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 <asm/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 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #ifdef CONFIG_SYSCTL
 104 static void addrconf_sysctl_register(struct inet6_dev *idev);
 105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 106 #else
 107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 108 {
 109 }
 110
 111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 112 {
 113 }
 114 #endif
 115
 116 #ifdef CONFIG_IPV6_PRIVACY
 117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 119 static void ipv6_regen_rndid(unsigned long data);
 120
 121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 122 #endif
 123
 124 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 125 static int ipv6_count_addresses(struct inet6_dev *idev);
 126
 127 /*
 128  *      Configured unicast address hash table
 129  */
 130 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 131 static DEFINE_RWLOCK(addrconf_hash_lock);
 132
 133 static void addrconf_verify(unsigned long);
 134
 135 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 136 static DEFINE_SPINLOCK(addrconf_verify_lock);
 137
 138 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 139 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 140
 141 static void addrconf_bonding_change(struct net_device *dev,
 142                                     unsigned long event);
 143 static int addrconf_ifdown(struct net_device *dev, int how);
 144
 145 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 146 static void addrconf_dad_timer(unsigned long data);
 147 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 148 static void addrconf_dad_run(struct inet6_dev *idev);
 149 static void addrconf_rs_timer(unsigned long data);
 150 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 152
 153 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 154                                 struct prefix_info *pinfo);
 155 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 156                               struct net_device *dev);
 157
 158 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 159
 160 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 161         .forwarding             = 0,
 162         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 163         .mtu6                   = IPV6_MIN_MTU,
 164         .accept_ra              = 1,
 165         .accept_redirects       = 1,
 166         .autoconf               = 1,
 167         .force_mld_version      = 0,
 168         .dad_transmits          = 1,
 169         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 170         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 171         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 172 #ifdef CONFIG_IPV6_PRIVACY
 173         .use_tempaddr           = 0,
 174         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 175         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 176         .regen_max_retry        = REGEN_MAX_RETRY,
 177         .max_desync_factor      = MAX_DESYNC_FACTOR,
 178 #endif
 179         .max_addresses          = IPV6_MAX_ADDRESSES,
 180         .accept_ra_defrtr       = 1,
 181         .accept_ra_pinfo        = 1,
 182 #ifdef CONFIG_IPV6_ROUTER_PREF
 183         .accept_ra_rtr_pref     = 1,
 184         .rtr_probe_interval     = 60 * HZ,
 185 #ifdef CONFIG_IPV6_ROUTE_INFO
 186         .accept_ra_rt_info_max_plen = 0,
 187 #endif
 188 #endif
 189         .proxy_ndp              = 0,
 190         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 191         .disable_ipv6           = 0,
 192         .accept_dad             = 1,
 193 };
 194
 195 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 196         .forwarding             = 0,
 197         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 198         .mtu6                   = IPV6_MIN_MTU,
 199         .accept_ra              = 1,
 200         .accept_redirects       = 1,
 201         .autoconf               = 1,
 202         .dad_transmits          = 1,
 203         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 204         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 205         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 206 #ifdef CONFIG_IPV6_PRIVACY
 207         .use_tempaddr           = 0,
 208         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 209         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 210         .regen_max_retry        = REGEN_MAX_RETRY,
 211         .max_desync_factor      = MAX_DESYNC_FACTOR,
 212 #endif
 213         .max_addresses          = IPV6_MAX_ADDRESSES,
 214         .accept_ra_defrtr       = 1,
 215         .accept_ra_pinfo        = 1,
 216 #ifdef CONFIG_IPV6_ROUTER_PREF
 217         .accept_ra_rtr_pref     = 1,
 218         .rtr_probe_interval     = 60 * HZ,
 219 #ifdef CONFIG_IPV6_ROUTE_INFO
 220         .accept_ra_rt_info_max_plen = 0,
 221 #endif
 222 #endif
 223         .proxy_ndp              = 0,
 224         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 225         .disable_ipv6           = 0,
 226         .accept_dad             = 1,
 227 };
 228
 229 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 230 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 231 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 232 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 233 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 234
 235 /* Check if a valid qdisc is available */
 236 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 237 {
 238         return !qdisc_tx_is_noop(dev);
 239 }
 240
 241 /* Check if a route is valid prefix route */
 242 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 243 {
 244         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 245 }
 246
 247 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 248 {
 249         if (del_timer(&ifp->timer))
 250                 __in6_ifa_put(ifp);
 251 }
 252
 253 enum addrconf_timer_t
 254 {
 255         AC_NONE,
 256         AC_DAD,
 257         AC_RS,
 258 };
 259
 260 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 261                                enum addrconf_timer_t what,
 262                                unsigned long when)
 263 {
 264         if (!del_timer(&ifp->timer))
 265                 in6_ifa_hold(ifp);
 266
 267         switch (what) {
 268         case AC_DAD:
 269                 ifp->timer.function = addrconf_dad_timer;
 270                 break;
 271         case AC_RS:
 272                 ifp->timer.function = addrconf_rs_timer;
 273                 break;
 274         default:;
 275         }
 276         ifp->timer.expires = jiffies + when;
 277         add_timer(&ifp->timer);
 278 }
 279
 280 static int snmp6_alloc_dev(struct inet6_dev *idev)
 281 {
 282         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 283                           sizeof(struct ipstats_mib)) < 0)
 284                 goto err_ip;
 285         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 286                           sizeof(struct icmpv6_mib)) < 0)
 287                 goto err_icmp;
 288         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 289                           sizeof(struct icmpv6msg_mib)) < 0)
 290                 goto err_icmpmsg;
 291
 292         return 0;
 293
 294 err_icmpmsg:
 295         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 296 err_icmp:
 297         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 298 err_ip:
 299         return -ENOMEM;
 300 }
 301
 302 static void snmp6_free_dev(struct inet6_dev *idev)
 303 {
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 305         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 }
 308
 309 /* Nobody refers to this device, we may destroy it. */
 310
 311 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 312 {
 313         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 314         kfree(idev);
 315 }
 316
 317 void in6_dev_finish_destroy(struct inet6_dev *idev)
 318 {
 319         struct net_device *dev = idev->dev;
 320
 321         WARN_ON(idev->addr_list != NULL);
 322         WARN_ON(idev->mc_list != NULL);
 323
 324 #ifdef NET_REFCNT_DEBUG
 325         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 326 #endif
 327         dev_put(dev);
 328         if (!idev->dead) {
 329                 printk("Freeing alive inet6 device %p\n", idev);
 330                 return;
 331         }
 332         snmp6_free_dev(idev);
 333         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 334 }
 335
 336 EXPORT_SYMBOL(in6_dev_finish_destroy);
 337
 338 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 339 {
 340         struct inet6_dev *ndev;
 341
 342         ASSERT_RTNL();
 343
 344         if (dev->mtu < IPV6_MIN_MTU)
 345                 return NULL;
 346
 347         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 348
 349         if (ndev == NULL)
 350                 return NULL;
 351
 352         rwlock_init(&ndev->lock);
 353         ndev->dev = dev;
 354         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 355         ndev->cnf.mtu6 = dev->mtu;
 356         ndev->cnf.sysctl = NULL;
 357         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 358         if (ndev->nd_parms == NULL) {
 359                 kfree(ndev);
 360                 return NULL;
 361         }
 362         if (ndev->cnf.forwarding)
 363                 dev_disable_lro(dev);
 364         /* We refer to the device */
 365         dev_hold(dev);
 366
 367         if (snmp6_alloc_dev(ndev) < 0) {
 368                 ADBG((KERN_WARNING
 369                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 370                         __func__, dev->name));
 371                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 372                 ndev->dead = 1;
 373                 in6_dev_finish_destroy(ndev);
 374                 return NULL;
 375         }
 376
 377         if (snmp6_register_dev(ndev) < 0) {
 378                 ADBG((KERN_WARNING
 379                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 380                         __func__, dev->name));
 381                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 382                 ndev->dead = 1;
 383                 in6_dev_finish_destroy(ndev);
 384                 return NULL;
 385         }
 386
 387         /* One reference from device.  We must do this before
 388          * we invoke __ipv6_regen_rndid().
 389          */
 390         in6_dev_hold(ndev);
 391
 392         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 393                 ndev->cnf.accept_dad = -1;
 394
 395 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 396         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 397                 printk(KERN_INFO
 398                        "%s: Disabled Multicast RS\n",
 399                        dev->name);
 400                 ndev->cnf.rtr_solicits = 0;
 401         }
 402 #endif
 403
 404 #ifdef CONFIG_IPV6_PRIVACY
 405         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 406         if ((dev->flags&IFF_LOOPBACK) ||
 407             dev->type == ARPHRD_TUNNEL ||
 408             dev->type == ARPHRD_TUNNEL6 ||
 409             dev->type == ARPHRD_SIT ||
 410             dev->type == ARPHRD_NONE) {
 411                 printk(KERN_INFO
 412                        "%s: Disabled Privacy Extensions\n",
 413                        dev->name);
 414                 ndev->cnf.use_tempaddr = -1;
 415         } else {
 416                 in6_dev_hold(ndev);
 417                 ipv6_regen_rndid((unsigned long) ndev);
 418         }
 419 #endif
 420
 421         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 422                 ndev->if_flags |= IF_READY;
 423
 424         ipv6_mc_init_dev(ndev);
 425         ndev->tstamp = jiffies;
 426         addrconf_sysctl_register(ndev);
 427         /* protected by rtnl_lock */
 428         rcu_assign_pointer(dev->ip6_ptr, ndev);
 429
 430         /* Join all-node multicast group */
 431         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 432
 433         return ndev;
 434 }
 435
 436 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 437 {
 438         struct inet6_dev *idev;
 439
 440         ASSERT_RTNL();
 441
 442         if ((idev = __in6_dev_get(dev)) == NULL) {
 443                 if ((idev = ipv6_add_dev(dev)) == NULL)
 444                         return NULL;
 445         }
 446
 447         if (dev->flags&IFF_UP)
 448                 ipv6_mc_up(idev);
 449         return idev;
 450 }
 451
 452 #ifdef CONFIG_SYSCTL
 453 static void dev_forward_change(struct inet6_dev *idev)
 454 {
 455         struct net_device *dev;
 456         struct inet6_ifaddr *ifa;
 457
 458         if (!idev)
 459                 return;
 460         dev = idev->dev;
 461         if (idev->cnf.forwarding)
 462                 dev_disable_lro(dev);
 463         if (dev && (dev->flags & IFF_MULTICAST)) {
 464                 if (idev->cnf.forwarding)
 465                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 466                 else
 467                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 468         }
 469         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 470                 if (ifa->flags&IFA_F_TENTATIVE)
 471                         continue;
 472                 if (idev->cnf.forwarding)
 473                         addrconf_join_anycast(ifa);
 474                 else
 475                         addrconf_leave_anycast(ifa);
 476         }
 477 }
 478
 479
 480 static void addrconf_forward_change(struct net *net, __s32 newf)
 481 {
 482         struct net_device *dev;
 483         struct inet6_dev *idev;
 484
 485         rcu_read_lock();
 486         for_each_netdev_rcu(net, dev) {
 487                 idev = __in6_dev_get(dev);
 488                 if (idev) {
 489                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 490                         idev->cnf.forwarding = newf;
 491                         if (changed)
 492                                 dev_forward_change(idev);
 493                 }
 494         }
 495         rcu_read_unlock();
 496 }
 497
 498 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 499 {
 500         struct net *net;
 501
 502         net = (struct net *)table->extra2;
 503         if (p == &net->ipv6.devconf_dflt->forwarding)
 504                 return 0;
 505
 506         if (!rtnl_trylock()) {
 507                 /* Restore the original values before restarting */
 508                 *p = old;
 509                 return restart_syscall();
 510         }
 511
 512         if (p == &net->ipv6.devconf_all->forwarding) {
 513                 __s32 newf = net->ipv6.devconf_all->forwarding;
 514                 net->ipv6.devconf_dflt->forwarding = newf;
 515                 addrconf_forward_change(net, newf);
 516         } else if ((!*p) ^ (!old))
 517                 dev_forward_change((struct inet6_dev *)table->extra1);
 518         rtnl_unlock();
 519
 520         if (*p)
 521                 rt6_purge_dflt_routers(net);
 522         return 1;
 523 }
 524 #endif
 525
 526 /* Nobody refers to this ifaddr, destroy it */
 527
 528 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 529 {
 530         WARN_ON(ifp->if_next != NULL);
 531         WARN_ON(ifp->lst_next != NULL);
 532
 533 #ifdef NET_REFCNT_DEBUG
 534         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 535 #endif
 536
 537         in6_dev_put(ifp->idev);
 538
 539         if (del_timer(&ifp->timer))
 540                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 541
 542         if (!ifp->dead) {
 543                 printk("Freeing alive inet6 address %p\n", ifp);
 544                 return;
 545         }
 546         dst_release(&ifp->rt->u.dst);
 547
 548         kfree(ifp);
 549 }
 550
 551 static void
 552 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 553 {
 554         struct inet6_ifaddr *ifa, **ifap;
 555         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 556
 557         /*
 558          * Each device address list is sorted in order of scope -
 559          * global before linklocal.
 560          */
 561         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 562              ifap = &ifa->if_next) {
 563                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 564                         break;
 565         }
 566
 567         ifp->if_next = *ifap;
 568         *ifap = ifp;
 569 }
 570
 571 /*
 572  *      Hash function taken from net_alias.c
 573  */
 574 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 575 {
 576         __u32 word;
 577
 578         /*
 579          * We perform the hash function over the last 64 bits of the address
 580          * This will include the IEEE address token on links that support it.
