git.oblomov.eu Git - 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/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/in6.h>
  47 #include <linux/netdevice.h>
  48 #include <linux/if_addr.h>
  49 #include <linux/if_arp.h>
  50 #include <linux/if_arcnet.h>
  51 #include <linux/if_infiniband.h>
  52 #include <linux/route.h>
  53 #include <linux/inetdevice.h>
  54 #include <linux/init.h>
  55 #ifdef CONFIG_SYSCTL
  56 #include <linux/sysctl.h>
  57 #endif
  58 #include <linux/capability.h>
  59 #include <linux/delay.h>
  60 #include <linux/notifier.h>
  61 #include <linux/string.h>
  62
  63 #include <net/net_namespace.h>
  64 #include <net/sock.h>
  65 #include <net/snmp.h>
  66
  67 #include <net/ipv6.h>
  68 #include <net/protocol.h>
  69 #include <net/ndisc.h>
  70 #include <net/ip6_route.h>
  71 #include <net/addrconf.h>
  72 #include <net/tcp.h>
  73 #include <net/ip.h>
  74 #include <net/netlink.h>
  75 #include <net/pkt_sched.h>
  76 #include <linux/if_tunnel.h>
  77 #include <linux/rtnetlink.h>
  78
  79 #ifdef CONFIG_IPV6_PRIVACY
  80 #include <linux/random.h>
  81 #endif
  82
  83 #include <asm/uaccess.h>
  84 #include <asm/unaligned.h>
  85
  86 #include <linux/proc_fs.h>
  87 #include <linux/seq_file.h>
  88
  89 /* Set to 3 to get tracing... */
  90 #define ACONF_DEBUG 2
  91
  92 #if ACONF_DEBUG >= 3
  93 #define ADBG(x) printk x
  94 #else
  95 #define ADBG(x)
  96 #endif
  97
  98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 100
 101 #ifdef CONFIG_SYSCTL
 102 static void addrconf_sysctl_register(struct inet6_dev *idev);
 103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 104 #else
 105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 106 {
 107 }
 108
 109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 110 {
 111 }
 112 #endif
 113
 114 #ifdef CONFIG_IPV6_PRIVACY
 115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 117 static void ipv6_regen_rndid(unsigned long data);
 118
 119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 120 #endif
 121
 122 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 123 static int ipv6_count_addresses(struct inet6_dev *idev);
 124
 125 /*
 126  *      Configured unicast address hash table
 127  */
 128 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 129 static DEFINE_RWLOCK(addrconf_hash_lock);
 130
 131 static void addrconf_verify(unsigned long);
 132
 133 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 134 static DEFINE_SPINLOCK(addrconf_verify_lock);
 135
 136 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 137 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 138
 139 static int addrconf_ifdown(struct net_device *dev, int how);
 140
 141 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 142 static void addrconf_dad_timer(unsigned long data);
 143 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 144 static void addrconf_dad_run(struct inet6_dev *idev);
 145 static void addrconf_rs_timer(unsigned long data);
 146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 147 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148
 149 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 150                                 struct prefix_info *pinfo);
 151 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 152                               struct net_device *dev);
 153
 154 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 155
 156 struct ipv6_devconf ipv6_devconf __read_mostly = {
 157         .forwarding             = 0,
 158         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 159         .mtu6                   = IPV6_MIN_MTU,
 160         .accept_ra              = 1,
 161         .accept_redirects       = 1,
 162         .autoconf               = 1,
 163         .force_mld_version      = 0,
 164         .dad_transmits          = 1,
 165         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 166         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 167         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 168 #ifdef CONFIG_IPV6_PRIVACY
 169         .use_tempaddr           = 0,
 170         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 171         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 172         .regen_max_retry        = REGEN_MAX_RETRY,
 173         .max_desync_factor      = MAX_DESYNC_FACTOR,
 174 #endif
 175         .max_addresses          = IPV6_MAX_ADDRESSES,
 176         .accept_ra_defrtr       = 1,
 177         .accept_ra_pinfo        = 1,
 178 #ifdef CONFIG_IPV6_ROUTER_PREF
 179         .accept_ra_rtr_pref     = 1,
 180         .rtr_probe_interval     = 60 * HZ,
 181 #ifdef CONFIG_IPV6_ROUTE_INFO
 182         .accept_ra_rt_info_max_plen = 0,
 183 #endif
 184 #endif
 185         .proxy_ndp              = 0,
 186         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 187         .disable_ipv6           = 0,
 188         .accept_dad             = 1,
 189 };
 190
 191 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 192         .forwarding             = 0,
 193         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 194         .mtu6                   = IPV6_MIN_MTU,
 195         .accept_ra              = 1,
 196         .accept_redirects       = 1,
 197         .autoconf               = 1,
 198         .dad_transmits          = 1,
 199         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 200         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 201         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 202 #ifdef CONFIG_IPV6_PRIVACY
 203         .use_tempaddr           = 0,
 204         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 205         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 206         .regen_max_retry        = REGEN_MAX_RETRY,
 207         .max_desync_factor      = MAX_DESYNC_FACTOR,
 208 #endif
 209         .max_addresses          = IPV6_MAX_ADDRESSES,
 210         .accept_ra_defrtr       = 1,
 211         .accept_ra_pinfo        = 1,
 212 #ifdef CONFIG_IPV6_ROUTER_PREF
 213         .accept_ra_rtr_pref     = 1,
 214         .rtr_probe_interval     = 60 * HZ,
 215 #ifdef CONFIG_IPV6_ROUTE_INFO
 216         .accept_ra_rt_info_max_plen = 0,
 217 #endif
 218 #endif
 219         .proxy_ndp              = 0,
 220         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 221         .disable_ipv6           = 0,
 222         .accept_dad             = 1,
 223 };
 224
 225 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 226 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 227 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 228 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 229 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 230
 231 /* Check if a valid qdisc is available */
 232 static inline int addrconf_qdisc_ok(struct net_device *dev)
 233 {
 234         return (dev->qdisc != &noop_qdisc);
 235 }
 236
 237 /* Check if a route is valid prefix route */
 238 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 239 {
 240         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 241 }
 242
 243 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 244 {
 245         if (del_timer(&ifp->timer))
 246                 __in6_ifa_put(ifp);
 247 }
 248
 249 enum addrconf_timer_t
 250 {
 251         AC_NONE,
 252         AC_DAD,
 253         AC_RS,
 254 };
 255
 256 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 257                                enum addrconf_timer_t what,
 258                                unsigned long when)
 259 {
 260         if (!del_timer(&ifp->timer))
 261                 in6_ifa_hold(ifp);
 262
 263         switch (what) {
 264         case AC_DAD:
 265                 ifp->timer.function = addrconf_dad_timer;
 266                 break;
 267         case AC_RS:
 268                 ifp->timer.function = addrconf_rs_timer;
 269                 break;
 270         default:;
 271         }
 272         ifp->timer.expires = jiffies + when;
 273         add_timer(&ifp->timer);
 274 }
 275
 276 static int snmp6_alloc_dev(struct inet6_dev *idev)
 277 {
 278         if (snmp_mib_init((void **)idev->stats.ipv6,
 279                           sizeof(struct ipstats_mib)) < 0)
 280                 goto err_ip;
 281         if (snmp_mib_init((void **)idev->stats.icmpv6,
 282                           sizeof(struct icmpv6_mib)) < 0)
 283                 goto err_icmp;
 284         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 285                           sizeof(struct icmpv6msg_mib)) < 0)
 286                 goto err_icmpmsg;
 287
 288         return 0;
 289
 290 err_icmpmsg:
 291         snmp_mib_free((void **)idev->stats.icmpv6);
 292 err_icmp:
 293         snmp_mib_free((void **)idev->stats.ipv6);
 294 err_ip:
 295         return -ENOMEM;
 296 }
 297
 298 static void snmp6_free_dev(struct inet6_dev *idev)
 299 {
 300         snmp_mib_free((void **)idev->stats.icmpv6msg);
 301         snmp_mib_free((void **)idev->stats.icmpv6);
 302         snmp_mib_free((void **)idev->stats.ipv6);
 303 }
 304
 305 /* Nobody refers to this device, we may destroy it. */
 306
 307 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 308 {
 309         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 310         kfree(idev);
 311 }
 312
 313 void in6_dev_finish_destroy(struct inet6_dev *idev)
 314 {
 315         struct net_device *dev = idev->dev;
 316         BUG_TRAP(idev->addr_list==NULL);
 317         BUG_TRAP(idev->mc_list==NULL);
 318 #ifdef NET_REFCNT_DEBUG
 319         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 320 #endif
 321         dev_put(dev);
 322         if (!idev->dead) {
 323                 printk("Freeing alive inet6 device %p\n", idev);
 324                 return;
 325         }
 326         snmp6_free_dev(idev);
 327         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 328 }
 329
 330 EXPORT_SYMBOL(in6_dev_finish_destroy);
 331
 332 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 333 {
 334         struct inet6_dev *ndev;
 335
 336         ASSERT_RTNL();
 337
 338         if (dev->mtu < IPV6_MIN_MTU)
 339                 return NULL;
 340
 341         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 342
 343         if (ndev == NULL)
 344                 return NULL;
 345
 346         rwlock_init(&ndev->lock);
 347         ndev->dev = dev;
 348         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 349         ndev->cnf.mtu6 = dev->mtu;
 350         ndev->cnf.sysctl = NULL;
 351         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 352         if (ndev->nd_parms == NULL) {
 353                 kfree(ndev);
 354                 return NULL;
 355         }
 356         if (ndev->cnf.forwarding)
 357                 dev_disable_lro(dev);
 358         /* We refer to the device */
 359         dev_hold(dev);
 360
 361         if (snmp6_alloc_dev(ndev) < 0) {
 362                 ADBG((KERN_WARNING
 363                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 364                         __func__, dev->name));
 365                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 366                 ndev->dead = 1;
 367                 in6_dev_finish_destroy(ndev);
 368                 return NULL;
 369         }
 370
 371         if (snmp6_register_dev(ndev) < 0) {
 372                 ADBG((KERN_WARNING
 373                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 374                         __func__, dev->name));
 375                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 376                 ndev->dead = 1;
 377                 in6_dev_finish_destroy(ndev);
 378                 return NULL;
 379         }
 380
 381         /* One reference from device.  We must do this before
 382          * we invoke __ipv6_regen_rndid().
