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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct 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 int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188         .disable_ipv6           = 0,
 189         .accept_dad             = 1,
 190 };
 191
 192 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 193         .forwarding             = 0,
 194         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 195         .mtu6                   = IPV6_MIN_MTU,
 196         .accept_ra              = 1,
 197         .accept_redirects       = 1,
 198         .autoconf               = 1,
 199         .dad_transmits          = 1,
 200         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 201         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 202         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 203 #ifdef CONFIG_IPV6_PRIVACY
 204         .use_tempaddr           = 0,
 205         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 206         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 207         .regen_max_retry        = REGEN_MAX_RETRY,
 208         .max_desync_factor      = MAX_DESYNC_FACTOR,
 209 #endif
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 1,
 224 };
 225
 226 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 227 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 228 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 229 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 230 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 231
 232 /* Check if a valid qdisc is available */
 233 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 234 {
 235         return !qdisc_tx_is_noop(dev);
 236 }
 237
 238 /* Check if a route is valid prefix route */
 239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 240 {
 241         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 242 }
 243
 244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 245 {
 246         if (del_timer(&ifp->timer))
 247                 __in6_ifa_put(ifp);
 248 }
 249
 250 enum addrconf_timer_t
 251 {
 252         AC_NONE,
 253         AC_DAD,
 254         AC_RS,
 255 };
 256
 257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 258                                enum addrconf_timer_t what,
 259                                unsigned long when)
 260 {
 261         if (!del_timer(&ifp->timer))
 262                 in6_ifa_hold(ifp);
 263
 264         switch (what) {
 265         case AC_DAD:
 266                 ifp->timer.function = addrconf_dad_timer;
 267                 break;
 268         case AC_RS:
 269                 ifp->timer.function = addrconf_rs_timer;
 270                 break;
 271         default:;
 272         }
 273         ifp->timer.expires = jiffies + when;
 274         add_timer(&ifp->timer);
 275 }
 276
 277 static int snmp6_alloc_dev(struct inet6_dev *idev)
 278 {
 279         if (snmp_mib_init((void **)idev->stats.ipv6,
 280                           sizeof(struct ipstats_mib)) < 0)
 281                 goto err_ip;
 282         if (snmp_mib_init((void **)idev->stats.icmpv6,
 283                           sizeof(struct icmpv6_mib)) < 0)
 284                 goto err_icmp;
 285         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 286                           sizeof(struct icmpv6msg_mib)) < 0)
 287                 goto err_icmpmsg;
 288
 289         return 0;
 290
 291 err_icmpmsg:
 292         snmp_mib_free((void **)idev->stats.icmpv6);
 293 err_icmp:
 294         snmp_mib_free((void **)idev->stats.ipv6);
 295 err_ip:
 296         return -ENOMEM;
 297 }
 298
 299 static void snmp6_free_dev(struct inet6_dev *idev)
 300 {
 301         snmp_mib_free((void **)idev->stats.icmpv6msg);
 302         snmp_mib_free((void **)idev->stats.icmpv6);
 303         snmp_mib_free((void **)idev->stats.ipv6);
 304 }
 305
 306 /* Nobody refers to this device, we may destroy it. */
 307
 308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 309 {
 310         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 311         kfree(idev);
 312 }
 313
 314 void in6_dev_finish_destroy(struct inet6_dev *idev)
 315 {
 316         struct net_device *dev = idev->dev;
 317
 318         WARN_ON(idev->addr_list != NULL);
 319         WARN_ON(idev->mc_list != NULL);
 320
 321 #ifdef NET_REFCNT_DEBUG
 322         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 323 #endif
 324         dev_put(dev);
 325         if (!idev->dead) {
 326                 printk("Freeing alive inet6 device %p\n", idev);
 327                 return;
 328         }
 329         snmp6_free_dev(idev);
 330         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 331 }
 332
 333 EXPORT_SYMBOL(in6_dev_finish_destroy);
 334
 335 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338
 339         ASSERT_RTNL();
 340
 341         if (dev->mtu < IPV6_MIN_MTU)
 342                 return NULL;
 343
 344         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 345
 346         if (ndev == NULL)
 347                 return NULL;
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 352         ndev->cnf.mtu6 = dev->mtu;
 353         ndev->cnf.sysctl = NULL;
 354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 355         if (ndev->nd_parms == NULL) {
 356                 kfree(ndev);
 357                 return NULL;
 358         }
 359         if (ndev->cnf.forwarding)
 360                 dev_disable_lro(dev);
 361         /* We refer to the device */
 362         dev_hold(dev);
 363
 364         if (snmp6_alloc_dev(ndev) < 0) {
 365                 ADBG((KERN_WARNING
 366                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 367                         __func__, dev->name));
 368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 369                 ndev->dead = 1;
 370                 in6_dev_finish_destroy(ndev);
 371                 return NULL;
 372         }
 373
 374         if (snmp6_register_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         /* One reference from device.  We must do this before
 385          * we invoke __ipv6_regen_rndid().
 386          */
 387         in6_dev_hold(ndev);
 388
 389         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 390                 ndev->cnf.accept_dad = -1;
 391
 392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 393         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 394                 printk(KERN_INFO
 395                        "%s: Disabled Multicast RS\n",
 396                        dev->name);
 397                 ndev->cnf.rtr_solicits = 0;
 398         }
 399 #endif
 400
 401 #ifdef CONFIG_IPV6_PRIVACY
 402         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 403         if ((dev->flags&IFF_LOOPBACK) ||
 404             dev->type == ARPHRD_TUNNEL ||
 405             dev->type == ARPHRD_TUNNEL6 ||
 406             dev->type == ARPHRD_SIT ||
 407             dev->type == ARPHRD_NONE) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Privacy Extensions\n",
 410                        dev->name);
 411                 ndev->cnf.use_tempaddr = -1;
 412         } else {
 413                 in6_dev_hold(ndev);
 414                 ipv6_regen_rndid((unsigned long) ndev);
 415         }
 416 #endif
 417
 418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 419                 ndev->if_flags |= IF_READY;
 420
 421         ipv6_mc_init_dev(ndev);
 422         ndev->tstamp = jiffies;
 423         addrconf_sysctl_register(ndev);
 424         /* protected by rtnl_lock */
 425         rcu_assign_pointer(dev->ip6_ptr, ndev);
 426
 427         /* Join all-node multicast group */
 428         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 429
 430         return ndev;
 431 }
 432
 433 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 434 {
 435         struct inet6_dev *idev;
 436
 437         ASSERT_RTNL();
 438
 439         if ((idev = __in6_dev_get(dev)) == NULL) {
 440                 if ((idev = ipv6_add_dev(dev)) == NULL)
 441                         return NULL;
 442         }
 443
 444         if (dev->flags&IFF_UP)
 445                 ipv6_mc_up(idev);
 446         return idev;
 447 }
 448
 449 #ifdef CONFIG_SYSCTL
 450 static void dev_forward_change(struct inet6_dev *idev)
 451 {
 452         struct net_device *dev;
 453         struct inet6_ifaddr *ifa;
 454
 455         if (!idev)
 456                 return;
 457         dev = idev->dev;
 458         if (idev->cnf.forwarding)
 459                 dev_disable_lro(dev);
 460         if (dev && (dev->flags & IFF_MULTICAST)) {
 461                 if (idev->cnf.forwarding)
 462                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463                 else
 464                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 465         }
 466         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 467                 if (ifa->flags&IFA_F_TENTATIVE)
 468                         continue;
 469                 if (idev->cnf.forwarding)
 470                         addrconf_join_anycast(ifa);
 471                 else
 472                         addrconf_leave_anycast(ifa);
 473         }
 474 }
 475
 476
 477 static void addrconf_forward_change(struct net *net, __s32 newf)
 478 {
 479         struct net_device *dev;
 480         struct inet6_dev *idev;
 481
 482         read_lock(&dev_base_lock);
 483         for_each_netdev(net, dev) {
 484                 rcu_read_lock();
 485                 idev = __in6_dev_get(dev);
 486                 if (idev) {
 487                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 488                         idev->cnf.forwarding = newf;
 489                         if (changed)
 490                                 dev_forward_change(idev);
 491                 }
 492                 rcu_read_unlock();
 493         }
 494         read_unlock(&dev_base_lock);
 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                 return -ERESTARTSYS;
 507
 508         if (p == &net->ipv6.devconf_all->forwarding) {
 509                 __s32 newf = net->ipv6.devconf_all->forwarding;
 510                 net->ipv6.devconf_dflt->forwarding = newf;
 511                 addrconf_forward_change(net, newf);
 512         } else if ((!*p) ^ (!old))
 513                 dev_forward_change((struct inet6_dev *)table->extra1);
 514         rtnl_unlock();
 515
 516         if (*p)
 517                 rt6_purge_dflt_routers(net);
 518         return 1;
 519 }
 520 #endif
 521
 522 /* Nobody refers to this ifaddr, destroy it */
 523
 524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 525 {
 526         WARN_ON(ifp->if_next != NULL);
 527         WARN_ON(ifp->lst_next != NULL);
 528
 529 #ifdef NET_REFCNT_DEBUG
 530         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 531 #endif
 532
 533         in6_dev_put(ifp->idev);
 534
 535         if (del_timer(&ifp->timer))
 536                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 537
 538         if (!ifp->dead) {
 539                 printk("Freeing alive inet6 address %p\n", ifp);
 540                 return;
 541         }
 542         dst_release(&ifp->rt->u.dst);
 543
 544         kfree(ifp);
 545 }
 546
 547 static void
 548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 549 {
 550         struct inet6_ifaddr *ifa, **ifap;
 551         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 552
 553         /*
 554          * Each device address list is sorted in order of scope -
 555          * global before linklocal.
 556          */
 557         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 558              ifap = &ifa->if_next) {
 559                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 560                         break;
 561         }
 562
 563         ifp->if_next = *ifap;
 564         *ifap = ifp;
 565 }
 566
 567 /*
 568  *      Hash function taken from net_alias.c
 569  */
 570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 571 {
 572         __u32 word;
 573
 574         /*
 575          * We perform the hash function over the last 64 bits of the address
 576          * This will include the IEEE address token on links that support it.
 577          */
 578
 579         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 580         word ^= (word >> 16);
 581         word ^= (word >> 8);
 582
 583         return ((word ^ (word >> 4)) & 0x0f);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         struct net *net = dev_net(idev->dev);
 595         int hash;
 596         int err = 0;
 597         int addr_type = ipv6_addr_type(addr);
 598
 599         if (addr_type == IPV6_ADDR_ANY ||
 600             addr_type & IPV6_ADDR_MULTICAST ||
 601             (!(idev->dev->flags & IFF_LOOPBACK) &&
 602              addr_type & IPV6_ADDR_LOOPBACK))
 603                 return ERR_PTR(-EADDRNOTAVAIL);
 604
 605         rcu_read_lock_bh();
 606         if (idev->dead) {
 607                 err = -ENODEV;                  /*XXX*/
 608                 goto out2;
 609         }
 610
 611         if (idev->cnf.disable_ipv6 || net->ipv6.devconf_all->disable_ipv6) {
 612                 err = -EACCES;
 613                 goto out2;
 614         }
 615
 616         write_lock(&addrconf_hash_lock);
 617
 618         /* Ignore adding duplicate addresses on an interface */
 619         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 620                 ADBG(("ipv6_add_addr: already assigned\n"));
 621                 err = -EEXIST;
 622                 goto out;
 623         }
 624
 625         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 626
 627         if (ifa == NULL) {
 628                 ADBG(("ipv6_add_addr: malloc failed\n"));
 629                 err = -ENOBUFS;
 630                 goto out;
 631         }
 632
 633         rt = addrconf_dst_alloc(idev, addr, 0);
 634         if (IS_ERR(rt)) {
 635                 err = PTR_ERR(rt);
 636                 goto out;
 637         }
 638
 639         ipv6_addr_copy(&ifa->addr, addr);
 640
 641         spin_lock_init(&ifa->lock);
 642         init_timer(&ifa->timer);
 643         ifa->timer.data = (unsigned long) ifa;
 644         ifa->scope = scope;
 645         ifa->prefix_len = pfxlen;
 646         ifa->flags = flags | IFA_F_TENTATIVE;
 647         ifa->cstamp = ifa->tstamp = jiffies;
 648
 649         ifa->rt = rt;
 650
 651         /*
 652          * part one of RFC 4429, section 3.3
 653          * We should not configure an address as
 654          * optimistic if we do not yet know the link
 655          * layer address of our nexhop router
 656          */
 657
 658         if (rt->rt6i_nexthop == NULL)
 659                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 660
 661         ifa->idev = idev;
 662         in6_dev_hold(idev);
 663         /* For caller */
 664         in6_ifa_hold(ifa);
 665
 666         /* Add to big hash table */
 667         hash = ipv6_addr_hash(addr);
 668
 669         ifa->lst_next = inet6_addr_lst[hash];
 670         inet6_addr_lst[hash] = ifa;
 671         in6_ifa_hold(ifa);
 672         write_unlock(&addrconf_hash_lock);
 673
 674         write_lock(&idev->lock);
 675         /* Add to inet6_dev unicast addr list. */
 676         ipv6_link_dev_addr(idev, ifa);
 677
 678 #ifdef CONFIG_IPV6_PRIVACY
 679         if (ifa->flags&IFA_F_TEMPORARY) {
 680                 ifa->tmp_next = idev->tempaddr_list;
 681                 idev->tempaddr_list = ifa;
 682                 in6_ifa_hold(ifa);
 683         }
 684 #endif
 685
 686         in6_ifa_hold(ifa);
 687         write_unlock(&idev->lock);
 688 out2:
 689         rcu_read_unlock_bh();
 690
 691         if (likely(err == 0))
 692                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 693         else {
 694                 kfree(ifa);
 695                 ifa = ERR_PTR(err);
 696         }
 697
 698         return ifa;
 699 out:
 700         write_unlock(&addrconf_hash_lock);
 701         goto out2;
 702 }
 703
 704 /* This function wants to get referenced ifp and releases it before return */
 705
 706 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 707 {
 708         struct inet6_ifaddr *ifa, **ifap;
 709         struct inet6_dev *idev = ifp->idev;
 710         int hash;
 711         int deleted = 0, onlink = 0;
 712         unsigned long expires = jiffies;
 713
 714         hash = ipv6_addr_hash(&ifp->addr);
 715
 716         ifp->dead = 1;
 717
 718         write_lock_bh(&addrconf_hash_lock);
 719         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 720              ifap = &ifa->lst_next) {
 721                 if (ifa == ifp) {
 722                         *ifap = ifa->lst_next;
 723                         __in6_ifa_put(ifp);
 724                         ifa->lst_next = NULL;
 725                         break;
 726                 }
 727         }
 728         write_unlock_bh(&addrconf_hash_lock);
 729
 730         write_lock_bh(&idev->lock);
 731 #ifdef CONFIG_IPV6_PRIVACY
 732         if (ifp->flags&IFA_F_TEMPORARY) {
 733                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 734                      ifap = &ifa->tmp_next) {
 735                         if (ifa == ifp) {
 736                                 *ifap = ifa->tmp_next;
 737                                 if (ifp->ifpub) {
 738                                         in6_ifa_put(ifp->ifpub);
 739                                         ifp->ifpub = NULL;
 740                                 }
 741                                 __in6_ifa_put(ifp);
 742                                 ifa->tmp_next = NULL;
 743                                 break;
 744                         }
 745                 }
 746         }
 747 #endif
 748
 749         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 750                 if (ifa == ifp) {
 751                         *ifap = ifa->if_next;
 752                         __in6_ifa_put(ifp);
 753                         ifa->if_next = NULL;
 754                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 755                                 break;
 756                         deleted = 1;
 757                         continue;
 758                 } else if (ifp->flags & IFA_F_PERMANENT) {
 759                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 760                                               ifp->prefix_len)) {
 761                                 if (ifa->flags & IFA_F_PERMANENT) {
 762                                         onlink = 1;
 763                                         if (deleted)
 764                                                 break;
 765                                 } else {
 766                                         unsigned long lifetime;
 767
 768                                         if (!onlink)
 769                                                 onlink = -1;
 770
 771                                         spin_lock(&ifa->lock);
 772
 773                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 774                                         /*
 775                                          * Note: Because this address is
 776                                          * not permanent, lifetime <
 777                                          * LONG_MAX / HZ here.
