net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static void addrconf_bonding_change(struct net_device *dev,
 141                                     unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 405         if ((dev->flags&IFF_LOOPBACK) ||
 406             dev->type == ARPHRD_TUNNEL ||
 407             dev->type == ARPHRD_TUNNEL6 ||
 408             dev->type == ARPHRD_SIT ||
 409             dev->type == ARPHRD_NONE) {
 410                 printk(KERN_INFO
 411                        "%s: Disabled Privacy Extensions\n",
 412                        dev->name);
 413                 ndev->cnf.use_tempaddr = -1;
 414         } else {
 415                 in6_dev_hold(ndev);
 416                 ipv6_regen_rndid((unsigned long) ndev);
 417         }
 418 #endif
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         addrconf_sysctl_register(ndev);
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join all-node multicast group */
 430         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 431
 432         return ndev;
 433 }
 434
 435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 436 {
 437         struct inet6_dev *idev;
 438
 439         ASSERT_RTNL();
 440
 441         if ((idev = __in6_dev_get(dev)) == NULL) {
 442                 if ((idev = ipv6_add_dev(dev)) == NULL)
 443                         return NULL;
 444         }
 445
 446         if (dev->flags&IFF_UP)
 447                 ipv6_mc_up(idev);
 448         return idev;
 449 }
 450
 451 #ifdef CONFIG_SYSCTL
 452 static void dev_forward_change(struct inet6_dev *idev)
 453 {
 454         struct net_device *dev;
 455         struct inet6_ifaddr *ifa;
 456
 457         if (!idev)
 458                 return;
 459         dev = idev->dev;
 460         if (idev->cnf.forwarding)
 461                 dev_disable_lro(dev);
 462         if (dev && (dev->flags & IFF_MULTICAST)) {
 463                 if (idev->cnf.forwarding)
 464                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 465                 else
 466                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 467         }
 468         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 469                 if (ifa->flags&IFA_F_TENTATIVE)
 470                         continue;
 471                 if (idev->cnf.forwarding)
 472                         addrconf_join_anycast(ifa);
 473                 else
 474                         addrconf_leave_anycast(ifa);
 475         }
 476 }
 477
 478
 479 static void addrconf_forward_change(struct net *net, __s32 newf)
 480 {
 481         struct net_device *dev;
 482         struct inet6_dev *idev;
 483
 484         rcu_read_lock();
 485         for_each_netdev_rcu(net, dev) {
 486                 idev = __in6_dev_get(dev);
 487                 if (idev) {
 488                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 489                         idev->cnf.forwarding = newf;
 490                         if (changed)
 491                                 dev_forward_change(idev);
 492                 }
 493         }
 494         rcu_read_unlock();
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock()) {
 506                 /* Restore the original values before restarting */
 507                 *p = old;
 508                 return restart_syscall();
 509         }
 510
 511         if (p == &net->ipv6.devconf_all->forwarding) {
 512                 __s32 newf = net->ipv6.devconf_all->forwarding;
 513                 net->ipv6.devconf_dflt->forwarding = newf;
 514                 addrconf_forward_change(net, newf);
 515         } else if ((!*p) ^ (!old))
 516                 dev_forward_change((struct inet6_dev *)table->extra1);
 517         rtnl_unlock();
 518
 519         if (*p)
 520                 rt6_purge_dflt_routers(net);
 521         return 1;
 522 }
 523 #endif
 524
 525 /* Nobody refers to this ifaddr, destroy it */
 526
 527 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 528 {
 529         WARN_ON(ifp->if_next != NULL);
 530         WARN_ON(ifp->lst_next != NULL);
 531
 532 #ifdef NET_REFCNT_DEBUG
 533         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 534 #endif
 535
 536         in6_dev_put(ifp->idev);
 537
 538         if (del_timer(&ifp->timer))
 539                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 540
 541         if (!ifp->dead) {
 542                 printk("Freeing alive inet6 address %p\n", ifp);
 543                 return;
 544         }
 545         dst_release(&ifp->rt->u.dst);
 546
 547         kfree(ifp);
 548 }
 549
 550 static void
 551 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 552 {
 553         struct inet6_ifaddr *ifa, **ifap;
 554         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 555
 556         /*
 557          * Each device address list is sorted in order of scope -
 558          * global before linklocal.
 559          */
 560         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 561              ifap = &ifa->if_next) {
 562                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 563                         break;
 564         }
 565
 566         ifp->if_next = *ifap;
 567         *ifap = ifp;
 568 }
 569
 570 /*
 571  *      Hash function taken from net_alias.c
 572  */
 573 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 574 {
 575         __u32 word;
 576
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581
 582         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 583         word ^= (word >> 16);
 584         word ^= (word >> 8);
 585
 586         return ((word ^ (word >> 4)) & 0x0f);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         write_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         init_timer(&ifa->timer);
 645         ifa->timer.data = (unsigned long) ifa;
 646         ifa->scope = scope;
 647         ifa->prefix_len = pfxlen;
 648         ifa->flags = flags | IFA_F_TENTATIVE;
 649         ifa->cstamp = ifa->tstamp = jiffies;
 650
 651         ifa->rt = rt;
 652
 653         /*
 654          * part one of RFC 4429, section 3.3
 655          * We should not configure an address as
 656          * optimistic if we do not yet know the link
 657          * layer address of our nexhop router
 658          */
 659
 660         if (rt->rt6i_nexthop == NULL)
 661                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 662
 663         ifa->idev = idev;
 664         in6_dev_hold(idev);
 665         /* For caller */
 666         in6_ifa_hold(ifa);
 667
 668         /* Add to big hash table */
 669         hash = ipv6_addr_hash(addr);
 670
 671         ifa->lst_next = inet6_addr_lst[hash];
 672         inet6_addr_lst[hash] = ifa;
 673         in6_ifa_hold(ifa);
 674         write_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 ifa->tmp_next = idev->tempaddr_list;
 683                 idev->tempaddr_list = ifa;
 684                 in6_ifa_hold(ifa);
 685         }
 686 #endif
 687
 688         in6_ifa_hold(ifa);
 689         write_unlock(&idev->lock);
 690 out2:
 691         rcu_read_unlock_bh();
 692
 693         if (likely(err == 0))
 694                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 695         else {
 696                 kfree(ifa);
 697                 ifa = ERR_PTR(err);
 698         }
 699
 700         return ifa;
 701 out:
 702         write_unlock(&addrconf_hash_lock);
 703         goto out2;
 704 }
 705
 706 /* This function wants to get referenced ifp and releases it before return */
 707
 708 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 709 {
 710         struct inet6_ifaddr *ifa, **ifap;
 711         struct inet6_dev *idev = ifp->idev;
 712         int hash;
 713         int deleted = 0, onlink = 0;
 714         unsigned long expires = jiffies;
 715
 716         hash = ipv6_addr_hash(&ifp->addr);
 717
 718         ifp->dead = 1;
 719
 720         write_lock_bh(&addrconf_hash_lock);
 721         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 722              ifap = &ifa->lst_next) {
 723                 if (ifa == ifp) {
 724                         *ifap = ifa->lst_next;
 725                         __in6_ifa_put(ifp);
 726                         ifa->lst_next = NULL;
 727                         break;
 728                 }
 729         }
 730         write_unlock_bh(&addrconf_hash_lock);
 731
 732         write_lock_bh(&idev->lock);
 733 #ifdef CONFIG_IPV6_PRIVACY
 734         if (ifp->flags&IFA_F_TEMPORARY) {
 735                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 736                      ifap = &ifa->tmp_next) {
 737                         if (ifa == ifp) {
 738                                 *ifap = ifa->tmp_next;
 739                                 if (ifp->ifpub) {
 740                                         in6_ifa_put(ifp->ifpub);
 741                                         ifp->ifpub = NULL;
 742                                 }
 743                                 __in6_ifa_put(ifp);
 744                                 ifa->tmp_next = NULL;
 745                                 break;
 746                         }
 747                 }
 748         }
 749 #endif
 750
 751         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 752                 if (ifa == ifp) {
 753                         *ifap = ifa->if_next;
 754                         __in6_ifa_put(ifp);
 755                         ifa->if_next = NULL;
 756                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 757                                 break;
 758                         deleted = 1;
 759                         continue;
 760                 } else if (ifp->flags & IFA_F_PERMANENT) {
 761                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 762                                               ifp->prefix_len)) {
 763                                 if (ifa->flags & IFA_F_PERMANENT) {
 764                                         onlink = 1;
 765                                         if (deleted)
 766                                                 break;
 767                                 } else {
 768                                         unsigned long lifetime;
 769
 770                                         if (!onlink)
 771                                                 onlink = -1;
 772
 773                                         spin_lock(&ifa->lock);
 774
 775                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 776                                         /*
 777                                          * Note: Because this address is
 778                                          * not permanent, lifetime <
 779                                          * LONG_MAX / HZ here.
 780                                          */
 781                                         if (time_before(expires,
 782                                                         ifa->tstamp + lifetime * HZ))
 783                                                 expires = ifa->tstamp + lifetime * HZ;
 784                                         spin_unlock(&ifa->lock);
 785                                 }
 786                         }
 787                 }
 788                 ifap = &ifa->if_next;
 789         }
 790         write_unlock_bh(&idev->lock);
 791
 792         addrconf_del_timer(ifp);
 793
 794         ipv6_ifa_notify(RTM_DELADDR, ifp);
 795
 796         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 797
 798         /*
 799          * Purge or update corresponding prefix
 800          *
 801          * 1) we don't purge prefix here if address was not permanent.
 802          *    prefix is managed by its own lifetime.
 803          * 2) if there're no addresses, delete prefix.
 804          * 3) if there're still other permanent address(es),
 805          *    corresponding prefix is still permanent.
 806          * 4) otherwise, update prefix lifetime to the
 807          *    longest valid lifetime among the corresponding
 808          *    addresses on the device.
 809          *    Note: subsequent RA will update lifetime.
 810          *
 811          * --yoshfuji
 812          */
 813         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 814                 struct in6_addr prefix;
 815                 struct rt6_info *rt;
 816                 struct net *net = dev_net(ifp->idev->dev);
 817                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 818                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 819
 820                 if (rt && addrconf_is_prefix_route(rt)) {
 821                         if (onlink == 0) {
 822                                 ip6_del_rt(rt);
 823                                 rt = NULL;
 824                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 825                                 rt->rt6i_expires = expires;
 826                                 rt->rt6i_flags |= RTF_EXPIRES;
 827                         }
 828                 }
 829                 dst_release(&rt->u.dst);
 830         }
 831
 832         in6_ifa_put(ifp);
 833 }
 834
 835 #ifdef CONFIG_IPV6_PRIVACY
 836 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 837 {
 838         struct inet6_dev *idev = ifp->idev;
 839         struct in6_addr addr, *tmpaddr;
 840         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 841         unsigned long regen_advance;
 842         int tmp_plen;
 843         int ret = 0;
 844         int max_addresses;
 845         u32 addr_flags;
 846
 847         write_lock(&idev->lock);
 848         if (ift) {
 849                 spin_lock_bh(&ift->lock);
 850                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 851                 spin_unlock_bh(&ift->lock);
 852                 tmpaddr = &addr;
 853         } else {
 854                 tmpaddr = NULL;
 855         }
 856 retry:
 857         in6_dev_hold(idev);
 858         if (idev->cnf.use_tempaddr <= 0) {
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_INFO
 861                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         spin_lock_bh(&ifp->lock);
 867         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 868                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 873                 in6_dev_put(idev);
 874                 ret = -1;
 875                 goto out;
 876         }
 877         in6_ifa_hold(ifp);
 878         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 879         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 880                 spin_unlock_bh(&ifp->lock);
 881                 write_unlock(&idev->lock);
 882                 printk(KERN_WARNING
 883                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 884                 in6_ifa_put(ifp);
 885                 in6_dev_put(idev);
 886                 ret = -1;
 887                 goto out;
 888         }
 889         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 890         tmp_valid_lft = min_t(__u32,
 891                               ifp->valid_lft,
 892                               idev->cnf.temp_valid_lft);
 893         tmp_prefered_lft = min_t(__u32,
 894                                  ifp->prefered_lft,
 895                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 896         tmp_plen = ifp->prefix_len;
 897         max_addresses = idev->cnf.max_addresses;
 898         tmp_cstamp = ifp->cstamp;
 899         tmp_tstamp = ifp->tstamp;
 900         spin_unlock_bh(&ifp->lock);
 901
 902         regen_advance = idev->cnf.regen_max_retry *
 903                         idev->cnf.dad_transmits *
 904                         idev->nd_parms->retrans_time / HZ;
 905         write_unlock(&idev->lock);
 906
 907         /* A temporary address is created only if this calculated Preferred
 908          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 909          * an implementation must not create a temporary address with a zero
 910          * Preferred Lifetime.
 911          */
 912         if (tmp_prefered_lft <= regen_advance) {
 913                 in6_ifa_put(ifp);
 914                 in6_dev_put(idev);
 915                 ret = -1;
 916                 goto out;
 917         }
 918
 919         addr_flags = IFA_F_TEMPORARY;
 920         /* set in addrconf_prefix_rcv() */
 921         if (ifp->flags & IFA_F_OPTIMISTIC)
 922                 addr_flags |= IFA_F_OPTIMISTIC;
 923
 924         ift = !max_addresses ||
 925               ipv6_count_addresses(idev) < max_addresses ?
 926                 ipv6_add_addr(idev, &addr, tmp_plen,
 927                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 928                               addr_flags) : NULL;
 929         if (!ift || IS_ERR(ift)) {
 930                 in6_ifa_put(ifp);
 931                 in6_dev_put(idev);
 932                 printk(KERN_INFO
 933                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 934                 tmpaddr = &addr;
 935                 write_lock(&idev->lock);
 936                 goto retry;
 937         }
 938
 939         spin_lock_bh(&ift->lock);
 940         ift->ifpub = ifp;
 941         ift->valid_lft = tmp_valid_lft;
 942         ift->prefered_lft = tmp_prefered_lft;
 943         ift->cstamp = tmp_cstamp;
 944         ift->tstamp = tmp_tstamp;
 945         spin_unlock_bh(&ift->lock);
 946
 947         addrconf_dad_start(ift, 0);
 948         in6_ifa_put(ift);
 949         in6_dev_put(idev);
 950 out:
 951         return ret;
 952 }
 953 #endif
 954
 955 /*
 956  *      Choose an appropriate source address (RFC3484)
 957  */
 958 enum {
 959         IPV6_SADDR_RULE_INIT = 0,
 960         IPV6_SADDR_RULE_LOCAL,
 961         IPV6_SADDR_RULE_SCOPE,
 962         IPV6_SADDR_RULE_PREFERRED,
 963 #ifdef CONFIG_IPV6_MIP6
 964         IPV6_SADDR_RULE_HOA,
 965 #endif
 966         IPV6_SADDR_RULE_OIF,
 967         IPV6_SADDR_RULE_LABEL,
 968 #ifdef CONFIG_IPV6_PRIVACY
 969         IPV6_SADDR_RULE_PRIVACY,
 970 #endif
 971         IPV6_SADDR_RULE_ORCHID,
 972         IPV6_SADDR_RULE_PREFIX,
 973         IPV6_SADDR_RULE_MAX
 974 };
 975
 976 struct ipv6_saddr_score {
 977         int                     rule;
 978         int                     addr_type;
 979         struct inet6_ifaddr     *ifa;
 980         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 981         int                     scopedist;
 982         int                     matchlen;
 983 };
 984
 985 struct ipv6_saddr_dst {
 986         const struct in6_addr *addr;
 987         int ifindex;
 988         int scope;
 989         int label;
 990         unsigned int prefs;
 991 };
 992
 993 static inline int ipv6_saddr_preferred(int type)
 994 {
 995         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 996                 return 1;
 997         return 0;
 998 }
 999
1000 static int ipv6_get_saddr_eval(struct net *net,
1001                                struct ipv6_saddr_score *score,
1002                                struct ipv6_saddr_dst *dst,
1003                                int i)
1004 {
1005         int ret;
1006
1007         if (i <= score->rule) {
1008                 switch (i) {
1009                 case IPV6_SADDR_RULE_SCOPE:
1010                         ret = score->scopedist;
1011                         break;
1012                 case IPV6_SADDR_RULE_PREFIX:
1013                         ret = score->matchlen;
1014                         break;
1015                 default:
1016                         ret = !!test_bit(i, score->scorebits);
1017                 }
1018                 goto out;
1019         }
1020
1021         switch (i) {
1022         case IPV6_SADDR_RULE_INIT:
1023                 /* Rule 0: remember if hiscore is not ready yet */
1024                 ret = !!score->ifa;
1025                 break;
1026         case IPV6_SADDR_RULE_LOCAL:
1027                 /* Rule 1: Prefer same address */
1028                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1029                 break;
1030         case IPV6_SADDR_RULE_SCOPE:
1031                 /* Rule 2: Prefer appropriate scope
1032                  *
1033                  *      ret
1034                  *       ^
1035                  *    -1 |  d 15
1036                  *    ---+--+-+---> scope
1037                  *       |
1038                  *       |             d is scope of the destination.
