net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 110
 111 #ifdef CONFIG_SYSCTL
 112 static void addrconf_sysctl_register(struct inet6_dev *idev);
 113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 114 #else
 115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 116 {
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 #ifdef CONFIG_IPV6_PRIVACY
 125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 127 static void ipv6_regen_rndid(unsigned long data);
 128 #endif
 129
 130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 131 static int ipv6_count_addresses(struct inet6_dev *idev);
 132
 133 /*
 134  *      Configured unicast address hash table
 135  */
 136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 137 static DEFINE_SPINLOCK(addrconf_hash_lock);
 138
 139 static void addrconf_verify(unsigned long);
 140
 141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 142 static DEFINE_SPINLOCK(addrconf_verify_lock);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 152 static void addrconf_dad_timer(unsigned long data);
 153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_run(struct inet6_dev *idev);
 155 static void addrconf_rs_timer(unsigned long data);
 156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158
 159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 160                                 struct prefix_info *pinfo);
 161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 162                                struct net_device *dev);
 163
 164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .dad_transmits          = 1,
 175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 178 #ifdef CONFIG_IPV6_PRIVACY
 179         .use_tempaddr           = 0,
 180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 182         .regen_max_retry        = REGEN_MAX_RETRY,
 183         .max_desync_factor      = MAX_DESYNC_FACTOR,
 184 #endif
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199 };
 200
 201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 202         .forwarding             = 0,
 203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 204         .mtu6                   = IPV6_MIN_MTU,
 205         .accept_ra              = 1,
 206         .accept_redirects       = 1,
 207         .autoconf               = 1,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212 #ifdef CONFIG_IPV6_PRIVACY
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218 #endif
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233 };
 234
 235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 240
 241 /* Check if a valid qdisc is available */
 242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 243 {
 244         return !qdisc_tx_is_noop(dev);
 245 }
 246
 247 /* Check if a route is valid prefix route */
 248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 249 {
 250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 251 }
 252
 253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 254 {
 255         if (del_timer(&ifp->timer))
 256                 __in6_ifa_put(ifp);
 257 }
 258
 259 enum addrconf_timer_t {
 260         AC_NONE,
 261         AC_DAD,
 262         AC_RS,
 263 };
 264
 265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 266                                enum addrconf_timer_t what,
 267                                unsigned long when)
 268 {
 269         if (!del_timer(&ifp->timer))
 270                 in6_ifa_hold(ifp);
 271
 272         switch (what) {
 273         case AC_DAD:
 274                 ifp->timer.function = addrconf_dad_timer;
 275                 break;
 276         case AC_RS:
 277                 ifp->timer.function = addrconf_rs_timer;
 278                 break;
 279         default:
 280                 break;
 281         }
 282         ifp->timer.expires = jiffies + when;
 283         add_timer(&ifp->timer);
 284 }
 285
 286 static int snmp6_alloc_dev(struct inet6_dev *idev)
 287 {
 288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 289                           sizeof(struct ipstats_mib),
 290                           __alignof__(struct ipstats_mib)) < 0)
 291                 goto err_ip;
 292         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static void snmp6_free_dev(struct inet6_dev *idev)
 312 {
 313         kfree(idev->stats.icmpv6msgdev);
 314         kfree(idev->stats.icmpv6dev);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 void in6_dev_finish_destroy(struct inet6_dev *idev)
 321 {
 322         struct net_device *dev = idev->dev;
 323
 324         WARN_ON(!list_empty(&idev->addr_list));
 325         WARN_ON(idev->mc_list != NULL);
 326
 327 #ifdef NET_REFCNT_DEBUG
 328         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 329 #endif
 330         dev_put(dev);
 331         if (!idev->dead) {
 332                 pr_warning("Freeing alive inet6 device %p\n", idev);
 333                 return;
 334         }
 335         snmp6_free_dev(idev);
 336         kfree_rcu(idev, rcu);
 337 }
 338
 339 EXPORT_SYMBOL(in6_dev_finish_destroy);
 340
 341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 342 {
 343         struct inet6_dev *ndev;
 344
 345         ASSERT_RTNL();
 346
 347         if (dev->mtu < IPV6_MIN_MTU)
 348                 return NULL;
 349
 350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 351
 352         if (ndev == NULL)
 353                 return NULL;
 354
 355         rwlock_init(&ndev->lock);
 356         ndev->dev = dev;
 357         INIT_LIST_HEAD(&ndev->addr_list);
 358
 359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 360         ndev->cnf.mtu6 = dev->mtu;
 361         ndev->cnf.sysctl = NULL;
 362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 363         if (ndev->nd_parms == NULL) {
 364                 kfree(ndev);
 365                 return NULL;
 366         }
 367         if (ndev->cnf.forwarding)
 368                 dev_disable_lro(dev);
 369         /* We refer to the device */
 370         dev_hold(dev);
 371
 372         if (snmp6_alloc_dev(ndev) < 0) {
 373                 ADBG((KERN_WARNING
 374                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 375                         __func__, dev->name));
 376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 377                 ndev->dead = 1;
 378                 in6_dev_finish_destroy(ndev);
 379                 return NULL;
 380         }
 381
 382         if (snmp6_register_dev(ndev) < 0) {
 383                 ADBG((KERN_WARNING
 384                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 385                         __func__, dev->name));
 386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 387                 ndev->dead = 1;
 388                 in6_dev_finish_destroy(ndev);
 389                 return NULL;
 390         }
 391
 392         /* One reference from device.  We must do this before
 393          * we invoke __ipv6_regen_rndid().
 394          */
 395         in6_dev_hold(ndev);
 396
 397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 398                 ndev->cnf.accept_dad = -1;
 399
 400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 402                 printk(KERN_INFO
 403                        "%s: Disabled Multicast RS\n",
 404                        dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423
 424         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 425                 ndev->if_flags |= IF_READY;
 426
 427         ipv6_mc_init_dev(ndev);
 428         ndev->tstamp = jiffies;
 429         addrconf_sysctl_register(ndev);
 430         /* protected by rtnl_lock */
 431         rcu_assign_pointer(dev->ip6_ptr, ndev);
 432
 433         /* Join all-node multicast group */
 434         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 435
 436         return ndev;
 437 }
 438
 439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 440 {
 441         struct inet6_dev *idev;
 442
 443         ASSERT_RTNL();
 444
 445         idev = __in6_dev_get(dev);
 446         if (!idev) {
 447                 idev = ipv6_add_dev(dev);
 448                 if (!idev)
 449                         return NULL;
 450         }
 451
 452         if (dev->flags&IFF_UP)
 453                 ipv6_mc_up(idev);
 454         return idev;
 455 }
 456
 457 #ifdef CONFIG_SYSCTL
 458 static void dev_forward_change(struct inet6_dev *idev)
 459 {
 460         struct net_device *dev;
 461         struct inet6_ifaddr *ifa;
 462
 463         if (!idev)
 464                 return;
 465         dev = idev->dev;
 466         if (idev->cnf.forwarding)
 467                 dev_disable_lro(dev);
 468         if (dev && (dev->flags & IFF_MULTICAST)) {
 469                 if (idev->cnf.forwarding)
 470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471                 else
 472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 473         }
 474
 475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 476                 if (ifa->flags&IFA_F_TENTATIVE)
 477                         continue;
 478                 if (idev->cnf.forwarding)
 479                         addrconf_join_anycast(ifa);
 480                 else
 481                         addrconf_leave_anycast(ifa);
 482         }
 483 }
 484
 485
 486 static void addrconf_forward_change(struct net *net, __s32 newf)
 487 {
 488         struct net_device *dev;
 489         struct inet6_dev *idev;
 490
 491         rcu_read_lock();
 492         for_each_netdev_rcu(net, dev) {
 493                 idev = __in6_dev_get(dev);
 494                 if (idev) {
 495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 496                         idev->cnf.forwarding = newf;
 497                         if (changed)
 498                                 dev_forward_change(idev);
 499                 }
 500         }
 501         rcu_read_unlock();
 502 }
 503
 504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 505 {
 506         struct net *net;
 507
 508         net = (struct net *)table->extra2;
 509         if (p == &net->ipv6.devconf_dflt->forwarding)
 510                 return 0;
 511
 512         if (!rtnl_trylock()) {
 513                 /* Restore the original values before restarting */
 514                 *p = old;
 515                 return restart_syscall();
 516         }
 517
 518         if (p == &net->ipv6.devconf_all->forwarding) {
 519                 __s32 newf = net->ipv6.devconf_all->forwarding;
 520                 net->ipv6.devconf_dflt->forwarding = newf;
 521                 addrconf_forward_change(net, newf);
 522         } else if ((!*p) ^ (!old))
 523                 dev_forward_change((struct inet6_dev *)table->extra1);
 524         rtnl_unlock();
 525
 526         if (*p)
 527                 rt6_purge_dflt_routers(net);
 528         return 1;
 529 }
 530 #endif
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 536
 537 #ifdef NET_REFCNT_DEBUG
 538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 539 #endif
 540
 541         in6_dev_put(ifp->idev);
 542
 543         if (del_timer(&ifp->timer))
 544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 545
 546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 548                 return;
 549         }
 550         dst_release(&ifp->rt->dst);
 551
 552         kfree_rcu(ifp, rcu);
 553 }
 554
 555 static void
 556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 557 {
 558         struct list_head *p;
 559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 560
 561         /*
 562          * Each device address list is sorted in order of scope -
 563          * global before linklocal.
 564          */
 565         list_for_each(p, &idev->addr_list) {
 566                 struct inet6_ifaddr *ifa
 567                         = list_entry(p, struct inet6_ifaddr, if_list);
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         list_add_tail(&ifp->if_list, p);
 573 }
 574
 575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 576 {
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581         return jhash_2words((__force u32)addr->s6_addr32[2],
 582                             (__force u32)addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         /*
 653          * part one of RFC 4429, section 3.3
 654          * We should not configure an address as
 655          * optimistic if we do not yet know the link
 656          * layer address of our nexhop router
 657          */
 658
 659         if (rt->rt6i_nexthop == NULL)
 660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 661
 662         ifa->idev = idev;
 663         in6_dev_hold(idev);
 664         /* For caller */
 665         in6_ifa_hold(ifa);
 666
 667         /* Add to big hash table */
 668         hash = ipv6_addr_hash(addr);
 669
 670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, *ifn;
 707         struct inet6_dev *idev = ifp->idev;
 708         int state;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         spin_lock_bh(&ifp->state_lock);
 713         state = ifp->state;
 714         ifp->state = INET6_IFADDR_STATE_DEAD;
 715         spin_unlock_bh(&ifp->state_lock);
 716
 717         if (state == INET6_IFADDR_STATE_DEAD)
 718                 goto out;
 719
 720         spin_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init_rcu(&ifp->addr_lst);
 722         spin_unlock_bh(&addrconf_hash_lock);
 723
 724         write_lock_bh(&idev->lock);
 725 #ifdef CONFIG_IPV6_PRIVACY
 726         if (ifp->flags&IFA_F_TEMPORARY) {
 727                 list_del(&ifp->tmp_list);
 728                 if (ifp->ifpub) {
 729                         in6_ifa_put(ifp->ifpub);
 730                         ifp->ifpub = NULL;
 731                 }
 732                 __in6_ifa_put(ifp);
 733         }
 734 #endif
 735
 736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 737                 if (ifa == ifp) {
 738                         list_del_init(&ifp->if_list);
 739                         __in6_ifa_put(ifp);
 740
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773         }
 774         write_unlock_bh(&idev->lock);
 775
 776         addrconf_del_timer(ifp);
 777
 778         ipv6_ifa_notify(RTM_DELADDR, ifp);
 779
 780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 781
 782         /*
 783          * Purge or update corresponding prefix
 784          *
 785          * 1) we don't purge prefix here if address was not permanent.
 786          *    prefix is managed by its own lifetime.
 787          * 2) if there're no addresses, delete prefix.
 788          * 3) if there're still other permanent address(es),
 789          *    corresponding prefix is still permanent.
 790          * 4) otherwise, update prefix lifetime to the
 791          *    longest valid lifetime among the corresponding
 792          *    addresses on the device.
 793          *    Note: subsequent RA will update lifetime.
 794          *
 795          * --yoshfuji
 796          */
 797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 798                 struct in6_addr prefix;
 799                 struct rt6_info *rt;
 800                 struct net *net = dev_net(ifp->idev->dev);
 801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 803
 804                 if (rt && addrconf_is_prefix_route(rt)) {
 805                         if (onlink == 0) {
 806                                 ip6_del_rt(rt);
 807                                 rt = NULL;
 808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 809                                 rt->rt6i_expires = expires;
 810                                 rt->rt6i_flags |= RTF_EXPIRES;
 811                         }
 812                 }
 813                 dst_release(&rt->dst);
 814         }
 815
 816         /* clean up prefsrc entries */
 817         rt6_remove_prefsrc(ifp);
 818 out:
 819         in6_ifa_put(ifp);
 820 }
 821
 822 #ifdef CONFIG_IPV6_PRIVACY
 823 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 824 {
 825         struct inet6_dev *idev = ifp->idev;
 826         struct in6_addr addr, *tmpaddr;
 827         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 828         unsigned long regen_advance;
 829         int tmp_plen;
 830         int ret = 0;
 831         int max_addresses;
 832         u32 addr_flags;
 833
 834         write_lock(&idev->lock);
 835         if (ift) {
 836                 spin_lock_bh(&ift->lock);
 837                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 838                 spin_unlock_bh(&ift->lock);
 839                 tmpaddr = &addr;
 840         } else {
 841                 tmpaddr = NULL;
 842         }
 843 retry:
 844         in6_dev_hold(idev);
 845         if (idev->cnf.use_tempaddr <= 0) {
 846                 write_unlock(&idev->lock);
 847                 printk(KERN_INFO
 848                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         spin_lock_bh(&ifp->lock);
 854         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 855                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 856                 spin_unlock_bh(&ifp->lock);
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_WARNING
 859                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         in6_ifa_hold(ifp);
 865         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 866         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 871                 in6_ifa_put(ifp);
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 877         age = (jiffies - ifp->tstamp) / HZ;
 878         tmp_valid_lft = min_t(__u32,
 879                               ifp->valid_lft,
 880                               idev->cnf.temp_valid_lft + age);
 881         tmp_prefered_lft = min_t(__u32,
 882                                  ifp->prefered_lft,
 883                                  idev->cnf.temp_prefered_lft + age -
 884                                  idev->cnf.max_desync_factor);
 885         tmp_plen = ifp->prefix_len;
 886         max_addresses = idev->cnf.max_addresses;
 887         tmp_cstamp = ifp->cstamp;
 888         tmp_tstamp = ifp->tstamp;
 889         spin_unlock_bh(&ifp->lock);
 890
 891         regen_advance = idev->cnf.regen_max_retry *
 892                         idev->cnf.dad_transmits *
 893                         idev->nd_parms->retrans_time / HZ;
 894         write_unlock(&idev->lock);
 895
 896         /* A temporary address is created only if this calculated Preferred
 897          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 898          * an implementation must not create a temporary address with a zero
 899          * Preferred Lifetime.
 900          */
 901         if (tmp_prefered_lft <= regen_advance) {
 902                 in6_ifa_put(ifp);
 903                 in6_dev_put(idev);
 904                 ret = -1;
 905                 goto out;
 906         }
 907
 908         addr_flags = IFA_F_TEMPORARY;
 909         /* set in addrconf_prefix_rcv() */
 910         if (ifp->flags & IFA_F_OPTIMISTIC)
 911                 addr_flags |= IFA_F_OPTIMISTIC;
 912
 913         ift = !max_addresses ||
 914               ipv6_count_addresses(idev) < max_addresses ?