 778                                          */
 779                                         if (time_before(expires,
 780                                                         ifa->tstamp + lifetime * HZ))
 781                                                 expires = ifa->tstamp + lifetime * HZ;
 782                                         spin_unlock(&ifa->lock);
 783                                 }
 784                         }
 785                 }
 786                 ifap = &ifa->if_next;
 787         }
 788         write_unlock_bh(&idev->lock);
 789
 790         addrconf_del_timer(ifp);
 791
 792         ipv6_ifa_notify(RTM_DELADDR, ifp);
 793
 794         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 795
 796         /*
 797          * Purge or update corresponding prefix
 798          *
 799          * 1) we don't purge prefix here if address was not permanent.
 800          *    prefix is managed by its own lifetime.
 801          * 2) if there're no addresses, delete prefix.
 802          * 3) if there're still other permanent address(es),
 803          *    corresponding prefix is still permanent.
 804          * 4) otherwise, update prefix lifetime to the
 805          *    longest valid lifetime among the corresponding
 806          *    addresses on the device.
 807          *    Note: subsequent RA will update lifetime.
 808          *
 809          * --yoshfuji
 810          */
 811         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 812                 struct in6_addr prefix;
 813                 struct rt6_info *rt;
 814                 struct net *net = dev_net(ifp->idev->dev);
 815                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 816                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 817
 818                 if (rt && addrconf_is_prefix_route(rt)) {
 819                         if (onlink == 0) {
 820                                 ip6_del_rt(rt);
 821                                 rt = NULL;
 822                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 823                                 rt->rt6i_expires = expires;
 824                                 rt->rt6i_flags |= RTF_EXPIRES;
 825                         }
 826                 }
 827                 dst_release(&rt->u.dst);
 828         }
 829
 830         in6_ifa_put(ifp);
 831 }
 832
 833 #ifdef CONFIG_IPV6_PRIVACY
 834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 835 {
 836         struct inet6_dev *idev = ifp->idev;
 837         struct in6_addr addr, *tmpaddr;
 838         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 839         unsigned long regen_advance;
 840         int tmp_plen;
 841         int ret = 0;
 842         int max_addresses;
 843         u32 addr_flags;
 844
 845         write_lock(&idev->lock);
 846         if (ift) {
 847                 spin_lock_bh(&ift->lock);
 848                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 849                 spin_unlock_bh(&ift->lock);
 850                 tmpaddr = &addr;
 851         } else {
 852                 tmpaddr = NULL;
 853         }
 854 retry:
 855         in6_dev_hold(idev);
 856         if (idev->cnf.use_tempaddr <= 0) {
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_INFO
 859                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         spin_lock_bh(&ifp->lock);
 865         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 866                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         in6_ifa_hold(ifp);
 876         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 877         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 878                 spin_unlock_bh(&ifp->lock);
 879                 write_unlock(&idev->lock);
 880                 printk(KERN_WARNING
 881                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 882                 in6_ifa_put(ifp);
 883                 in6_dev_put(idev);
 884                 ret = -1;
 885                 goto out;
 886         }
 887         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 894         tmp_plen = ifp->prefix_len;
 895         max_addresses = idev->cnf.max_addresses;
 896         tmp_cstamp = ifp->cstamp;
 897         tmp_tstamp = ifp->tstamp;
 898         spin_unlock_bh(&ifp->lock);
 899
 900         regen_advance = idev->cnf.regen_max_retry *
 901                         idev->cnf.dad_transmits *
 902                         idev->nd_parms->retrans_time / HZ;
 903         write_unlock(&idev->lock);
 904
 905         /* A temporary address is created only if this calculated Preferred
 906          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 907          * an implementation must not create a temporary address with a zero
 908          * Preferred Lifetime.
 909          */
 910         if (tmp_prefered_lft <= regen_advance) {
 911                 in6_ifa_put(ifp);
 912                 in6_dev_put(idev);
 913                 ret = -1;
 914                 goto out;
 915         }
 916
 917         addr_flags = IFA_F_TEMPORARY;
 918         /* set in addrconf_prefix_rcv() */
 919         if (ifp->flags & IFA_F_OPTIMISTIC)
 920                 addr_flags |= IFA_F_OPTIMISTIC;
 921
 922         ift = !max_addresses ||
 923               ipv6_count_addresses(idev) < max_addresses ?
 924                 ipv6_add_addr(idev, &addr, tmp_plen,
 925                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 926                               addr_flags) : NULL;
 927         if (!ift || IS_ERR(ift)) {
 928                 in6_ifa_put(ifp);
 929                 in6_dev_put(idev);
 930                 printk(KERN_INFO
 931                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 932                 tmpaddr = &addr;
 933                 write_lock(&idev->lock);
 934                 goto retry;
 935         }
 936
 937         spin_lock_bh(&ift->lock);
 938         ift->ifpub = ifp;
 939         ift->valid_lft = tmp_valid_lft;
 940         ift->prefered_lft = tmp_prefered_lft;
 941         ift->cstamp = tmp_cstamp;
 942         ift->tstamp = tmp_tstamp;
 943         spin_unlock_bh(&ift->lock);
 944
 945         addrconf_dad_start(ift, 0);
 946         in6_ifa_put(ift);
 947         in6_dev_put(idev);
 948 out:
 949         return ret;
 950 }
 951 #endif
 952
 953 /*
 954  *      Choose an appropriate source address (RFC3484)
 955  */
 956 enum {
 957         IPV6_SADDR_RULE_INIT = 0,
 958         IPV6_SADDR_RULE_LOCAL,
 959         IPV6_SADDR_RULE_SCOPE,
 960         IPV6_SADDR_RULE_PREFERRED,
 961 #ifdef CONFIG_IPV6_MIP6
 962         IPV6_SADDR_RULE_HOA,
 963 #endif
 964         IPV6_SADDR_RULE_OIF,
 965         IPV6_SADDR_RULE_LABEL,
 966 #ifdef CONFIG_IPV6_PRIVACY
 967         IPV6_SADDR_RULE_PRIVACY,
 968 #endif
 969         IPV6_SADDR_RULE_ORCHID,
 970         IPV6_SADDR_RULE_PREFIX,
 971         IPV6_SADDR_RULE_MAX
 972 };
 973
 974 struct ipv6_saddr_score {
 975         int                     rule;
 976         int                     addr_type;
 977         struct inet6_ifaddr     *ifa;
 978         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 979         int                     scopedist;
 980         int                     matchlen;
 981 };
 982
 983 struct ipv6_saddr_dst {
 984         const struct in6_addr *addr;
 985         int ifindex;
 986         int scope;
 987         int label;
 988         unsigned int prefs;
 989 };
 990
 991 static inline int ipv6_saddr_preferred(int type)
 992 {
 993         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 994                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 995                 return 1;
 996         return 0;
 997 }
 998
 999 static int ipv6_get_saddr_eval(struct net *net,
1000                                struct ipv6_saddr_score *score,
1001                                struct ipv6_saddr_dst *dst,
1002                                int i)
1003 {
1004         int ret;
1005
1006         if (i <= score->rule) {
1007                 switch (i) {
1008                 case IPV6_SADDR_RULE_SCOPE:
1009                         ret = score->scopedist;
1010                         break;
1011                 case IPV6_SADDR_RULE_PREFIX:
1012                         ret = score->matchlen;
1013                         break;
1014                 default:
1015                         ret = !!test_bit(i, score->scorebits);
1016                 }
1017                 goto out;
1018         }
1019
1020         switch (i) {
1021         case IPV6_SADDR_RULE_INIT:
1022                 /* Rule 0: remember if hiscore is not ready yet */
1023                 ret = !!score->ifa;
1024                 break;
1025         case IPV6_SADDR_RULE_LOCAL:
1026                 /* Rule 1: Prefer same address */
1027                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1028                 break;
1029         case IPV6_SADDR_RULE_SCOPE:
1030                 /* Rule 2: Prefer appropriate scope
1031                  *
1032                  *      ret
1033                  *       ^
1034                  *    -1 |  d 15
1035                  *    ---+--+-+---> scope
1036                  *       |
1037                  *       |             d is scope of the destination.
1038                  *  B-d  |  \
1039                  *       |   \      <- smaller scope is better if
1040                  *  B-15 |    \        if scope is enough for destinaion.
1041                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1042                  * d-C-1 | /
1043                  *       |/         <- greater is better
1044                  *   -C  /             if scope is not enough for destination.
1045                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1046                  *
1047                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1048                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1049                  * Assume B = 0 and we get C > 29.
1050                  */
1051                 ret = __ipv6_addr_src_scope(score->addr_type);
1052                 if (ret >= dst->scope)
1053                         ret = -ret;
1054                 else
1055                         ret -= 128;     /* 30 is enough */
1056                 score->scopedist = ret;
1057                 break;
1058         case IPV6_SADDR_RULE_PREFERRED:
1059                 /* Rule 3: Avoid deprecated and optimistic addresses */
1060                 ret = ipv6_saddr_preferred(score->addr_type) ||
1061                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1062                 break;
1063 #ifdef CONFIG_IPV6_MIP6
1064         case IPV6_SADDR_RULE_HOA:
1065             {
1066                 /* Rule 4: Prefer home address */
1067                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1068                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1069                 break;
1070             }
1071 #endif
1072         case IPV6_SADDR_RULE_OIF:
1073                 /* Rule 5: Prefer outgoing interface */
1074                 ret = (!dst->ifindex ||
1075                        dst->ifindex == score->ifa->idev->dev->ifindex);
1076                 break;
1077         case IPV6_SADDR_RULE_LABEL:
1078                 /* Rule 6: Prefer matching label */
1079                 ret = ipv6_addr_label(net,
1080                                       &score->ifa->addr, score->addr_type,
1081                                       score->ifa->idev->dev->ifindex) == dst->label;
1082                 break;
1083 #ifdef CONFIG_IPV6_PRIVACY
1084         case IPV6_SADDR_RULE_PRIVACY:
1085             {
1086                 /* Rule 7: Prefer public address
1087                  * Note: prefer temprary address if use_tempaddr >= 2
1088                  */
1089                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1090                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1091                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1092                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1093                 break;
1094             }
1095 #endif
1096         case IPV6_SADDR_RULE_ORCHID:
1097                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1098                  *          non-ORCHID vs non-ORCHID
1099                  */
1100                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1101                         ipv6_addr_orchid(dst->addr));
1102                 break;
1103         case IPV6_SADDR_RULE_PREFIX:
1104                 /* Rule 8: Use longest matching prefix */
1105                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1106                                                        dst->addr);
1107                 break;
1108         default:
1109                 ret = 0;
1110         }
1111
1112         if (ret)
1113                 __set_bit(i, score->scorebits);
1114         score->rule = i;
1115 out:
1116         return ret;
1117 }
1118
1119 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1120                        const struct in6_addr *daddr, unsigned int prefs,
1121                        struct in6_addr *saddr)
1122 {
1123         struct ipv6_saddr_score scores[2],
1124                                 *score = &scores[0], *hiscore = &scores[1];
1125         struct ipv6_saddr_dst dst;
1126         struct net_device *dev;
1127         int dst_type;
1128
1129         dst_type = __ipv6_addr_type(daddr);
1130         dst.addr = daddr;
1131         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1132         dst.scope = __ipv6_addr_src_scope(dst_type);
1133         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1134         dst.prefs = prefs;
1135
1136         hiscore->rule = -1;
1137         hiscore->ifa = NULL;
1138
1139         read_lock(&dev_base_lock);
1140         rcu_read_lock();
1141
1142         for_each_netdev(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         read_unlock(&dev_base_lock);
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         if ((idev = __in6_dev_get(dev)) != NULL) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr * ifp;
1289         u8 hash = ipv6_addr_hash(addr);
1290
1291         read_lock_bh(&addrconf_hash_lock);
1292         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE)) {
1297                         if (dev == NULL || ifp->idev->dev == dev ||
1298                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299                                 break;
1300                 }
1301         }
1302         read_unlock_bh(&addrconf_hash_lock);
1303         return ifp != NULL;
1304 }
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1306
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                        struct net_device *dev)
1310 {
1311         struct inet6_ifaddr * ifp;
1312         u8 hash = ipv6_addr_hash(addr);
1313
1314         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315                 if (!net_eq(dev_net(ifp->idev->dev), net))
1316                         continue;
1317                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318                         if (dev == NULL || ifp->idev->dev == dev)
1319                                 break;
1320                 }
1321         }
1322         return ifp != NULL;
1323 }
1324
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 {
1327         struct inet6_dev *idev;
1328         struct inet6_ifaddr *ifa;
1329         int     onlink;
1330
1331         onlink = 0;
1332         rcu_read_lock();
1333         idev = __in6_dev_get(dev);
1334         if (idev) {
1335                 read_lock_bh(&idev->lock);
1336                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338                                                    ifa->prefix_len);
1339                         if (onlink)
1340                                 break;
1341                 }
1342                 read_unlock_bh(&idev->lock);
1343         }
1344         rcu_read_unlock();
1345         return onlink;
1346 }
1347
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1349
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351                                      struct net_device *dev, int strict)
1352 {
1353         struct inet6_ifaddr * ifp;
1354         u8 hash = ipv6_addr_hash(addr);
1355
1356         read_lock_bh(&addrconf_hash_lock);
1357         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358                 if (!net_eq(dev_net(ifp->idev->dev), net))
1359                         continue;
1360                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361                         if (dev == NULL || ifp->idev->dev == dev ||
1362                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363                                 in6_ifa_hold(ifp);
1364                                 break;
1365                         }
1366                 }
1367         }
1368         read_unlock_bh(&addrconf_hash_lock);
1369
1370         return ifp;
1371 }
1372
1373 /* Gets referenced address, destroys ifaddr */
1374
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1376 {
1377         if (ifp->flags&IFA_F_PERMANENT) {
1378                 spin_lock_bh(&ifp->lock);
1379                 addrconf_del_timer(ifp);
1380                 ifp->flags |= IFA_F_TENTATIVE;
1381                 spin_unlock_bh(&ifp->lock);
1382                 in6_ifa_put(ifp);
1383 #ifdef CONFIG_IPV6_PRIVACY
1384         } else if (ifp->flags&IFA_F_TEMPORARY) {
1385                 struct inet6_ifaddr *ifpub;
1386                 spin_lock_bh(&ifp->lock);
1387                 ifpub = ifp->ifpub;
1388                 if (ifpub) {
1389                         in6_ifa_hold(ifpub);
1390                         spin_unlock_bh(&ifp->lock);
1391                         ipv6_create_tempaddr(ifpub, ifp);
1392                         in6_ifa_put(ifpub);
1393                 } else {
1394                         spin_unlock_bh(&ifp->lock);
1395                 }
1396                 ipv6_del_addr(ifp);
1397 #endif
1398         } else
1399                 ipv6_del_addr(ifp);
1400 }
1401
1402 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1403 {
1404         struct inet6_dev *idev = ifp->idev;
1405
1406         if (net_ratelimit())
1407                 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1408                         ifp->idev->dev->name);
1409
1410         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1411                 struct in6_addr addr;
1412
1413                 addr.s6_addr32[0] = htonl(0xfe800000);
1414                 addr.s6_addr32[1] = 0;
1415
1416                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1417                     ipv6_addr_equal(&ifp->addr, &addr)) {
1418                         /* DAD failed for link-local based on MAC address */
1419                         idev->cnf.disable_ipv6 = 1;
1420
1421                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1422                                 ifp->idev->dev->name);
1423                 }
1424         }
1425
1426         addrconf_dad_stop(ifp);
1427 }
1428
1429 /* Join to solicited addr multicast group. */
1430
1431 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1432 {
1433         struct in6_addr maddr;
1434
1435         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1436                 return;
1437
1438         addrconf_addr_solict_mult(addr, &maddr);
1439         ipv6_dev_mc_inc(dev, &maddr);
1440 }
1441
1442 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1443 {
1444         struct in6_addr maddr;
1445
1446         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1447                 return;
1448
1449         addrconf_addr_solict_mult(addr, &maddr);
1450         __ipv6_dev_mc_dec(idev, &maddr);
1451 }
1452
1453 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1454 {
1455         struct in6_addr addr;
1456         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1457         if (ipv6_addr_any(&addr))
1458                 return;
1459         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1460 }
1461
1462 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1463 {
1464         struct in6_addr addr;
1465         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1466         if (ipv6_addr_any(&addr))
1467                 return;
1468         __ipv6_dev_ac_dec(ifp->idev, &addr);
1469 }
1470
1471 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1472 {
1473         if (dev->addr_len != ETH_ALEN)
1474                 return -1;
1475         memcpy(eui, dev->dev_addr, 3);
1476         memcpy(eui + 5, dev->dev_addr + 3, 3);
1477
1478         /*
1479          * The zSeries OSA network cards can be shared among various
1480          * OS instances, but the OSA cards have only one MAC address.