 383          */
 384         in6_dev_hold(ndev);
 385
 386         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 387                 ndev->cnf.accept_dad = -1;
 388
 389 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 390         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 391                 printk(KERN_INFO
 392                        "%s: Disabled Multicast RS\n",
 393                        dev->name);
 394                 ndev->cnf.rtr_solicits = 0;
 395         }
 396 #endif
 397
 398 #ifdef CONFIG_IPV6_PRIVACY
 399         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 400         if ((dev->flags&IFF_LOOPBACK) ||
 401             dev->type == ARPHRD_TUNNEL ||
 402             dev->type == ARPHRD_TUNNEL6 ||
 403             dev->type == ARPHRD_SIT ||
 404             dev->type == ARPHRD_NONE) {
 405                 printk(KERN_INFO
 406                        "%s: Disabled Privacy Extensions\n",
 407                        dev->name);
 408                 ndev->cnf.use_tempaddr = -1;
 409         } else {
 410                 in6_dev_hold(ndev);
 411                 ipv6_regen_rndid((unsigned long) ndev);
 412         }
 413 #endif
 414
 415         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 416                 ndev->if_flags |= IF_READY;
 417
 418         ipv6_mc_init_dev(ndev);
 419         ndev->tstamp = jiffies;
 420         addrconf_sysctl_register(ndev);
 421         /* protected by rtnl_lock */
 422         rcu_assign_pointer(dev->ip6_ptr, ndev);
 423
 424         /* Join all-node multicast group */
 425         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 426
 427         return ndev;
 428 }
 429
 430 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 431 {
 432         struct inet6_dev *idev;
 433
 434         ASSERT_RTNL();
 435
 436         if ((idev = __in6_dev_get(dev)) == NULL) {
 437                 if ((idev = ipv6_add_dev(dev)) == NULL)
 438                         return NULL;
 439         }
 440
 441         if (dev->flags&IFF_UP)
 442                 ipv6_mc_up(idev);
 443         return idev;
 444 }
 445
 446 #ifdef CONFIG_SYSCTL
 447 static void dev_forward_change(struct inet6_dev *idev)
 448 {
 449         struct net_device *dev;
 450         struct inet6_ifaddr *ifa;
 451
 452         if (!idev)
 453                 return;
 454         dev = idev->dev;
 455         if (idev->cnf.forwarding)
 456                 dev_disable_lro(dev);
 457         if (dev && (dev->flags & IFF_MULTICAST)) {
 458                 if (idev->cnf.forwarding)
 459                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 460                 else
 461                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 462         }
 463         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 464                 if (ifa->flags&IFA_F_TENTATIVE)
 465                         continue;
 466                 if (idev->cnf.forwarding)
 467                         addrconf_join_anycast(ifa);
 468                 else
 469                         addrconf_leave_anycast(ifa);
 470         }
 471 }
 472
 473
 474 static void addrconf_forward_change(struct net *net, __s32 newf)
 475 {
 476         struct net_device *dev;
 477         struct inet6_dev *idev;
 478
 479         read_lock(&dev_base_lock);
 480         for_each_netdev(net, dev) {
 481                 rcu_read_lock();
 482                 idev = __in6_dev_get(dev);
 483                 if (idev) {
 484                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 485                         idev->cnf.forwarding = newf;
 486                         if (changed)
 487                                 dev_forward_change(idev);
 488                 }
 489                 rcu_read_unlock();
 490         }
 491         read_unlock(&dev_base_lock);
 492 }
 493
 494 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 495 {
 496         struct net *net;
 497
 498         net = (struct net *)table->extra2;
 499         if (p == &net->ipv6.devconf_dflt->forwarding)
 500                 return;
 501
 502         rtnl_lock();
 503         if (p == &net->ipv6.devconf_all->forwarding) {
 504                 __s32 newf = net->ipv6.devconf_all->forwarding;
 505                 net->ipv6.devconf_dflt->forwarding = newf;
 506                 addrconf_forward_change(net, newf);
 507         } else if ((!*p) ^ (!old))
 508                 dev_forward_change((struct inet6_dev *)table->extra1);
 509         rtnl_unlock();
 510
 511         if (*p)
 512                 rt6_purge_dflt_routers(net);
 513 }
 514 #endif
 515
 516 /* Nobody refers to this ifaddr, destroy it */
 517
 518 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 519 {
 520         BUG_TRAP(ifp->if_next==NULL);
 521         BUG_TRAP(ifp->lst_next==NULL);
 522 #ifdef NET_REFCNT_DEBUG
 523         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 524 #endif
 525
 526         in6_dev_put(ifp->idev);
 527
 528         if (del_timer(&ifp->timer))
 529                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 530
 531         if (!ifp->dead) {
 532                 printk("Freeing alive inet6 address %p\n", ifp);
 533                 return;
 534         }
 535         dst_release(&ifp->rt->u.dst);
 536
 537         kfree(ifp);
 538 }
 539
 540 static void
 541 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 542 {
 543         struct inet6_ifaddr *ifa, **ifap;
 544         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 545
 546         /*
 547          * Each device address list is sorted in order of scope -
 548          * global before linklocal.
 549          */
 550         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 551              ifap = &ifa->if_next) {
 552                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 553                         break;
 554         }
 555
 556         ifp->if_next = *ifap;
 557         *ifap = ifp;
 558 }
 559
 560 /*
 561  *      Hash function taken from net_alias.c
 562  */
 563 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 564 {
 565         __u32 word;
 566
 567         /*
 568          * We perform the hash function over the last 64 bits of the address
 569          * This will include the IEEE address token on links that support it.
 570          */
 571
 572         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 573         word ^= (word >> 16);
 574         word ^= (word >> 8);
 575
 576         return ((word ^ (word >> 4)) & 0x0f);
 577 }
 578
 579 /* On success it returns ifp with increased reference count */
 580
 581 static struct inet6_ifaddr *
 582 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 583               int scope, u32 flags)
 584 {
 585         struct inet6_ifaddr *ifa = NULL;
 586         struct rt6_info *rt;
 587         int hash;
 588         int err = 0;
 589         int addr_type = ipv6_addr_type(addr);
 590
 591         if (addr_type == IPV6_ADDR_ANY ||
 592             addr_type & IPV6_ADDR_MULTICAST ||
 593             (!(idev->dev->flags & IFF_LOOPBACK) &&
 594              addr_type & IPV6_ADDR_LOOPBACK))
 595                 return ERR_PTR(-EADDRNOTAVAIL);
 596
 597         rcu_read_lock_bh();
 598         if (idev->dead) {
 599                 err = -ENODEV;                  /*XXX*/
 600                 goto out2;
 601         }
 602
 603         write_lock(&addrconf_hash_lock);
 604
 605         /* Ignore adding duplicate addresses on an interface */
 606         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 607                 ADBG(("ipv6_add_addr: already assigned\n"));
 608                 err = -EEXIST;
 609                 goto out;
 610         }
 611
 612         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 613
 614         if (ifa == NULL) {
 615                 ADBG(("ipv6_add_addr: malloc failed\n"));
 616                 err = -ENOBUFS;
 617                 goto out;
 618         }
 619
 620         rt = addrconf_dst_alloc(idev, addr, 0);
 621         if (IS_ERR(rt)) {
 622                 err = PTR_ERR(rt);
 623                 goto out;
 624         }
 625
 626         ipv6_addr_copy(&ifa->addr, addr);
 627
 628         spin_lock_init(&ifa->lock);
 629         init_timer(&ifa->timer);
 630         ifa->timer.data = (unsigned long) ifa;
 631         ifa->scope = scope;
 632         ifa->prefix_len = pfxlen;
 633         ifa->flags = flags | IFA_F_TENTATIVE;
 634         ifa->cstamp = ifa->tstamp = jiffies;
 635
 636         ifa->rt = rt;
 637
 638         /*
 639          * part one of RFC 4429, section 3.3
 640          * We should not configure an address as
 641          * optimistic if we do not yet know the link
 642          * layer address of our nexhop router
 643          */
 644
 645         if (rt->rt6i_nexthop == NULL)
 646                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 647
 648         ifa->idev = idev;
 649         in6_dev_hold(idev);
 650         /* For caller */
 651         in6_ifa_hold(ifa);
 652
 653         /* Add to big hash table */
 654         hash = ipv6_addr_hash(addr);
 655
 656         ifa->lst_next = inet6_addr_lst[hash];
 657         inet6_addr_lst[hash] = ifa;
 658         in6_ifa_hold(ifa);
 659         write_unlock(&addrconf_hash_lock);
 660
 661         write_lock(&idev->lock);
 662         /* Add to inet6_dev unicast addr list. */
 663         ipv6_link_dev_addr(idev, ifa);
 664
 665 #ifdef CONFIG_IPV6_PRIVACY
 666         if (ifa->flags&IFA_F_TEMPORARY) {
 667                 ifa->tmp_next = idev->tempaddr_list;
 668                 idev->tempaddr_list = ifa;
 669                 in6_ifa_hold(ifa);
 670         }
 671 #endif
 672
 673         in6_ifa_hold(ifa);
 674         write_unlock(&idev->lock);
 675 out2:
 676         rcu_read_unlock_bh();
 677
 678         if (likely(err == 0))
 679                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 680         else {
 681                 kfree(ifa);
 682                 ifa = ERR_PTR(err);
 683         }
 684
 685         return ifa;
 686 out:
 687         write_unlock(&addrconf_hash_lock);
 688         goto out2;
 689 }
 690
 691 /* This function wants to get referenced ifp and releases it before return */
 692
 693 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 694 {
 695         struct inet6_ifaddr *ifa, **ifap;
 696         struct inet6_dev *idev = ifp->idev;
 697         int hash;
 698         int deleted = 0, onlink = 0;
 699         unsigned long expires = jiffies;
 700
 701         hash = ipv6_addr_hash(&ifp->addr);
 702
 703         ifp->dead = 1;
 704
 705         write_lock_bh(&addrconf_hash_lock);
 706         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 707              ifap = &ifa->lst_next) {
 708                 if (ifa == ifp) {
 709                         *ifap = ifa->lst_next;
 710                         __in6_ifa_put(ifp);
 711                         ifa->lst_next = NULL;
 712                         break;
 713                 }
 714         }
 715         write_unlock_bh(&addrconf_hash_lock);
 716
 717         write_lock_bh(&idev->lock);
 718 #ifdef CONFIG_IPV6_PRIVACY
 719         if (ifp->flags&IFA_F_TEMPORARY) {
 720                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 721                      ifap = &ifa->tmp_next) {
 722                         if (ifa == ifp) {
 723                                 *ifap = ifa->tmp_next;
 724                                 if (ifp->ifpub) {
 725                                         in6_ifa_put(ifp->ifpub);
 726                                         ifp->ifpub = NULL;
 727                                 }
 728                                 __in6_ifa_put(ifp);
 729                                 ifa->tmp_next = NULL;
 730                                 break;
 731                         }
 732                 }
 733         }
 734 #endif
 735
 736         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 737                 if (ifa == ifp) {
 738                         *ifap = ifa->if_next;
 739                         __in6_ifa_put(ifp);
 740                         ifa->if_next = NULL;
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773                 ifap = &ifa->if_next;
 774         }
 775         write_unlock_bh(&idev->lock);
 776
 777         addrconf_del_timer(ifp);
 778
 779         ipv6_ifa_notify(RTM_DELADDR, ifp);
 780
 781         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 782
 783         /*
 784          * Purge or update corresponding prefix
 785          *
 786          * 1) we don't purge prefix here if address was not permanent.
 787          *    prefix is managed by its own lifetime.
 788          * 2) if there're no addresses, delete prefix.
 789          * 3) if there're still other permanent address(es),
 790          *    corresponding prefix is still permanent.
 791          * 4) otherwise, update prefix lifetime to the
 792          *    longest valid lifetime among the corresponding
 793          *    addresses on the device.
 794          *    Note: subsequent RA will update lifetime.