 581          */
 582
 583         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 584         word ^= (word >> 16);
 585         word ^= (word >> 8);
 586
 587         return ((word ^ (word >> 4)) & 0x0f);
 588 }
 589
 590 /* On success it returns ifp with increased reference count */
 591
 592 static struct inet6_ifaddr *
 593 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 594               int scope, u32 flags)
 595 {
 596         struct inet6_ifaddr *ifa = NULL;
 597         struct rt6_info *rt;
 598         int hash;
 599         int err = 0;
 600         int addr_type = ipv6_addr_type(addr);
 601
 602         if (addr_type == IPV6_ADDR_ANY ||
 603             addr_type & IPV6_ADDR_MULTICAST ||
 604             (!(idev->dev->flags & IFF_LOOPBACK) &&
 605              addr_type & IPV6_ADDR_LOOPBACK))
 606                 return ERR_PTR(-EADDRNOTAVAIL);
 607
 608         rcu_read_lock_bh();
 609         if (idev->dead) {
 610                 err = -ENODEV;                  /*XXX*/
 611                 goto out2;
 612         }
 613
 614         if (idev->cnf.disable_ipv6) {
 615                 err = -EACCES;
 616                 goto out2;
 617         }
 618
 619         write_lock(&addrconf_hash_lock);
 620
 621         /* Ignore adding duplicate addresses on an interface */
 622         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 623                 ADBG(("ipv6_add_addr: already assigned\n"));
 624                 err = -EEXIST;
 625                 goto out;
 626         }
 627
 628         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 629
 630         if (ifa == NULL) {
 631                 ADBG(("ipv6_add_addr: malloc failed\n"));
 632                 err = -ENOBUFS;
 633                 goto out;
 634         }
 635
 636         rt = addrconf_dst_alloc(idev, addr, 0);
 637         if (IS_ERR(rt)) {
 638                 err = PTR_ERR(rt);
 639                 goto out;
 640         }
 641
 642         ipv6_addr_copy(&ifa->addr, addr);
 643
 644         spin_lock_init(&ifa->lock);
 645         init_timer(&ifa->timer);
 646         ifa->timer.data = (unsigned long) ifa;
 647         ifa->scope = scope;
 648         ifa->prefix_len = pfxlen;
 649         ifa->flags = flags | IFA_F_TENTATIVE;
 650         ifa->cstamp = ifa->tstamp = jiffies;
 651
 652         ifa->rt = rt;
 653
 654         /*
 655          * part one of RFC 4429, section 3.3
 656          * We should not configure an address as
 657          * optimistic if we do not yet know the link
 658          * layer address of our nexhop router
 659          */
 660
 661         if (rt->rt6i_nexthop == NULL)
 662                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 663
 664         ifa->idev = idev;
 665         in6_dev_hold(idev);
 666         /* For caller */
 667         in6_ifa_hold(ifa);
 668
 669         /* Add to big hash table */
 670         hash = ipv6_addr_hash(addr);
 671
 672         ifa->lst_next = inet6_addr_lst[hash];
 673         inet6_addr_lst[hash] = ifa;
 674         in6_ifa_hold(ifa);
 675         write_unlock(&addrconf_hash_lock);
 676
 677         write_lock(&idev->lock);
 678         /* Add to inet6_dev unicast addr list. */
 679         ipv6_link_dev_addr(idev, ifa);
 680
 681 #ifdef CONFIG_IPV6_PRIVACY
 682         if (ifa->flags&IFA_F_TEMPORARY) {
 683                 ifa->tmp_next = idev->tempaddr_list;
 684                 idev->tempaddr_list = ifa;
 685                 in6_ifa_hold(ifa);
 686         }
 687 #endif
 688
 689         in6_ifa_hold(ifa);
 690         write_unlock(&idev->lock);
 691 out2:
 692         rcu_read_unlock_bh();
 693
 694         if (likely(err == 0))
 695                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 696         else {
 697                 kfree(ifa);
 698                 ifa = ERR_PTR(err);
 699         }
 700
 701         return ifa;
 702 out:
 703         write_unlock(&addrconf_hash_lock);
 704         goto out2;
 705 }
 706
 707 /* This function wants to get referenced ifp and releases it before return */
 708
 709 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 710 {
 711         struct inet6_ifaddr *ifa, **ifap;
 712         struct inet6_dev *idev = ifp->idev;
 713         int hash;
 714         int deleted = 0, onlink = 0;
 715         unsigned long expires = jiffies;
 716
 717         hash = ipv6_addr_hash(&ifp->addr);
 718
 719         ifp->dead = 1;
 720
 721         write_lock_bh(&addrconf_hash_lock);
 722         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 723              ifap = &ifa->lst_next) {
 724                 if (ifa == ifp) {
 725                         *ifap = ifa->lst_next;
 726                         __in6_ifa_put(ifp);
 727                         ifa->lst_next = NULL;
 728                         break;
 729                 }
 730         }
 731         write_unlock_bh(&addrconf_hash_lock);
 732
 733         write_lock_bh(&idev->lock);
 734 #ifdef CONFIG_IPV6_PRIVACY
 735         if (ifp->flags&IFA_F_TEMPORARY) {
 736                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 737                      ifap = &ifa->tmp_next) {
 738                         if (ifa == ifp) {
 739                                 *ifap = ifa->tmp_next;
 740                                 if (ifp->ifpub) {
 741                                         in6_ifa_put(ifp->ifpub);
 742                                         ifp->ifpub = NULL;
 743                                 }
 744                                 __in6_ifa_put(ifp);
 745                                 ifa->tmp_next = NULL;
 746                                 break;
 747                         }
 748                 }
 749         }
 750 #endif
 751
 752         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 753                 if (ifa == ifp) {
 754                         *ifap = ifa->if_next;
 755                         __in6_ifa_put(ifp);
 756                         ifa->if_next = NULL;
 757                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 758                                 break;
 759                         deleted = 1;
 760                         continue;
 761                 } else if (ifp->flags & IFA_F_PERMANENT) {
 762                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 763                                               ifp->prefix_len)) {
 764                                 if (ifa->flags & IFA_F_PERMANENT) {
 765                                         onlink = 1;
 766                                         if (deleted)
 767                                                 break;
 768                                 } else {
 769                                         unsigned long lifetime;
 770
 771                                         if (!onlink)
 772                                                 onlink = -1;
 773
 774                                         spin_lock(&ifa->lock);
 775
 776                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 777                                         /*
 778                                          * Note: Because this address is
 779                                          * not permanent, lifetime <
 780                                          * LONG_MAX / HZ here.
 781                                          */
 782                                         if (time_before(expires,
 783                                                         ifa->tstamp + lifetime * HZ))
 784                                                 expires = ifa->tstamp + lifetime * HZ;
 785                                         spin_unlock(&ifa->lock);
 786                                 }
 787                         }
 788                 }
 789                 ifap = &ifa->if_next;
 790         }
 791         write_unlock_bh(&idev->lock);
 792
 793         addrconf_del_timer(ifp);
 794
 795         ipv6_ifa_notify(RTM_DELADDR, ifp);
 796
 797         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 798
 799         /*
 800          * Purge or update corresponding prefix
 801          *
 802          * 1) we don't purge prefix here if address was not permanent.
 803          *    prefix is managed by its own lifetime.
 804          * 2) if there're no addresses, delete prefix.
 805          * 3) if there're still other permanent address(es),
 806          *    corresponding prefix is still permanent.
 807          * 4) otherwise, update prefix lifetime to the
 808          *    longest valid lifetime among the corresponding
 809          *    addresses on the device.
 810          *    Note: subsequent RA will update lifetime.
 811          *
 812          * --yoshfuji
 813          */
 814         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 815                 struct in6_addr prefix;
 816                 struct rt6_info *rt;
 817                 struct net *net = dev_net(ifp->idev->dev);
 818                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 819                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 820
 821                 if (rt && addrconf_is_prefix_route(rt)) {
 822                         if (onlink == 0) {
 823                                 ip6_del_rt(rt);
 824                                 rt = NULL;
 825                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 826                                 rt->rt6i_expires = expires;
 827                                 rt->rt6i_flags |= RTF_EXPIRES;
 828                         }
 829                 }
 830                 dst_release(&rt->u.dst);
 831         }
 832
 833         in6_ifa_put(ifp);
 834 }
 835
 836 #ifdef CONFIG_IPV6_PRIVACY
 837 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 838 {
 839         struct inet6_dev *idev = ifp->idev;
 840         struct in6_addr addr, *tmpaddr;
 841         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 842         unsigned long regen_advance;
 843         int tmp_plen;
 844         int ret = 0;
 845         int max_addresses;
 846         u32 addr_flags;
 847
 848         write_lock(&idev->lock);
 849         if (ift) {
 850                 spin_lock_bh(&ift->lock);
 851                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 852                 spin_unlock_bh(&ift->lock);
 853                 tmpaddr = &addr;
 854         } else {
 855                 tmpaddr = NULL;
 856         }
 857 retry:
 858         in6_dev_hold(idev);
 859         if (idev->cnf.use_tempaddr <= 0) {
 860                 write_unlock(&idev->lock);
 861                 printk(KERN_INFO
 862                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 863                 in6_dev_put(idev);
 864                 ret = -1;
 865                 goto out;
 866         }
 867         spin_lock_bh(&ifp->lock);
 868         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 869                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 870                 spin_unlock_bh(&ifp->lock);
 871                 write_unlock(&idev->lock);
 872                 printk(KERN_WARNING
 873                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 874                 in6_dev_put(idev);
 875                 ret = -1;
 876                 goto out;
 877         }
 878         in6_ifa_hold(ifp);
 879         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 880         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 881                 spin_unlock_bh(&ifp->lock);
 882                 write_unlock(&idev->lock);
 883                 printk(KERN_WARNING
 884                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 885                 in6_ifa_put(ifp);
 886                 in6_dev_put(idev);
 887                 ret = -1;
 888                 goto out;
 889         }
 890         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 891         tmp_valid_lft = min_t(__u32,
 892                               ifp->valid_lft,
 893                               idev->cnf.temp_valid_lft);
 894         tmp_prefered_lft = min_t(__u32,
 895                                  ifp->prefered_lft,
 896                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 897         tmp_plen = ifp->prefix_len;
 898         max_addresses = idev->cnf.max_addresses;
 899         tmp_cstamp = ifp->cstamp;
 900         tmp_tstamp = ifp->tstamp;
 901         spin_unlock_bh(&ifp->lock);
 902
 903         regen_advance = idev->cnf.regen_max_retry *
 904                         idev->cnf.dad_transmits *
 905                         idev->nd_parms->retrans_time / HZ;
 906         write_unlock(&idev->lock);
 907
 908         /* A temporary address is created only if this calculated Preferred
 909          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 910          * an implementation must not create a temporary address with a zero
 911          * Preferred Lifetime.
 912          */
 913         if (tmp_prefered_lft <= regen_advance) {
 914                 in6_ifa_put(ifp);
 915                 in6_dev_put(idev);
 916                 ret = -1;
 917                 goto out;
 918         }
 919
 920         addr_flags = IFA_F_TEMPORARY;
 921         /* set in addrconf_prefix_rcv() */
 922         if (ifp->flags & IFA_F_OPTIMISTIC)
 923                 addr_flags |= IFA_F_OPTIMISTIC;
 924
 925         ift = !max_addresses ||
 926               ipv6_count_addresses(idev) < max_addresses ?
 927                 ipv6_add_addr(idev, &addr, tmp_plen,
 928                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 929                               addr_flags) : NULL;
 930         if (!ift || IS_ERR(ift)) {
 931                 in6_ifa_put(ifp);
 932                 in6_dev_put(idev);
 933                 printk(KERN_INFO
 934                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 935                 tmpaddr = &addr;
 936                 write_lock(&idev->lock);
 937                 goto retry;
 938         }
 939
 940         spin_lock_bh(&ift->lock);
 941         ift->ifpub = ifp;
 942         ift->valid_lft = tmp_valid_lft;
 943         ift->prefered_lft = tmp_prefered_lft;
 944         ift->cstamp = tmp_cstamp;
 945         ift->tstamp = tmp_tstamp;
 946         spin_unlock_bh(&ift->lock);
 947
 948         addrconf_dad_start(ift, 0);
 949         in6_ifa_put(ift);
 950         in6_dev_put(idev);
 951 out:
 952         return ret;
 953 }
 954 #endif
 955
 956 /*
 957  *      Choose an appropriate source address (RFC3484)
 958  */
 959 enum {
 960         IPV6_SADDR_RULE_INIT = 0,
 961         IPV6_SADDR_RULE_LOCAL,
 962         IPV6_SADDR_RULE_SCOPE,
 963         IPV6_SADDR_RULE_PREFERRED,
 964 #ifdef CONFIG_IPV6_MIP6
 965         IPV6_SADDR_RULE_HOA,
 966 #endif
 967         IPV6_SADDR_RULE_OIF,
 968         IPV6_SADDR_RULE_LABEL,
 969 #ifdef CONFIG_IPV6_PRIVACY
 970         IPV6_SADDR_RULE_PRIVACY,
 971 #endif
 972         IPV6_SADDR_RULE_ORCHID,
 973         IPV6_SADDR_RULE_PREFIX,
 974         IPV6_SADDR_RULE_MAX
 975 };
 976
 977 struct ipv6_saddr_score {
 978         int                     rule;
 979         int                     addr_type;
 980         struct inet6_ifaddr     *ifa;
 981         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 982         int                     scopedist;
 983         int                     matchlen;
 984 };
 985
 986 struct ipv6_saddr_dst {
 987         const struct in6_addr *addr;
 988         int ifindex;
 989         int scope;
 990         int label;
 991         unsigned int prefs;
 992 };
 993
 994 static inline int ipv6_saddr_preferred(int type)
 995 {
 996         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 997                 return 1;
 998         return 0;
 999 }
1000
1001 static int ipv6_get_saddr_eval(struct net *net,
1002                                struct ipv6_saddr_score *score,
1003                                struct ipv6_saddr_dst *dst,
1004                                int i)
1005 {
1006         int ret;
1007
1008         if (i <= score->rule) {
1009                 switch (i) {
1010                 case IPV6_SADDR_RULE_SCOPE:
1011                         ret = score->scopedist;
1012                         break;
1013                 case IPV6_SADDR_RULE_PREFIX:
1014                         ret = score->matchlen;
1015                         break;
1016                 default:
1017                         ret = !!test_bit(i, score->scorebits);
1018                 }
1019                 goto out;
1020         }
1021
1022         switch (i) {
1023         case IPV6_SADDR_RULE_INIT:
1024                 /* Rule 0: remember if hiscore is not ready yet */
1025                 ret = !!score->ifa;
1026                 break;
1027         case IPV6_SADDR_RULE_LOCAL:
1028                 /* Rule 1: Prefer same address */
1029                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030                 break;
1031         case IPV6_SADDR_RULE_SCOPE:
1032                 /* Rule 2: Prefer appropriate scope
1033                  *
1034                  *      ret
1035                  *       ^
1036                  *    -1 |  d 15
1037                  *    ---+--+-+---> scope
1038                  *       |
1039                  *       |             d is scope of the destination.
1040                  *  B-d  |  \
1041                  *       |   \      <- smaller scope is better if
1042                  *  B-15 |    \        if scope is enough for destinaion.
1043                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1044                  * d-C-1 | /
1045                  *       |/         <- greater is better
1046                  *   -C  /             if scope is not enough for destination.
1047                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1048                  *
1049                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051                  * Assume B = 0 and we get C > 29.
1052                  */
1053                 ret = __ipv6_addr_src_scope(score->addr_type);
1054                 if (ret >= dst->scope)
1055                         ret = -ret;
1056                 else
1057                         ret -= 128;     /* 30 is enough */
1058                 score->scopedist = ret;
1059                 break;
1060         case IPV6_SADDR_RULE_PREFERRED:
1061                 /* Rule 3: Avoid deprecated and optimistic addresses */
1062                 ret = ipv6_saddr_preferred(score->addr_type) ||
1063                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064                 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066         case IPV6_SADDR_RULE_HOA:
1067             {
1068                 /* Rule 4: Prefer home address */
1069                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071                 break;
1072             }
1073 #endif
1074         case IPV6_SADDR_RULE_OIF:
1075                 /* Rule 5: Prefer outgoing interface */
1076                 ret = (!dst->ifindex ||
1077                        dst->ifindex == score->ifa->idev->dev->ifindex);
1078                 break;
1079         case IPV6_SADDR_RULE_LABEL:
1080                 /* Rule 6: Prefer matching label */
1081                 ret = ipv6_addr_label(net,
1082                                       &score->ifa->addr, score->addr_type,
1083                                       score->ifa->idev->dev->ifindex) == dst->label;
1084                 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086         case IPV6_SADDR_RULE_PRIVACY:
1087             {
1088                 /* Rule 7: Prefer public address
1089                  * Note: prefer temprary address if use_tempaddr >= 2
1090                  */
1091                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1094                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095                 break;
1096             }
1097 #endif
1098         case IPV6_SADDR_RULE_ORCHID:
1099                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100                  *          non-ORCHID vs non-ORCHID
1101                  */
1102                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103                         ipv6_addr_orchid(dst->addr));
1104                 break;
1105         case IPV6_SADDR_RULE_PREFIX:
1106                 /* Rule 8: Use longest matching prefix */
1107                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108                                                        dst->addr);
1109                 break;
1110         default:
1111                 ret = 0;
1112         }
1113
1114         if (ret)
1115                 __set_bit(i, score->scorebits);
1116         score->rule = i;
1117 out:
1118         return ret;
1119 }
1120
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122                        const struct in6_addr *daddr, unsigned int prefs,
1123                        struct in6_addr *saddr)
1124 {
1125         struct ipv6_saddr_score scores[2],
1126                                 *score = &scores[0], *hiscore = &scores[1];
1127         struct ipv6_saddr_dst dst;
1128         struct net_device *dev;
1129         int dst_type;
1130
1131         dst_type = __ipv6_addr_type(daddr);
1132         dst.addr = daddr;
1133         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134         dst.scope = __ipv6_addr_src_scope(dst_type);
1135         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136         dst.prefs = prefs;
1137
1138         hiscore->rule = -1;
1139         hiscore->ifa = NULL;
1140
1141         rcu_read_lock();
1142
1143         for_each_netdev_rcu(net, dev) {
1144                 struct inet6_dev *idev;
1145
1146                 /* Candidate Source Address (section 4)
1147                  *  - multicast and link-local destination address,
1148                  *    the set of candidate source address MUST only
1149                  *    include addresses assigned to interfaces
1150                  *    belonging to the same link as the outgoing
1151                  *    interface.
1152                  * (- For site-local destination addresses, the
1153                  *    set of candidate source addresses MUST only
1154                  *    include addresses assigned to interfaces
1155                  *    belonging to the same site as the outgoing
1156                  *    interface.)
1157                  */
1158                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160                     dst.ifindex && dev->ifindex != dst.ifindex)
1161                         continue;
1162
1163                 idev = __in6_dev_get(dev);
1164                 if (!idev)
1165                         continue;
1166
1167                 read_lock_bh(&idev->lock);
1168                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1169                         int i;
1170
1171                         /*
1172                          * - Tentative Address (RFC2462 section 5.4)
1173                          *  - A tentative address is not considered
1174                          *    "assigned to an interface" in the traditional
1175                          *    sense, unless it is also flagged as optimistic.
1176                          * - Candidate Source Address (section 4)
1177                          *  - In any case, anycast addresses, multicast
1178                          *    addresses, and the unspecified address MUST
1179                          *    NOT be included in a candidate set.