1481          * This leads to duplicate address conflicts in conjunction
1482          * with IPv6 if more than one instance uses the same card.
1483          *
1484          * The driver for these cards can deliver a unique 16-bit
1485          * identifier for each instance sharing the same card.  It is
1486          * placed instead of 0xFFFE in the interface identifier.  The
1487          * "u" bit of the interface identifier is not inverted in this
1488          * case.  Hence the resulting interface identifier has local
1489          * scope according to RFC2373.
1490          */
1491         if (dev->dev_id) {
1492                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1493                 eui[4] = dev->dev_id & 0xFF;
1494         } else {
1495                 eui[3] = 0xFF;
1496                 eui[4] = 0xFE;
1497                 eui[0] ^= 2;
1498         }
1499         return 0;
1500 }
1501
1502 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1503 {
1504         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1505         if (dev->addr_len != ARCNET_ALEN)
1506                 return -1;
1507         memset(eui, 0, 7);
1508         eui[7] = *(u8*)dev->dev_addr;
1509         return 0;
1510 }
1511
1512 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1513 {
1514         if (dev->addr_len != INFINIBAND_ALEN)
1515                 return -1;
1516         memcpy(eui, dev->dev_addr + 12, 8);
1517         eui[0] |= 2;
1518         return 0;
1519 }
1520
1521 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1522 {
1523         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1524                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1525                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1526                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1527                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1528                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1529         eui[1] = 0;
1530         eui[2] = 0x5E;
1531         eui[3] = 0xFE;
1532         memcpy(eui + 4, &addr, 4);
1533         return 0;
1534 }
1535 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1536
1537 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1538 {
1539         if (dev->priv_flags & IFF_ISATAP)
1540                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1541         return -1;
1542 }
1543
1544 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1545 {
1546         switch (dev->type) {
1547         case ARPHRD_ETHER:
1548         case ARPHRD_FDDI:
1549         case ARPHRD_IEEE802_TR:
1550                 return addrconf_ifid_eui48(eui, dev);
1551         case ARPHRD_ARCNET:
1552                 return addrconf_ifid_arcnet(eui, dev);
1553         case ARPHRD_INFINIBAND:
1554                 return addrconf_ifid_infiniband(eui, dev);
1555         case ARPHRD_SIT:
1556                 return addrconf_ifid_sit(eui, dev);
1557         }
1558         return -1;
1559 }
1560
1561 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1562 {
1563         int err = -1;
1564         struct inet6_ifaddr *ifp;
1565
1566         read_lock_bh(&idev->lock);
1567         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1568                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1569                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1570                         err = 0;
1571                         break;
1572                 }
1573         }
1574         read_unlock_bh(&idev->lock);
1575         return err;
1576 }
1577
1578 #ifdef CONFIG_IPV6_PRIVACY
1579 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1580 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1581 {
1582 regen:
1583         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1584         idev->rndid[0] &= ~0x02;
1585
1586         /*
1587          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1588          * check if generated address is not inappropriate
1589          *
1590          *  - Reserved subnet anycast (RFC 2526)
1591          *      11111101 11....11 1xxxxxxx
1592          *  - ISATAP (RFC4214) 6.1
1593          *      00-00-5E-FE-xx-xx-xx-xx
1594          *  - value 0
1595          *  - XXX: already assigned to an address on the device
1596          */
1597         if (idev->rndid[0] == 0xfd &&
1598             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1599             (idev->rndid[7]&0x80))
1600                 goto regen;
1601         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1602                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1603                         goto regen;
1604                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1605                         goto regen;
1606         }
1607
1608         return 0;
1609 }
1610
1611 static void ipv6_regen_rndid(unsigned long data)
1612 {
1613         struct inet6_dev *idev = (struct inet6_dev *) data;
1614         unsigned long expires;
1615
1616         rcu_read_lock_bh();
1617         write_lock_bh(&idev->lock);
1618
1619         if (idev->dead)
1620                 goto out;
1621
1622         if (__ipv6_regen_rndid(idev) < 0)
1623                 goto out;
1624
1625         expires = jiffies +
1626                 idev->cnf.temp_prefered_lft * HZ -
1627                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1628         if (time_before(expires, jiffies)) {
1629                 printk(KERN_WARNING
1630                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1631                         idev->dev->name);
1632                 goto out;
1633         }
1634
1635         if (!mod_timer(&idev->regen_timer, expires))
1636                 in6_dev_hold(idev);
1637
1638 out:
1639         write_unlock_bh(&idev->lock);
1640         rcu_read_unlock_bh();
1641         in6_dev_put(idev);
1642 }
1643
1644 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1645         int ret = 0;
1646
1647         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1648                 ret = __ipv6_regen_rndid(idev);
1649         return ret;
1650 }
1651 #endif
1652
1653 /*
1654  *      Add prefix route.
1655  */
1656
1657 static void
1658 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1659                       unsigned long expires, u32 flags)
1660 {
1661         struct fib6_config cfg = {
1662                 .fc_table = RT6_TABLE_PREFIX,
1663                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1664                 .fc_ifindex = dev->ifindex,
1665                 .fc_expires = expires,
1666                 .fc_dst_len = plen,
1667                 .fc_flags = RTF_UP | flags,
1668                 .fc_nlinfo.nl_net = dev_net(dev),
1669                 .fc_protocol = RTPROT_KERNEL,
1670         };
1671
1672         ipv6_addr_copy(&cfg.fc_dst, pfx);
1673
1674         /* Prevent useless cloning on PtP SIT.
1675            This thing is done here expecting that the whole
1676            class of non-broadcast devices need not cloning.
1677          */
1678 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1679         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1680                 cfg.fc_flags |= RTF_NONEXTHOP;
1681 #endif
1682
1683         ip6_route_add(&cfg);
1684 }
1685
1686 /* Create "default" multicast route to the interface */
1687
1688 static void addrconf_add_mroute(struct net_device *dev)
1689 {
1690         struct fib6_config cfg = {
1691                 .fc_table = RT6_TABLE_LOCAL,
1692                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1693                 .fc_ifindex = dev->ifindex,
1694                 .fc_dst_len = 8,
1695                 .fc_flags = RTF_UP,
1696                 .fc_nlinfo.nl_net = dev_net(dev),
1697         };
1698
1699         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1700
1701         ip6_route_add(&cfg);
1702 }
1703
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705 static void sit_route_add(struct net_device *dev)
1706 {
1707         struct fib6_config cfg = {
1708                 .fc_table = RT6_TABLE_MAIN,
1709                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710                 .fc_ifindex = dev->ifindex,
1711                 .fc_dst_len = 96,
1712                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1713                 .fc_nlinfo.nl_net = dev_net(dev),
1714         };
1715
1716         /* prefix length - 96 bits "::d.d.d.d" */
1717         ip6_route_add(&cfg);
1718 }
1719 #endif
1720
1721 static void addrconf_add_lroute(struct net_device *dev)
1722 {
1723         struct in6_addr addr;
1724
1725         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1726         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1727 }
1728
1729 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1730 {
1731         struct inet6_dev *idev;
1732
1733         ASSERT_RTNL();
1734
1735         if ((idev = ipv6_find_idev(dev)) == NULL)
1736                 return NULL;
1737
1738         /* Add default multicast route */
1739         addrconf_add_mroute(dev);
1740
1741         /* Add link local route */
1742         addrconf_add_lroute(dev);
1743         return idev;
1744 }
1745
1746 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1747 {
1748         struct prefix_info *pinfo;
1749         __u32 valid_lft;
1750         __u32 prefered_lft;
1751         int addr_type;
1752         struct inet6_dev *in6_dev;
1753
1754         pinfo = (struct prefix_info *) opt;
1755
1756         if (len < sizeof(struct prefix_info)) {
1757                 ADBG(("addrconf: prefix option too short\n"));
1758                 return;
1759         }
1760
1761         /*
1762          *      Validation checks ([ADDRCONF], page 19)
1763          */
1764
1765         addr_type = ipv6_addr_type(&pinfo->prefix);
1766
1767         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1768                 return;
1769
1770         valid_lft = ntohl(pinfo->valid);
1771         prefered_lft = ntohl(pinfo->prefered);
1772
1773         if (prefered_lft > valid_lft) {
1774                 if (net_ratelimit())
1775                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1776                 return;
1777         }
1778
1779         in6_dev = in6_dev_get(dev);
1780
1781         if (in6_dev == NULL) {
1782                 if (net_ratelimit())
1783                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1784                 return;
1785         }
1786
1787         /*
1788          *      Two things going on here:
1789          *      1) Add routes for on-link prefixes
1790          *      2) Configure prefixes with the auto flag set
1791          */
1792
1793         if (pinfo->onlink) {
1794                 struct rt6_info *rt;
1795                 unsigned long rt_expires;
1796
1797                 /* Avoid arithmetic overflow. Really, we could
1798                  * save rt_expires in seconds, likely valid_lft,
1799                  * but it would require division in fib gc, that it
1800                  * not good.
1801                  */
1802                 if (HZ > USER_HZ)
1803                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1804                 else
1805                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1806
1807                 if (addrconf_finite_timeout(rt_expires))
1808                         rt_expires *= HZ;
1809
1810                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1811                                 dev->ifindex, 1);
1812
1813                 if (rt && addrconf_is_prefix_route(rt)) {
1814                         /* Autoconf prefix route */
1815                         if (valid_lft == 0) {
1816                                 ip6_del_rt(rt);
1817                                 rt = NULL;
1818                         } else if (addrconf_finite_timeout(rt_expires)) {
1819                                 /* not infinity */
1820                                 rt->rt6i_expires = jiffies + rt_expires;
1821                                 rt->rt6i_flags |= RTF_EXPIRES;
1822                         } else {
1823                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1824                                 rt->rt6i_expires = 0;
1825                         }
1826                 } else if (valid_lft) {
1827                         clock_t expires = 0;
1828                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1829                         if (addrconf_finite_timeout(rt_expires)) {
1830                                 /* not infinity */
1831                                 flags |= RTF_EXPIRES;
1832                                 expires = jiffies_to_clock_t(rt_expires);
1833                         }
1834                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1835                                               dev, expires, flags);
1836                 }
1837                 if (rt)
1838                         dst_release(&rt->u.dst);
1839         }
1840
1841         /* Try to figure out our local address for this prefix */
1842
1843         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1844                 struct inet6_ifaddr * ifp;
1845                 struct in6_addr addr;
1846                 int create = 0, update_lft = 0;
1847                 struct net *net = dev_net(dev);
1848
1849                 if (pinfo->prefix_len == 64) {
1850                         memcpy(&addr, &pinfo->prefix, 8);
1851                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1852                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1853                                 in6_dev_put(in6_dev);
1854                                 return;
1855                         }
1856                         goto ok;
1857                 }
1858                 if (net_ratelimit())
1859                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1860                                pinfo->prefix_len);
1861                 in6_dev_put(in6_dev);
1862                 return;
1863
1864 ok:
1865
1866                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1867
1868                 if (ifp == NULL && valid_lft) {
1869                         int max_addresses = in6_dev->cnf.max_addresses;
1870                         u32 addr_flags = 0;
1871
1872 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1873                         if (in6_dev->cnf.optimistic_dad &&
1874                             !net->ipv6.devconf_all->forwarding)
1875                                 addr_flags = IFA_F_OPTIMISTIC;
1876 #endif
1877
1878                         /* Do not allow to create too much of autoconfigured
1879                          * addresses; this would be too easy way to crash kernel.
