlock_bh(&idev->lock);
1640         rcu_read_unlock_bh();
1641         in6_dev_put(idev);
1642 }
1643
1644 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1645         int ret = 0;
1646
1647         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1648                 ret = __ipv6_regen_rndid(idev);
1649         return ret;
1650 }
1651 #endif
1652
1653 /*
1654  *      Add prefix route.
1655  */
1656
1657 static void
1658 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1659                       unsigned long expires, u32 flags)
1660 {
1661         struct fib6_config cfg = {
1662                 .fc_table = RT6_TABLE_PREFIX,
1663                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1664                 .fc_ifindex = dev->ifindex,
1665                 .fc_expires = expires,
1666                 .fc_dst_len = plen,
1667                 .fc_flags = RTF_UP | flags,
1668                 .fc_nlinfo.nl_net = dev_net(dev),
1669                 .fc_protocol = RTPROT_KERNEL,
1670         };
1671
1672         ipv6_addr_copy(&cfg.fc_dst, pfx);
1673
1674         /* Prevent useless cloning on PtP SIT.
1675            This thing is done here expecting that the whole
1676            class of non-broadcast devices need not cloning.
1677          */
1678 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1679         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1680                 cfg.fc_flags |= RTF_NONEXTHOP;
1681 #endif
1682
1683         ip6_route_add(&cfg);
1684 }
1685
1686 /* Create "default" multicast route to the interface */
1687
1688 static void addrconf_add_mroute(struct net_device *dev)
1689 {
1690         struct fib6_config cfg = {
1691                 .fc_table = RT6_TABLE_LOCAL,
1692                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1693                 .fc_ifindex = dev->ifindex,
1694                 .fc_dst_len = 8,
1695                 .fc_flags = RTF_UP,
1696                 .fc_nlinfo.nl_net = dev_net(dev),
1697         };
1698
1699         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1700
1701         ip6_route_add(&cfg);
1702 }
1703
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705 static void sit_route_add(struct net_device *dev)
1706 {
1707         struct fib6_config cfg = {
1708                 .fc_table = RT6_TABLE_MAIN,
1709                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710                 .fc_ifindex = dev->ifindex,
1711                 .fc_dst_len = 96,
1712                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1713                 .fc_nlinfo.nl_net = dev_net(dev),
1714         };
1715
1716         /* prefix length - 96 bits "::d.d.d.d" */
1717         ip6_route_add(&cfg);
1718 }
1719 #endif
1720
1721 static void addrconf_add_lroute(struct net_device *dev)
1722 {
1723         struct in6_addr addr;
1724
1725         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1726         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1727 }
1728
1729 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1730 {
1731         struct inet6_dev *idev;
1732
1733         ASSERT_RTNL();
1734
1735         if ((idev = ipv6_find_idev(dev)) == NULL)
1736                 return NULL;
1737
1738         /* Add default multicast route */
1739         addrconf_add_mroute(dev);
1740
1741         /* Add link local route */
1742         addrconf_add_lroute(dev);
1743         return idev;
1744 }
1745
1746 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1747 {
1748         struct prefix_info *pinfo;
1749         __u32 valid_lft;
1750         __u32 prefered_lft;
1751         int addr_type;
1752         struct inet6_dev *in6_dev;
1753         struct net *net = dev_net(dev);
1754
1755         pinfo = (struct prefix_info *) opt;
1756
1757         if (len < sizeof(struct prefix_info)) {
1758                 ADBG(("addrconf: prefix option too short\n"));
1759                 return;
1760         }
1761
1762         /*
1763          *      Validation checks ([ADDRCONF], page 19)
1764          */
1765
1766         addr_type = ipv6_addr_type(&pinfo->prefix);
1767
1768         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1769                 return;
1770
1771         valid_lft = ntohl(pinfo->valid);
1772         prefered_lft = ntohl(pinfo->prefered);
1773
1774         if (prefered_lft > valid_lft) {
1775                 if (net_ratelimit())
1776                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1777                 return;
1778         }
1779
1780         in6_dev = in6_dev_get(dev);
1781
1782         if (in6_dev == NULL) {
1783                 if (net_ratelimit())
1784                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1785                 return;
1786         }
1787
1788         /*
1789          *      Two things going on here:
1790          *      1) Add routes for on-link prefixes
1791          *      2) Configure prefixes with the auto flag set
1792          */
1793
1794         if (pinfo->onlink) {
1795                 struct rt6_info *rt;
1796                 unsigned long rt_expires;
1797
1798                 /* Avoid arithmetic overflow. Really, we could
1799                  * save rt_expires in seconds, likely valid_lft,
1800                  * but it would require division in fib gc, that it
1801                  * not good.
1802                  */
1803                 if (HZ > USER_HZ)
1804                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1805                 else
1806                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1807
1808                 if (addrconf_finite_timeout(rt_expires))
1809                         rt_expires *= HZ;
1810
1811                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1812                                 dev->ifindex, 1);
1813
1814                 if (rt && addrconf_is_prefix_route(rt)) {
1815                         /* Autoconf prefix route */
1816                         if (valid_lft == 0) {
1817                                 ip6_del_rt(rt);
1818                                 rt = NULL;
1819                         } else if (addrconf_finite_timeout(rt_expires)) {
1820                                 /* not infinity */
1821                                 rt->rt6i_expires = jiffies + rt_expires;
1822                                 rt->rt6i_flags |= RTF_EXPIRES;
1823                         } else {
1824                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1825                                 rt->rt6i_expires = 0;
1826                         }
1827                 } else if (valid_lft) {
1828                         clock_t expires = 0;
1829                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1830                         if (addrconf_finite_timeout(rt_expires)) {
1831                                 /* not infinity */
1832                                 flags |= RTF_EXPIRES;
1833                                 expires = jiffies_to_clock_t(rt_expires);
1834                         }
1835                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1836                                               dev, expires, flags);
1837                 }
1838                 if (rt)
1839                         dst_release(&rt->u.dst);
1840         }
1841
1842         /* Try to figure out our local address for this prefix */
1843
1844         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1845                 struct inet6_ifaddr * ifp;
1846                 struct in6_addr addr;
1847                 int create = 0, update_lft = 0;
1848
1849                 if (pinfo->prefix_len == 64) {
1850                         memcpy(&addr, &pinfo->prefix, 8);
1851                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1852                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1853                                 in6_dev_put(in6_dev);
1854                                 return;
1855                         }
1856                         goto ok;
1857                 }
1858                 if (net_ratelimit())
1859                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1860                                pinfo->prefix_len);
1861                 in6_dev_put(in6_dev);
1862                 return;
1863
1864 ok:
1865
1866                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1867
1868                 if (ifp == NULL && valid_lft) {
1869                         int max_addresses = in6_dev->cnf.max_addresses;
1870                         u32 addr_flags = 0;
1871
1872 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1873                         if (in6_dev->cnf.optimistic_dad &&
1874                             !net->ipv6.devconf_all->forwarding)
1875                                 addr_flags = IFA_F_OPTIMISTIC;
1876 #endif
1877
1878                         /* Do not allow to create too much of autoconfigured
1879                          * addresses; this would be too easy way to crash kernel.
1880                          */
1881                         if (!max_addresses ||
1882                             ipv6_count_addresses(in6_dev) < max_addresses)
1883                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1884                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1885                                                     addr_flags);
1886
1887                         if (!ifp || IS_ERR(ifp)) {
1888                                 in6_dev_put(in6_dev);
1889                                 return;
1890                         }
1891
1892                         update_lft = create = 1;
1893                         ifp->cstamp = jiffies;
1894                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1895                 }
1896
1897                 if (ifp) {
1898                         int flags;
1899                         unsigned long now;
1900 #ifdef CONFIG_IPV6_PRIVACY
1901                         struct inet6_ifaddr *ift;
1902 #endif
1903                         u32 stored_lft;
1904
1905                         /* update lifetime (RFC2462 5.5.3 e) */
1906                         spin_lock(&ifp->lock);
1907                         now = jiffies;
1908                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1909                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1910                         else
1911                                 stored_lft = 0;
1912                         if (!update_lft && stored_lft) {
1913                                 if (valid_lft > MIN_VALID_LIFETIME ||
1914                                     valid_lft > stored_lft)
1915                                         update_lft = 1;
1916                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1917                                         /* valid_lft <= stored_lft is always true */
1918                                         /*
1919                                          * RFC 4862 Section 5.5.3e:
1920                                          * "Note that the preferred lifetime of
1921                                          *  the corresponding address is always
1922                                          *  reset to the Preferred Lifetime in
1923                                          *  the received Prefix Information
1924                                          *  option, regardless of whether the
1925                                          *  valid lifetime is also reset or
1926                                          *  ignored."
1927                                          *
1928                                          *  So if the preferred lifetime in
1929                                          *  this advertisement is different
1930                                          *  than what we have stored, but the
1931                                          *  valid lifetime is invalid, just
1932                                          *  reset prefered_lft.
1933                                          *
1934                                          *  We must set the valid lifetime
1935                                          *  to the stored lifetime since we'll
1936                                          *  be updating the timestamp below,
1937                                          *  else we'll set it back to the
1938                                          *  minumum.
1939                                          */
1940                                         if (prefered_lft != ifp->prefered_lft) {
1941                                                 valid_lft = stored_lft;
1942                                                 update_lft = 1;
1943                                         }
1944                                 } else {
1945                                         valid_lft = MIN_VALID_LIFETIME;
1946                                         if (valid_lft < prefered_lft)
1947                                                 prefered_lft = valid_lft;
1948                                         update_lft = 1;
1949                                 }
1950                         }
1951
1952                         if (update_lft) {
1953                                 ifp->valid_lft = valid_lft;
1954                                 ifp->prefered_lft = prefered_lft;
1955                                 ifp->tstamp = now;
1956                                 flags = ifp->flags;
1957                                 ifp->flags &= ~IFA_F_DEPRECATED;
1958                                 spin_unlock(&ifp->lock);
1959
1960                                 if (!(flags&IFA_F_TENTATIVE))
1961                                         ipv6_ifa_notify(0, ifp);
1962                         } else
1963                                 spin_unlock(&ifp->lock);
1964
1965 #ifdef CONFIG_IPV6_PRIVACY
1966                         read_lock_bh(&in6_dev->lock);
1967                         /* update all temporary addresses in the list */
1968                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1969                                 /*
1970                                  * When adjusting the lifetimes of an existing
1971                                  * temporary address, only lower the lifetimes.
1972                                  * Implementations must not increase the
1973                                  * lifetimes of an existing temporary address
1974                                  * when processing a Prefix Information Option.
1975                                  */
1976                                 if (ifp != ift->ifpub)
1977                                         continue;
1978
1979                                 spin_lock(&ift->lock);
1980                                 flags = ift->flags;
1981                                 if (ift->valid_lft > valid_lft &&
1982                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1983                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1984                                 if (ift->prefered_lft > prefered_lft &&
1985                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1986                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1987                                 spin_unlock(&ift->lock);
1988                                 if (!(flags&IFA_F_TENTATIVE))
1989                                         ipv6_ifa_notify(0, ift);
1990                         }
1991
1992                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1993                                 /*
1994                                  * When a new public address is created as described in [ADDRCONF],
1995                                  * also create a new temporary address.