 915                 ipv6_add_addr(idev, &addr, tmp_plen,
 916                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 917                               addr_flags) : NULL;
 918         if (!ift || IS_ERR(ift)) {
 919                 in6_ifa_put(ifp);
 920                 in6_dev_put(idev);
 921                 printk(KERN_INFO
 922                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 923                 tmpaddr = &addr;
 924                 write_lock(&idev->lock);
 925                 goto retry;
 926         }
 927
 928         spin_lock_bh(&ift->lock);
 929         ift->ifpub = ifp;
 930         ift->valid_lft = tmp_valid_lft;
 931         ift->prefered_lft = tmp_prefered_lft;
 932         ift->cstamp = tmp_cstamp;
 933         ift->tstamp = tmp_tstamp;
 934         spin_unlock_bh(&ift->lock);
 935
 936         addrconf_dad_start(ift, 0);
 937         in6_ifa_put(ift);
 938         in6_dev_put(idev);
 939 out:
 940         return ret;
 941 }
 942 #endif
 943
 944 /*
 945  *      Choose an appropriate source address (RFC3484)
 946  */
 947 enum {
 948         IPV6_SADDR_RULE_INIT = 0,
 949         IPV6_SADDR_RULE_LOCAL,
 950         IPV6_SADDR_RULE_SCOPE,
 951         IPV6_SADDR_RULE_PREFERRED,
 952 #ifdef CONFIG_IPV6_MIP6
 953         IPV6_SADDR_RULE_HOA,
 954 #endif
 955         IPV6_SADDR_RULE_OIF,
 956         IPV6_SADDR_RULE_LABEL,
 957 #ifdef CONFIG_IPV6_PRIVACY
 958         IPV6_SADDR_RULE_PRIVACY,
 959 #endif
 960         IPV6_SADDR_RULE_ORCHID,
 961         IPV6_SADDR_RULE_PREFIX,
 962         IPV6_SADDR_RULE_MAX
 963 };
 964
 965 struct ipv6_saddr_score {
 966         int                     rule;
 967         int                     addr_type;
 968         struct inet6_ifaddr     *ifa;
 969         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 970         int                     scopedist;
 971         int                     matchlen;
 972 };
 973
 974 struct ipv6_saddr_dst {
 975         const struct in6_addr *addr;
 976         int ifindex;
 977         int scope;
 978         int label;
 979         unsigned int prefs;
 980 };
 981
 982 static inline int ipv6_saddr_preferred(int type)
 983 {
 984         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 985                 return 1;
 986         return 0;
 987 }
 988
 989 static int ipv6_get_saddr_eval(struct net *net,
 990                                struct ipv6_saddr_score *score,
 991                                struct ipv6_saddr_dst *dst,
 992                                int i)
 993 {
 994         int ret;
 995
 996         if (i <= score->rule) {
 997                 switch (i) {
 998                 case IPV6_SADDR_RULE_SCOPE:
 999                         ret = score->scopedist;
1000                         break;
1001                 case IPV6_SADDR_RULE_PREFIX:
1002                         ret = score->matchlen;
1003                         break;
1004                 default:
1005                         ret = !!test_bit(i, score->scorebits);
1006                 }
1007                 goto out;
1008         }
1009
1010         switch (i) {
1011         case IPV6_SADDR_RULE_INIT:
1012                 /* Rule 0: remember if hiscore is not ready yet */
1013                 ret = !!score->ifa;
1014                 break;
1015         case IPV6_SADDR_RULE_LOCAL:
1016                 /* Rule 1: Prefer same address */
1017                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018                 break;
1019         case IPV6_SADDR_RULE_SCOPE:
1020                 /* Rule 2: Prefer appropriate scope
1021                  *
1022                  *      ret
1023                  *       ^
1024                  *    -1 |  d 15
1025                  *    ---+--+-+---> scope
1026                  *       |
1027                  *       |             d is scope of the destination.
1028                  *  B-d  |  \
1029                  *       |   \      <- smaller scope is better if
1030                  *  B-15 |    \        if scope is enough for destinaion.
1031                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032                  * d-C-1 | /
1033                  *       |/         <- greater is better
1034                  *   -C  /             if scope is not enough for destination.
1035                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036                  *
1037                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039                  * Assume B = 0 and we get C > 29.
1040                  */
1041                 ret = __ipv6_addr_src_scope(score->addr_type);
1042                 if (ret >= dst->scope)
1043                         ret = -ret;
1044                 else
1045                         ret -= 128;     /* 30 is enough */
1046                 score->scopedist = ret;
1047                 break;
1048         case IPV6_SADDR_RULE_PREFERRED:
1049                 /* Rule 3: Avoid deprecated and optimistic addresses */
1050                 ret = ipv6_saddr_preferred(score->addr_type) ||
1051                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052                 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054         case IPV6_SADDR_RULE_HOA:
1055             {
1056                 /* Rule 4: Prefer home address */
1057                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059                 break;
1060             }
1061 #endif
1062         case IPV6_SADDR_RULE_OIF:
1063                 /* Rule 5: Prefer outgoing interface */
1064                 ret = (!dst->ifindex ||
1065                        dst->ifindex == score->ifa->idev->dev->ifindex);
1066                 break;
1067         case IPV6_SADDR_RULE_LABEL:
1068                 /* Rule 6: Prefer matching label */
1069                 ret = ipv6_addr_label(net,
1070                                       &score->ifa->addr, score->addr_type,
1071                                       score->ifa->idev->dev->ifindex) == dst->label;
1072                 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074         case IPV6_SADDR_RULE_PRIVACY:
1075             {
1076                 /* Rule 7: Prefer public address
1077                  * Note: prefer temporary address if use_tempaddr >= 2
1078                  */
1079                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1082                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083                 break;
1084             }
1085 #endif
1086         case IPV6_SADDR_RULE_ORCHID:
1087                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088                  *          non-ORCHID vs non-ORCHID
1089                  */
1090                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091                         ipv6_addr_orchid(dst->addr));
1092                 break;
1093         case IPV6_SADDR_RULE_PREFIX:
1094                 /* Rule 8: Use longest matching prefix */
1095                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096                                                        dst->addr);
1097                 break;
1098         default:
1099                 ret = 0;
1100         }
1101
1102         if (ret)
1103                 __set_bit(i, score->scorebits);
1104         score->rule = i;
1105 out:
1106         return ret;
1107 }
1108
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110                        const struct in6_addr *daddr, unsigned int prefs,
1111                        struct in6_addr *saddr)
1112 {
1113         struct ipv6_saddr_score scores[2],
1114                                 *score = &scores[0], *hiscore = &scores[1];
1115         struct ipv6_saddr_dst dst;
1116         struct net_device *dev;
1117         int dst_type;
1118
1119         dst_type = __ipv6_addr_type(daddr);
1120         dst.addr = daddr;
1121         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122         dst.scope = __ipv6_addr_src_scope(dst_type);
1123         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124         dst.prefs = prefs;
1125
1126         hiscore->rule = -1;
1127         hiscore->ifa = NULL;
1128
1129         rcu_read_lock();
1130
1131         for_each_netdev_rcu(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         if (hiscore->ifa)
1209                                                 in6_ifa_put(hiscore->ifa);
1210
1211                                         in6_ifa_hold(score->ifa);
1212
1213                                         swap(hiscore, score);
1214
1215                                         /* restore our iterator */
1216                                         score->ifa = hiscore->ifa;
1217
1218                                         break;
1219                                 }
1220                         }
1221                 }
1222 try_nextdev:
1223                 read_unlock_bh(&idev->lock);
1224         }
1225         rcu_read_unlock();
1226
1227         if (!hiscore->ifa)
1228                 return -EADDRNOTAVAIL;
1229
1230         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231         in6_ifa_put(hiscore->ifa);
1232         return 0;
1233 }
1234 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237                     unsigned char banned_flags)
1238 {
1239         struct inet6_dev *idev;
1240         int err = -EADDRNOTAVAIL;
1241
1242         rcu_read_lock();
1243         idev = __in6_dev_get(dev);
1244         if (idev) {
1245                 struct inet6_ifaddr *ifp;
1246
1247                 read_lock_bh(&idev->lock);
1248                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                         if (ifp->scope == IFA_LINK &&
1250                             !(ifp->flags & banned_flags)) {
1251                                 ipv6_addr_copy(addr, &ifp->addr);
1252                                 err = 0;
1253                                 break;
1254                         }
1255                 }
1256                 read_unlock_bh(&idev->lock);
1257         }
1258         rcu_read_unlock();
1259         return err;
1260 }
1261
1262 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 {
1264         int cnt = 0;
1265         struct inet6_ifaddr *ifp;
1266
1267         read_lock_bh(&idev->lock);
1268         list_for_each_entry(ifp, &idev->addr_list, if_list)
1269                 cnt++;
1270         read_unlock_bh(&idev->lock);
1271         return cnt;
1272 }
1273
1274 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275                   struct net_device *dev, int strict)
1276 {
1277         struct inet6_ifaddr *ifp;
1278         struct hlist_node *node;
1279         unsigned int hash = ipv6_addr_hash(addr);
1280
1281         rcu_read_lock_bh();
1282         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283                 if (!net_eq(dev_net(ifp->idev->dev), net))
1284                         continue;
1285                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286                     !(ifp->flags&IFA_F_TENTATIVE) &&
1287                     (dev == NULL || ifp->idev->dev == dev ||
1288                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289                         rcu_read_unlock_bh();
1290                         return 1;
1291                 }
1292         }
1293
1294         rcu_read_unlock_bh();
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300                                struct net_device *dev)
1301 {
1302         unsigned int hash = ipv6_addr_hash(addr);
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305
1306         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 return true;
1312                 }
1313         }
1314         return false;
1315 }
1316
1317 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr *ifp, *result = NULL;
1346         unsigned int hash = ipv6_addr_hash(addr);
1347         struct hlist_node *node;
1348
1349         rcu_read_lock_bh();
1350         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 result = ifp;
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return result;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401 {
1402         int err = -ENOENT;
1403
1404         spin_lock(&ifp->state_lock);
1405         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407                 err = 0;
1408         }
1409         spin_unlock(&ifp->state_lock);
1410
1411         return err;
1412 }
1413
1414 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415 {
1416         struct inet6_dev *idev = ifp->idev;
1417
1418         if (addrconf_dad_end(ifp)) {
1419                 in6_ifa_put(ifp);
1420                 return;
1421         }
1422
1423         if (net_ratelimit())
1424                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425                         ifp->idev->dev->name, &ifp->addr);
1426
1427         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428                 struct in6_addr addr;
1429
1430                 addr.s6_addr32[0] = htonl(0xfe800000);
1431                 addr.s6_addr32[1] = 0;
1432
1433                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434                     ipv6_addr_equal(&ifp->addr, &addr)) {
1435                         /* DAD failed for link-local based on MAC address */
1436                         idev->cnf.disable_ipv6 = 1;
1437
1438                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439                                 ifp->idev->dev->name);
1440                 }
1441         }
1442
1443         addrconf_dad_stop(ifp, 1);
1444 }
1445
1446 /* Join to solicited addr multicast group. */
1447
1448 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449 {
1450         struct in6_addr maddr;
1451
1452         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453                 return;
1454
1455         addrconf_addr_solict_mult(addr, &maddr);
1456         ipv6_dev_mc_inc(dev, &maddr);
1457 }
1458
1459 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         __ipv6_dev_mc_dec(idev, &maddr);
1468 }
1469
1470 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471 {
1472         struct in6_addr addr;
1473         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1474         if (ipv6_addr_any(&addr))
1475                 return;
1476         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1477 }
1478
1479 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1480 {
1481         struct in6_addr addr;
1482         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483         if (ipv6_addr_any(&addr))
1484                 return;
1485         __ipv6_dev_ac_dec(ifp->idev, &addr);
1486 }
1487
1488 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1489 {
1490         if (dev->addr_len != ETH_ALEN)
1491                 return -1;
1492         memcpy(eui, dev->dev_addr, 3);
1493         memcpy(eui + 5, dev->dev_addr + 3, 3);
1494
1495         /*
1496          * The zSeries OSA network cards can be shared among various
1497          * OS instances, but the OSA cards have only one MAC address.
1498          * This leads to duplicate address conflicts in conjunction
1499          * with IPv6 if more than one instance uses the same card.
1500          *
1501          * The driver for these cards can deliver a unique 16-bit
1502          * identifier for each instance sharing the same card.  It is
1503          * placed instead of 0xFFFE in the interface identifier.  The
1504          * "u" bit of the interface identifier is not inverted in this
1505          * case.  Hence the resulting interface identifier has local
1506          * scope according to RFC2373.
1507          */
1508         if (dev->dev_id) {
1509                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1510                 eui[4] = dev->dev_id & 0xFF;
1511         } else {
1512                 eui[3] = 0xFF;
1513                 eui[4] = 0xFE;
1514                 eui[0] ^= 2;
1515         }
1516         return 0;
1517 }
1518
1519 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1520 {
1521         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1522         if (dev->addr_len != ARCNET_ALEN)
1523                 return -1;
1524         memset(eui, 0, 7);
1525         eui[7] = *(u8*)dev->dev_addr;
1526         return 0;
1527 }
1528
1529 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1530 {
1531         if (dev->addr_len != INFINIBAND_ALEN)
1532                 return -1;
1533         memcpy(eui, dev->dev_addr + 12, 8);
1534         eui[0] |= 2;
1535         return 0;
1536 }
1537
1538 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1539 {
1540         if (addr == 0)
1541                 return -1;
1542         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1543                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1544                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1545                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1546                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1547                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1548         eui[1] = 0;
1549         eui[2] = 0x5E;
1550         eui[3] = 0xFE;
1551         memcpy(eui + 4, &addr, 4);
1552         return 0;
1553 }
1554
1555 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1556 {
1557         if (dev->priv_flags & IFF_ISATAP)
1558                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1559         return -1;
1560 }
1561
1562 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1563 {
1564         switch (dev->type) {
1565         case ARPHRD_ETHER:
1566         case ARPHRD_FDDI:
1567         case ARPHRD_IEEE802_TR:
1568                 return addrconf_ifid_eui48(eui, dev);
1569         case ARPHRD_ARCNET:
1570                 return addrconf_ifid_arcnet(eui, dev);
1571         case ARPHRD_INFINIBAND:
1572                 return addrconf_ifid_infiniband(eui, dev);
1573         case ARPHRD_SIT:
1574                 return addrconf_ifid_sit(eui, dev);
1575         }
1576         return -1;
1577 }
1578
1579 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1580 {
1581         int err = -1;
1582         struct inet6_ifaddr *ifp;
1583
1584         read_lock_bh(&idev->lock);
1585         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1586                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1587                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1588                         err = 0;
1589                         break;
1590                 }
1591         }
1592         read_unlock_bh(&idev->lock);
1593         return err;
1594 }
1595
1596 #ifdef CONFIG_IPV6_PRIVACY
1597 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1598 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1599 {
1600 regen:
1601         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1602         idev->rndid[0] &= ~0x02;
1603
1604         /*
1605          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1606          * check if generated address is not inappropriate
1607          *
1608          *  - Reserved subnet anycast (RFC 2526)
1609          *      11111101 11....11 1xxxxxxx
1610          *  - ISATAP (RFC4214) 6.1
1611          *      00-00-5E-FE-xx-xx-xx-xx
1612          *  - value 0
1613          *  - XXX: already assigned to an address on the device
1614          */
1615         if (idev->rndid[0] == 0xfd &&
1616             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1617             (idev->rndid[7]&0x80))
1618                 goto regen;
1619         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1620                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1621                         goto regen;
1622                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1623                         goto regen;
1624         }
1625
1626         return 0;
1627 }
1628
1629 static void ipv6_regen_rndid(unsigned long data)
1630 {
1631         struct inet6_dev *idev = (struct inet6_dev *) data;
1632         unsigned long expires;
1633
1634         rcu_read_lock_bh();
1635         write_lock_bh(&idev->lock);
1636
1637         if (idev->dead)
1638                 goto out;
1639
1640         if (__ipv6_regen_rndid(idev) < 0)
1641                 goto out;
1642
1643         expires = jiffies +
1644                 idev->cnf.temp_prefered_lft * HZ -
1645                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1646                 idev->cnf.max_desync_factor * HZ;
1647         if (time_before(expires, jiffies)) {
1648                 printk(KERN_WARNING
1649                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1650                         idev->dev->name);
1651                 goto out;
1652         }
1653
1654         if (!mod_timer(&idev->regen_timer, expires))
1655                 in6_dev_hold(idev);
1656
1657 out:
1658         write_unlock_bh(&idev->lock);
1659         rcu_read_unlock_bh();
1660         in6_dev_put(idev);
1661 }
1662
1663 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1664         int ret = 0;
1665
1666         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1667                 ret = __ipv6_regen_rndid(idev);
1668         return ret;
1669 }
1670 #endif
1671
1672 /*
1673  *      Add prefix route.