1880                          */
1881                         if (!max_addresses ||
1882                             ipv6_count_addresses(in6_dev) < max_addresses)
1883                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1884                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1885                                                     addr_flags);
1886
1887                         if (!ifp || IS_ERR(ifp)) {
1888                                 in6_dev_put(in6_dev);
1889                                 return;
1890                         }
1891
1892                         update_lft = create = 1;
1893                         ifp->cstamp = jiffies;
1894                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1895                 }
1896
1897                 if (ifp) {
1898                         int flags;
1899                         unsigned long now;
1900 #ifdef CONFIG_IPV6_PRIVACY
1901                         struct inet6_ifaddr *ift;
1902 #endif
1903                         u32 stored_lft;
1904
1905                         /* update lifetime (RFC2462 5.5.3 e) */
1906                         spin_lock(&ifp->lock);
1907                         now = jiffies;
1908                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1909                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1910                         else
1911                                 stored_lft = 0;
1912                         if (!update_lft && stored_lft) {
1913                                 if (valid_lft > MIN_VALID_LIFETIME ||
1914                                     valid_lft > stored_lft)
1915                                         update_lft = 1;
1916                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1917                                         /* valid_lft <= stored_lft is always true */
1918                                         /* XXX: IPsec */
1919                                         update_lft = 0;
1920                                 } else {
1921                                         valid_lft = MIN_VALID_LIFETIME;
1922                                         if (valid_lft < prefered_lft)
1923                                                 prefered_lft = valid_lft;
1924                                         update_lft = 1;
1925                                 }
1926                         }
1927
1928                         if (update_lft) {
1929                                 ifp->valid_lft = valid_lft;
1930                                 ifp->prefered_lft = prefered_lft;
1931                                 ifp->tstamp = now;
1932                                 flags = ifp->flags;
1933                                 ifp->flags &= ~IFA_F_DEPRECATED;
1934                                 spin_unlock(&ifp->lock);
1935
1936                                 if (!(flags&IFA_F_TENTATIVE))
1937                                         ipv6_ifa_notify(0, ifp);
1938                         } else
1939                                 spin_unlock(&ifp->lock);
1940
1941 #ifdef CONFIG_IPV6_PRIVACY
1942                         read_lock_bh(&in6_dev->lock);
1943                         /* update all temporary addresses in the list */
1944                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1945                                 /*
1946                                  * When adjusting the lifetimes of an existing
1947                                  * temporary address, only lower the lifetimes.
1948                                  * Implementations must not increase the
1949                                  * lifetimes of an existing temporary address
1950                                  * when processing a Prefix Information Option.
1951                                  */
1952                                 if (ifp != ift->ifpub)
1953                                         continue;
1954
1955                                 spin_lock(&ift->lock);
1956                                 flags = ift->flags;
1957                                 if (ift->valid_lft > valid_lft &&
1958                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1959                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1960                                 if (ift->prefered_lft > prefered_lft &&
1961                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1962                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1963                                 spin_unlock(&ift->lock);
1964                                 if (!(flags&IFA_F_TENTATIVE))
1965                                         ipv6_ifa_notify(0, ift);
1966                         }
1967
1968                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1969                                 /*
1970                                  * When a new public address is created as described in [ADDRCONF],
1971                                  * also create a new temporary address.
1972                                  */
1973                                 read_unlock_bh(&in6_dev->lock);
1974                                 ipv6_create_tempaddr(ifp, NULL);
1975                         } else {
1976                                 read_unlock_bh(&in6_dev->lock);
1977                         }
1978 #endif
1979                         in6_ifa_put(ifp);
1980                         addrconf_verify(0);
1981                 }
1982         }
1983         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1984         in6_dev_put(in6_dev);
1985 }
1986
1987 /*
1988  *      Set destination address.
1989  *      Special case for SIT interfaces where we create a new "virtual"
1990  *      device.
1991  */
1992 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1993 {
1994         struct in6_ifreq ireq;
1995         struct net_device *dev;
1996         int err = -EINVAL;
1997
1998         rtnl_lock();
1999
2000         err = -EFAULT;
2001         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2002                 goto err_exit;
2003
2004         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2005
2006         err = -ENODEV;
2007         if (dev == NULL)
2008                 goto err_exit;
2009
2010 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2011         if (dev->type == ARPHRD_SIT) {
2012                 const struct net_device_ops *ops = dev->netdev_ops;
2013                 struct ifreq ifr;
2014                 struct ip_tunnel_parm p;
2015
2016                 err = -EADDRNOTAVAIL;
2017                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2018                         goto err_exit;
2019
2020                 memset(&p, 0, sizeof(p));
2021                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2022                 p.iph.saddr = 0;
2023                 p.iph.version = 4;
2024                 p.iph.ihl = 5;
2025                 p.iph.protocol = IPPROTO_IPV6;
2026                 p.iph.ttl = 64;
2027                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2028
2029                 if (ops->ndo_do_ioctl) {
2030                         mm_segment_t oldfs = get_fs();
2031
2032                         set_fs(KERNEL_DS);
2033                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2034                         set_fs(oldfs);
2035                 } else
2036                         err = -EOPNOTSUPP;
2037
2038                 if (err == 0) {
2039                         err = -ENOBUFS;
2040                         dev = __dev_get_by_name(net, p.name);
2041                         if (!dev)
2042                                 goto err_exit;
2043                         err = dev_open(dev);
2044                 }
2045         }
2046 #endif
2047
2048 err_exit:
2049         rtnl_unlock();
2050         return err;
2051 }
2052
2053 /*
2054  *      Manual configuration of address on an interface
2055  */
2056 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2057                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2058                           __u32 valid_lft)
2059 {
2060         struct inet6_ifaddr *ifp;
2061         struct inet6_dev *idev;
2062         struct net_device *dev;
2063         int scope;
2064         u32 flags;
2065         clock_t expires;
2066         unsigned long timeout;
2067
2068         ASSERT_RTNL();
2069
2070         if (plen > 128)
2071                 return -EINVAL;
2072
2073         /* check the lifetime */
2074         if (!valid_lft || prefered_lft > valid_lft)
2075                 return -EINVAL;
2076
2077         dev = __dev_get_by_index(net, ifindex);
2078         if (!dev)
2079                 return -ENODEV;
2080
2081         if ((idev = addrconf_add_dev(dev)) == NULL)
2082                 return -ENOBUFS;
2083
2084         scope = ipv6_addr_scope(pfx);
2085
2086         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2087         if (addrconf_finite_timeout(timeout)) {
2088                 expires = jiffies_to_clock_t(timeout * HZ);
2089                 valid_lft = timeout;
2090                 flags = RTF_EXPIRES;
2091         } else {
2092                 expires = 0;
2093                 flags = 0;
2094                 ifa_flags |= IFA_F_PERMANENT;
2095         }
2096
2097         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2098         if (addrconf_finite_timeout(timeout)) {
2099                 if (timeout == 0)
2100                         ifa_flags |= IFA_F_DEPRECATED;
2101                 prefered_lft = timeout;
2102         }
2103
2104         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2105
2106         if (!IS_ERR(ifp)) {
2107                 spin_lock_bh(&ifp->lock);
2108                 ifp->valid_lft = valid_lft;
2109                 ifp->prefered_lft = prefered_lft;
2110                 ifp->tstamp = jiffies;
2111                 spin_unlock_bh(&ifp->lock);
2112
2113                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2114                                       expires, flags);
2115                 /*
2116                  * Note that section 3.1 of RFC 4429 indicates
2117                  * that the Optimistic flag should not be set for
2118                  * manually configured addresses
2119                  */
2120                 addrconf_dad_start(ifp, 0);
2121                 in6_ifa_put(ifp);
2122                 addrconf_verify(0);
2123                 return 0;
2124         }
2125
2126         return PTR_ERR(ifp);
2127 }
2128
2129 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2130                           unsigned int plen)
2131 {
2132         struct inet6_ifaddr *ifp;
2133         struct inet6_dev *idev;
2134         struct net_device *dev;
2135
2136         if (plen > 128)
2137                 return -EINVAL;
2138
2139         dev = __dev_get_by_index(net, ifindex);
2140         if (!dev)
2141                 return -ENODEV;
2142
2143         if ((idev = __in6_dev_get(dev)) == NULL)
2144                 return -ENXIO;
2145
2146         read_lock_bh(&idev->lock);
2147         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2148                 if (ifp->prefix_len == plen &&
2149                     ipv6_addr_equal(pfx, &ifp->addr)) {
2150                         in6_ifa_hold(ifp);
2151                         read_unlock_bh(&idev->lock);
2152
2153                         ipv6_del_addr(ifp);
2154
2155                         /* If the last address is deleted administratively,
2156                            disable IPv6 on this interface.
2157                          */
2158                         if (idev->addr_list == NULL)
2159                                 addrconf_ifdown(idev->dev, 1);
2160                         return 0;
2161                 }
2162         }
2163         read_unlock_bh(&idev->lock);
2164         return -EADDRNOTAVAIL;
2165 }
2166
2167
2168 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2169 {
2170         struct in6_ifreq ireq;
2171         int err;
2172
2173         if (!capable(CAP_NET_ADMIN))
2174                 return -EPERM;
2175
2176         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2177                 return -EFAULT;
2178
2179         rtnl_lock();
2180         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2181                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2182                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2183         rtnl_unlock();
2184         return err;
2185 }
2186
2187 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2188 {
2189         struct in6_ifreq ireq;
2190         int err;
2191
2192         if (!capable(CAP_NET_ADMIN))
2193                 return -EPERM;
2194
2195         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2196                 return -EFAULT;
2197
2198         rtnl_lock();
2199         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2200                              ireq.ifr6_prefixlen);
2201         rtnl_unlock();
2202         return err;
2203 }
2204
2205 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2206                      int plen, int scope)
2207 {
2208         struct inet6_ifaddr *ifp;
2209
2210         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2211         if (!IS_ERR(ifp)) {
2212                 spin_lock_bh(&ifp->lock);
2213                 ifp->flags &= ~IFA_F_TENTATIVE;
2214                 spin_unlock_bh(&ifp->lock);
2215                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2216                 in6_ifa_put(ifp);
2217         }
2218 }
2219
2220 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2221 static void sit_add_v4_addrs(struct inet6_dev *idev)
2222 {
2223         struct in6_addr addr;
2224         struct net_device *dev;
2225         struct net *net = dev_net(idev->dev);
2226         int scope;
2227
2228         ASSERT_RTNL();
2229
2230         memset(&addr, 0, sizeof(struct in6_addr));
2231         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2232
2233         if (idev->dev->flags&IFF_POINTOPOINT) {
2234                 addr.s6_addr32[0] = htonl(0xfe800000);
2235                 scope = IFA_LINK;
2236         } else {
2237                 scope = IPV6_ADDR_COMPATv4;
2238         }
2239
2240         if (addr.s6_addr32[3]) {
2241                 add_addr(idev, &addr, 128, scope);
2242                 return;
2243         }
2244
2245         for_each_netdev(net, dev) {
2246                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2247                 if (in_dev && (dev->flags & IFF_UP)) {
2248                         struct in_ifaddr * ifa;
2249
2250                         int flag = scope;
2251
2252                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2253                                 int plen;
2254
2255                                 addr.s6_addr32[3] = ifa->ifa_local;
2256
2257                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2258                                         continue;
2259                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2260                                         if (idev->dev->flags&IFF_POINTOPOINT)
2261                                                 continue;
2262                                         flag |= IFA_HOST;
2263                                 }
2264                                 if (idev->dev->flags&IFF_POINTOPOINT)
2265                                         plen = 64;
2266                                 else
2267                                         plen = 96;
2268
2269                                 add_addr(idev, &addr, plen, flag);
2270                         }
2271                 }
2272         }
2273 }
2274 #endif
2275
2276 static void init_loopback(struct net_device *dev)
2277 {
2278         struct inet6_dev  *idev;
2279
2280         /* ::1 */
2281
2282         ASSERT_RTNL();
2283
2284         if ((idev = ipv6_find_idev(dev)) == NULL) {
2285                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2286                 return;
2287         }
2288
2289         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2290 }
2291
2292 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2293 {
2294         struct inet6_ifaddr * ifp;
2295         u32 addr_flags = IFA_F_PERMANENT;
2296
2297 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2298         if (idev->cnf.optimistic_dad &&
2299             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2300                 addr_flags |= IFA_F_OPTIMISTIC;
2301 #endif
2302
2303
2304         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2305         if (!IS_ERR(ifp)) {
2306                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2307                 addrconf_dad_start(ifp, 0);
2308                 in6_ifa_put(ifp);
2309         }
2310 }
2311
2312 static void addrconf_dev_config(struct net_device *dev)
2313 {
2314         struct in6_addr addr;
2315         struct inet6_dev    * idev;
2316
2317         ASSERT_RTNL();
2318
2319         if ((dev->type != ARPHRD_ETHER) &&
2320             (dev->type != ARPHRD_FDDI) &&
2321             (dev->type != ARPHRD_IEEE802_TR) &&
2322             (dev->type != ARPHRD_ARCNET) &&
2323             (dev->type != ARPHRD_INFINIBAND)) {
2324                 /* Alas, we support only Ethernet autoconfiguration. */
2325                 return;
2326         }
2327
2328         idev = addrconf_add_dev(dev);
2329         if (idev == NULL)
2330                 return;
2331
2332         memset(&addr, 0, sizeof(struct in6_addr));
2333         addr.s6_addr32[0] = htonl(0xFE800000);
2334
2335         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2336                 addrconf_add_linklocal(idev, &addr);
2337 }
2338
2339 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2340 static void addrconf_sit_config(struct net_device *dev)
2341 {
2342         struct inet6_dev *idev;
2343
2344         ASSERT_RTNL();
2345
2346         /*
2347          * Configure the tunnel with one of our IPv4
2348          * addresses... we should configure all of
2349          * our v4 addrs in the tunnel
2350          */
2351
2352         if ((idev = ipv6_find_idev(dev)) == NULL) {
2353                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2354                 return;
2355         }
2356
2357         if (dev->priv_flags & IFF_ISATAP) {
2358                 struct in6_addr addr;
2359
2360                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2361                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2362                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2363                         addrconf_add_linklocal(idev, &addr);
2364                 return;
2365         }
2366
2367         sit_add_v4_addrs(idev);
2368
2369         if (dev->flags&IFF_POINTOPOINT) {
2370                 addrconf_add_mroute(dev);
2371                 addrconf_add_lroute(dev);
2372         } else
2373                 sit_route_add(dev);
2374 }
2375 #endif
2376
2377 static inline int
2378 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2379 {
2380         struct in6_addr lladdr;
2381
2382         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2383                 addrconf_add_linklocal(idev, &lladdr);
2384                 return 0;
2385         }
2386         return -1;
2387 }
2388
2389 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2390 {
2391         struct net_device *link_dev;
2392         struct net *net = dev_net(idev->dev);
2393
2394         /* first try to inherit the link-local address from the link device */
2395         if (idev->dev->iflink &&
2396             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2397                 if (!ipv6_inherit_linklocal(idev, link_dev))
2398                         return;
2399         }
2400         /* then try to inherit it from any device */
2401         for_each_netdev(net, link_dev) {
2402                 if (!ipv6_inherit_linklocal(idev, link_dev))
2403                         return;
2404         }
2405         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2406 }
2407
2408 /*
2409  * Autoconfigure tunnel with a link-local address so routing protocols,
2410  * DHCPv6, MLD etc. can be run over the virtual link
2411  */
2412
2413 static void addrconf_ip6_tnl_config(struct net_device *dev)
2414 {
2415         struct inet6_dev *idev;
2416
2417         ASSERT_RTNL();
2418
2419         if ((idev = addrconf_add_dev(dev)) == NULL) {
2420                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2421                 return;
2422         }
2423         ip6_tnl_add_linklocal(idev);
2424 }
2425
2426 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2427                            void * data)
2428 {
2429         struct net_device *dev = (struct net_device *) data;
2430         struct inet6_dev *idev = __in6_dev_get(dev);
2431         int run_pending = 0;
2432         int err;
2433
2434         switch(event) {
2435         case NETDEV_REGISTER:
2436                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2437                         idev = ipv6_add_dev(dev);
2438                         if (!idev)
2439                                 return notifier_from_errno(-ENOMEM);
2440                 }
2441                 break;
2442         case NETDEV_UP:
2443         case NETDEV_CHANGE:
2444                 if (dev->flags & IFF_SLAVE)
2445                         break;
2446
2447                 if (event == NETDEV_UP) {
2448                         if (!addrconf_qdisc_ok(dev)) {
2449                                 /* device is not ready yet. */
2450                                 printk(KERN_INFO
2451                                         "ADDRCONF(NETDEV_UP): %s: "
2452                                         "link is not ready\n",
2453                                         dev->name);
2454                                 break;
2455                         }
2456
2457                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2458                                 idev = ipv6_add_dev(dev);
2459
2460                         if (idev) {
2461                                 idev->if_flags |= IF_READY;
2462                                 run_pending = 1;
2463                         }
2464                 } else {
2465                         if (!addrconf_qdisc_ok(dev)) {
2466                                 /* device is still not ready. */
2467                                 break;
2468                         }
2469
2470                         if (idev) {
2471                                 if (idev->if_flags & IF_READY) {
2472                                         /* device is already configured. */
2473                                         break;
2474                                 }
2475                                 idev->if_flags |= IF_READY;
2476                         }
2477
2478                         printk(KERN_INFO
2479                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2480                                         "link becomes ready\n",
2481                                         dev->name);
2482
2483                         run_pending = 1;
2484                 }
2485
2486                 switch(dev->type) {
2487 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2488                 case ARPHRD_SIT:
2489                         addrconf_sit_config(dev);
2490                         break;
2491 #endif
2492                 case ARPHRD_TUNNEL6:
2493                         addrconf_ip6_tnl_config(dev);
2494                         break;
2495                 case ARPHRD_LOOPBACK:
2496                         init_loopback(dev);
2497                         break;
2498
2499                 default:
2500                         addrconf_dev_config(dev);
2501                         break;
2502                 }
2503                 if (idev) {
2504                         if (run_pending)
2505                                 addrconf_dad_run(idev);
2506
2507                         /* If the MTU changed during the interface down, when the
2508                            interface up, the changed MTU must be reflected in the
2509                            idev as well as routers.