1039                  *  B-d  |  \
1040                  *       |   \      <- smaller scope is better if
1041                  *  B-15 |    \        if scope is enough for destinaion.
1042                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1043                  * d-C-1 | /
1044                  *       |/         <- greater is better
1045                  *   -C  /             if scope is not enough for destination.
1046                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1047                  *
1048                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1049                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1050                  * Assume B = 0 and we get C > 29.
1051                  */
1052                 ret = __ipv6_addr_src_scope(score->addr_type);
1053                 if (ret >= dst->scope)
1054                         ret = -ret;
1055                 else
1056                         ret -= 128;     /* 30 is enough */
1057                 score->scopedist = ret;
1058                 break;
1059         case IPV6_SADDR_RULE_PREFERRED:
1060                 /* Rule 3: Avoid deprecated and optimistic addresses */
1061                 ret = ipv6_saddr_preferred(score->addr_type) ||
1062                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1063                 break;
1064 #ifdef CONFIG_IPV6_MIP6
1065         case IPV6_SADDR_RULE_HOA:
1066             {
1067                 /* Rule 4: Prefer home address */
1068                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1069                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1070                 break;
1071             }
1072 #endif
1073         case IPV6_SADDR_RULE_OIF:
1074                 /* Rule 5: Prefer outgoing interface */
1075                 ret = (!dst->ifindex ||
1076                        dst->ifindex == score->ifa->idev->dev->ifindex);
1077                 break;
1078         case IPV6_SADDR_RULE_LABEL:
1079                 /* Rule 6: Prefer matching label */
1080                 ret = ipv6_addr_label(net,
1081                                       &score->ifa->addr, score->addr_type,
1082                                       score->ifa->idev->dev->ifindex) == dst->label;
1083                 break;
1084 #ifdef CONFIG_IPV6_PRIVACY
1085         case IPV6_SADDR_RULE_PRIVACY:
1086             {
1087                 /* Rule 7: Prefer public address
1088                  * Note: prefer temprary address if use_tempaddr >= 2
1089                  */
1090                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1091                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1092                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1093                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1094                 break;
1095             }
1096 #endif
1097         case IPV6_SADDR_RULE_ORCHID:
1098                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1099                  *          non-ORCHID vs non-ORCHID
1100                  */
1101                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1102                         ipv6_addr_orchid(dst->addr));
1103                 break;
1104         case IPV6_SADDR_RULE_PREFIX:
1105                 /* Rule 8: Use longest matching prefix */
1106                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1107                                                        dst->addr);
1108                 break;
1109         default:
1110                 ret = 0;
1111         }
1112
1113         if (ret)
1114                 __set_bit(i, score->scorebits);
1115         score->rule = i;
1116 out:
1117         return ret;
1118 }
1119
1120 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1121                        const struct in6_addr *daddr, unsigned int prefs,
1122                        struct in6_addr *saddr)
1123 {
1124         struct ipv6_saddr_score scores[2],
1125                                 *score = &scores[0], *hiscore = &scores[1];
1126         struct ipv6_saddr_dst dst;
1127         struct net_device *dev;
1128         int dst_type;
1129
1130         dst_type = __ipv6_addr_type(daddr);
1131         dst.addr = daddr;
1132         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1133         dst.scope = __ipv6_addr_src_scope(dst_type);
1134         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1135         dst.prefs = prefs;
1136
1137         hiscore->rule = -1;
1138         hiscore->ifa = NULL;
1139
1140         rcu_read_lock();
1141
1142         for_each_netdev_rcu(net, dev) {
1143                 struct inet6_dev *idev;
1144
1145                 /* Candidate Source Address (section 4)
1146                  *  - multicast and link-local destination address,
1147                  *    the set of candidate source address MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same link as the outgoing
1150                  *    interface.
1151                  * (- For site-local destination addresses, the
1152                  *    set of candidate source addresses MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same site as the outgoing
1155                  *    interface.)
1156                  */
1157                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159                     dst.ifindex && dev->ifindex != dst.ifindex)
1160                         continue;
1161
1162                 idev = __in6_dev_get(dev);
1163                 if (!idev)
1164                         continue;
1165
1166                 read_lock_bh(&idev->lock);
1167                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1168                         int i;
1169
1170                         /*
1171                          * - Tentative Address (RFC2462 section 5.4)
1172                          *  - A tentative address is not considered
1173                          *    "assigned to an interface" in the traditional
1174                          *    sense, unless it is also flagged as optimistic.
1175                          * - Candidate Source Address (section 4)
1176                          *  - In any case, anycast addresses, multicast
1177                          *    addresses, and the unspecified address MUST
1178                          *    NOT be included in a candidate set.
1179                          */
1180                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182                                 continue;
1183
1184                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185
1186                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1188                                 LIMIT_NETDEBUG(KERN_DEBUG
1189                                                "ADDRCONF: unspecified / multicast address "
1190                                                "assigned as unicast address on %s",
1191                                                dev->name);
1192                                 continue;
1193                         }
1194
1195                         score->rule = -1;
1196                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197
1198                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199                                 int minihiscore, miniscore;
1200
1201                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203
1204                                 if (minihiscore > miniscore) {
1205                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1206                                             score->scopedist > 0) {
1207                                                 /*
1208                                                  * special case:
1209                                                  * each remaining entry
1210                                                  * has too small (not enough)
1211                                                  * scope, because ifa entries
1212                                                  * are sorted by their scope
1213                                                  * values.
1214                                                  */
1215                                                 goto try_nextdev;
1216                                         }
1217                                         break;
1218                                 } else if (minihiscore < miniscore) {
1219                                         if (hiscore->ifa)
1220                                                 in6_ifa_put(hiscore->ifa);
1221
1222                                         in6_ifa_hold(score->ifa);
1223
1224                                         swap(hiscore, score);
1225
1226                                         /* restore our iterator */
1227                                         score->ifa = hiscore->ifa;
1228
1229                                         break;
1230                                 }
1231                         }
1232                 }
1233 try_nextdev:
1234                 read_unlock_bh(&idev->lock);
1235         }
1236         rcu_read_unlock();
1237
1238         if (!hiscore->ifa)
1239                 return -EADDRNOTAVAIL;
1240
1241         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242         in6_ifa_put(hiscore->ifa);
1243         return 0;
1244 }
1245
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249                     unsigned char banned_flags)
1250 {
1251         struct inet6_dev *idev;
1252         int err = -EADDRNOTAVAIL;
1253
1254         rcu_read_lock();
1255         if ((idev = __in6_dev_get(dev)) != NULL) {
1256                 struct inet6_ifaddr *ifp;
1257
1258                 read_lock_bh(&idev->lock);
1259                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1260                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1261                                 ipv6_addr_copy(addr, &ifp->addr);
1262                                 err = 0;
1263                                 break;
1264                         }
1265                 }
1266                 read_unlock_bh(&idev->lock);
1267         }
1268         rcu_read_unlock();
1269         return err;
1270 }
1271
1272 static int ipv6_count_addresses(struct inet6_dev *idev)
1273 {
1274         int cnt = 0;
1275         struct inet6_ifaddr *ifp;
1276
1277         read_lock_bh(&idev->lock);
1278         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1279                 cnt++;
1280         read_unlock_bh(&idev->lock);
1281         return cnt;
1282 }
1283
1284 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1285                   struct net_device *dev, int strict)
1286 {
1287         struct inet6_ifaddr * ifp;
1288         u8 hash = ipv6_addr_hash(addr);
1289
1290         read_lock_bh(&addrconf_hash_lock);
1291         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1292                 if (!net_eq(dev_net(ifp->idev->dev), net))
1293                         continue;
1294                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295                     !(ifp->flags&IFA_F_TENTATIVE)) {
1296                         if (dev == NULL || ifp->idev->dev == dev ||
1297                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1298                                 break;
1299                 }
1300         }
1301         read_unlock_bh(&addrconf_hash_lock);
1302         return ifp != NULL;
1303 }
1304 EXPORT_SYMBOL(ipv6_chk_addr);
1305
1306 static
1307 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1308                        struct net_device *dev)
1309 {
1310         struct inet6_ifaddr * ifp;
1311         u8 hash = ipv6_addr_hash(addr);
1312
1313         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1314                 if (!net_eq(dev_net(ifp->idev->dev), net))
1315                         continue;
1316                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1317                         if (dev == NULL || ifp->idev->dev == dev)
1318                                 break;
1319                 }
1320         }
1321         return ifp != NULL;
1322 }
1323
1324 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1325 {
1326         struct inet6_dev *idev;
1327         struct inet6_ifaddr *ifa;
1328         int     onlink;
1329
1330         onlink = 0;
1331         rcu_read_lock();
1332         idev = __in6_dev_get(dev);
1333         if (idev) {
1334                 read_lock_bh(&idev->lock);
1335                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1336                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1337                                                    ifa->prefix_len);
1338                         if (onlink)
1339                                 break;
1340                 }
1341                 read_unlock_bh(&idev->lock);
1342         }
1343         rcu_read_unlock();
1344         return onlink;
1345 }
1346
1347 EXPORT_SYMBOL(ipv6_chk_prefix);
1348
1349 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1350                                      struct net_device *dev, int strict)
1351 {
1352         struct inet6_ifaddr * ifp;
1353         u8 hash = ipv6_addr_hash(addr);
1354
1355         read_lock_bh(&addrconf_hash_lock);
1356         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1357                 if (!net_eq(dev_net(ifp->idev->dev), net))
1358                         continue;
1359                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1360                         if (dev == NULL || ifp->idev->dev == dev ||
1361                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362                                 in6_ifa_hold(ifp);
1363                                 break;
1364                         }
1365                 }
1366         }
1367         read_unlock_bh(&addrconf_hash_lock);
1368
1369         return ifp;
1370 }
1371
1372 /* Gets referenced address, destroys ifaddr */
1373
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1375 {
1376         if (ifp->flags&IFA_F_PERMANENT) {
1377                 spin_lock_bh(&ifp->lock);
1378                 addrconf_del_timer(ifp);
1379                 ifp->flags |= IFA_F_TENTATIVE;
1380                 if (dad_failed)
1381                         ifp->flags |= IFA_F_DADFAILED;
1382                 spin_unlock_bh(&ifp->lock);
1383                 if (dad_failed)
1384                         ipv6_ifa_notify(0, ifp);
1385                 in6_ifa_put(ifp);
1386 #ifdef CONFIG_IPV6_PRIVACY
1387         } else if (ifp->flags&IFA_F_TEMPORARY) {
1388                 struct inet6_ifaddr *ifpub;
1389                 spin_lock_bh(&ifp->lock);
1390                 ifpub = ifp->ifpub;
1391                 if (ifpub) {
1392                         in6_ifa_hold(ifpub);
1393                         spin_unlock_bh(&ifp->lock);
1394                         ipv6_create_tempaddr(ifpub, ifp);
1395                         in6_ifa_put(ifpub);
1396                 } else {
1397                         spin_unlock_bh(&ifp->lock);
1398                 }
1399                 ipv6_del_addr(ifp);
1400 #endif
1401         } else
1402                 ipv6_del_addr(ifp);
1403 }
1404
1405 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1406 {
1407         struct inet6_dev *idev = ifp->idev;
1408
1409         if (net_ratelimit())
1410                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1411                         ifp->idev->dev->name, &ifp->addr);
1412
1413         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1414                 struct in6_addr addr;
1415
1416                 addr.s6_addr32[0] = htonl(0xfe800000);
1417                 addr.s6_addr32[1] = 0;
1418
1419                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1420                     ipv6_addr_equal(&ifp->addr, &addr)) {
1421                         /* DAD failed for link-local based on MAC address */
1422                         idev->cnf.disable_ipv6 = 1;
1423
1424                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1425                                 ifp->idev->dev->name);
1426                 }
1427         }
1428
1429         addrconf_dad_stop(ifp, 1);
1430 }
1431
1432 /* Join to solicited addr multicast group. */
1433
1434 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1435 {
1436         struct in6_addr maddr;
1437
1438         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1439                 return;
1440
1441         addrconf_addr_solict_mult(addr, &maddr);
1442         ipv6_dev_mc_inc(dev, &maddr);
1443 }
1444
1445 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1446 {
1447         struct in6_addr maddr;
1448
1449         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1450                 return;
1451
1452         addrconf_addr_solict_mult(addr, &maddr);
1453         __ipv6_dev_mc_dec(idev, &maddr);
1454 }
1455
1456 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1457 {
1458         struct in6_addr addr;
1459         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1460         if (ipv6_addr_any(&addr))
1461                 return;
1462         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1463 }
1464
1465 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1466 {
1467         struct in6_addr addr;
1468         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1469         if (ipv6_addr_any(&addr))
1470                 return;
1471         __ipv6_dev_ac_dec(ifp->idev, &addr);
1472 }
1473
1474 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1475 {
1476         if (dev->addr_len != ETH_ALEN)
1477                 return -1;
1478         memcpy(eui, dev->dev_addr, 3);
1479         memcpy(eui + 5, dev->dev_addr + 3, 3);
1480
1481         /*
1482          * The zSeries OSA network cards can be shared among various
1483          * OS instances, but the OSA cards have only one MAC address.
1484          * This leads to duplicate address conflicts in conjunction
1485          * with IPv6 if more than one instance uses the same card.
1486          *
1487          * The driver for these cards can deliver a unique 16-bit
1488          * identifier for each instance sharing the same card.  It is
1489          * placed instead of 0xFFFE in the interface identifier.  The
1490          * "u" bit of the interface identifier is not inverted in this
1491          * case.  Hence the resulting interface identifier has local
1492          * scope according to RFC2373.