1180                          */
1181                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183                                 continue;
1184
1185                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186
1187                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1189                                 LIMIT_NETDEBUG(KERN_DEBUG
1190                                                "ADDRCONF: unspecified / multicast address "
1191                                                "assigned as unicast address on %s",
1192                                                dev->name);
1193                                 continue;
1194                         }
1195
1196                         score->rule = -1;
1197                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198
1199                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200                                 int minihiscore, miniscore;
1201
1202                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204
1205                                 if (minihiscore > miniscore) {
1206                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1207                                             score->scopedist > 0) {
1208                                                 /*
1209                                                  * special case:
1210                                                  * each remaining entry
1211                                                  * has too small (not enough)
1212                                                  * scope, because ifa entries
1213                                                  * are sorted by their scope
1214                                                  * values.
1215                                                  */
1216                                                 goto try_nextdev;
1217                                         }
1218                                         break;
1219                                 } else if (minihiscore < miniscore) {
1220                                         if (hiscore->ifa)
1221                                                 in6_ifa_put(hiscore->ifa);
1222
1223                                         in6_ifa_hold(score->ifa);
1224
1225                                         swap(hiscore, score);
1226
1227                                         /* restore our iterator */
1228                                         score->ifa = hiscore->ifa;
1229
1230                                         break;
1231                                 }
1232                         }
1233                 }
1234 try_nextdev:
1235                 read_unlock_bh(&idev->lock);
1236         }
1237         rcu_read_unlock();
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         if ((idev = __in6_dev_get(dev)) != NULL) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr * ifp;
1289         u8 hash = ipv6_addr_hash(addr);
1290
1291         read_lock_bh(&addrconf_hash_lock);
1292         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE)) {
1297                         if (dev == NULL || ifp->idev->dev == dev ||
1298                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299                                 break;
1300                 }
1301         }
1302         read_unlock_bh(&addrconf_hash_lock);
1303         return ifp != NULL;
1304 }
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1306
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                        struct net_device *dev)
1310 {
1311         struct inet6_ifaddr * ifp;
1312         u8 hash = ipv6_addr_hash(addr);
1313
1314         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315                 if (!net_eq(dev_net(ifp->idev->dev), net))
1316                         continue;
1317                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318                         if (dev == NULL || ifp->idev->dev == dev)
1319                                 break;
1320                 }
1321         }
1322         return ifp != NULL;
1323 }
1324
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 {
1327         struct inet6_dev *idev;
1328         struct inet6_ifaddr *ifa;
1329         int     onlink;
1330
1331         onlink = 0;
1332         rcu_read_lock();
1333         idev = __in6_dev_get(dev);
1334         if (idev) {
1335                 read_lock_bh(&idev->lock);
1336                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338                                                    ifa->prefix_len);
1339                         if (onlink)
1340                                 break;
1341                 }
1342                 read_unlock_bh(&idev->lock);
1343         }
1344         rcu_read_unlock();
1345         return onlink;
1346 }
1347
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1349
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351                                      struct net_device *dev, int strict)
1352 {
1353         struct inet6_ifaddr * ifp;
1354         u8 hash = ipv6_addr_hash(addr);
1355
1356         read_lock_bh(&addrconf_hash_lock);
1357         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358                 if (!net_eq(dev_net(ifp->idev->dev), net))
1359                         continue;
1360                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361                         if (dev == NULL || ifp->idev->dev == dev ||
1362                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363                                 in6_ifa_hold(ifp);
1364                                 break;
1365                         }
1366                 }
1367         }
1368         read_unlock_bh(&addrconf_hash_lock);
1369
1370         return ifp;
1371 }
1372
1373 /* Gets referenced address, destroys ifaddr */
1374
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1376 {
1377         if (ifp->flags&IFA_F_PERMANENT) {
1378                 spin_lock_bh(&ifp->lock);
1379                 addrconf_del_timer(ifp);
1380                 ifp->flags |= IFA_F_TENTATIVE;
1381                 if (dad_failed)
1382                         ifp->flags |= IFA_F_DADFAILED;
1383                 spin_unlock_bh(&ifp->lock);
1384                 if (dad_failed)
1385                         ipv6_ifa_notify(0, ifp);
1386                 in6_ifa_put(ifp);
1387 #ifdef CONFIG_IPV6_PRIVACY
1388         } else if (ifp->flags&IFA_F_TEMPORARY) {
1389                 struct inet6_ifaddr *ifpub;
1390                 spin_lock_bh(&ifp->lock);
1391                 ifpub = ifp->ifpub;
1392                 if (ifpub) {
1393                         in6_ifa_hold(ifpub);
1394                         spin_unlock_bh(&ifp->lock);
1395                         ipv6_create_tempaddr(ifpub, ifp);
1396                         in6_ifa_put(ifpub);
1397                 } else {
1398                         spin_unlock_bh(&ifp->lock);
1399                 }
1400                 ipv6_del_addr(ifp);
1401 #endif
1402         } else
1403                 ipv6_del_addr(ifp);
1404 }
1405
1406 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1407 {
1408         struct inet6_dev *idev = ifp->idev;
1409
1410         if (net_ratelimit())
1411                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1412                         ifp->idev->dev->name, &ifp->addr);
1413
1414         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1415                 struct in6_addr addr;
1416
1417                 addr.s6_addr32[0] = htonl(0xfe800000);
1418                 addr.s6_addr32[1] = 0;
1419
1420                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1421                     ipv6_addr_equal(&ifp->addr, &addr)) {
1422                         /* DAD failed for link-local based on MAC address */
1423                         idev->cnf.disable_ipv6 = 1;
1424
1425                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1426                                 ifp->idev->dev->name);
1427                 }
1428         }
1429
1430         addrconf_dad_stop(ifp, 1);
1431 }
1432
1433 /* Join to solicited addr multicast group. */
1434
1435 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1436 {
1437         struct in6_addr maddr;
1438
1439         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1440                 return;
1441
1442         addrconf_addr_solict_mult(addr, &maddr);
1443         ipv6_dev_mc_inc(dev, &maddr);
1444 }
1445
1446 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1447 {
1448         struct in6_addr maddr;
1449
1450         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1451                 return;
1452
1453         addrconf_addr_solict_mult(addr, &maddr);
1454         __ipv6_dev_mc_dec(idev, &maddr);
1455 }
1456
1457 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1458 {
1459         struct in6_addr addr;
1460         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1461         if (ipv6_addr_any(&addr))
1462                 return;
1463         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1464 }
1465
1466 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1467 {
1468         struct in6_addr addr;
1469         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1470         if (ipv6_addr_any(&addr))
1471                 return;
1472         __ipv6_dev_ac_dec(ifp->idev, &addr);
1473 }
1474
1475 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1476 {
1477         if (dev->addr_len != ETH_ALEN)
1478                 return -1;
1479         memcpy(eui, dev->dev_addr, 3);
1480         memcpy(eui + 5, dev->dev_addr + 3, 3);
1481
1482         /*
1483          * The zSeries OSA network cards can be shared among various
1484          * OS instances, but the OSA cards have only one MAC address.
1485          * This leads to duplicate address conflicts in conjunction
1486          * with IPv6 if more than one instance uses the same card.
1487          *
1488          * The driver for these cards can deliver a unique 16-bit
1489          * identifier for each instance sharing the same card.  It is
1490          * placed instead of 0xFFFE in the interface identifier.  The
1491          * "u" bit of the interface identifier is not inverted in this
1492          * case.  Hence the resulting interface identifier has local
1493          * scope according to RFC2373.
1494          */
1495         if (dev->dev_id) {
1496                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1497                 eui[4] = dev->dev_id & 0xFF;
1498         } else {
1499                 eui[3] = 0xFF;
1500                 eui[4] = 0xFE;
1501                 eui[0] ^= 2;
1502         }
1503         return 0;
1504 }
1505
1506 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1507 {
1508         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1509         if (dev->addr_len != ARCNET_ALEN)
1510                 return -1;
1511         memset(eui, 0, 7);
1512         eui[7] = *(u8*)dev->dev_addr;
1513         return 0;
1514 }
1515
1516 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1517 {
1518         if (dev->addr_len != INFINIBAND_ALEN)
1519                 return -1;
1520         memcpy(eui, dev->dev_addr + 12, 8);
1521         eui[0] |= 2;
1522         return 0;
1523 }
1524
1525 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1526 {
1527         if (addr == 0)
1528                 return -1;
1529         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1530                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1531                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1532                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1533                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1534                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1535         eui[1] = 0;
1536         eui[2] = 0x5E;
1537         eui[3] = 0xFE;
1538         memcpy(eui + 4, &addr, 4);
1539         return 0;
1540 }
1541 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1542
1543 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1544 {
1545         if (dev->priv_flags & IFF_ISATAP)
1546                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1547         return -1;
1548 }
1549
1550 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1551 {
1552         switch (dev->type) {
1553         case ARPHRD_ETHER:
1554         case ARPHRD_FDDI:
1555         case ARPHRD_IEEE802_TR:
1556                 return addrconf_ifid_eui48(eui, dev);
1557         case ARPHRD_ARCNET:
1558                 return addrconf_ifid_arcnet(eui, dev);
1559         case ARPHRD_INFINIBAND:
1560                 return addrconf_ifid_infiniband(eui, dev);
1561         case ARPHRD_SIT:
1562                 return addrconf_ifid_sit(eui, dev);
1563         }
1564         return -1;
1565 }
1566
1567 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1568 {
1569         int err = -1;
1570         struct inet6_ifaddr *ifp;
1571
1572         read_lock_bh(&idev->lock);
1573         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1574                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1575                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1576                         err = 0;
1577                         break;
1578                 }
1579         }
1580         read_unlock_bh(&idev->lock);
1581         return err;
1582 }
1583
1584 #ifdef CONFIG_IPV6_PRIVACY
1585 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1586 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1587 {
1588 regen:
1589         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1590         idev->rndid[0] &= ~0x02;
1591
1592         /*
1593          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1594          * check if generated address is not inappropriate
1595          *
1596          *  - Reserved subnet anycast (RFC 2526)
1597          *      11111101 11....11 1xxxxxxx
1598          *  - ISATAP (RFC4214) 6.1
1599          *      00-00-5E-FE-xx-xx-xx-xx
1600          *  - value 0
1601          *  - XXX: already assigned to an address on the device
1602          */
1603         if (idev->rndid[0] == 0xfd &&
1604             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1605             (idev->rndid[7]&0x80))
1606                 goto regen;
1607         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1608                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1609                         goto regen;
1610                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1611                         goto regen;
1612         }
1613
1614         return 0;
1615 }
1616
1617 static void ipv6_regen_rndid(unsigned long data)
1618 {
1619         struct inet6_dev *idev = (struct inet6_dev *) data;
1620         unsigned long expires;
1621
1622         rcu_read_lock_bh();
1623         write_lock_bh(&idev->lock);
1624
1625         if (idev->dead)
1626                 goto out;
1627
1628         if (__ipv6_regen_rndid(idev) < 0)
1629                 goto out;
1630
1631         expires = jiffies +
1632                 idev->cnf.temp_prefered_lft * HZ -
1633                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1634         if (time_before(expires, jiffies)) {
1635                 printk(KERN_WARNING
1636                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1637                         idev->dev->name);
1638                 goto out;
1639         }
1640
1641         if (!mod_timer(&idev->regen_timer, expires))
1642                 in6_dev_hold(idev);
1643
1644 out:
1645         write_unlock_bh(&idev->lock);
1646         rcu_read_unlock_bh();
1647         in6_dev_put(idev);
1648 }
1649
1650 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1651         int ret = 0;
1652
1653         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1654                 ret = __ipv6_regen_rndid(idev);
1655         return ret;
1656 }
1657 #endif
1658
1659 /*
1660  *      Add prefix route.
1661  */
1662
1663 static void
1664 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1665                       unsigned long expires, u32 flags)
1666 {
1667         struct fib6_config cfg = {
1668                 .fc_table = RT6_TABLE_PREFIX,
1669                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1670                 .fc_ifindex = dev->ifindex,
1671                 .fc_expires = expires,
1672                 .fc_dst_len = plen,
1673                 .fc_flags = RTF_UP | flags,
1674                 .fc_nlinfo.nl_net = dev_net(dev),
1675                 .fc_protocol = RTPROT_KERNEL,
1676         };
1677
1678         ipv6_addr_copy(&cfg.fc_dst, pfx);
1679
1680         /* Prevent useless cloning on PtP SIT.
1681            This thing is done here expecting that the whole
1682            class of non-broadcast devices need not cloning.
1683          */
1684 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1685         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1686                 cfg.fc_flags |= RTF_NONEXTHOP;
1687 #endif
1688
1689         ip6_route_add(&cfg);
1690 }
1691
1692 /* Create "default" multicast route to the interface */
1693
1694 static void addrconf_add_mroute(struct net_device *dev)
1695 {
1696         struct fib6_config cfg = {
1697                 .fc_table = RT6_TABLE_LOCAL,
1698                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1699                 .fc_ifindex = dev->ifindex,
1700                 .fc_dst_len = 8,
1701                 .fc_flags = RTF_UP,
1702                 .fc_nlinfo.nl_net = dev_net(dev),
1703         };
1704
1705         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1706
1707         ip6_route_add(&cfg);
1708 }
1709
1710 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1711 static void sit_route_add(struct net_device *dev)
1712 {
1713         struct fib6_config cfg = {
1714                 .fc_table = RT6_TABLE_MAIN,
1715                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1716                 .fc_ifindex = dev->ifindex,
1717                 .fc_dst_len = 96,
1718                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1719                 .fc_nlinfo.nl_net = dev_net(dev),
1720         };
1721
1722         /* prefix length - 96 bits "::d.d.d.d" */
1723         ip6_route_add(&cfg);
1724 }
1725 #endif
1726
1727 static void addrconf_add_lroute(struct net_device *dev)
1728 {
1729         struct in6_addr addr;
1730
1731         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1732         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1733 }
1734
1735 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1736 {
1737         struct inet6_dev *idev;
1738
1739         ASSERT_RTNL();
1740
1741         if ((idev = ipv6_find_idev(dev)) == NULL)
1742                 return NULL;
1743
1744         /* Add default multicast route */
1745         addrconf_add_mroute(dev);
1746
1747         /* Add link local route */
1748         addrconf_add_lroute(dev);
1749         return idev;
1750 }
1751
1752 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1753 {
1754         struct prefix_info *pinfo;
1755         __u32 valid_lft;
1756         __u32 prefered_lft;
1757         int addr_type;
1758         struct inet6_dev *in6_dev;
1759         struct net *net = dev_net(dev);
1760
1761         pinfo = (struct prefix_info *) opt;
1762
1763         if (len < sizeof(struct prefix_info)) {
1764                 ADBG(("addrconf: prefix option too short\n"));
1765                 return;
1766         }
1767
1768         /*
1769          *      Validation checks ([ADDRCONF], page 19)
1770          */
1771
1772         addr_type = ipv6_addr_type(&pinfo->prefix);
1773
1774         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1775                 return;
1776
1777         valid_lft = ntohl(pinfo->valid);
1778         prefered_lft = ntohl(pinfo->prefered);
1779
1780         if (prefered_lft > valid_lft) {
1781                 if (net_ratelimit())
1782                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1783                 return;
1784         }
1785
1786         in6_dev = in6_dev_get(dev);
1787
1788         if (in6_dev == NULL) {
1789                 if (net_ratelimit())
1790                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1791                 return;
1792         }
1793
1794         /*
1795          *      Two things going on here:
1796          *      1) Add routes for on-link prefixes
1797          *      2) Configure prefixes with the auto flag set
1798          */
1799
1800         if (pinfo->onlink) {
1801                 struct rt6_info *rt;
1802                 unsigned long rt_expires;
1803
1804                 /* Avoid arithmetic overflow. Really, we could
1805                  * save rt_expires in seconds, likely valid_lft,
1806                  * but it would require division in fib gc, that it
1807                  * not good.