1674  */
1675
1676 static void
1677 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1678                       unsigned long expires, u32 flags)
1679 {
1680         struct fib6_config cfg = {
1681                 .fc_table = RT6_TABLE_PREFIX,
1682                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1683                 .fc_ifindex = dev->ifindex,
1684                 .fc_expires = expires,
1685                 .fc_dst_len = plen,
1686                 .fc_flags = RTF_UP | flags,
1687                 .fc_nlinfo.nl_net = dev_net(dev),
1688                 .fc_protocol = RTPROT_KERNEL,
1689         };
1690
1691         ipv6_addr_copy(&cfg.fc_dst, pfx);
1692
1693         /* Prevent useless cloning on PtP SIT.
1694            This thing is done here expecting that the whole
1695            class of non-broadcast devices need not cloning.
1696          */
1697 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1698         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1699                 cfg.fc_flags |= RTF_NONEXTHOP;
1700 #endif
1701
1702         ip6_route_add(&cfg);
1703 }
1704
1705 /* Create "default" multicast route to the interface */
1706
1707 static void addrconf_add_mroute(struct net_device *dev)
1708 {
1709         struct fib6_config cfg = {
1710                 .fc_table = RT6_TABLE_LOCAL,
1711                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1712                 .fc_ifindex = dev->ifindex,
1713                 .fc_dst_len = 8,
1714                 .fc_flags = RTF_UP,
1715                 .fc_nlinfo.nl_net = dev_net(dev),
1716         };
1717
1718         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1719
1720         ip6_route_add(&cfg);
1721 }
1722
1723 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1724 static void sit_route_add(struct net_device *dev)
1725 {
1726         struct fib6_config cfg = {
1727                 .fc_table = RT6_TABLE_MAIN,
1728                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1729                 .fc_ifindex = dev->ifindex,
1730                 .fc_dst_len = 96,
1731                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1732                 .fc_nlinfo.nl_net = dev_net(dev),
1733         };
1734
1735         /* prefix length - 96 bits "::d.d.d.d" */
1736         ip6_route_add(&cfg);
1737 }
1738 #endif
1739
1740 static void addrconf_add_lroute(struct net_device *dev)
1741 {
1742         struct in6_addr addr;
1743
1744         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1745         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1746 }
1747
1748 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1749 {
1750         struct inet6_dev *idev;
1751
1752         ASSERT_RTNL();
1753
1754         idev = ipv6_find_idev(dev);
1755         if (!idev)
1756                 return ERR_PTR(-ENOBUFS);
1757
1758         if (idev->cnf.disable_ipv6)
1759                 return ERR_PTR(-EACCES);
1760
1761         /* Add default multicast route */
1762         addrconf_add_mroute(dev);
1763
1764         /* Add link local route */
1765         addrconf_add_lroute(dev);
1766         return idev;
1767 }
1768
1769 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1770 {
1771         struct prefix_info *pinfo;
1772         __u32 valid_lft;
1773         __u32 prefered_lft;
1774         int addr_type;
1775         struct inet6_dev *in6_dev;
1776         struct net *net = dev_net(dev);
1777
1778         pinfo = (struct prefix_info *) opt;
1779
1780         if (len < sizeof(struct prefix_info)) {
1781                 ADBG(("addrconf: prefix option too short\n"));
1782                 return;
1783         }
1784
1785         /*
1786          *      Validation checks ([ADDRCONF], page 19)
1787          */
1788
1789         addr_type = ipv6_addr_type(&pinfo->prefix);
1790
1791         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1792                 return;
1793
1794         valid_lft = ntohl(pinfo->valid);
1795         prefered_lft = ntohl(pinfo->prefered);
1796
1797         if (prefered_lft > valid_lft) {
1798                 if (net_ratelimit())
1799                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1800                 return;
1801         }
1802
1803         in6_dev = in6_dev_get(dev);
1804
1805         if (in6_dev == NULL) {
1806                 if (net_ratelimit())
1807                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1808                 return;
1809         }
1810
1811         /*
1812          *      Two things going on here:
1813          *      1) Add routes for on-link prefixes
1814          *      2) Configure prefixes with the auto flag set
1815          */
1816
1817         if (pinfo->onlink) {
1818                 struct rt6_info *rt;
1819                 unsigned long rt_expires;
1820
1821                 /* Avoid arithmetic overflow. Really, we could
1822                  * save rt_expires in seconds, likely valid_lft,
1823                  * but it would require division in fib gc, that it
1824                  * not good.
1825                  */
1826                 if (HZ > USER_HZ)
1827                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1828                 else
1829                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1830
1831                 if (addrconf_finite_timeout(rt_expires))
1832                         rt_expires *= HZ;
1833
1834                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1835                                 dev->ifindex, 1);
1836
1837                 if (rt && addrconf_is_prefix_route(rt)) {
1838                         /* Autoconf prefix route */
1839                         if (valid_lft == 0) {
1840                                 ip6_del_rt(rt);
1841                                 rt = NULL;
1842                         } else if (addrconf_finite_timeout(rt_expires)) {
1843                                 /* not infinity */
1844                                 rt->rt6i_expires = jiffies + rt_expires;
1845                                 rt->rt6i_flags |= RTF_EXPIRES;
1846                         } else {
1847                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1848                                 rt->rt6i_expires = 0;
1849                         }
1850                 } else if (valid_lft) {
1851                         clock_t expires = 0;
1852                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1853                         if (addrconf_finite_timeout(rt_expires)) {
1854                                 /* not infinity */
1855                                 flags |= RTF_EXPIRES;
1856                                 expires = jiffies_to_clock_t(rt_expires);
1857                         }
1858                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1859                                               dev, expires, flags);
1860                 }
1861                 if (rt)
1862                         dst_release(&rt->dst);
1863         }
1864
1865         /* Try to figure out our local address for this prefix */
1866
1867         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1868                 struct inet6_ifaddr * ifp;
1869                 struct in6_addr addr;
1870                 int create = 0, update_lft = 0;
1871
1872                 if (pinfo->prefix_len == 64) {
1873                         memcpy(&addr, &pinfo->prefix, 8);
1874                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1875                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1876                                 in6_dev_put(in6_dev);
1877                                 return;
1878                         }
1879                         goto ok;
1880                 }
1881                 if (net_ratelimit())
1882                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1883                                pinfo->prefix_len);
1884                 in6_dev_put(in6_dev);
1885                 return;
1886
1887 ok:
1888
1889                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1890
1891                 if (ifp == NULL && valid_lft) {
1892                         int max_addresses = in6_dev->cnf.max_addresses;
1893                         u32 addr_flags = 0;
1894
1895 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1896                         if (in6_dev->cnf.optimistic_dad &&
1897                             !net->ipv6.devconf_all->forwarding)
1898                                 addr_flags = IFA_F_OPTIMISTIC;
1899 #endif
1900
1901                         /* Do not allow to create too much of autoconfigured
1902                          * addresses; this would be too easy way to crash kernel.
1903                          */
1904                         if (!max_addresses ||
1905                             ipv6_count_addresses(in6_dev) < max_addresses)
1906                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1907                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1908                                                     addr_flags);
1909
1910                         if (!ifp || IS_ERR(ifp)) {
1911                                 in6_dev_put(in6_dev);
1912                                 return;
1913                         }
1914
1915                         update_lft = create = 1;
1916                         ifp->cstamp = jiffies;
1917                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1918                 }
1919
1920                 if (ifp) {
1921                         int flags;
1922                         unsigned long now;
1923 #ifdef CONFIG_IPV6_PRIVACY
1924                         struct inet6_ifaddr *ift;
1925 #endif
1926                         u32 stored_lft;
1927
1928                         /* update lifetime (RFC2462 5.5.3 e) */
1929                         spin_lock(&ifp->lock);
1930                         now = jiffies;
1931                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1932                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1933                         else
1934                                 stored_lft = 0;
1935                         if (!update_lft && stored_lft) {
1936                                 if (valid_lft > MIN_VALID_LIFETIME ||
1937                                     valid_lft > stored_lft)
1938                                         update_lft = 1;
1939                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1940                                         /* valid_lft <= stored_lft is always true */
1941                                         /*
1942                                          * RFC 4862 Section 5.5.3e:
1943                                          * "Note that the preferred lifetime of
1944                                          *  the corresponding address is always
1945                                          *  reset to the Preferred Lifetime in
1946                                          *  the received Prefix Information
1947                                          *  option, regardless of whether the
1948                                          *  valid lifetime is also reset or
1949                                          *  ignored."
1950                                          *
1951                                          *  So if the preferred lifetime in
1952                                          *  this advertisement is different
1953                                          *  than what we have stored, but the
1954                                          *  valid lifetime is invalid, just
1955                                          *  reset prefered_lft.
1956                                          *
1957                                          *  We must set the valid lifetime
1958                                          *  to the stored lifetime since we'll
1959                                          *  be updating the timestamp below,
1960                                          *  else we'll set it back to the
1961                                          *  minimum.
1962                                          */
1963                                         if (prefered_lft != ifp->prefered_lft) {
1964                                                 valid_lft = stored_lft;
1965                                                 update_lft = 1;
1966                                         }
1967                                 } else {
1968                                         valid_lft = MIN_VALID_LIFETIME;
1969                                         if (valid_lft < prefered_lft)
1970                                                 prefered_lft = valid_lft;
1971                                         update_lft = 1;
1972                                 }
1973                         }
1974
1975                         if (update_lft) {
1976                                 ifp->valid_lft = valid_lft;
1977                                 ifp->prefered_lft = prefered_lft;
1978                                 ifp->tstamp = now;
1979                                 flags = ifp->flags;
1980                                 ifp->flags &= ~IFA_F_DEPRECATED;
1981                                 spin_unlock(&ifp->lock);
1982
1983                                 if (!(flags&IFA_F_TENTATIVE))
1984                                         ipv6_ifa_notify(0, ifp);
1985                         } else
1986                                 spin_unlock(&ifp->lock);
1987
1988 #ifdef CONFIG_IPV6_PRIVACY
1989                         read_lock_bh(&in6_dev->lock);
1990                         /* update all temporary addresses in the list */
1991                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1992                                 /*
1993                                  * When adjusting the lifetimes of an existing
1994                                  * temporary address, only lower the lifetimes.
1995                                  * Implementations must not increase the
1996                                  * lifetimes of an existing temporary address
1997                                  * when processing a Prefix Information Option.
1998                                  */
1999                                 if (ifp != ift->ifpub)
2000                                         continue;
2001
2002                                 spin_lock(&ift->lock);
2003                                 flags = ift->flags;
2004                                 if (ift->valid_lft > valid_lft &&
2005                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2006                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2007                                 if (ift->prefered_lft > prefered_lft &&
2008                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2009                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2010                                 spin_unlock(&ift->lock);
2011                                 if (!(flags&IFA_F_TENTATIVE))
2012                                         ipv6_ifa_notify(0, ift);
2013                         }
2014
2015                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2016                                 /*
2017                                  * When a new public address is created as described in [ADDRCONF],
2018                                  * also create a new temporary address. Also create a temporary
2019                                  * address if it's enabled but no temporary address currently exists.
2020                                  */
2021                                 read_unlock_bh(&in6_dev->lock);
2022                                 ipv6_create_tempaddr(ifp, NULL);
2023                         } else {
2024                                 read_unlock_bh(&in6_dev->lock);
2025                         }
2026 #endif
2027                         in6_ifa_put(ifp);
2028                         addrconf_verify(0);
2029                 }
2030         }
2031         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2032         in6_dev_put(in6_dev);
2033 }
2034
2035 /*
2036  *      Set destination address.
2037  *      Special case for SIT interfaces where we create a new "virtual"
2038  *      device.