2510                          */
2511                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2512                                 rt6_mtu_change(dev, dev->mtu);
2513                                 idev->cnf.mtu6 = dev->mtu;
2514                         }
2515                         idev->tstamp = jiffies;
2516                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2517                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2518                            stop IPv6 on this interface.
2519                          */
2520                         if (dev->mtu < IPV6_MIN_MTU)
2521                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2522                 }
2523                 break;
2524
2525         case NETDEV_CHANGEMTU:
2526                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2527                         rt6_mtu_change(dev, dev->mtu);
2528                         idev->cnf.mtu6 = dev->mtu;
2529                         break;
2530                 }
2531
2532                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2533                         idev = ipv6_add_dev(dev);
2534                         if (idev)
2535                                 break;
2536                 }
2537
2538                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2539
2540         case NETDEV_DOWN:
2541         case NETDEV_UNREGISTER:
2542                 /*
2543                  *      Remove all addresses from this interface.
2544                  */
2545                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2546                 break;
2547
2548         case NETDEV_CHANGENAME:
2549                 if (idev) {
2550                         snmp6_unregister_dev(idev);
2551                         addrconf_sysctl_unregister(idev);
2552                         addrconf_sysctl_register(idev);
2553                         err = snmp6_register_dev(idev);
2554                         if (err)
2555                                 return notifier_from_errno(err);
2556                 }
2557                 break;
2558         }
2559
2560         return NOTIFY_OK;
2561 }
2562
2563 /*
2564  *      addrconf module should be notified of a device going up
2565  */
2566 static struct notifier_block ipv6_dev_notf = {
2567         .notifier_call = addrconf_notify,
2568         .priority = 0
2569 };
2570
2571 static int addrconf_ifdown(struct net_device *dev, int how)
2572 {
2573         struct inet6_dev *idev;
2574         struct inet6_ifaddr *ifa, **bifa;
2575         struct net *net = dev_net(dev);
2576         int i;
2577
2578         ASSERT_RTNL();
2579
2580         rt6_ifdown(net, dev);
2581         neigh_ifdown(&nd_tbl, dev);
2582
2583         idev = __in6_dev_get(dev);
2584         if (idev == NULL)
2585                 return -ENODEV;
2586
2587         /* Step 1: remove reference to ipv6 device from parent device.
2588                    Do not dev_put!
2589          */
2590         if (how) {
2591                 idev->dead = 1;
2592
2593                 /* protected by rtnl_lock */
2594                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2595
2596                 /* Step 1.5: remove snmp6 entry */
2597                 snmp6_unregister_dev(idev);
2598
2599         }
2600
2601         /* Step 2: clear hash table */
2602         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2603                 bifa = &inet6_addr_lst[i];
2604
2605                 write_lock_bh(&addrconf_hash_lock);
2606                 while ((ifa = *bifa) != NULL) {
2607                         if (ifa->idev == idev) {
2608                                 *bifa = ifa->lst_next;
2609                                 ifa->lst_next = NULL;
2610                                 addrconf_del_timer(ifa);
2611                                 in6_ifa_put(ifa);
2612                                 continue;
2613                         }
2614                         bifa = &ifa->lst_next;
2615                 }
2616                 write_unlock_bh(&addrconf_hash_lock);
2617         }
2618
2619         write_lock_bh(&idev->lock);
2620
2621         /* Step 3: clear flags for stateless addrconf */
2622         if (!how)
2623                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2624
2625         /* Step 4: clear address list */
2626 #ifdef CONFIG_IPV6_PRIVACY
2627         if (how && del_timer(&idev->regen_timer))
2628                 in6_dev_put(idev);
2629
2630         /* clear tempaddr list */
2631         while ((ifa = idev->tempaddr_list) != NULL) {
2632                 idev->tempaddr_list = ifa->tmp_next;
2633                 ifa->tmp_next = NULL;
2634                 ifa->dead = 1;
2635                 write_unlock_bh(&idev->lock);
2636                 spin_lock_bh(&ifa->lock);
2637
2638                 if (ifa->ifpub) {
2639                         in6_ifa_put(ifa->ifpub);
2640                         ifa->ifpub = NULL;
2641                 }
2642                 spin_unlock_bh(&ifa->lock);
2643                 in6_ifa_put(ifa);
2644                 write_lock_bh(&idev->lock);
2645         }
2646 #endif
2647         while ((ifa = idev->addr_list) != NULL) {
2648                 idev->addr_list = ifa->if_next;
2649                 ifa->if_next = NULL;
2650                 ifa->dead = 1;
2651                 addrconf_del_timer(ifa);
2652                 write_unlock_bh(&idev->lock);
2653
2654                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2655                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2656                 in6_ifa_put(ifa);
2657
2658                 write_lock_bh(&idev->lock);
2659         }
2660         write_unlock_bh(&idev->lock);
2661
2662         /* Step 5: Discard multicast list */
2663
2664         if (how)
2665                 ipv6_mc_destroy_dev(idev);
2666         else
2667                 ipv6_mc_down(idev);
2668
2669         idev->tstamp = jiffies;
2670
2671         /* Shot the device (if unregistered) */
2672
2673         if (how) {
2674                 addrconf_sysctl_unregister(idev);
2675                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2676                 neigh_ifdown(&nd_tbl, dev);
2677                 in6_dev_put(idev);
2678         }
2679         return 0;
2680 }
2681
2682 static void addrconf_rs_timer(unsigned long data)
2683 {
2684         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2685
2686         if (ifp->idev->cnf.forwarding)
2687                 goto out;
2688
2689         if (ifp->idev->if_flags & IF_RA_RCVD) {
2690                 /*
2691                  *      Announcement received after solicitation
2692                  *      was sent
2693                  */
2694                 goto out;
2695         }
2696
2697         spin_lock(&ifp->lock);
2698         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2699                 /* The wait after the last probe can be shorter */
2700                 addrconf_mod_timer(ifp, AC_RS,
2701                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2702                                    ifp->idev->cnf.rtr_solicit_delay :
2703                                    ifp->idev->cnf.rtr_solicit_interval);
2704                 spin_unlock(&ifp->lock);
2705
2706                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2707         } else {
2708                 spin_unlock(&ifp->lock);
2709                 /*
2710                  * Note: we do not support deprecated "all on-link"
2711                  * assumption any longer.
2712                  */
2713                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2714                        ifp->idev->dev->name);
2715         }
2716
2717 out:
2718         in6_ifa_put(ifp);
2719 }
2720
2721 /*
2722  *      Duplicate Address Detection
2723  */
2724 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2725 {
2726         unsigned long rand_num;
2727         struct inet6_dev *idev = ifp->idev;
2728
2729         if (ifp->flags & IFA_F_OPTIMISTIC)
2730                 rand_num = 0;
2731         else
2732                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2733
2734         ifp->probes = idev->cnf.dad_transmits;
2735         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2736 }
2737
2738 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2739 {
2740         struct inet6_dev *idev = ifp->idev;
2741         struct net_device *dev = idev->dev;
2742
2743         addrconf_join_solict(dev, &ifp->addr);
2744
2745         net_srandom(ifp->addr.s6_addr32[3]);
2746
2747         read_lock_bh(&idev->lock);
2748         if (ifp->dead)
2749                 goto out;
2750         spin_lock_bh(&ifp->lock);
2751
2752         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2753             idev->cnf.accept_dad < 1 ||
2754             !(ifp->flags&IFA_F_TENTATIVE) ||
2755             ifp->flags & IFA_F_NODAD) {
2756                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2757                 spin_unlock_bh(&ifp->lock);
2758                 read_unlock_bh(&idev->lock);
2759
2760                 addrconf_dad_completed(ifp);
2761                 return;
2762         }
2763
2764         if (!(idev->if_flags & IF_READY)) {
2765                 spin_unlock_bh(&ifp->lock);
2766                 read_unlock_bh(&idev->lock);
2767                 /*
2768                  * If the defice is not ready:
2769                  * - keep it tentative if it is a permanent address.
2770                  * - otherwise, kill it.
2771                  */
2772                 in6_ifa_hold(ifp);
2773                 addrconf_dad_stop(ifp);
2774                 return;
2775         }
2776
2777         /*
2778          * Optimistic nodes can start receiving
2779          * Frames right away
2780          */
2781         if(ifp->flags & IFA_F_OPTIMISTIC)
2782                 ip6_ins_rt(ifp->rt);
2783
2784         addrconf_dad_kick(ifp);
2785         spin_unlock_bh(&ifp->lock);
2786 out:
2787         read_unlock_bh(&idev->lock);
2788 }
2789
2790 static void addrconf_dad_timer(unsigned long data)
2791 {
2792         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2793         struct inet6_dev *idev = ifp->idev;
2794         struct in6_addr mcaddr;
2795
2796         read_lock_bh(&idev->lock);
2797         if (idev->dead) {
2798                 read_unlock_bh(&idev->lock);
2799                 goto out;
2800         }
2801         spin_lock_bh(&ifp->lock);
2802         if (ifp->probes == 0) {
2803                 /*
2804                  * DAD was successful
2805                  */
2806
2807                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2808                 spin_unlock_bh(&ifp->lock);
2809                 read_unlock_bh(&idev->lock);
2810
2811                 addrconf_dad_completed(ifp);
2812
2813                 goto out;
2814         }
2815
2816         ifp->probes--;
2817         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2818         spin_unlock_bh(&ifp->lock);
2819         read_unlock_bh(&idev->lock);
2820
2821         /* send a neighbour solicitation for our addr */
2822         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2823         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2824 out:
2825         in6_ifa_put(ifp);
2826 }
2827
2828 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2829 {
2830         struct net_device *     dev = ifp->idev->dev;
2831
2832         /*
2833          *      Configure the address for reception. Now it is valid.
2834          */
2835
2836         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2837
2838         /* If added prefix is link local and forwarding is off,
2839            start sending router solicitations.