1808                  */
1809                 if (HZ > USER_HZ)
1810                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1811                 else
1812                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1813
1814                 if (addrconf_finite_timeout(rt_expires))
1815                         rt_expires *= HZ;
1816
1817                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1818                                 dev->ifindex, 1);
1819
1820                 if (rt && addrconf_is_prefix_route(rt)) {
1821                         /* Autoconf prefix route */
1822                         if (valid_lft == 0) {
1823                                 ip6_del_rt(rt);
1824                                 rt = NULL;
1825                         } else if (addrconf_finite_timeout(rt_expires)) {
1826                                 /* not infinity */
1827                                 rt->rt6i_expires = jiffies + rt_expires;
1828                                 rt->rt6i_flags |= RTF_EXPIRES;
1829                         } else {
1830                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1831                                 rt->rt6i_expires = 0;
1832                         }
1833                 } else if (valid_lft) {
1834                         clock_t expires = 0;
1835                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1836                         if (addrconf_finite_timeout(rt_expires)) {
1837                                 /* not infinity */
1838                                 flags |= RTF_EXPIRES;
1839                                 expires = jiffies_to_clock_t(rt_expires);
1840                         }
1841                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1842                                               dev, expires, flags);
1843                 }
1844                 if (rt)
1845                         dst_release(&rt->u.dst);
1846         }
1847
1848         /* Try to figure out our local address for this prefix */
1849
1850         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1851                 struct inet6_ifaddr * ifp;
1852                 struct in6_addr addr;
1853                 int create = 0, update_lft = 0;
1854
1855                 if (pinfo->prefix_len == 64) {
1856                         memcpy(&addr, &pinfo->prefix, 8);
1857                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1858                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1859                                 in6_dev_put(in6_dev);
1860                                 return;
1861                         }
1862                         goto ok;
1863                 }
1864                 if (net_ratelimit())
1865                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1866                                pinfo->prefix_len);
1867                 in6_dev_put(in6_dev);
1868                 return;
1869
1870 ok:
1871
1872                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1873
1874                 if (ifp == NULL && valid_lft) {
1875                         int max_addresses = in6_dev->cnf.max_addresses;
1876                         u32 addr_flags = 0;
1877
1878 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1879                         if (in6_dev->cnf.optimistic_dad &&
1880                             !net->ipv6.devconf_all->forwarding)
1881                                 addr_flags = IFA_F_OPTIMISTIC;
1882 #endif
1883
1884                         /* Do not allow to create too much of autoconfigured
1885                          * addresses; this would be too easy way to crash kernel.
1886                          */
1887                         if (!max_addresses ||
1888                             ipv6_count_addresses(in6_dev) < max_addresses)
1889                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1890                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1891                                                     addr_flags);
1892
1893                         if (!ifp || IS_ERR(ifp)) {
1894                                 in6_dev_put(in6_dev);
1895                                 return;
1896                         }
1897
1898                         update_lft = create = 1;
1899                         ifp->cstamp = jiffies;
1900                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1901                 }
1902
1903                 if (ifp) {
1904                         int flags;
1905                         unsigned long now;
1906 #ifdef CONFIG_IPV6_PRIVACY
1907                         struct inet6_ifaddr *ift;
1908 #endif
1909                         u32 stored_lft;
1910
1911                         /* update lifetime (RFC2462 5.5.3 e) */
1912                         spin_lock(&ifp->lock);
1913                         now = jiffies;
1914                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1915                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1916                         else
1917                                 stored_lft = 0;
1918                         if (!update_lft && stored_lft) {
1919                                 if (valid_lft > MIN_VALID_LIFETIME ||
1920                                     valid_lft > stored_lft)
1921                                         update_lft = 1;
1922                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1923                                         /* valid_lft <= stored_lft is always true */
1924                                         /*
1925                                          * RFC 4862 Section 5.5.3e:
1926                                          * "Note that the preferred lifetime of
1927                                          *  the corresponding address is always
1928                                          *  reset to the Preferred Lifetime in
1929                                          *  the received Prefix Information
1930                                          *  option, regardless of whether the
1931                                          *  valid lifetime is also reset or
1932                                          *  ignored."
1933                                          *
1934                                          *  So if the preferred lifetime in
1935                                          *  this advertisement is different
1936                                          *  than what we have stored, but the
1937                                          *  valid lifetime is invalid, just
1938                                          *  reset prefered_lft.
1939                                          *
1940                                          *  We must set the valid lifetime
1941                                          *  to the stored lifetime since we'll
1942                                          *  be updating the timestamp below,
1943                                          *  else we'll set it back to the
1944                                          *  minumum.
1945                                          */
1946                                         if (prefered_lft != ifp->prefered_lft) {
1947                                                 valid_lft = stored_lft;
1948                                                 update_lft = 1;
1949                                         }
1950                                 } else {
1951                                         valid_lft = MIN_VALID_LIFETIME;
1952                                         if (valid_lft < prefered_lft)
1953                                                 prefered_lft = valid_lft;
1954                                         update_lft = 1;
1955                                 }
1956                         }
1957
1958                         if (update_lft) {
1959                                 ifp->valid_lft = valid_lft;
1960                                 ifp->prefered_lft = prefered_lft;
1961                                 ifp->tstamp = now;
1962                                 flags = ifp->flags;
1963                                 ifp->flags &= ~IFA_F_DEPRECATED;
1964                                 spin_unlock(&ifp->lock);
1965
1966                                 if (!(flags&IFA_F_TENTATIVE))
1967                                         ipv6_ifa_notify(0, ifp);
1968                         } else
1969                                 spin_unlock(&ifp->lock);
1970
1971 #ifdef CONFIG_IPV6_PRIVACY
1972                         read_lock_bh(&in6_dev->lock);
1973                         /* update all temporary addresses in the list */
1974                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1975                                 /*
1976                                  * When adjusting the lifetimes of an existing
1977                                  * temporary address, only lower the lifetimes.
1978                                  * Implementations must not increase the
1979                                  * lifetimes of an existing temporary address
1980                                  * when processing a Prefix Information Option.
1981                                  */
1982                                 if (ifp != ift->ifpub)
1983                                         continue;
1984
1985                                 spin_lock(&ift->lock);
1986                                 flags = ift->flags;
1987                                 if (ift->valid_lft > valid_lft &&
1988                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1989                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1990                                 if (ift->prefered_lft > prefered_lft &&
1991                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1992                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1993                                 spin_unlock(&ift->lock);
1994                                 if (!(flags&IFA_F_TENTATIVE))
1995                                         ipv6_ifa_notify(0, ift);
1996                         }
1997
1998                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1999                                 /*
2000                                  * When a new public address is created as described in [ADDRCONF],
2001                                  * also create a new temporary address.
2002                                  */
2003                                 read_unlock_bh(&in6_dev->lock);
2004                                 ipv6_create_tempaddr(ifp, NULL);
2005                         } else {
2006                                 read_unlock_bh(&in6_dev->lock);
2007                         }
2008 #endif
2009                         in6_ifa_put(ifp);
2010                         addrconf_verify(0);
2011                 }
2012         }
2013         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2014         in6_dev_put(in6_dev);
2015 }
2016
2017 /*
2018  *      Set destination address.
2019  *      Special case for SIT interfaces where we create a new "virtual"
2020  *      device.
2021  */
2022 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2023 {
2024         struct in6_ifreq ireq;
2025         struct net_device *dev;
2026         int err = -EINVAL;
2027
2028         rtnl_lock();
2029
2030         err = -EFAULT;
2031         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2032                 goto err_exit;
2033
2034         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2035
2036         err = -ENODEV;
2037         if (dev == NULL)
2038                 goto err_exit;
2039
2040 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2041         if (dev->type == ARPHRD_SIT) {
2042                 const struct net_device_ops *ops = dev->netdev_ops;
2043                 struct ifreq ifr;
2044                 struct ip_tunnel_parm p;
2045
2046                 err = -EADDRNOTAVAIL;
2047                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2048                         goto err_exit;
2049
2050                 memset(&p, 0, sizeof(p));
2051                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2052                 p.iph.saddr = 0;
2053                 p.iph.version = 4;
2054                 p.iph.ihl = 5;
2055                 p.iph.protocol = IPPROTO_IPV6;
2056                 p.iph.ttl = 64;
2057                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2058
2059                 if (ops->ndo_do_ioctl) {
2060                         mm_segment_t oldfs = get_fs();
2061
2062                         set_fs(KERNEL_DS);
2063                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2064                         set_fs(oldfs);
2065                 } else
2066                         err = -EOPNOTSUPP;
2067
2068                 if (err == 0) {
2069                         err = -ENOBUFS;
2070                         dev = __dev_get_by_name(net, p.name);
2071                         if (!dev)
2072                                 goto err_exit;
2073                         err = dev_open(dev);
2074                 }
2075         }
2076 #endif
2077
2078 err_exit:
2079         rtnl_unlock();
2080         return err;
2081 }
2082
2083 /*
2084  *      Manual configuration of address on an interface
2085  */
2086 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2087                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2088                           __u32 valid_lft)
2089 {
2090         struct inet6_ifaddr *ifp;
2091         struct inet6_dev *idev;
2092         struct net_device *dev;
2093         int scope;
2094         u32 flags;
2095         clock_t expires;
2096         unsigned long timeout;
2097
2098         ASSERT_RTNL();
2099
2100         if (plen > 128)
2101                 return -EINVAL;
2102
2103         /* check the lifetime */
2104         if (!valid_lft || prefered_lft > valid_lft)
2105                 return -EINVAL;
2106
2107         dev = __dev_get_by_index(net, ifindex);
2108         if (!dev)
2109                 return -ENODEV;
2110
2111         if ((idev = addrconf_add_dev(dev)) == NULL)
2112                 return -ENOBUFS;
2113
2114         scope = ipv6_addr_scope(pfx);
2115
2116         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2117         if (addrconf_finite_timeout(timeout)) {
2118                 expires = jiffies_to_clock_t(timeout * HZ);
2119                 valid_lft = timeout;
2120                 flags = RTF_EXPIRES;
2121         } else {
2122                 expires = 0;
2123                 flags = 0;
2124                 ifa_flags |= IFA_F_PERMANENT;
2125         }
2126
2127         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2128         if (addrconf_finite_timeout(timeout)) {
2129                 if (timeout == 0)
2130                         ifa_flags |= IFA_F_DEPRECATED;
2131                 prefered_lft = timeout;
2132         }
2133
2134         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2135
2136         if (!IS_ERR(ifp)) {
2137                 spin_lock_bh(&ifp->lock);
2138                 ifp->valid_lft = valid_lft;
2139                 ifp->prefered_lft = prefered_lft;
2140                 ifp->tstamp = jiffies;
2141                 spin_unlock_bh(&ifp->lock);
2142
2143                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2144                                       expires, flags);
2145                 /*
2146                  * Note that section 3.1 of RFC 4429 indicates
2147                  * that the Optimistic flag should not be set for
2148                  * manually configured addresses
2149                  */
2150                 addrconf_dad_start(ifp, 0);
2151                 in6_ifa_put(ifp);
2152                 addrconf_verify(0);
2153                 return 0;
2154         }
2155
2156         return PTR_ERR(ifp);
2157 }
2158
2159 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2160                           unsigned int plen)
2161 {
2162         struct inet6_ifaddr *ifp;
2163         struct inet6_dev *idev;
2164         struct net_device *dev;
2165
2166         if (plen > 128)
2167                 return -EINVAL;
2168
2169         dev = __dev_get_by_index(net, ifindex);
2170         if (!dev)
2171                 return -ENODEV;
2172
2173         if ((idev = __in6_dev_get(dev)) == NULL)
2174                 return -ENXIO;
2175
2176         read_lock_bh(&idev->lock);
2177         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2178                 if (ifp->prefix_len == plen &&
2179                     ipv6_addr_equal(pfx, &ifp->addr)) {
2180                         in6_ifa_hold(ifp);
2181                         read_unlock_bh(&idev->lock);
2182
2183                         ipv6_del_addr(ifp);
2184
2185                         /* If the last address is deleted administratively,
2186                            disable IPv6 on this interface.
2187                          */
2188                         if (idev->addr_list == NULL)
2189                                 addrconf_ifdown(idev->dev, 1);
2190                         return 0;
2191                 }
2192         }
2193         read_unlock_bh(&idev->lock);
2194         return -EADDRNOTAVAIL;
2195 }
2196
2197
2198 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2199 {
2200         struct in6_ifreq ireq;
2201         int err;
2202
2203         if (!capable(CAP_NET_ADMIN))
2204                 return -EPERM;
2205
2206         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2207                 return -EFAULT;
2208
2209         rtnl_lock();
2210         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2211                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2212                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2213         rtnl_unlock();
2214         return err;
2215 }
2216
2217 int addrconf_del_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_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2230                              ireq.ifr6_prefixlen);
2231         rtnl_unlock();
2232         return err;
2233 }
2234
2235 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2236                      int plen, int scope)
2237 {
2238         struct inet6_ifaddr *ifp;
2239
2240         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2241         if (!IS_ERR(ifp)) {
2242                 spin_lock_bh(&ifp->lock);
2243                 ifp->flags &= ~IFA_F_TENTATIVE;
2244                 spin_unlock_bh(&ifp->lock);
2245                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2246                 in6_ifa_put(ifp);
2247         }
2248 }
2249
2250 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2251 static void sit_add_v4_addrs(struct inet6_dev *idev)
2252 {
2253         struct in6_addr addr;
2254         struct net_device *dev;
2255         struct net *net = dev_net(idev->dev);
2256         int scope;
2257
2258         ASSERT_RTNL();
2259
2260         memset(&addr, 0, sizeof(struct in6_addr));
2261         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2262
2263         if (idev->dev->flags&IFF_POINTOPOINT) {
2264                 addr.s6_addr32[0] = htonl(0xfe800000);
2265                 scope = IFA_LINK;
2266         } else {
2267                 scope = IPV6_ADDR_COMPATv4;
2268         }
2269
2270         if (addr.s6_addr32[3]) {
2271                 add_addr(idev, &addr, 128, scope);
2272                 return;
2273         }
2274
2275         for_each_netdev(net, dev) {
2276                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2277                 if (in_dev && (dev->flags & IFF_UP)) {
2278                         struct in_ifaddr * ifa;
2279
2280                         int flag = scope;
2281
2282                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2283                                 int plen;
2284
2285                                 addr.s6_addr32[3] = ifa->ifa_local;
2286
2287                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2288                                         continue;
2289                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2290                                         if (idev->dev->flags&IFF_POINTOPOINT)
2291                                                 continue;
2292                                         flag |= IFA_HOST;
2293                                 }
2294                                 if (idev->dev->flags&IFF_POINTOPOINT)
2295                                         plen = 64;
2296                                 else
2297                                         plen = 96;
2298
2299                                 add_addr(idev, &addr, plen, flag);
2300                         }
2301                 }
2302         }
2303 }
2304 #endif
2305
2306 static void init_loopback(struct net_device *dev)
2307 {
2308         struct inet6_dev  *idev;
2309
2310         /* ::1 */
2311
2312         ASSERT_RTNL();
2313
2314         if ((idev = ipv6_find_idev(dev)) == NULL) {
2315                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2316                 return;
2317         }
2318
2319         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2320 }
2321
2322 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2323 {
2324         struct inet6_ifaddr * ifp;
2325         u32 addr_flags = IFA_F_PERMANENT;
2326
2327 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2328         if (idev->cnf.optimistic_dad &&
2329             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2330                 addr_flags |= IFA_F_OPTIMISTIC;
2331 #endif
2332
2333
2334         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2335         if (!IS_ERR(ifp)) {
2336                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2337                 addrconf_dad_start(ifp, 0);
2338                 in6_ifa_put(ifp);
2339         }
2340 }
2341
2342 static void addrconf_dev_config(struct net_device *dev)
2343 {
2344         struct in6_addr addr;
2345         struct inet6_dev    * idev;
2346
2347         ASSERT_RTNL();
2348
2349         if ((dev->type != ARPHRD_ETHER) &&
2350             (dev->type != ARPHRD_FDDI) &&
2351             (dev->type != ARPHRD_IEEE802_TR) &&
2352             (dev->type != ARPHRD_ARCNET) &&
2353             (dev->type != ARPHRD_INFINIBAND)) {
2354                 /* Alas, we support only Ethernet autoconfiguration. */
2355                 return;
2356         }
2357
2358         idev = addrconf_add_dev(dev);
2359         if (idev == NULL)
2360                 return;
2361
2362         memset(&addr, 0, sizeof(struct in6_addr));
2363         addr.s6_addr32[0] = htonl(0xFE800000);
2364
2365         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2366                 addrconf_add_linklocal(idev, &addr);
2367 }
2368
2369 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2370 static void addrconf_sit_config(struct net_device *dev)
2371 {
2372         struct inet6_dev *idev;
2373
2374         ASSERT_RTNL();
2375
2376         /*
2377          * Configure the tunnel with one of our IPv4
2378          * addresses... we should configure all of
2379          * our v4 addrs in the tunnel
2380          */
2381
2382         if ((idev = ipv6_find_idev(dev)) == NULL) {
2383                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2384                 return;
2385         }
2386
2387         if (dev->priv_flags & IFF_ISATAP) {
2388                 struct in6_addr addr;
2389
2390                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2391                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2392                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2393                         addrconf_add_linklocal(idev, &addr);
2394                 return;
2395         }
2396
2397         sit_add_v4_addrs(idev);
2398
2399         if (dev->flags&IFF_POINTOPOINT) {
2400                 addrconf_add_mroute(dev);
2401                 addrconf_add_lroute(dev);
2402         } else
2403                 sit_route_add(dev);
2404 }
2405 #endif
2406
2407 static inline int
2408 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2409 {
2410         struct in6_addr lladdr;
2411
2412         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2413                 addrconf_add_linklocal(idev, &lladdr);
2414                 return 0;
2415         }
2416         return -1;
2417 }
2418
2419 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2420 {
2421         struct net_device *link_dev;
2422         struct net *net = dev_net(idev->dev);
2423
2424         /* first try to inherit the link-local address from the link device */
2425         if (idev->dev->iflink &&
2426             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2427                 if (!ipv6_inherit_linklocal(idev, link_dev))
2428                         return;
2429         }
2430         /* then try to inherit it from any device */
2431         for_each_netdev(net, link_dev) {
2432                 if (!ipv6_inherit_linklocal(idev, link_dev))
2433                         return;
2434         }
2435         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2436 }
2437
2438 /*
2439  * Autoconfigure tunnel with a link-local address so routing protocols,
2440  * DHCPv6, MLD etc. can be run over the virtual link
2441  */
2442
2443 static void addrconf_ip6_tnl_config(struct net_device *dev)
2444 {
2445         struct inet6_dev *idev;
2446
2447         ASSERT_RTNL();
2448
2449         if ((idev = addrconf_add_dev(dev)) == NULL) {
2450                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2451                 return;
2452         }
2453         ip6_tnl_add_linklocal(idev);
2454 }
2455
2456 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2457                            void * data)
2458 {
2459         struct net_device *dev = (struct net_device *) data;
2460         struct inet6_dev *idev = __in6_dev_get(dev);
2461         int run_pending = 0;
2462         int err;
2463
2464         switch(event) {
2465         case NETDEV_REGISTER:
2466                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2467                         idev = ipv6_add_dev(dev);
2468                         if (!idev)
2469                                 return notifier_from_errno(-ENOMEM);
2470                 }
2471                 break;
2472         case NETDEV_UP:
2473         case NETDEV_CHANGE:
2474                 if (dev->flags & IFF_SLAVE)
2475                         break;
2476
2477                 if (event == NETDEV_UP) {
2478                         if (!addrconf_qdisc_ok(dev)) {
2479                                 /* device is not ready yet. */
2480                                 printk(KERN_INFO
2481                                         "ADDRCONF(NETDEV_UP): %s: "
2482                                         "link is not ready\n",
2483                                         dev->name);
2484                                 break;
2485                         }
2486
2487                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2488                                 idev = ipv6_add_dev(dev);
2489
2490                         if (idev) {
2491                                 idev->if_flags |= IF_READY;
2492                                 run_pending = 1;
2493                         }
2494                 } else {
2495                         if (!addrconf_qdisc_ok(dev)) {
2496                                 /* device is still not ready. */
2497                                 break;
2498                         }
2499
2500                         if (idev) {
2501                                 if (idev->if_flags & IF_READY) {
2502                                         /* device is already configured. */
2503                                         break;
2504                                 }
2505                                 idev->if_flags |= IF_READY;
2506                         }
2507
2508                         printk(KERN_INFO
2509                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2510                                         "link becomes ready\n",
2511                                         dev->name);
2512
2513                         run_pending = 1;
2514                 }
2515
2516                 switch(dev->type) {
2517 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2518                 case ARPHRD_SIT:
2519                         addrconf_sit_config(dev);
2520                         break;
2521 #endif
2522                 case ARPHRD_TUNNEL6:
2523                         addrconf_ip6_tnl_config(dev);
2524                         break;
2525                 case ARPHRD_LOOPBACK:
2526                         init_loopback(dev);
2527                         break;
2528
2529                 default:
2530                         addrconf_dev_config(dev);
2531                         break;
2532                 }
2533                 if (idev) {
2534                         if (run_pending)
2535                                 addrconf_dad_run(idev);
2536
2537                         /* If the MTU changed during the interface down, when the
2538                            interface up, the changed MTU must be reflected in the
2539                            idev as well as routers.