1996                                  */
1997                                 read_unlock_bh(&in6_dev->lock);
1998                                 ipv6_create_tempaddr(ifp, NULL);
1999                         } else {
2000                                 read_unlock_bh(&in6_dev->lock);
2001                         }
2002 #endif
2003                         in6_ifa_put(ifp);
2004                         addrconf_verify(0);
2005                 }
2006         }
2007         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2008         in6_dev_put(in6_dev);
2009 }
2010
2011 /*
2012  *      Set destination address.
2013  *      Special case for SIT interfaces where we create a new "virtual"
2014  *      device.
2015  */
2016 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2017 {
2018         struct in6_ifreq ireq;
2019         struct net_device *dev;
2020         int err = -EINVAL;
2021
2022         rtnl_lock();
2023
2024         err = -EFAULT;
2025         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2026                 goto err_exit;
2027
2028         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2029
2030         err = -ENODEV;
2031         if (dev == NULL)
2032                 goto err_exit;
2033
2034 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2035         if (dev->type == ARPHRD_SIT) {
2036                 const struct net_device_ops *ops = dev->netdev_ops;
2037                 struct ifreq ifr;
2038                 struct ip_tunnel_parm p;
2039
2040                 err = -EADDRNOTAVAIL;
2041                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2042                         goto err_exit;
2043
2044                 memset(&p, 0, sizeof(p));
2045                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2046                 p.iph.saddr = 0;
2047                 p.iph.version = 4;
2048                 p.iph.ihl = 5;
2049                 p.iph.protocol = IPPROTO_IPV6;
2050                 p.iph.ttl = 64;
2051                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2052
2053                 if (ops->ndo_do_ioctl) {
2054                         mm_segment_t oldfs = get_fs();
2055
2056                         set_fs(KERNEL_DS);
2057                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2058                         set_fs(oldfs);
2059                 } else
2060                         err = -EOPNOTSUPP;
2061
2062                 if (err == 0) {
2063                         err = -ENOBUFS;
2064                         dev = __dev_get_by_name(net, p.name);
2065                         if (!dev)
2066                                 goto err_exit;
2067                         err = dev_open(dev);
2068                 }
2069         }
2070 #endif
2071
2072 err_exit:
2073         rtnl_unlock();
2074         return err;
2075 }
2076
2077 /*
2078  *      Manual configuration of address on an interface
2079  */
2080 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2081                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2082                           __u32 valid_lft)
2083 {
2084         struct inet6_ifaddr *ifp;
2085         struct inet6_dev *idev;
2086         struct net_device *dev;
2087         int scope;
2088         u32 flags;
2089         clock_t expires;
2090         unsigned long timeout;
2091
2092         ASSERT_RTNL();
2093
2094         if (plen > 128)
2095                 return -EINVAL;
2096
2097         /* check the lifetime */
2098         if (!valid_lft || prefered_lft > valid_lft)
2099                 return -EINVAL;
2100
2101         dev = __dev_get_by_index(net, ifindex);
2102         if (!dev)
2103                 return -ENODEV;
2104
2105         if ((idev = addrconf_add_dev(dev)) == NULL)
2106                 return -ENOBUFS;
2107
2108         scope = ipv6_addr_scope(pfx);
2109
2110         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2111         if (addrconf_finite_timeout(timeout)) {
2112                 expires = jiffies_to_clock_t(timeout * HZ);
2113                 valid_lft = timeout;
2114                 flags = RTF_EXPIRES;
2115         } else {
2116                 expires = 0;
2117                 flags = 0;
2118                 ifa_flags |= IFA_F_PERMANENT;
2119         }
2120
2121         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2122         if (addrconf_finite_timeout(timeout)) {
2123                 if (timeout == 0)
2124                         ifa_flags |= IFA_F_DEPRECATED;
2125                 prefered_lft = timeout;
2126         }
2127
2128         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2129
2130         if (!IS_ERR(ifp)) {
2131                 spin_lock_bh(&ifp->lock);
2132                 ifp->valid_lft = valid_lft;
2133                 ifp->prefered_lft = prefered_lft;
2134                 ifp->tstamp = jiffies;
2135                 spin_unlock_bh(&ifp->lock);
2136
2137                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2138                                       expires, flags);
2139                 /*
2140                  * Note that section 3.1 of RFC 4429 indicates
2141                  * that the Optimistic flag should not be set for
2142                  * manually configured addresses
2143                  */
2144                 addrconf_dad_start(ifp, 0);
2145                 in6_ifa_put(ifp);
2146                 addrconf_verify(0);
2147                 return 0;
2148         }
2149
2150         return PTR_ERR(ifp);
2151 }
2152
2153 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2154                           unsigned int plen)
2155 {
2156         struct inet6_ifaddr *ifp;
2157         struct inet6_dev *idev;
2158         struct net_device *dev;
2159
2160         if (plen > 128)
2161                 return -EINVAL;
2162
2163         dev = __dev_get_by_index(net, ifindex);
2164         if (!dev)
2165                 return -ENODEV;
2166
2167         if ((idev = __in6_dev_get(dev)) == NULL)
2168                 return -ENXIO;
2169
2170         read_lock_bh(&idev->lock);
2171         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2172                 if (ifp->prefix_len == plen &&
2173                     ipv6_addr_equal(pfx, &ifp->addr)) {
2174                         in6_ifa_hold(ifp);
2175                         read_unlock_bh(&idev->lock);
2176
2177                         ipv6_del_addr(ifp);
2178
2179                         /* If the last address is deleted administratively,
2180                            disable IPv6 on this interface.
2181                          */
2182                         if (idev->addr_list == NULL)
2183                                 addrconf_ifdown(idev->dev, 1);
2184                         return 0;
2185                 }
2186         }
2187         read_unlock_bh(&idev->lock);
2188         return -EADDRNOTAVAIL;
2189 }
2190
2191
2192 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2193 {
2194         struct in6_ifreq ireq;
2195         int err;
2196
2197         if (!capable(CAP_NET_ADMIN))
2198                 return -EPERM;
2199
2200         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2201                 return -EFAULT;
2202
2203         rtnl_lock();
2204         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2205                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2206                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2207         rtnl_unlock();
2208         return err;
2209 }
2210
2211 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2212 {
2213         struct in6_ifreq ireq;
2214         int err;
2215
2216         if (!capable(CAP_NET_ADMIN))
2217                 return -EPERM;
2218
2219         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2220                 return -EFAULT;
2221
2222         rtnl_lock();
2223         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2224                              ireq.ifr6_prefixlen);
2225         rtnl_unlock();
2226         return err;
2227 }
2228
2229 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2230                      int plen, int scope)
2231 {
2232         struct inet6_ifaddr *ifp;
2233
2234         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2235         if (!IS_ERR(ifp)) {
2236                 spin_lock_bh(&ifp->lock);
2237                 ifp->flags &= ~IFA_F_TENTATIVE;
2238                 spin_unlock_bh(&ifp->lock);
2239                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2240                 in6_ifa_put(ifp);
2241         }
2242 }
2243
2244 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2245 static void sit_add_v4_addrs(struct inet6_dev *idev)
2246 {
2247         struct in6_addr addr;
2248         struct net_device *dev;
2249         struct net *net = dev_net(idev->dev);
2250         int scope;
2251
2252         ASSERT_RTNL();
2253
2254         memset(&addr, 0, sizeof(struct in6_addr));
2255         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2256
2257         if (idev->dev->flags&IFF_POINTOPOINT) {
2258                 addr.s6_addr32[0] = htonl(0xfe800000);
2259                 scope = IFA_LINK;
2260         } else {
2261                 scope = IPV6_ADDR_COMPATv4;
2262         }
2263
2264         if (addr.s6_addr32[3]) {
2265                 add_addr(idev, &addr, 128, scope);
2266                 return;
2267         }
2268
2269         for_each_netdev(net, dev) {
2270                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2271                 if (in_dev && (dev->flags & IFF_UP)) {
2272                         struct in_ifaddr * ifa;
2273
2274                         int flag = scope;
2275
2276                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2277                                 int plen;
2278
2279                                 addr.s6_addr32[3] = ifa->ifa_local;
2280
2281                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2282                                         continue;
2283                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2284                                         if (idev->dev->flags&IFF_POINTOPOINT)
2285                                                 continue;
2286                                         flag |= IFA_HOST;
2287                                 }
2288                                 if (idev->dev->flags&IFF_POINTOPOINT)
2289                                         plen = 64;
2290                                 else
2291                                         plen = 96;
2292
2293                                 add_addr(idev, &addr, plen, flag);
2294                         }
2295                 }
2296         }
2297 }
2298 #endif
2299
2300 static void init_loopback(struct net_device *dev)
2301 {
2302         struct inet6_dev  *idev;
2303
2304         /* ::1 */
2305
2306         ASSERT_RTNL();
2307
2308         if ((idev = ipv6_find_idev(dev)) == NULL) {
2309                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2310                 return;
2311         }
2312
2313         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2314 }
2315
2316 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2317 {
2318         struct inet6_ifaddr * ifp;
2319         u32 addr_flags = IFA_F_PERMANENT;
2320
2321 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2322         if (idev->cnf.optimistic_dad &&
2323             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2324                 addr_flags |= IFA_F_OPTIMISTIC;
2325 #endif
2326
2327
2328         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2329         if (!IS_ERR(ifp)) {
2330                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2331                 addrconf_dad_start(ifp, 0);
2332                 in6_ifa_put(ifp);
2333         }
2334 }
2335
2336 static void addrconf_dev_config(struct net_device *dev)
2337 {
2338         struct in6_addr addr;
2339         struct inet6_dev    * idev;
2340
2341         ASSERT_RTNL();
2342
2343         if ((dev->type != ARPHRD_ETHER) &&
2344             (dev->type != ARPHRD_FDDI) &&
2345             (dev->type != ARPHRD_IEEE802_TR) &&
2346             (dev->type != ARPHRD_ARCNET) &&
2347             (dev->type != ARPHRD_INFINIBAND)) {
2348                 /* Alas, we support only Ethernet autoconfiguration. */
2349                 return;
2350         }
2351
2352         idev = addrconf_add_dev(dev);
2353         if (idev == NULL)
2354                 return;
2355
2356         memset(&addr, 0, sizeof(struct in6_addr));
2357         addr.s6_addr32[0] = htonl(0xFE800000);
2358
2359         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2360                 addrconf_add_linklocal(idev, &addr);
2361 }
2362
2363 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2364 static void addrconf_sit_config(struct net_device *dev)
2365 {
2366         struct inet6_dev *idev;
2367
2368         ASSERT_RTNL();
2369
2370         /*
2371          * Configure the tunnel with one of our IPv4
2372          * addresses... we should configure all of
2373          * our v4 addrs in the tunnel
2374          */
2375
2376         if ((idev = ipv6_find_idev(dev)) == NULL) {
2377                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2378                 return;
2379         }
2380
2381         if (dev->priv_flags & IFF_ISATAP) {
2382                 struct in6_addr addr;
2383
2384                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2385                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2386                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2387                         addrconf_add_linklocal(idev, &addr);
2388                 return;
2389         }
2390
2391         sit_add_v4_addrs(idev);
2392
2393         if (dev->flags&IFF_POINTOPOINT) {
2394                 addrconf_add_mroute(dev);
2395                 addrconf_add_lroute(dev);
2396         } else
2397                 sit_route_add(dev);
2398 }
2399 #endif
2400
2401 static inline int
2402 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2403 {
2404         struct in6_addr lladdr;
2405