1493          */
1494         if (dev->dev_id) {
1495                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1496                 eui[4] = dev->dev_id & 0xFF;
1497         } else {
1498                 eui[3] = 0xFF;
1499                 eui[4] = 0xFE;
1500                 eui[0] ^= 2;
1501         }
1502         return 0;
1503 }
1504
1505 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1506 {
1507         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1508         if (dev->addr_len != ARCNET_ALEN)
1509                 return -1;
1510         memset(eui, 0, 7);
1511         eui[7] = *(u8*)dev->dev_addr;
1512         return 0;
1513 }
1514
1515 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1516 {
1517         if (dev->addr_len != INFINIBAND_ALEN)
1518                 return -1;
1519         memcpy(eui, dev->dev_addr + 12, 8);
1520         eui[0] |= 2;
1521         return 0;
1522 }
1523
1524 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1525 {
1526         if (addr == 0)
1527                 return -1;
1528         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1529                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1530                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1531                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1532                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1533                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1534         eui[1] = 0;
1535         eui[2] = 0x5E;
1536         eui[3] = 0xFE;
1537         memcpy(eui + 4, &addr, 4);
1538         return 0;
1539 }
1540 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1541
1542 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1543 {
1544         if (dev->priv_flags & IFF_ISATAP)
1545                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1546         return -1;
1547 }
1548
1549 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1550 {
1551         switch (dev->type) {
1552         case ARPHRD_ETHER:
1553         case ARPHRD_FDDI:
1554         case ARPHRD_IEEE802_TR:
1555                 return addrconf_ifid_eui48(eui, dev);
1556         case ARPHRD_ARCNET:
1557                 return addrconf_ifid_arcnet(eui, dev);
1558         case ARPHRD_INFINIBAND:
1559                 return addrconf_ifid_infiniband(eui, dev);
1560         case ARPHRD_SIT:
1561                 return addrconf_ifid_sit(eui, dev);
1562         }
1563         return -1;
1564 }
1565
1566 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1567 {
1568         int err = -1;
1569         struct inet6_ifaddr *ifp;
1570
1571         read_lock_bh(&idev->lock);
1572         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1573                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1574                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1575                         err = 0;
1576                         break;
1577                 }
1578         }
1579         read_unlock_bh(&idev->lock);
1580         return err;
1581 }
1582
1583 #ifdef CONFIG_IPV6_PRIVACY
1584 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1585 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1586 {
1587 regen:
1588         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1589         idev->rndid[0] &= ~0x02;
1590
1591         /*
1592          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1593          * check if generated address is not inappropriate
1594          *
1595          *  - Reserved subnet anycast (RFC 2526)
1596          *      11111101 11....11 1xxxxxxx
1597          *  - ISATAP (RFC4214) 6.1
1598          *      00-00-5E-FE-xx-xx-xx-xx
1599          *  - value 0
1600          *  - XXX: already assigned to an address on the device
1601          */
1602         if (idev->rndid[0] == 0xfd &&
1603             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1604             (idev->rndid[7]&0x80))
1605                 goto regen;
1606         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1607                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1608                         goto regen;
1609                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1610                         goto regen;
1611         }
1612
1613         return 0;
1614 }
1615
1616 static void ipv6_regen_rndid(unsigned long data)
1617 {
1618         struct inet6_dev *idev = (struct inet6_dev *) data;
1619         unsigned long expires;
1620
1621         rcu_read_lock_bh();
1622         write_lock_bh(&idev->lock);
1623
1624         if (idev->dead)
1625                 goto out;
1626
1627         if (__ipv6_regen_rndid(idev) < 0)
1628                 goto out;
1629
1630         expires = jiffies +
1631                 idev->cnf.temp_prefered_lft * HZ -
1632                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1633         if (time_before(expires, jiffies)) {
1634                 printk(KERN_WARNING
1635                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1636                         idev->dev->name);
1637                 goto out;
1638         }
1639
1640         if (!mod_timer(&idev->regen_timer, expires))
1641                 in6_dev_hold(idev);
1642
1643 out:
1644         write_unlock_bh(&idev->lock);
1645         rcu_read_unlock_bh();
1646         in6_dev_put(idev);
1647 }
1648
1649 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1650         int ret = 0;
1651
1652         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1653                 ret = __ipv6_regen_rndid(idev);
1654         return ret;
1655 }
1656 #endif
1657
1658 /*
1659  *      Add prefix route.
1660  */
1661
1662 static void
1663 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1664                       unsigned long expires, u32 flags)
1665 {
1666         struct fib6_config cfg = {
1667                 .fc_table = RT6_TABLE_PREFIX,
1668                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1669                 .fc_ifindex = dev->ifindex,
1670                 .fc_expires = expires,
1671                 .fc_dst_len = plen,
1672                 .fc_flags = RTF_UP | flags,
1673                 .fc_nlinfo.nl_net = dev_net(dev),
1674                 .fc_protocol = RTPROT_KERNEL,
1675         };
1676
1677         ipv6_addr_copy(&cfg.fc_dst, pfx);
1678
1679         /* Prevent useless cloning on PtP SIT.
1680            This thing is done here expecting that the whole
1681            class of non-broadcast devices need not cloning.
1682          */
1683 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1684         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1685                 cfg.fc_flags |= RTF_NONEXTHOP;
1686 #endif
1687
1688         ip6_route_add(&cfg);
1689 }
1690
1691 /* Create "default" multicast route to the interface */
1692
1693 static void addrconf_add_mroute(struct net_device *dev)
1694 {
1695         struct fib6_config cfg = {
1696                 .fc_table = RT6_TABLE_LOCAL,
1697                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1698                 .fc_ifindex = dev->ifindex,
1699                 .fc_dst_len = 8,
1700                 .fc_flags = RTF_UP,
1701                 .fc_nlinfo.nl_net = dev_net(dev),
1702         };
1703
1704         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1705
1706         ip6_route_add(&cfg);
1707 }
1708
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710 static void sit_route_add(struct net_device *dev)
1711 {
1712         struct fib6_config cfg = {
1713                 .fc_table = RT6_TABLE_MAIN,
1714                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1715                 .fc_ifindex = dev->ifindex,
1716                 .fc_dst_len = 96,
1717                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1718                 .fc_nlinfo.nl_net = dev_net(dev),
1719         };
1720
1721         /* prefix length - 96 bits "::d.d.d.d" */
1722         ip6_route_add(&cfg);
1723 }
1724 #endif
1725
1726 static void addrconf_add_lroute(struct net_device *dev)
1727 {
1728         struct in6_addr addr;
1729
1730         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1731         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1732 }
1733
1734 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1735 {
1736         struct inet6_dev *idev;
1737
1738         ASSERT_RTNL();
1739
1740         if ((idev = ipv6_find_idev(dev)) == NULL)
1741                 return NULL;
1742
1743         /* Add default multicast route */
1744         addrconf_add_mroute(dev);
1745
1746         /* Add link local route */
1747         addrconf_add_lroute(dev);
1748         return idev;
1749 }
1750
1751 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1752 {
1753         struct prefix_info *pinfo;
1754         __u32 valid_lft;
1755         __u32 prefered_lft;
1756         int addr_type;
1757         struct inet6_dev *in6_dev;
1758         struct net *net = dev_net(dev);
1759
1760         pinfo = (struct prefix_info *) opt;
1761
1762         if (len < sizeof(struct prefix_info)) {
1763                 ADBG(("addrconf: prefix option too short\n"));
1764                 return;
1765         }
1766
1767         /*
1768          *      Validation checks ([ADDRCONF], page 19)
1769          */
1770
1771         addr_type = ipv6_addr_type(&pinfo->prefix);
1772
1773         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1774                 return;
1775
1776         valid_lft = ntohl(pinfo->valid);
1777         prefered_lft = ntohl(pinfo->prefered);
1778
1779         if (prefered_lft > valid_lft) {
1780                 if (net_ratelimit())
1781                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1782                 return;
1783         }
1784
1785         in6_dev = in6_dev_get(dev);
1786
1787         if (in6_dev == NULL) {
1788                 if (net_ratelimit())
1789                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1790                 return;
1791         }
1792
1793         /*
1794          *      Two things going on here:
1795          *      1) Add routes for on-link prefixes
1796          *      2) Configure prefixes with the auto flag set
1797          */
1798
1799         if (pinfo->onlink) {
1800                 struct rt6_info *rt;
1801                 unsigned long rt_expires;
1802
1803                 /* Avoid arithmetic overflow. Really, we could
1804                  * save rt_expires in seconds, likely valid_lft,
1805                  * but it would require division in fib gc, that it
1806                  * not good.
1807                  */
1808                 if (HZ > USER_HZ)
1809                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1810                 else
1811                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1812
1813                 if (addrconf_finite_timeout(rt_expires))
1814                         rt_expires *= HZ;
1815
1816                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1817                                 dev->ifindex, 1);
1818
1819                 if (rt && addrconf_is_prefix_route(rt)) {
1820                         /* Autoconf prefix route */
1821                         if (valid_lft == 0) {
1822                                 ip6_del_rt(rt);
1823                                 rt = NULL;
1824                         } else if (addrconf_finite_timeout(rt_expires)) {
1825                                 /* not infinity */
1826                                 rt->rt6i_expires = jiffies + rt_expires;
1827                                 rt->rt6i_flags |= RTF_EXPIRES;
1828                         } else {
1829                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1830                                 rt->rt6i_expires = 0;
1831                         }
1832                 } else if (valid_lft) {
1833                         clock_t expires = 0;
1834                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1835                         if (addrconf_finite_timeout(rt_expires)) {
1836                                 /* not infinity */
1837                                 flags |= RTF_EXPIRES;
1838                                 expires = jiffies_to_clock_t(rt_expires);
1839                         }
1840                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1841                                               dev, expires, flags);
1842                 }
1843                 if (rt)
1844                         dst_release(&rt->u.dst);
1845         }
1846
1847         /* Try to figure out our local address for this prefix */
1848
1849         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1850                 struct inet6_ifaddr * ifp;
1851                 struct in6_addr addr;
1852                 int create = 0, update_lft = 0;
1853
1854                 if (pinfo->prefix_len == 64) {
1855                         memcpy(&addr, &pinfo->prefix, 8);
1856                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1857                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1858                                 in6_dev_put(in6_dev);
1859                                 return;
1860                         }
1861                         goto ok;
1862                 }
1863                 if (net_ratelimit())
1864                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1865                                pinfo->prefix_len);
1866                 in6_dev_put(in6_dev);
1867                 return;
1868
1869 ok:
1870
1871                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1872
1873                 if (ifp == NULL && valid_lft) {
1874                         int max_addresses = in6_dev->cnf.max_addresses;
1875                         u32 addr_flags = 0;
1876
1877 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1878                         if (in6_dev->cnf.optimistic_dad &&
1879                             !net->ipv6.devconf_all->forwarding)
1880                                 addr_flags = IFA_F_OPTIMISTIC;
1881 #endif
1882
1883                         /* Do not allow to create too much of autoconfigured
1884                          * addresses; this would be too easy way to crash kernel.
1885                          */
1886                         if (!max_addresses ||
1887                             ipv6_count_addresses(in6_dev) < max_addresses)
1888                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1889                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1890                                                     addr_flags);
1891
1892                         if (!ifp || IS_ERR(ifp)) {
1893                                 in6_dev_put(in6_dev);
1894                                 return;
1895                         }
1896
1897                         update_lft = create = 1;
1898                         ifp->cstamp = jiffies;
1899                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1900                 }
1901
1902                 if (ifp) {
1903                         int flags;
1904                         unsigned long now;
1905 #ifdef CONFIG_IPV6_PRIVACY
1906                         struct inet6_ifaddr *ift;
1907 #endif
1908                         u32 stored_lft;
1909
1910                         /* update lifetime (RFC2462 5.5.3 e) */
1911                         spin_lock(&ifp->lock);
1912                         now = jiffies;
1913                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1914                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1915                         else
1916                                 stored_lft = 0;
1917                         if (!update_lft && stored_lft) {
1918                                 if (valid_lft > MIN_VALID_LIFETIME ||
1919                                     valid_lft > stored_lft)
1920                                         update_lft = 1;
1921                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1922                                         /* valid_lft <= stored_lft is always true */
1923                                         /*
1924                                          * RFC 4862 Section 5.5.3e:
1925                                          * "Note that the preferred lifetime of
1926                                          *  the corresponding address is always
1927                                          *  reset to the Preferred Lifetime in
1928                                          *  the received Prefix Information
1929                                          *  option, regardless of whether the
1930                                          *  valid lifetime is also reset or
1931                                          *  ignored."
1932                                          *
1933                                          *  So if the preferred lifetime in
1934                                          *  this advertisement is different
1935                                          *  than what we have stored, but the
1936                                          *  valid lifetime is invalid, just
1937                                          *  reset prefered_lft.
1938                                          *
1939                                          *  We must set the valid lifetime
1940                                          *  to the stored lifetime since we'll
1941                                          *  be updating the timestamp below,
1942                                          *  else we'll set it back to the
1943                                          *  minumum.
1944                                          */
1945                                         if (prefered_lft != ifp->prefered_lft) {
1946                                                 valid_lft = stored_lft;
1947                                                 update_lft = 1;
1948                                         }
1949                                 } else {
1950                                         valid_lft = MIN_VALID_LIFETIME;
1951                                         if (valid_lft < prefered_lft)
1952                                                 prefered_lft = valid_lft;
1953                                         update_lft = 1;
1954                                 }
1955                         }
1956
1957                         if (update_lft) {
1958                                 ifp->valid_lft = valid_lft;
1959                                 ifp->prefered_lft = prefered_lft;
1960                                 ifp->tstamp = now;
1961                                 flags = ifp->flags;
1962                                 ifp->flags &= ~IFA_F_DEPRECATED;
1963                                 spin_unlock(&ifp->lock);
1964
1965                                 if (!(flags&IFA_F_TENTATIVE))
1966                                         ipv6_ifa_notify(0, ifp);
1967                         } else
1968                                 spin_unlock(&ifp->lock);
1969
1970 #ifdef CONFIG_IPV6_PRIVACY
1971                         read_lock_bh(&in6_dev->lock);
1972                         /* update all temporary addresses in the list */
1973                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1974                                 /*
1975                                  * When adjusting the lifetimes of an existing
1976                                  * temporary address, only lower the lifetimes.
1977                                  * Implementations must not increase the
1978                                  * lifetimes of an existing temporary address
1979                                  * when processing a Prefix Information Option.
1980                                  */
1981                                 if (ifp != ift->ifpub)
1982                                         continue;
1983
1984                                 spin_lock(&ift->lock);
1985                                 flags = ift->flags;
1986                                 if (ift->valid_lft > valid_lft &&
1987                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1988                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1989                                 if (ift->prefered_lft > prefered_lft &&
1990                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1991                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1992                                 spin_unlock(&ift->lock);
1993                                 if (!(flags&IFA_F_TENTATIVE))
1994                                         ipv6_ifa_notify(0, ift);
1995                         }
1996
1997                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1998                                 /*
1999                                  * When a new public address is created as described in [ADDRCONF],
2000                                  * also create a new temporary address.
2001                                  */
2002                                 read_unlock_bh(&in6_dev->lock);
2003                                 ipv6_create_tempaddr(ifp, NULL);
2004                         } else {
2005                                 read_unlock_bh(&in6_dev->lock);
2006                         }
2007 #endif
2008                         in6_ifa_put(ifp);
2009                         addrconf_verify(0);
2010                 }
2011         }
2012         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2013         in6_dev_put(in6_dev);
2014 }
2015
2016 /*
2017  *      Set destination address.