 795          *
 796          * --yoshfuji
 797          */
 798         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 799                 struct in6_addr prefix;
 800                 struct rt6_info *rt;
 801                 struct net *net = dev_net(ifp->idev->dev);
 802                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 803                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 804
 805                 if (rt && addrconf_is_prefix_route(rt)) {
 806                         if (onlink == 0) {
 807                                 ip6_del_rt(rt);
 808                                 rt = NULL;
 809                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 810                                 rt->rt6i_expires = expires;
 811                                 rt->rt6i_flags |= RTF_EXPIRES;
 812                         }
 813                 }
 814                 dst_release(&rt->u.dst);
 815         }
 816
 817         in6_ifa_put(ifp);
 818 }
 819
 820 #ifdef CONFIG_IPV6_PRIVACY
 821 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 822 {
 823         struct inet6_dev *idev = ifp->idev;
 824         struct in6_addr addr, *tmpaddr;
 825         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 826         unsigned long regen_advance;
 827         int tmp_plen;
 828         int ret = 0;
 829         int max_addresses;
 830         u32 addr_flags;
 831
 832         write_lock(&idev->lock);
 833         if (ift) {
 834                 spin_lock_bh(&ift->lock);
 835                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 836                 spin_unlock_bh(&ift->lock);
 837                 tmpaddr = &addr;
 838         } else {
 839                 tmpaddr = NULL;
 840         }
 841 retry:
 842         in6_dev_hold(idev);
 843         if (idev->cnf.use_tempaddr <= 0) {
 844                 write_unlock(&idev->lock);
 845                 printk(KERN_INFO
 846                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 847                 in6_dev_put(idev);
 848                 ret = -1;
 849                 goto out;
 850         }
 851         spin_lock_bh(&ifp->lock);
 852         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 853                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 854                 spin_unlock_bh(&ifp->lock);
 855                 write_unlock(&idev->lock);
 856                 printk(KERN_WARNING
 857                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 858                 in6_dev_put(idev);
 859                 ret = -1;
 860                 goto out;
 861         }
 862         in6_ifa_hold(ifp);
 863         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 864         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 865                 spin_unlock_bh(&ifp->lock);
 866                 write_unlock(&idev->lock);
 867                 printk(KERN_WARNING
 868                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 869                 in6_ifa_put(ifp);
 870                 in6_dev_put(idev);
 871                 ret = -1;
 872                 goto out;
 873         }
 874         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 875         tmp_valid_lft = min_t(__u32,
 876                               ifp->valid_lft,
 877                               idev->cnf.temp_valid_lft);
 878         tmp_prefered_lft = min_t(__u32,
 879                                  ifp->prefered_lft,
 880                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 881         tmp_plen = ifp->prefix_len;
 882         max_addresses = idev->cnf.max_addresses;
 883         tmp_cstamp = ifp->cstamp;
 884         tmp_tstamp = ifp->tstamp;
 885         spin_unlock_bh(&ifp->lock);
 886
 887         regen_advance = idev->cnf.regen_max_retry *
 888                         idev->cnf.dad_transmits *
 889                         idev->nd_parms->retrans_time / HZ;
 890         write_unlock(&idev->lock);
 891
 892         /* A temporary address is created only if this calculated Preferred
 893          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 894          * an implementation must not create a temporary address with a zero
 895          * Preferred Lifetime.
 896          */
 897         if (tmp_prefered_lft <= regen_advance) {
 898                 in6_ifa_put(ifp);
 899                 in6_dev_put(idev);
 900                 ret = -1;
 901                 goto out;
 902         }
 903
 904         addr_flags = IFA_F_TEMPORARY;
 905         /* set in addrconf_prefix_rcv() */
 906         if (ifp->flags & IFA_F_OPTIMISTIC)
 907                 addr_flags |= IFA_F_OPTIMISTIC;
 908
 909         ift = !max_addresses ||
 910               ipv6_count_addresses(idev) < max_addresses ?
 911                 ipv6_add_addr(idev, &addr, tmp_plen,
 912                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 913                               addr_flags) : NULL;
 914         if (!ift || IS_ERR(ift)) {
 915                 in6_ifa_put(ifp);
 916                 in6_dev_put(idev);
 917                 printk(KERN_INFO
 918                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 919                 tmpaddr = &addr;
 920                 write_lock(&idev->lock);
 921                 goto retry;
 922         }
 923
 924         spin_lock_bh(&ift->lock);
 925         ift->ifpub = ifp;
 926         ift->valid_lft = tmp_valid_lft;
 927         ift->prefered_lft = tmp_prefered_lft;
 928         ift->cstamp = tmp_cstamp;
 929         ift->tstamp = tmp_tstamp;
 930         spin_unlock_bh(&ift->lock);
 931
 932         addrconf_dad_start(ift, 0);
 933         in6_ifa_put(ift);
 934         in6_dev_put(idev);
 935 out:
 936         return ret;
 937 }
 938 #endif
 939
 940 /*
 941  *      Choose an appropriate source address (RFC3484)
 942  */
 943 enum {
 944         IPV6_SADDR_RULE_INIT = 0,
 945         IPV6_SADDR_RULE_LOCAL,
 946         IPV6_SADDR_RULE_SCOPE,
 947         IPV6_SADDR_RULE_PREFERRED,
 948 #ifdef CONFIG_IPV6_MIP6
 949         IPV6_SADDR_RULE_HOA,
 950 #endif
 951         IPV6_SADDR_RULE_OIF,
 952         IPV6_SADDR_RULE_LABEL,
 953 #ifdef CONFIG_IPV6_PRIVACY
 954         IPV6_SADDR_RULE_PRIVACY,
 955 #endif
 956         IPV6_SADDR_RULE_ORCHID,
 957         IPV6_SADDR_RULE_PREFIX,
 958         IPV6_SADDR_RULE_MAX
 959 };
 960
 961 struct ipv6_saddr_score {
 962         int                     rule;
 963         int                     addr_type;
 964         struct inet6_ifaddr     *ifa;
 965         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 966         int                     scopedist;
 967         int                     matchlen;
 968 };
 969
 970 struct ipv6_saddr_dst {
 971         const struct in6_addr *addr;
 972         int ifindex;
 973         int scope;
 974         int label;
 975         unsigned int prefs;
 976 };
 977
 978 static inline int ipv6_saddr_preferred(int type)
 979 {
 980         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 981                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 982                 return 1;
 983         return 0;
 984 }
 985
 986 static int ipv6_get_saddr_eval(struct net *net,
 987                                struct ipv6_saddr_score *score,
 988                                struct ipv6_saddr_dst *dst,
 989                                int i)
 990 {
 991         int ret;
 992
 993         if (i <= score->rule) {
 994                 switch (i) {
 995                 case IPV6_SADDR_RULE_SCOPE:
 996                         ret = score->scopedist;
 997                         break;
 998                 case IPV6_SADDR_RULE_PREFIX:
 999                         ret = score->matchlen;
1000                         break;
1001                 default:
1002                         ret = !!test_bit(i, score->scorebits);
1003                 }
1004                 goto out;
1005         }
1006
1007         switch (i) {
1008         case IPV6_SADDR_RULE_INIT:
1009                 /* Rule 0: remember if hiscore is not ready yet */
1010                 ret = !!score->ifa;
1011                 break;
1012         case IPV6_SADDR_RULE_LOCAL:
1013                 /* Rule 1: Prefer same address */
1014                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1015                 break;
1016         case IPV6_SADDR_RULE_SCOPE:
1017                 /* Rule 2: Prefer appropriate scope
1018                  *
1019                  *      ret
1020                  *       ^
1021                  *    -1 |  d 15
1022                  *    ---+--+-+---> scope
1023                  *       |
1024                  *       |             d is scope of the destination.
1025                  *  B-d  |  \
1026                  *       |   \      <- smaller scope is better if
1027                  *  B-15 |    \        if scope is enough for destinaion.
1028                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1029                  * d-C-1 | /
1030                  *       |/         <- greater is better
1031                  *   -C  /             if scope is not enough for destination.
1032                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1033                  *
1034                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1035                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1036                  * Assume B = 0 and we get C > 29.
1037                  */
1038                 ret = __ipv6_addr_src_scope(score->addr_type);
1039                 if (ret >= dst->scope)
1040                         ret = -ret;
1041                 else
1042                         ret -= 128;     /* 30 is enough */
1043                 score->scopedist = ret;
1044                 break;
1045         case IPV6_SADDR_RULE_PREFERRED:
1046                 /* Rule 3: Avoid deprecated and optimistic addresses */
1047                 ret = ipv6_saddr_preferred(score->addr_type) ||
1048                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1049                 break;
1050 #ifdef CONFIG_IPV6_MIP6
1051         case IPV6_SADDR_RULE_HOA:
1052             {
1053                 /* Rule 4: Prefer home address */
1054                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1055                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1056                 break;
1057             }
1058 #endif
1059         case IPV6_SADDR_RULE_OIF:
1060                 /* Rule 5: Prefer outgoing interface */
1061                 ret = (!dst->ifindex ||
1062                        dst->ifindex == score->ifa->idev->dev->ifindex);
1063                 break;
1064         case IPV6_SADDR_RULE_LABEL:
1065                 /* Rule 6: Prefer matching label */
1066                 ret = ipv6_addr_label(net,
1067                                       &score->ifa->addr, score->addr_type,
1068                                       score->ifa->idev->dev->ifindex) == dst->label;
1069                 break;
1070 #ifdef CONFIG_IPV6_PRIVACY
1071         case IPV6_SADDR_RULE_PRIVACY:
1072             {
1073                 /* Rule 7: Prefer public address
1074                  * Note: prefer temprary address if use_tempaddr >= 2
1075                  */
1076                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1077                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1078                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1079                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1080                 break;
1081             }
1082 #endif
1083         case IPV6_SADDR_RULE_ORCHID:
1084                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1085                  *          non-ORCHID vs non-ORCHID
1086                  */
1087                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1088                         ipv6_addr_orchid(dst->addr));
1089                 break;
1090         case IPV6_SADDR_RULE_PREFIX:
1091                 /* Rule 8: Use longest matching prefix */
1092                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1093                                                        dst->addr);
1094                 break;
1095         default:
1096                 ret = 0;
1097         }
1098
1099         if (ret)
1100                 __set_bit(i, score->scorebits);
1101         score->rule = i;
1102 out:
1103         return ret;
1104 }
1105
1106 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1107                        const struct in6_addr *daddr, unsigned int prefs,
1108                        struct in6_addr *saddr)
1109 {
1110         struct ipv6_saddr_score scores[2],
1111                                 *score = &scores[0], *hiscore = &scores[1];
1112         struct net *net = dev_net(dst_dev);
1113         struct ipv6_saddr_dst dst;
1114         struct net_device *dev;
1115         int dst_type;
1116
1117         dst_type = __ipv6_addr_type(daddr);
1118         dst.addr = daddr;
1119         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1120         dst.scope = __ipv6_addr_src_scope(dst_type);
1121         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1122         dst.prefs = prefs;
1123
1124         hiscore->rule = -1;
1125         hiscore->ifa = NULL;
1126
1127         read_lock(&dev_base_lock);
1128         rcu_read_lock();
1129
1130         for_each_netdev(net, dev) {
1131                 struct inet6_dev *idev;
1132
1133                 /* Candidate Source Address (section 4)
1134                  *  - multicast and link-local destination address,
1135                  *    the set of candidate source address MUST only
1136                  *    include addresses assigned to interfaces
1137                  *    belonging to the same link as the outgoing
1138                  *    interface.