2540                          */
2541                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2542                                 rt6_mtu_change(dev, dev->mtu);
2543                                 idev->cnf.mtu6 = dev->mtu;
2544                         }
2545                         idev->tstamp = jiffies;
2546                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2547                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2548                            stop IPv6 on this interface.
2549                          */
2550                         if (dev->mtu < IPV6_MIN_MTU)
2551                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2552                 }
2553                 break;
2554
2555         case NETDEV_CHANGEMTU:
2556                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2557                         rt6_mtu_change(dev, dev->mtu);
2558                         idev->cnf.mtu6 = dev->mtu;
2559                         break;
2560                 }
2561
2562                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2563                         idev = ipv6_add_dev(dev);
2564                         if (idev)
2565                                 break;
2566                 }
2567
2568                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2569
2570         case NETDEV_DOWN:
2571         case NETDEV_UNREGISTER:
2572                 /*
2573                  *      Remove all addresses from this interface.
2574                  */
2575                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2576                 break;
2577
2578         case NETDEV_CHANGENAME:
2579                 if (idev) {
2580                         snmp6_unregister_dev(idev);
2581                         addrconf_sysctl_unregister(idev);
2582                         addrconf_sysctl_register(idev);
2583                         err = snmp6_register_dev(idev);
2584                         if (err)
2585                                 return notifier_from_errno(err);
2586                 }
2587                 break;
2588         case NETDEV_BONDING_OLDTYPE:
2589         case NETDEV_BONDING_NEWTYPE:
2590                 addrconf_bonding_change(dev, event);
2591                 break;
2592         }
2593
2594         return NOTIFY_OK;
2595 }
2596
2597 /*
2598  *      addrconf module should be notified of a device going up
2599  */
2600 static struct notifier_block ipv6_dev_notf = {
2601         .notifier_call = addrconf_notify,
2602         .priority = 0
2603 };
2604
2605 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2606 {
2607         struct inet6_dev *idev;
2608         ASSERT_RTNL();
2609
2610         idev = __in6_dev_get(dev);
2611
2612         if (event == NETDEV_BONDING_NEWTYPE)
2613                 ipv6_mc_remap(idev);
2614         else if (event == NETDEV_BONDING_OLDTYPE)
2615                 ipv6_mc_unmap(idev);
2616 }
2617
2618 static int addrconf_ifdown(struct net_device *dev, int how)
2619 {
2620         struct inet6_dev *idev;
2621         struct inet6_ifaddr *ifa, *keep_list, **bifa;
2622         struct net *net = dev_net(dev);
2623         int i;
2624
2625         ASSERT_RTNL();
2626
2627         rt6_ifdown(net, dev);
2628         neigh_ifdown(&nd_tbl, dev);
2629
2630         idev = __in6_dev_get(dev);
2631         if (idev == NULL)
2632                 return -ENODEV;
2633
2634         /* Step 1: remove reference to ipv6 device from parent device.
2635                    Do not dev_put!
2636          */
2637         if (how) {
2638                 idev->dead = 1;
2639
2640                 /* protected by rtnl_lock */
2641                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2642
2643                 /* Step 1.5: remove snmp6 entry */
2644                 snmp6_unregister_dev(idev);
2645
2646         }
2647
2648         /* Step 2: clear hash table */
2649         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2650                 bifa = &inet6_addr_lst[i];
2651
2652                 write_lock_bh(&addrconf_hash_lock);
2653                 while ((ifa = *bifa) != NULL) {
2654                         if (ifa->idev == idev &&
2655                             (how || !(ifa->flags&IFA_F_PERMANENT) ||
2656                              ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2657                                 *bifa = ifa->lst_next;
2658                                 ifa->lst_next = NULL;
2659                                 __in6_ifa_put(ifa);
2660                                 continue;
2661                         }
2662                         bifa = &ifa->lst_next;
2663                 }
2664                 write_unlock_bh(&addrconf_hash_lock);
2665         }
2666
2667         write_lock_bh(&idev->lock);
2668
2669         /* Step 3: clear flags for stateless addrconf */
2670         if (!how)
2671                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2672
2673         /* Step 4: clear address list */
2674 #ifdef CONFIG_IPV6_PRIVACY
2675         if (how && del_timer(&idev->regen_timer))
2676                 in6_dev_put(idev);
2677
2678         /* clear tempaddr list */
2679         while ((ifa = idev->tempaddr_list) != NULL) {
2680                 idev->tempaddr_list = ifa->tmp_next;
2681                 ifa->tmp_next = NULL;
2682                 ifa->dead = 1;
2683                 write_unlock_bh(&idev->lock);
2684                 spin_lock_bh(&ifa->lock);
2685
2686                 if (ifa->ifpub) {
2687                         in6_ifa_put(ifa->ifpub);
2688                         ifa->ifpub = NULL;
2689                 }
2690                 spin_unlock_bh(&ifa->lock);
2691                 in6_ifa_put(ifa);
2692                 write_lock_bh(&idev->lock);
2693         }
2694 #endif
2695         keep_list = NULL;
2696         bifa = &keep_list;
2697         while ((ifa = idev->addr_list) != NULL) {
2698                 idev->addr_list = ifa->if_next;
2699                 ifa->if_next = NULL;
2700
2701                 addrconf_del_timer(ifa);
2702
2703                 /* If just doing link down, and address is permanent
2704                    and not link-local, then retain it. */
2705                 if (how == 0 &&
2706                     (ifa->flags&IFA_F_PERMANENT) &&
2707                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2708
2709                         /* Move to holding list */
2710                         *bifa = ifa;
2711                         bifa = &ifa->if_next;
2712
2713                         /* If not doing DAD on this address, just keep it. */
2714                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2715                             idev->cnf.accept_dad <= 0 ||
2716                             (ifa->flags & IFA_F_NODAD))
2717                                 continue;
2718
2719                         /* If it was tentative already, no need to notify */
2720                         if (ifa->flags & IFA_F_TENTATIVE)
2721                                 continue;
2722
2723                         /* Flag it for later restoration when link comes up */
2724                         ifa->flags |= IFA_F_TENTATIVE;
2725                         in6_ifa_hold(ifa);
2726                 } else {
2727                         ifa->dead = 1;
2728                 }
2729                 write_unlock_bh(&idev->lock);
2730
2731                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2732                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2733                 in6_ifa_put(ifa);
2734
2735                 write_lock_bh(&idev->lock);
2736         }
2737
2738         idev->addr_list = keep_list;
2739
2740         write_unlock_bh(&idev->lock);
2741
2742         /* Step 5: Discard multicast list */
2743
2744         if (how)
2745                 ipv6_mc_destroy_dev(idev);
2746         else
2747                 ipv6_mc_down(idev);
2748
2749         idev->tstamp = jiffies;
2750
2751         /* Shot the device (if unregistered) */
2752
2753         if (how) {
2754                 addrconf_sysctl_unregister(idev);
2755                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2756                 neigh_ifdown(&nd_tbl, dev);
2757                 in6_dev_put(idev);
2758         }
2759         return 0;
2760 }
2761
2762 static void addrconf_rs_timer(unsigned long data)
2763 {
2764         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2765         struct inet6_dev *idev = ifp->idev;
2766
2767         read_lock(&idev->lock);
2768         if (idev->dead || !(idev->if_flags & IF_READY))
2769                 goto out;
2770
2771         if (idev->cnf.forwarding)
2772                 goto out;
2773
2774         /* Announcement received after solicitation was sent */
2775         if (idev->if_flags & IF_RA_RCVD)
2776                 goto out;
2777
2778         spin_lock(&ifp->lock);
2779         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2780                 /* The wait after the last probe can be shorter */
2781                 addrconf_mod_timer(ifp, AC_RS,
2782                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2783                                    idev->cnf.rtr_solicit_delay :
2784                                    idev->cnf.rtr_solicit_interval);
2785                 spin_unlock(&ifp->lock);
2786
2787                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2788         } else {
2789                 spin_unlock(&ifp->lock);
2790                 /*
2791                  * Note: we do not support deprecated "all on-link"
2792                  * assumption any longer.
2793                  */
2794                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2795                        idev->dev->name);
2796         }
2797
2798 out:
2799         read_unlock(&idev->lock);
2800         in6_ifa_put(ifp);
2801 }
2802
2803 /*
2804  *      Duplicate Address Detection
2805  */
2806 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2807 {
2808         unsigned long rand_num;
2809         struct inet6_dev *idev = ifp->idev;
2810
2811         if (ifp->flags & IFA_F_OPTIMISTIC)
2812                 rand_num = 0;
2813         else
2814                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2815
2816         ifp->probes = idev->cnf.dad_transmits;
2817         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2818 }
2819
2820 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2821 {
2822         struct inet6_dev *idev = ifp->idev;
2823         struct net_device *dev = idev->dev;
2824
2825         addrconf_join_solict(dev, &ifp->addr);
2826
2827         net_srandom(ifp->addr.s6_addr32[3]);
2828
2829         read_lock_bh(&idev->lock);
2830         if (ifp->dead)
2831                 goto out;
2832
2833         spin_lock(&ifp->lock);
2834         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2835             idev->cnf.accept_dad < 1 ||
2836             !(ifp->flags&IFA_F_TENTATIVE) ||
2837             ifp->flags & IFA_F_NODAD) {
2838                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2839                 spin_unlock(&ifp->lock);
2840                 read_unlock_bh(&idev->lock);
2841
2842                 addrconf_dad_completed(ifp);
2843                 return;
2844         }
2845
2846         if (!(idev->if_flags & IF_READY)) {
2847                 spin_unlock(&ifp->lock);
2848                 read_unlock_bh(&idev->lock);
2849                 /*
2850                  * If the device is not ready:
2851                  * - keep it tentative if it is a permanent address.
2852                  * - otherwise, kill it.
2853                  */
2854                 in6_ifa_hold(ifp);
2855                 addrconf_dad_stop(ifp, 0);
2856                 return;
2857         }
2858
2859         /*
2860          * Optimistic nodes can start receiving
2861          * Frames right away
2862          */
2863         if(ifp->flags & IFA_F_OPTIMISTIC)
2864                 ip6_ins_rt(ifp->rt);
2865
2866         addrconf_dad_kick(ifp);
2867         spin_unlock(&ifp->lock);
2868 out:
2869         read_unlock_bh(&idev->lock);
2870 }
2871
2872 static void addrconf_dad_timer(unsigned long data)
2873 {
2874         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2875         struct inet6_dev *idev = ifp->idev;
2876         struct in6_addr mcaddr;
2877
2878         read_lock(&idev->lock);
2879         if (idev->dead || !(idev->if_flags & IF_READY)) {
2880                 read_unlock(&idev->lock);
2881                 goto out;
2882         }
2883
2884         spin_lock(&ifp->lock);
2885         if (ifp->probes == 0) {
2886                 /*
2887                  * DAD was successful
2888                  */
2889
2890                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2891                 spin_unlock(&ifp->lock);
2892                 read_unlock(&idev->lock);
2893
2894                 addrconf_dad_completed(ifp);
2895
2896                 goto out;
2897         }
2898
2899         ifp->probes--;
2900         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2901         spin_unlock(&ifp->lock);
2902         read_unlock(&idev->lock);
2903
2904         /* send a neighbour solicitation for our addr */
2905         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2906         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2907 out:
2908         in6_ifa_put(ifp);
2909 }
2910
2911 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2912 {
2913         struct net_device *     dev = ifp->idev->dev;
2914
2915         /*
2916          *      Configure the address for reception. Now it is valid.
2917          */
2918
2919         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2920
2921         /* If added prefix is link local and forwarding is off,
2922            start sending router solicitations.