2039  */
2040 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2041 {
2042         struct in6_ifreq ireq;
2043         struct net_device *dev;
2044         int err = -EINVAL;
2045
2046         rtnl_lock();
2047
2048         err = -EFAULT;
2049         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2050                 goto err_exit;
2051
2052         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2053
2054         err = -ENODEV;
2055         if (dev == NULL)
2056                 goto err_exit;
2057
2058 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2059         if (dev->type == ARPHRD_SIT) {
2060                 const struct net_device_ops *ops = dev->netdev_ops;
2061                 struct ifreq ifr;
2062                 struct ip_tunnel_parm p;
2063
2064                 err = -EADDRNOTAVAIL;
2065                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2066                         goto err_exit;
2067
2068                 memset(&p, 0, sizeof(p));
2069                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2070                 p.iph.saddr = 0;
2071                 p.iph.version = 4;
2072                 p.iph.ihl = 5;
2073                 p.iph.protocol = IPPROTO_IPV6;
2074                 p.iph.ttl = 64;
2075                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2076
2077                 if (ops->ndo_do_ioctl) {
2078                         mm_segment_t oldfs = get_fs();
2079
2080                         set_fs(KERNEL_DS);
2081                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2082                         set_fs(oldfs);
2083                 } else
2084                         err = -EOPNOTSUPP;
2085
2086                 if (err == 0) {
2087                         err = -ENOBUFS;
2088                         dev = __dev_get_by_name(net, p.name);
2089                         if (!dev)
2090                                 goto err_exit;
2091                         err = dev_open(dev);
2092                 }
2093         }
2094 #endif
2095
2096 err_exit:
2097         rtnl_unlock();
2098         return err;
2099 }
2100
2101 /*
2102  *      Manual configuration of address on an interface
2103  */
2104 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2105                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2106                           __u32 valid_lft)
2107 {
2108         struct inet6_ifaddr *ifp;
2109         struct inet6_dev *idev;
2110         struct net_device *dev;
2111         int scope;
2112         u32 flags;
2113         clock_t expires;
2114         unsigned long timeout;
2115
2116         ASSERT_RTNL();
2117
2118         if (plen > 128)
2119                 return -EINVAL;
2120
2121         /* check the lifetime */
2122         if (!valid_lft || prefered_lft > valid_lft)
2123                 return -EINVAL;
2124
2125         dev = __dev_get_by_index(net, ifindex);
2126         if (!dev)
2127                 return -ENODEV;
2128
2129         idev = addrconf_add_dev(dev);
2130         if (IS_ERR(idev))
2131                 return PTR_ERR(idev);
2132
2133         scope = ipv6_addr_scope(pfx);
2134
2135         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2136         if (addrconf_finite_timeout(timeout)) {
2137                 expires = jiffies_to_clock_t(timeout * HZ);
2138                 valid_lft = timeout;
2139                 flags = RTF_EXPIRES;
2140         } else {
2141                 expires = 0;
2142                 flags = 0;
2143                 ifa_flags |= IFA_F_PERMANENT;
2144         }
2145
2146         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2147         if (addrconf_finite_timeout(timeout)) {
2148                 if (timeout == 0)
2149                         ifa_flags |= IFA_F_DEPRECATED;
2150                 prefered_lft = timeout;
2151         }
2152
2153         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2154
2155         if (!IS_ERR(ifp)) {
2156                 spin_lock_bh(&ifp->lock);
2157                 ifp->valid_lft = valid_lft;
2158                 ifp->prefered_lft = prefered_lft;
2159                 ifp->tstamp = jiffies;
2160                 spin_unlock_bh(&ifp->lock);
2161
2162                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2163                                       expires, flags);
2164                 /*
2165                  * Note that section 3.1 of RFC 4429 indicates
2166                  * that the Optimistic flag should not be set for
2167                  * manually configured addresses
2168                  */
2169                 addrconf_dad_start(ifp, 0);
2170                 in6_ifa_put(ifp);
2171                 addrconf_verify(0);
2172                 return 0;
2173         }
2174
2175         return PTR_ERR(ifp);
2176 }
2177
2178 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2179                           unsigned int plen)
2180 {
2181         struct inet6_ifaddr *ifp;
2182         struct inet6_dev *idev;
2183         struct net_device *dev;
2184
2185         if (plen > 128)
2186                 return -EINVAL;
2187
2188         dev = __dev_get_by_index(net, ifindex);
2189         if (!dev)
2190                 return -ENODEV;
2191
2192         if ((idev = __in6_dev_get(dev)) == NULL)
2193                 return -ENXIO;
2194
2195         read_lock_bh(&idev->lock);
2196         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2197                 if (ifp->prefix_len == plen &&
2198                     ipv6_addr_equal(pfx, &ifp->addr)) {
2199                         in6_ifa_hold(ifp);
2200                         read_unlock_bh(&idev->lock);
2201
2202                         ipv6_del_addr(ifp);
2203
2204                         /* If the last address is deleted administratively,
2205                            disable IPv6 on this interface.
2206                          */
2207                         if (list_empty(&idev->addr_list))
2208                                 addrconf_ifdown(idev->dev, 1);
2209                         return 0;
2210                 }
2211         }
2212         read_unlock_bh(&idev->lock);
2213         return -EADDRNOTAVAIL;
2214 }
2215
2216
2217 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2218 {
2219         struct in6_ifreq ireq;
2220         int err;
2221
2222         if (!capable(CAP_NET_ADMIN))
2223                 return -EPERM;
2224
2225         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2226                 return -EFAULT;
2227
2228         rtnl_lock();
2229         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2230                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2231                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2232         rtnl_unlock();
2233         return err;
2234 }
2235
2236 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2237 {
2238         struct in6_ifreq ireq;
2239         int err;
2240
2241         if (!capable(CAP_NET_ADMIN))
2242                 return -EPERM;
2243
2244         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2245                 return -EFAULT;
2246
2247         rtnl_lock();
2248         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2249                              ireq.ifr6_prefixlen);
2250         rtnl_unlock();
2251         return err;
2252 }
2253
2254 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2255                      int plen, int scope)
2256 {
2257         struct inet6_ifaddr *ifp;
2258
2259         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2260         if (!IS_ERR(ifp)) {
2261                 spin_lock_bh(&ifp->lock);
2262                 ifp->flags &= ~IFA_F_TENTATIVE;
2263                 spin_unlock_bh(&ifp->lock);
2264                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2265                 in6_ifa_put(ifp);
2266         }
2267 }
2268
2269 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2270 static void sit_add_v4_addrs(struct inet6_dev *idev)
2271 {
2272         struct in6_addr addr;
2273         struct net_device *dev;
2274         struct net *net = dev_net(idev->dev);
2275         int scope;
2276
2277         ASSERT_RTNL();
2278
2279         memset(&addr, 0, sizeof(struct in6_addr));
2280         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2281
2282         if (idev->dev->flags&IFF_POINTOPOINT) {
2283                 addr.s6_addr32[0] = htonl(0xfe800000);
2284                 scope = IFA_LINK;
2285         } else {
2286                 scope = IPV6_ADDR_COMPATv4;
2287         }
2288
2289         if (addr.s6_addr32[3]) {
2290                 add_addr(idev, &addr, 128, scope);
2291                 return;
2292         }
2293
2294         for_each_netdev(net, dev) {
2295                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2296                 if (in_dev && (dev->flags & IFF_UP)) {
2297                         struct in_ifaddr * ifa;
2298
2299                         int flag = scope;
2300
2301                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2302                                 int plen;
2303
2304                                 addr.s6_addr32[3] = ifa->ifa_local;
2305
2306                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2307                                         continue;
2308                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2309                                         if (idev->dev->flags&IFF_POINTOPOINT)
2310                                                 continue;
2311                                         flag |= IFA_HOST;
2312                                 }
2313                                 if (idev->dev->flags&IFF_POINTOPOINT)
2314                                         plen = 64;
2315                                 else
2316                                         plen = 96;
2317
2318                                 add_addr(idev, &addr, plen, flag);
2319                         }
2320                 }
2321         }
2322 }
2323 #endif
2324
2325 static void init_loopback(struct net_device *dev)
2326 {
2327         struct inet6_dev  *idev;
2328
2329         /* ::1 */
2330
2331         ASSERT_RTNL();
2332
2333         if ((idev = ipv6_find_idev(dev)) == NULL) {
2334                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2335                 return;
2336         }
2337
2338         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2339 }
2340
2341 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2342 {
2343         struct inet6_ifaddr * ifp;
2344         u32 addr_flags = IFA_F_PERMANENT;
2345
2346 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2347         if (idev->cnf.optimistic_dad &&
2348             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2349                 addr_flags |= IFA_F_OPTIMISTIC;
2350 #endif
2351
2352
2353         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2354         if (!IS_ERR(ifp)) {
2355                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2356                 addrconf_dad_start(ifp, 0);
2357                 in6_ifa_put(ifp);
2358         }
2359 }
2360
2361 static void addrconf_dev_config(struct net_device *dev)
2362 {
2363         struct in6_addr addr;
2364         struct inet6_dev    * idev;
2365
2366         ASSERT_RTNL();
2367
2368         if ((dev->type != ARPHRD_ETHER) &&
2369             (dev->type != ARPHRD_FDDI) &&
2370             (dev->type != ARPHRD_IEEE802_TR) &&
2371             (dev->type != ARPHRD_ARCNET) &&
2372             (dev->type != ARPHRD_INFINIBAND)) {
2373                 /* Alas, we support only Ethernet autoconfiguration. */
2374                 return;
2375         }
2376
2377         idev = addrconf_add_dev(dev);
2378         if (IS_ERR(idev))
2379                 return;
2380
2381         memset(&addr, 0, sizeof(struct in6_addr));
2382         addr.s6_addr32[0] = htonl(0xFE800000);
2383
2384         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2385                 addrconf_add_linklocal(idev, &addr);
2386 }
2387
2388 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2389 static void addrconf_sit_config(struct net_device *dev)
2390 {
2391         struct inet6_dev *idev;
2392
2393         ASSERT_RTNL();
2394
2395         /*
2396          * Configure the tunnel with one of our IPv4
2397          * addresses... we should configure all of
2398          * our v4 addrs in the tunnel
2399          */
2400
2401         if ((idev = ipv6_find_idev(dev)) == NULL) {
2402                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2403                 return;
2404         }
2405
2406         if (dev->priv_flags & IFF_ISATAP) {
2407                 struct in6_addr addr;
2408
2409                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2410                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2411                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2412                         addrconf_add_linklocal(idev, &addr);
2413                 return;
2414         }
2415
2416         sit_add_v4_addrs(idev);
2417
2418         if (dev->flags&IFF_POINTOPOINT) {
2419                 addrconf_add_mroute(dev);
2420                 addrconf_add_lroute(dev);
2421         } else
2422                 sit_route_add(dev);
2423 }
2424 #endif
2425
2426 static inline int
2427 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2428 {
2429         struct in6_addr lladdr;
2430
2431         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2432                 addrconf_add_linklocal(idev, &lladdr);
2433                 return 0;
2434         }
2435         return -1;
2436 }
2437
2438 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2439 {
2440         struct net_device *link_dev;
2441         struct net *net = dev_net(idev->dev);
2442
2443         /* first try to inherit the link-local address from the link device */
2444         if (idev->dev->iflink &&
2445             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2446                 if (!ipv6_inherit_linklocal(idev, link_dev))
2447                         return;
2448         }
2449         /* then try to inherit it from any device */
2450         for_each_netdev(net, link_dev) {
2451                 if (!ipv6_inherit_linklocal(idev, link_dev))
2452                         return;
2453         }
2454         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2455 }
2456
2457 /*
2458  * Autoconfigure tunnel with a link-local address so routing protocols,
2459  * DHCPv6, MLD etc. can be run over the virtual link
2460  */
2461
2462 static void addrconf_ip6_tnl_config(struct net_device *dev)
2463 {
2464         struct inet6_dev *idev;
2465
2466         ASSERT_RTNL();
2467
2468         idev = addrconf_add_dev(dev);
2469         if (IS_ERR(idev)) {
2470                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2471                 return;
2472         }
2473         ip6_tnl_add_linklocal(idev);
2474 }
2475
2476 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2477                            void * data)
2478 {
2479         struct net_device *dev = (struct net_device *) data;
2480         struct inet6_dev *idev = __in6_dev_get(dev);
2481         int run_pending = 0;
2482         int err;
2483
2484         switch (event) {
2485         case NETDEV_REGISTER:
2486                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2487                         idev = ipv6_add_dev(dev);
2488                         if (!idev)
2489                                 return notifier_from_errno(-ENOMEM);
2490                 }
2491                 break;
2492
2493         case NETDEV_UP:
2494         case NETDEV_CHANGE:
2495                 if (dev->flags & IFF_SLAVE)
2496                         break;
2497
2498                 if (event == NETDEV_UP) {
2499                         if (!addrconf_qdisc_ok(dev)) {
2500                                 /* device is not ready yet. */
2501                                 printk(KERN_INFO
2502                                         "ADDRCONF(NETDEV_UP): %s: "
2503                                         "link is not ready\n",
2504                                         dev->name);
2505                                 break;
2506                         }
2507
2508                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2509                                 idev = ipv6_add_dev(dev);
2510
2511                         if (idev) {
2512                                 idev->if_flags |= IF_READY;
2513                                 run_pending = 1;
2514                         }
2515                 } else {
2516                         if (!addrconf_qdisc_ok(dev)) {
2517                                 /* device is still not ready. */
2518                                 break;
2519                         }
2520
2521                         if (idev) {
2522                                 if (idev->if_flags & IF_READY)
2523                                         /* device is already configured. */
2524                                         break;
2525                                 idev->if_flags |= IF_READY;
2526                         }
2527
2528                         printk(KERN_INFO
2529                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2530                                         "link becomes ready\n",
2531                                         dev->name);
2532
2533                         run_pending = 1;
2534                 }
2535
2536                 switch (dev->type) {
2537 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2538                 case ARPHRD_SIT:
2539                         addrconf_sit_config(dev);
2540                         break;
2541 #endif
2542                 case ARPHRD_TUNNEL6:
2543                         addrconf_ip6_tnl_config(dev);
2544                         break;
2545                 case ARPHRD_LOOPBACK:
2546                         init_loopback(dev);
2547                         break;
2548
2549                 default:
2550                         addrconf_dev_config(dev);
2551                         break;
2552                 }
2553
2554                 if (idev) {
2555                         if (run_pending)
2556                                 addrconf_dad_run(idev);
2557
2558                         /*
2559                          * If the MTU changed during the interface down,
2560                          * when the interface up, the changed MTU must be
2561                          * reflected in the idev as well as routers.
2562                          */
2563                         if (idev->cnf.mtu6 != dev->mtu &&
2564                             dev->mtu >= IPV6_MIN_MTU) {
2565                                 rt6_mtu_change(dev, dev->mtu);
2566                                 idev->cnf.mtu6 = dev->mtu;
2567                         }
2568                         idev->tstamp = jiffies;
2569                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2570
2571                         /*
2572                          * If the changed mtu during down is lower than
2573                          * IPV6_MIN_MTU stop IPv6 on this interface.
2574                          */
2575                         if (dev->mtu < IPV6_MIN_MTU)
2576                                 addrconf_ifdown(dev, 1);
2577                 }
2578                 break;
2579
2580         case NETDEV_CHANGEMTU:
2581                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2582                         rt6_mtu_change(dev, dev->mtu);
2583                         idev->cnf.mtu6 = dev->mtu;
2584                         break;
2585                 }
2586
2587                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2588                         idev = ipv6_add_dev(dev);
2589                         if (idev)
2590                                 break;
2591                 }
2592
2593                 /*
2594                  * MTU falled under IPV6_MIN_MTU.
2595                  * Stop IPv6 on this interface.
2596                  */
2597
2598         case NETDEV_DOWN:
2599         case NETDEV_UNREGISTER:
2600                 /*
2601                  *      Remove all addresses from this interface.
2602                  */
2603                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2604                 break;
2605
2606         case NETDEV_CHANGENAME:
2607                 if (idev) {
2608                         snmp6_unregister_dev(idev);
2609                         addrconf_sysctl_unregister(idev);
2610                         addrconf_sysctl_register(idev);
2611                         err = snmp6_register_dev(idev);
2612                         if (err)
2613                                 return notifier_from_errno(err);
2614                 }
2615                 break;
2616
2617         case NETDEV_PRE_TYPE_CHANGE:
2618         case NETDEV_POST_TYPE_CHANGE:
2619                 addrconf_type_change(dev, event);
2620                 break;
2621         }
2622
2623         return NOTIFY_OK;
2624 }
2625
2626 /*
2627  *      addrconf module should be notified of a device going up
2628  */
2629 static struct notifier_block ipv6_dev_notf = {
2630         .notifier_call = addrconf_notify,
2631 };
2632
2633 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2634 {
2635         struct inet6_dev *idev;
2636         ASSERT_RTNL();
2637
2638         idev = __in6_dev_get(dev);
2639
2640         if (event == NETDEV_POST_TYPE_CHANGE)
2641                 ipv6_mc_remap(idev);
2642         else if (event == NETDEV_PRE_TYPE_CHANGE)
2643                 ipv6_mc_unmap(idev);
2644 }
2645
2646 static int addrconf_ifdown(struct net_device *dev, int how)
2647 {
2648         struct net *net = dev_net(dev);
2649         struct inet6_dev *idev;
2650         struct inet6_ifaddr *ifa;
2651         int state, i;
2652
2653         ASSERT_RTNL();
2654
2655         rt6_ifdown(net, dev);
2656         neigh_ifdown(&nd_tbl, dev);
2657
2658         idev = __in6_dev_get(dev);
2659         if (idev == NULL)
2660                 return -ENODEV;
2661
2662         /*
2663          * Step 1: remove reference to ipv6 device from parent device.