1139                  * (- For site-local destination addresses, the
1140                  *    set of candidate source addresses MUST only
1141                  *    include addresses assigned to interfaces
1142                  *    belonging to the same site as the outgoing
1143                  *    interface.)
1144                  */
1145                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1146                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1147                     dst.ifindex && dev->ifindex != dst.ifindex)
1148                         continue;
1149
1150                 idev = __in6_dev_get(dev);
1151                 if (!idev)
1152                         continue;
1153
1154                 read_lock_bh(&idev->lock);
1155                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1156                         int i;
1157
1158                         /*
1159                          * - Tentative Address (RFC2462 section 5.4)
1160                          *  - A tentative address is not considered
1161                          *    "assigned to an interface" in the traditional
1162                          *    sense, unless it is also flagged as optimistic.
1163                          * - Candidate Source Address (section 4)
1164                          *  - In any case, anycast addresses, multicast
1165                          *    addresses, and the unspecified address MUST
1166                          *    NOT be included in a candidate set.
1167                          */
1168                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1169                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1170                                 continue;
1171
1172                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1173
1174                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1175                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1176                                 LIMIT_NETDEBUG(KERN_DEBUG
1177                                                "ADDRCONF: unspecified / multicast address "
1178                                                "assigned as unicast address on %s",
1179                                                dev->name);
1180                                 continue;
1181                         }
1182
1183                         score->rule = -1;
1184                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1185
1186                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1187                                 int minihiscore, miniscore;
1188
1189                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1190                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1191
1192                                 if (minihiscore > miniscore) {
1193                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1194                                             score->scopedist > 0) {
1195                                                 /*
1196                                                  * special case:
1197                                                  * each remaining entry
1198                                                  * has too small (not enough)
1199                                                  * scope, because ifa entries
1200                                                  * are sorted by their scope
1201                                                  * values.
1202                                                  */
1203                                                 goto try_nextdev;
1204                                         }
1205                                         break;
1206                                 } else if (minihiscore < miniscore) {
1207                                         struct ipv6_saddr_score *tmp;
1208
1209                                         if (hiscore->ifa)
1210                                                 in6_ifa_put(hiscore->ifa);
1211
1212                                         in6_ifa_hold(score->ifa);
1213
1214                                         tmp = hiscore;
1215                                         hiscore = score;
1216                                         score = tmp;
1217
1218                                         /* restore our iterator */
1219                                         score->ifa = hiscore->ifa;
1220
1221                                         break;
1222                                 }
1223                         }
1224                 }
1225 try_nextdev:
1226                 read_unlock_bh(&idev->lock);
1227         }
1228         rcu_read_unlock();
1229         read_unlock(&dev_base_lock);
1230
1231         if (!hiscore->ifa)
1232                 return -EADDRNOTAVAIL;
1233
1234         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1235         in6_ifa_put(hiscore->ifa);
1236         return 0;
1237 }
1238
1239 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1240
1241 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1242                     unsigned char banned_flags)
1243 {
1244         struct inet6_dev *idev;
1245         int err = -EADDRNOTAVAIL;
1246
1247         rcu_read_lock();
1248         if ((idev = __in6_dev_get(dev)) != NULL) {
1249                 struct inet6_ifaddr *ifp;
1250
1251                 read_lock_bh(&idev->lock);
1252                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1253                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1254                                 ipv6_addr_copy(addr, &ifp->addr);
1255                                 err = 0;
1256                                 break;
1257                         }
1258                 }
1259                 read_unlock_bh(&idev->lock);
1260         }
1261         rcu_read_unlock();
1262         return err;
1263 }
1264
1265 static int ipv6_count_addresses(struct inet6_dev *idev)
1266 {
1267         int cnt = 0;
1268         struct inet6_ifaddr *ifp;
1269
1270         read_lock_bh(&idev->lock);
1271         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1272                 cnt++;
1273         read_unlock_bh(&idev->lock);
1274         return cnt;
1275 }
1276
1277 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1278                   struct net_device *dev, int strict)
1279 {
1280         struct inet6_ifaddr * ifp;
1281         u8 hash = ipv6_addr_hash(addr);
1282
1283         read_lock_bh(&addrconf_hash_lock);
1284         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1285                 if (!net_eq(dev_net(ifp->idev->dev), net))
1286                         continue;
1287                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1288                     !(ifp->flags&IFA_F_TENTATIVE)) {
1289                         if (dev == NULL || ifp->idev->dev == dev ||
1290                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1291                                 break;
1292                 }
1293         }
1294         read_unlock_bh(&addrconf_hash_lock);
1295         return ifp != NULL;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static
1300 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301                        struct net_device *dev)
1302 {
1303         struct inet6_ifaddr * ifp;
1304         u8 hash = ipv6_addr_hash(addr);
1305
1306         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 break;
1312                 }
1313         }
1314         return ifp != NULL;
1315 }
1316
1317 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr * ifp;
1346         u8 hash = ipv6_addr_hash(addr);
1347
1348         read_lock_bh(&addrconf_hash_lock);
1349         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1350                 if (!net_eq(dev_net(ifp->idev->dev), net))
1351                         continue;
1352                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1353                         if (dev == NULL || ifp->idev->dev == dev ||
1354                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1355                                 in6_ifa_hold(ifp);
1356                                 break;
1357                         }
1358                 }
1359         }
1360         read_unlock_bh(&addrconf_hash_lock);
1361
1362         return ifp;
1363 }
1364
1365 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1366 {
1367         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1368         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1369         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1370         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1371         int sk_ipv6only = ipv6_only_sock(sk);
1372         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1373         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1374         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1375
1376         if (!sk2_rcv_saddr && !sk_ipv6only)
1377                 return 1;
1378
1379         if (addr_type2 == IPV6_ADDR_ANY &&
1380             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1381                 return 1;
1382
1383         if (addr_type == IPV6_ADDR_ANY &&
1384             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1385                 return 1;
1386
1387         if (sk2_rcv_saddr6 &&
1388             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1389                 return 1;
1390
1391         if (addr_type == IPV6_ADDR_MAPPED &&
1392             !sk2_ipv6only &&
1393             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1394                 return 1;
1395
1396         return 0;
1397 }
1398
1399 /* Gets referenced address, destroys ifaddr */
1400
1401 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1402 {
1403         if (ifp->flags&IFA_F_PERMANENT) {
1404                 spin_lock_bh(&ifp->lock);
1405                 addrconf_del_timer(ifp);
1406                 ifp->flags |= IFA_F_TENTATIVE;
1407                 spin_unlock_bh(&ifp->lock);
1408                 in6_ifa_put(ifp);
1409 #ifdef CONFIG_IPV6_PRIVACY
1410         } else if (ifp->flags&IFA_F_TEMPORARY) {
1411                 struct inet6_ifaddr *ifpub;
1412                 spin_lock_bh(&ifp->lock);
1413                 ifpub = ifp->ifpub;
1414                 if (ifpub) {
1415                         in6_ifa_hold(ifpub);
1416                         spin_unlock_bh(&ifp->lock);
1417                         ipv6_create_tempaddr(ifpub, ifp);
1418                         in6_ifa_put(ifpub);
1419                 } else {
1420                         spin_unlock_bh(&ifp->lock);
1421                 }
1422                 ipv6_del_addr(ifp);
1423 #endif
1424         } else
1425                 ipv6_del_addr(ifp);
1426 }
1427
1428 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1429 {
1430         struct inet6_dev *idev = ifp->idev;
1431         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1432                 struct in6_addr addr;
1433
1434                 addr.s6_addr32[0] = htonl(0xfe800000);
1435                 addr.s6_addr32[1] = 0;
1436
1437                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1438                     ipv6_addr_equal(&ifp->addr, &addr)) {
1439                         /* DAD failed for link-local based on MAC address */
1440                         idev->cnf.disable_ipv6 = 1;
1441                 }
1442         }
1443
1444         if (net_ratelimit())
1445                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1446         addrconf_dad_stop(ifp);
1447 }
1448
1449 /* Join to solicited addr multicast group. */
1450
1451 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1452 {
1453         struct in6_addr maddr;
1454
1455         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1456                 return;
1457
1458         addrconf_addr_solict_mult(addr, &maddr);
1459         ipv6_dev_mc_inc(dev, &maddr);
1460 }
1461
1462 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1463 {
1464         struct in6_addr maddr;
1465
1466         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1467                 return;
1468
1469         addrconf_addr_solict_mult(addr, &maddr);
1470         __ipv6_dev_mc_dec(idev, &maddr);
1471 }
1472
1473 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1474 {
1475         struct in6_addr addr;
1476         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1477         if (ipv6_addr_any(&addr))
1478                 return;
1479         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1480 }
1481
1482 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1483 {
1484         struct in6_addr addr;
1485         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1486         if (ipv6_addr_any(&addr))
1487                 return;
1488         __ipv6_dev_ac_dec(ifp->idev, &addr);
1489 }
1490
1491 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1492 {
1493         if (dev->addr_len != ETH_ALEN)
1494                 return -1;
1495         memcpy(eui, dev->dev_addr, 3);
1496         memcpy(eui + 5, dev->dev_addr + 3, 3);
1497
1498         /*
1499          * The zSeries OSA network cards can be shared among various
1500          * OS instances, but the OSA cards have only one MAC address.
1501          * This leads to duplicate address conflicts in conjunction
1502          * with IPv6 if more than one instance uses the same card.
1503          *
1504          * The driver for these cards can deliver a unique 16-bit
1505          * identifier for each instance sharing the same card.  It is
1506          * placed instead of 0xFFFE in the interface identifier.  The
1507          * "u" bit of the interface identifier is not inverted in this
1508          * case.  Hence the resulting interface identifier has local
1509          * scope according to RFC2373.