2018  *      Special case for SIT interfaces where we create a new "virtual"
2019  *      device.
2020  */
2021 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2022 {
2023         struct in6_ifreq ireq;
2024         struct net_device *dev;
2025         int err = -EINVAL;
2026
2027         rtnl_lock();
2028
2029         err = -EFAULT;
2030         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2031                 goto err_exit;
2032
2033         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2034
2035         err = -ENODEV;
2036         if (dev == NULL)
2037                 goto err_exit;
2038
2039 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2040         if (dev->type == ARPHRD_SIT) {
2041                 const struct net_device_ops *ops = dev->netdev_ops;
2042                 struct ifreq ifr;
2043                 struct ip_tunnel_parm p;
2044
2045                 err = -EADDRNOTAVAIL;
2046                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2047                         goto err_exit;
2048
2049                 memset(&p, 0, sizeof(p));
2050                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2051                 p.iph.saddr = 0;
2052                 p.iph.version = 4;
2053                 p.iph.ihl = 5;
2054                 p.iph.protocol = IPPROTO_IPV6;
2055                 p.iph.ttl = 64;
2056                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2057
2058                 if (ops->ndo_do_ioctl) {
2059                         mm_segment_t oldfs = get_fs();
2060
2061                         set_fs(KERNEL_DS);
2062                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2063                         set_fs(oldfs);
2064                 } else
2065                         err = -EOPNOTSUPP;
2066
2067                 if (err == 0) {
2068                         err = -ENOBUFS;
2069                         dev = __dev_get_by_name(net, p.name);
2070                         if (!dev)
2071                                 goto err_exit;
2072                         err = dev_open(dev);
2073                 }
2074         }
2075 #endif
2076
2077 err_exit:
2078         rtnl_unlock();
2079         return err;
2080 }
2081
2082 /*
2083  *      Manual configuration of address on an interface
2084  */
2085 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2086                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2087                           __u32 valid_lft)
2088 {
2089         struct inet6_ifaddr *ifp;
2090         struct inet6_dev *idev;
2091         struct net_device *dev;
2092         int scope;
2093         u32 flags;
2094         clock_t expires;
2095         unsigned long timeout;
2096
2097         ASSERT_RTNL();
2098
2099         if (plen > 128)
2100                 return -EINVAL;
2101
2102         /* check the lifetime */
2103         if (!valid_lft || prefered_lft > valid_lft)
2104                 return -EINVAL;
2105
2106         dev = __dev_get_by_index(net, ifindex);
2107         if (!dev)
2108                 return -ENODEV;
2109
2110         if ((idev = addrconf_add_dev(dev)) == NULL)
2111                 return -ENOBUFS;
2112
2113         scope = ipv6_addr_scope(pfx);
2114
2115         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2116         if (addrconf_finite_timeout(timeout)) {
2117                 expires = jiffies_to_clock_t(timeout * HZ);
2118                 valid_lft = timeout;
2119                 flags = RTF_EXPIRES;
2120         } else {
2121                 expires = 0;
2122                 flags = 0;
2123                 ifa_flags |= IFA_F_PERMANENT;
2124         }
2125
2126         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2127         if (addrconf_finite_timeout(timeout)) {
2128                 if (timeout == 0)
2129                         ifa_flags |= IFA_F_DEPRECATED;
2130                 prefered_lft = timeout;
2131         }
2132
2133         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2134
2135         if (!IS_ERR(ifp)) {
2136                 spin_lock_bh(&ifp->lock);
2137                 ifp->valid_lft = valid_lft;
2138                 ifp->prefered_lft = prefered_lft;
2139                 ifp->tstamp = jiffies;
2140                 spin_unlock_bh(&ifp->lock);
2141
2142                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2143                                       expires, flags);
2144                 /*
2145                  * Note that section 3.1 of RFC 4429 indicates
2146                  * that the Optimistic flag should not be set for
2147                  * manually configured addresses
2148                  */
2149                 addrconf_dad_start(ifp, 0);
2150                 in6_ifa_put(ifp);
2151                 addrconf_verify(0);
2152                 return 0;
2153         }
2154
2155         return PTR_ERR(ifp);
2156 }
2157
2158 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2159                           unsigned int plen)
2160 {
2161         struct inet6_ifaddr *ifp;
2162         struct inet6_dev *idev;
2163         struct net_device *dev;
2164
2165         if (plen > 128)
2166                 return -EINVAL;
2167
2168         dev = __dev_get_by_index(net, ifindex);
2169         if (!dev)
2170                 return -ENODEV;
2171
2172         if ((idev = __in6_dev_get(dev)) == NULL)
2173                 return -ENXIO;
2174
2175         read_lock_bh(&idev->lock);
2176         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2177                 if (ifp->prefix_len == plen &&
2178                     ipv6_addr_equal(pfx, &ifp->addr)) {
2179                         in6_ifa_hold(ifp);
2180                         read_unlock_bh(&idev->lock);
2181
2182                         ipv6_del_addr(ifp);
2183
2184                         /* If the last address is deleted administratively,
2185                            disable IPv6 on this interface.
2186                          */
2187                         if (idev->addr_list == NULL)
2188                                 addrconf_ifdown(idev->dev, 1);
2189                         return 0;
2190                 }
2191         }
2192         read_unlock_bh(&idev->lock);
2193         return -EADDRNOTAVAIL;
2194 }
2195
2196
2197 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2198 {
2199         struct in6_ifreq ireq;
2200         int err;
2201
2202         if (!capable(CAP_NET_ADMIN))
2203                 return -EPERM;
2204
2205         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2206                 return -EFAULT;
2207
2208         rtnl_lock();
2209         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2210                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2211                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2212         rtnl_unlock();
2213         return err;
2214 }
2215
2216 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2217 {
2218         struct in6_ifreq ireq;
2219         int err;
2220
2221         if (!capable(CAP_NET_ADMIN))
2222                 return -EPERM;
2223
2224         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2225                 return -EFAULT;
2226
2227         rtnl_lock();
2228         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2229                              ireq.ifr6_prefixlen);
2230         rtnl_unlock();
2231         return err;
2232 }
2233
2234 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2235                      int plen, int scope)
2236 {
2237         struct inet6_ifaddr *ifp;
2238
2239         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2240         if (!IS_ERR(ifp)) {
2241                 spin_lock_bh(&ifp->lock);
2242                 ifp->flags &= ~IFA_F_TENTATIVE;
2243                 spin_unlock_bh(&ifp->lock);
2244                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2245                 in6_ifa_put(ifp);
2246         }
2247 }
2248
2249 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2250 static void sit_add_v4_addrs(struct inet6_dev *idev)
2251 {
2252         struct in6_addr addr;
2253         struct net_device *dev;
2254         struct net *net = dev_net(idev->dev);
2255         int scope;
2256
2257         ASSERT_RTNL();
2258
2259         memset(&addr, 0, sizeof(struct in6_addr));
2260         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2261
2262         if (idev->dev->flags&IFF_POINTOPOINT) {
2263                 addr.s6_addr32[0] = htonl(0xfe800000);
2264                 scope = IFA_LINK;
2265         } else {
2266                 scope = IPV6_ADDR_COMPATv4;
2267         }
2268
2269         if (addr.s6_addr32[3]) {
2270                 add_addr(idev, &addr, 128, scope);
2271                 return;
2272         }
2273
2274         for_each_netdev(net, dev) {
2275                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2276                 if (in_dev && (dev->flags & IFF_UP)) {
2277                         struct in_ifaddr * ifa;
2278
2279                         int flag = scope;
2280
2281                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2282                                 int plen;
2283
2284                                 addr.s6_addr32[3] = ifa->ifa_local;
2285
2286                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2287                                         continue;
2288                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2289                                         if (idev->dev->flags&IFF_POINTOPOINT)
2290                                                 continue;
2291                                         flag |= IFA_HOST;
2292                                 }
2293                                 if (idev->dev->flags&IFF_POINTOPOINT)
2294                                         plen = 64;
2295                                 else
2296                                         plen = 96;
2297
2298                                 add_addr(idev, &addr, plen, flag);
2299                         }
2300                 }
2301         }
2302 }
2303 #endif
2304
2305 static void init_loopback(struct net_device *dev)
2306 {
2307         struct inet6_dev  *idev;
2308
2309         /* ::1 */
2310
2311         ASSERT_RTNL();
2312
2313         if ((idev = ipv6_find_idev(dev)) == NULL) {
2314                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2315                 return;
2316         }
2317
2318         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2319 }
2320
2321 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2322 {
2323         struct inet6_ifaddr * ifp;
2324         u32 addr_flags = IFA_F_PERMANENT;
2325
2326 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2327         if (idev->cnf.optimistic_dad &&
2328             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2329                 addr_flags |= IFA_F_OPTIMISTIC;
2330 #endif
2331
2332
2333         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2334         if (!IS_ERR(ifp)) {
2335                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2336                 addrconf_dad_start(ifp, 0);
2337                 in6_ifa_put(ifp);
2338         }
2339 }
2340
2341 static void addrconf_dev_config(struct net_device *dev)
2342 {
2343         struct in6_addr addr;
2344         struct inet6_dev    * idev;
2345
2346         ASSERT_RTNL();
2347
2348         if ((dev->type != ARPHRD_ETHER) &&
2349             (dev->type != ARPHRD_FDDI) &&
2350             (dev->type != ARPHRD_IEEE802_TR) &&
2351             (dev->type != ARPHRD_ARCNET) &&
2352             (dev->type != ARPHRD_INFINIBAND)) {
2353                 /* Alas, we support only Ethernet autoconfiguration. */
2354                 return;
2355         }
2356
2357         idev = addrconf_add_dev(dev);
2358         if (idev == NULL)
2359                 return;
2360
2361         memset(&addr, 0, sizeof(struct in6_addr));
2362         addr.s6_addr32[0] = htonl(0xFE800000);
2363
2364         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2365                 addrconf_add_linklocal(idev, &addr);
2366 }
2367
2368 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2369 static void addrconf_sit_config(struct net_device *dev)
2370 {
2371         struct inet6_dev *idev;
2372
2373         ASSERT_RTNL();
2374
2375         /*
2376          * Configure the tunnel with one of our IPv4
2377          * addresses... we should configure all of
2378          * our v4 addrs in the tunnel
2379          */
2380
2381         if ((idev = ipv6_find_idev(dev)) == NULL) {
2382                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2383                 return;
2384         }
2385
2386         if (dev->priv_flags & IFF_ISATAP) {
2387                 struct in6_addr addr;
2388
2389                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2390                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2391                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2392                         addrconf_add_linklocal(idev, &addr);
2393                 return;
2394         }
2395
2396         sit_add_v4_addrs(idev);
2397
2398         if (dev->flags&IFF_POINTOPOINT) {
2399                 addrconf_add_mroute(dev);
2400                 addrconf_add_lroute(dev);
2401         } else
2402                 sit_route_add(dev);
2403 }
2404 #endif
2405
2406 static inline int
2407 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2408 {
2409         struct in6_addr lladdr;
2410
2411         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2412                 addrconf_add_linklocal(idev, &lladdr);
2413                 return 0;
2414         }
2415         return -1;
2416 }
2417
2418 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2419 {
2420         struct net_device *link_dev;
2421         struct net *net = dev_net(idev->dev);
2422
2423         /* first try to inherit the link-local address from the link device */
2424         if (idev->dev->iflink &&
2425             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2426                 if (!ipv6_inherit_linklocal(idev, link_dev))
2427                         return;
2428         }
2429         /* then try to inherit it from any device */
2430         for_each_netdev(net, link_dev) {
2431                 if (!ipv6_inherit_linklocal(idev, link_dev))
2432                         return;
2433         }
2434         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2435 }
2436
2437 /*
2438  * Autoconfigure tunnel with a link-local address so routing protocols,
2439  * DHCPv6, MLD etc. can be run over the virtual link
2440  */
2441
2442 static void addrconf_ip6_tnl_config(struct net_device *dev)
2443 {
2444         struct inet6_dev *idev;
2445
2446         ASSERT_RTNL();
2447
2448         if ((idev = addrconf_add_dev(dev)) == NULL) {
2449                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2450                 return;
2451         }
2452         ip6_tnl_add_linklocal(idev);
2453 }
2454
2455 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2456                            void * data)
2457 {
2458         struct net_device *dev = (struct net_device *) data;
2459         struct inet6_dev *idev = __in6_dev_get(dev);
2460         int run_pending = 0;
2461         int err;
2462
2463         switch(event) {
2464         case NETDEV_REGISTER:
2465                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2466                         idev = ipv6_add_dev(dev);
2467                         if (!idev)
2468                                 return notifier_from_errno(-ENOMEM);
2469                 }
2470                 break;
2471         case NETDEV_UP:
2472         case NETDEV_CHANGE:
2473                 if (dev->flags & IFF_SLAVE)
2474                         break;
2475
2476                 if (event == NETDEV_UP) {
2477                         if (!addrconf_qdisc_ok(dev)) {
2478                                 /* device is not ready yet. */
2479                                 printk(KERN_INFO
2480                                         "ADDRCONF(NETDEV_UP): %s: "
2481                                         "link is not ready\n",
2482                                         dev->name);
2483                                 break;
2484                         }
2485
2486                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2487                                 idev = ipv6_add_dev(dev);
2488
2489                         if (idev) {
2490                                 idev->if_flags |= IF_READY;
2491                                 run_pending = 1;
2492                         }
2493                 } else {
2494                         if (!addrconf_qdisc_ok(dev)) {
2495                                 /* device is still not ready. */
2496                                 break;
2497                         }
2498
2499                         if (idev) {
2500                                 if (idev->if_flags & IF_READY) {
2501                                         /* device is already configured. */
2502                                         break;
2503                                 }
2504                                 idev->if_flags |= IF_READY;
2505                         }
2506
2507                         printk(KERN_INFO
2508                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2509                                         "link becomes ready\n",
2510                                         dev->name);
2511
2512                         run_pending = 1;
2513                 }
2514
2515                 switch(dev->type) {
2516 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2517                 case ARPHRD_SIT:
2518                         addrconf_sit_config(dev);
2519                         break;
2520 #endif
2521                 case ARPHRD_TUNNEL6:
2522                         addrconf_ip6_tnl_config(dev);
2523                         break;
2524                 case ARPHRD_LOOPBACK:
2525                         init_loopback(dev);
2526                         break;
2527
2528                 default:
2529                         addrconf_dev_config(dev);
2530                         break;
2531                 }
2532                 if (idev) {
2533                         if (run_pending)
2534                                 addrconf_dad_run(idev);
2535
2536                         /* If the MTU changed during the interface down, when the
2537                            interface up, the changed MTU must be reflected in the
2538                            idev as well as routers.
2539                          */
2540                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2541                                 rt6_mtu_change(dev, dev->mtu);
2542                                 idev->cnf.mtu6 = dev->mtu;
2543                         }
2544                         idev->tstamp = jiffies;
2545                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2546                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2547                            stop IPv6 on this interface.
2548                          */
2549                         if (dev->mtu < IPV6_MIN_MTU)
2550                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2551                 }
2552                 break;
2553
2554         case NETDEV_CHANGEMTU:
2555                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2556                         rt6_mtu_change(dev, dev->mtu);
2557                         idev->cnf.mtu6 = dev->mtu;
2558                         break;
2559                 }
2560
2561                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2562                         idev = ipv6_add_dev(dev);
2563                         if (idev)
2564                                 break;
2565                 }
2566
2567                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2568
2569         case NETDEV_DOWN:
2570         case NETDEV_UNREGISTER:
2571                 /*
2572                  *      Remove all addresses from this interface.