2664          *         Do not dev_put!
2665          */
2666         if (how) {
2667                 idev->dead = 1;
2668
2669                 /* protected by rtnl_lock */
2670                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2671
2672                 /* Step 1.5: remove snmp6 entry */
2673                 snmp6_unregister_dev(idev);
2674
2675         }
2676
2677         /* Step 2: clear hash table */
2678         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2679                 struct hlist_head *h = &inet6_addr_lst[i];
2680                 struct hlist_node *n;
2681
2682                 spin_lock_bh(&addrconf_hash_lock);
2683         restart:
2684                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2685                         if (ifa->idev == idev) {
2686                                 hlist_del_init_rcu(&ifa->addr_lst);
2687                                 addrconf_del_timer(ifa);
2688                                 goto restart;
2689                         }
2690                 }
2691                 spin_unlock_bh(&addrconf_hash_lock);
2692         }
2693
2694         write_lock_bh(&idev->lock);
2695
2696         /* Step 2: clear flags for stateless addrconf */
2697         if (!how)
2698                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2699
2700 #ifdef CONFIG_IPV6_PRIVACY
2701         if (how && del_timer(&idev->regen_timer))
2702                 in6_dev_put(idev);
2703
2704         /* Step 3: clear tempaddr list */
2705         while (!list_empty(&idev->tempaddr_list)) {
2706                 ifa = list_first_entry(&idev->tempaddr_list,
2707                                        struct inet6_ifaddr, tmp_list);
2708                 list_del(&ifa->tmp_list);
2709                 write_unlock_bh(&idev->lock);
2710                 spin_lock_bh(&ifa->lock);
2711
2712                 if (ifa->ifpub) {
2713                         in6_ifa_put(ifa->ifpub);
2714                         ifa->ifpub = NULL;
2715                 }
2716                 spin_unlock_bh(&ifa->lock);
2717                 in6_ifa_put(ifa);
2718                 write_lock_bh(&idev->lock);
2719         }
2720 #endif
2721
2722         while (!list_empty(&idev->addr_list)) {
2723                 ifa = list_first_entry(&idev->addr_list,
2724                                        struct inet6_ifaddr, if_list);
2725                 addrconf_del_timer(ifa);
2726
2727                 list_del(&ifa->if_list);
2728
2729                 write_unlock_bh(&idev->lock);
2730
2731                 spin_lock_bh(&ifa->state_lock);
2732                 state = ifa->state;
2733                 ifa->state = INET6_IFADDR_STATE_DEAD;
2734                 spin_unlock_bh(&ifa->state_lock);
2735
2736                 if (state != INET6_IFADDR_STATE_DEAD) {
2737                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2738                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2739                 }
2740                 in6_ifa_put(ifa);
2741
2742                 write_lock_bh(&idev->lock);
2743         }
2744
2745         write_unlock_bh(&idev->lock);
2746
2747         /* Step 5: Discard multicast list */
2748         if (how)
2749                 ipv6_mc_destroy_dev(idev);
2750         else
2751                 ipv6_mc_down(idev);
2752
2753         idev->tstamp = jiffies;
2754
2755         /* Last: Shot the device (if unregistered) */
2756         if (how) {
2757                 addrconf_sysctl_unregister(idev);
2758                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2759                 neigh_ifdown(&nd_tbl, dev);
2760                 in6_dev_put(idev);
2761         }
2762         return 0;
2763 }
2764
2765 static void addrconf_rs_timer(unsigned long data)
2766 {
2767         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2768         struct inet6_dev *idev = ifp->idev;
2769
2770         read_lock(&idev->lock);
2771         if (idev->dead || !(idev->if_flags & IF_READY))
2772                 goto out;
2773
2774         if (idev->cnf.forwarding)
2775                 goto out;
2776
2777         /* Announcement received after solicitation was sent */
2778         if (idev->if_flags & IF_RA_RCVD)
2779                 goto out;
2780
2781         spin_lock(&ifp->lock);
2782         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2783                 /* The wait after the last probe can be shorter */
2784                 addrconf_mod_timer(ifp, AC_RS,
2785                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2786                                    idev->cnf.rtr_solicit_delay :
2787                                    idev->cnf.rtr_solicit_interval);
2788                 spin_unlock(&ifp->lock);
2789
2790                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2791         } else {
2792                 spin_unlock(&ifp->lock);
2793                 /*
2794                  * Note: we do not support deprecated "all on-link"
2795                  * assumption any longer.
2796                  */
2797                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2798                        idev->dev->name);
2799         }
2800
2801 out:
2802         read_unlock(&idev->lock);
2803         in6_ifa_put(ifp);
2804 }
2805
2806 /*
2807  *      Duplicate Address Detection
2808  */
2809 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2810 {
2811         unsigned long rand_num;
2812         struct inet6_dev *idev = ifp->idev;
2813
2814         if (ifp->flags & IFA_F_OPTIMISTIC)
2815                 rand_num = 0;
2816         else
2817                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2818
2819         ifp->probes = idev->cnf.dad_transmits;
2820         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2821 }
2822
2823 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2824 {
2825         struct inet6_dev *idev = ifp->idev;
2826         struct net_device *dev = idev->dev;
2827
2828         addrconf_join_solict(dev, &ifp->addr);
2829
2830         net_srandom(ifp->addr.s6_addr32[3]);
2831
2832         read_lock_bh(&idev->lock);
2833         spin_lock(&ifp->lock);
2834         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2835                 goto out;
2836
2837         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2838             idev->cnf.accept_dad < 1 ||
2839             !(ifp->flags&IFA_F_TENTATIVE) ||
2840             ifp->flags & IFA_F_NODAD) {
2841                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2842                 spin_unlock(&ifp->lock);
2843                 read_unlock_bh(&idev->lock);
2844
2845                 addrconf_dad_completed(ifp);
2846                 return;
2847         }
2848
2849         if (!(idev->if_flags & IF_READY)) {
2850                 spin_unlock(&ifp->lock);
2851                 read_unlock_bh(&idev->lock);
2852                 /*
2853                  * If the device is not ready:
2854                  * - keep it tentative if it is a permanent address.
2855                  * - otherwise, kill it.
2856                  */
2857                 in6_ifa_hold(ifp);
2858                 addrconf_dad_stop(ifp, 0);
2859                 return;
2860         }
2861
2862         /*
2863          * Optimistic nodes can start receiving
2864          * Frames right away
2865          */
2866         if (ifp->flags & IFA_F_OPTIMISTIC)
2867                 ip6_ins_rt(ifp->rt);
2868
2869         addrconf_dad_kick(ifp);
2870 out:
2871         spin_unlock(&ifp->lock);
2872         read_unlock_bh(&idev->lock);
2873 }
2874
2875 static void addrconf_dad_timer(unsigned long data)
2876 {
2877         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2878         struct inet6_dev *idev = ifp->idev;
2879         struct in6_addr mcaddr;
2880
2881         if (!ifp->probes && addrconf_dad_end(ifp))
2882                 goto out;
2883
2884         read_lock(&idev->lock);
2885         if (idev->dead || !(idev->if_flags & IF_READY)) {
2886                 read_unlock(&idev->lock);
2887                 goto out;
2888         }
2889
2890         spin_lock(&ifp->lock);
2891         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2892                 spin_unlock(&ifp->lock);
2893                 read_unlock(&idev->lock);
2894                 goto out;
2895         }
2896
2897         if (ifp->probes == 0) {
2898                 /*
2899                  * DAD was successful
2900                  */
2901
2902                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2903                 spin_unlock(&ifp->lock);
2904                 read_unlock(&idev->lock);
2905
2906                 addrconf_dad_completed(ifp);
2907
2908                 goto out;
2909         }
2910
2911         ifp->probes--;
2912         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2913         spin_unlock(&ifp->lock);
2914         read_unlock(&idev->lock);
2915
2916         /* send a neighbour solicitation for our addr */
2917         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2918         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2919 out:
2920         in6_ifa_put(ifp);
2921 }
2922
2923 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2924 {
2925         struct net_device *dev = ifp->idev->dev;
2926
2927         /*
2928          *      Configure the address for reception. Now it is valid.
2929          */
2930
2931         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2932
2933         /* If added prefix is link local and forwarding is off,
2934            start sending router solicitations.
2935          */
2936
2937         if ((ifp->idev->cnf.forwarding == 0 ||
2938              ifp->idev->cnf.forwarding == 2) &&
2939             ifp->idev->cnf.rtr_solicits > 0 &&
2940             (dev->flags&IFF_LOOPBACK) == 0 &&
2941             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2942                 /*
2943                  *      If a host as already performed a random delay
2944                  *      [...] as part of DAD [...] there is no need
2945                  *      to delay again before sending the first RS
2946                  */
2947                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2948
2949                 spin_lock_bh(&ifp->lock);
2950                 ifp->probes = 1;
2951                 ifp->idev->if_flags |= IF_RS_SENT;
2952                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2953                 spin_unlock_bh(&ifp->lock);
2954         }
2955 }
2956
2957 static void addrconf_dad_run(struct inet6_dev *idev)
2958 {
2959         struct inet6_ifaddr *ifp;
2960
2961         read_lock_bh(&idev->lock);
2962         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2963                 spin_lock(&ifp->lock);
2964                 if (ifp->flags & IFA_F_TENTATIVE &&
2965                     ifp->state == INET6_IFADDR_STATE_DAD)
2966                         addrconf_dad_kick(ifp);
2967                 spin_unlock(&ifp->lock);
2968         }
2969         read_unlock_bh(&idev->lock);
2970 }
2971
2972 #ifdef CONFIG_PROC_FS
2973 struct if6_iter_state {
2974         struct seq_net_private p;
2975         int bucket;
2976 };
2977
2978 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2979 {
2980         struct inet6_ifaddr *ifa = NULL;
2981         struct if6_iter_state *state = seq->private;
2982         struct net *net = seq_file_net(seq);
2983
2984         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2985                 struct hlist_node *n;
2986                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2987                                          addr_lst)
2988                         if (net_eq(dev_net(ifa->idev->dev), net))
2989                                 return ifa;
2990         }
2991         return NULL;
2992 }
2993
2994 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2995                                          struct inet6_ifaddr *ifa)
2996 {
2997         struct if6_iter_state *state = seq->private;
2998         struct net *net = seq_file_net(seq);
2999         struct hlist_node *n = &ifa->addr_lst;
3000
3001         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3002                 if (net_eq(dev_net(ifa->idev->dev), net))
3003                         return ifa;
3004
3005         while (++state->bucket < IN6_ADDR_HSIZE) {
3006                 hlist_for_each_entry_rcu_bh(ifa, n,
3007                                      &inet6_addr_lst[state->bucket], addr_lst) {
3008                         if (net_eq(dev_net(ifa->idev->dev), net))
3009                                 return ifa;
3010                 }
3011         }
3012
3013         return NULL;
3014 }
3015
3016 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3017 {
3018         struct inet6_ifaddr *ifa = if6_get_first(seq);
3019
3020         if (ifa)
3021                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3022                         --pos;
3023         return pos ? NULL : ifa;
3024 }
3025
3026 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3027         __acquires(rcu_bh)
3028 {
3029         rcu_read_lock_bh();
3030         return if6_get_idx(seq, *pos);
3031 }
3032
3033 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3034 {
3035         struct inet6_ifaddr *ifa;
3036
3037         ifa = if6_get_next(seq, v);
3038         ++*pos;
3039         return ifa;
3040 }
3041
3042 static void if6_seq_stop(struct seq_file *seq, void *v)
3043         __releases(rcu_bh)
3044 {
3045         rcu_read_unlock_bh();
3046 }
3047
3048 static int if6_seq_show(struct seq_file *seq, void *v)
3049 {
3050         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3051         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3052                    &ifp->addr,
3053                    ifp->idev->dev->ifindex,
3054                    ifp->prefix_len,
3055                    ifp->scope,
3056                    ifp->flags,
3057                    ifp->idev->dev->name);
3058         return 0;
3059 }
3060
3061 static const struct seq_operations if6_seq_ops = {
3062         .start  = if6_seq_start,
3063         .next   = if6_seq_next,
3064         .show   = if6_seq_show,
3065         .stop   = if6_seq_stop,
3066 };
3067
3068 static int if6_seq_open(struct inode *inode, struct file *file)
3069 {
3070         return seq_open_net(inode, file, &if6_seq_ops,
3071                             sizeof(struct if6_iter_state));
3072 }
3073
3074 static const struct file_operations if6_fops = {
3075         .owner          = THIS_MODULE,
3076         .open           = if6_seq_open,
3077         .read           = seq_read,
3078         .llseek         = seq_lseek,
3079         .release        = seq_release_net,
3080 };
3081
3082 static int __net_init if6_proc_net_init(struct net *net)
3083 {
3084         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3085                 return -ENOMEM;
3086         return 0;
3087 }
3088
3089 static void __net_exit if6_proc_net_exit(struct net *net)
3090 {
3091        proc_net_remove(net, "if_inet6");
3092 }
3093
3094 static struct pernet_operations if6_proc_net_ops = {
3095        .init = if6_proc_net_init,
3096        .exit = if6_proc_net_exit,
3097 };
3098
3099 int __init if6_proc_init(void)
3100 {
3101         return register_pernet_subsys(&if6_proc_net_ops);
3102 }
3103
3104 void if6_proc_exit(void)
3105 {
3106         unregister_pernet_subsys(&if6_proc_net_ops);
3107 }
3108 #endif  /* CONFIG_PROC_FS */
3109
3110 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3111 /* Check if address is a home address configured on any interface. */
3112 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3113 {
3114         int ret = 0;
3115         struct inet6_ifaddr *ifp = NULL;
3116         struct hlist_node *n;
3117         unsigned int hash = ipv6_addr_hash(addr);
3118
3119         rcu_read_lock_bh();
3120         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3121                 if (!net_eq(dev_net(ifp->idev->dev), net))
3122                         continue;
3123                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3124                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3125                         ret = 1;
3126                         break;
3127                 }
3128         }
3129         rcu_read_unlock_bh();
3130         return ret;
3131 }
3132 #endif
3133
3134 /*
3135  *      Periodic address status verification
3136  */
3137
3138 static void addrconf_verify(unsigned long foo)
3139 {
3140         unsigned long now, next, next_sec, next_sched;
3141         struct inet6_ifaddr *ifp;
3142         struct hlist_node *node;
3143         int i;
3144
3145         rcu_read_lock_bh();
3146         spin_lock(&addrconf_verify_lock);
3147         now = jiffies;
3148         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3149
3150         del_timer(&addr_chk_timer);
3151
3152         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3153 restart:
3154                 hlist_for_each_entry_rcu_bh(ifp, node,
3155                                          &inet6_addr_lst[i], addr_lst) {
3156                         unsigned long age;
3157
3158                         if (ifp->flags & IFA_F_PERMANENT)
3159                                 continue;
3160
3161                         spin_lock(&ifp->lock);
3162                         /* We try to batch several events at once. */
3163                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3164
3165                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3166                             age >= ifp->valid_lft) {
3167                                 spin_unlock(&ifp->lock);
3168                                 in6_ifa_hold(ifp);
3169                                 ipv6_del_addr(ifp);
3170                                 goto restart;
3171                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3172                                 spin_unlock(&ifp->lock);
3173                                 continue;
3174                         } else if (age >= ifp->prefered_lft) {
3175                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3176                                 int deprecate = 0;
3177
3178                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3179                                         deprecate = 1;
3180                                         ifp->flags |= IFA_F_DEPRECATED;
3181                                 }
3182
3183                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3184                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3185
3186                                 spin_unlock(&ifp->lock);
3187
3188                                 if (deprecate) {
3189                                         in6_ifa_hold(ifp);
3190
3191                                         ipv6_ifa_notify(0, ifp);
3192                                         in6_ifa_put(ifp);
3193                                         goto restart;
3194                                 }
3195 #ifdef CONFIG_IPV6_PRIVACY
3196                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3197                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3198                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3199                                         ifp->idev->cnf.dad_transmits *
3200                                         ifp->idev->nd_parms->retrans_time / HZ;
3201
3202                                 if (age >= ifp->prefered_lft - regen_advance) {