1510          */
1511         if (dev->dev_id) {
1512                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1513                 eui[4] = dev->dev_id & 0xFF;
1514         } else {
1515                 eui[3] = 0xFF;
1516                 eui[4] = 0xFE;
1517                 eui[0] ^= 2;
1518         }
1519         return 0;
1520 }
1521
1522 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1523 {
1524         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1525         if (dev->addr_len != ARCNET_ALEN)
1526                 return -1;
1527         memset(eui, 0, 7);
1528         eui[7] = *(u8*)dev->dev_addr;
1529         return 0;
1530 }
1531
1532 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1533 {
1534         if (dev->addr_len != INFINIBAND_ALEN)
1535                 return -1;
1536         memcpy(eui, dev->dev_addr + 12, 8);
1537         eui[0] |= 2;
1538         return 0;
1539 }
1540
1541 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1542 {
1543         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1544                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1545                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1546                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1547                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1548                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1549         eui[1] = 0;
1550         eui[2] = 0x5E;
1551         eui[3] = 0xFE;
1552         memcpy(eui + 4, &addr, 4);
1553         return 0;
1554 }
1555 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1556
1557 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1558 {
1559         if (dev->priv_flags & IFF_ISATAP)
1560                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1561         return -1;
1562 }
1563
1564 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1565 {
1566         switch (dev->type) {
1567         case ARPHRD_ETHER:
1568         case ARPHRD_FDDI:
1569         case ARPHRD_IEEE802_TR:
1570                 return addrconf_ifid_eui48(eui, dev);
1571         case ARPHRD_ARCNET:
1572                 return addrconf_ifid_arcnet(eui, dev);
1573         case ARPHRD_INFINIBAND:
1574                 return addrconf_ifid_infiniband(eui, dev);
1575         case ARPHRD_SIT:
1576                 return addrconf_ifid_sit(eui, dev);
1577         }
1578         return -1;
1579 }
1580
1581 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1582 {
1583         int err = -1;
1584         struct inet6_ifaddr *ifp;
1585
1586         read_lock_bh(&idev->lock);
1587         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1588                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1589                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1590                         err = 0;
1591                         break;
1592                 }
1593         }
1594         read_unlock_bh(&idev->lock);
1595         return err;
1596 }
1597
1598 #ifdef CONFIG_IPV6_PRIVACY
1599 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1600 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1601 {
1602 regen:
1603         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1604         idev->rndid[0] &= ~0x02;
1605
1606         /*
1607          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1608          * check if generated address is not inappropriate
1609          *
1610          *  - Reserved subnet anycast (RFC 2526)
1611          *      11111101 11....11 1xxxxxxx
1612          *  - ISATAP (RFC4214) 6.1
1613          *      00-00-5E-FE-xx-xx-xx-xx
1614          *  - value 0
1615          *  - XXX: already assigned to an address on the device
1616          */
1617         if (idev->rndid[0] == 0xfd &&
1618             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1619             (idev->rndid[7]&0x80))
1620                 goto regen;
1621         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1622                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1623                         goto regen;
1624                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1625                         goto regen;
1626         }
1627
1628         return 0;
1629 }
1630
1631 static void ipv6_regen_rndid(unsigned long data)
1632 {
1633         struct inet6_dev *idev = (struct inet6_dev *) data;
1634         unsigned long expires;
1635
1636         rcu_read_lock_bh();
1637         write_lock_bh(&idev->lock);
1638
1639         if (idev->dead)
1640                 goto out;
1641
1642         if (__ipv6_regen_rndid(idev) < 0)
1643                 goto out;
1644
1645         expires = jiffies +
1646                 idev->cnf.temp_prefered_lft * HZ -
1647                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1648         if (time_before(expires, jiffies)) {
1649                 printk(KERN_WARNING
1650                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1651                         idev->dev->name);
1652                 goto out;
1653         }
1654
1655         if (!mod_timer(&idev->regen_timer, expires))
1656                 in6_dev_hold(idev);
1657
1658 out:
1659         write_unlock_bh(&idev->lock);
1660         rcu_read_unlock_bh();
1661         in6_dev_put(idev);
1662 }
1663
1664 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1665         int ret = 0;
1666
1667         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1668                 ret = __ipv6_regen_rndid(idev);
1669         return ret;
1670 }
1671 #endif
1672
1673 /*
1674  *      Add prefix route.
1675  */
1676
1677 static void
1678 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1679                       unsigned long expires, u32 flags)
1680 {
1681         struct fib6_config cfg = {
1682                 .fc_table = RT6_TABLE_PREFIX,
1683                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1684                 .fc_ifindex = dev->ifindex,
1685                 .fc_expires = expires,
1686                 .fc_dst_len = plen,
1687                 .fc_flags = RTF_UP | flags,
1688                 .fc_nlinfo.nl_net = dev_net(dev),
1689         };
1690
1691         ipv6_addr_copy(&cfg.fc_dst, pfx);
1692
1693         /* Prevent useless cloning on PtP SIT.
1694            This thing is done here expecting that the whole
1695            class of non-broadcast devices need not cloning.
1696          */
1697 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1698         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1699                 cfg.fc_flags |= RTF_NONEXTHOP;
1700 #endif
1701
1702         ip6_route_add(&cfg);
1703 }
1704
1705 /* Create "default" multicast route to the interface */
1706
1707 static void addrconf_add_mroute(struct net_device *dev)
1708 {
1709         struct fib6_config cfg = {
1710                 .fc_table = RT6_TABLE_LOCAL,
1711                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1712                 .fc_ifindex = dev->ifindex,
1713                 .fc_dst_len = 8,
1714                 .fc_flags = RTF_UP,
1715                 .fc_nlinfo.nl_net = dev_net(dev),
1716         };
1717
1718         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1719
1720         ip6_route_add(&cfg);
1721 }
1722
1723 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1724 static void sit_route_add(struct net_device *dev)
1725 {
1726         struct fib6_config cfg = {
1727                 .fc_table = RT6_TABLE_MAIN,
1728                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1729                 .fc_ifindex = dev->ifindex,
1730                 .fc_dst_len = 96,
1731                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1732                 .fc_nlinfo.nl_net = dev_net(dev),
1733         };
1734
1735         /* prefix length - 96 bits "::d.d.d.d" */
1736         ip6_route_add(&cfg);
1737 }
1738 #endif
1739
1740 static void addrconf_add_lroute(struct net_device *dev)
1741 {
1742         struct in6_addr addr;
1743
1744         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1745         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1746 }
1747
1748 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1749 {
1750         struct inet6_dev *idev;
1751
1752         ASSERT_RTNL();
1753
1754         if ((idev = ipv6_find_idev(dev)) == NULL)
1755                 return NULL;
1756
1757         /* Add default multicast route */
1758         addrconf_add_mroute(dev);
1759
1760         /* Add link local route */
1761         addrconf_add_lroute(dev);
1762         return idev;
1763 }
1764
1765 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1766 {
1767         struct prefix_info *pinfo;
1768         __u32 valid_lft;
1769         __u32 prefered_lft;
1770         int addr_type;
1771         struct inet6_dev *in6_dev;
1772
1773         pinfo = (struct prefix_info *) opt;
1774
1775         if (len < sizeof(struct prefix_info)) {
1776                 ADBG(("addrconf: prefix option too short\n"));
1777                 return;
1778         }
1779
1780         /*
1781          *      Validation checks ([ADDRCONF], page 19)
1782          */
1783
1784         addr_type = ipv6_addr_type(&pinfo->prefix);
1785
1786         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1787                 return;
1788
1789         valid_lft = ntohl(pinfo->valid);
1790         prefered_lft = ntohl(pinfo->prefered);
1791
1792         if (prefered_lft > valid_lft) {
1793                 if (net_ratelimit())
1794                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1795                 return;
1796         }
1797
1798         in6_dev = in6_dev_get(dev);
1799
1800         if (in6_dev == NULL) {
1801                 if (net_ratelimit())
1802                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1803                 return;
1804         }
1805
1806         /*
1807          *      Two things going on here:
1808          *      1) Add routes for on-link prefixes
1809          *      2) Configure prefixes with the auto flag set
1810          */
1811
1812         if (pinfo->onlink) {
1813                 struct rt6_info *rt;
1814                 unsigned long rt_expires;
1815
1816                 /* Avoid arithmetic overflow. Really, we could
1817                  * save rt_expires in seconds, likely valid_lft,
1818                  * but it would require division in fib gc, that it
1819                  * not good.
1820                  */
1821                 if (HZ > USER_HZ)
1822                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1823                 else
1824                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1825
1826                 if (addrconf_finite_timeout(rt_expires))
1827                         rt_expires *= HZ;
1828
1829                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1830                                 dev->ifindex, 1);
1831
1832                 if (rt && addrconf_is_prefix_route(rt)) {
1833                         /* Autoconf prefix route */
1834                         if (valid_lft == 0) {
1835                                 ip6_del_rt(rt);
1836                                 rt = NULL;
1837                         } else if (addrconf_finite_timeout(rt_expires)) {
1838                                 /* not infinity */
1839                                 rt->rt6i_expires = jiffies + rt_expires;
1840                                 rt->rt6i_flags |= RTF_EXPIRES;
1841                         } else {
1842                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1843                                 rt->rt6i_expires = 0;
1844                         }
1845                 } else if (valid_lft) {
1846                         clock_t expires = 0;
1847                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1848                         if (addrconf_finite_timeout(rt_expires)) {
1849                                 /* not infinity */
1850                                 flags |= RTF_EXPIRES;
1851                                 expires = jiffies_to_clock_t(rt_expires);
1852                         }
1853                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1854                                               dev, expires, flags);
1855                 }
1856                 if (rt)
1857                         dst_release(&rt->u.dst);
1858         }
1859
1860         /* Try to figure out our local address for this prefix */
1861
1862         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1863                 struct inet6_ifaddr * ifp;
1864                 struct in6_addr addr;
1865                 int create = 0, update_lft = 0;
1866
1867                 if (pinfo->prefix_len == 64) {
1868                         memcpy(&addr, &pinfo->prefix, 8);
1869                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1870                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1871                                 in6_dev_put(in6_dev);
1872                                 return;
1873                         }
1874                         goto ok;
1875                 }
1876                 if (net_ratelimit())
1877                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1878                                pinfo->prefix_len);
1879                 in6_dev_put(in6_dev);
1880                 return;
1881
1882 ok:
1883
1884                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1885
1886                 if (ifp == NULL && valid_lft) {
1887                         int max_addresses = in6_dev->cnf.max_addresses;
1888                         u32 addr_flags = 0;
1889
1890 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1891                         if (in6_dev->cnf.optimistic_dad &&
1892                             !ipv6_devconf.forwarding)
1893                                 addr_flags = IFA_F_OPTIMISTIC;
1894 #endif
1895
1896                         /* Do not allow to create too much of autoconfigured
1897                          * addresses; this would be too easy way to crash kernel.
1898                          */
1899                         if (!max_addresses ||
1900                             ipv6_count_addresses(in6_dev) < max_addresses)
1901                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1902                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1903                                                     addr_flags);
1904
1905                         if (!ifp || IS_ERR(ifp)) {
1906                                 in6_dev_put(in6_dev);
1907                                 return;
1908                         }
1909
1910                         update_lft = create = 1;
1911                         ifp->cstamp = jiffies;
1912                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1913                 }
1914
1915                 if (ifp) {
1916                         int flags;
1917                         unsigned long now;
1918 #ifdef CONFIG_IPV6_PRIVACY
1919                         struct inet6_ifaddr *ift;
1920 #endif
1921                         u32 stored_lft;
1922
1923                         /* update lifetime (RFC2462 5.5.3 e) */
1924                         spin_lock(&ifp->lock);
1925                         now = jiffies;
1926                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1927                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1928                         else
1929                                 stored_lft = 0;
1930                         if (!update_lft && stored_lft) {
1931                                 if (valid_lft > MIN_VALID_LIFETIME ||
1932                                     valid_lft > stored_lft)
1933                                         update_lft = 1;
1934                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1935                                         /* valid_lft <= stored_lft is always true */
1936                                         /* XXX: IPsec */
1937                                         update_lft = 0;
1938                                 } else {
1939                                         valid_lft = MIN_VALID_LIFETIME;
1940                                         if (valid_lft < prefered_lft)
1941                                                 prefered_lft = valid_lft;
1942                                         update_lft = 1;
1943                                 }
1944                         }
1945
1946                         if (update_lft) {
1947                                 ifp->valid_lft = valid_lft;
1948                                 ifp->prefered_lft = prefered_lft;
1949                                 ifp->tstamp = now;
1950                                 flags = ifp->flags;
1951                                 ifp->flags &= ~IFA_F_DEPRECATED;
1952                                 spin_unlock(&ifp->lock);
1953
1954                                 if (!(flags&IFA_F_TENTATIVE))
1955                                         ipv6_ifa_notify(0, ifp);
1956                         } else
1957                                 spin_unlock(&ifp->lock);
1958
1959 #ifdef CONFIG_IPV6_PRIVACY
1960                         read_lock_bh(&in6_dev->lock);
1961                         /* update all temporary addresses in the list */
1962                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1963                                 /*
1964                                  * When adjusting the lifetimes of an existing
1965                                  * temporary address, only lower the lifetimes.