2406         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2407                 addrconf_add_linklocal(idev, &lladdr);
2408                 return 0;
2409         }
2410         return -1;
2411 }
2412
2413 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2414 {
2415         struct net_device *link_dev;
2416         struct net *net = dev_net(idev->dev);
2417
2418         /* first try to inherit the link-local address from the link device */
2419         if (idev->dev->iflink &&
2420             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2421                 if (!ipv6_inherit_linklocal(idev, link_dev))
2422                         return;
2423         }
2424         /* then try to inherit it from any device */
2425         for_each_netdev(net, link_dev) {
2426                 if (!ipv6_inherit_linklocal(idev, link_dev))
2427                         return;
2428         }
2429         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2430 }
2431
2432 /*
2433  * Autoconfigure tunnel with a link-local address so routing protocols,
2434  * DHCPv6, MLD etc. can be run over the virtual link
2435  */
2436
2437 static void addrconf_ip6_tnl_config(struct net_device *dev)
2438 {
2439         struct inet6_dev *idev;
2440
2441         ASSERT_RTNL();
2442
2443         if ((idev = addrconf_add_dev(dev)) == NULL) {
2444                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2445                 return;
2446         }
2447         ip6_tnl_add_linklocal(idev);
2448 }
2449
2450 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2451                            void * data)
2452 {
2453         struct net_device *dev = (struct net_device *) data;
2454         struct inet6_dev *idev = __in6_dev_get(dev);
2455         int run_pending = 0;
2456         int err;
2457
2458         switch(event) {
2459         case NETDEV_REGISTER:
2460                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2461                         idev = ipv6_add_dev(dev);
2462                         if (!idev)
2463                                 return notifier_from_errno(-ENOMEM);
2464                 }
2465                 break;
2466         case NETDEV_UP:
2467         case NETDEV_CHANGE:
2468                 if (dev->flags & IFF_SLAVE)
2469                         break;
2470
2471                 if (event == NETDEV_UP) {
2472                         if (!addrconf_qdisc_ok(dev)) {
2473                                 /* device is not ready yet. */
2474                                 printk(KERN_INFO
2475                                         "ADDRCONF(NETDEV_UP): %s: "
2476                                         "link is not ready\n",
2477                                         dev->name);
2478                                 break;
2479                         }
2480
2481                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2482                                 idev = ipv6_add_dev(dev);
2483
2484                         if (idev) {
2485                                 idev->if_flags |= IF_READY;
2486                                 run_pending = 1;
2487                         }
2488                 } else {
2489                         if (!addrconf_qdisc_ok(dev)) {
2490                                 /* device is still not ready. */
2491                                 break;
2492                         }
2493
2494                         if (idev) {
2495                                 if (idev->if_flags & IF_READY) {
2496                                         /* device is already configured. */
2497                                         break;
2498                                 }
2499                                 idev->if_flags |= IF_READY;
2500                         }
2501
2502                         printk(KERN_INFO
2503                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2504                                         "link becomes ready\n",
2505                                         dev->name);
2506
2507                         run_pending = 1;
2508                 }
2509
2510                 switch(dev->type) {
2511 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2512                 case ARPHRD_SIT:
2513                         addrconf_sit_config(dev);
2514                         break;
2515 #endif
2516                 case ARPHRD_TUNNEL6:
2517                         addrconf_ip6_tnl_config(dev);
2518                         break;
2519                 case ARPHRD_LOOPBACK:
2520                         init_loopback(dev);
2521                         break;
2522
2523                 default:
2524                         addrconf_dev_config(dev);
2525                         break;
2526                 }
2527                 if (idev) {
2528                         if (run_pending)
2529                                 addrconf_dad_run(idev);
2530
2531                         /* If the MTU changed during the interface down, when the
2532                            interface up, the changed MTU must be reflected in the
2533                            idev as well as routers.
2534                          */
2535                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2536                                 rt6_mtu_change(dev, dev->mtu);
2537                                 idev->cnf.mtu6 = dev->mtu;
2538                         }
2539                         idev->tstamp = jiffies;
2540                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2541                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2542                            stop IPv6 on this interface.
2543                          */
2544                         if (dev->mtu < IPV6_MIN_MTU)
2545                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2546                 }
2547                 break;
2548
2549         case NETDEV_CHANGEMTU:
2550                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2551                         rt6_mtu_change(dev, dev->mtu);
2552                         idev->cnf.mtu6 = dev->mtu;
2553                         break;
2554                 }
2555
2556                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2557                         idev = ipv6_add_dev(dev);
2558                         if (idev)
2559                                 break;
2560                 }
2561
2562                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2563
2564         case NETDEV_DOWN:
2565         case NETDEV_UNREGISTER:
2566                 /*
2567                  *      Remove all addresses from this interface.
2568                  */
2569                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2570                 break;
2571
2572         case NETDEV_CHANGENAME:
2573                 if (idev) {
2574                         snmp6_unregister_dev(idev);
2575                         addrconf_sysctl_unregister(idev);
2576                         addrconf_sysctl_register(idev);
2577                         err = snmp6_register_dev(idev);
2578                         if (err)
2579                                 return notifier_from_errno(err);
2580                 }
2581                 break;
2582         case NETDEV_PRE_TYPE_CHANGE:
2583         case NETDEV_POST_TYPE_CHANGE:
2584                 addrconf_type_change(dev, event);
2585                 break;
2586         }
2587
2588         return NOTIFY_OK;
2589 }
2590
2591 /*
2592  *      addrconf module should be notified of a device going up
2593  */
2594 static struct notifier_block ipv6_dev_notf = {
2595         .notifier_call = addrconf_notify,
2596         .priority = 0
2597 };
2598
2599 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2600 {
2601         struct inet6_dev *idev;
2602         ASSERT_RTNL();
2603
2604         idev = __in6_dev_get(dev);
2605
2606         if (event == NETDEV_POST_TYPE_CHANGE)
2607                 ipv6_mc_remap(idev);
2608         else if (event == NETDEV_PRE_TYPE_CHANGE)
2609                 ipv6_mc_unmap(idev);
2610 }
2611
2612 static int addrconf_ifdown(struct net_device *dev, int how)
2613 {
2614         struct inet6_dev *idev;
2615         struct inet6_ifaddr *ifa, *keep_list, **bifa;
2616         struct net *net = dev_net(dev);
2617
2618         ASSERT_RTNL();
2619
2620         rt6_ifdown(net, dev);
2621         neigh_ifdown(&nd_tbl, dev);
2622
2623         idev = __in6_dev_get(dev);
2624         if (idev == NULL)
2625                 return -ENODEV;
2626
2627         /* Step 1: remove reference to ipv6 device from parent device.
2628                    Do not dev_put!
2629          */
2630         if (how) {
2631                 idev->dead = 1;
2632
2633                 /* protected by rtnl_lock */
2634                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2635
2636                 /* Step 1.5: remove snmp6 entry */
2637                 snmp6_unregister_dev(idev);
2638
2639         }
2640
2641         write_lock_bh(&idev->lock);
2642
2643         /* Step 3: clear flags for stateless addrconf */
2644         if (!how)
2645                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2646
2647         /* Step 4: clear address list */
2648 #ifdef CONFIG_IPV6_PRIVACY
2649         if (how && del_timer(&idev->regen_timer))
2650                 in6_dev_put(idev);
2651
2652         /* clear tempaddr list */
2653         while (!list_empty(&idev->tempaddr_list)) {
2654                 ifa = list_first_entry(&idev->tempaddr_list,
2655                                        struct inet6_ifaddr, tmp_list);
2656                 list_del(&ifa->tmp_list);
2657                 ifa->dead = 1;
2658                 write_unlock_bh(&idev->lock);
2659                 spin_lock_bh(&ifa->lock);
2660
2661                 if (ifa->ifpub) {
2662                         in6_ifa_put(ifa->ifpub);
2663                         ifa->ifpub = NULL;
2664                 }
2665                 spin_unlock_bh(&ifa->lock);
2666                 in6_ifa_put(ifa);
2667                 write_lock_bh(&idev->lock);
2668         }
2669 #endif
2670         keep_list = NULL;
2671         bifa = &keep_list;
2672         while ((ifa = idev->addr_list) != NULL) {
2673                 idev->addr_list = ifa->if_next;
2674                 ifa->if_next = NULL;
2675
2676                 addrconf_del_timer(ifa);
2677
2678                 /* If just doing link down, and address is permanent
2679                    and not link-local, then retain it. */
2680                 if (how == 0 &&
2681                     (ifa->flags&IFA_F_PERMANENT) &&
2682                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2683
2684                         /* Move to holding list */
2685                         *bifa = ifa;
2686                         bifa = &ifa->if_next;
2687
2688                         /* If not doing DAD on this address, just keep it. */
2689                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2690                             idev->cnf.accept_dad <= 0 ||
2691                             (ifa->flags & IFA_F_NODAD))
2692                                 continue;
2693
2694                         /* If it was tentative already, no need to notify */
2695                         if (ifa->flags & IFA_F_TENTATIVE)
2696                                 continue;
2697
2698                         /* Flag it for later restoration when link comes up */
2699                         ifa->flags |= IFA_F_TENTATIVE;
2700                         in6_ifa_hold(ifa);
2701                 } else {
2702                         ifa->dead = 1;
2703                 }
2704                 write_unlock_bh(&idev->lock);
2705
2706                 /* clear hash table */
2707                 spin_lock_bh(&addrconf_hash_lock);
2708                 hlist_del_init_rcu(&ifa->addr_lst);
2709                 __in6_ifa_put(ifa);
2710                 spin_unlock_bh(&addrconf_hash_lock);
2711
2712                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2713                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2714                 in6_ifa_put(ifa);
2715
2716                 write_lock_bh(&idev->lock);
2717         }
2718
2719         idev->addr_list = keep_list;
2720
2721         write_unlock_bh(&idev->lock);
2722
2723         /* Step 5: Discard multicast list */
2724
2725         if (how)
2726                 ipv6_mc_destroy_dev(idev);
2727         else
2728                 ipv6_mc_down(idev);
2729
2730         idev->tstamp = jiffies;
2731
2732         /* Shot the device (if unregistered) */
2733
2734         if (how) {
2735                 addrconf_sysctl_unregister(idev);
2736                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2737                 neigh_ifdown(&nd_tbl, dev);
2738                 in6_dev_put(idev);
2739         }
2740         return 0;
2741 }
2742
2743 static void addrconf_rs_timer(unsigned long data)
2744 {
2745         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2746         struct inet6_dev *idev = ifp->idev;
2747
2748         read_lock(&idev->lock);
2749         if (idev->dead || !(idev->if_flags & IF_READY))
2750                 goto out;
2751
2752         if (idev->cnf.forwarding)
2753                 goto out;
2754
2755         /* Announcement received after solicitation was sent */
2756         if (idev->if_flags & IF_RA_RCVD)
2757                 goto out;
2758
2759         spin_lock(&ifp->lock);
2760         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2761                 /* The wait after the last probe can be shorter */
2762                 addrconf_mod_timer(ifp, AC_RS,
2763                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2764                                    idev->cnf.rtr_solicit_delay :
2765                                    idev->cnf.rtr_solicit_interval);
2766                 spin_unlock(&ifp->lock);
2767
2768                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2769         } else {
2770                 spin_unlock(&ifp->lock);
2771                 /*
2772                  * Note: we do not support deprecated "all on-link"
2773                  * assumption any longer.