2923          */
2924
2925         if (ifp->idev->cnf.forwarding == 0 &&
2926             ifp->idev->cnf.rtr_solicits > 0 &&
2927             (dev->flags&IFF_LOOPBACK) == 0 &&
2928             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2929                 /*
2930                  *      If a host as already performed a random delay
2931                  *      [...] as part of DAD [...] there is no need
2932                  *      to delay again before sending the first RS
2933                  */
2934                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2935
2936                 spin_lock_bh(&ifp->lock);
2937                 ifp->probes = 1;
2938                 ifp->idev->if_flags |= IF_RS_SENT;
2939                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2940                 spin_unlock_bh(&ifp->lock);
2941         }
2942 }
2943
2944 static void addrconf_dad_run(struct inet6_dev *idev) {
2945         struct inet6_ifaddr *ifp;
2946
2947         read_lock_bh(&idev->lock);
2948         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2949                 spin_lock(&ifp->lock);
2950                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2951                         spin_unlock(&ifp->lock);
2952                         continue;
2953                 }
2954                 spin_unlock(&ifp->lock);
2955                 addrconf_dad_kick(ifp);
2956         }
2957         read_unlock_bh(&idev->lock);
2958 }
2959
2960 #ifdef CONFIG_PROC_FS
2961 struct if6_iter_state {
2962         struct seq_net_private p;
2963         int bucket;
2964 };
2965
2966 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2967 {
2968         struct inet6_ifaddr *ifa = NULL;
2969         struct if6_iter_state *state = seq->private;
2970         struct net *net = seq_file_net(seq);
2971
2972         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2973                 ifa = inet6_addr_lst[state->bucket];
2974
2975                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2976                         ifa = ifa->lst_next;
2977                 if (ifa)
2978                         break;
2979         }
2980         return ifa;
2981 }
2982
2983 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2984 {
2985         struct if6_iter_state *state = seq->private;
2986         struct net *net = seq_file_net(seq);
2987
2988         ifa = ifa->lst_next;
2989 try_again:
2990         if (ifa) {
2991                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2992                         ifa = ifa->lst_next;
2993                         goto try_again;
2994                 }
2995         }
2996
2997         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2998                 ifa = inet6_addr_lst[state->bucket];
2999                 goto try_again;
3000         }
3001
3002         return ifa;
3003 }
3004
3005 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3006 {
3007         struct inet6_ifaddr *ifa = if6_get_first(seq);
3008
3009         if (ifa)
3010                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3011                         --pos;
3012         return pos ? NULL : ifa;
3013 }
3014
3015 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3016         __acquires(addrconf_hash_lock)
3017 {
3018         read_lock_bh(&addrconf_hash_lock);
3019         return if6_get_idx(seq, *pos);
3020 }
3021
3022 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3023 {
3024         struct inet6_ifaddr *ifa;
3025
3026         ifa = if6_get_next(seq, v);
3027         ++*pos;
3028         return ifa;
3029 }
3030
3031 static void if6_seq_stop(struct seq_file *seq, void *v)
3032         __releases(addrconf_hash_lock)
3033 {
3034         read_unlock_bh(&addrconf_hash_lock);
3035 }
3036
3037 static int if6_seq_show(struct seq_file *seq, void *v)
3038 {
3039         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3040         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3041                    &ifp->addr,
3042                    ifp->idev->dev->ifindex,
3043                    ifp->prefix_len,
3044                    ifp->scope,
3045                    ifp->flags,
3046                    ifp->idev->dev->name);
3047         return 0;
3048 }
3049
3050 static const struct seq_operations if6_seq_ops = {
3051         .start  = if6_seq_start,
3052         .next   = if6_seq_next,
3053         .show   = if6_seq_show,
3054         .stop   = if6_seq_stop,
3055 };
3056
3057 static int if6_seq_open(struct inode *inode, struct file *file)
3058 {
3059         return seq_open_net(inode, file, &if6_seq_ops,
3060                             sizeof(struct if6_iter_state));
3061 }
3062
3063 static const struct file_operations if6_fops = {
3064         .owner          = THIS_MODULE,
3065         .open           = if6_seq_open,
3066         .read           = seq_read,
3067         .llseek         = seq_lseek,
3068         .release        = seq_release_net,
3069 };
3070
3071 static int __net_init if6_proc_net_init(struct net *net)
3072 {
3073         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3074                 return -ENOMEM;
3075         return 0;
3076 }
3077
3078 static void __net_exit if6_proc_net_exit(struct net *net)
3079 {
3080        proc_net_remove(net, "if_inet6");
3081 }
3082
3083 static struct pernet_operations if6_proc_net_ops = {
3084        .init = if6_proc_net_init,
3085        .exit = if6_proc_net_exit,
3086 };
3087
3088 int __init if6_proc_init(void)
3089 {
3090         return register_pernet_subsys(&if6_proc_net_ops);
3091 }
3092
3093 void if6_proc_exit(void)
3094 {
3095         unregister_pernet_subsys(&if6_proc_net_ops);
3096 }
3097 #endif  /* CONFIG_PROC_FS */
3098
3099 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3100 /* Check if address is a home address configured on any interface. */
3101 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3102 {
3103         int ret = 0;
3104         struct inet6_ifaddr * ifp;
3105         u8 hash = ipv6_addr_hash(addr);
3106         read_lock_bh(&addrconf_hash_lock);
3107         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3108                 if (!net_eq(dev_net(ifp->idev->dev), net))
3109                         continue;
3110                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3111                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3112                         ret = 1;
3113                         break;
3114                 }
3115         }
3116         read_unlock_bh(&addrconf_hash_lock);
3117         return ret;
3118 }
3119 #endif
3120
3121 /*
3122  *      Periodic address status verification
3123  */
3124
3125 static void addrconf_verify(unsigned long foo)
3126 {
3127         struct inet6_ifaddr *ifp;
3128         unsigned long now, next;
3129         int i;
3130
3131         spin_lock_bh(&addrconf_verify_lock);
3132         now = jiffies;
3133         next = now + ADDR_CHECK_FREQUENCY;
3134
3135         del_timer(&addr_chk_timer);
3136
3137         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3138
3139 restart:
3140                 read_lock(&addrconf_hash_lock);
3141                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3142                         unsigned long age;
3143 #ifdef CONFIG_IPV6_PRIVACY
3144                         unsigned long regen_advance;
3145 #endif
3146
3147                         if (ifp->flags & IFA_F_PERMANENT)
3148                                 continue;
3149
3150                         spin_lock(&ifp->lock);
3151                         age = (now - ifp->tstamp) / HZ;
3152
3153 #ifdef CONFIG_IPV6_PRIVACY
3154                         regen_advance = ifp->idev->cnf.regen_max_retry *
3155                                         ifp->idev->cnf.dad_transmits *
3156                                         ifp->idev->nd_parms->retrans_time / HZ;
3157 #endif
3158
3159                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3160                             age >= ifp->valid_lft) {
3161                                 spin_unlock(&ifp->lock);
3162                                 in6_ifa_hold(ifp);
3163                                 read_unlock(&addrconf_hash_lock);
3164                                 ipv6_del_addr(ifp);
3165                                 goto restart;
3166                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3167                                 spin_unlock(&ifp->lock);
3168                                 continue;
3169                         } else if (age >= ifp->prefered_lft) {
3170                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3171                                 int deprecate = 0;
3172
3173                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3174                                         deprecate = 1;
3175                                         ifp->flags |= IFA_F_DEPRECATED;
3176                                 }
3177
3178                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3179                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3180
3181                                 spin_unlock(&ifp->lock);
3182
3183                                 if (deprecate) {
3184                                         in6_ifa_hold(ifp);
3185                                         read_unlock(&addrconf_hash_lock);
3186
3187                                         ipv6_ifa_notify(0, ifp);
3188                                         in6_ifa_put(ifp);
3189                                         goto restart;
3190                                 }
3191 #ifdef CONFIG_IPV6_PRIVACY
3192                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3193                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3194                                 if (age >= ifp->prefered_lft - regen_advance) {
3195                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3196                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3197                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3198                                         if (!ifp->regen_count && ifpub) {
3199                                                 ifp->regen_count++;
3200                                                 in6_ifa_hold(ifp);
3201                                                 in6_ifa_hold(ifpub);
3202                                                 spin_unlock(&ifp->lock);
3203                                                 read_unlock(&addrconf_hash_lock);
3204                                                 spin_lock(&ifpub->lock);
3205                                                 ifpub->regen_count = 0;
3206                                                 spin_unlock(&ifpub->lock);
3207                                                 ipv6_create_tempaddr(ifpub, ifp);
3208                                                 in6_ifa_put(ifpub);
3209                                                 in6_ifa_put(ifp);
3210                                                 goto restart;
3211                                         }
3212                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3213                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3214                                 spin_unlock(&ifp->lock);
3215 #endif
3216                         } else {
3217                                 /* ifp->prefered_lft <= ifp->valid_lft */
3218                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3219                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3220                                 spin_unlock(&ifp->lock);
3221                         }
3222                 }
3223                 read_unlock(&addrconf_hash_lock);
3224         }
3225
3226         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3227         add_timer(&addr_chk_timer);
3228         spin_unlock_bh(&addrconf_verify_lock);
3229 }
3230
3231 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3232 {
3233         struct in6_addr *pfx = NULL;
3234
3235         if (addr)
3236                 pfx = nla_data(addr);
3237
3238         if (local) {
3239                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3240                         pfx = NULL;
3241                 else
3242                         pfx = nla_data(local);
3243         }
3244
3245         return pfx;
3246 }
3247
3248 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3249         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3250         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3251         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3252 };
3253
3254 static int
3255 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3256 {
3257         struct net *net = sock_net(skb->sk);
3258         struct ifaddrmsg *ifm;
3259         struct nlattr *tb[IFA_MAX+1];
3260         struct in6_addr *pfx;
3261         int err;
3262
3263         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3264         if (err < 0)
3265                 return err;
3266
3267         ifm = nlmsg_data(nlh);
3268         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3269         if (pfx == NULL)
3270                 return -EINVAL;
3271
3272         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3273 }
3274
3275 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3276                              u32 prefered_lft, u32 valid_lft)
3277 {
3278         u32 flags;
3279         clock_t expires;
3280         unsigned long timeout;
3281
3282         if (!valid_lft || (prefered_lft > valid_lft))
3283                 return -EINVAL;
3284
3285         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3286         if (addrconf_finite_timeout(timeout)) {
3287                 expires = jiffies_to_clock_t(timeout * HZ);
3288                 valid_lft = timeout;
3289                 flags = RTF_EXPIRES;
3290         } else {
3291                 expires = 0;
3292                 flags = 0;
3293                 ifa_flags |= IFA_F_PERMANENT;
3294         }
3295
3296         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3297         if (addrconf_finite_timeout(timeout)) {
3298                 if (timeout == 0)
3299                         ifa_flags |= IFA_F_DEPRECATED;
3300                 prefered_lft = timeout;
3301         }
3302
3303         spin_lock_bh(&ifp->lock);
3304         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3305         ifp->tstamp = jiffies;
3306         ifp->valid_lft = valid_lft;
3307         ifp->prefered_lft = prefered_lft;
3308
3309         spin_unlock_bh(&ifp->lock);
3310         if (!(ifp->flags&IFA_F_TENTATIVE))
3311                 ipv6_ifa_notify(0, ifp);
3312
3313         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3314                               expires, flags);
3315         addrconf_verify(0);
3316
3317         return 0;
3318 }
3319
3320 static int
3321 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3322 {
3323         struct net *net = sock_net(skb->sk);
3324         struct ifaddrmsg *ifm;
3325         struct nlattr *tb[IFA_MAX+1];
3326         struct in6_addr *pfx;
3327         struct inet6_ifaddr *ifa;
3328         struct net_device *dev;
3329         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3330         u8 ifa_flags;
3331         int err;
3332
3333         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3334         if (err < 0)
3335                 return err;
3336
3337         ifm = nlmsg_data(nlh);
3338         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3339         if (pfx == NULL)
3340                 return -EINVAL;
3341
3342         if (tb[IFA_CACHEINFO]) {
3343                 struct ifa_cacheinfo *ci;
3344
3345                 ci = nla_data(tb[IFA_CACHEINFO]);
3346                 valid_lft = ci->ifa_valid;
3347                 preferred_lft = ci->ifa_prefered;
3348         } else {
3349                 preferred_lft = INFINITY_LIFE_TIME;
3350                 valid_lft = INFINITY_LIFE_TIME;
3351         }
3352
3353         dev =  __dev_get_by_index(net, ifm->ifa_index);
3354         if (dev == NULL)
3355                 return -ENODEV;
3356
3357         /* We ignore other flags so far. */
3358         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3359
3360         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3361         if (ifa == NULL) {
3362                 /*
3363                  * It would be best to check for !NLM_F_CREATE here but
3364                  * userspace alreay relies on not having to provide this.