1966                                  * Implementations must not increase the
1967                                  * lifetimes of an existing temporary address
1968                                  * when processing a Prefix Information Option.
1969                                  */
1970                                 if (ifp != ift->ifpub)
1971                                         continue;
1972
1973                                 spin_lock(&ift->lock);
1974                                 flags = ift->flags;
1975                                 if (ift->valid_lft > valid_lft &&
1976                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1977                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1978                                 if (ift->prefered_lft > prefered_lft &&
1979                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1980                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1981                                 spin_unlock(&ift->lock);
1982                                 if (!(flags&IFA_F_TENTATIVE))
1983                                         ipv6_ifa_notify(0, ift);
1984                         }
1985
1986                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1987                                 /*
1988                                  * When a new public address is created as described in [ADDRCONF],
1989                                  * also create a new temporary address.
1990                                  */
1991                                 read_unlock_bh(&in6_dev->lock);
1992                                 ipv6_create_tempaddr(ifp, NULL);
1993                         } else {
1994                                 read_unlock_bh(&in6_dev->lock);
1995                         }
1996 #endif
1997                         in6_ifa_put(ifp);
1998                         addrconf_verify(0);
1999                 }
2000         }
2001         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2002         in6_dev_put(in6_dev);
2003 }
2004
2005 /*
2006  *      Set destination address.
2007  *      Special case for SIT interfaces where we create a new "virtual"
2008  *      device.
2009  */
2010 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2011 {
2012         struct in6_ifreq ireq;
2013         struct net_device *dev;
2014         int err = -EINVAL;
2015
2016         rtnl_lock();
2017
2018         err = -EFAULT;
2019         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2020                 goto err_exit;
2021
2022         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2023
2024         err = -ENODEV;
2025         if (dev == NULL)
2026                 goto err_exit;
2027
2028 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2029         if (dev->type == ARPHRD_SIT) {
2030                 struct ifreq ifr;
2031                 mm_segment_t    oldfs;
2032                 struct ip_tunnel_parm p;
2033
2034                 err = -EADDRNOTAVAIL;
2035                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2036                         goto err_exit;
2037
2038                 memset(&p, 0, sizeof(p));
2039                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2040                 p.iph.saddr = 0;
2041                 p.iph.version = 4;
2042                 p.iph.ihl = 5;
2043                 p.iph.protocol = IPPROTO_IPV6;
2044                 p.iph.ttl = 64;
2045                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2046
2047                 oldfs = get_fs(); set_fs(KERNEL_DS);
2048                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2049                 set_fs(oldfs);
2050
2051                 if (err == 0) {
2052                         err = -ENOBUFS;
2053                         dev = __dev_get_by_name(net, p.name);
2054                         if (!dev)
2055                                 goto err_exit;
2056                         err = dev_open(dev);
2057                 }
2058         }
2059 #endif
2060
2061 err_exit:
2062         rtnl_unlock();
2063         return err;
2064 }
2065
2066 /*
2067  *      Manual configuration of address on an interface
2068  */
2069 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2070                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2071                           __u32 valid_lft)
2072 {
2073         struct inet6_ifaddr *ifp;
2074         struct inet6_dev *idev;
2075         struct net_device *dev;
2076         int scope;
2077         u32 flags;
2078         clock_t expires;
2079         unsigned long timeout;
2080
2081         ASSERT_RTNL();
2082
2083         if (plen > 128)
2084                 return -EINVAL;
2085
2086         /* check the lifetime */
2087         if (!valid_lft || prefered_lft > valid_lft)
2088                 return -EINVAL;
2089
2090         dev = __dev_get_by_index(net, ifindex);
2091         if (!dev)
2092                 return -ENODEV;
2093
2094         if ((idev = addrconf_add_dev(dev)) == NULL)
2095                 return -ENOBUFS;
2096
2097         scope = ipv6_addr_scope(pfx);
2098
2099         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2100         if (addrconf_finite_timeout(timeout)) {
2101                 expires = jiffies_to_clock_t(timeout * HZ);
2102                 valid_lft = timeout;
2103                 flags = RTF_EXPIRES;
2104         } else {
2105                 expires = 0;
2106                 flags = 0;
2107                 ifa_flags |= IFA_F_PERMANENT;
2108         }
2109
2110         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2111         if (addrconf_finite_timeout(timeout)) {
2112                 if (timeout == 0)
2113                         ifa_flags |= IFA_F_DEPRECATED;
2114                 prefered_lft = timeout;
2115         }
2116
2117         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2118
2119         if (!IS_ERR(ifp)) {
2120                 spin_lock_bh(&ifp->lock);
2121                 ifp->valid_lft = valid_lft;
2122                 ifp->prefered_lft = prefered_lft;
2123                 ifp->tstamp = jiffies;
2124                 spin_unlock_bh(&ifp->lock);
2125
2126                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2127                                       expires, flags);
2128                 /*
2129                  * Note that section 3.1 of RFC 4429 indicates
2130                  * that the Optimistic flag should not be set for
2131                  * manually configured addresses
2132                  */
2133                 addrconf_dad_start(ifp, 0);
2134                 in6_ifa_put(ifp);
2135                 addrconf_verify(0);
2136                 return 0;
2137         }
2138
2139         return PTR_ERR(ifp);
2140 }
2141
2142 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2143                           unsigned int plen)
2144 {
2145         struct inet6_ifaddr *ifp;
2146         struct inet6_dev *idev;
2147         struct net_device *dev;
2148
2149         if (plen > 128)
2150                 return -EINVAL;
2151
2152         dev = __dev_get_by_index(net, ifindex);
2153         if (!dev)
2154                 return -ENODEV;
2155
2156         if ((idev = __in6_dev_get(dev)) == NULL)
2157                 return -ENXIO;
2158
2159         read_lock_bh(&idev->lock);
2160         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2161                 if (ifp->prefix_len == plen &&
2162                     ipv6_addr_equal(pfx, &ifp->addr)) {
2163                         in6_ifa_hold(ifp);
2164                         read_unlock_bh(&idev->lock);
2165
2166                         ipv6_del_addr(ifp);
2167
2168                         /* If the last address is deleted administratively,
2169                            disable IPv6 on this interface.
2170                          */
2171                         if (idev->addr_list == NULL)
2172                                 addrconf_ifdown(idev->dev, 1);
2173                         return 0;
2174                 }
2175         }
2176         read_unlock_bh(&idev->lock);
2177         return -EADDRNOTAVAIL;
2178 }
2179
2180
2181 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2182 {
2183         struct in6_ifreq ireq;
2184         int err;
2185
2186         if (!capable(CAP_NET_ADMIN))
2187                 return -EPERM;
2188
2189         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2190                 return -EFAULT;
2191
2192         rtnl_lock();
2193         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2194                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2195                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2196         rtnl_unlock();
2197         return err;
2198 }
2199
2200 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2201 {
2202         struct in6_ifreq ireq;
2203         int err;
2204
2205         if (!capable(CAP_NET_ADMIN))
2206                 return -EPERM;
2207
2208         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2209                 return -EFAULT;
2210
2211         rtnl_lock();
2212         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2213                              ireq.ifr6_prefixlen);
2214         rtnl_unlock();
2215         return err;
2216 }
2217
2218 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2219 static void sit_add_v4_addrs(struct inet6_dev *idev)
2220 {
2221         struct inet6_ifaddr * ifp;
2222         struct in6_addr addr;
2223         struct net_device *dev;
2224         struct net *net = dev_net(idev->dev);
2225         int scope;
2226
2227         ASSERT_RTNL();
2228
2229         memset(&addr, 0, sizeof(struct in6_addr));
2230         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2231
2232         if (idev->dev->flags&IFF_POINTOPOINT) {
2233                 addr.s6_addr32[0] = htonl(0xfe800000);
2234                 scope = IFA_LINK;
2235         } else {
2236                 scope = IPV6_ADDR_COMPATv4;
2237         }
2238
2239         if (addr.s6_addr32[3]) {
2240                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2241                 if (!IS_ERR(ifp)) {
2242                         spin_lock_bh(&ifp->lock);
2243                         ifp->flags &= ~IFA_F_TENTATIVE;
2244                         spin_unlock_bh(&ifp->lock);
2245                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2246                         in6_ifa_put(ifp);
2247                 }
2248                 return;
2249         }
2250
2251         for_each_netdev(net, dev) {
2252                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2253                 if (in_dev && (dev->flags & IFF_UP)) {
2254                         struct in_ifaddr * ifa;
2255
2256                         int flag = scope;
2257
2258                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2259                                 int plen;
2260
2261                                 addr.s6_addr32[3] = ifa->ifa_local;
2262
2263                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2264                                         continue;
2265                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2266                                         if (idev->dev->flags&IFF_POINTOPOINT)
2267                                                 continue;
2268                                         flag |= IFA_HOST;
2269                                 }
2270                                 if (idev->dev->flags&IFF_POINTOPOINT)
2271                                         plen = 64;
2272                                 else
2273                                         plen = 96;
2274
2275                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2276                                                     IFA_F_PERMANENT);
2277                                 if (!IS_ERR(ifp)) {
2278                                         spin_lock_bh(&ifp->lock);
2279                                         ifp->flags &= ~IFA_F_TENTATIVE;
2280                                         spin_unlock_bh(&ifp->lock);
2281                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2282                                         in6_ifa_put(ifp);
2283                                 }
2284                         }
2285                 }
2286         }
2287 }
2288 #endif
2289
2290 static void init_loopback(struct net_device *dev)
2291 {
2292         struct inet6_dev  *idev;
2293         struct inet6_ifaddr * ifp;
2294
2295         /* ::1 */
2296
2297         ASSERT_RTNL();
2298
2299         if ((idev = ipv6_find_idev(dev)) == NULL) {
2300                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2301                 return;
2302         }
2303
2304         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2305         if (!IS_ERR(ifp)) {
2306                 spin_lock_bh(&ifp->lock);
2307                 ifp->flags &= ~IFA_F_TENTATIVE;
2308                 spin_unlock_bh(&ifp->lock);
2309                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2310                 in6_ifa_put(ifp);
2311         }
2312 }
2313
2314 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2315 {
2316         struct inet6_ifaddr * ifp;
2317         u32 addr_flags = IFA_F_PERMANENT;
2318
2319 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2320         if (idev->cnf.optimistic_dad &&
2321             !ipv6_devconf.forwarding)
2322                 addr_flags |= IFA_F_OPTIMISTIC;
2323 #endif
2324
2325
2326         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2327         if (!IS_ERR(ifp)) {
2328                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2329                 addrconf_dad_start(ifp, 0);
2330                 in6_ifa_put(ifp);
2331         }
2332 }
2333
2334 static void addrconf_dev_config(struct net_device *dev)
2335 {
2336         struct in6_addr addr;
2337         struct inet6_dev    * idev;
2338
2339         ASSERT_RTNL();
2340
2341         if ((dev->type != ARPHRD_ETHER) &&
2342             (dev->type != ARPHRD_FDDI) &&
2343             (dev->type != ARPHRD_IEEE802_TR) &&
2344             (dev->type != ARPHRD_ARCNET) &&
2345             (dev->type != ARPHRD_INFINIBAND)) {
2346                 /* Alas, we support only Ethernet autoconfiguration. */
2347                 return;
2348         }
2349
2350         idev = addrconf_add_dev(dev);