2774                  */
2775                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2776                        idev->dev->name);
2777         }
2778
2779 out:
2780         read_unlock(&idev->lock);
2781         in6_ifa_put(ifp);
2782 }
2783
2784 /*
2785  *      Duplicate Address Detection
2786  */
2787 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2788 {
2789         unsigned long rand_num;
2790         struct inet6_dev *idev = ifp->idev;
2791
2792         if (ifp->flags & IFA_F_OPTIMISTIC)
2793                 rand_num = 0;
2794         else
2795                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2796
2797         ifp->probes = idev->cnf.dad_transmits;
2798         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2799 }
2800
2801 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2802 {
2803         struct inet6_dev *idev = ifp->idev;
2804         struct net_device *dev = idev->dev;
2805
2806         addrconf_join_solict(dev, &ifp->addr);
2807
2808         net_srandom(ifp->addr.s6_addr32[3]);
2809
2810         read_lock_bh(&idev->lock);
2811         if (ifp->dead)
2812                 goto out;
2813
2814         spin_lock(&ifp->lock);
2815         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2816             idev->cnf.accept_dad < 1 ||
2817             !(ifp->flags&IFA_F_TENTATIVE) ||
2818             ifp->flags & IFA_F_NODAD) {
2819                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2820                 spin_unlock(&ifp->lock);
2821                 read_unlock_bh(&idev->lock);
2822
2823                 addrconf_dad_completed(ifp);
2824                 return;
2825         }
2826
2827         if (!(idev->if_flags & IF_READY)) {
2828                 spin_unlock(&ifp->lock);
2829                 read_unlock_bh(&idev->lock);
2830                 /*
2831                  * If the device is not ready:
2832                  * - keep it tentative if it is a permanent address.
2833                  * - otherwise, kill it.
2834                  */
2835                 in6_ifa_hold(ifp);
2836                 addrconf_dad_stop(ifp, 0);
2837                 return;
2838         }
2839
2840         /*
2841          * Optimistic nodes can start receiving
2842          * Frames right away
2843          */
2844         if(ifp->flags & IFA_F_OPTIMISTIC)
2845                 ip6_ins_rt(ifp->rt);
2846
2847         addrconf_dad_kick(ifp);
2848         spin_unlock(&ifp->lock);
2849 out:
2850         read_unlock_bh(&idev->lock);
2851 }
2852
2853 static void addrconf_dad_timer(unsigned long data)
2854 {
2855         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2856         struct inet6_dev *idev = ifp->idev;
2857         struct in6_addr mcaddr;
2858
2859         read_lock(&idev->lock);
2860         if (idev->dead || !(idev->if_flags & IF_READY)) {
2861                 read_unlock(&idev->lock);
2862                 goto out;
2863         }
2864
2865         spin_lock(&ifp->lock);
2866         if (ifp->probes == 0) {
2867                 /*
2868                  * DAD was successful
2869                  */
2870
2871                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2872                 spin_unlock(&ifp->lock);
2873                 read_unlock(&idev->lock);
2874
2875                 addrconf_dad_completed(ifp);
2876
2877                 goto out;
2878         }
2879
2880         ifp->probes--;
2881         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2882         spin_unlock(&ifp->lock);
2883         read_unlock(&idev->lock);
2884
2885         /* send a neighbour solicitation for our addr */
2886         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2887         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2888 out:
2889         in6_ifa_put(ifp);
2890 }
2891
2892 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2893 {
2894         struct net_device *     dev = ifp->idev->dev;
2895
2896         /*
2897          *      Configure the address for reception. Now it is valid.
2898          */
2899
2900         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2901
2902         /* If added prefix is link local and forwarding is off,
2903            start sending router solicitations.
2904          */
2905
2906         if (ifp->idev->cnf.forwarding == 0 &&
2907             ifp->idev->cnf.rtr_solicits > 0 &&
2908             (dev->flags&IFF_LOOPBACK) == 0 &&
2909             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2910                 /*
2911                  *      If a host as already performed a random delay
2912                  *      [...] as part of DAD [...] there is no need
2913                  *      to delay again before sending the first RS
2914                  */
2915                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2916
2917                 spin_lock_bh(&ifp->lock);
2918                 ifp->probes = 1;
2919                 ifp->idev->if_flags |= IF_RS_SENT;
2920                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2921                 spin_unlock_bh(&ifp->lock);
2922         }
2923 }
2924
2925 static void addrconf_dad_run(struct inet6_dev *idev) {
2926         struct inet6_ifaddr *ifp;
2927
2928         read_lock_bh(&idev->lock);
2929         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2930                 spin_lock(&ifp->lock);
2931                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2932                         spin_unlock(&ifp->lock);
2933                         continue;
2934                 }
2935                 spin_unlock(&ifp->lock);
2936                 addrconf_dad_kick(ifp);
2937         }
2938         read_unlock_bh(&idev->lock);
2939 }
2940
2941 #ifdef CONFIG_PROC_FS
2942 struct if6_iter_state {
2943         struct seq_net_private p;
2944         int bucket;
2945 };
2946
2947 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2948 {
2949         struct inet6_ifaddr *ifa = NULL;
2950         struct if6_iter_state *state = seq->private;
2951         struct net *net = seq_file_net(seq);
2952
2953         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2954                 struct hlist_node *n;
2955                 hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
2956                                          addr_lst)
2957                         if (net_eq(dev_net(ifa->idev->dev), net))
2958                                 return ifa;
2959         }
2960         return NULL;
2961 }
2962
2963 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2964                                          struct inet6_ifaddr *ifa)
2965 {
2966         struct if6_iter_state *state = seq->private;
2967         struct net *net = seq_file_net(seq);
2968         struct hlist_node *n = &ifa->addr_lst;
2969
2970         hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
2971                 if (net_eq(dev_net(ifa->idev->dev), net))
2972                         return ifa;
2973
2974         while (++state->bucket < IN6_ADDR_HSIZE) {
2975                 hlist_for_each_entry(ifa, n,
2976                                      &inet6_addr_lst[state->bucket], addr_lst) {
2977                         if (net_eq(dev_net(ifa->idev->dev), net))
2978                                 return ifa;
2979                 }
2980         }
2981
2982         return NULL;
2983 }
2984
2985 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2986 {
2987         struct inet6_ifaddr *ifa = if6_get_first(seq);
2988
2989         if (ifa)
2990                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2991                         --pos;
2992         return pos ? NULL : ifa;
2993 }
2994
2995 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2996         __acquires(rcu)
2997 {
2998         rcu_read_lock_bh();
2999         return if6_get_idx(seq, *pos);
3000 }
3001
3002 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3003 {
3004         struct inet6_ifaddr *ifa;
3005
3006         ifa = if6_get_next(seq, v);
3007         ++*pos;
3008         return ifa;
3009 }
3010
3011 static void if6_seq_stop(struct seq_file *seq, void *v)
3012         __releases(rcu)
3013 {
3014         rcu_read_unlock_bh();
3015 }
3016
3017 static int if6_seq_show(struct seq_file *seq, void *v)
3018 {
3019         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3020         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3021                    &ifp->addr,
3022                    ifp->idev->dev->ifindex,
3023                    ifp->prefix_len,
3024                    ifp->scope,
3025                    ifp->flags,
3026                    ifp->idev->dev->name);
3027         return 0;
3028 }
3029
3030 static const struct seq_operations if6_seq_ops = {
3031         .start  = if6_seq_start,
3032         .next   = if6_seq_next,
3033         .show   = if6_seq_show,
3034         .stop   = if6_seq_stop,
3035 };
3036
3037 static int if6_seq_open(struct inode *inode, struct file *file)
3038 {
3039         return seq_open_net(inode, file, &if6_seq_ops,
3040                             sizeof(struct if6_iter_state));
3041 }
3042
3043 static const struct file_operations if6_fops = {
3044         .owner          = THIS_MODULE,
3045         .open           = if6_seq_open,
3046         .read           = seq_read,
3047         .llseek         = seq_lseek,
3048         .release        = seq_release_net,
3049 };
3050
3051 static int __net_init if6_proc_net_init(struct net *net)
3052 {
3053         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3054                 return -ENOMEM;
3055         return 0;
3056 }
3057
3058 static void __net_exit if6_proc_net_exit(struct net *net)
3059 {
3060        proc_net_remove(net, "if_inet6");
3061 }
3062
3063 static struct pernet_operations if6_proc_net_ops = {
3064        .init = if6_proc_net_init,
3065        .exit = if6_proc_net_exit,
3066 };
3067
3068 int __init if6_proc_init(void)
3069 {
3070         return register_pernet_subsys(&if6_proc_net_ops);
3071 }
3072
3073 void if6_proc_exit(void)
3074 {
3075         unregister_pernet_subsys(&if6_proc_net_ops);
3076 }
3077 #endif  /* CONFIG_PROC_FS */
3078
3079 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3080 /* Check if address is a home address configured on any interface. */
3081 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3082 {
3083         int ret = 0;
3084         struct inet6_ifaddr *ifp = NULL;
3085         struct hlist_node *n;
3086         u8 hash = ipv6_addr_hash(addr);
3087
3088         rcu_read_lock_bh();
3089         hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3090                 if (!net_eq(dev_net(ifp->idev->dev), net))
3091                         continue;
3092                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3093                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3094                         ret = 1;
3095                         break;
3096                 }
3097         }
3098         rcu_read_unlock_bh();
3099         return ret;
3100 }
3101 #endif
3102
3103 /*
3104  *      Periodic address status verification
3105  */
3106
3107 static void addrconf_verify(unsigned long foo)
3108 {
3109         struct inet6_ifaddr *ifp;
3110         struct hlist_node *node;
3111         unsigned long now, next;
3112         int i;
3113
3114         rcu_read_lock_bh();
3115         spin_lock(&addrconf_verify_lock);
3116         now = jiffies;
3117         next = now + ADDR_CHECK_FREQUENCY;
3118
3119         del_timer(&addr_chk_timer);
3120
3121         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3122
3123 restart:
3124                 hlist_for_each_entry_rcu(ifp, node,
3125                                          &inet6_addr_lst[i], addr_lst) {
3126                         unsigned long age;
3127 #ifdef CONFIG_IPV6_PRIVACY
3128                         unsigned long regen_advance;
3129 #endif
3130
3131                         if (ifp->flags & IFA_F_PERMANENT)
3132                                 continue;
3133
3134                         spin_lock(&ifp->lock);
3135                         age = (now - ifp->tstamp) / HZ;
3136
3137 #ifdef CONFIG_IPV6_PRIVACY
3138                         regen_advance = ifp->idev->cnf.regen_max_retry *
3139                                         ifp->idev->cnf.dad_transmits *
3140                                         ifp->idev->nd_parms->retrans_time / HZ;
3141 #endif
3142
3143                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3144                             age >= ifp->valid_lft) {
3145                                 spin_unlock(&ifp->lock);
3146                                 in6_ifa_hold(ifp);
3147                                 ipv6_del_addr(ifp);
3148                                 goto restart;
3149                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3150                                 spin_unlock(&ifp->lock);
3151                                 continue;
3152                         } else if (age >= ifp->prefered_lft) {
3153                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3154                                 int deprecate = 0;
3155
3156                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3157                                         deprecate = 1;
3158                                         ifp->flags |= IFA_F_DEPRECATED;
3159                                 }
3160
3161                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3162                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3163
3164                                 spin_unlock(&ifp->lock);
3165
3166                                 if (deprecate) {
3167                                         in6_ifa_hold(ifp);
3168
3169                                         ipv6_ifa_notify(0, ifp);
3170                                         in6_ifa_put(ifp);
3171                                         goto restart;
3172                                 }
3173 #ifdef CONFIG_IPV6_PRIVACY
3174                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3175                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3176                                 if (age >= ifp->prefered_lft - regen_advance) {