3365                  */
3366                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3367                                       ifm->ifa_prefixlen, ifa_flags,
3368                                       preferred_lft, valid_lft);
3369         }
3370
3371         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3372             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3373                 err = -EEXIST;
3374         else
3375                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3376
3377         in6_ifa_put(ifa);
3378
3379         return err;
3380 }
3381
3382 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3383                           u8 scope, int ifindex)
3384 {
3385         struct ifaddrmsg *ifm;
3386
3387         ifm = nlmsg_data(nlh);
3388         ifm->ifa_family = AF_INET6;
3389         ifm->ifa_prefixlen = prefixlen;
3390         ifm->ifa_flags = flags;
3391         ifm->ifa_scope = scope;
3392         ifm->ifa_index = ifindex;
3393 }
3394
3395 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3396                          unsigned long tstamp, u32 preferred, u32 valid)
3397 {
3398         struct ifa_cacheinfo ci;
3399
3400         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3401                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3402         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3403                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3404         ci.ifa_prefered = preferred;
3405         ci.ifa_valid = valid;
3406
3407         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3408 }
3409
3410 static inline int rt_scope(int ifa_scope)
3411 {
3412         if (ifa_scope & IFA_HOST)
3413                 return RT_SCOPE_HOST;
3414         else if (ifa_scope & IFA_LINK)
3415                 return RT_SCOPE_LINK;
3416         else if (ifa_scope & IFA_SITE)
3417                 return RT_SCOPE_SITE;
3418         else
3419                 return RT_SCOPE_UNIVERSE;
3420 }
3421
3422 static inline int inet6_ifaddr_msgsize(void)
3423 {
3424         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3425                + nla_total_size(16) /* IFA_ADDRESS */
3426                + nla_total_size(sizeof(struct ifa_cacheinfo));
3427 }
3428
3429 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3430                              u32 pid, u32 seq, int event, unsigned int flags)
3431 {
3432         struct nlmsghdr  *nlh;
3433         u32 preferred, valid;
3434
3435         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3436         if (nlh == NULL)
3437                 return -EMSGSIZE;
3438
3439         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3440                       ifa->idev->dev->ifindex);
3441
3442         if (!(ifa->flags&IFA_F_PERMANENT)) {
3443                 preferred = ifa->prefered_lft;
3444                 valid = ifa->valid_lft;
3445                 if (preferred != INFINITY_LIFE_TIME) {
3446                         long tval = (jiffies - ifa->tstamp)/HZ;
3447                         if (preferred > tval)
3448                                 preferred -= tval;
3449                         else
3450                                 preferred = 0;
3451                         if (valid != INFINITY_LIFE_TIME)
3452                                 valid -= tval;
3453                 }
3454         } else {
3455                 preferred = INFINITY_LIFE_TIME;
3456                 valid = INFINITY_LIFE_TIME;
3457         }
3458
3459         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3460             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3461                 nlmsg_cancel(skb, nlh);
3462                 return -EMSGSIZE;
3463         }
3464
3465         return nlmsg_end(skb, nlh);
3466 }
3467
3468 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3469                                 u32 pid, u32 seq, int event, u16 flags)
3470 {
3471         struct nlmsghdr  *nlh;
3472         u8 scope = RT_SCOPE_UNIVERSE;
3473         int ifindex = ifmca->idev->dev->ifindex;
3474
3475         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3476                 scope = RT_SCOPE_SITE;
3477
3478         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3479         if (nlh == NULL)
3480                 return -EMSGSIZE;
3481
3482         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3483         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3484             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3485                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3486                 nlmsg_cancel(skb, nlh);
3487                 return -EMSGSIZE;
3488         }
3489
3490         return nlmsg_end(skb, nlh);
3491 }
3492
3493 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3494                                 u32 pid, u32 seq, int event, unsigned int flags)
3495 {
3496         struct nlmsghdr  *nlh;
3497         u8 scope = RT_SCOPE_UNIVERSE;
3498         int ifindex = ifaca->aca_idev->dev->ifindex;
3499
3500         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3501                 scope = RT_SCOPE_SITE;
3502
3503         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3504         if (nlh == NULL)
3505                 return -EMSGSIZE;
3506
3507         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3508         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3509             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3510                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3511                 nlmsg_cancel(skb, nlh);
3512                 return -EMSGSIZE;
3513         }
3514
3515         return nlmsg_end(skb, nlh);
3516 }
3517
3518 enum addr_type_t
3519 {
3520         UNICAST_ADDR,
3521         MULTICAST_ADDR,
3522         ANYCAST_ADDR,
3523 };
3524
3525 /* called with rcu_read_lock() */
3526 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3527                           struct netlink_callback *cb, enum addr_type_t type,
3528                           int s_ip_idx, int *p_ip_idx)
3529 {
3530         struct inet6_ifaddr *ifa;
3531         struct ifmcaddr6 *ifmca;
3532         struct ifacaddr6 *ifaca;
3533         int err = 1;
3534         int ip_idx = *p_ip_idx;
3535
3536         read_lock_bh(&idev->lock);
3537         switch (type) {
3538         case UNICAST_ADDR:
3539                 /* unicast address incl. temp addr */
3540                 for (ifa = idev->addr_list; ifa;
3541                      ifa = ifa->if_next, ip_idx++) {
3542                         if (ip_idx < s_ip_idx)
3543                                 continue;
3544                         err = inet6_fill_ifaddr(skb, ifa,
3545                                                 NETLINK_CB(cb->skb).pid,
3546                                                 cb->nlh->nlmsg_seq,
3547                                                 RTM_NEWADDR,
3548                                                 NLM_F_MULTI);
3549                         if (err <= 0)
3550                                 break;
3551                 }
3552                 break;
3553         case MULTICAST_ADDR:
3554                 /* multicast address */
3555                 for (ifmca = idev->mc_list; ifmca;
3556                      ifmca = ifmca->next, ip_idx++) {
3557                         if (ip_idx < s_ip_idx)
3558                                 continue;
3559                         err = inet6_fill_ifmcaddr(skb, ifmca,
3560                                                   NETLINK_CB(cb->skb).pid,
3561                                                   cb->nlh->nlmsg_seq,
3562                                                   RTM_GETMULTICAST,
3563                                                   NLM_F_MULTI);
3564                         if (err <= 0)
3565                                 break;
3566                 }
3567                 break;
3568         case ANYCAST_ADDR:
3569                 /* anycast address */
3570                 for (ifaca = idev->ac_list; ifaca;
3571                      ifaca = ifaca->aca_next, ip_idx++) {
3572                         if (ip_idx < s_ip_idx)
3573                                 continue;
3574                         err = inet6_fill_ifacaddr(skb, ifaca,
3575                                                   NETLINK_CB(cb->skb).pid,
3576                                                   cb->nlh->nlmsg_seq,
3577                                                   RTM_GETANYCAST,
3578                                                   NLM_F_MULTI);
3579                         if (err <= 0)
3580                                 break;
3581                 }
3582                 break;
3583         default:
3584                 break;
3585         }
3586         read_unlock_bh(&idev->lock);
3587         *p_ip_idx = ip_idx;
3588         return err;
3589 }
3590
3591 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3592                            enum addr_type_t type)
3593 {
3594         struct net *net = sock_net(skb->sk);
3595         int h, s_h;
3596         int idx, ip_idx;
3597         int s_idx, s_ip_idx;
3598         struct net_device *dev;
3599         struct inet6_dev *idev;
3600         struct hlist_head *head;
3601         struct hlist_node *node;
3602
3603         s_h = cb->args[0];
3604         s_idx = idx = cb->args[1];
3605         s_ip_idx = ip_idx = cb->args[2];
3606
3607         rcu_read_lock();
3608         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3609                 idx = 0;
3610                 head = &net->dev_index_head[h];
3611                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3612                         if (idx < s_idx)
3613                                 goto cont;
3614                         if (h > s_h || idx > s_idx)
3615                                 s_ip_idx = 0;
3616                         ip_idx = 0;
3617                         if ((idev = __in6_dev_get(dev)) == NULL)
3618                                 goto cont;
3619
3620                         if (in6_dump_addrs(idev, skb, cb, type,
3621                                            s_ip_idx, &ip_idx) <= 0)
3622                                 goto done;
3623 cont:
3624                         idx++;
3625                 }
3626         }
3627 done:
3628         rcu_read_unlock();
3629         cb->args[0] = h;
3630         cb->args[1] = idx;
3631         cb->args[2] = ip_idx;
3632
3633         return skb->len;
3634 }
3635
3636 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3637 {
3638         enum addr_type_t type = UNICAST_ADDR;
3639
3640         return inet6_dump_addr(skb, cb, type);
3641 }
3642
3643 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3644 {
3645         enum addr_type_t type = MULTICAST_ADDR;
3646
3647         return inet6_dump_addr(skb, cb, type);
3648 }
3649
3650
3651 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3652 {
3653         enum addr_type_t type = ANYCAST_ADDR;
3654
3655         return inet6_dump_addr(skb, cb, type);
3656 }
3657
3658 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3659                              void *arg)
3660 {
3661         struct net *net = sock_net(in_skb->sk);
3662         struct ifaddrmsg *ifm;
3663         struct nlattr *tb[IFA_MAX+1];
3664         struct in6_addr *addr = NULL;
3665         struct net_device *dev = NULL;
3666         struct inet6_ifaddr *ifa;
3667         struct sk_buff *skb;
3668         int err;
3669
3670         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3671         if (err < 0)
3672                 goto errout;
3673
3674         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3675         if (addr == NULL) {
3676                 err = -EINVAL;
3677                 goto errout;
3678         }
3679
3680         ifm = nlmsg_data(nlh);
3681         if (ifm->ifa_index)
3682                 dev = __dev_get_by_index(net, ifm->ifa_index);
3683
3684         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3685                 err = -EADDRNOTAVAIL;
3686                 goto errout;
3687         }
3688
3689         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3690                 err = -ENOBUFS;
3691                 goto errout_ifa;
3692         }
3693
3694         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3695                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3696         if (err < 0) {
3697                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3698                 WARN_ON(err == -EMSGSIZE);
3699                 kfree_skb(skb);
3700                 goto errout_ifa;
3701         }
3702         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3703 errout_ifa:
3704         in6_ifa_put(ifa);
3705 errout:
3706         return err;
3707 }
3708
3709 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3710 {
3711         struct sk_buff *skb;
3712         struct net *net = dev_net(ifa->idev->dev);
3713         int err = -ENOBUFS;
3714
3715         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3716         if (skb == NULL)
3717                 goto errout;
3718
3719         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3720         if (err < 0) {
3721                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3722                 WARN_ON(err == -EMSGSIZE);
3723                 kfree_skb(skb);
3724                 goto errout;
3725         }
3726         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3727         return;
3728 errout:
3729         if (err < 0)
3730                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3731 }
3732
3733 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3734                                 __s32 *array, int bytes)
3735 {
3736         BUG_ON(bytes < (DEVCONF_MAX * 4));
3737
3738         memset(array, 0, bytes);
3739         array[DEVCONF_FORWARDING] = cnf->forwarding;
3740         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3741         array[DEVCONF_MTU6] = cnf->mtu6;
3742         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3743         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3744         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3745         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3746         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3747         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3748         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3749         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3750 #ifdef CONFIG_IPV6_PRIVACY
3751         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3752         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3753         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3754         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3755         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3756 #endif
3757         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3758         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3759         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3760 #ifdef CONFIG_IPV6_ROUTER_PREF
3761         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3762         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3763 #ifdef CONFIG_IPV6_ROUTE_INFO
3764         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3765 #endif
3766 #endif
3767         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3768         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3769 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3770         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3771 #endif
3772 #ifdef CONFIG_IPV6_MROUTE
3773         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3774 #endif
3775         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3776         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3777         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3778 }
3779
3780 static inline size_t inet6_if_nlmsg_size(void)
3781 {
3782         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3783                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3784                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3785                + nla_total_size(4) /* IFLA_MTU */
3786                + nla_total_size(4) /* IFLA_LINK */
3787                + nla_total_size( /* IFLA_PROTINFO */
3788                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3789                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3790                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3791                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3792                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3793                  );
3794 }
3795
3796 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3797                                       int items, int bytes)
3798 {
3799         int i;
3800         int pad = bytes - sizeof(u64) * items;
3801         BUG_ON(pad < 0);
3802
3803         /* Use put_unaligned() because stats may not be aligned for u64. */
3804         put_unaligned(items, &stats[0]);
3805         for (i = 1; i < items; i++)
3806                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3807
3808         memset(&stats[items], 0, pad);
3809 }
3810
3811 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3812                              int bytes)
3813 {
3814         switch(attrtype) {
3815         case IFLA_INET6_STATS:
3816                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3817                 break;
3818         case IFLA_INET6_ICMP6STATS:
3819                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3820                 break;
3821         }
3822 }
3823
3824 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3825                              u32 pid, u32 seq, int event, unsigned int flags)
3826 {
3827         struct net_device *dev = idev->dev;
3828         struct nlattr *nla;
3829         struct ifinfomsg *hdr;
3830         struct nlmsghdr *nlh;
3831         void *protoinfo;
3832         struct ifla_cacheinfo ci;
3833
3834         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3835         if (nlh == NULL)
3836                 return -EMSGSIZE;
3837
3838         hdr = nlmsg_data(nlh);
3839         hdr->ifi_family = AF_INET6;
3840         hdr->__ifi_pad = 0;
3841         hdr->ifi_type = dev->type;
3842         hdr->ifi_index = dev->ifindex;
3843         hdr->ifi_flags = dev_get_flags(dev);
3844         hdr->ifi_change = 0;
3845
3846         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3847
3848         if (dev->addr_len)
3849                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3850
3851         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3852         if (dev->ifindex != dev->iflink)
3853                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3854
3855         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3856         if (protoinfo == NULL)
3857                 goto nla_put_failure;
3858
3859         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3860
3861         ci.max_reasm_len = IPV6_MAXPLEN;
3862         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3863                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3864         ci.reachable_time = idev->nd_parms->reachable_time;
3865         ci.retrans_time = idev->nd_parms->retrans_time;
3866         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3867
3868         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3869         if (nla == NULL)
3870                 goto nla_put_failure;
3871         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3872
3873         /* XXX - MC not implemented */
3874
3875         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3876         if (nla == NULL)
3877                 goto nla_put_failure;
3878         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3879
3880         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3881         if (nla == NULL)
3882                 goto nla_put_failure;
3883         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3884
3885         nla_nest_end(skb, protoinfo);
3886         return nlmsg_end(skb, nlh);
3887
3888 nla_put_failure:
3889         nlmsg_cancel(skb, nlh);
3890         return -EMSGSIZE;
3891 }
3892
3893 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3894 {
3895         struct net *net = sock_net(skb->sk);
3896         int h, s_h;
3897         int idx = 0, s_idx;
3898         struct net_device *dev;
3899         struct inet6_dev *idev;
3900         struct hlist_head *head;
3901         struct hlist_node *node;
3902
3903         s_h = cb->args[0];
3904         s_idx = cb->args[1];
3905
3906         rcu_read_lock();
3907         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3908                 idx = 0;
3909                 head = &net->dev_index_head[h];
3910                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3911                         if (idx < s_idx)
3912                                 goto cont;
3913                         idev = __in6_dev_get(dev);
3914                         if (!idev)
3915                                 goto cont;
3916                         if (inet6_fill_ifinfo(skb, idev,
3917                                               NETLINK_CB(cb->skb).pid,
3918                                               cb->nlh->nlmsg_seq,
3919                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3920                                 goto out;
3921 cont:
3922                         idx++;
3923                 }
3924         }
3925 out:
3926         rcu_read_unlock();
3927         cb->args[1] = idx;
3928         cb->args[0] = h;
3929
3930         return skb->len;
3931 }
3932
3933 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3934 {
3935         struct sk_buff *skb;
3936         struct net *net = dev_net(idev->dev);
3937         int err = -ENOBUFS;
3938
3939         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3940         if (skb == NULL)
3941                 goto errout;
3942
3943         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3944         if (err < 0) {
3945                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3946                 WARN_ON(err == -EMSGSIZE);
3947                 kfree_skb(skb);
3948                 goto errout;
3949         }
3950         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3951         return;
3952 errout:
3953         if (err < 0)
3954                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3955 }
3956
3957 static inline size_t inet6_prefix_nlmsg_size(void)
3958 {
3959         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3960                + nla_total_size(sizeof(struct in6_addr))
3961                + nla_total_size(sizeof(struct prefix_cacheinfo));
3962 }
3963
3964 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3965                              struct prefix_info *pinfo, u32 pid, u32 seq,
3966                              int event, unsigned int flags)
3967 {
3968         struct prefixmsg *pmsg;
3969         struct nlmsghdr *nlh;
3970         struct prefix_cacheinfo ci;
3971
3972         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3973         if (nlh == NULL)
3974                 return -EMSGSIZE;
3975
3976         pmsg = nlmsg_data(nlh);
3977         pmsg->prefix_family = AF_INET6;
3978         pmsg->prefix_pad1 = 0;
3979         pmsg->prefix_pad2 = 0;
3980         pmsg->prefix_ifindex = idev->dev->ifindex;
3981         pmsg->prefix_len = pinfo->prefix_len;
3982         pmsg->prefix_type = pinfo->type;
3983         pmsg->prefix_pad3 = 0;
3984         pmsg->prefix_flags = 0;
3985         if (pinfo->onlink)
3986                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3987         if (pinfo->autoconf)
3988                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3989
3990         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3991
3992         ci.preferred_time = ntohl(pinfo->prefered);
3993         ci.valid_time = ntohl(pinfo->valid);
3994         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3995
3996         return nlmsg_end(skb, nlh);
3997
3998 nla_put_failure:
3999         nlmsg_cancel(skb, nlh);
4000         return -EMSGSIZE;
4001 }
4002
4003 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4004                          struct prefix_info *pinfo)
4005 {
4006         struct sk_buff *skb;
4007         struct net *net = dev_net(idev->dev);
4008         int err = -ENOBUFS;
4009
4010         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4011         if (skb == NULL)
4012                 goto errout;
4013
4014         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4015         if (err < 0) {
4016                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4017                 WARN_ON(err == -EMSGSIZE);
4018                 kfree_skb(skb);
4019                 goto errout;
4020         }
4021         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4022         return;
4023 errout:
4024         if (err < 0)
4025                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4026 }
4027