2840          */
2841
2842         if (ifp->idev->cnf.forwarding == 0 &&
2843             ifp->idev->cnf.rtr_solicits > 0 &&
2844             (dev->flags&IFF_LOOPBACK) == 0 &&
2845             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2846                 /*
2847                  *      If a host as already performed a random delay
2848                  *      [...] as part of DAD [...] there is no need
2849                  *      to delay again before sending the first RS
2850                  */
2851                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2852
2853                 spin_lock_bh(&ifp->lock);
2854                 ifp->probes = 1;
2855                 ifp->idev->if_flags |= IF_RS_SENT;
2856                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2857                 spin_unlock_bh(&ifp->lock);
2858         }
2859 }
2860
2861 static void addrconf_dad_run(struct inet6_dev *idev) {
2862         struct inet6_ifaddr *ifp;
2863
2864         read_lock_bh(&idev->lock);
2865         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2866                 spin_lock_bh(&ifp->lock);
2867                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2868                         spin_unlock_bh(&ifp->lock);
2869                         continue;
2870                 }
2871                 spin_unlock_bh(&ifp->lock);
2872                 addrconf_dad_kick(ifp);
2873         }
2874         read_unlock_bh(&idev->lock);
2875 }
2876
2877 #ifdef CONFIG_PROC_FS
2878 struct if6_iter_state {
2879         struct seq_net_private p;
2880         int bucket;
2881 };
2882
2883 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2884 {
2885         struct inet6_ifaddr *ifa = NULL;
2886         struct if6_iter_state *state = seq->private;
2887         struct net *net = seq_file_net(seq);
2888
2889         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2890                 ifa = inet6_addr_lst[state->bucket];
2891
2892                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2893                         ifa = ifa->lst_next;
2894                 if (ifa)
2895                         break;
2896         }
2897         return ifa;
2898 }
2899
2900 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2901 {
2902         struct if6_iter_state *state = seq->private;
2903         struct net *net = seq_file_net(seq);
2904
2905         ifa = ifa->lst_next;
2906 try_again:
2907         if (ifa) {
2908                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2909                         ifa = ifa->lst_next;
2910                         goto try_again;
2911                 }
2912         }
2913
2914         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2915                 ifa = inet6_addr_lst[state->bucket];
2916                 goto try_again;
2917         }
2918
2919         return ifa;
2920 }
2921
2922 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2923 {
2924         struct inet6_ifaddr *ifa = if6_get_first(seq);
2925
2926         if (ifa)
2927                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2928                         --pos;
2929         return pos ? NULL : ifa;
2930 }
2931
2932 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2933         __acquires(addrconf_hash_lock)
2934 {
2935         read_lock_bh(&addrconf_hash_lock);
2936         return if6_get_idx(seq, *pos);
2937 }
2938
2939 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2940 {
2941         struct inet6_ifaddr *ifa;
2942
2943         ifa = if6_get_next(seq, v);
2944         ++*pos;
2945         return ifa;
2946 }
2947
2948 static void if6_seq_stop(struct seq_file *seq, void *v)
2949         __releases(addrconf_hash_lock)
2950 {
2951         read_unlock_bh(&addrconf_hash_lock);
2952 }
2953
2954 static int if6_seq_show(struct seq_file *seq, void *v)
2955 {
2956         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2957         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2958                    &ifp->addr,
2959                    ifp->idev->dev->ifindex,
2960                    ifp->prefix_len,
2961                    ifp->scope,
2962                    ifp->flags,
2963                    ifp->idev->dev->name);
2964         return 0;
2965 }
2966
2967 static const struct seq_operations if6_seq_ops = {
2968         .start  = if6_seq_start,
2969         .next   = if6_seq_next,
2970         .show   = if6_seq_show,
2971         .stop   = if6_seq_stop,
2972 };
2973
2974 static int if6_seq_open(struct inode *inode, struct file *file)
2975 {
2976         return seq_open_net(inode, file, &if6_seq_ops,
2977                             sizeof(struct if6_iter_state));
2978 }
2979
2980 static const struct file_operations if6_fops = {
2981         .owner          = THIS_MODULE,
2982         .open           = if6_seq_open,
2983         .read           = seq_read,
2984         .llseek         = seq_lseek,
2985         .release        = seq_release_net,
2986 };
2987
2988 static int if6_proc_net_init(struct net *net)
2989 {
2990         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2991                 return -ENOMEM;
2992         return 0;
2993 }
2994
2995 static void if6_proc_net_exit(struct net *net)
2996 {
2997        proc_net_remove(net, "if_inet6");
2998 }
2999
3000 static struct pernet_operations if6_proc_net_ops = {
3001        .init = if6_proc_net_init,
3002        .exit = if6_proc_net_exit,
3003 };
3004
3005 int __init if6_proc_init(void)
3006 {
3007         return register_pernet_subsys(&if6_proc_net_ops);
3008 }
3009
3010 void if6_proc_exit(void)
3011 {
3012         unregister_pernet_subsys(&if6_proc_net_ops);
3013 }
3014 #endif  /* CONFIG_PROC_FS */
3015
3016 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3017 /* Check if address is a home address configured on any interface. */
3018 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3019 {
3020         int ret = 0;
3021         struct inet6_ifaddr * ifp;
3022         u8 hash = ipv6_addr_hash(addr);
3023         read_lock_bh(&addrconf_hash_lock);
3024         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3025                 if (!net_eq(dev_net(ifp->idev->dev), net))
3026                         continue;
3027                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3028                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3029                         ret = 1;
3030                         break;
3031                 }
3032         }
3033         read_unlock_bh(&addrconf_hash_lock);
3034         return ret;
3035 }
3036 #endif
3037
3038 /*
3039  *      Periodic address status verification
3040  */
3041
3042 static void addrconf_verify(unsigned long foo)
3043 {
3044         struct inet6_ifaddr *ifp;
3045         unsigned long now, next;
3046         int i;
3047
3048         spin_lock_bh(&addrconf_verify_lock);
3049         now = jiffies;
3050         next = now + ADDR_CHECK_FREQUENCY;
3051
3052         del_timer(&addr_chk_timer);
3053
3054         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3055
3056 restart:
3057                 read_lock(&addrconf_hash_lock);
3058                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3059                         unsigned long age;
3060 #ifdef CONFIG_IPV6_PRIVACY
3061                         unsigned long regen_advance;
3062 #endif
3063
3064                         if (ifp->flags & IFA_F_PERMANENT)
3065                                 continue;
3066
3067                         spin_lock(&ifp->lock);
3068                         age = (now - ifp->tstamp) / HZ;
3069
3070 #ifdef CONFIG_IPV6_PRIVACY
3071                         regen_advance = ifp->idev->cnf.regen_max_retry *
3072                                         ifp->idev->cnf.dad_transmits *
3073                                         ifp->idev->nd_parms->retrans_time / HZ;
3074 #endif
3075
3076                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3077                             age >= ifp->valid_lft) {
3078                                 spin_unlock(&ifp->lock);
3079                                 in6_ifa_hold(ifp);
3080                                 read_unlock(&addrconf_hash_lock);
3081                                 ipv6_del_addr(ifp);
3082                                 goto restart;
3083                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3084                                 spin_unlock(&ifp->lock);
3085                                 continue;
3086                         } else if (age >= ifp->prefered_lft) {
3087                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3088                                 int deprecate = 0;
3089
3090                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3091                                         deprecate = 1;
3092                                         ifp->flags |= IFA_F_DEPRECATED;
3093                                 }
3094
3095                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3096                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3097
3098                                 spin_unlock(&ifp->lock);
3099
3100                                 if (deprecate) {
3101                                         in6_ifa_hold(ifp);
3102                                         read_unlock(&addrconf_hash_lock);
3103
3104                                         ipv6_ifa_notify(0, ifp);
3105                                         in6_ifa_put(ifp);
3106                                         goto restart;
3107                                 }
3108 #ifdef CONFIG_IPV6_PRIVACY
3109                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3110                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3111                                 if (age >= ifp->prefered_lft - regen_advance) {
3112                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3113                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3114                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3115                                         if (!ifp->regen_count && ifpub) {
3116                                                 ifp->regen_count++;
3117                                                 in6_ifa_hold(ifp);
3118                                                 in6_ifa_hold(ifpub);
3119                                                 spin_unlock(&ifp->lock);
3120                                                 read_unlock(&addrconf_hash_lock);
3121                                                 spin_lock(&ifpub->lock);
3122                                                 ifpub->regen_count = 0;
3123                                                 spin_unlock(&ifpub->lock);
3124                                                 ipv6_create_tempaddr(ifpub, ifp);
3125                                                 in6_ifa_put(ifpub);
3126                                                 in6_ifa_put(ifp);
3127                                                 goto restart;
3128                                         }
3129                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3130                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3131                                 spin_unlock(&ifp->lock);
3132 #endif
3133                         } else {
3134                                 /* ifp->prefered_lft <= ifp->valid_lft */
3135                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3136                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3137                                 spin_unlock(&ifp->lock);
3138                         }
3139                 }
3140                 read_unlock(&addrconf_hash_lock);
3141         }
3142
3143         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3144         add_timer(&addr_chk_timer);
3145         spin_unlock_bh(&addrconf_verify_lock);
3146 }
3147
3148 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3149 {
3150         struct in6_addr *pfx = NULL;
3151
3152         if (addr)
3153                 pfx = nla_data(addr);
3154
3155         if (local) {
3156                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3157                         pfx = NULL;
3158                 else
3159                         pfx = nla_data(local);
3160         }
3161
3162         return pfx;
3163 }
3164
3165 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3166         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3167         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3168         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3169 };
3170
3171 static int
3172 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3173 {
3174         struct net *net = sock_net(skb->sk);
3175         struct ifaddrmsg *ifm;
3176         struct nlattr *tb[IFA_MAX+1];
3177         struct in6_addr *pfx;
3178         int err;
3179
3180         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3181         if (err < 0)
3182                 return err;
3183
3184         ifm = nlmsg_data(nlh);
3185         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3186         if (pfx == NULL)
3187                 return -EINVAL;
3188
3189         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3190 }
3191
3192 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3193                              u32 prefered_lft, u32 valid_lft)
3194 {
3195         u32 flags;
3196         clock_t expires;
3197         unsigned long timeout;
3198
3199         if (!valid_lft || (prefered_lft > valid_lft))
3200                 return -EINVAL;
3201
3202         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3203         if (addrconf_finite_timeout(timeout)) {
3204                 expires = jiffies_to_clock_t(timeout * HZ);
3205                 valid_lft = timeout;
3206                 flags = RTF_EXPIRES;
3207         } else {
3208                 expires = 0;
3209                 flags = 0;
3210                 ifa_flags |= IFA_F_PERMANENT;
3211         }
3212
3213         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3214         if (addrconf_finite_timeout(timeout)) {
3215                 if (timeout == 0)
3216                         ifa_flags |= IFA_F_DEPRECATED;
3217                 prefered_lft = timeout;
3218         }
3219
3220         spin_lock_bh(&ifp->lock);
3221         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3222         ifp->tstamp = jiffies;
3223         ifp->valid_lft = valid_lft;
3224         ifp->prefered_lft = prefered_lft;
3225
3226         spin_unlock_bh(&ifp->lock);
3227         if (!(ifp->flags&IFA_F_TENTATIVE))
3228                 ipv6_ifa_notify(0, ifp);
3229
3230         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3231                               expires, flags);
3232         addrconf_verify(0);
3233
3234         return 0;
3235 }
3236
3237 static int
3238 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3239 {
3240         struct net *net = sock_net(skb->sk);
3241         struct ifaddrmsg *ifm;
3242         struct nlattr *tb[IFA_MAX+1];
3243         struct in6_addr *pfx;
3244         struct inet6_ifaddr *ifa;
3245         struct net_device *dev;
3246         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3247         u8 ifa_flags;
3248         int err;
3249
3250         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3251         if (err < 0)
3252                 return err;
3253
3254         ifm = nlmsg_data(nlh);
3255         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3256         if (pfx == NULL)
3257                 return -EINVAL;
3258
3259         if (tb[IFA_CACHEINFO]) {
3260                 struct ifa_cacheinfo *ci;
3261
3262                 ci = nla_data(tb[IFA_CACHEINFO]);
3263                 valid_lft = ci->ifa_valid;
3264                 preferred_lft = ci->ifa_prefered;
3265         } else {
3266                 preferred_lft = INFINITY_LIFE_TIME;
3267                 valid_lft = INFINITY_LIFE_TIME;
3268         }
3269
3270         dev =  __dev_get_by_index(net, ifm->ifa_index);
3271         if (dev == NULL)
3272                 return -ENODEV;
3273
3274         /* We ignore other flags so far. */
3275         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3276
3277         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3278         if (ifa == NULL) {
3279                 /*
3280                  * It would be best to check for !NLM_F_CREATE here but
3281                  * userspace alreay relies on not having to provide this.
3282                  */
3283                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3284                                       ifm->ifa_prefixlen, ifa_flags,
3285                                       preferred_lft, valid_lft);
3286         }
3287
3288         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3289             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3290                 err = -EEXIST;
3291         else
3292                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3293
3294         in6_ifa_put(ifa);
3295
3296         return err;
3297 }
3298
3299 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3300                           u8 scope, int ifindex)
3301 {
3302         struct ifaddrmsg *ifm;
3303
3304         ifm = nlmsg_data(nlh);
3305         ifm->ifa_family = AF_INET6;
3306         ifm->ifa_prefixlen = prefixlen;
3307         ifm->ifa_flags = flags;
3308         ifm->ifa_scope = scope;
3309         ifm->ifa_index = ifindex;
3310 }
3311
3312 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3313                          unsigned long tstamp, u32 preferred, u32 valid)
3314 {
3315         struct ifa_cacheinfo ci;
3316
3317         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3318                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3319         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3320                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3321         ci.ifa_prefered = preferred;
3322         ci.ifa_valid = valid;
3323
3324         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3325 }
3326
3327 static inline int rt_scope(int ifa_scope)
3328 {
3329         if (ifa_scope & IFA_HOST)
3330                 return RT_SCOPE_HOST;
3331         else if (ifa_scope & IFA_LINK)
3332                 return RT_SCOPE_LINK;
3333         else if (ifa_scope & IFA_SITE)
3334                 return RT_SCOPE_SITE;
3335         else
3336                 return RT_SCOPE_UNIVERSE;
3337 }
3338
3339 static inline int inet6_ifaddr_msgsize(void)
3340 {
3341         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3342                + nla_total_size(16) /* IFA_ADDRESS */
3343                + nla_total_size(sizeof(struct ifa_cacheinfo));
3344 }
3345
3346 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3347                              u32 pid, u32 seq, int event, unsigned int flags)
3348 {
3349         struct nlmsghdr  *nlh;
3350         u32 preferred, valid;
3351
3352         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3353         if (nlh == NULL)
3354                 return -EMSGSIZE;
3355
3356         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3357                       ifa->idev->dev->ifindex);
3358
3359         if (!(ifa->flags&IFA_F_PERMANENT)) {
3360                 preferred = ifa->prefered_lft;
3361                 valid = ifa->valid_lft;
3362                 if (preferred != INFINITY_LIFE_TIME) {
3363                         long tval = (jiffies - ifa->tstamp)/HZ;
3364                         preferred -= tval;
3365                         if (valid != INFINITY_LIFE_TIME)
3366                                 valid -= tval;
3367                 }
3368         } else {
3369                 preferred = INFINITY_LIFE_TIME;
3370                 valid = INFINITY_LIFE_TIME;
3371         }
3372
3373         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3374             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3375                 nlmsg_cancel(skb, nlh);
3376                 return -EMSGSIZE;
3377         }
3378
3379         return nlmsg_end(skb, nlh);
3380 }
3381
3382 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3383                                 u32 pid, u32 seq, int event, u16 flags)
3384 {
3385         struct nlmsghdr  *nlh;
3386         u8 scope = RT_SCOPE_UNIVERSE;
3387         int ifindex = ifmca->idev->dev->ifindex;
3388
3389         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3390                 scope = RT_SCOPE_SITE;
3391
3392         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3393         if (nlh == NULL)
3394                 return -EMSGSIZE;
3395
3396         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3397         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3398             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3399                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3400                 nlmsg_cancel(skb, nlh);
3401                 return -EMSGSIZE;
3402         }
3403
3404         return nlmsg_end(skb, nlh);
3405 }
3406
3407 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3408                                 u32 pid, u32 seq, int event, unsigned int flags)
3409 {
3410         struct nlmsghdr  *nlh;
3411         u8 scope = RT_SCOPE_UNIVERSE;
3412         int ifindex = ifaca->aca_idev->dev->ifindex;
3413
3414         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3415                 scope = RT_SCOPE_SITE;
3416
3417         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3418         if (nlh == NULL)
3419                 return -EMSGSIZE;
3420
3421         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3422         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3423             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3424                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3425                 nlmsg_cancel(skb, nlh);
3426                 return -EMSGSIZE;
3427         }
3428
3429         return nlmsg_end(skb, nlh);
3430 }
3431
3432 enum addr_type_t
3433 {
3434         UNICAST_ADDR,
3435         MULTICAST_ADDR,