2351         if (idev == NULL)
2352                 return;
2353
2354         memset(&addr, 0, sizeof(struct in6_addr));
2355         addr.s6_addr32[0] = htonl(0xFE800000);
2356
2357         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2358                 addrconf_add_linklocal(idev, &addr);
2359 }
2360
2361 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2362 static void addrconf_sit_config(struct net_device *dev)
2363 {
2364         struct inet6_dev *idev;
2365
2366         ASSERT_RTNL();
2367
2368         /*
2369          * Configure the tunnel with one of our IPv4
2370          * addresses... we should configure all of
2371          * our v4 addrs in the tunnel
2372          */
2373
2374         if ((idev = ipv6_find_idev(dev)) == NULL) {
2375                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2376                 return;
2377         }
2378
2379         if (dev->priv_flags & IFF_ISATAP) {
2380                 struct in6_addr addr;
2381
2382                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2383                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2384                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2385                         addrconf_add_linklocal(idev, &addr);
2386                 return;
2387         }
2388
2389         sit_add_v4_addrs(idev);
2390
2391         if (dev->flags&IFF_POINTOPOINT) {
2392                 addrconf_add_mroute(dev);
2393                 addrconf_add_lroute(dev);
2394         } else
2395                 sit_route_add(dev);
2396 }
2397 #endif
2398
2399 static inline int
2400 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2401 {
2402         struct in6_addr lladdr;
2403
2404         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2405                 addrconf_add_linklocal(idev, &lladdr);
2406                 return 0;
2407         }
2408         return -1;
2409 }
2410
2411 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2412 {
2413         struct net_device *link_dev;
2414         struct net *net = dev_net(idev->dev);
2415
2416         /* first try to inherit the link-local address from the link device */
2417         if (idev->dev->iflink &&
2418             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2419                 if (!ipv6_inherit_linklocal(idev, link_dev))
2420                         return;
2421         }
2422         /* then try to inherit it from any device */
2423         for_each_netdev(net, link_dev) {
2424                 if (!ipv6_inherit_linklocal(idev, link_dev))
2425                         return;
2426         }
2427         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2428 }
2429
2430 /*
2431  * Autoconfigure tunnel with a link-local address so routing protocols,
2432  * DHCPv6, MLD etc. can be run over the virtual link
2433  */
2434
2435 static void addrconf_ip6_tnl_config(struct net_device *dev)
2436 {
2437         struct inet6_dev *idev;
2438
2439         ASSERT_RTNL();
2440
2441         if ((idev = addrconf_add_dev(dev)) == NULL) {
2442                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2443                 return;
2444         }
2445         ip6_tnl_add_linklocal(idev);
2446 }
2447
2448 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2449                            void * data)
2450 {
2451         struct net_device *dev = (struct net_device *) data;
2452         struct inet6_dev *idev = __in6_dev_get(dev);
2453         int run_pending = 0;
2454         int err;
2455
2456         switch(event) {
2457         case NETDEV_REGISTER:
2458                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2459                         idev = ipv6_add_dev(dev);
2460                         if (!idev)
2461                                 return notifier_from_errno(-ENOMEM);
2462                 }
2463                 break;
2464         case NETDEV_UP:
2465         case NETDEV_CHANGE:
2466                 if (dev->flags & IFF_SLAVE)
2467                         break;
2468
2469                 if (event == NETDEV_UP) {
2470                         if (!addrconf_qdisc_ok(dev)) {
2471                                 /* device is not ready yet. */
2472                                 printk(KERN_INFO
2473                                         "ADDRCONF(NETDEV_UP): %s: "
2474                                         "link is not ready\n",
2475                                         dev->name);
2476                                 break;
2477                         }
2478
2479                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2480                                 idev = ipv6_add_dev(dev);
2481
2482                         if (idev)
2483                                 idev->if_flags |= IF_READY;
2484                 } else {
2485                         if (!addrconf_qdisc_ok(dev)) {
2486                                 /* device is still not ready. */
2487                                 break;
2488                         }
2489
2490                         if (idev) {
2491                                 if (idev->if_flags & IF_READY) {
2492                                         /* device is already configured. */
2493                                         break;
2494                                 }
2495                                 idev->if_flags |= IF_READY;
2496                         }
2497
2498                         printk(KERN_INFO
2499                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2500                                         "link becomes ready\n",
2501                                         dev->name);
2502
2503                         run_pending = 1;
2504                 }
2505
2506                 switch(dev->type) {
2507 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2508                 case ARPHRD_SIT:
2509                         addrconf_sit_config(dev);
2510                         break;
2511 #endif
2512                 case ARPHRD_TUNNEL6:
2513                         addrconf_ip6_tnl_config(dev);
2514                         break;
2515                 case ARPHRD_LOOPBACK:
2516                         init_loopback(dev);
2517                         break;
2518
2519                 default:
2520                         addrconf_dev_config(dev);
2521                         break;
2522                 }
2523                 if (idev) {
2524                         if (run_pending)
2525                                 addrconf_dad_run(idev);
2526
2527                         /* If the MTU changed during the interface down, when the
2528                            interface up, the changed MTU must be reflected in the
2529                            idev as well as routers.
2530                          */
2531                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2532                                 rt6_mtu_change(dev, dev->mtu);
2533                                 idev->cnf.mtu6 = dev->mtu;
2534                         }
2535                         idev->tstamp = jiffies;
2536                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2537                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2538                            stop IPv6 on this interface.
2539                          */
2540                         if (dev->mtu < IPV6_MIN_MTU)
2541                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2542                 }
2543                 break;
2544
2545         case NETDEV_CHANGEMTU:
2546                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2547                         rt6_mtu_change(dev, dev->mtu);
2548                         idev->cnf.mtu6 = dev->mtu;
2549                         break;
2550                 }
2551
2552                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2553                         idev = ipv6_add_dev(dev);
2554                         if (idev)
2555                                 break;
2556                 }
2557
2558                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2559
2560         case NETDEV_DOWN:
2561         case NETDEV_UNREGISTER:
2562                 /*
2563                  *      Remove all addresses from this interface.
2564                  */
2565                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2566                 break;
2567
2568         case NETDEV_CHANGENAME:
2569                 if (idev) {
2570                         snmp6_unregister_dev(idev);
2571                         addrconf_sysctl_unregister(idev);
2572                         addrconf_sysctl_register(idev);
2573                         err = snmp6_register_dev(idev);
2574                         if (err)
2575                                 return notifier_from_errno(err);
2576                 }
2577                 break;
2578         }
2579
2580         return NOTIFY_OK;
2581 }
2582
2583 /*
2584  *      addrconf module should be notified of a device going up
2585  */
2586 static struct notifier_block ipv6_dev_notf = {
2587         .notifier_call = addrconf_notify,
2588         .priority = 0
2589 };
2590
2591 static int addrconf_ifdown(struct net_device *dev, int how)
2592 {
2593         struct inet6_dev *idev;
2594         struct inet6_ifaddr *ifa, **bifa;
2595         struct net *net = dev_net(dev);
2596         int i;
2597
2598         ASSERT_RTNL();
2599
2600         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2601                 how = 0;
2602
2603         rt6_ifdown(net, dev);
2604         neigh_ifdown(&nd_tbl, dev);
2605
2606         idev = __in6_dev_get(dev);
2607         if (idev == NULL)
2608                 return -ENODEV;
2609
2610         /* Step 1: remove reference to ipv6 device from parent device.
2611                    Do not dev_put!
2612          */
2613         if (how) {
2614                 idev->dead = 1;
2615
2616                 /* protected by rtnl_lock */
2617                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2618
2619                 /* Step 1.5: remove snmp6 entry */
2620                 snmp6_unregister_dev(idev);
2621
2622         }
2623
2624         /* Step 2: clear hash table */
2625         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2626                 bifa = &inet6_addr_lst[i];
2627
2628                 write_lock_bh(&addrconf_hash_lock);
2629                 while ((ifa = *bifa) != NULL) {
2630                         if (ifa->idev == idev) {
2631                                 *bifa = ifa->lst_next;
2632                                 ifa->lst_next = NULL;
2633                                 addrconf_del_timer(ifa);
2634                                 in6_ifa_put(ifa);
2635                                 continue;
2636                         }
2637                         bifa = &ifa->lst_next;
2638                 }
2639                 write_unlock_bh(&addrconf_hash_lock);
2640         }
2641
2642         write_lock_bh(&idev->lock);
2643
2644         /* Step 3: clear flags for stateless addrconf */
2645         if (!how)
2646                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2647
2648         /* Step 4: clear address list */
2649 #ifdef CONFIG_IPV6_PRIVACY
2650         if (how && del_timer(&idev->regen_timer))
2651                 in6_dev_put(idev);
2652
2653         /* clear tempaddr list */
2654         while ((ifa = idev->tempaddr_list) != NULL) {
2655                 idev->tempaddr_list = ifa->tmp_next;
2656                 ifa->tmp_next = NULL;
2657                 ifa->dead = 1;
2658                 write_unlock_bh(&idev->lock);
2659                 spin_lock_bh(&ifa->lock);
2660
2661                 if (ifa->ifpub) {
2662                         in6_ifa_put(ifa->ifpub);
2663                         ifa->ifpub = NULL;
2664                 }
2665                 spin_unlock_bh(&ifa->lock);
2666                 in6_ifa_put(ifa);
2667                 write_lock_bh(&idev->lock);
2668         }
2669 #endif
2670         while ((ifa = idev->addr_list) != NULL) {
2671                 idev->addr_list = ifa->if_next;
2672                 ifa->if_next = NULL;
2673                 ifa->dead = 1;
2674                 addrconf_del_timer(ifa);
2675                 write_unlock_bh(&idev->lock);
2676
2677                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2678                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2679                 in6_ifa_put(ifa);
2680
2681                 write_lock_bh(&idev->lock);
2682         }
2683         write_unlock_bh(&idev->lock);
2684
2685         /* Step 5: Discard multicast list */
2686
2687         if (how)
2688                 ipv6_mc_destroy_dev(idev);
2689         else
2690                 ipv6_mc_down(idev);
2691
2692         idev->tstamp = jiffies;
2693
2694         /* Shot the device (if unregistered) */
2695
2696         if (how) {
2697                 addrconf_sysctl_unregister(idev);
2698                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2699                 neigh_ifdown(&nd_tbl, dev);
2700                 in6_dev_put(idev);
2701         }
2702         return 0;
2703 }
2704
2705 static void addrconf_rs_timer(unsigned long data)
2706 {
2707         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2708
2709         if (ifp->idev->cnf.forwarding)
2710                 goto out;
2711
2712         if (ifp->idev->if_flags & IF_RA_RCVD) {
2713                 /*
2714                  *      Announcement received after solicitation
2715                  *      was sent
2716                  */
2717                 goto out;
2718         }
2719
2720         spin_lock(&ifp->lock);
2721         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2722                 /* The wait after the last probe can be shorter */
2723                 addrconf_mod_timer(ifp, AC_RS,
2724                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2725                                    ifp->idev->cnf.rtr_solicit_delay :
2726                                    ifp->idev->cnf.rtr_solicit_interval);
2727                 spin_unlock(&ifp->lock);
2728
2729                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2730         } else {
2731                 spin_unlock(&ifp->lock);
2732                 /*
2733                  * Note: we do not support deprecated "all on-link"
2734                  * assumption any longer.