2573                  */
2574                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2575                 break;
2576
2577         case NETDEV_CHANGENAME:
2578                 if (idev) {
2579                         snmp6_unregister_dev(idev);
2580                         addrconf_sysctl_unregister(idev);
2581                         addrconf_sysctl_register(idev);
2582                         err = snmp6_register_dev(idev);
2583                         if (err)
2584                                 return notifier_from_errno(err);
2585                 }
2586                 break;
2587         case NETDEV_BONDING_OLDTYPE:
2588         case NETDEV_BONDING_NEWTYPE:
2589                 addrconf_bonding_change(dev, event);
2590                 break;
2591         }
2592
2593         return NOTIFY_OK;
2594 }
2595
2596 /*
2597  *      addrconf module should be notified of a device going up
2598  */
2599 static struct notifier_block ipv6_dev_notf = {
2600         .notifier_call = addrconf_notify,
2601         .priority = 0
2602 };
2603
2604 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2605 {
2606         struct inet6_dev *idev;
2607         ASSERT_RTNL();
2608
2609         idev = __in6_dev_get(dev);
2610
2611         if (event == NETDEV_BONDING_NEWTYPE)
2612                 ipv6_mc_remap(idev);
2613         else if (event == NETDEV_BONDING_OLDTYPE)
2614                 ipv6_mc_unmap(idev);
2615 }
2616
2617 static int addrconf_ifdown(struct net_device *dev, int how)
2618 {
2619         struct inet6_dev *idev;
2620         struct inet6_ifaddr *ifa, *keep_list, **bifa;
2621         struct net *net = dev_net(dev);
2622         int i;
2623
2624         ASSERT_RTNL();
2625
2626         rt6_ifdown(net, dev);
2627         neigh_ifdown(&nd_tbl, dev);
2628
2629         idev = __in6_dev_get(dev);
2630         if (idev == NULL)
2631                 return -ENODEV;
2632
2633         /* Step 1: remove reference to ipv6 device from parent device.
2634                    Do not dev_put!
2635          */
2636         if (how) {
2637                 idev->dead = 1;
2638
2639                 /* protected by rtnl_lock */
2640                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2641
2642                 /* Step 1.5: remove snmp6 entry */
2643                 snmp6_unregister_dev(idev);
2644
2645         }
2646
2647         /* Step 2: clear hash table */
2648         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2649                 bifa = &inet6_addr_lst[i];
2650
2651                 write_lock_bh(&addrconf_hash_lock);
2652                 while ((ifa = *bifa) != NULL) {
2653                         if (ifa->idev == idev &&
2654                             (how || !(ifa->flags&IFA_F_PERMANENT) ||
2655                              ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2656                                 *bifa = ifa->lst_next;
2657                                 ifa->lst_next = NULL;
2658                                 __in6_ifa_put(ifa);
2659                                 continue;
2660                         }
2661                         bifa = &ifa->lst_next;
2662                 }
2663                 write_unlock_bh(&addrconf_hash_lock);
2664         }
2665
2666         write_lock_bh(&idev->lock);
2667
2668         /* Step 3: clear flags for stateless addrconf */
2669         if (!how)
2670                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2671
2672         /* Step 4: clear address list */
2673 #ifdef CONFIG_IPV6_PRIVACY
2674         if (how && del_timer(&idev->regen_timer))
2675                 in6_dev_put(idev);
2676
2677         /* clear tempaddr list */
2678         while ((ifa = idev->tempaddr_list) != NULL) {
2679                 idev->tempaddr_list = ifa->tmp_next;
2680                 ifa->tmp_next = NULL;
2681                 ifa->dead = 1;
2682                 write_unlock_bh(&idev->lock);
2683                 spin_lock_bh(&ifa->lock);
2684
2685                 if (ifa->ifpub) {
2686                         in6_ifa_put(ifa->ifpub);
2687                         ifa->ifpub = NULL;
2688                 }
2689                 spin_unlock_bh(&ifa->lock);
2690                 in6_ifa_put(ifa);
2691                 write_lock_bh(&idev->lock);
2692         }
2693 #endif
2694         keep_list = NULL;
2695         bifa = &keep_list;
2696         while ((ifa = idev->addr_list) != NULL) {
2697                 idev->addr_list = ifa->if_next;
2698                 ifa->if_next = NULL;
2699
2700                 addrconf_del_timer(ifa);
2701
2702                 /* If just doing link down, and address is permanent
2703                    and not link-local, then retain it. */
2704                 if (how == 0 &&
2705                     (ifa->flags&IFA_F_PERMANENT) &&
2706                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2707
2708                         /* Move to holding list */
2709                         *bifa = ifa;
2710                         bifa = &ifa->if_next;
2711
2712                         /* If not doing DAD on this address, just keep it. */
2713                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2714                             idev->cnf.accept_dad <= 0 ||
2715                             (ifa->flags & IFA_F_NODAD))
2716                                 continue;
2717
2718                         /* If it was tentative already, no need to notify */
2719                         if (ifa->flags & IFA_F_TENTATIVE)
2720                                 continue;
2721
2722                         /* Flag it for later restoration when link comes up */
2723                         ifa->flags |= IFA_F_TENTATIVE;
2724                         in6_ifa_hold(ifa);
2725                 } else {
2726                         ifa->dead = 1;
2727                 }
2728                 write_unlock_bh(&idev->lock);
2729
2730                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2731                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2732                 in6_ifa_put(ifa);
2733
2734                 write_lock_bh(&idev->lock);
2735         }
2736
2737         idev->addr_list = keep_list;
2738
2739         write_unlock_bh(&idev->lock);
2740
2741         /* Step 5: Discard multicast list */
2742
2743         if (how)
2744                 ipv6_mc_destroy_dev(idev);
2745         else
2746                 ipv6_mc_down(idev);
2747
2748         idev->tstamp = jiffies;
2749
2750         /* Shot the device (if unregistered) */
2751
2752         if (how) {
2753                 addrconf_sysctl_unregister(idev);
2754                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2755                 neigh_ifdown(&nd_tbl, dev);
2756                 in6_dev_put(idev);
2757         }
2758         return 0;
2759 }
2760
2761 static void addrconf_rs_timer(unsigned long data)
2762 {
2763         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2764         struct inet6_dev *idev = ifp->idev;
2765
2766         read_lock(&idev->lock);
2767         if (idev->dead || !(idev->if_flags & IF_READY))
2768                 goto out;
2769
2770         if (idev->cnf.forwarding)
2771                 goto out;
2772
2773         /* Announcement received after solicitation was sent */
2774         if (idev->if_flags & IF_RA_RCVD)
2775                 goto out;
2776
2777         spin_lock(&ifp->lock);
2778         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2779                 /* The wait after the last probe can be shorter */
2780                 addrconf_mod_timer(ifp, AC_RS,
2781                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2782                                    idev->cnf.rtr_solicit_delay :
2783                                    idev->cnf.rtr_solicit_interval);
2784                 spin_unlock(&ifp->lock);
2785
2786                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2787         } else {
2788                 spin_unlock(&ifp->lock);
2789                 /*
2790                  * Note: we do not support deprecated "all on-link"
2791                  * assumption any longer.
2792                  */
2793                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2794                        idev->dev->name);
2795         }
2796
2797 out:
2798         read_unlock(&idev->lock);
2799         in6_ifa_put(ifp);
2800 }
2801
2802 /*
2803  *      Duplicate Address Detection
2804  */
2805 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2806 {
2807         unsigned long rand_num;
2808         struct inet6_dev *idev = ifp->idev;
2809
2810         if (ifp->flags & IFA_F_OPTIMISTIC)
2811                 rand_num = 0;
2812         else
2813                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2814
2815         ifp->probes = idev->cnf.dad_transmits;
2816         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2817 }
2818
2819 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2820 {
2821         struct inet6_dev *idev = ifp->idev;
2822         struct net_device *dev = idev->dev;
2823
2824         addrconf_join_solict(dev, &ifp->addr);
2825
2826         net_srandom(ifp->addr.s6_addr32[3]);
2827
2828         read_lock_bh(&idev->lock);
2829         if (ifp->dead)
2830                 goto out;
2831
2832         spin_lock(&ifp->lock);
2833         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2834             idev->cnf.accept_dad < 1 ||
2835             !(ifp->flags&IFA_F_TENTATIVE) ||
2836             ifp->flags & IFA_F_NODAD) {
2837                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2838                 spin_unlock(&ifp->lock);
2839                 read_unlock_bh(&idev->lock);
2840
2841                 addrconf_dad_completed(ifp);
2842                 return;
2843         }
2844
2845         if (!(idev->if_flags & IF_READY)) {
2846                 spin_unlock(&ifp->lock);
2847                 read_unlock_bh(&idev->lock);
2848                 /*
2849                  * If the device is not ready:
2850                  * - keep it tentative if it is a permanent address.
2851                  * - otherwise, kill it.
2852                  */
2853                 in6_ifa_hold(ifp);
2854                 addrconf_dad_stop(ifp, 0);
2855                 return;
2856         }
2857
2858         /*
2859          * Optimistic nodes can start receiving
2860          * Frames right away
2861          */
2862         if(ifp->flags & IFA_F_OPTIMISTIC)
2863                 ip6_ins_rt(ifp->rt);
2864
2865         addrconf_dad_kick(ifp);
2866         spin_unlock(&ifp->lock);
2867 out:
2868         read_unlock_bh(&idev->lock);
2869 }
2870
2871 static void addrconf_dad_timer(unsigned long data)
2872 {
2873         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2874         struct inet6_dev *idev = ifp->idev;
2875         struct in6_addr mcaddr;
2876
2877         read_lock(&idev->lock);
2878         if (idev->dead || !(idev->if_flags & IF_READY)) {
2879                 read_unlock(&idev->lock);
2880                 goto out;
2881         }
2882
2883         spin_lock(&ifp->lock);
2884         if (ifp->probes == 0) {
2885                 /*
2886                  * DAD was successful
2887                  */
2888
2889                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2890                 spin_unlock(&ifp->lock);
2891                 read_unlock(&idev->lock);
2892
2893                 addrconf_dad_completed(ifp);
2894
2895                 goto out;
2896         }
2897
2898         ifp->probes--;
2899         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2900         spin_unlock(&ifp->lock);
2901         read_unlock(&idev->lock);
2902
2903         /* send a neighbour solicitation for our addr */
2904         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2905         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2906 out:
2907         in6_ifa_put(ifp);
2908 }
2909
2910 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2911 {
2912         struct net_device *     dev = ifp->idev->dev;
2913
2914         /*
2915          *      Configure the address for reception. Now it is valid.
2916          */
2917
2918         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2919
2920         /* If added prefix is link local and forwarding is off,
2921            start sending router solicitations.
2922          */
2923
2924         if (ifp->idev->cnf.forwarding == 0 &&
2925             ifp->idev->cnf.rtr_solicits > 0 &&
2926             (dev->flags&IFF_LOOPBACK) == 0 &&
2927             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2928                 /*
2929                  *      If a host as already performed a random delay
2930                  *      [...] as part of DAD [...] there is no need
2931                  *      to delay again before sending the first RS
2932                  */
2933                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2934
2935                 spin_lock_bh(&ifp->lock);
2936                 ifp->probes = 1;
2937                 ifp->idev->if_flags |= IF_RS_SENT;
2938                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2939                 spin_unlock_bh(&ifp->lock);
2940         }
2941 }
2942
2943 static void addrconf_dad_run(struct inet6_dev *idev) {
2944         struct inet6_ifaddr *ifp;
2945
2946         read_lock_bh(&idev->lock);
2947         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2948                 spin_lock(&ifp->lock);
2949                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2950                         spin_unlock(&ifp->lock);
2951                         continue;
2952                 }
2953                 spin_unlock(&ifp->lock);
2954                 addrconf_dad_kick(ifp);
2955         }
2956         read_unlock_bh(&idev->lock);
2957 }
2958
2959 #ifdef CONFIG_PROC_FS
2960 struct if6_iter_state {
2961         struct seq_net_private p;
2962         int bucket;
2963 };
2964
2965 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2966 {
2967         struct inet6_ifaddr *ifa = NULL;
2968         struct if6_iter_state *state = seq->private;
2969         struct net *net = seq_file_net(seq);
2970
2971         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2972                 ifa = inet6_addr_lst[state->bucket];
2973
2974                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2975                         ifa = ifa->lst_next;
2976                 if (ifa)
2977                         break;
2978         }
2979         return ifa;
2980 }
2981
2982 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2983 {
2984         struct if6_iter_state *state = seq->private;
2985         struct net *net = seq_file_net(seq);
2986
2987         ifa = ifa->lst_next;
2988 try_again:
2989         if (ifa) {
2990                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2991                         ifa = ifa->lst_next;
2992                         goto try_again;
2993                 }
2994         }
2995
2996         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2997                 ifa = inet6_addr_lst[state->bucket];
2998                 goto try_again;
2999         }
3000
3001         return ifa;
3002 }
3003
3004 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3005 {
3006         struct inet6_ifaddr *ifa = if6_get_first(seq);
3007
3008         if (ifa)
3009                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3010                         --pos;
3011         return pos ? NULL : ifa;
3012 }
3013
3014 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3015         __acquires(addrconf_hash_lock)
3016 {
3017         read_lock_bh(&addrconf_hash_lock);
3018         return if6_get_idx(seq, *pos);
3019 }
3020
3021 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3022 {
3023         struct inet6_ifaddr *ifa;
3024
3025         ifa = if6_get_next(seq, v);
3026         ++*pos;
3027         return ifa;
3028 }
3029
3030 static void if6_seq_stop(struct seq_file *seq, void *v)
3031         __releases(addrconf_hash_lock)
3032 {
3033         read_unlock_bh(&addrconf_hash_lock);
3034 }
3035
3036 static int if6_seq_show(struct seq_file *seq, void *v)
3037 {
3038         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3039         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3040                    &ifp->addr,
3041                    ifp->idev->dev->ifindex,
3042                    ifp->prefix_len,
3043                    ifp->scope,
3044                    ifp->flags,
3045                    ifp->idev->dev->name);
3046         return 0;
3047 }
3048
3049 static const struct seq_operations if6_seq_ops = {
3050         .start  = if6_seq_start,
3051         .next   = if6_seq_next,
3052         .show   = if6_seq_show,
3053         .stop   = if6_seq_stop,
3054 };
3055
3056 static int if6_seq_open(struct inode *inode, struct file *file)
3057 {
3058         return seq_open_net(inode, file, &if6_seq_ops,
3059                             sizeof(struct if6_iter_state));
3060 }
3061
3062 static const struct file_operations if6_fops = {
3063         .owner          = THIS_MODULE,
3064         .open           = if6_seq_open,
3065         .read           = seq_read,
3066         .llseek         = seq_lseek,
3067         .release        = seq_release_net,
3068 };
3069
3070 static int __net_init if6_proc_net_init(struct net *net)
3071 {
3072         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3073                 return -ENOMEM;
3074         return 0;
3075 }
3076
3077 static void __net_exit if6_proc_net_exit(struct net *net)
3078 {
3079        proc_net_remove(net, "if_inet6");
3080 }
3081
3082 static struct pernet_operations if6_proc_net_ops = {
3083        .init = if6_proc_net_init,
3084        .exit = if6_proc_net_exit,
3085 };
3086
3087 int __init if6_proc_init(void)
3088 {
3089         return register_pernet_subsys(&if6_proc_net_ops);
3090 }
3091
3092 void if6_proc_exit(void)
3093 {
3094         unregister_pernet_subsys(&if6_proc_net_ops);
3095 }
3096 #endif  /* CONFIG_PROC_FS */
3097
3098 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3099 /* Check if address is a home address configured on any interface. */
3100 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3101 {
3102         int ret = 0;
3103         struct inet6_ifaddr * ifp;
3104         u8 hash = ipv6_addr_hash(addr);
3105         read_lock_bh(&addrconf_hash_lock);
3106         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3107                 if (!net_eq(dev_net(ifp->idev->dev), net))
3108                         continue;
3109                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3110                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3111                         ret = 1;
3112                         break;
3113                 }
3114         }
3115         read_unlock_bh(&addrconf_hash_lock);
3116         return ret;
3117 }
3118 #endif
3119
3120 /*
3121  *      Periodic address status verification
3122  */
3123
3124 static void addrconf_verify(unsigned long foo)
3125 {
3126         struct inet6_ifaddr *ifp;
3127         unsigned long now, next;
3128         int i;
3129
3130         spin_lock_bh(&addrconf_verify_lock);
3131         now = jiffies;
3132         next = now + ADDR_CHECK_FREQUENCY;
3133
3134         del_timer(&addr_chk_timer);
3135
3136         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3137
3138 restart:
3139                 read_lock(&addrconf_hash_lock);
3140                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3141                         unsigned long age;
3142 #ifdef CONFIG_IPV6_PRIVACY
3143                         unsigned long regen_advance;
3144 #endif
3145
3146                         if (ifp->flags & IFA_F_PERMANENT)
3147                                 continue;
3148
3149                         spin_lock(&ifp->lock);
3150                         age = (now - ifp->tstamp) / HZ;