3436         ANYCAST_ADDR,
3437 };
3438
3439 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3440                            enum addr_type_t type)
3441 {
3442         int idx, ip_idx;
3443         int s_idx, s_ip_idx;
3444         int err = 1;
3445         struct net_device *dev;
3446         struct inet6_dev *idev = NULL;
3447         struct inet6_ifaddr *ifa;
3448         struct ifmcaddr6 *ifmca;
3449         struct ifacaddr6 *ifaca;
3450         struct net *net = sock_net(skb->sk);
3451
3452         s_idx = cb->args[0];
3453         s_ip_idx = ip_idx = cb->args[1];
3454
3455         idx = 0;
3456         for_each_netdev(net, dev) {
3457                 if (idx < s_idx)
3458                         goto cont;
3459                 if (idx > s_idx)
3460                         s_ip_idx = 0;
3461                 ip_idx = 0;
3462                 if ((idev = in6_dev_get(dev)) == NULL)
3463                         goto cont;
3464                 read_lock_bh(&idev->lock);
3465                 switch (type) {
3466                 case UNICAST_ADDR:
3467                         /* unicast address incl. temp addr */
3468                         for (ifa = idev->addr_list; ifa;
3469                              ifa = ifa->if_next, ip_idx++) {
3470                                 if (ip_idx < s_ip_idx)
3471                                         continue;
3472                                 err = inet6_fill_ifaddr(skb, ifa,
3473                                                         NETLINK_CB(cb->skb).pid,
3474                                                         cb->nlh->nlmsg_seq,
3475                                                         RTM_NEWADDR,
3476                                                         NLM_F_MULTI);
3477                         }
3478                         break;
3479                 case MULTICAST_ADDR:
3480                         /* multicast address */
3481                         for (ifmca = idev->mc_list; ifmca;
3482                              ifmca = ifmca->next, ip_idx++) {
3483                                 if (ip_idx < s_ip_idx)
3484                                         continue;
3485                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3486                                                           NETLINK_CB(cb->skb).pid,
3487                                                           cb->nlh->nlmsg_seq,
3488                                                           RTM_GETMULTICAST,
3489                                                           NLM_F_MULTI);
3490                         }
3491                         break;
3492                 case ANYCAST_ADDR:
3493                         /* anycast address */
3494                         for (ifaca = idev->ac_list; ifaca;
3495                              ifaca = ifaca->aca_next, ip_idx++) {
3496                                 if (ip_idx < s_ip_idx)
3497                                         continue;
3498                                 err = inet6_fill_ifacaddr(skb, ifaca,
3499                                                           NETLINK_CB(cb->skb).pid,
3500                                                           cb->nlh->nlmsg_seq,
3501                                                           RTM_GETANYCAST,
3502                                                           NLM_F_MULTI);
3503                         }
3504                         break;
3505                 default:
3506                         break;
3507                 }
3508                 read_unlock_bh(&idev->lock);
3509                 in6_dev_put(idev);
3510
3511                 if (err <= 0)
3512                         break;
3513 cont:
3514                 idx++;
3515         }
3516         cb->args[0] = idx;
3517         cb->args[1] = ip_idx;
3518         return skb->len;
3519 }
3520
3521 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3522 {
3523         enum addr_type_t type = UNICAST_ADDR;
3524
3525         return inet6_dump_addr(skb, cb, type);
3526 }
3527
3528 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3529 {
3530         enum addr_type_t type = MULTICAST_ADDR;
3531
3532         return inet6_dump_addr(skb, cb, type);
3533 }
3534
3535
3536 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3537 {
3538         enum addr_type_t type = ANYCAST_ADDR;
3539
3540         return inet6_dump_addr(skb, cb, type);
3541 }
3542
3543 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3544                              void *arg)
3545 {
3546         struct net *net = sock_net(in_skb->sk);
3547         struct ifaddrmsg *ifm;
3548         struct nlattr *tb[IFA_MAX+1];
3549         struct in6_addr *addr = NULL;
3550         struct net_device *dev = NULL;
3551         struct inet6_ifaddr *ifa;
3552         struct sk_buff *skb;
3553         int err;
3554
3555         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3556         if (err < 0)
3557                 goto errout;
3558
3559         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3560         if (addr == NULL) {
3561                 err = -EINVAL;
3562                 goto errout;
3563         }
3564
3565         ifm = nlmsg_data(nlh);
3566         if (ifm->ifa_index)
3567                 dev = __dev_get_by_index(net, ifm->ifa_index);
3568
3569         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3570                 err = -EADDRNOTAVAIL;
3571                 goto errout;
3572         }
3573
3574         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3575                 err = -ENOBUFS;
3576                 goto errout_ifa;
3577         }
3578
3579         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3580                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3581         if (err < 0) {
3582                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3583                 WARN_ON(err == -EMSGSIZE);
3584                 kfree_skb(skb);
3585                 goto errout_ifa;
3586         }
3587         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3588 errout_ifa:
3589         in6_ifa_put(ifa);
3590 errout:
3591         return err;
3592 }
3593
3594 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3595 {
3596         struct sk_buff *skb;
3597         struct net *net = dev_net(ifa->idev->dev);
3598         int err = -ENOBUFS;
3599
3600         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3601         if (skb == NULL)
3602                 goto errout;
3603
3604         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3605         if (err < 0) {
3606                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3607                 WARN_ON(err == -EMSGSIZE);
3608                 kfree_skb(skb);
3609                 goto errout;
3610         }
3611         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3612         return;
3613 errout:
3614         if (err < 0)
3615                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3616 }
3617
3618 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3619                                 __s32 *array, int bytes)
3620 {
3621         BUG_ON(bytes < (DEVCONF_MAX * 4));
3622
3623         memset(array, 0, bytes);
3624         array[DEVCONF_FORWARDING] = cnf->forwarding;
3625         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3626         array[DEVCONF_MTU6] = cnf->mtu6;
3627         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3628         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3629         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3630         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3631         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3632         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3633         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3634         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3635 #ifdef CONFIG_IPV6_PRIVACY
3636         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3637         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3638         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3639         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3640         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3641 #endif
3642         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3643         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3644         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3645 #ifdef CONFIG_IPV6_ROUTER_PREF
3646         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3647         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3648 #ifdef CONFIG_IPV6_ROUTE_INFO
3649         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3650 #endif
3651 #endif
3652         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3653         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3654 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3655         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3656 #endif
3657 #ifdef CONFIG_IPV6_MROUTE
3658         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3659 #endif
3660         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3661         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3662 }
3663
3664 static inline size_t inet6_if_nlmsg_size(void)
3665 {
3666         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3667                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3668                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3669                + nla_total_size(4) /* IFLA_MTU */
3670                + nla_total_size(4) /* IFLA_LINK */
3671                + nla_total_size( /* IFLA_PROTINFO */
3672                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3673                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3674                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3675                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3676                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3677                  );
3678 }
3679
3680 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3681                                       int bytes)
3682 {
3683         int i;
3684         int pad = bytes - sizeof(u64) * items;
3685         BUG_ON(pad < 0);
3686
3687         /* Use put_unaligned() because stats may not be aligned for u64. */
3688         put_unaligned(items, &stats[0]);
3689         for (i = 1; i < items; i++)
3690                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3691
3692         memset(&stats[items], 0, pad);
3693 }
3694
3695 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3696                              int bytes)
3697 {
3698         switch(attrtype) {
3699         case IFLA_INET6_STATS:
3700                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3701                 break;
3702         case IFLA_INET6_ICMP6STATS:
3703                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3704                 break;
3705         }
3706 }
3707
3708 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3709                              u32 pid, u32 seq, int event, unsigned int flags)
3710 {
3711         struct net_device *dev = idev->dev;
3712         struct nlattr *nla;
3713         struct ifinfomsg *hdr;
3714         struct nlmsghdr *nlh;
3715         void *protoinfo;
3716         struct ifla_cacheinfo ci;
3717
3718         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3719         if (nlh == NULL)
3720                 return -EMSGSIZE;
3721
3722         hdr = nlmsg_data(nlh);
3723         hdr->ifi_family = AF_INET6;
3724         hdr->__ifi_pad = 0;
3725         hdr->ifi_type = dev->type;
3726         hdr->ifi_index = dev->ifindex;
3727         hdr->ifi_flags = dev_get_flags(dev);
3728         hdr->ifi_change = 0;
3729
3730         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3731
3732         if (dev->addr_len)
3733                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3734
3735         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3736         if (dev->ifindex != dev->iflink)
3737                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3738
3739         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3740         if (protoinfo == NULL)
3741                 goto nla_put_failure;
3742
3743         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3744
3745         ci.max_reasm_len = IPV6_MAXPLEN;
3746         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3747                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3748         ci.reachable_time = idev->nd_parms->reachable_time;
3749         ci.retrans_time = idev->nd_parms->retrans_time;
3750         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3751
3752         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3753         if (nla == NULL)
3754                 goto nla_put_failure;
3755         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3756
3757         /* XXX - MC not implemented */
3758
3759         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3760         if (nla == NULL)
3761                 goto nla_put_failure;
3762         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3763
3764         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3765         if (nla == NULL)
3766                 goto nla_put_failure;
3767         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3768
3769         nla_nest_end(skb, protoinfo);
3770         return nlmsg_end(skb, nlh);
3771
3772 nla_put_failure:
3773         nlmsg_cancel(skb, nlh);
3774         return -EMSGSIZE;
3775 }
3776
3777 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3778 {
3779         struct net *net = sock_net(skb->sk);
3780         int idx, err;
3781         int s_idx = cb->args[0];
3782         struct net_device *dev;
3783         struct inet6_dev *idev;
3784
3785         read_lock(&dev_base_lock);
3786         idx = 0;
3787         for_each_netdev(net, dev) {
3788                 if (idx < s_idx)
3789                         goto cont;
3790                 if ((idev = in6_dev_get(dev)) == NULL)
3791                         goto cont;
3792                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3793                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3794                 in6_dev_put(idev);
3795                 if (err <= 0)
3796                         break;
3797 cont:
3798                 idx++;
3799         }
3800         read_unlock(&dev_base_lock);
3801         cb->args[0] = idx;
3802
3803         return skb->len;
3804 }
3805
3806 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3807 {
3808         struct sk_buff *skb;
3809         struct net *net = dev_net(idev->dev);
3810         int err = -ENOBUFS;
3811
3812         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3813         if (skb == NULL)
3814                 goto errout;
3815
3816         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3817         if (err < 0) {
3818                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3819                 WARN_ON(err == -EMSGSIZE);
3820                 kfree_skb(skb);
3821                 goto errout;
3822         }
3823         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3824         return;
3825 errout:
3826         if (err < 0)
3827                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3828 }
3829
3830 static inline size_t inet6_prefix_nlmsg_size(void)
3831 {
3832         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3833                + nla_total_size(sizeof(struct in6_addr))
3834                + nla_total_size(sizeof(struct prefix_cacheinfo));
3835 }
3836
3837 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3838                              struct prefix_info *pinfo, u32 pid, u32 seq,
3839                              int event, unsigned int flags)
3840 {
3841         struct prefixmsg *pmsg;
3842         struct nlmsghdr *nlh;
3843         struct prefix_cacheinfo ci;
3844
3845         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3846         if (nlh == NULL)
3847                 return -EMSGSIZE;
3848
3849         pmsg = nlmsg_data(nlh);
3850         pmsg->prefix_family = AF_INET6;
3851         pmsg->prefix_pad1 = 0;
3852         pmsg->prefix_pad2 = 0;
3853         pmsg->prefix_ifindex = idev->dev->ifindex;
3854         pmsg->prefix_len = pinfo->prefix_len;
3855         pmsg->prefix_type = pinfo->type;
3856         pmsg->prefix_pad3 = 0;
3857         pmsg->prefix_flags = 0;
3858         if (pinfo->onlink)
3859                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3860         if (pinfo->autoconf)
3861                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3862
3863         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3864
3865         ci.preferred_time = ntohl(pinfo->prefered);
3866         ci.valid_time = ntohl(pinfo->valid);
3867         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3868
3869         return nlmsg_end(skb, nlh);
3870
3871 nla_put_failure:
3872         nlmsg_cancel(skb, nlh);
3873         return -EMSGSIZE;
3874 }
3875
3876 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3877                          struct prefix_info *pinfo)
3878 {
3879         struct sk_buff *skb;
3880         struct net *net = dev_net(idev->dev);
3881         int err = -ENOBUFS;
3882
3883         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3884         if (skb == NULL)
3885                 goto errout;
3886
3887         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3888         if (err < 0) {
3889                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3890                 WARN_ON(err == -EMSGSIZE);
3891                 kfree_skb(skb);
3892                 goto errout;
3893         }
3894         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3895         return;
3896 errout:
3897         if (err < 0)
3898                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3899 }
3900
3901 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3902 {
3903         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3904
3905         switch (event) {
3906         case RTM_NEWADDR:
3907                 /*
3908                  * If the address was optimistic
3909                  * we inserted the route at the start of
3910                  * our DAD process, so we don't need
3911                  * to do it again
3912                  */
3913                 if (!(ifp->rt->rt6i_node))
3914                         ip6_ins_rt(ifp->rt);
3915                 if (ifp->idev->cnf.forwarding)
3916                         addrconf_join_anycast(ifp);
3917                 break;
3918         case RTM_DELADDR:
3919                 if (ifp->idev->cnf.forwarding)
3920                         addrconf_leave_anycast(ifp);
3921                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3922                 dst_hold(&ifp->rt->u.dst);
3923                 if (ip6_del_rt(ifp->rt))
3924                         dst_free(&ifp->rt->u.dst);
3925                 break;
3926         }
3927 }
3928
3929 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3930 {
3931         rcu_read_lock_bh();
3932         if (likely(ifp->idev->dead == 0))
3933                 __ipv6_ifa_notify(event, ifp);
3934         rcu_read_unlock_bh();
3935 }
3936
3937 #ifdef CONFIG_SYSCTL
3938
3939 static
3940 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3941                            void __user *buffer, size_t *lenp, loff_t *ppos)
3942 {
3943         int *valp = ctl->data;
3944         int val = *valp;
3945         int ret;
3946
3947         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3948
3949         if (write)
3950                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3951         return ret;
3952 }
3953
3954 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3955                                             void __user *oldval,
3956                                             size_t __user *oldlenp,
3957                                             void __user *newval, size_t newlen)
3958 {
3959         int *valp = table->data;
3960         int val = *valp;
3961         int new;
3962
3963         if (!newval || !newlen)
3964                 return 0;
3965         if (newlen != sizeof(int))
3966                 return -EINVAL;
3967         if (get_user(new, (int __user *)newval))
3968                 return -EFAULT;
3969         if (new == *valp)
3970                 return 0;
3971         if (oldval && oldlenp) {
3972                 size_t len;
3973                 if (get_user(len, oldlenp))
3974                         return -EFAULT;
3975                 if (len) {
3976                         if (len > table->maxlen)
3977                                 len = table->maxlen;
3978                         if (copy_to_user(oldval, valp, len))
3979                                 return -EFAULT;
3980                         if (put_user(len, oldlenp))
3981                                 return -EFAULT;
3982                 }
3983         }
3984
3985         *valp = new;
3986         return addrconf_fixup_forwarding(table, valp, val);
3987 }
3988
3989 static struct addrconf_sysctl_table
3990 {
3991         struct ctl_table_header *sysctl_header;
3992         ctl_table addrconf_vars[DEVCONF_MAX+1];
3993         char *dev_name;
3994 } addrconf_sysctl __read_mostly = {
3995         .sysctl_header = NULL,
3996         .addrconf_vars = {
3997                 {
3998                         .ctl_name       =       NET_IPV6_FORWARDING,
3999                         .procname       =       "forwarding",
4000                         .data           =       &ipv6_devconf.forwarding,
4001                         .maxlen         =       sizeof(int),
4002                         .mode           =       0644,
4003                         .proc_handler   =       addrconf_sysctl_forward,
4004                         .strategy       =       addrconf_sysctl_forward_strategy,
4005                 },
4006                 {