4028 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4029 {
4030         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4031
4032         switch (event) {
4033         case RTM_NEWADDR:
4034                 /*
4035                  * If the address was optimistic
4036                  * we inserted the route at the start of
4037                  * our DAD process, so we don't need
4038                  * to do it again
4039                  */
4040                 if (!(ifp->rt->rt6i_node))
4041                         ip6_ins_rt(ifp->rt);
4042                 if (ifp->idev->cnf.forwarding)
4043                         addrconf_join_anycast(ifp);
4044                 break;
4045         case RTM_DELADDR:
4046                 if (ifp->idev->cnf.forwarding)
4047                         addrconf_leave_anycast(ifp);
4048                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4049                 dst_hold(&ifp->rt->u.dst);
4050                 if (ip6_del_rt(ifp->rt))
4051                         dst_free(&ifp->rt->u.dst);
4052                 break;
4053         }
4054 }
4055
4056 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4057 {
4058         rcu_read_lock_bh();
4059         if (likely(ifp->idev->dead == 0))
4060                 __ipv6_ifa_notify(event, ifp);
4061         rcu_read_unlock_bh();
4062 }
4063
4064 #ifdef CONFIG_SYSCTL
4065
4066 static
4067 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4068                            void __user *buffer, size_t *lenp, loff_t *ppos)
4069 {
4070         int *valp = ctl->data;
4071         int val = *valp;
4072         loff_t pos = *ppos;
4073         int ret;
4074
4075         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4076
4077         if (write)
4078                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4079         if (ret)
4080                 *ppos = pos;
4081         return ret;
4082 }
4083
4084 static void dev_disable_change(struct inet6_dev *idev)
4085 {
4086         if (!idev || !idev->dev)
4087                 return;
4088
4089         if (idev->cnf.disable_ipv6)
4090                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4091         else
4092                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4093 }
4094
4095 static void addrconf_disable_change(struct net *net, __s32 newf)
4096 {
4097         struct net_device *dev;
4098         struct inet6_dev *idev;
4099
4100         rcu_read_lock();
4101         for_each_netdev_rcu(net, dev) {
4102                 idev = __in6_dev_get(dev);
4103                 if (idev) {
4104                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4105                         idev->cnf.disable_ipv6 = newf;
4106                         if (changed)
4107                                 dev_disable_change(idev);
4108                 }
4109         }
4110         rcu_read_unlock();
4111 }
4112
4113 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4114 {
4115         struct net *net;
4116
4117         net = (struct net *)table->extra2;
4118
4119         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4120                 return 0;
4121
4122         if (!rtnl_trylock()) {
4123                 /* Restore the original values before restarting */
4124                 *p = old;
4125                 return restart_syscall();
4126         }
4127
4128         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4129                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4130                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4131                 addrconf_disable_change(net, newf);
4132         } else if ((!*p) ^ (!old))
4133                 dev_disable_change((struct inet6_dev *)table->extra1);
4134
4135         rtnl_unlock();
4136         return 0;
4137 }
4138
4139 static
4140 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4141                             void __user *buffer, size_t *lenp, loff_t *ppos)
4142 {
4143         int *valp = ctl->data;
4144         int val = *valp;
4145         loff_t pos = *ppos;
4146         int ret;
4147
4148         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4149
4150         if (write)
4151                 ret = addrconf_disable_ipv6(ctl, valp, val);
4152         if (ret)
4153                 *ppos = pos;
4154         return ret;
4155 }
4156
4157 static struct addrconf_sysctl_table
4158 {
4159         struct ctl_table_header *sysctl_header;
4160         ctl_table addrconf_vars[DEVCONF_MAX+1];
4161         char *dev_name;
4162 } addrconf_sysctl __read_mostly = {
4163         .sysctl_header = NULL,
4164         .addrconf_vars = {
4165                 {
4166                         .procname       =       "forwarding",
4167                         .data           =       &ipv6_devconf.forwarding,
4168                         .maxlen         =       sizeof(int),
4169                         .mode           =       0644,
4170                         .proc_handler   =       addrconf_sysctl_forward,
4171                 },
4172                 {
4173                         .procname       =       "hop_limit",
4174                         .data           =       &ipv6_devconf.hop_limit,
4175                         .maxlen         =       sizeof(int),
4176                         .mode           =       0644,
4177                         .proc_handler   =       proc_dointvec,
4178                 },
4179                 {
4180                         .procname       =       "mtu",
4181                         .data           =       &ipv6_devconf.mtu6,
4182                         .maxlen         =       sizeof(int),
4183                         .mode           =       0644,
4184                         .proc_handler   =       proc_dointvec,
4185                 },
4186                 {
4187                         .procname       =       "accept_ra",
4188                         .data           =       &ipv6_devconf.accept_ra,
4189                         .maxlen         =       sizeof(int),
4190                         .mode           =       0644,
4191                         .proc_handler   =       proc_dointvec,
4192                 },
4193                 {
4194                         .procname       =       "accept_redirects",
4195                         .data           =       &ipv6_devconf.accept_redirects,
4196                         .maxlen         =       sizeof(int),
4197                         .mode           =       0644,
4198                         .proc_handler   =       proc_dointvec,
4199                 },
4200                 {
4201                         .procname       =       "autoconf",
4202                         .data           =       &ipv6_devconf.autoconf,
4203                         .maxlen         =       sizeof(int),
4204                         .mode           =       0644,
4205                         .proc_handler   =       proc_dointvec,
4206                 },
4207                 {
4208                         .procname       =       "dad_transmits",
4209                         .data           =       &ipv6_devconf.dad_transmits,
4210                         .maxlen         =       sizeof(int),
4211                         .mode           =       0644,
4212                         .proc_handler   =       proc_dointvec,
4213                 },
4214                 {
4215                         .procname       =       "router_solicitations",
4216                         .data           =       &ipv6_devconf.rtr_solicits,
4217                         .maxlen         =       sizeof(int),
4218                         .mode           =       0644,
4219                         .proc_handler   =       proc_dointvec,
4220                 },
4221                 {
4222                         .procname       =       "router_solicitation_interval",
4223                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       proc_dointvec_jiffies,
4227                 },
4228                 {
4229                         .procname       =       "router_solicitation_delay",
4230                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4231                         .maxlen         =       sizeof(int),
4232                         .mode           =       0644,
4233                         .proc_handler   =       proc_dointvec_jiffies,
4234                 },
4235                 {
4236                         .procname       =       "force_mld_version",
4237                         .data           =       &ipv6_devconf.force_mld_version,
4238                         .maxlen         =       sizeof(int),
4239                         .mode           =       0644,
4240                         .proc_handler   =       proc_dointvec,
4241                 },
4242 #ifdef CONFIG_IPV6_PRIVACY
4243                 {
4244                         .procname       =       "use_tempaddr",
4245                         .data           =       &ipv6_devconf.use_tempaddr,
4246                         .maxlen         =       sizeof(int),
4247                         .mode           =       0644,
4248                         .proc_handler   =       proc_dointvec,
4249                 },
4250                 {
4251                         .procname       =       "temp_valid_lft",
4252                         .data           =       &ipv6_devconf.temp_valid_lft,
4253                         .maxlen         =       sizeof(int),
4254                         .mode           =       0644,
4255                         .proc_handler   =       proc_dointvec,
4256                 },
4257                 {
4258                         .procname       =       "temp_prefered_lft",
4259                         .data           =       &ipv6_devconf.temp_prefered_lft,
4260                         .maxlen         =       sizeof(int),
4261                         .mode           =       0644,
4262                         .proc_handler   =       proc_dointvec,
4263                 },
4264                 {
4265                         .procname       =       "regen_max_retry",
4266                         .data           =       &ipv6_devconf.regen_max_retry,
4267                         .maxlen         =       sizeof(int),
4268                         .mode           =       0644,
4269                         .proc_handler   =       proc_dointvec,
4270                 },
4271                 {
4272                         .procname       =       "max_desync_factor",
4273                         .data           =       &ipv6_devconf.max_desync_factor,
4274                         .maxlen         =       sizeof(int),
4275                         .mode           =       0644,
4276                         .proc_handler   =       proc_dointvec,
4277                 },
4278 #endif
4279                 {
4280                         .procname       =       "max_addresses",
4281                         .data           =       &ipv6_devconf.max_addresses,
4282                         .maxlen         =       sizeof(int),
4283                         .mode           =       0644,
4284                         .proc_handler   =       proc_dointvec,
4285                 },
4286                 {
4287                         .procname       =       "accept_ra_defrtr",
4288                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4289                         .maxlen         =       sizeof(int),
4290                         .mode           =       0644,
4291                         .proc_handler   =       proc_dointvec,
4292                 },
4293                 {
4294                         .procname       =       "accept_ra_pinfo",
4295                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4296                         .maxlen         =       sizeof(int),
4297                         .mode           =       0644,
4298                         .proc_handler   =       proc_dointvec,
4299                 },
4300 #ifdef CONFIG_IPV6_ROUTER_PREF
4301                 {
4302                         .procname       =       "accept_ra_rtr_pref",
4303                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4304                         .maxlen         =       sizeof(int),
4305                         .mode           =       0644,
4306                         .proc_handler   =       proc_dointvec,
4307                 },
4308                 {
4309                         .procname       =       "router_probe_interval",
4310                         .data           =       &ipv6_devconf.rtr_probe_interval,
4311                         .maxlen         =       sizeof(int),
4312                         .mode           =       0644,
4313                         .proc_handler   =       proc_dointvec_jiffies,
4314                 },
4315 #ifdef CONFIG_IPV6_ROUTE_INFO
4316                 {
4317                         .procname       =       "accept_ra_rt_info_max_plen",
4318                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4319                         .maxlen         =       sizeof(int),
4320                         .mode           =       0644,
4321                         .proc_handler   =       proc_dointvec,
4322                 },
4323 #endif
4324 #endif
4325                 {
4326                         .procname       =       "proxy_ndp",
4327                         .data           =       &ipv6_devconf.proxy_ndp,
4328                         .maxlen         =       sizeof(int),
4329                         .mode           =       0644,
4330                         .proc_handler   =       proc_dointvec,
4331                 },
4332                 {
4333                         .procname       =       "accept_source_route",
4334                         .data           =       &ipv6_devconf.accept_source_route,
4335                         .maxlen         =       sizeof(int),
4336                         .mode           =       0644,
4337                         .proc_handler   =       proc_dointvec,
4338                 },
4339 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4340                 {
4341                         .procname       =       "optimistic_dad",
4342                         .data           =       &ipv6_devconf.optimistic_dad,
4343                         .maxlen         =       sizeof(int),
4344                         .mode           =       0644,
4345                         .proc_handler   =       proc_dointvec,
4346
4347                 },
4348 #endif
4349 #ifdef CONFIG_IPV6_MROUTE
4350                 {
4351                         .procname       =       "mc_forwarding",
4352                         .data           =       &ipv6_devconf.mc_forwarding,
4353                         .maxlen         =       sizeof(int),
4354                         .mode           =       0444,
4355                         .proc_handler   =       proc_dointvec,
4356                 },
4357 #endif
4358                 {
4359                         .procname       =       "disable_ipv6",
4360                         .data           =       &ipv6_devconf.disable_ipv6,
4361                         .maxlen         =       sizeof(int),
4362                         .mode           =       0644,
4363                         .proc_handler   =       addrconf_sysctl_disable,
4364                 },
4365                 {
4366                         .procname       =       "accept_dad",
4367                         .data           =       &ipv6_devconf.accept_dad,
4368                         .maxlen         =       sizeof(int),
4369                         .mode           =       0644,
4370                         .proc_handler   =       proc_dointvec,
4371                 },
4372                 {
4373                         .procname       = "force_tllao",
4374                         .data           = &ipv6_devconf.force_tllao,
4375                         .maxlen         = sizeof(int),
4376                         .mode           = 0644,
4377                         .proc_handler   = proc_dointvec
4378                 },
4379                 {
4380                         /* sentinel */
4381                 }
4382         },
4383 };
4384
4385 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4386                 struct inet6_dev *idev, struct ipv6_devconf *p)
4387 {
4388         int i;
4389         struct addrconf_sysctl_table *t;
4390
4391 #define ADDRCONF_CTL_PATH_DEV   3
4392
4393         struct ctl_path addrconf_ctl_path[] = {
4394                 { .procname = "net", },
4395                 { .procname = "ipv6", },
4396                 { .procname = "conf", },
4397                 { /* to be set */ },
4398                 { },
4399         };
4400
4401
4402         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4403         if (t == NULL)
4404                 goto out;
4405
4406         for (i=0; t->addrconf_vars[i].data; i++) {
4407                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4408                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4409                 t->addrconf_vars[i].extra2 = net;
4410         }
4411
4412         /*
4413          * Make a copy of dev_name, because '.procname' is regarded as const
4414          * by sysctl and we wouldn't want anyone to change it under our feet
4415          * (see SIOCSIFNAME).
4416          */
4417         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4418         if (!t->dev_name)
4419                 goto free;
4420
4421         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4422
4423         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4424                         t->addrconf_vars);
4425         if (t->sysctl_header == NULL)
4426                 goto free_procname;
4427
4428         p->sysctl = t;
4429         return 0;
4430
4431 free_procname:
4432         kfree(t->dev_name);
4433 free:
4434         kfree(t);
4435 out:
4436         return -ENOBUFS;
4437 }
4438
4439 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4440 {
4441         struct addrconf_sysctl_table *t;
4442
4443         if (p->sysctl == NULL)
4444                 return;
4445
4446         t = p->sysctl;
4447         p->sysctl = NULL;
4448         unregister_sysctl_table(t->sysctl_header);
4449         kfree(t->dev_name);
4450         kfree(t);
4451 }
4452
4453 static void addrconf_sysctl_register(struct inet6_dev *idev)
4454 {
4455         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4456                               &ndisc_ifinfo_sysctl_change);
4457         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4458                                         idev, &idev->cnf);
4459 }
4460
4461 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4462 {
4463         __addrconf_sysctl_unregister(&idev->cnf);
4464         neigh_sysctl_unregister(idev->nd_parms);
4465 }
4466
4467
4468 #endif
4469
4470 static int __net_init addrconf_init_net(struct net *net)
4471 {
4472         int err;
4473         struct ipv6_devconf *all, *dflt;
4474
4475         err = -ENOMEM;
4476         all = &ipv6_devconf;
4477         dflt = &ipv6_devconf_dflt;
4478
4479         if (!net_eq(net, &init_net)) {
4480                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4481                 if (all == NULL)
4482                         goto err_alloc_all;
4483
4484                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4485                 if (dflt == NULL)
4486                         goto err_alloc_dflt;
4487         } else {
4488                 /* these will be inherited by all namespaces */
4489                 dflt->autoconf = ipv6_defaults.autoconf;
4490                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4491         }
4492
4493         net->ipv6.devconf_all = all;
4494         net->ipv6.devconf_dflt = dflt;
4495
4496 #ifdef CONFIG_SYSCTL
4497         err = __addrconf_sysctl_register(net, "all", NULL, all);
4498         if (err < 0)
4499                 goto err_reg_all;
4500
4501         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4502         if (err < 0)
4503                 goto err_reg_dflt;
4504 #endif
4505         return 0;
4506
4507 #ifdef CONFIG_SYSCTL
4508 err_reg_dflt:
4509         __addrconf_sysctl_unregister(all);
4510 err_reg_all:
4511         kfree(dflt);
4512 #endif
4513 err_alloc_dflt:
4514         kfree(all);
4515 err_alloc_all:
4516         return err;
4517 }
4518
4519 static void __net_exit addrconf_exit_net(struct net *net)
4520 {
4521 #ifdef CONFIG_SYSCTL
4522         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4523         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4524 #endif
4525         if (!net_eq(net, &init_net)) {
4526                 kfree(net->ipv6.devconf_dflt);
4527                 kfree(net->ipv6.devconf_all);
4528         }
4529 }
4530
4531 static struct pernet_operations addrconf_ops = {
4532         .init = addrconf_init_net,
4533         .exit = addrconf_exit_net,
4534 };
4535
4536 /*
4537  *      Device notifier
4538  */
4539
4540 int register_inet6addr_notifier(struct notifier_block *nb)
4541 {
4542         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4543 }
4544
4545 EXPORT_SYMBOL(register_inet6addr_notifier);
4546
4547 int unregister_inet6addr_notifier(struct notifier_block *nb)
4548 {
4549         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4550 }
4551
4552 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4553
4554 /*
4555  *      Init / cleanup code
4556  */
4557
4558 int __init addrconf_init(void)
4559 {
4560         int err;
4561
4562         if ((err = ipv6_addr_label_init()) < 0) {
4563                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4564                         err);
4565                 return err;
4566         }
4567
4568         register_pernet_subsys(&addrconf_ops);
4569
4570         /* The addrconf netdev notifier requires that loopback_dev
4571          * has it's ipv6 private information allocated and setup
4572          * before it can bring up and give link-local addresses
4573          * to other devices which are up.
4574          *
4575          * Unfortunately, loopback_dev is not necessarily the first
4576          * entry in the global dev_base list of net devices.  In fact,
4577          * it is likely to be the very last entry on that list.
4578          * So this causes the notifier registry below to try and
4579          * give link-local addresses to all devices besides loopback_dev
4580          * first, then loopback_dev, which cases all the non-loopback_dev
4581          * devices to fail to get a link-local address.
4582          *
4583          * So, as a temporary fix, allocate the ipv6 structure for
4584          * loopback_dev first by hand.
4585          * Longer term, all of the dependencies ipv6 has upon the loopback
4586          * device and it being up should be removed.
4587          */
4588         rtnl_lock();
4589         if (!ipv6_add_dev(init_net.loopback_dev))
4590                 err = -ENOMEM;
4591         rtnl_unlock();
4592         if (err)
4593                 goto errlo;
4594
4595         register_netdevice_notifier(&ipv6_dev_notf);
4596
4597         addrconf_verify(0);
4598
4599         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4600         if (err < 0)
4601                 goto errout;
4602
4603         /* Only the first call to __rtnl_register can fail */
4604         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4605         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4606         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4607         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4608         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4609
4610         ipv6_addr_label_rtnl_register();
4611
4612         return 0;
4613 errout:
4614         unregister_netdevice_notifier(&ipv6_dev_notf);
4615 errlo:
4616         unregister_pernet_subsys(&addrconf_ops);
4617
4618         return err;
4619 }
4620
4621 void addrconf_cleanup(void)
4622 {
4623         struct inet6_ifaddr *ifa;
4624         struct net_device *dev;
4625         int i;
4626
4627         unregister_netdevice_notifier(&ipv6_dev_notf);
4628         unregister_pernet_subsys(&addrconf_ops);
4629
4630         rtnl_lock();
4631
4632         /* clean dev list */
4633         for_each_netdev(&init_net, dev) {
4634                 if (__in6_dev_get(dev) == NULL)
4635                         continue;
4636                 addrconf_ifdown(dev, 1);
4637         }
4638         addrconf_ifdown(init_net.loopback_dev, 2);
4639
4640         /*
4641          *      Check hash table.
4642          */
4643         write_lock_bh(&addrconf_hash_lock);
4644         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4645                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4646                         struct inet6_ifaddr *bifa;
4647
4648                         bifa = ifa;
4649                         ifa = ifa->lst_next;
4650                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4651                         /* Do not free it; something is wrong.
4652                            Now we can investigate it with debugger.
4653                          */
4654                 }
4655         }
4656         write_unlock_bh(&addrconf_hash_lock);
4657
4658         del_timer(&addr_chk_timer);
4659         rtnl_unlock();
4660 }