3151
3152 #ifdef CONFIG_IPV6_PRIVACY
3153                         regen_advance = ifp->idev->cnf.regen_max_retry *
3154                                         ifp->idev->cnf.dad_transmits *
3155                                         ifp->idev->nd_parms->retrans_time / HZ;
3156 #endif
3157
3158                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3159                             age >= ifp->valid_lft) {
3160                                 spin_unlock(&ifp->lock);
3161                                 in6_ifa_hold(ifp);
3162                                 read_unlock(&addrconf_hash_lock);
3163                                 ipv6_del_addr(ifp);
3164                                 goto restart;
3165                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3166                                 spin_unlock(&ifp->lock);
3167                                 continue;
3168                         } else if (age >= ifp->prefered_lft) {
3169                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3170                                 int deprecate = 0;
3171
3172                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3173                                         deprecate = 1;
3174                                         ifp->flags |= IFA_F_DEPRECATED;
3175                                 }
3176
3177                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3178                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3179
3180                                 spin_unlock(&ifp->lock);
3181
3182                                 if (deprecate) {
3183                                         in6_ifa_hold(ifp);
3184                                         read_unlock(&addrconf_hash_lock);
3185
3186                                         ipv6_ifa_notify(0, ifp);
3187                                         in6_ifa_put(ifp);
3188                                         goto restart;
3189                                 }
3190 #ifdef CONFIG_IPV6_PRIVACY
3191                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3192                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3193                                 if (age >= ifp->prefered_lft - regen_advance) {
3194                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3195                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3196                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3197                                         if (!ifp->regen_count && ifpub) {
3198                                                 ifp->regen_count++;
3199                                                 in6_ifa_hold(ifp);
3200                                                 in6_ifa_hold(ifpub);
3201                                                 spin_unlock(&ifp->lock);
3202                                                 read_unlock(&addrconf_hash_lock);
3203                                                 spin_lock(&ifpub->lock);
3204                                                 ifpub->regen_count = 0;
3205                                                 spin_unlock(&ifpub->lock);
3206                                                 ipv6_create_tempaddr(ifpub, ifp);
3207                                                 in6_ifa_put(ifpub);
3208                                                 in6_ifa_put(ifp);
3209                                                 goto restart;
3210                                         }
3211                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3212                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3213                                 spin_unlock(&ifp->lock);
3214 #endif
3215                         } else {
3216                                 /* ifp->prefered_lft <= ifp->valid_lft */
3217                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3218                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3219                                 spin_unlock(&ifp->lock);
3220                         }
3221                 }
3222                 read_unlock(&addrconf_hash_lock);
3223         }
3224
3225         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3226         add_timer(&addr_chk_timer);
3227         spin_unlock_bh(&addrconf_verify_lock);
3228 }
3229
3230 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3231 {
3232         struct in6_addr *pfx = NULL;
3233
3234         if (addr)
3235                 pfx = nla_data(addr);
3236
3237         if (local) {
3238                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3239                         pfx = NULL;
3240                 else
3241                         pfx = nla_data(local);
3242         }
3243
3244         return pfx;
3245 }
3246
3247 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3248         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3249         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3250         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3251 };
3252
3253 static int
3254 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3255 {
3256         struct net *net = sock_net(skb->sk);
3257         struct ifaddrmsg *ifm;
3258         struct nlattr *tb[IFA_MAX+1];
3259         struct in6_addr *pfx;
3260         int err;
3261
3262         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3263         if (err < 0)
3264                 return err;
3265
3266         ifm = nlmsg_data(nlh);
3267         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3268         if (pfx == NULL)
3269                 return -EINVAL;
3270
3271         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3272 }
3273
3274 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3275                              u32 prefered_lft, u32 valid_lft)
3276 {
3277         u32 flags;
3278         clock_t expires;
3279         unsigned long timeout;
3280
3281         if (!valid_lft || (prefered_lft > valid_lft))
3282                 return -EINVAL;
3283
3284         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3285         if (addrconf_finite_timeout(timeout)) {
3286                 expires = jiffies_to_clock_t(timeout * HZ);
3287                 valid_lft = timeout;
3288                 flags = RTF_EXPIRES;
3289         } else {
3290                 expires = 0;
3291                 flags = 0;
3292                 ifa_flags |= IFA_F_PERMANENT;
3293         }
3294
3295         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3296         if (addrconf_finite_timeout(timeout)) {
3297                 if (timeout == 0)
3298                         ifa_flags |= IFA_F_DEPRECATED;
3299                 prefered_lft = timeout;
3300         }
3301
3302         spin_lock_bh(&ifp->lock);
3303         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3304         ifp->tstamp = jiffies;
3305         ifp->valid_lft = valid_lft;
3306         ifp->prefered_lft = prefered_lft;
3307
3308         spin_unlock_bh(&ifp->lock);
3309         if (!(ifp->flags&IFA_F_TENTATIVE))
3310                 ipv6_ifa_notify(0, ifp);
3311
3312         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3313                               expires, flags);
3314         addrconf_verify(0);
3315
3316         return 0;
3317 }
3318
3319 static int
3320 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3321 {
3322         struct net *net = sock_net(skb->sk);
3323         struct ifaddrmsg *ifm;
3324         struct nlattr *tb[IFA_MAX+1];
3325         struct in6_addr *pfx;
3326         struct inet6_ifaddr *ifa;
3327         struct net_device *dev;
3328         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3329         u8 ifa_flags;
3330         int err;
3331
3332         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3333         if (err < 0)
3334                 return err;
3335
3336         ifm = nlmsg_data(nlh);
3337         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3338         if (pfx == NULL)
3339                 return -EINVAL;
3340
3341         if (tb[IFA_CACHEINFO]) {
3342                 struct ifa_cacheinfo *ci;
3343
3344                 ci = nla_data(tb[IFA_CACHEINFO]);
3345                 valid_lft = ci->ifa_valid;
3346                 preferred_lft = ci->ifa_prefered;
3347         } else {
3348                 preferred_lft = INFINITY_LIFE_TIME;
3349                 valid_lft = INFINITY_LIFE_TIME;
3350         }
3351
3352         dev =  __dev_get_by_index(net, ifm->ifa_index);
3353         if (dev == NULL)
3354                 return -ENODEV;
3355
3356         /* We ignore other flags so far. */
3357         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3358
3359         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3360         if (ifa == NULL) {
3361                 /*
3362                  * It would be best to check for !NLM_F_CREATE here but
3363                  * userspace alreay relies on not having to provide this.
3364                  */
3365                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3366                                       ifm->ifa_prefixlen, ifa_flags,
3367                                       preferred_lft, valid_lft);
3368         }
3369
3370         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3371             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3372                 err = -EEXIST;
3373         else
3374                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3375
3376         in6_ifa_put(ifa);
3377
3378         return err;
3379 }
3380
3381 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3382                           u8 scope, int ifindex)
3383 {
3384         struct ifaddrmsg *ifm;
3385
3386         ifm = nlmsg_data(nlh);
3387         ifm->ifa_family = AF_INET6;
3388         ifm->ifa_prefixlen = prefixlen;
3389         ifm->ifa_flags = flags;
3390         ifm->ifa_scope = scope;
3391         ifm->ifa_index = ifindex;
3392 }
3393
3394 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3395                          unsigned long tstamp, u32 preferred, u32 valid)
3396 {
3397         struct ifa_cacheinfo ci;
3398
3399         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3400                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3401         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3402                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3403         ci.ifa_prefered = preferred;
3404         ci.ifa_valid = valid;
3405
3406         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3407 }
3408
3409 static inline int rt_scope(int ifa_scope)
3410 {
3411         if (ifa_scope & IFA_HOST)
3412                 return RT_SCOPE_HOST;
3413         else if (ifa_scope & IFA_LINK)
3414                 return RT_SCOPE_LINK;
3415         else if (ifa_scope & IFA_SITE)
3416                 return RT_SCOPE_SITE;
3417         else
3418                 return RT_SCOPE_UNIVERSE;
3419 }
3420
3421 static inline int inet6_ifaddr_msgsize(void)
3422 {
3423         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3424                + nla_total_size(16) /* IFA_ADDRESS */
3425                + nla_total_size(sizeof(struct ifa_cacheinfo));
3426 }
3427
3428 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3429                              u32 pid, u32 seq, int event, unsigned int flags)
3430 {
3431         struct nlmsghdr  *nlh;
3432         u32 preferred, valid;
3433
3434         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3435         if (nlh == NULL)
3436                 return -EMSGSIZE;
3437
3438         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3439                       ifa->idev->dev->ifindex);
3440
3441         if (!(ifa->flags&IFA_F_PERMANENT)) {
3442                 preferred = ifa->prefered_lft;
3443                 valid = ifa->valid_lft;
3444                 if (preferred != INFINITY_LIFE_TIME) {
3445                         long tval = (jiffies - ifa->tstamp)/HZ;
3446                         if (preferred > tval)
3447                                 preferred -= tval;
3448                         else
3449                                 preferred = 0;
3450                         if (valid != INFINITY_LIFE_TIME)
3451                                 valid -= tval;
3452                 }
3453         } else {
3454                 preferred = INFINITY_LIFE_TIME;
3455                 valid = INFINITY_LIFE_TIME;
3456         }
3457
3458         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3459             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3460                 nlmsg_cancel(skb, nlh);
3461                 return -EMSGSIZE;
3462         }
3463
3464         return nlmsg_end(skb, nlh);
3465 }
3466
3467 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3468                                 u32 pid, u32 seq, int event, u16 flags)
3469 {
3470         struct nlmsghdr  *nlh;
3471         u8 scope = RT_SCOPE_UNIVERSE;
3472         int ifindex = ifmca->idev->dev->ifindex;
3473
3474         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3475                 scope = RT_SCOPE_SITE;
3476
3477         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3478         if (nlh == NULL)
3479                 return -EMSGSIZE;
3480
3481         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3482         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3483             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3484                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3485                 nlmsg_cancel(skb, nlh);
3486                 return -EMSGSIZE;
3487         }
3488
3489         return nlmsg_end(skb, nlh);
3490 }
3491
3492 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3493                                 u32 pid, u32 seq, int event, unsigned int flags)
3494 {
3495         struct nlmsghdr  *nlh;
3496         u8 scope = RT_SCOPE_UNIVERSE;
3497         int ifindex = ifaca->aca_idev->dev->ifindex;
3498
3499         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3500                 scope = RT_SCOPE_SITE;
3501
3502         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3503         if (nlh == NULL)
3504                 return -EMSGSIZE;
3505
3506         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3507         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3508             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3509                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3510                 nlmsg_cancel(skb, nlh);
3511                 return -EMSGSIZE;
3512         }
3513
3514         return nlmsg_end(skb, nlh);
3515 }
3516
3517 enum addr_type_t
3518 {
3519         UNICAST_ADDR,
3520         MULTICAST_ADDR,
3521         ANYCAST_ADDR,
3522 };
3523
3524 /* called with rcu_read_lock() */
3525 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3526                           struct netlink_callback *cb, enum addr_type_t type,
3527                           int s_ip_idx, int *p_ip_idx)
3528 {
3529         struct inet6_ifaddr *ifa;
3530         struct ifmcaddr6 *ifmca;
3531         struct ifacaddr6 *ifaca;
3532         int err = 1;
3533         int ip_idx = *p_ip_idx;
3534
3535         read_lock_bh(&idev->lock);
3536         switch (type) {
3537         case UNICAST_ADDR:
3538                 /* unicast address incl. temp addr */
3539                 for (ifa = idev->addr_list; ifa;
3540                      ifa = ifa->if_next, ip_idx++) {
3541                         if (ip_idx < s_ip_idx)
3542                                 continue;
3543                         err = inet6_fill_ifaddr(skb, ifa,
3544                                                 NETLINK_CB(cb->skb).pid,
3545                                                 cb->nlh->nlmsg_seq,
3546                                                 RTM_NEWADDR,
3547                                                 NLM_F_MULTI);
3548                         if (err <= 0)
3549                                 break;
3550                 }
3551                 break;
3552         case MULTICAST_ADDR:
3553                 /* multicast address */
3554                 for (ifmca = idev->mc_list; ifmca;
3555                      ifmca = ifmca->next, ip_idx++) {
3556                         if (ip_idx < s_ip_idx)
3557                                 continue;
3558                         err = inet6_fill_ifmcaddr(skb, ifmca,
3559                                                   NETLINK_CB(cb->skb).pid,
3560                                                   cb->nlh->nlmsg_seq,
3561                                                   RTM_GETMULTICAST,
3562                                                   NLM_F_MULTI);
3563                         if (err <= 0)
3564                                 break;
3565                 }
3566                 break;
3567         case ANYCAST_ADDR:
3568                 /* anycast address */
3569                 for (ifaca = idev->ac_list; ifaca;
3570                      ifaca = ifaca->aca_next, ip_idx++) {
3571                         if (ip_idx < s_ip_idx)
3572                                 continue;
3573                         err = inet6_fill_ifacaddr(skb, ifaca,
3574                                                   NETLINK_CB(cb->skb).pid,
3575                                                   cb->nlh->nlmsg_seq,
3576                                                   RTM_GETANYCAST,
3577                                                   NLM_F_MULTI);
3578                         if (err <= 0)
3579                                 break;
3580                 }
3581                 break;
3582         default:
3583                 break;
3584         }
3585         read_unlock_bh(&idev->lock);
3586         *p_ip_idx = ip_idx;
3587         return err;
3588 }
3589
3590 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3591                            enum addr_type_t type)
3592 {
3593         struct net *net = sock_net(skb->sk);
3594         int h, s_h;
3595         int idx, ip_idx;
3596         int s_idx, s_ip_idx;
3597         struct net_device *dev;
3598         struct inet6_dev *idev;
3599         struct hlist_head *head;
3600         struct hlist_node *node;
3601
3602         s_h = cb->args[0];
3603         s_idx = idx = cb->args[1];
3604         s_ip_idx = ip_idx = cb->args[2];
3605
3606         rcu_read_lock();
3607         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3608                 idx = 0;
3609                 head = &net->dev_index_head[h];
3610                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3611                         if (idx < s_idx)
3612                                 goto cont;
3613                         if (idx > s_idx)
3614                                 s_ip_idx = 0;
3615                         ip_idx = 0;
3616                         if ((idev = __in6_dev_get(dev)) == NULL)
3617                                 goto cont;
3618
3619                         if (in6_dump_addrs(idev, skb, cb, type,
3620                                            s_ip_idx, &ip_idx) <= 0)
3621                                 goto done;
3622 cont:
3623                         idx++;
3624                 }
3625         }
3626 done:
3627         rcu_read_unlock();
3628         cb->args[0] = h;
3629         cb->args[1] = idx;
3630         cb->args[2] = ip_idx;
3631
3632         return skb->len;
3633 }
3634
3635 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3636 {
3637         enum addr_type_t type = UNICAST_ADDR;
3638
3639         return inet6_dump_addr(skb, cb, type);
3640 }
3641
3642 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3643 {
3644         enum addr_type_t type = MULTICAST_ADDR;
3645
3646         return inet6_dump_addr(skb, cb, type);
3647 }
3648
3649
3650 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3651 {
3652         enum addr_type_t type = ANYCAST_ADDR;