4007                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4008                         .procname       =       "hop_limit",
4009                         .data           =       &ipv6_devconf.hop_limit,
4010                         .maxlen         =       sizeof(int),
4011                         .mode           =       0644,
4012                         .proc_handler   =       proc_dointvec,
4013                 },
4014                 {
4015                         .ctl_name       =       NET_IPV6_MTU,
4016                         .procname       =       "mtu",
4017                         .data           =       &ipv6_devconf.mtu6,
4018                         .maxlen         =       sizeof(int),
4019                         .mode           =       0644,
4020                         .proc_handler   =       proc_dointvec,
4021                 },
4022                 {
4023                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4024                         .procname       =       "accept_ra",
4025                         .data           =       &ipv6_devconf.accept_ra,
4026                         .maxlen         =       sizeof(int),
4027                         .mode           =       0644,
4028                         .proc_handler   =       proc_dointvec,
4029                 },
4030                 {
4031                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4032                         .procname       =       "accept_redirects",
4033                         .data           =       &ipv6_devconf.accept_redirects,
4034                         .maxlen         =       sizeof(int),
4035                         .mode           =       0644,
4036                         .proc_handler   =       proc_dointvec,
4037                 },
4038                 {
4039                         .ctl_name       =       NET_IPV6_AUTOCONF,
4040                         .procname       =       "autoconf",
4041                         .data           =       &ipv6_devconf.autoconf,
4042                         .maxlen         =       sizeof(int),
4043                         .mode           =       0644,
4044                         .proc_handler   =       proc_dointvec,
4045                 },
4046                 {
4047                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4048                         .procname       =       "dad_transmits",
4049                         .data           =       &ipv6_devconf.dad_transmits,
4050                         .maxlen         =       sizeof(int),
4051                         .mode           =       0644,
4052                         .proc_handler   =       proc_dointvec,
4053                 },
4054                 {
4055                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4056                         .procname       =       "router_solicitations",
4057                         .data           =       &ipv6_devconf.rtr_solicits,
4058                         .maxlen         =       sizeof(int),
4059                         .mode           =       0644,
4060                         .proc_handler   =       proc_dointvec,
4061                 },
4062                 {
4063                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4064                         .procname       =       "router_solicitation_interval",
4065                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4066                         .maxlen         =       sizeof(int),
4067                         .mode           =       0644,
4068                         .proc_handler   =       proc_dointvec_jiffies,
4069                         .strategy       =       sysctl_jiffies,
4070                 },
4071                 {
4072                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4073                         .procname       =       "router_solicitation_delay",
4074                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4075                         .maxlen         =       sizeof(int),
4076                         .mode           =       0644,
4077                         .proc_handler   =       proc_dointvec_jiffies,
4078                         .strategy       =       sysctl_jiffies,
4079                 },
4080                 {
4081                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4082                         .procname       =       "force_mld_version",
4083                         .data           =       &ipv6_devconf.force_mld_version,
4084                         .maxlen         =       sizeof(int),
4085                         .mode           =       0644,
4086                         .proc_handler   =       proc_dointvec,
4087                 },
4088 #ifdef CONFIG_IPV6_PRIVACY
4089                 {
4090                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4091                         .procname       =       "use_tempaddr",
4092                         .data           =       &ipv6_devconf.use_tempaddr,
4093                         .maxlen         =       sizeof(int),
4094                         .mode           =       0644,
4095                         .proc_handler   =       proc_dointvec,
4096                 },
4097                 {
4098                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4099                         .procname       =       "temp_valid_lft",
4100                         .data           =       &ipv6_devconf.temp_valid_lft,
4101                         .maxlen         =       sizeof(int),
4102                         .mode           =       0644,
4103                         .proc_handler   =       proc_dointvec,
4104                 },
4105                 {
4106                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4107                         .procname       =       "temp_prefered_lft",
4108                         .data           =       &ipv6_devconf.temp_prefered_lft,
4109                         .maxlen         =       sizeof(int),
4110                         .mode           =       0644,
4111                         .proc_handler   =       proc_dointvec,
4112                 },
4113                 {
4114                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4115                         .procname       =       "regen_max_retry",
4116                         .data           =       &ipv6_devconf.regen_max_retry,
4117                         .maxlen         =       sizeof(int),
4118                         .mode           =       0644,
4119                         .proc_handler   =       proc_dointvec,
4120                 },
4121                 {
4122                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4123                         .procname       =       "max_desync_factor",
4124                         .data           =       &ipv6_devconf.max_desync_factor,
4125                         .maxlen         =       sizeof(int),
4126                         .mode           =       0644,
4127                         .proc_handler   =       proc_dointvec,
4128                 },
4129 #endif
4130                 {
4131                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4132                         .procname       =       "max_addresses",
4133                         .data           =       &ipv6_devconf.max_addresses,
4134                         .maxlen         =       sizeof(int),
4135                         .mode           =       0644,
4136                         .proc_handler   =       proc_dointvec,
4137                 },
4138                 {
4139                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4140                         .procname       =       "accept_ra_defrtr",
4141                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4142                         .maxlen         =       sizeof(int),
4143                         .mode           =       0644,
4144                         .proc_handler   =       proc_dointvec,
4145                 },
4146                 {
4147                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4148                         .procname       =       "accept_ra_pinfo",
4149                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4150                         .maxlen         =       sizeof(int),
4151                         .mode           =       0644,
4152                         .proc_handler   =       proc_dointvec,
4153                 },
4154 #ifdef CONFIG_IPV6_ROUTER_PREF
4155                 {
4156                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4157                         .procname       =       "accept_ra_rtr_pref",
4158                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4159                         .maxlen         =       sizeof(int),
4160                         .mode           =       0644,
4161                         .proc_handler   =       proc_dointvec,
4162                 },
4163                 {
4164                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4165                         .procname       =       "router_probe_interval",
4166                         .data           =       &ipv6_devconf.rtr_probe_interval,
4167                         .maxlen         =       sizeof(int),
4168                         .mode           =       0644,
4169                         .proc_handler   =       proc_dointvec_jiffies,
4170                         .strategy       =       sysctl_jiffies,
4171                 },
4172 #ifdef CONFIG_IPV6_ROUTE_INFO
4173                 {
4174                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4175                         .procname       =       "accept_ra_rt_info_max_plen",
4176                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4177                         .maxlen         =       sizeof(int),
4178                         .mode           =       0644,
4179                         .proc_handler   =       proc_dointvec,
4180                 },
4181 #endif
4182 #endif
4183                 {
4184                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4185                         .procname       =       "proxy_ndp",
4186                         .data           =       &ipv6_devconf.proxy_ndp,
4187                         .maxlen         =       sizeof(int),
4188                         .mode           =       0644,
4189                         .proc_handler   =       proc_dointvec,
4190                 },
4191                 {
4192                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4193                         .procname       =       "accept_source_route",
4194                         .data           =       &ipv6_devconf.accept_source_route,
4195                         .maxlen         =       sizeof(int),
4196                         .mode           =       0644,
4197                         .proc_handler   =       proc_dointvec,
4198                 },
4199 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4200                 {
4201                         .ctl_name       =       CTL_UNNUMBERED,
4202                         .procname       =       "optimistic_dad",
4203                         .data           =       &ipv6_devconf.optimistic_dad,
4204                         .maxlen         =       sizeof(int),
4205                         .mode           =       0644,
4206                         .proc_handler   =       proc_dointvec,
4207
4208                 },
4209 #endif
4210 #ifdef CONFIG_IPV6_MROUTE
4211                 {
4212                         .ctl_name       =       CTL_UNNUMBERED,
4213                         .procname       =       "mc_forwarding",
4214                         .data           =       &ipv6_devconf.mc_forwarding,
4215                         .maxlen         =       sizeof(int),
4216                         .mode           =       0444,
4217                         .proc_handler   =       proc_dointvec,
4218                 },
4219 #endif
4220                 {
4221                         .ctl_name       =       CTL_UNNUMBERED,
4222                         .procname       =       "disable_ipv6",
4223                         .data           =       &ipv6_devconf.disable_ipv6,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       proc_dointvec,
4227                 },
4228                 {
4229                         .ctl_name       =       CTL_UNNUMBERED,
4230                         .procname       =       "accept_dad",
4231                         .data           =       &ipv6_devconf.accept_dad,
4232                         .maxlen         =       sizeof(int),
4233                         .mode           =       0644,
4234                         .proc_handler   =       proc_dointvec,
4235                 },
4236                 {
4237                         .ctl_name       =       0,      /* sentinel */
4238                 }
4239         },
4240 };
4241
4242 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4243                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4244 {
4245         int i;
4246         struct addrconf_sysctl_table *t;
4247
4248 #define ADDRCONF_CTL_PATH_DEV   3
4249
4250         struct ctl_path addrconf_ctl_path[] = {
4251                 { .procname = "net", .ctl_name = CTL_NET, },
4252                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4253                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4254                 { /* to be set */ },
4255                 { },
4256         };
4257
4258
4259         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4260         if (t == NULL)
4261                 goto out;
4262
4263         for (i=0; t->addrconf_vars[i].data; i++) {
4264                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4265                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4266                 t->addrconf_vars[i].extra2 = net;
4267         }
4268
4269         /*
4270          * Make a copy of dev_name, because '.procname' is regarded as const
4271          * by sysctl and we wouldn't want anyone to change it under our feet
4272          * (see SIOCSIFNAME).
4273          */
4274         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4275         if (!t->dev_name)
4276                 goto free;
4277
4278         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4279         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4280
4281         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4282                         t->addrconf_vars);
4283         if (t->sysctl_header == NULL)
4284                 goto free_procname;
4285
4286         p->sysctl = t;
4287         return 0;
4288
4289 free_procname:
4290         kfree(t->dev_name);
4291 free:
4292         kfree(t);
4293 out:
4294         return -ENOBUFS;
4295 }
4296
4297 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4298 {
4299         struct addrconf_sysctl_table *t;
4300
4301         if (p->sysctl == NULL)
4302                 return;
4303
4304         t = p->sysctl;
4305         p->sysctl = NULL;
4306         unregister_sysctl_table(t->sysctl_header);
4307         kfree(t->dev_name);
4308         kfree(t);
4309 }
4310
4311 static void addrconf_sysctl_register(struct inet6_dev *idev)
4312 {
4313         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4314                               NET_IPV6_NEIGH, "ipv6",
4315                               &ndisc_ifinfo_sysctl_change,
4316                               ndisc_ifinfo_sysctl_strategy);
4317         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4318                         idev->dev->ifindex, idev, &idev->cnf);
4319 }
4320
4321 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4322 {
4323         __addrconf_sysctl_unregister(&idev->cnf);
4324         neigh_sysctl_unregister(idev->nd_parms);
4325 }
4326
4327
4328 #endif
4329
4330 static int addrconf_init_net(struct net *net)
4331 {
4332         int err;
4333         struct ipv6_devconf *all, *dflt;
4334
4335         err = -ENOMEM;
4336         all = &ipv6_devconf;
4337         dflt = &ipv6_devconf_dflt;
4338
4339         if (net != &init_net) {
4340                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4341                 if (all == NULL)
4342                         goto err_alloc_all;
4343
4344                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4345                 if (dflt == NULL)
4346                         goto err_alloc_dflt;
4347         }
4348
4349         net->ipv6.devconf_all = all;
4350         net->ipv6.devconf_dflt = dflt;
4351
4352 #ifdef CONFIG_SYSCTL
4353         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4354                         NULL, all);
4355         if (err < 0)
4356                 goto err_reg_all;
4357
4358         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4359                         NULL, dflt);
4360         if (err < 0)
4361                 goto err_reg_dflt;
4362 #endif
4363         return 0;
4364
4365 #ifdef CONFIG_SYSCTL
4366 err_reg_dflt:
4367         __addrconf_sysctl_unregister(all);
4368 err_reg_all:
4369         kfree(dflt);
4370 #endif
4371 err_alloc_dflt:
4372         kfree(all);
4373 err_alloc_all:
4374         return err;
4375 }
4376
4377 static void addrconf_exit_net(struct net *net)
4378 {
4379 #ifdef CONFIG_SYSCTL
4380         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4381         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4382 #endif
4383         if (net != &init_net) {
4384                 kfree(net->ipv6.devconf_dflt);
4385                 kfree(net->ipv6.devconf_all);
4386         }
4387 }
4388
4389 static struct pernet_operations addrconf_ops = {
4390         .init = addrconf_init_net,
4391         .exit = addrconf_exit_net,
4392 };
4393
4394 /*
4395  *      Device notifier
4396  */
4397
4398 int register_inet6addr_notifier(struct notifier_block *nb)
4399 {
4400         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4401 }
4402
4403 EXPORT_SYMBOL(register_inet6addr_notifier);
4404
4405 int unregister_inet6addr_notifier(struct notifier_block *nb)
4406 {
4407         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4408 }
4409
4410 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4411
4412 /*
4413  *      Init / cleanup code
4414  */
4415
4416 int __init addrconf_init(void)
4417 {
4418         int err;
4419
4420         if ((err = ipv6_addr_label_init()) < 0) {
4421                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4422                         err);
4423                 return err;
4424         }
4425
4426         register_pernet_subsys(&addrconf_ops);
4427
4428         /* The addrconf netdev notifier requires that loopback_dev
4429          * has it's ipv6 private information allocated and setup
4430          * before it can bring up and give link-local addresses
4431          * to other devices which are up.
4432          *
4433          * Unfortunately, loopback_dev is not necessarily the first
4434          * entry in the global dev_base list of net devices.  In fact,
4435          * it is likely to be the very last entry on that list.
4436          * So this causes the notifier registry below to try and
4437          * give link-local addresses to all devices besides loopback_dev
4438          * first, then loopback_dev, which cases all the non-loopback_dev
4439          * devices to fail to get a link-local address.
4440          *
4441          * So, as a temporary fix, allocate the ipv6 structure for
4442          * loopback_dev first by hand.
4443          * Longer term, all of the dependencies ipv6 has upon the loopback
4444          * device and it being up should be removed.
4445          */
4446         rtnl_lock();
4447         if (!ipv6_add_dev(init_net.loopback_dev))
4448                 err = -ENOMEM;
4449         rtnl_unlock();
4450         if (err)
4451                 goto errlo;
4452
4453         register_netdevice_notifier(&ipv6_dev_notf);
4454
4455         addrconf_verify(0);
4456
4457         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4458         if (err < 0)
4459                 goto errout;
4460
4461         /* Only the first call to __rtnl_register can fail */
4462         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4463         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4464         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4465         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4466         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4467
4468         ipv6_addr_label_rtnl_register();
4469
4470         return 0;
4471 errout:
4472         unregister_netdevice_notifier(&ipv6_dev_notf);
4473 errlo:
4474         unregister_pernet_subsys(&addrconf_ops);
4475
4476         return err;
4477 }
4478
4479 void addrconf_cleanup(void)
4480 {
4481         struct inet6_ifaddr *ifa;
4482         struct net_device *dev;
4483         int i;
4484
4485         unregister_netdevice_notifier(&ipv6_dev_notf);
4486         unregister_pernet_subsys(&addrconf_ops);
4487
4488         rtnl_lock();
4489
4490         /* clean dev list */
4491         for_each_netdev(&init_net, dev) {
4492                 if (__in6_dev_get(dev) == NULL)
4493                         continue;
4494                 addrconf_ifdown(dev, 1);
4495         }
4496         addrconf_ifdown(init_net.loopback_dev, 2);
4497
4498         /*
4499          *      Check hash table.
4500          */
4501         write_lock_bh(&addrconf_hash_lock);
4502         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4503                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4504                         struct inet6_ifaddr *bifa;
4505
4506                         bifa = ifa;
4507                         ifa = ifa->lst_next;
4508                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4509                         /* Do not free it; something is wrong.
4510                            Now we can investigate it with debugger.
4511                          */
4512                 }
4513         }
4514         write_unlock_bh(&addrconf_hash_lock);
4515
4516         del_timer(&addr_chk_timer);
4517         rtnl_unlock();
4518 }