3177                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3178                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3179                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3180                                         if (!ifp->regen_count && ifpub) {
3181                                                 ifp->regen_count++;
3182                                                 in6_ifa_hold(ifp);
3183                                                 in6_ifa_hold(ifpub);
3184                                                 spin_unlock(&ifp->lock);
3185
3186                                                 spin_lock(&ifpub->lock);
3187                                                 ifpub->regen_count = 0;
3188                                                 spin_unlock(&ifpub->lock);
3189                                                 ipv6_create_tempaddr(ifpub, ifp);
3190                                                 in6_ifa_put(ifpub);
3191                                                 in6_ifa_put(ifp);
3192                                                 goto restart;
3193                                         }
3194                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3195                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3196                                 spin_unlock(&ifp->lock);
3197 #endif
3198                         } else {
3199                                 /* ifp->prefered_lft <= ifp->valid_lft */
3200                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3201                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3202                                 spin_unlock(&ifp->lock);
3203                         }
3204                 }
3205         }
3206
3207         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3208         add_timer(&addr_chk_timer);
3209         spin_unlock(&addrconf_verify_lock);
3210         rcu_read_unlock_bh();
3211 }
3212
3213 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3214 {
3215         struct in6_addr *pfx = NULL;
3216
3217         if (addr)
3218                 pfx = nla_data(addr);
3219
3220         if (local) {
3221                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3222                         pfx = NULL;
3223                 else
3224                         pfx = nla_data(local);
3225         }
3226
3227         return pfx;
3228 }
3229
3230 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3231         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3232         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3233         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3234 };
3235
3236 static int
3237 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3238 {
3239         struct net *net = sock_net(skb->sk);
3240         struct ifaddrmsg *ifm;
3241         struct nlattr *tb[IFA_MAX+1];
3242         struct in6_addr *pfx;
3243         int err;
3244
3245         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3246         if (err < 0)
3247                 return err;
3248
3249         ifm = nlmsg_data(nlh);
3250         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3251         if (pfx == NULL)
3252                 return -EINVAL;
3253
3254         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3255 }
3256
3257 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3258                              u32 prefered_lft, u32 valid_lft)
3259 {
3260         u32 flags;
3261         clock_t expires;
3262         unsigned long timeout;
3263
3264         if (!valid_lft || (prefered_lft > valid_lft))
3265                 return -EINVAL;
3266
3267         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3268         if (addrconf_finite_timeout(timeout)) {
3269                 expires = jiffies_to_clock_t(timeout * HZ);
3270                 valid_lft = timeout;
3271                 flags = RTF_EXPIRES;
3272         } else {
3273                 expires = 0;
3274                 flags = 0;
3275                 ifa_flags |= IFA_F_PERMANENT;
3276         }
3277
3278         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3279         if (addrconf_finite_timeout(timeout)) {
3280                 if (timeout == 0)
3281                         ifa_flags |= IFA_F_DEPRECATED;
3282                 prefered_lft = timeout;
3283         }
3284
3285         spin_lock_bh(&ifp->lock);
3286         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3287         ifp->tstamp = jiffies;
3288         ifp->valid_lft = valid_lft;
3289         ifp->prefered_lft = prefered_lft;
3290
3291         spin_unlock_bh(&ifp->lock);
3292         if (!(ifp->flags&IFA_F_TENTATIVE))
3293                 ipv6_ifa_notify(0, ifp);
3294
3295         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3296                               expires, flags);
3297         addrconf_verify(0);
3298
3299         return 0;
3300 }
3301
3302 static int
3303 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3304 {
3305         struct net *net = sock_net(skb->sk);
3306         struct ifaddrmsg *ifm;
3307         struct nlattr *tb[IFA_MAX+1];
3308         struct in6_addr *pfx;
3309         struct inet6_ifaddr *ifa;
3310         struct net_device *dev;
3311         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3312         u8 ifa_flags;
3313         int err;
3314
3315         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3316         if (err < 0)
3317                 return err;
3318
3319         ifm = nlmsg_data(nlh);
3320         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3321         if (pfx == NULL)
3322                 return -EINVAL;
3323
3324         if (tb[IFA_CACHEINFO]) {
3325                 struct ifa_cacheinfo *ci;
3326
3327                 ci = nla_data(tb[IFA_CACHEINFO]);
3328                 valid_lft = ci->ifa_valid;
3329                 preferred_lft = ci->ifa_prefered;
3330         } else {
3331                 preferred_lft = INFINITY_LIFE_TIME;
3332                 valid_lft = INFINITY_LIFE_TIME;
3333         }
3334
3335         dev =  __dev_get_by_index(net, ifm->ifa_index);
3336         if (dev == NULL)
3337                 return -ENODEV;
3338
3339         /* We ignore other flags so far. */
3340         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3341
3342         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3343         if (ifa == NULL) {
3344                 /*
3345                  * It would be best to check for !NLM_F_CREATE here but
3346                  * userspace alreay relies on not having to provide this.
3347                  */
3348                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3349                                       ifm->ifa_prefixlen, ifa_flags,
3350                                       preferred_lft, valid_lft);
3351         }
3352
3353         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3354             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3355                 err = -EEXIST;
3356         else
3357                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3358
3359         in6_ifa_put(ifa);
3360
3361         return err;
3362 }
3363
3364 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3365                           u8 scope, int ifindex)
3366 {
3367         struct ifaddrmsg *ifm;
3368
3369         ifm = nlmsg_data(nlh);
3370         ifm->ifa_family = AF_INET6;
3371         ifm->ifa_prefixlen = prefixlen;
3372         ifm->ifa_flags = flags;
3373         ifm->ifa_scope = scope;
3374         ifm->ifa_index = ifindex;
3375 }
3376
3377 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3378                          unsigned long tstamp, u32 preferred, u32 valid)
3379 {
3380         struct ifa_cacheinfo ci;
3381
3382         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3383                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3384         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3385                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3386         ci.ifa_prefered = preferred;
3387         ci.ifa_valid = valid;
3388
3389         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3390 }
3391
3392 static inline int rt_scope(int ifa_scope)
3393 {
3394         if (ifa_scope & IFA_HOST)
3395                 return RT_SCOPE_HOST;
3396         else if (ifa_scope & IFA_LINK)
3397                 return RT_SCOPE_LINK;
3398         else if (ifa_scope & IFA_SITE)
3399                 return RT_SCOPE_SITE;
3400         else
3401                 return RT_SCOPE_UNIVERSE;
3402 }
3403
3404 static inline int inet6_ifaddr_msgsize(void)
3405 {
3406         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3407                + nla_total_size(16) /* IFA_ADDRESS */
3408                + nla_total_size(sizeof(struct ifa_cacheinfo));
3409 }
3410
3411 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3412                              u32 pid, u32 seq, int event, unsigned int flags)
3413 {
3414         struct nlmsghdr  *nlh;
3415         u32 preferred, valid;
3416
3417         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3418         if (nlh == NULL)
3419                 return -EMSGSIZE;
3420
3421         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3422                       ifa->idev->dev->ifindex);
3423
3424         if (!(ifa->flags&IFA_F_PERMANENT)) {
3425                 preferred = ifa->prefered_lft;
3426                 valid = ifa->valid_lft;
3427                 if (preferred != INFINITY_LIFE_TIME) {
3428                         long tval = (jiffies - ifa->tstamp)/HZ;
3429                         if (preferred > tval)
3430                                 preferred -= tval;
3431                         else
3432                                 preferred = 0;
3433                         if (valid != INFINITY_LIFE_TIME)
3434                                 valid -= tval;
3435                 }
3436         } else {
3437                 preferred = INFINITY_LIFE_TIME;
3438                 valid = INFINITY_LIFE_TIME;
3439         }
3440
3441         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3442             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3443                 nlmsg_cancel(skb, nlh);
3444                 return -EMSGSIZE;
3445         }
3446
3447         return nlmsg_end(skb, nlh);
3448 }
3449
3450 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3451                                 u32 pid, u32 seq, int event, u16 flags)
3452 {
3453         struct nlmsghdr  *nlh;
3454         u8 scope = RT_SCOPE_UNIVERSE;
3455         int ifindex = ifmca->idev->dev->ifindex;
3456
3457         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3458                 scope = RT_SCOPE_SITE;
3459
3460         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3461         if (nlh == NULL)
3462                 return -EMSGSIZE;
3463
3464         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3465         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3466             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3467                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3468                 nlmsg_cancel(skb, nlh);
3469                 return -EMSGSIZE;
3470         }
3471
3472         return nlmsg_end(skb, nlh);
3473 }
3474
3475 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3476                                 u32 pid, u32 seq, int event, unsigned int flags)
3477 {
3478         struct nlmsghdr  *nlh;
3479         u8 scope = RT_SCOPE_UNIVERSE;
3480         int ifindex = ifaca->aca_idev->dev->ifindex;
3481
3482         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3483                 scope = RT_SCOPE_SITE;
3484
3485         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3486         if (nlh == NULL)
3487                 return -EMSGSIZE;
3488
3489         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3490         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3491             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3492                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3493                 nlmsg_cancel(skb, nlh);
3494                 return -EMSGSIZE;
3495         }
3496
3497         return nlmsg_end(skb, nlh);
3498 }
3499
3500 enum addr_type_t
3501 {
3502         UNICAST_ADDR,
3503         MULTICAST_ADDR,
3504         ANYCAST_ADDR,
3505 };
3506
3507 /* called with rcu_read_lock() */
3508 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3509                           struct netlink_callback *cb, enum addr_type_t type,
3510                           int s_ip_idx, int *p_ip_idx)
3511 {
3512         struct inet6_ifaddr *ifa;
3513         struct ifmcaddr6 *ifmca;
3514         struct ifacaddr6 *ifaca;
3515         int err = 1;
3516         int ip_idx = *p_ip_idx;
3517
3518         read_lock_bh(&idev->lock);
3519         switch (type) {
3520         case UNICAST_ADDR:
3521                 /* unicast address incl. temp addr */
3522                 for (ifa = idev->addr_list; ifa;
3523                      ifa = ifa->if_next, ip_idx++) {
3524                         if (ip_idx < s_ip_idx)
3525                                 continue;
3526                         err = inet6_fill_ifaddr(skb, ifa,
3527                                                 NETLINK_CB(cb->skb).pid,
3528                                                 cb->nlh->nlmsg_seq,
3529                                                 RTM_NEWADDR,
3530                                                 NLM_F_MULTI);
3531                         if (err <= 0)
3532                                 break;
3533                 }
3534                 break;
3535         case MULTICAST_ADDR:
3536                 /* multicast address */
3537                 for (ifmca = idev->mc_list; ifmca;
3538                      ifmca = ifmca->next, ip_idx++) {
3539                         if (ip_idx < s_ip_idx)
3540                                 continue;
3541                         err = inet6_fill_ifmcaddr(skb, ifmca,
3542                                                   NETLINK_CB(cb->skb).pid,
3543                                                   cb->nlh->nlmsg_seq,
3544                                                   RTM_GETMULTICAST,