3653
3654         return inet6_dump_addr(skb, cb, type);
3655 }
3656
3657 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3658                              void *arg)
3659 {
3660         struct net *net = sock_net(in_skb->sk);
3661         struct ifaddrmsg *ifm;
3662         struct nlattr *tb[IFA_MAX+1];
3663         struct in6_addr *addr = NULL;
3664         struct net_device *dev = NULL;
3665         struct inet6_ifaddr *ifa;
3666         struct sk_buff *skb;
3667         int err;
3668
3669         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3670         if (err < 0)
3671                 goto errout;
3672
3673         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3674         if (addr == NULL) {
3675                 err = -EINVAL;
3676                 goto errout;
3677         }
3678
3679         ifm = nlmsg_data(nlh);
3680         if (ifm->ifa_index)
3681                 dev = __dev_get_by_index(net, ifm->ifa_index);
3682
3683         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3684                 err = -EADDRNOTAVAIL;
3685                 goto errout;
3686         }
3687
3688         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3689                 err = -ENOBUFS;
3690                 goto errout_ifa;
3691         }
3692
3693         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3694                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3695         if (err < 0) {
3696                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3697                 WARN_ON(err == -EMSGSIZE);
3698                 kfree_skb(skb);
3699                 goto errout_ifa;
3700         }
3701         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3702 errout_ifa:
3703         in6_ifa_put(ifa);
3704 errout:
3705         return err;
3706 }
3707
3708 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3709 {
3710         struct sk_buff *skb;
3711         struct net *net = dev_net(ifa->idev->dev);
3712         int err = -ENOBUFS;
3713
3714         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3715         if (skb == NULL)
3716                 goto errout;
3717
3718         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3719         if (err < 0) {
3720                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3721                 WARN_ON(err == -EMSGSIZE);
3722                 kfree_skb(skb);
3723                 goto errout;
3724         }
3725         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3726         return;
3727 errout:
3728         if (err < 0)
3729                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3730 }
3731
3732 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3733                                 __s32 *array, int bytes)
3734 {
3735         BUG_ON(bytes < (DEVCONF_MAX * 4));
3736
3737         memset(array, 0, bytes);
3738         array[DEVCONF_FORWARDING] = cnf->forwarding;
3739         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3740         array[DEVCONF_MTU6] = cnf->mtu6;
3741         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3742         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3743         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3744         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3745         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3746         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3747         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3748         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3749 #ifdef CONFIG_IPV6_PRIVACY
3750         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3751         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3752         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3753         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3754         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3755 #endif
3756         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3757         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3758         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3759 #ifdef CONFIG_IPV6_ROUTER_PREF
3760         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3761         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3762 #ifdef CONFIG_IPV6_ROUTE_INFO
3763         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3764 #endif
3765 #endif
3766         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3767         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3768 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3769         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3770 #endif
3771 #ifdef CONFIG_IPV6_MROUTE
3772         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3773 #endif
3774         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3775         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3776         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3777 }
3778
3779 static inline size_t inet6_if_nlmsg_size(void)
3780 {
3781         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3782                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3783                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3784                + nla_total_size(4) /* IFLA_MTU */
3785                + nla_total_size(4) /* IFLA_LINK */
3786                + nla_total_size( /* IFLA_PROTINFO */
3787                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3788                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3789                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3790                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3791                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3792                  );
3793 }
3794
3795 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3796                                       int items, int bytes)
3797 {
3798         int i;
3799         int pad = bytes - sizeof(u64) * items;
3800         BUG_ON(pad < 0);
3801
3802         /* Use put_unaligned() because stats may not be aligned for u64. */
3803         put_unaligned(items, &stats[0]);
3804         for (i = 1; i < items; i++)
3805                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3806
3807         memset(&stats[items], 0, pad);
3808 }
3809
3810 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3811                              int bytes)
3812 {
3813         switch(attrtype) {
3814         case IFLA_INET6_STATS:
3815                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3816                 break;
3817         case IFLA_INET6_ICMP6STATS:
3818                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3819                 break;
3820         }
3821 }
3822
3823 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3824                              u32 pid, u32 seq, int event, unsigned int flags)
3825 {
3826         struct net_device *dev = idev->dev;
3827         struct nlattr *nla;
3828         struct ifinfomsg *hdr;
3829         struct nlmsghdr *nlh;
3830         void *protoinfo;
3831         struct ifla_cacheinfo ci;
3832
3833         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3834         if (nlh == NULL)
3835                 return -EMSGSIZE;
3836
3837         hdr = nlmsg_data(nlh);
3838         hdr->ifi_family = AF_INET6;
3839         hdr->__ifi_pad = 0;
3840         hdr->ifi_type = dev->type;
3841         hdr->ifi_index = dev->ifindex;
3842         hdr->ifi_flags = dev_get_flags(dev);
3843         hdr->ifi_change = 0;
3844
3845         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3846
3847         if (dev->addr_len)
3848                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3849
3850         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3851         if (dev->ifindex != dev->iflink)
3852                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3853
3854         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3855         if (protoinfo == NULL)
3856                 goto nla_put_failure;
3857
3858         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3859
3860         ci.max_reasm_len = IPV6_MAXPLEN;
3861         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3862                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3863         ci.reachable_time = idev->nd_parms->reachable_time;
3864         ci.retrans_time = idev->nd_parms->retrans_time;
3865         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3866
3867         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3868         if (nla == NULL)
3869                 goto nla_put_failure;
3870         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3871
3872         /* XXX - MC not implemented */
3873
3874         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3875         if (nla == NULL)
3876                 goto nla_put_failure;
3877         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3878
3879         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3880         if (nla == NULL)
3881                 goto nla_put_failure;
3882         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3883
3884         nla_nest_end(skb, protoinfo);
3885         return nlmsg_end(skb, nlh);
3886
3887 nla_put_failure:
3888         nlmsg_cancel(skb, nlh);
3889         return -EMSGSIZE;
3890 }
3891
3892 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3893 {
3894         struct net *net = sock_net(skb->sk);
3895         int h, s_h;
3896         int idx = 0, s_idx;
3897         struct net_device *dev;
3898         struct inet6_dev *idev;
3899         struct hlist_head *head;
3900         struct hlist_node *node;
3901
3902         s_h = cb->args[0];
3903         s_idx = cb->args[1];
3904
3905         rcu_read_lock();
3906         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3907                 idx = 0;
3908                 head = &net->dev_index_head[h];
3909                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3910                         if (idx < s_idx)
3911                                 goto cont;
3912                         idev = __in6_dev_get(dev);
3913                         if (!idev)
3914                                 goto cont;
3915                         if (inet6_fill_ifinfo(skb, idev,
3916                                               NETLINK_CB(cb->skb).pid,
3917                                               cb->nlh->nlmsg_seq,
3918                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3919                                 goto out;
3920 cont:
3921                         idx++;
3922                 }
3923         }
3924 out:
3925         rcu_read_unlock();
3926         cb->args[1] = idx;
3927         cb->args[0] = h;
3928
3929         return skb->len;
3930 }
3931
3932 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3933 {
3934         struct sk_buff *skb;
3935         struct net *net = dev_net(idev->dev);
3936         int err = -ENOBUFS;
3937
3938         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3939         if (skb == NULL)
3940                 goto errout;
3941
3942         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3943         if (err < 0) {
3944                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3945                 WARN_ON(err == -EMSGSIZE);
3946                 kfree_skb(skb);
3947                 goto errout;
3948         }
3949         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3950         return;
3951 errout:
3952         if (err < 0)
3953                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3954 }
3955
3956 static inline size_t inet6_prefix_nlmsg_size(void)
3957 {
3958         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3959                + nla_total_size(sizeof(struct in6_addr))
3960                + nla_total_size(sizeof(struct prefix_cacheinfo));
3961 }
3962
3963 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3964                              struct prefix_info *pinfo, u32 pid, u32 seq,
3965                              int event, unsigned int flags)
3966 {
3967         struct prefixmsg *pmsg;
3968         struct nlmsghdr *nlh;
3969         struct prefix_cacheinfo ci;
3970
3971         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3972         if (nlh == NULL)
3973                 return -EMSGSIZE;
3974
3975         pmsg = nlmsg_data(nlh);
3976         pmsg->prefix_family = AF_INET6;
3977         pmsg->prefix_pad1 = 0;
3978         pmsg->prefix_pad2 = 0;
3979         pmsg->prefix_ifindex = idev->dev->ifindex;
3980         pmsg->prefix_len = pinfo->prefix_len;
3981         pmsg->prefix_type = pinfo->type;
3982         pmsg->prefix_pad3 = 0;
3983         pmsg->prefix_flags = 0;
3984         if (pinfo->onlink)
3985                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3986         if (pinfo->autoconf)
3987                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3988
3989         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3990
3991         ci.preferred_time = ntohl(pinfo->prefered);
3992         ci.valid_time = ntohl(pinfo->valid);
3993         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3994
3995         return nlmsg_end(skb, nlh);
3996
3997 nla_put_failure:
3998         nlmsg_cancel(skb, nlh);
3999         return -EMSGSIZE;
4000 }
4001
4002 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4003                          struct prefix_info *pinfo)
4004 {
4005         struct sk_buff *skb;
4006         struct net *net = dev_net(idev->dev);
4007         int err = -ENOBUFS;
4008
4009         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4010         if (skb == NULL)
4011                 goto errout;
4012
4013         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4014         if (err < 0) {
4015                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4016                 WARN_ON(err == -EMSGSIZE);
4017                 kfree_skb(skb);
4018                 goto errout;
4019         }
4020         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4021         return;
4022 errout:
4023         if (err < 0)
4024                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4025 }
4026
4027 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4028 {
4029         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4030
4031         switch (event) {
4032         case RTM_NEWADDR:
4033                 /*
4034                  * If the address was optimistic
4035                  * we inserted the route at the start of
4036                  * our DAD process, so we don't need
4037                  * to do it again
4038                  */
4039                 if (!(ifp->rt->rt6i_node))
4040                         ip6_ins_rt(ifp->rt);
4041                 if (ifp->idev->cnf.forwarding)
4042                         addrconf_join_anycast(ifp);
4043                 break;
4044         case RTM_DELADDR:
4045                 if (ifp->idev->cnf.forwarding)
4046                         addrconf_leave_anycast(ifp);
4047                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4048                 dst_hold(&ifp->rt->u.dst);
4049                 if (ip6_del_rt(ifp->rt))
4050                         dst_free(&ifp->rt->u.dst);
4051                 break;
4052         }
4053 }
4054
4055 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4056 {
4057         rcu_read_lock_bh();
4058         if (likely(ifp->idev->dead == 0))
4059                 __ipv6_ifa_notify(event, ifp);
4060         rcu_read_unlock_bh();
4061 }
4062
4063 #ifdef CONFIG_SYSCTL
4064
4065 static
4066 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4067                            void __user *buffer, size_t *lenp, loff_t *ppos)
4068 {
4069         int *valp = ctl->data;
4070         int val = *valp;
4071         loff_t pos = *ppos;
4072         int ret;
4073
4074         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4075
4076         if (write)
4077                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4078         if (ret)
4079                 *ppos = pos;
4080         return ret;
4081 }
4082
4083 static void dev_disable_change(struct inet6_dev *idev)
4084 {
4085         if (!idev || !idev->dev)
4086                 return;
4087
4088         if (idev->cnf.disable_ipv6)
4089                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4090         else
4091                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4092 }
4093
4094 static void addrconf_disable_change(struct net *net, __s32 newf)
4095 {
4096         struct net_device *dev;
4097         struct inet6_dev *idev;
4098
4099         rcu_read_lock();
4100         for_each_netdev_rcu(net, dev) {
4101                 idev = __in6_dev_get(dev);
4102                 if (idev) {
4103                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4104                         idev->cnf.disable_ipv6 = newf;
4105                         if (changed)
4106                                 dev_disable_change(idev);
4107                 }
4108         }
4109         rcu_read_unlock();
4110 }
4111
4112 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4113 {
4114         struct net *net;
4115
4116         net = (struct net *)table->extra2;
4117
4118         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4119                 return 0;
4120
4121         if (!rtnl_trylock()) {
4122                 /* Restore the original values before restarting */
4123                 *p = old;
4124                 return restart_syscall();
4125         }
4126
4127         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4128                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4129                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4130                 addrconf_disable_change(net, newf);
4131         } else if ((!*p) ^ (!old))
4132                 dev_disable_change((struct inet6_dev *)table->extra1);
4133
4134         rtnl_unlock();
4135         return 0;
4136 }
4137
4138 static
4139 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4140                             void __user *buffer, size_t *lenp, loff_t *ppos)
4141 {
4142         int *valp = ctl->data;
4143         int val = *valp;
4144         loff_t pos = *ppos;
4145         int ret;
4146
4147         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4148
4149         if (write)
4150                 ret = addrconf_disable_ipv6(ctl, valp, val);
4151         if (ret)
4152                 *ppos = pos;
4153         return ret;
4154 }
4155
4156 static struct addrconf_sysctl_table
4157 {
4158         struct ctl_table_header *sysctl_header;
4159         ctl_table addrconf_vars[DEVCONF_MAX+1];
4160         char *dev_name;
4161 } addrconf_sysctl __read_mostly = {
4162         .sysctl_header = NULL,
4163         .addrconf_vars = {
4164                 {
4165                         .procname       =       "forwarding",
4166                         .data           =       &ipv6_devconf.forwarding,
4167                         .maxlen         =       sizeof(int),
4168                         .mode           =       0644,
4169                         .proc_handler   =       addrconf_sysctl_forward,
4170                 },
4171                 {
4172                         .procname       =       "hop_limit",
4173                         .data           =       &ipv6_devconf.hop_limit,
4174                         .maxlen         =       sizeof(int),
4175                         .mode           =       0644,
4176                         .proc_handler   =       proc_dointvec,
4177                 },
4178                 {
4179                         .procname       =       "mtu",
4180                         .data           =       &ipv6_devconf.mtu6,