3203                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3204                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3205                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3206                                         if (!ifp->regen_count && ifpub) {
3207                                                 ifp->regen_count++;
3208                                                 in6_ifa_hold(ifp);
3209                                                 in6_ifa_hold(ifpub);
3210                                                 spin_unlock(&ifp->lock);
3211
3212                                                 spin_lock(&ifpub->lock);
3213                                                 ifpub->regen_count = 0;
3214                                                 spin_unlock(&ifpub->lock);
3215                                                 ipv6_create_tempaddr(ifpub, ifp);
3216                                                 in6_ifa_put(ifpub);
3217                                                 in6_ifa_put(ifp);
3218                                                 goto restart;
3219                                         }
3220                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3221                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3222                                 spin_unlock(&ifp->lock);
3223 #endif
3224                         } else {
3225                                 /* ifp->prefered_lft <= ifp->valid_lft */
3226                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3227                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3228                                 spin_unlock(&ifp->lock);
3229                         }
3230                 }
3231         }
3232
3233         next_sec = round_jiffies_up(next);
3234         next_sched = next;
3235
3236         /* If rounded timeout is accurate enough, accept it. */
3237         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3238                 next_sched = next_sec;
3239
3240         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3241         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3242                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3243
3244         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3245               now, next, next_sec, next_sched));
3246
3247         addr_chk_timer.expires = next_sched;
3248         add_timer(&addr_chk_timer);
3249         spin_unlock(&addrconf_verify_lock);
3250         rcu_read_unlock_bh();
3251 }
3252
3253 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3254 {
3255         struct in6_addr *pfx = NULL;
3256
3257         if (addr)
3258                 pfx = nla_data(addr);
3259
3260         if (local) {
3261                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3262                         pfx = NULL;
3263                 else
3264                         pfx = nla_data(local);
3265         }
3266
3267         return pfx;
3268 }
3269
3270 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3271         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3272         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3273         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3274 };
3275
3276 static int
3277 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3278 {
3279         struct net *net = sock_net(skb->sk);
3280         struct ifaddrmsg *ifm;
3281         struct nlattr *tb[IFA_MAX+1];
3282         struct in6_addr *pfx;
3283         int err;
3284
3285         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3286         if (err < 0)
3287                 return err;
3288
3289         ifm = nlmsg_data(nlh);
3290         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3291         if (pfx == NULL)
3292                 return -EINVAL;
3293
3294         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3295 }
3296
3297 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3298                              u32 prefered_lft, u32 valid_lft)
3299 {
3300         u32 flags;
3301         clock_t expires;
3302         unsigned long timeout;
3303
3304         if (!valid_lft || (prefered_lft > valid_lft))
3305                 return -EINVAL;
3306
3307         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3308         if (addrconf_finite_timeout(timeout)) {
3309                 expires = jiffies_to_clock_t(timeout * HZ);
3310                 valid_lft = timeout;
3311                 flags = RTF_EXPIRES;
3312         } else {
3313                 expires = 0;
3314                 flags = 0;
3315                 ifa_flags |= IFA_F_PERMANENT;
3316         }
3317
3318         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3319         if (addrconf_finite_timeout(timeout)) {
3320                 if (timeout == 0)
3321                         ifa_flags |= IFA_F_DEPRECATED;
3322                 prefered_lft = timeout;
3323         }
3324
3325         spin_lock_bh(&ifp->lock);
3326         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3327         ifp->tstamp = jiffies;
3328         ifp->valid_lft = valid_lft;
3329         ifp->prefered_lft = prefered_lft;
3330
3331         spin_unlock_bh(&ifp->lock);
3332         if (!(ifp->flags&IFA_F_TENTATIVE))
3333                 ipv6_ifa_notify(0, ifp);
3334
3335         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3336                               expires, flags);
3337         addrconf_verify(0);
3338
3339         return 0;
3340 }
3341
3342 static int
3343 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3344 {
3345         struct net *net = sock_net(skb->sk);
3346         struct ifaddrmsg *ifm;
3347         struct nlattr *tb[IFA_MAX+1];
3348         struct in6_addr *pfx;
3349         struct inet6_ifaddr *ifa;
3350         struct net_device *dev;
3351         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3352         u8 ifa_flags;
3353         int err;
3354
3355         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3356         if (err < 0)
3357                 return err;
3358
3359         ifm = nlmsg_data(nlh);
3360         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3361         if (pfx == NULL)
3362                 return -EINVAL;
3363
3364         if (tb[IFA_CACHEINFO]) {
3365                 struct ifa_cacheinfo *ci;
3366
3367                 ci = nla_data(tb[IFA_CACHEINFO]);
3368                 valid_lft = ci->ifa_valid;
3369                 preferred_lft = ci->ifa_prefered;
3370         } else {
3371                 preferred_lft = INFINITY_LIFE_TIME;
3372                 valid_lft = INFINITY_LIFE_TIME;
3373         }
3374
3375         dev =  __dev_get_by_index(net, ifm->ifa_index);
3376         if (dev == NULL)
3377                 return -ENODEV;
3378
3379         /* We ignore other flags so far. */
3380         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3381
3382         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3383         if (ifa == NULL) {
3384                 /*
3385                  * It would be best to check for !NLM_F_CREATE here but
3386                  * userspace alreay relies on not having to provide this.
3387                  */
3388                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3389                                       ifm->ifa_prefixlen, ifa_flags,
3390                                       preferred_lft, valid_lft);
3391         }
3392
3393         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3394             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3395                 err = -EEXIST;
3396         else
3397                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3398
3399         in6_ifa_put(ifa);
3400
3401         return err;
3402 }
3403
3404 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3405                           u8 scope, int ifindex)
3406 {
3407         struct ifaddrmsg *ifm;
3408
3409         ifm = nlmsg_data(nlh);
3410         ifm->ifa_family = AF_INET6;
3411         ifm->ifa_prefixlen = prefixlen;
3412         ifm->ifa_flags = flags;
3413         ifm->ifa_scope = scope;
3414         ifm->ifa_index = ifindex;
3415 }
3416
3417 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3418                          unsigned long tstamp, u32 preferred, u32 valid)
3419 {
3420         struct ifa_cacheinfo ci;
3421
3422         ci.cstamp = cstamp_delta(cstamp);
3423         ci.tstamp = cstamp_delta(tstamp);
3424         ci.ifa_prefered = preferred;
3425         ci.ifa_valid = valid;
3426
3427         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3428 }
3429
3430 static inline int rt_scope(int ifa_scope)
3431 {
3432         if (ifa_scope & IFA_HOST)
3433                 return RT_SCOPE_HOST;
3434         else if (ifa_scope & IFA_LINK)
3435                 return RT_SCOPE_LINK;
3436         else if (ifa_scope & IFA_SITE)
3437                 return RT_SCOPE_SITE;
3438         else
3439                 return RT_SCOPE_UNIVERSE;
3440 }
3441
3442 static inline int inet6_ifaddr_msgsize(void)
3443 {
3444         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3445                + nla_total_size(16) /* IFA_ADDRESS */
3446                + nla_total_size(sizeof(struct ifa_cacheinfo));
3447 }
3448
3449 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3450                              u32 pid, u32 seq, int event, unsigned int flags)
3451 {
3452         struct nlmsghdr  *nlh;
3453         u32 preferred, valid;
3454
3455         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3456         if (nlh == NULL)
3457                 return -EMSGSIZE;
3458
3459         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3460                       ifa->idev->dev->ifindex);
3461
3462         if (!(ifa->flags&IFA_F_PERMANENT)) {
3463                 preferred = ifa->prefered_lft;
3464                 valid = ifa->valid_lft;
3465                 if (preferred != INFINITY_LIFE_TIME) {
3466                         long tval = (jiffies - ifa->tstamp)/HZ;
3467                         if (preferred > tval)
3468                                 preferred -= tval;
3469                         else
3470                                 preferred = 0;
3471                         if (valid != INFINITY_LIFE_TIME) {
3472                                 if (valid > tval)
3473                                         valid -= tval;
3474                                 else
3475                                         valid = 0;
3476                         }
3477                 }
3478         } else {
3479                 preferred = INFINITY_LIFE_TIME;
3480                 valid = INFINITY_LIFE_TIME;
3481         }
3482
3483         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3484             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3485                 nlmsg_cancel(skb, nlh);
3486                 return -EMSGSIZE;
3487         }
3488
3489         return nlmsg_end(skb, nlh);
3490 }
3491
3492 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3493                                 u32 pid, u32 seq, int event, u16 flags)
3494 {
3495         struct nlmsghdr  *nlh;
3496         u8 scope = RT_SCOPE_UNIVERSE;
3497         int ifindex = ifmca->idev->dev->ifindex;
3498
3499         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3500                 scope = RT_SCOPE_SITE;
3501
3502         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3503         if (nlh == NULL)
3504                 return -EMSGSIZE;
3505
3506         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3507         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3508             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3509                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3510                 nlmsg_cancel(skb, nlh);
3511                 return -EMSGSIZE;
3512         }
3513
3514         return nlmsg_end(skb, nlh);
3515 }
3516
3517 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3518                                 u32 pid, u32 seq, int event, unsigned int flags)
3519 {
3520         struct nlmsghdr  *nlh;
3521         u8 scope = RT_SCOPE_UNIVERSE;
3522         int ifindex = ifaca->aca_idev->dev->ifindex;
3523
3524         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3525                 scope = RT_SCOPE_SITE;
3526
3527         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3528         if (nlh == NULL)
3529                 return -EMSGSIZE;
3530
3531         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3532         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3533             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3534                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3535                 nlmsg_cancel(skb, nlh);
3536                 return -EMSGSIZE;
3537         }
3538
3539         return nlmsg_end(skb, nlh);
3540 }
3541
3542 enum addr_type_t {
3543         UNICAST_ADDR,
3544         MULTICAST_ADDR,
3545         ANYCAST_ADDR,
3546 };
3547
3548 /* called with rcu_read_lock() */
3549 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3550                           struct netlink_callback *cb, enum addr_type_t type,
3551                           int s_ip_idx, int *p_ip_idx)
3552 {
3553         struct ifmcaddr6 *ifmca;
3554         struct ifacaddr6 *ifaca;
3555         int err = 1;
3556         int ip_idx = *p_ip_idx;
3557
3558         read_lock_bh(&idev->lock);
3559         switch (type) {
3560         case UNICAST_ADDR: {
3561                 struct inet6_ifaddr *ifa;
3562
3563                 /* unicast address incl. temp addr */
3564                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3565                         if (++ip_idx < s_ip_idx)
3566                                 continue;
3567                         err = inet6_fill_ifaddr(skb, ifa,
3568                                                 NETLINK_CB(cb->skb).pid,
3569                                                 cb->nlh->nlmsg_seq,
3570                                                 RTM_NEWADDR,
3571                                                 NLM_F_MULTI);
3572                         if (err <= 0)
3573                                 break;
3574                 }
3575                 break;
3576         }
3577         case MULTICAST_ADDR:
3578                 /* multicast address */
3579                 for (ifmca = idev->mc_list; ifmca;
3580                      ifmca = ifmca->next, ip_idx++) {
3581                         if (ip_idx < s_ip_idx)
3582                                 continue;
3583                         err = inet6_fill_ifmcaddr(skb, ifmca,
3584                                                   NETLINK_CB(cb->skb).pid,
3585                                                   cb->nlh->nlmsg_seq,
3586                                                   RTM_GETMULTICAST,
3587                                                   NLM_F_MULTI);
3588                         if (err <= 0)
3589                                 break;
3590                 }
3591                 break;
3592         case ANYCAST_ADDR:
3593                 /* anycast address */
3594                 for (ifaca = idev->ac_list; ifaca;
3595                      ifaca = ifaca->aca_next, ip_idx++) {
3596                         if (ip_idx < s_ip_idx)
3597                                 continue;
3598                         err = inet6_fill_ifacaddr(skb, ifaca,
3599                                                   NETLINK_CB(cb->skb).pid,
3600                                                   cb->nlh->nlmsg_seq,
3601                                                   RTM_GETANYCAST,
3602                                                   NLM_F_MULTI);
3603                         if (err <= 0)
3604                                 break;
3605                 }
3606                 break;
3607         default:
3608                 break;
3609         }
3610         read_unlock_bh(&idev->lock);
3611         *p_ip_idx = ip_idx;
3612         return err;
3613 }
3614
3615 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3616                            enum addr_type_t type)
3617 {
3618         struct net *net = sock_net(skb->sk);
3619         int h, s_h;
3620         int idx, ip_idx;
3621         int s_idx, s_ip_idx;
3622         struct net_device *dev;
3623         struct inet6_dev *idev;
3624         struct hlist_head *head;
3625         struct hlist_node *node;
3626
3627         s_h = cb->args[0];
3628         s_idx = idx = cb->args[1];
3629         s_ip_idx = ip_idx = cb->args[2];
3630
3631         rcu_read_lock();
3632         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3633                 idx = 0;
3634                 head = &net->dev_index_head[h];
3635                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3636                         if (idx < s_idx)
3637                                 goto cont;
3638                         if (h > s_h || idx > s_idx)
3639                                 s_ip_idx = 0;
3640                         ip_idx = 0;
3641                         idev = __in6_dev_get(dev);
3642                         if (!idev)
3643                                 goto cont;
3644
3645                         if (in6_dump_addrs(idev, skb, cb, type,
3646                                            s_ip_idx, &ip_idx) <= 0)
3647                                 goto done;
3648 cont:
3649                         idx++;
3650                 }
3651         }
3652 done:
3653         rcu_read_unlock();
3654         cb->args[0] = h;
3655         cb->args[1] = idx;
3656         cb->args[2] = ip_idx;
3657
3658         return skb->len;
3659 }
3660
3661 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3662 {
3663         enum addr_type_t type = UNICAST_ADDR;
3664
3665         return inet6_dump_addr(skb, cb, type);
3666 }
3667
3668 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3669 {
3670         enum addr_type_t type = MULTICAST_ADDR;
3671
3672         return inet6_dump_addr(skb, cb, type);
3673 }
3674
3675
3676 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3677 {
3678         enum addr_type_t type = ANYCAST_ADDR;
3679
3680         return inet6_dump_addr(skb, cb, type);
3681 }
3682
3683 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3684                              void *arg)
3685 {
3686         struct net *net = sock_net(in_skb->sk);
3687         struct ifaddrmsg *ifm;
3688         struct nlattr *tb[IFA_MAX+1];
3689         struct in6_addr *addr = NULL;
3690         struct net_device *dev = NULL;
3691         struct inet6_ifaddr *ifa;
3692         struct sk_buff *skb;
3693         int err;
3694
3695         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3696         if (err < 0)
3697                 goto errout;
3698
3699         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3700         if (addr == NULL) {
3701                 err = -EINVAL;
3702                 goto errout;
3703         }
3704
3705         ifm = nlmsg_data(nlh);
3706         if (ifm->ifa_index)
3707                 dev = __dev_get_by_index(net, ifm->ifa_index);
3708
3709         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3710         if (!ifa) {
3711                 err = -EADDRNOTAVAIL;
3712                 goto errout;
3713         }
3714
3715         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3716         if (!skb) {
3717                 err = -ENOBUFS;
3718                 goto errout_ifa;
3719         }
3720
3721         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3722                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3723         if (err < 0) {