3545                                                   NLM_F_MULTI);
3546                         if (err <= 0)
3547                                 break;
3548                 }
3549                 break;
3550         case ANYCAST_ADDR:
3551                 /* anycast address */
3552                 for (ifaca = idev->ac_list; ifaca;
3553                      ifaca = ifaca->aca_next, ip_idx++) {
3554                         if (ip_idx < s_ip_idx)
3555                                 continue;
3556                         err = inet6_fill_ifacaddr(skb, ifaca,
3557                                                   NETLINK_CB(cb->skb).pid,
3558                                                   cb->nlh->nlmsg_seq,
3559                                                   RTM_GETANYCAST,
3560                                                   NLM_F_MULTI);
3561                         if (err <= 0)
3562                                 break;
3563                 }
3564                 break;
3565         default:
3566                 break;
3567         }
3568         read_unlock_bh(&idev->lock);
3569         *p_ip_idx = ip_idx;
3570         return err;
3571 }
3572
3573 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3574                            enum addr_type_t type)
3575 {
3576         struct net *net = sock_net(skb->sk);
3577         int h, s_h;
3578         int idx, ip_idx;
3579         int s_idx, s_ip_idx;
3580         struct net_device *dev;
3581         struct inet6_dev *idev;
3582         struct hlist_head *head;
3583         struct hlist_node *node;
3584
3585         s_h = cb->args[0];
3586         s_idx = idx = cb->args[1];
3587         s_ip_idx = ip_idx = cb->args[2];
3588
3589         rcu_read_lock();
3590         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3591                 idx = 0;
3592                 head = &net->dev_index_head[h];
3593                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3594                         if (idx < s_idx)
3595                                 goto cont;
3596                         if (idx > s_idx)
3597                                 s_ip_idx = 0;
3598                         ip_idx = 0;
3599                         if ((idev = __in6_dev_get(dev)) == NULL)
3600                                 goto cont;
3601
3602                         if (in6_dump_addrs(idev, skb, cb, type,
3603                                            s_ip_idx, &ip_idx) <= 0)
3604                                 goto done;
3605 cont:
3606                         idx++;
3607                 }
3608         }
3609 done:
3610         rcu_read_unlock();
3611         cb->args[0] = h;
3612         cb->args[1] = idx;
3613         cb->args[2] = ip_idx;
3614
3615         return skb->len;
3616 }
3617
3618 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3619 {
3620         enum addr_type_t type = UNICAST_ADDR;
3621
3622         return inet6_dump_addr(skb, cb, type);
3623 }
3624
3625 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3626 {
3627         enum addr_type_t type = MULTICAST_ADDR;
3628
3629         return inet6_dump_addr(skb, cb, type);
3630 }
3631
3632
3633 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3634 {
3635         enum addr_type_t type = ANYCAST_ADDR;
3636
3637         return inet6_dump_addr(skb, cb, type);
3638 }
3639
3640 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3641                              void *arg)
3642 {
3643         struct net *net = sock_net(in_skb->sk);
3644         struct ifaddrmsg *ifm;
3645         struct nlattr *tb[IFA_MAX+1];
3646         struct in6_addr *addr = NULL;
3647         struct net_device *dev = NULL;
3648         struct inet6_ifaddr *ifa;
3649         struct sk_buff *skb;
3650         int err;
3651
3652         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3653         if (err < 0)
3654                 goto errout;
3655
3656         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3657         if (addr == NULL) {
3658                 err = -EINVAL;
3659                 goto errout;
3660         }
3661
3662         ifm = nlmsg_data(nlh);
3663         if (ifm->ifa_index)
3664                 dev = __dev_get_by_index(net, ifm->ifa_index);
3665
3666         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3667                 err = -EADDRNOTAVAIL;
3668                 goto errout;
3669         }
3670
3671         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3672                 err = -ENOBUFS;
3673                 goto errout_ifa;
3674         }
3675
3676         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3677                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3678         if (err < 0) {
3679                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3680                 WARN_ON(err == -EMSGSIZE);
3681                 kfree_skb(skb);
3682                 goto errout_ifa;
3683         }
3684         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3685 errout_ifa:
3686         in6_ifa_put(ifa);
3687 errout:
3688         return err;
3689 }
3690
3691 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3692 {
3693         struct sk_buff *skb;
3694         struct net *net = dev_net(ifa->idev->dev);
3695         int err = -ENOBUFS;
3696
3697         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3698         if (skb == NULL)
3699                 goto errout;
3700
3701         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3702         if (err < 0) {
3703                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3704                 WARN_ON(err == -EMSGSIZE);
3705                 kfree_skb(skb);
3706                 goto errout;
3707         }
3708         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3709         return;
3710 errout:
3711         if (err < 0)
3712                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3713 }
3714
3715 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3716                                 __s32 *array, int bytes)
3717 {
3718         BUG_ON(bytes < (DEVCONF_MAX * 4));
3719
3720         memset(array, 0, bytes);
3721         array[DEVCONF_FORWARDING] = cnf->forwarding;
3722         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3723         array[DEVCONF_MTU6] = cnf->mtu6;
3724         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3725         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3726         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3727         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3728         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3729         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3730         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3731         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3732 #ifdef CONFIG_IPV6_PRIVACY
3733         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3734         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3735         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3736         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3737         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3738 #endif
3739         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3740         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3741         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3742 #ifdef CONFIG_IPV6_ROUTER_PREF
3743         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3744         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3745 #ifdef CONFIG_IPV6_ROUTE_INFO
3746         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3747 #endif
3748 #endif
3749         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3750         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3751 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3752         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3753 #endif
3754 #ifdef CONFIG_IPV6_MROUTE
3755         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3756 #endif
3757         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3758         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3759         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3760 }
3761
3762 static inline size_t inet6_if_nlmsg_size(void)
3763 {
3764         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3765                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3766                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3767                + nla_total_size(4) /* IFLA_MTU */
3768                + nla_total_size(4) /* IFLA_LINK */
3769                + nla_total_size( /* IFLA_PROTINFO */
3770                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3771                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3772                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3773                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3774                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3775                  );
3776 }
3777
3778 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3779                                       int items, int bytes)
3780 {
3781         int i;
3782         int pad = bytes - sizeof(u64) * items;
3783         BUG_ON(pad < 0);
3784
3785         /* Use put_unaligned() because stats may not be aligned for u64. */
3786         put_unaligned(items, &stats[0]);
3787         for (i = 1; i < items; i++)
3788                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3789
3790         memset(&stats[items], 0, pad);
3791 }
3792
3793 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3794                              int bytes)
3795 {
3796         switch(attrtype) {
3797         case IFLA_INET6_STATS:
3798                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3799                 break;
3800         case IFLA_INET6_ICMP6STATS:
3801                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3802                 break;
3803         }
3804 }
3805
3806 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3807                              u32 pid, u32 seq, int event, unsigned int flags)
3808 {
3809         struct net_device *dev = idev->dev;
3810         struct nlattr *nla;
3811         struct ifinfomsg *hdr;
3812         struct nlmsghdr *nlh;
3813         void *protoinfo;
3814         struct ifla_cacheinfo ci;
3815
3816         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3817         if (nlh == NULL)
3818                 return -EMSGSIZE;
3819
3820         hdr = nlmsg_data(nlh);
3821         hdr->ifi_family = AF_INET6;
3822         hdr->__ifi_pad = 0;
3823         hdr->ifi_type = dev->type;
3824         hdr->ifi_index = dev->ifindex;
3825         hdr->ifi_flags = dev_get_flags(dev);
3826         hdr->ifi_change = 0;
3827
3828         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3829
3830         if (dev->addr_len)
3831                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3832
3833         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3834         if (dev->ifindex != dev->iflink)
3835                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3836
3837         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3838         if (protoinfo == NULL)
3839                 goto nla_put_failure;
3840
3841         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3842
3843         ci.max_reasm_len = IPV6_MAXPLEN;
3844         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3845                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3846         ci.reachable_time = idev->nd_parms->reachable_time;
3847         ci.retrans_time = idev->nd_parms->retrans_time;
3848         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3849
3850         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3851         if (nla == NULL)
3852                 goto nla_put_failure;
3853         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3854
3855         /* XXX - MC not implemented */
3856
3857         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3858         if (nla == NULL)
3859                 goto nla_put_failure;
3860         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3861
3862         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3863         if (nla == NULL)
3864                 goto nla_put_failure;
3865         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3866
3867         nla_nest_end(skb, protoinfo);
3868         return nlmsg_end(skb, nlh);
3869
3870 nla_put_failure:
3871         nlmsg_cancel(skb, nlh);
3872         return -EMSGSIZE;
3873 }
3874
3875 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3876 {
3877         struct net *net = sock_net(skb->sk);
3878         int h, s_h;
3879         int idx = 0, s_idx;
3880         struct net_device *dev;
3881         struct inet6_dev *idev;
3882         struct hlist_head *head;
3883         struct hlist_node *node;
3884
3885         s_h = cb->args[0];
3886         s_idx = cb->args[1];
3887
3888         rcu_read_lock();
3889         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3890                 idx = 0;
3891                 head = &net->dev_index_head[h];
3892                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3893                         if (idx < s_idx)
3894                                 goto cont;
3895                         idev = __in6_dev_get(dev);
3896                         if (!idev)
3897                                 goto cont;
3898                         if (inet6_fill_ifinfo(skb, idev,
3899                                               NETLINK_CB(cb->skb).pid,
3900                                               cb->nlh->nlmsg_seq,
3901                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3902                                 goto out;
3903 cont:
3904                         idx++;
3905                 }
3906         }
3907 out:
3908         rcu_read_unlock();
3909         cb->args[1] = idx;
3910         cb->args[0] = h;
3911
3912         return skb->len;
3913 }
3914
3915 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3916 {
3917         struct sk_buff *skb;