4181                         .maxlen         =       sizeof(int),
4182                         .mode           =       0644,
4183                         .proc_handler   =       proc_dointvec,
4184                 },
4185                 {
4186                         .procname       =       "accept_ra",
4187                         .data           =       &ipv6_devconf.accept_ra,
4188                         .maxlen         =       sizeof(int),
4189                         .mode           =       0644,
4190                         .proc_handler   =       proc_dointvec,
4191                 },
4192                 {
4193                         .procname       =       "accept_redirects",
4194                         .data           =       &ipv6_devconf.accept_redirects,
4195                         .maxlen         =       sizeof(int),
4196                         .mode           =       0644,
4197                         .proc_handler   =       proc_dointvec,
4198                 },
4199                 {
4200                         .procname       =       "autoconf",
4201                         .data           =       &ipv6_devconf.autoconf,
4202                         .maxlen         =       sizeof(int),
4203                         .mode           =       0644,
4204                         .proc_handler   =       proc_dointvec,
4205                 },
4206                 {
4207                         .procname       =       "dad_transmits",
4208                         .data           =       &ipv6_devconf.dad_transmits,
4209                         .maxlen         =       sizeof(int),
4210                         .mode           =       0644,
4211                         .proc_handler   =       proc_dointvec,
4212                 },
4213                 {
4214                         .procname       =       "router_solicitations",
4215                         .data           =       &ipv6_devconf.rtr_solicits,
4216                         .maxlen         =       sizeof(int),
4217                         .mode           =       0644,
4218                         .proc_handler   =       proc_dointvec,
4219                 },
4220                 {
4221                         .procname       =       "router_solicitation_interval",
4222                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4223                         .maxlen         =       sizeof(int),
4224                         .mode           =       0644,
4225                         .proc_handler   =       proc_dointvec_jiffies,
4226                 },
4227                 {
4228                         .procname       =       "router_solicitation_delay",
4229                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4230                         .maxlen         =       sizeof(int),
4231                         .mode           =       0644,
4232                         .proc_handler   =       proc_dointvec_jiffies,
4233                 },
4234                 {
4235                         .procname       =       "force_mld_version",
4236                         .data           =       &ipv6_devconf.force_mld_version,
4237                         .maxlen         =       sizeof(int),
4238                         .mode           =       0644,
4239                         .proc_handler   =       proc_dointvec,
4240                 },
4241 #ifdef CONFIG_IPV6_PRIVACY
4242                 {
4243                         .procname       =       "use_tempaddr",
4244                         .data           =       &ipv6_devconf.use_tempaddr,
4245                         .maxlen         =       sizeof(int),
4246                         .mode           =       0644,
4247                         .proc_handler   =       proc_dointvec,
4248                 },
4249                 {
4250                         .procname       =       "temp_valid_lft",
4251                         .data           =       &ipv6_devconf.temp_valid_lft,
4252                         .maxlen         =       sizeof(int),
4253                         .mode           =       0644,
4254                         .proc_handler   =       proc_dointvec,
4255                 },
4256                 {
4257                         .procname       =       "temp_prefered_lft",
4258                         .data           =       &ipv6_devconf.temp_prefered_lft,
4259                         .maxlen         =       sizeof(int),
4260                         .mode           =       0644,
4261                         .proc_handler   =       proc_dointvec,
4262                 },
4263                 {
4264                         .procname       =       "regen_max_retry",
4265                         .data           =       &ipv6_devconf.regen_max_retry,
4266                         .maxlen         =       sizeof(int),
4267                         .mode           =       0644,
4268                         .proc_handler   =       proc_dointvec,
4269                 },
4270                 {
4271                         .procname       =       "max_desync_factor",
4272                         .data           =       &ipv6_devconf.max_desync_factor,
4273                         .maxlen         =       sizeof(int),
4274                         .mode           =       0644,
4275                         .proc_handler   =       proc_dointvec,
4276                 },
4277 #endif
4278                 {
4279                         .procname       =       "max_addresses",
4280                         .data           =       &ipv6_devconf.max_addresses,
4281                         .maxlen         =       sizeof(int),
4282                         .mode           =       0644,
4283                         .proc_handler   =       proc_dointvec,
4284                 },
4285                 {
4286                         .procname       =       "accept_ra_defrtr",
4287                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4288                         .maxlen         =       sizeof(int),
4289                         .mode           =       0644,
4290                         .proc_handler   =       proc_dointvec,
4291                 },
4292                 {
4293                         .procname       =       "accept_ra_pinfo",
4294                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4295                         .maxlen         =       sizeof(int),
4296                         .mode           =       0644,
4297                         .proc_handler   =       proc_dointvec,
4298                 },
4299 #ifdef CONFIG_IPV6_ROUTER_PREF
4300                 {
4301                         .procname       =       "accept_ra_rtr_pref",
4302                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4303                         .maxlen         =       sizeof(int),
4304                         .mode           =       0644,
4305                         .proc_handler   =       proc_dointvec,
4306                 },
4307                 {
4308                         .procname       =       "router_probe_interval",
4309                         .data           =       &ipv6_devconf.rtr_probe_interval,
4310                         .maxlen         =       sizeof(int),
4311                         .mode           =       0644,
4312                         .proc_handler   =       proc_dointvec_jiffies,
4313                 },
4314 #ifdef CONFIG_IPV6_ROUTE_INFO
4315                 {
4316                         .procname       =       "accept_ra_rt_info_max_plen",
4317                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4318                         .maxlen         =       sizeof(int),
4319                         .mode           =       0644,
4320                         .proc_handler   =       proc_dointvec,
4321                 },
4322 #endif
4323 #endif
4324                 {
4325                         .procname       =       "proxy_ndp",
4326                         .data           =       &ipv6_devconf.proxy_ndp,
4327                         .maxlen         =       sizeof(int),
4328                         .mode           =       0644,
4329                         .proc_handler   =       proc_dointvec,
4330                 },
4331                 {
4332                         .procname       =       "accept_source_route",
4333                         .data           =       &ipv6_devconf.accept_source_route,
4334                         .maxlen         =       sizeof(int),
4335                         .mode           =       0644,
4336                         .proc_handler   =       proc_dointvec,
4337                 },
4338 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4339                 {
4340                         .procname       =       "optimistic_dad",
4341                         .data           =       &ipv6_devconf.optimistic_dad,
4342                         .maxlen         =       sizeof(int),
4343                         .mode           =       0644,
4344                         .proc_handler   =       proc_dointvec,
4345
4346                 },
4347 #endif
4348 #ifdef CONFIG_IPV6_MROUTE
4349                 {
4350                         .procname       =       "mc_forwarding",
4351                         .data           =       &ipv6_devconf.mc_forwarding,
4352                         .maxlen         =       sizeof(int),
4353                         .mode           =       0444,
4354                         .proc_handler   =       proc_dointvec,
4355                 },
4356 #endif
4357                 {
4358                         .procname       =       "disable_ipv6",
4359                         .data           =       &ipv6_devconf.disable_ipv6,
4360                         .maxlen         =       sizeof(int),
4361                         .mode           =       0644,
4362                         .proc_handler   =       addrconf_sysctl_disable,
4363                 },
4364                 {
4365                         .procname       =       "accept_dad",
4366                         .data           =       &ipv6_devconf.accept_dad,
4367                         .maxlen         =       sizeof(int),
4368                         .mode           =       0644,
4369                         .proc_handler   =       proc_dointvec,
4370                 },
4371                 {
4372                         .procname       = "force_tllao",
4373                         .data           = &ipv6_devconf.force_tllao,
4374                         .maxlen         = sizeof(int),
4375                         .mode           = 0644,
4376                         .proc_handler   = proc_dointvec
4377                 },
4378                 {
4379                         /* sentinel */
4380                 }
4381         },
4382 };
4383
4384 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4385                 struct inet6_dev *idev, struct ipv6_devconf *p)
4386 {
4387         int i;
4388         struct addrconf_sysctl_table *t;
4389
4390 #define ADDRCONF_CTL_PATH_DEV   3
4391
4392         struct ctl_path addrconf_ctl_path[] = {
4393                 { .procname = "net", },
4394                 { .procname = "ipv6", },
4395                 { .procname = "conf", },
4396                 { /* to be set */ },
4397                 { },
4398         };
4399
4400
4401         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4402         if (t == NULL)
4403                 goto out;
4404
4405         for (i=0; t->addrconf_vars[i].data; i++) {
4406                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4407                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4408                 t->addrconf_vars[i].extra2 = net;
4409         }
4410
4411         /*
4412          * Make a copy of dev_name, because '.procname' is regarded as const
4413          * by sysctl and we wouldn't want anyone to change it under our feet
4414          * (see SIOCSIFNAME).
4415          */
4416         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4417         if (!t->dev_name)
4418                 goto free;
4419
4420         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4421
4422         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4423                         t->addrconf_vars);
4424         if (t->sysctl_header == NULL)
4425                 goto free_procname;
4426
4427         p->sysctl = t;
4428         return 0;
4429
4430 free_procname:
4431         kfree(t->dev_name);
4432 free:
4433         kfree(t);
4434 out:
4435         return -ENOBUFS;
4436 }
4437
4438 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4439 {
4440         struct addrconf_sysctl_table *t;
4441
4442         if (p->sysctl == NULL)
4443                 return;
4444
4445         t = p->sysctl;
4446         p->sysctl = NULL;
4447         unregister_sysctl_table(t->sysctl_header);
4448         kfree(t->dev_name);
4449         kfree(t);
4450 }
4451
4452 static void addrconf_sysctl_register(struct inet6_dev *idev)
4453 {
4454         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4455                               &ndisc_ifinfo_sysctl_change);
4456         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4457                                         idev, &idev->cnf);
4458 }
4459
4460 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4461 {
4462         __addrconf_sysctl_unregister(&idev->cnf);
4463         neigh_sysctl_unregister(idev->nd_parms);
4464 }
4465
4466
4467 #endif
4468
4469 static int __net_init addrconf_init_net(struct net *net)
4470 {
4471         int err;
4472         struct ipv6_devconf *all, *dflt;
4473
4474         err = -ENOMEM;
4475         all = &ipv6_devconf;
4476         dflt = &ipv6_devconf_dflt;
4477
4478         if (!net_eq(net, &init_net)) {
4479                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4480                 if (all == NULL)
4481                         goto err_alloc_all;
4482
4483                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4484                 if (dflt == NULL)
4485                         goto err_alloc_dflt;
4486         } else {
4487                 /* these will be inherited by all namespaces */
4488                 dflt->autoconf = ipv6_defaults.autoconf;
4489                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4490         }
4491
4492         net->ipv6.devconf_all = all;
4493         net->ipv6.devconf_dflt = dflt;
4494
4495 #ifdef CONFIG_SYSCTL
4496         err = __addrconf_sysctl_register(net, "all", NULL, all);
4497         if (err < 0)
4498                 goto err_reg_all;
4499
4500         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4501         if (err < 0)
4502                 goto err_reg_dflt;
4503 #endif
4504         return 0;
4505
4506 #ifdef CONFIG_SYSCTL
4507 err_reg_dflt:
4508         __addrconf_sysctl_unregister(all);
4509 err_reg_all:
4510         kfree(dflt);
4511 #endif
4512 err_alloc_dflt:
4513         kfree(all);
4514 err_alloc_all:
4515         return err;
4516 }
4517
4518 static void __net_exit addrconf_exit_net(struct net *net)
4519 {
4520 #ifdef CONFIG_SYSCTL
4521         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4522         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4523 #endif
4524         if (!net_eq(net, &init_net)) {
4525                 kfree(net->ipv6.devconf_dflt);
4526                 kfree(net->ipv6.devconf_all);
4527         }
4528 }
4529
4530 static struct pernet_operations addrconf_ops = {
4531         .init = addrconf_init_net,
4532         .exit = addrconf_exit_net,
4533 };
4534
4535 /*
4536  *      Device notifier
4537  */
4538
4539 int register_inet6addr_notifier(struct notifier_block *nb)
4540 {
4541         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4542 }
4543
4544 EXPORT_SYMBOL(register_inet6addr_notifier);
4545
4546 int unregister_inet6addr_notifier(struct notifier_block *nb)
4547 {
4548         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4549 }
4550
4551 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4552
4553 /*
4554  *      Init / cleanup code
4555  */
4556
4557 int __init addrconf_init(void)
4558 {
4559         int err;
4560
4561         if ((err = ipv6_addr_label_init()) < 0) {
4562                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4563                         err);
4564                 return err;
4565         }
4566
4567         register_pernet_subsys(&addrconf_ops);
4568
4569         /* The addrconf netdev notifier requires that loopback_dev
4570          * has it's ipv6 private information allocated and setup
4571          * before it can bring up and give link-local addresses
4572          * to other devices which are up.
4573          *
4574          * Unfortunately, loopback_dev is not necessarily the first
4575          * entry in the global dev_base list of net devices.  In fact,
4576          * it is likely to be the very last entry on that list.
4577          * So this causes the notifier registry below to try and
4578          * give link-local addresses to all devices besides loopback_dev
4579          * first, then loopback_dev, which cases all the non-loopback_dev
4580          * devices to fail to get a link-local address.
4581          *
4582          * So, as a temporary fix, allocate the ipv6 structure for
4583          * loopback_dev first by hand.
4584          * Longer term, all of the dependencies ipv6 has upon the loopback
4585          * device and it being up should be removed.
4586          */
4587         rtnl_lock();
4588         if (!ipv6_add_dev(init_net.loopback_dev))
4589                 err = -ENOMEM;
4590         rtnl_unlock();
4591         if (err)
4592                 goto errlo;
4593
4594         register_netdevice_notifier(&ipv6_dev_notf);
4595
4596         addrconf_verify(0);
4597
4598         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4599         if (err < 0)
4600                 goto errout;
4601
4602         /* Only the first call to __rtnl_register can fail */
4603         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4604         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4605         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4606         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4607         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4608
4609         ipv6_addr_label_rtnl_register();
4610
4611         return 0;
4612 errout:
4613         unregister_netdevice_notifier(&ipv6_dev_notf);
4614 errlo:
4615         unregister_pernet_subsys(&addrconf_ops);
4616
4617         return err;
4618 }
4619
4620 void addrconf_cleanup(void)
4621 {
4622         struct inet6_ifaddr *ifa;
4623         struct net_device *dev;
4624         int i;
4625
4626         unregister_netdevice_notifier(&ipv6_dev_notf);
4627         unregister_pernet_subsys(&addrconf_ops);
4628
4629         rtnl_lock();
4630
4631         /* clean dev list */
4632         for_each_netdev(&init_net, dev) {
4633                 if (__in6_dev_get(dev) == NULL)
4634                         continue;
4635                 addrconf_ifdown(dev, 1);
4636         }
4637         addrconf_ifdown(init_net.loopback_dev, 2);
4638
4639         /*
4640          *      Check hash table.
4641          */
4642         write_lock_bh(&addrconf_hash_lock);
4643         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4644                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4645                         struct inet6_ifaddr *bifa;
4646
4647                         bifa = ifa;
4648                         ifa = ifa->lst_next;
4649                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4650                         /* Do not free it; something is wrong.
4651                            Now we can investigate it with debugger.
4652                          */
4653                 }
4654         }
4655         write_unlock_bh(&addrconf_hash_lock);
4656
4657         del_timer(&addr_chk_timer);
4658         rtnl_unlock();
4659 }