3724                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3725                 WARN_ON(err == -EMSGSIZE);
3726                 kfree_skb(skb);
3727                 goto errout_ifa;
3728         }
3729         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3730 errout_ifa:
3731         in6_ifa_put(ifa);
3732 errout:
3733         return err;
3734 }
3735
3736 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3737 {
3738         struct sk_buff *skb;
3739         struct net *net = dev_net(ifa->idev->dev);
3740         int err = -ENOBUFS;
3741
3742         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3743         if (skb == NULL)
3744                 goto errout;
3745
3746         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3747         if (err < 0) {
3748                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3749                 WARN_ON(err == -EMSGSIZE);
3750                 kfree_skb(skb);
3751                 goto errout;
3752         }
3753         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3754         return;
3755 errout:
3756         if (err < 0)
3757                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3758 }
3759
3760 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3761                                 __s32 *array, int bytes)
3762 {
3763         BUG_ON(bytes < (DEVCONF_MAX * 4));
3764
3765         memset(array, 0, bytes);
3766         array[DEVCONF_FORWARDING] = cnf->forwarding;
3767         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3768         array[DEVCONF_MTU6] = cnf->mtu6;
3769         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3770         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3771         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3772         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3773         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3774         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3775                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3776         array[DEVCONF_RTR_SOLICIT_DELAY] =
3777                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3778         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3779 #ifdef CONFIG_IPV6_PRIVACY
3780         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3781         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3782         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3783         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3784         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3785 #endif
3786         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3787         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3788         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3789 #ifdef CONFIG_IPV6_ROUTER_PREF
3790         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3791         array[DEVCONF_RTR_PROBE_INTERVAL] =
3792                 jiffies_to_msecs(cnf->rtr_probe_interval);
3793 #ifdef CONFIG_IPV6_ROUTE_INFO
3794         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3795 #endif
3796 #endif
3797         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3798         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3799 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3800         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3801 #endif
3802 #ifdef CONFIG_IPV6_MROUTE
3803         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3804 #endif
3805         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3806         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3807         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3808 }
3809
3810 static inline size_t inet6_ifla6_size(void)
3811 {
3812         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3813              + nla_total_size(sizeof(struct ifla_cacheinfo))
3814              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3815              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3816              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3817 }
3818
3819 static inline size_t inet6_if_nlmsg_size(void)
3820 {
3821         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3822                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3823                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3824                + nla_total_size(4) /* IFLA_MTU */
3825                + nla_total_size(4) /* IFLA_LINK */
3826                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3827 }
3828
3829 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3830                                       int items, int bytes)
3831 {
3832         int i;
3833         int pad = bytes - sizeof(u64) * items;
3834         BUG_ON(pad < 0);
3835
3836         /* Use put_unaligned() because stats may not be aligned for u64. */
3837         put_unaligned(items, &stats[0]);
3838         for (i = 1; i < items; i++)
3839                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3840
3841         memset(&stats[items], 0, pad);
3842 }
3843
3844 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3845                                       int items, int bytes, size_t syncpoff)
3846 {
3847         int i;
3848         int pad = bytes - sizeof(u64) * items;
3849         BUG_ON(pad < 0);
3850
3851         /* Use put_unaligned() because stats may not be aligned for u64. */
3852         put_unaligned(items, &stats[0]);
3853         for (i = 1; i < items; i++)
3854                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3855
3856         memset(&stats[items], 0, pad);
3857 }
3858
3859 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3860                              int bytes)
3861 {
3862         switch (attrtype) {
3863         case IFLA_INET6_STATS:
3864                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3865                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3866                 break;
3867         case IFLA_INET6_ICMP6STATS:
3868                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3869                 break;
3870         }
3871 }
3872
3873 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3874 {
3875         struct nlattr *nla;
3876         struct ifla_cacheinfo ci;
3877
3878         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3879
3880         ci.max_reasm_len = IPV6_MAXPLEN;
3881         ci.tstamp = cstamp_delta(idev->tstamp);
3882         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3883         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3884         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3885
3886         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3887         if (nla == NULL)
3888                 goto nla_put_failure;
3889         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3890
3891         /* XXX - MC not implemented */
3892
3893         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3894         if (nla == NULL)
3895                 goto nla_put_failure;
3896         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3897
3898         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3899         if (nla == NULL)
3900                 goto nla_put_failure;
3901         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3902
3903         return 0;
3904
3905 nla_put_failure:
3906         return -EMSGSIZE;
3907 }
3908
3909 static size_t inet6_get_link_af_size(const struct net_device *dev)
3910 {
3911         if (!__in6_dev_get(dev))
3912                 return 0;
3913
3914         return inet6_ifla6_size();
3915 }
3916
3917 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3918 {
3919         struct inet6_dev *idev = __in6_dev_get(dev);
3920
3921         if (!idev)
3922                 return -ENODATA;
3923
3924         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3925                 return -EMSGSIZE;
3926
3927         return 0;
3928 }
3929
3930 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3931                              u32 pid, u32 seq, int event, unsigned int flags)
3932 {
3933         struct net_device *dev = idev->dev;
3934         struct ifinfomsg *hdr;
3935         struct nlmsghdr *nlh;
3936         void *protoinfo;
3937
3938         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3939         if (nlh == NULL)
3940                 return -EMSGSIZE;
3941
3942         hdr = nlmsg_data(nlh);
3943         hdr->ifi_family = AF_INET6;
3944         hdr->__ifi_pad = 0;
3945         hdr->ifi_type = dev->type;
3946         hdr->ifi_index = dev->ifindex;
3947         hdr->ifi_flags = dev_get_flags(dev);
3948         hdr->ifi_change = 0;
3949
3950         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3951
3952         if (dev->addr_len)
3953                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3954
3955         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3956         if (dev->ifindex != dev->iflink)
3957                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3958
3959         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3960         if (protoinfo == NULL)
3961                 goto nla_put_failure;
3962
3963         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3964                 goto nla_put_failure;
3965
3966         nla_nest_end(skb, protoinfo);
3967         return nlmsg_end(skb, nlh);
3968
3969 nla_put_failure:
3970         nlmsg_cancel(skb, nlh);
3971         return -EMSGSIZE;
3972 }
3973
3974 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3975 {
3976         struct net *net = sock_net(skb->sk);
3977         int h, s_h;
3978         int idx = 0, s_idx;
3979         struct net_device *dev;
3980         struct inet6_dev *idev;
3981         struct hlist_head *head;
3982         struct hlist_node *node;
3983
3984         s_h = cb->args[0];
3985         s_idx = cb->args[1];
3986
3987         rcu_read_lock();
3988         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3989                 idx = 0;
3990                 head = &net->dev_index_head[h];
3991                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3992                         if (idx < s_idx)
3993                                 goto cont;
3994                         idev = __in6_dev_get(dev);
3995                         if (!idev)
3996                                 goto cont;
3997                         if (inet6_fill_ifinfo(skb, idev,
3998                                               NETLINK_CB(cb->skb).pid,
3999                                               cb->nlh->nlmsg_seq,
4000                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4001                                 goto out;
4002 cont:
4003                         idx++;
4004                 }
4005         }
4006 out:
4007         rcu_read_unlock();
4008         cb->args[1] = idx;
4009         cb->args[0] = h;
4010
4011         return skb->len;
4012 }
4013
4014 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4015 {
4016         struct sk_buff *skb;
4017         struct net *net = dev_net(idev->dev);
4018         int err = -ENOBUFS;
4019
4020         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4021         if (skb == NULL)
4022                 goto errout;
4023
4024         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4025         if (err < 0) {
4026                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4027                 WARN_ON(err == -EMSGSIZE);
4028                 kfree_skb(skb);
4029                 goto errout;
4030         }
4031         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4032         return;
4033 errout:
4034         if (err < 0)
4035                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4036 }
4037
4038 static inline size_t inet6_prefix_nlmsg_size(void)
4039 {
4040         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4041                + nla_total_size(sizeof(struct in6_addr))
4042                + nla_total_size(sizeof(struct prefix_cacheinfo));
4043 }
4044
4045 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4046                              struct prefix_info *pinfo, u32 pid, u32 seq,
4047                              int event, unsigned int flags)
4048 {
4049         struct prefixmsg *pmsg;
4050         struct nlmsghdr *nlh;
4051         struct prefix_cacheinfo ci;
4052
4053         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4054         if (nlh == NULL)
4055                 return -EMSGSIZE;
4056
4057         pmsg = nlmsg_data(nlh);
4058         pmsg->prefix_family = AF_INET6;
4059         pmsg->prefix_pad1 = 0;
4060         pmsg->prefix_pad2 = 0;
4061         pmsg->prefix_ifindex = idev->dev->ifindex;
4062         pmsg->prefix_len = pinfo->prefix_len;
4063         pmsg->prefix_type = pinfo->type;
4064         pmsg->prefix_pad3 = 0;
4065         pmsg->prefix_flags = 0;
4066         if (pinfo->onlink)
4067                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4068         if (pinfo->autoconf)
4069                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4070
4071         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4072
4073         ci.preferred_time = ntohl(pinfo->prefered);
4074         ci.valid_time = ntohl(pinfo->valid);
4075         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4076
4077         return nlmsg_end(skb, nlh);
4078
4079 nla_put_failure:
4080         nlmsg_cancel(skb, nlh);
4081         return -EMSGSIZE;
4082 }
4083
4084 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4085                          struct prefix_info *pinfo)
4086 {
4087         struct sk_buff *skb;
4088         struct net *net = dev_net(idev->dev);
4089         int err = -ENOBUFS;
4090
4091         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4092         if (skb == NULL)
4093                 goto errout;
4094
4095         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4096         if (err < 0) {
4097                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4098                 WARN_ON(err == -EMSGSIZE);
4099                 kfree_skb(skb);
4100                 goto errout;
4101         }
4102         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4103         return;
4104 errout:
4105         if (err < 0)
4106                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4107 }
4108
4109 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4110 {
4111         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4112
4113         switch (event) {
4114         case RTM_NEWADDR:
4115                 /*
4116                  * If the address was optimistic
4117                  * we inserted the route at the start of
4118                  * our DAD process, so we don't need
4119                  * to do it again
4120                  */
4121                 if (!(ifp->rt->rt6i_node))
4122                         ip6_ins_rt(ifp->rt);
4123                 if (ifp->idev->cnf.forwarding)
4124                         addrconf_join_anycast(ifp);
4125                 break;
4126         case RTM_DELADDR:
4127                 if (ifp->idev->cnf.forwarding)
4128                         addrconf_leave_anycast(ifp);
4129                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4130                 dst_hold(&ifp->rt->dst);
4131
4132                 if (ip6_del_rt(ifp->rt))
4133                         dst_free(&ifp->rt->dst);
4134                 break;
4135         }
4136 }
4137
4138 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4139 {
4140         rcu_read_lock_bh();
4141         if (likely(ifp->idev->dead == 0))
4142                 __ipv6_ifa_notify(event, ifp);
4143         rcu_read_unlock_bh();
4144 }
4145
4146 #ifdef CONFIG_SYSCTL
4147
4148 static
4149 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4150                            void __user *buffer, size_t *lenp, loff_t *ppos)
4151 {
4152         int *valp = ctl->data;
4153         int val = *valp;
4154         loff_t pos = *ppos;
4155         int ret;
4156
4157         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4158
4159         if (write)
4160                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4161         if (ret)
4162                 *ppos = pos;
4163         return ret;
4164 }
4165
4166 static void dev_disable_change(struct inet6_dev *idev)
4167 {
4168         if (!idev || !idev->dev)
4169                 return;
4170
4171         if (idev->cnf.disable_ipv6)
4172                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4173         else
4174                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4175 }
4176
4177 static void addrconf_disable_change(struct net *net, __s32 newf)
4178 {
4179         struct net_device *dev;
4180         struct inet6_dev *idev;
4181
4182         rcu_read_lock();
4183         for_each_netdev_rcu(net, dev) {
4184                 idev = __in6_dev_get(dev);
4185                 if (idev) {
4186                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4187                         idev->cnf.disable_ipv6 = newf;
4188                         if (changed)
4189                                 dev_disable_change(idev);
4190                 }
4191         }
4192         rcu_read_unlock();
4193 }
4194
4195 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4196 {
4197         struct net *net;
4198
4199         net = (struct net *)table->extra2;
4200
4201         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4202                 return 0;
4203
4204         if (!rtnl_trylock()) {
4205                 /* Restore the original values before restarting */
4206                 *p = old;
4207                 return restart_syscall();
4208         }
4209
4210         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4211                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4212                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4213                 addrconf_disable_change(net, newf);
4214         } else if ((!*p) ^ (!old))
4215                 dev_disable_change((struct inet6_dev *)table->extra1);
4216
4217         rtnl_unlock();
4218         return 0;
4219 }
4220
4221 static
4222 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4223                             void __user *buffer, size_t *lenp, loff_t *ppos)
4224 {
4225         int *valp = ctl->data;
4226         int val = *valp;
4227         loff_t pos = *ppos;
4228         int ret;
4229
4230         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4231
4232         if (write)
4233                 ret = addrconf_disable_ipv6(ctl, valp, val);
4234         if (ret)
4235                 *ppos = pos;
4236         return ret;
4237 }
4238
4239 static struct addrconf_sysctl_table
4240 {
4241         struct ctl_table_header *sysctl_header;
4242         ctl_table addrconf_vars[DEVCONF_MAX+1];
4243         char *dev_name;
4244 } addrconf_sysctl __read_mostly = {
4245         .sysctl_header = NULL,
4246         .addrconf_vars = {
4247                 {
4248                         .procname       = "forwarding",
4249                         .data           = &ipv6_devconf.forwarding,
4250                         .maxlen         = sizeof(int),
4251                         .mode           = 0644,
4252                         .proc_handler   = addrconf_sysctl_forward,
4253                 },
4254                 {
4255                         .procname       = "hop_limit",
4256                         .data           = &ipv6_devconf.hop_limit,
4257                         .maxlen         = sizeof(int),