3918         struct net *net = dev_net(idev->dev);
3919         int err = -ENOBUFS;
3920
3921         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3922         if (skb == NULL)
3923                 goto errout;
3924
3925         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3926         if (err < 0) {
3927                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3928                 WARN_ON(err == -EMSGSIZE);
3929                 kfree_skb(skb);
3930                 goto errout;
3931         }
3932         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3933         return;
3934 errout:
3935         if (err < 0)
3936                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3937 }
3938
3939 static inline size_t inet6_prefix_nlmsg_size(void)
3940 {
3941         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3942                + nla_total_size(sizeof(struct in6_addr))
3943                + nla_total_size(sizeof(struct prefix_cacheinfo));
3944 }
3945
3946 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3947                              struct prefix_info *pinfo, u32 pid, u32 seq,
3948                              int event, unsigned int flags)
3949 {
3950         struct prefixmsg *pmsg;
3951         struct nlmsghdr *nlh;
3952         struct prefix_cacheinfo ci;
3953
3954         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3955         if (nlh == NULL)
3956                 return -EMSGSIZE;
3957
3958         pmsg = nlmsg_data(nlh);
3959         pmsg->prefix_family = AF_INET6;
3960         pmsg->prefix_pad1 = 0;
3961         pmsg->prefix_pad2 = 0;
3962         pmsg->prefix_ifindex = idev->dev->ifindex;
3963         pmsg->prefix_len = pinfo->prefix_len;
3964         pmsg->prefix_type = pinfo->type;
3965         pmsg->prefix_pad3 = 0;
3966         pmsg->prefix_flags = 0;
3967         if (pinfo->onlink)
3968                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3969         if (pinfo->autoconf)
3970                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3971
3972         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3973
3974         ci.preferred_time = ntohl(pinfo->prefered);
3975         ci.valid_time = ntohl(pinfo->valid);
3976         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3977
3978         return nlmsg_end(skb, nlh);
3979
3980 nla_put_failure:
3981         nlmsg_cancel(skb, nlh);
3982         return -EMSGSIZE;
3983 }
3984
3985 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3986                          struct prefix_info *pinfo)
3987 {
3988         struct sk_buff *skb;
3989         struct net *net = dev_net(idev->dev);
3990         int err = -ENOBUFS;
3991
3992         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3993         if (skb == NULL)
3994                 goto errout;
3995
3996         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3997         if (err < 0) {
3998                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3999                 WARN_ON(err == -EMSGSIZE);
4000                 kfree_skb(skb);
4001                 goto errout;
4002         }
4003         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4004         return;
4005 errout:
4006         if (err < 0)
4007                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4008 }
4009
4010 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4011 {
4012         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4013
4014         switch (event) {
4015         case RTM_NEWADDR:
4016                 /*
4017                  * If the address was optimistic
4018                  * we inserted the route at the start of
4019                  * our DAD process, so we don't need
4020                  * to do it again
4021                  */
4022                 if (!(ifp->rt->rt6i_node))
4023                         ip6_ins_rt(ifp->rt);
4024                 if (ifp->idev->cnf.forwarding)
4025                         addrconf_join_anycast(ifp);
4026                 break;
4027         case RTM_DELADDR:
4028                 if (ifp->idev->cnf.forwarding)
4029                         addrconf_leave_anycast(ifp);
4030                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4031                 dst_hold(&ifp->rt->u.dst);
4032                 if (ip6_del_rt(ifp->rt))
4033                         dst_free(&ifp->rt->u.dst);
4034                 break;
4035         }
4036 }
4037
4038 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4039 {
4040         rcu_read_lock_bh();
4041         if (likely(ifp->idev->dead == 0))
4042                 __ipv6_ifa_notify(event, ifp);
4043         rcu_read_unlock_bh();
4044 }
4045
4046 #ifdef CONFIG_SYSCTL
4047
4048 static
4049 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4050                            void __user *buffer, size_t *lenp, loff_t *ppos)
4051 {
4052         int *valp = ctl->data;
4053         int val = *valp;
4054         loff_t pos = *ppos;
4055         int ret;
4056
4057         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4058
4059         if (write)
4060                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4061         if (ret)
4062                 *ppos = pos;
4063         return ret;
4064 }
4065
4066 static void dev_disable_change(struct inet6_dev *idev)
4067 {
4068         if (!idev || !idev->dev)
4069                 return;
4070
4071         if (idev->cnf.disable_ipv6)
4072                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4073         else
4074                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4075 }
4076
4077 static void addrconf_disable_change(struct net *net, __s32 newf)
4078 {
4079         struct net_device *dev;
4080         struct inet6_dev *idev;
4081
4082         rcu_read_lock();
4083         for_each_netdev_rcu(net, dev) {
4084                 idev = __in6_dev_get(dev);
4085                 if (idev) {
4086                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4087                         idev->cnf.disable_ipv6 = newf;
4088                         if (changed)
4089                                 dev_disable_change(idev);
4090                 }
4091         }
4092         rcu_read_unlock();
4093 }
4094
4095 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4096 {
4097         struct net *net;
4098
4099         net = (struct net *)table->extra2;
4100
4101         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4102                 return 0;
4103
4104         if (!rtnl_trylock()) {
4105                 /* Restore the original values before restarting */
4106                 *p = old;
4107                 return restart_syscall();
4108         }
4109
4110         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4111                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4112                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4113                 addrconf_disable_change(net, newf);
4114         } else if ((!*p) ^ (!old))
4115                 dev_disable_change((struct inet6_dev *)table->extra1);
4116
4117         rtnl_unlock();
4118         return 0;
4119 }
4120
4121 static
4122 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4123                             void __user *buffer, size_t *lenp, loff_t *ppos)
4124 {
4125         int *valp = ctl->data;
4126         int val = *valp;
4127         loff_t pos = *ppos;
4128         int ret;
4129
4130         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4131
4132         if (write)
4133                 ret = addrconf_disable_ipv6(ctl, valp, val);
4134         if (ret)
4135                 *ppos = pos;
4136         return ret;
4137 }
4138
4139 static struct addrconf_sysctl_table
4140 {
4141         struct ctl_table_header *sysctl_header;
4142         ctl_table addrconf_vars[DEVCONF_MAX+1];
4143         char *dev_name;
4144 } addrconf_sysctl __read_mostly = {
4145         .sysctl_header = NULL,
4146         .addrconf_vars = {
4147                 {
4148                         .procname       =       "forwarding",
4149                         .data           =       &ipv6_devconf.forwarding,
4150                         .maxlen         =       sizeof(int),
4151                         .mode           =       0644,
4152                         .proc_handler   =       addrconf_sysctl_forward,
4153                 },
4154                 {
4155                         .procname       =       "hop_limit",
4156                         .data           =       &ipv6_devconf.hop_limit,
4157                         .maxlen         =       sizeof(int),
4158                         .mode           =       0644,
4159                         .proc_handler   =       proc_dointvec,
4160                 },
4161                 {
4162                         .procname       =       "mtu",
4163                         .data           =       &ipv6_devconf.mtu6,
4164                         .maxlen         =       sizeof(int),
4165                         .mode           =       0644,
4166                         .proc_handler   =       proc_dointvec,
4167                 },
4168                 {
4169                         .procname       =       "accept_ra",
4170                         .data           =       &ipv6_devconf.accept_ra,
4171                         .maxlen         =       sizeof(int),
4172                         .mode           =       0644,
4173                         .proc_handler   =       proc_dointvec,
4174                 },
4175                 {
4176                         .procname       =       "accept_redirects",
4177                         .data           =       &ipv6_devconf.accept_redirects,
4178                         .maxlen         =       sizeof(int),
4179                         .mode           =       0644,
4180                         .proc_handler   =       proc_dointvec,
4181                 },
4182                 {
4183                         .procname       =       "autoconf",
4184                         .data           =       &ipv6_devconf.autoconf,
4185                         .maxlen         =       sizeof(int),
4186                         .mode           =       0644,
4187                         .proc_handler   =       proc_dointvec,
4188                 },
4189                 {
4190                         .procname       =       "dad_transmits",
4191                         .data           =       &ipv6_devconf.dad_transmits,
4192                         .maxlen         =       sizeof(int),
4193                         .mode           =       0644,
4194                         .proc_handler   =       proc_dointvec,
4195                 },
4196                 {
4197                         .procname       =       "router_solicitations",
4198                         .data           =       &ipv6_devconf.rtr_solicits,
4199                         .maxlen         =       sizeof(int),
4200                         .mode           =       0644,
4201                         .proc_handler   =       proc_dointvec,
4202                 },
4203                 {
4204                         .procname       =       "router_solicitation_interval",
4205                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4206                         .maxlen         =       sizeof(int),
4207                         .mode           =       0644,
4208                         .proc_handler   =       proc_dointvec_jiffies,
4209                 },
4210                 {
4211                         .procname       =       "router_solicitation_delay",
4212                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4213                         .maxlen         =       sizeof(int),
4214                         .mode           =       0644,
4215                         .proc_handler   =       proc_dointvec_jiffies,
4216                 },
4217                 {
4218                         .procname       =       "force_mld_version",
4219                         .data           =       &ipv6_devconf.force_mld_version,
4220                         .maxlen         =       sizeof(int),
4221                         .mode           =       0644,
4222                         .proc_handler   =       proc_dointvec,
4223                 },
4224 #ifdef CONFIG_IPV6_PRIVACY
4225                 {
4226                         .procname       =       "use_tempaddr",
4227                         .data           =       &ipv6_devconf.use_tempaddr,
4228                         .maxlen         =       sizeof(int),
4229                         .mode           =       0644,
4230                         .proc_handler   =       proc_dointvec,
4231                 },
4232                 {
4233                         .procname       =       "temp_valid_lft",
4234                         .data           =       &ipv6_devconf.temp_valid_lft,
4235                         .maxlen         =       sizeof(int),
4236                         .mode           =       0644,
4237                         .proc_handler   =       proc_dointvec,
4238                 },
4239                 {
4240                         .procname       =       "temp_prefered_lft",
4241                         .data           =       &ipv6_devconf.temp_prefered_lft,
4242                         .maxlen         =       sizeof(int),
4243                         .mode           =       0644,
4244                         .proc_handler   =       proc_dointvec,
4245                 },
4246                 {
4247                         .procname       =       "regen_max_retry",
4248                         .data           =       &ipv6_devconf.regen_max_retry,
4249                         .maxlen         =       sizeof(int),
4250                         .mode           =       0644,
4251                         .proc_handler   =       proc_dointvec,
4252                 },
4253                 {
4254                         .procname       =       "max_desync_factor",
4255                         .data           =       &ipv6_devconf.max_desync_factor,
4256                         .maxlen         =       sizeof(int),
4257                         .mode           =       0644,
4258                         .proc_handler   =       proc_dointvec,
4259                 },
4260 #endif
4261                 {
4262                         .procname       =       "max_addresses",
4263                         .data           =       &ipv6_devconf.max_addresses,
4264                         .maxlen         =       sizeof(int),
4265                         .mode           =       0644,
4266                         .proc_handler   =       proc_dointvec,
4267                 },
4268                 {
4269                         .procname       =       "accept_ra_defrtr",
4270                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4271                         .maxlen         =       sizeof(int),
4272                         .mode           =       0644,
4273                         .proc_handler   =       proc_dointvec,
4274                 },
4275                 {
4276                         .procname       =       "accept_ra_pinfo",
4277                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4278                         .maxlen         =       sizeof(int),
4279                         .mode           =       0644,
4280                         .proc_handler   =       proc_dointvec,
4281                 },
4282 #ifdef CONFIG_IPV6_ROUTER_PREF
4283                 {
4284                         .procname       =       "accept_ra_rtr_pref",
4285                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4286                         .maxlen         =       sizeof(int),
4287                         .mode           =       0644,
4288                         .proc_handler   =       proc_dointvec,
4289                 },
4290                 {
4291                         .procname       =       "router_probe_interval",
4292                         .data           =       &ipv6_devconf.rtr_probe_interval,
4293                         .maxlen         =       sizeof(int),
4294                         .mode           =       0644,
4295                         .proc_handler   =       proc_dointvec_jiffies,
4296                 },
4297 #ifdef CONFIG_IPV6_ROUTE_INFO
4298                 {
4299                         .procname       =       "accept_ra_rt_info_max_plen",
4300                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4301                         .maxlen         =       sizeof(int),
4302                         .mode           =       0644,
4303                         .proc_handler   =       proc_dointvec,
4304                 },
4305 #endif
4306 #endif
4307                 {
4308                         .procname       =       "proxy_ndp",
4309                         .data           =       &ipv6_devconf.proxy_ndp,
4310                         .maxlen         =       sizeof(int),
4311                         .mode           =       0644,
4312                         .proc_handler   =       proc_dointvec,
4313                 },
4314                 {
4315                         .procname       =       "accept_source_route",
4316                         .data           =       &ipv6_devconf.accept_source_route,
4317                         .maxlen         =       sizeof(int),
4318                         .mode           =       0644,
4319                         .proc_handler   =       proc_dointvec,
4320                 },
4321 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4322                 {
4323                         .procname       =       "optimistic_dad",
4324                         .data           =       &ipv6_devconf.optimistic_dad,
4325                         .maxlen         =       sizeof(int),
4326                         .mode           =       0644,
4327                         .proc_handler   =       proc_dointvec,
4328
4329                 },
4330 #endif
4331 #ifdef CONFIG_IPV6_MROUTE
4332                 {
4333                         .procname       =       "mc_forwarding",
4334                         .data           =       &ipv6_devconf.mc_forwarding,
4335                         .maxlen         =       sizeof(int),
4336                         .mode           =       0444,
4337                         .proc_handler   =       proc_dointvec,
4338                 },
4339 #endif
4340                 {
4341                         .procname       =       "disable_ipv6",
4342                         .data           =       &ipv6_devconf.disable_ipv6,
4343                         .maxlen         =       sizeof(int),
4344                         .mode           =       0644,
4345                         .proc_handler   =       addrconf_sysctl_disable,
4346                 },
4347                 {
4348                         .procname       =       "accept_dad",
4349                         .data           =       &ipv6_devconf.accept_dad,
4350                         .maxlen         =       sizeof(int),
4351                         .mode           =       0644,
4352                         .proc_handler   =       proc_dointvec,
4353                 },
4354                 {
4355                         .procname       = "force_tllao",
4356                         .data           = &ipv6_devconf.force_tllao,
4357                         .maxlen         = sizeof(int),
4358                         .mode           = 0644,
4359                         .proc_handler   = proc_dointvec
4360                 },
4361                 {
4362                         /* sentinel */
4363                 }
4364         },
4365 };
4366
4367 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4368                 struct inet6_dev *idev, struct ipv6_devconf *p)
4369 {
4370         int i;
4371         struct addrconf_sysctl_table *t;
4372
4373 #define ADDRCONF_CTL_PATH_DEV   3
4374
4375         struct ctl_path addrconf_ctl_path[] = {
4376                 { .procname = "net", },
4377                 { .procname = "ipv6", },
4378                 { .procname = "conf", },
4379                 { /* to be set */ },
4380                 { },
4381         };
4382
4383
4384         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4385         if (t == NULL)
4386                 goto out;
4387
4388         for (i=0; t->addrconf_vars[i].data; i++) {
4389                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4390                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4391                 t->addrconf_vars[i].extra2 = net;
4392         }
4393
4394         /*
4395          * Make a copy of dev_name, because '.procname' is regarded as const
4396          * by sysctl and we wouldn't want anyone to change it under our feet
4397          * (see SIOCSIFNAME).
4398          */
4399         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4400         if (!t->dev_name)
4401                 goto free;
4402
4403         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4404
4405         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4406                         t->addrconf_vars);
4407         if (t->sysctl_header == NULL)
4408                 goto free_procname;
4409
4410         p->sysctl = t;
4411         return 0;
4412
4413 free_procname:
4414         kfree(t->dev_name);
4415 free:
4416         kfree(t);
4417 out:
4418         return -ENOBUFS;
4419 }
4420
4421 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4422 {
4423         struct addrconf_sysctl_table *t;
4424
4425         if (p->sysctl == NULL)
4426                 return;
4427
4428         t = p->sysctl;
4429         p->sysctl = NULL;
4430         unregister_sysctl_table(t->sysctl_header);
4431         kfree(t->dev_name);
4432         kfree(t);
4433 }
4434
4435 static void addrconf_sysctl_register(struct inet6_dev *idev)
4436 {
4437         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4438                               &ndisc_ifinfo_sysctl_change);
4439         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4440                                         idev, &idev->cnf);
4441 }
4442
4443 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4444 {
4445         __addrconf_sysctl_unregister(&idev->cnf);
4446         neigh_sysctl_unregister(idev->nd_parms);
4447 }
4448
4449
4450 #endif
4451
4452 static int __net_init addrconf_init_net(struct net *net)
4453 {
4454         int err;
4455         struct ipv6_devconf *all, *dflt;
4456
4457         err = -ENOMEM;
4458         all = &ipv6_devconf;
4459         dflt = &ipv6_devconf_dflt;
4460
4461         if (!net_eq(net, &init_net)) {
4462                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4463                 if (all == NULL)
4464                         goto err_alloc_all;
4465
4466                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4467                 if (dflt == NULL)
4468                         goto err_alloc_dflt;
4469         } else {
4470                 /* these will be inherited by all namespaces */
4471                 dflt->autoconf = ipv6_defaults.autoconf;
4472                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4473         }
4474
4475         net->ipv6.devconf_all = all;
4476         net->ipv6.devconf_dflt = dflt;
4477
4478 #ifdef CONFIG_SYSCTL
4479         err = __addrconf_sysctl_register(net, "all", NULL, all);
4480         if (err < 0)
4481                 goto err_reg_all;
4482
4483         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4484         if (err < 0)
4485                 goto err_reg_dflt;
4486 #endif
4487         return 0;
4488
4489 #ifdef CONFIG_SYSCTL
4490 err_reg_dflt:
4491         __addrconf_sysctl_unregister(all);
4492 err_reg_all:
4493         kfree(dflt);
4494 #endif
4495 err_alloc_dflt:
4496         kfree(all);
4497 err_alloc_all:
4498         return err;
4499 }
4500
4501 static void __net_exit addrconf_exit_net(struct net *net)
4502 {
4503 #ifdef CONFIG_SYSCTL
4504         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4505         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4506 #endif
4507         if (!net_eq(net, &init_net)) {
4508                 kfree(net->ipv6.devconf_dflt);
4509                 kfree(net->ipv6.devconf_all);
4510         }
4511 }
4512
4513 static struct pernet_operations addrconf_ops = {
4514         .init = addrconf_init_net,
4515         .exit = addrconf_exit_net,
4516 };
4517
4518 /*
4519  *      Device notifier
4520  */
4521
4522 int register_inet6addr_notifier(struct notifier_block *nb)
4523 {
4524         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4525 }
4526
4527 EXPORT_SYMBOL(register_inet6addr_notifier);
4528
4529 int unregister_inet6addr_notifier(struct notifier_block *nb)
4530 {
4531         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4532 }
4533
4534 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4535
4536 /*
4537  *      Init / cleanup code
4538  */
4539
4540 int __init addrconf_init(void)
4541 {
4542         int i, err;
4543
4544         if ((err = ipv6_addr_label_init()) < 0) {
4545                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4546                         err);
4547                 return err;
4548         }
4549
4550         register_pernet_subsys(&addrconf_ops);
4551
4552         /* The addrconf netdev notifier requires that loopback_dev
4553          * has it's ipv6 private information allocated and setup
4554          * before it can bring up and give link-local addresses
4555          * to other devices which are up.
4556          *
4557          * Unfortunately, loopback_dev is not necessarily the first
4558          * entry in the global dev_base list of net devices.  In fact,
4559          * it is likely to be the very last entry on that list.
4560          * So this causes the notifier registry below to try and
4561          * give link-local addresses to all devices besides loopback_dev
4562          * first, then loopback_dev, which cases all the non-loopback_dev
4563          * devices to fail to get a link-local address.
4564          *
4565          * So, as a temporary fix, allocate the ipv6 structure for
4566          * loopback_dev first by hand.
4567          * Longer term, all of the dependencies ipv6 has upon the loopback
4568          * device and it being up should be removed.
4569          */
4570         rtnl_lock();
4571         if (!ipv6_add_dev(init_net.loopback_dev))
4572                 err = -ENOMEM;
4573         rtnl_unlock();
4574         if (err)
4575                 goto errlo;
4576
4577         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4578                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4579
4580         register_netdevice_notifier(&ipv6_dev_notf);
4581
4582         addrconf_verify(0);
4583
4584         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4585         if (err < 0)
4586                 goto errout;
4587
4588         /* Only the first call to __rtnl_register can fail */
4589         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4590         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4591         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4592         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4593         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4594
4595         ipv6_addr_label_rtnl_register();
4596
4597         return 0;
4598 errout:
4599         unregister_netdevice_notifier(&ipv6_dev_notf);
4600 errlo:
4601         unregister_pernet_subsys(&addrconf_ops);
4602
4603         return err;
4604 }
4605
4606 void addrconf_cleanup(void)
4607 {
4608         struct net_device *dev;
4609         int i;
4610
4611         unregister_netdevice_notifier(&ipv6_dev_notf);
4612         unregister_pernet_subsys(&addrconf_ops);
4613
4614         rtnl_lock();
4615
4616         /* clean dev list */
4617         for_each_netdev(&init_net, dev) {
4618                 if (__in6_dev_get(dev) == NULL)
4619                         continue;
4620                 addrconf_ifdown(dev, 1);
4621         }
4622         addrconf_ifdown(init_net.loopback_dev, 2);
4623
4624         /*
4625          *      Check hash table.
4626          */
4627         spin_lock_bh(&addrconf_hash_lock);
4628         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4629                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4630         spin_unlock_bh(&addrconf_hash_lock);
4631
4632         del_timer(&addr_chk_timer);
4633         rtnl_unlock();
4634 }