4258                         .mode           = 0644,
4259                         .proc_handler   = proc_dointvec,
4260                 },
4261                 {
4262                         .procname       = "mtu",
4263                         .data           = &ipv6_devconf.mtu6,
4264                         .maxlen         = sizeof(int),
4265                         .mode           = 0644,
4266                         .proc_handler   = proc_dointvec,
4267                 },
4268                 {
4269                         .procname       = "accept_ra",
4270                         .data           = &ipv6_devconf.accept_ra,
4271                         .maxlen         = sizeof(int),
4272                         .mode           = 0644,
4273                         .proc_handler   = proc_dointvec,
4274                 },
4275                 {
4276                         .procname       = "accept_redirects",
4277                         .data           = &ipv6_devconf.accept_redirects,
4278                         .maxlen         = sizeof(int),
4279                         .mode           = 0644,
4280                         .proc_handler   = proc_dointvec,
4281                 },
4282                 {
4283                         .procname       = "autoconf",
4284                         .data           = &ipv6_devconf.autoconf,
4285                         .maxlen         = sizeof(int),
4286                         .mode           = 0644,
4287                         .proc_handler   = proc_dointvec,
4288                 },
4289                 {
4290                         .procname       = "dad_transmits",
4291                         .data           = &ipv6_devconf.dad_transmits,
4292                         .maxlen         = sizeof(int),
4293                         .mode           = 0644,
4294                         .proc_handler   = proc_dointvec,
4295                 },
4296                 {
4297                         .procname       = "router_solicitations",
4298                         .data           = &ipv6_devconf.rtr_solicits,
4299                         .maxlen         = sizeof(int),
4300                         .mode           = 0644,
4301                         .proc_handler   = proc_dointvec,
4302                 },
4303                 {
4304                         .procname       = "router_solicitation_interval",
4305                         .data           = &ipv6_devconf.rtr_solicit_interval,
4306                         .maxlen         = sizeof(int),
4307                         .mode           = 0644,
4308                         .proc_handler   = proc_dointvec_jiffies,
4309                 },
4310                 {
4311                         .procname       = "router_solicitation_delay",
4312                         .data           = &ipv6_devconf.rtr_solicit_delay,
4313                         .maxlen         = sizeof(int),
4314                         .mode           = 0644,
4315                         .proc_handler   = proc_dointvec_jiffies,
4316                 },
4317                 {
4318                         .procname       = "force_mld_version",
4319                         .data           = &ipv6_devconf.force_mld_version,
4320                         .maxlen         = sizeof(int),
4321                         .mode           = 0644,
4322                         .proc_handler   = proc_dointvec,
4323                 },
4324 #ifdef CONFIG_IPV6_PRIVACY
4325                 {
4326                         .procname       = "use_tempaddr",
4327                         .data           = &ipv6_devconf.use_tempaddr,
4328                         .maxlen         = sizeof(int),
4329                         .mode           = 0644,
4330                         .proc_handler   = proc_dointvec,
4331                 },
4332                 {
4333                         .procname       = "temp_valid_lft",
4334                         .data           = &ipv6_devconf.temp_valid_lft,
4335                         .maxlen         = sizeof(int),
4336                         .mode           = 0644,
4337                         .proc_handler   = proc_dointvec,
4338                 },
4339                 {
4340                         .procname       = "temp_prefered_lft",
4341                         .data           = &ipv6_devconf.temp_prefered_lft,
4342                         .maxlen         = sizeof(int),
4343                         .mode           = 0644,
4344                         .proc_handler   = proc_dointvec,
4345                 },
4346                 {
4347                         .procname       = "regen_max_retry",
4348                         .data           = &ipv6_devconf.regen_max_retry,
4349                         .maxlen         = sizeof(int),
4350                         .mode           = 0644,
4351                         .proc_handler   = proc_dointvec,
4352                 },
4353                 {
4354                         .procname       = "max_desync_factor",
4355                         .data           = &ipv6_devconf.max_desync_factor,
4356                         .maxlen         = sizeof(int),
4357                         .mode           = 0644,
4358                         .proc_handler   = proc_dointvec,
4359                 },
4360 #endif
4361                 {
4362                         .procname       = "max_addresses",
4363                         .data           = &ipv6_devconf.max_addresses,
4364                         .maxlen         = sizeof(int),
4365                         .mode           = 0644,
4366                         .proc_handler   = proc_dointvec,
4367                 },
4368                 {
4369                         .procname       = "accept_ra_defrtr",
4370                         .data           = &ipv6_devconf.accept_ra_defrtr,
4371                         .maxlen         = sizeof(int),
4372                         .mode           = 0644,
4373                         .proc_handler   = proc_dointvec,
4374                 },
4375                 {
4376                         .procname       = "accept_ra_pinfo",
4377                         .data           = &ipv6_devconf.accept_ra_pinfo,
4378                         .maxlen         = sizeof(int),
4379                         .mode           = 0644,
4380                         .proc_handler   = proc_dointvec,
4381                 },
4382 #ifdef CONFIG_IPV6_ROUTER_PREF
4383                 {
4384                         .procname       = "accept_ra_rtr_pref",
4385                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4386                         .maxlen         = sizeof(int),
4387                         .mode           = 0644,
4388                         .proc_handler   = proc_dointvec,
4389                 },
4390                 {
4391                         .procname       = "router_probe_interval",
4392                         .data           = &ipv6_devconf.rtr_probe_interval,
4393                         .maxlen         = sizeof(int),
4394                         .mode           = 0644,
4395                         .proc_handler   = proc_dointvec_jiffies,
4396                 },
4397 #ifdef CONFIG_IPV6_ROUTE_INFO
4398                 {
4399                         .procname       = "accept_ra_rt_info_max_plen",
4400                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405 #endif
4406 #endif
4407                 {
4408                         .procname       = "proxy_ndp",
4409                         .data           = &ipv6_devconf.proxy_ndp,
4410                         .maxlen         = sizeof(int),
4411                         .mode           = 0644,
4412                         .proc_handler   = proc_dointvec,
4413                 },
4414                 {
4415                         .procname       = "accept_source_route",
4416                         .data           = &ipv6_devconf.accept_source_route,
4417                         .maxlen         = sizeof(int),
4418                         .mode           = 0644,
4419                         .proc_handler   = proc_dointvec,
4420                 },
4421 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4422                 {
4423                         .procname       = "optimistic_dad",
4424                         .data           = &ipv6_devconf.optimistic_dad,
4425                         .maxlen         = sizeof(int),
4426                         .mode           = 0644,
4427                         .proc_handler   = proc_dointvec,
4428
4429                 },
4430 #endif
4431 #ifdef CONFIG_IPV6_MROUTE
4432                 {
4433                         .procname       = "mc_forwarding",
4434                         .data           = &ipv6_devconf.mc_forwarding,
4435                         .maxlen         = sizeof(int),
4436                         .mode           = 0444,
4437                         .proc_handler   = proc_dointvec,
4438                 },
4439 #endif
4440                 {
4441                         .procname       = "disable_ipv6",
4442                         .data           = &ipv6_devconf.disable_ipv6,
4443                         .maxlen         = sizeof(int),
4444                         .mode           = 0644,
4445                         .proc_handler   = addrconf_sysctl_disable,
4446                 },
4447                 {
4448                         .procname       = "accept_dad",
4449                         .data           = &ipv6_devconf.accept_dad,
4450                         .maxlen         = sizeof(int),
4451                         .mode           = 0644,
4452                         .proc_handler   = proc_dointvec,
4453                 },
4454                 {
4455                         .procname       = "force_tllao",
4456                         .data           = &ipv6_devconf.force_tllao,
4457                         .maxlen         = sizeof(int),
4458                         .mode           = 0644,
4459                         .proc_handler   = proc_dointvec
4460                 },
4461                 {
4462                         /* sentinel */
4463                 }
4464         },
4465 };
4466
4467 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4468                 struct inet6_dev *idev, struct ipv6_devconf *p)
4469 {
4470         int i;
4471         struct addrconf_sysctl_table *t;
4472
4473 #define ADDRCONF_CTL_PATH_DEV   3
4474
4475         struct ctl_path addrconf_ctl_path[] = {
4476                 { .procname = "net", },
4477                 { .procname = "ipv6", },
4478                 { .procname = "conf", },
4479                 { /* to be set */ },
4480                 { },
4481         };
4482
4483
4484         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4485         if (t == NULL)
4486                 goto out;
4487
4488         for (i = 0; t->addrconf_vars[i].data; i++) {
4489                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4490                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4491                 t->addrconf_vars[i].extra2 = net;
4492         }
4493
4494         /*
4495          * Make a copy of dev_name, because '.procname' is regarded as const
4496          * by sysctl and we wouldn't want anyone to change it under our feet
4497          * (see SIOCSIFNAME).
4498          */
4499         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4500         if (!t->dev_name)
4501                 goto free;
4502
4503         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4504
4505         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4506                         t->addrconf_vars);
4507         if (t->sysctl_header == NULL)
4508                 goto free_procname;
4509
4510         p->sysctl = t;
4511         return 0;
4512
4513 free_procname:
4514         kfree(t->dev_name);
4515 free:
4516         kfree(t);
4517 out:
4518         return -ENOBUFS;
4519 }
4520
4521 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4522 {
4523         struct addrconf_sysctl_table *t;
4524
4525         if (p->sysctl == NULL)
4526                 return;
4527
4528         t = p->sysctl;
4529         p->sysctl = NULL;
4530         unregister_net_sysctl_table(t->sysctl_header);
4531         kfree(t->dev_name);
4532         kfree(t);
4533 }
4534
4535 static void addrconf_sysctl_register(struct inet6_dev *idev)
4536 {
4537         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4538                               &ndisc_ifinfo_sysctl_change);
4539         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4540                                         idev, &idev->cnf);
4541 }
4542
4543 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4544 {
4545         __addrconf_sysctl_unregister(&idev->cnf);
4546         neigh_sysctl_unregister(idev->nd_parms);
4547 }
4548
4549
4550 #endif
4551
4552 static int __net_init addrconf_init_net(struct net *net)
4553 {
4554         int err;
4555         struct ipv6_devconf *all, *dflt;
4556
4557         err = -ENOMEM;
4558         all = &ipv6_devconf;
4559         dflt = &ipv6_devconf_dflt;
4560
4561         if (!net_eq(net, &init_net)) {
4562                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4563                 if (all == NULL)
4564                         goto err_alloc_all;
4565
4566                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4567                 if (dflt == NULL)
4568                         goto err_alloc_dflt;
4569         } else {
4570                 /* these will be inherited by all namespaces */
4571                 dflt->autoconf = ipv6_defaults.autoconf;
4572                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4573         }
4574
4575         net->ipv6.devconf_all = all;
4576         net->ipv6.devconf_dflt = dflt;
4577
4578 #ifdef CONFIG_SYSCTL
4579         err = __addrconf_sysctl_register(net, "all", NULL, all);
4580         if (err < 0)
4581                 goto err_reg_all;
4582
4583         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4584         if (err < 0)
4585                 goto err_reg_dflt;
4586 #endif
4587         return 0;
4588
4589 #ifdef CONFIG_SYSCTL
4590 err_reg_dflt:
4591         __addrconf_sysctl_unregister(all);
4592 err_reg_all:
4593         kfree(dflt);
4594 #endif
4595 err_alloc_dflt:
4596         kfree(all);
4597 err_alloc_all:
4598         return err;
4599 }
4600
4601 static void __net_exit addrconf_exit_net(struct net *net)
4602 {
4603 #ifdef CONFIG_SYSCTL
4604         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4605         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4606 #endif
4607         if (!net_eq(net, &init_net)) {
4608                 kfree(net->ipv6.devconf_dflt);
4609                 kfree(net->ipv6.devconf_all);
4610         }
4611 }
4612
4613 static struct pernet_operations addrconf_ops = {
4614         .init = addrconf_init_net,
4615         .exit = addrconf_exit_net,
4616 };
4617
4618 /*
4619  *      Device notifier
4620  */
4621
4622 int register_inet6addr_notifier(struct notifier_block *nb)
4623 {
4624         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4625 }
4626 EXPORT_SYMBOL(register_inet6addr_notifier);
4627
4628 int unregister_inet6addr_notifier(struct notifier_block *nb)
4629 {
4630         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4631 }
4632 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4633
4634 static struct rtnl_af_ops inet6_ops = {
4635         .family           = AF_INET6,
4636         .fill_link_af     = inet6_fill_link_af,
4637         .get_link_af_size = inet6_get_link_af_size,
4638 };
4639
4640 /*
4641  *      Init / cleanup code
4642  */
4643
4644 int __init addrconf_init(void)
4645 {
4646         int i, err;
4647
4648         err = ipv6_addr_label_init();
4649         if (err < 0) {
4650                 printk(KERN_CRIT "IPv6 Addrconf:"
4651                        " cannot initialize default policy table: %d.\n", err);
4652                 goto out;
4653         }
4654
4655         err = register_pernet_subsys(&addrconf_ops);
4656         if (err < 0)
4657                 goto out_addrlabel;
4658
4659         /* The addrconf netdev notifier requires that loopback_dev
4660          * has it's ipv6 private information allocated and setup
4661          * before it can bring up and give link-local addresses
4662          * to other devices which are up.
4663          *
4664          * Unfortunately, loopback_dev is not necessarily the first
4665          * entry in the global dev_base list of net devices.  In fact,
4666          * it is likely to be the very last entry on that list.
4667          * So this causes the notifier registry below to try and
4668          * give link-local addresses to all devices besides loopback_dev
4669          * first, then loopback_dev, which cases all the non-loopback_dev
4670          * devices to fail to get a link-local address.
4671          *
4672          * So, as a temporary fix, allocate the ipv6 structure for
4673          * loopback_dev first by hand.
4674          * Longer term, all of the dependencies ipv6 has upon the loopback
4675          * device and it being up should be removed.
4676          */
4677         rtnl_lock();
4678         if (!ipv6_add_dev(init_net.loopback_dev))
4679                 err = -ENOMEM;
4680         rtnl_unlock();
4681         if (err)
4682                 goto errlo;
4683
4684         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4685                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4686
4687         register_netdevice_notifier(&ipv6_dev_notf);
4688
4689         addrconf_verify(0);
4690
4691         err = rtnl_af_register(&inet6_ops);
4692         if (err < 0)
4693                 goto errout_af;
4694
4695         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4696         if (err < 0)
4697                 goto errout;
4698
4699         /* Only the first call to __rtnl_register can fail */
4700         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4701         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4702         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4703         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4704         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4705
4706         ipv6_addr_label_rtnl_register();
4707
4708         return 0;
4709 errout:
4710         rtnl_af_unregister(&inet6_ops);
4711 errout_af:
4712         unregister_netdevice_notifier(&ipv6_dev_notf);
4713 errlo:
4714         unregister_pernet_subsys(&addrconf_ops);
4715 out_addrlabel:
4716         ipv6_addr_label_cleanup();
4717 out:
4718         return err;
4719 }
4720
4721 void addrconf_cleanup(void)
4722 {
4723         struct net_device *dev;
4724         int i;
4725
4726         unregister_netdevice_notifier(&ipv6_dev_notf);
4727         unregister_pernet_subsys(&addrconf_ops);
4728         ipv6_addr_label_cleanup();
4729
4730         rtnl_lock();
4731
4732         __rtnl_af_unregister(&inet6_ops);
4733
4734         /* clean dev list */
4735         for_each_netdev(&init_net, dev) {
4736                 if (__in6_dev_get(dev) == NULL)
4737                         continue;
4738                 addrconf_ifdown(dev, 1);
4739         }
4740         addrconf_ifdown(init_net.loopback_dev, 2);
4741
4742         /*
4743          *      Check hash table.
4744          */
4745         spin_lock_bh(&addrconf_hash_lock);
4746         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4747                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4748         spin_unlock_bh(&addrconf_hash_lock);
4749
4750         del_timer(&addr_chk_timer);
4751         rtnl_unlock();
4752 }