2735                  */
2736                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2737                        ifp->idev->dev->name);
2738         }
2739
2740 out:
2741         in6_ifa_put(ifp);
2742 }
2743
2744 /*
2745  *      Duplicate Address Detection
2746  */
2747 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2748 {
2749         unsigned long rand_num;
2750         struct inet6_dev *idev = ifp->idev;
2751
2752         if (ifp->flags & IFA_F_OPTIMISTIC)
2753                 rand_num = 0;
2754         else
2755                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2756
2757         ifp->probes = idev->cnf.dad_transmits;
2758         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2759 }
2760
2761 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2762 {
2763         struct inet6_dev *idev = ifp->idev;
2764         struct net_device *dev = idev->dev;
2765
2766         addrconf_join_solict(dev, &ifp->addr);
2767
2768         net_srandom(ifp->addr.s6_addr32[3]);
2769
2770         read_lock_bh(&idev->lock);
2771         if (ifp->dead)
2772                 goto out;
2773         spin_lock_bh(&ifp->lock);
2774
2775         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2776             idev->cnf.accept_dad < 1 ||
2777             !(ifp->flags&IFA_F_TENTATIVE) ||
2778             ifp->flags & IFA_F_NODAD) {
2779                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2780                 spin_unlock_bh(&ifp->lock);
2781                 read_unlock_bh(&idev->lock);
2782
2783                 addrconf_dad_completed(ifp);
2784                 return;
2785         }
2786
2787         if (!(idev->if_flags & IF_READY)) {
2788                 spin_unlock_bh(&ifp->lock);
2789                 read_unlock_bh(&idev->lock);
2790                 /*
2791                  * If the defice is not ready:
2792                  * - keep it tentative if it is a permanent address.
2793                  * - otherwise, kill it.
2794                  */
2795                 in6_ifa_hold(ifp);
2796                 addrconf_dad_stop(ifp);
2797                 return;
2798         }
2799
2800         /*
2801          * Optimistic nodes can start receiving
2802          * Frames right away
2803          */
2804         if(ifp->flags & IFA_F_OPTIMISTIC)
2805                 ip6_ins_rt(ifp->rt);
2806
2807         addrconf_dad_kick(ifp);
2808         spin_unlock_bh(&ifp->lock);
2809 out:
2810         read_unlock_bh(&idev->lock);
2811 }
2812
2813 static void addrconf_dad_timer(unsigned long data)
2814 {
2815         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2816         struct inet6_dev *idev = ifp->idev;
2817         struct in6_addr mcaddr;
2818
2819         read_lock_bh(&idev->lock);
2820         if (idev->dead) {
2821                 read_unlock_bh(&idev->lock);
2822                 goto out;
2823         }
2824         if (idev->cnf.accept_dad > 1 && idev->cnf.disable_ipv6) {
2825                 read_unlock_bh(&idev->lock);
2826                 addrconf_dad_failure(ifp);
2827                 return;
2828         }
2829         spin_lock_bh(&ifp->lock);
2830         if (ifp->probes == 0) {
2831                 /*
2832                  * DAD was successful
2833                  */
2834
2835                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2836                 spin_unlock_bh(&ifp->lock);
2837                 read_unlock_bh(&idev->lock);
2838
2839                 addrconf_dad_completed(ifp);
2840
2841                 goto out;
2842         }
2843
2844         ifp->probes--;
2845         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2846         spin_unlock_bh(&ifp->lock);
2847         read_unlock_bh(&idev->lock);
2848
2849         /* send a neighbour solicitation for our addr */
2850         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2851         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2852 out:
2853         in6_ifa_put(ifp);
2854 }
2855
2856 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2857 {
2858         struct net_device *     dev = ifp->idev->dev;
2859
2860         /*
2861          *      Configure the address for reception. Now it is valid.
2862          */
2863
2864         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2865
2866         /* If added prefix is link local and forwarding is off,
2867            start sending router solicitations.
2868          */
2869
2870         if (ifp->idev->cnf.forwarding == 0 &&
2871             ifp->idev->cnf.rtr_solicits > 0 &&
2872             (dev->flags&IFF_LOOPBACK) == 0 &&
2873             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2874                 /*
2875                  *      If a host as already performed a random delay
2876                  *      [...] as part of DAD [...] there is no need
2877                  *      to delay again before sending the first RS
2878                  */
2879                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2880
2881                 spin_lock_bh(&ifp->lock);
2882                 ifp->probes = 1;
2883                 ifp->idev->if_flags |= IF_RS_SENT;
2884                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2885                 spin_unlock_bh(&ifp->lock);
2886         }
2887 }
2888
2889 static void addrconf_dad_run(struct inet6_dev *idev) {
2890         struct inet6_ifaddr *ifp;
2891
2892         read_lock_bh(&idev->lock);
2893         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2894                 spin_lock_bh(&ifp->lock);
2895                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2896                         spin_unlock_bh(&ifp->lock);
2897                         continue;
2898                 }
2899                 spin_unlock_bh(&ifp->lock);
2900                 addrconf_dad_kick(ifp);
2901         }
2902         read_unlock_bh(&idev->lock);
2903 }
2904
2905 #ifdef CONFIG_PROC_FS
2906 struct if6_iter_state {
2907         struct seq_net_private p;
2908         int bucket;
2909 };
2910
2911 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2912 {
2913         struct inet6_ifaddr *ifa = NULL;
2914         struct if6_iter_state *state = seq->private;
2915         struct net *net = seq_file_net(seq);
2916
2917         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2918                 ifa = inet6_addr_lst[state->bucket];
2919
2920                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2921                         ifa = ifa->lst_next;
2922                 if (ifa)
2923                         break;
2924         }
2925         return ifa;
2926 }
2927
2928 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2929 {
2930         struct if6_iter_state *state = seq->private;
2931         struct net *net = seq_file_net(seq);
2932
2933         ifa = ifa->lst_next;
2934 try_again:
2935         if (ifa) {
2936                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2937                         ifa = ifa->lst_next;
2938                         goto try_again;
2939                 }
2940         }
2941
2942         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2943                 ifa = inet6_addr_lst[state->bucket];
2944                 goto try_again;
2945         }
2946
2947         return ifa;
2948 }
2949
2950 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2951 {
2952         struct inet6_ifaddr *ifa = if6_get_first(seq);
2953
2954         if (ifa)
2955                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2956                         --pos;
2957         return pos ? NULL : ifa;
2958 }
2959
2960 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2961         __acquires(addrconf_hash_lock)
2962 {
2963         read_lock_bh(&addrconf_hash_lock);
2964         return if6_get_idx(seq, *pos);
2965 }
2966
2967 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2968 {
2969         struct inet6_ifaddr *ifa;
2970
2971         ifa = if6_get_next(seq, v);
2972         ++*pos;
2973         return ifa;
2974 }
2975
2976 static void if6_seq_stop(struct seq_file *seq, void *v)
2977         __releases(addrconf_hash_lock)
2978 {
2979         read_unlock_bh(&addrconf_hash_lock);
2980 }
2981
2982 static int if6_seq_show(struct seq_file *seq, void *v)
2983 {
2984         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2985         seq_printf(seq,
2986                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2987                    NIP6(ifp->addr),
2988                    ifp->idev->dev->ifindex,
2989                    ifp->prefix_len,
2990                    ifp->scope,
2991                    ifp->flags,
2992                    ifp->idev->dev->name);
2993         return 0;
2994 }
2995
2996 static const struct seq_operations if6_seq_ops = {
2997         .start  = if6_seq_start,
2998         .next   = if6_seq_next,
2999         .show   = if6_seq_show,
3000         .stop   = if6_seq_stop,
3001 };
3002
3003 static int if6_seq_open(struct inode *inode, struct file *file)
3004 {
3005         return seq_open_net(inode, file, &if6_seq_ops,
3006                             sizeof(struct if6_iter_state));
3007 }
3008
3009 static const struct file_operations if6_fops = {
3010         .owner          = THIS_MODULE,
3011         .open           = if6_seq_open,
3012         .read           = seq_read,
3013         .llseek         = seq_lseek,
3014         .release        = seq_release_net,
3015 };
3016
3017 static int if6_proc_net_init(struct net *net)
3018 {
3019         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3020                 return -ENOMEM;
3021         return 0;
3022 }
3023
3024 static void if6_proc_net_exit(struct net *net)
3025 {
3026        proc_net_remove(net, "if_inet6");
3027 }
3028
3029 static struct pernet_operations if6_proc_net_ops = {
3030        .init = if6_proc_net_init,
3031        .exit = if6_proc_net_exit,
3032 };
3033
3034 int __init if6_proc_init(void)
3035 {
3036         return register_pernet_subsys(&if6_proc_net_ops);
3037 }
3038
3039 void if6_proc_exit(void)
3040 {
3041         unregister_pernet_subsys(&if6_proc_net_ops);
3042 }
3043 #endif  /* CONFIG_PROC_FS */
3044
3045 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3046 /* Check if address is a home address configured on any interface. */
3047 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3048 {
3049         int ret = 0;
3050         struct inet6_ifaddr * ifp;
3051         u8 hash = ipv6_addr_hash(addr);
3052         read_lock_bh(&addrconf_hash_lock);
3053         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3054                 if (!net_eq(dev_net(ifp->idev->dev), net))
3055                         continue;
3056                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3057                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3058                         ret = 1;
3059                         break;
3060                 }
3061         }
3062         read_unlock_bh(&addrconf_hash_lock);
3063         return ret;
3064 }
3065 #endif
3066
3067 /*
3068  *      Periodic address status verification
3069  */
3070
3071 static void addrconf_verify(unsigned long foo)
3072 {
3073         struct inet6_ifaddr *ifp;
3074         unsigned long now, next;
3075         int i;
3076
3077         spin_lock_bh(&addrconf_verify_lock);
3078         now = jiffies;
3079         next = now + ADDR_CHECK_FREQUENCY;
3080
3081         del_timer(&addr_chk_timer);
3082
3083         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3084
3085 restart:
3086                 read_lock(&addrconf_hash_lock);
3087                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3088                         unsigned long age;
3089 #ifdef CONFIG_IPV6_PRIVACY
3090                         unsigned long regen_advance;
3091 #endif
3092
3093                         if (ifp->flags & IFA_F_PERMANENT)
3094                                 continue;
3095
3096                         spin_lock(&ifp->lock);
3097                         age = (now - ifp->tstamp) / HZ;
3098
3099 #ifdef CONFIG_IPV6_PRIVACY
3100                         regen_advance = ifp->idev->cnf.regen_max_retry *
3101                                         ifp->idev->cnf.dad_transmits *
3102                                         ifp->idev->nd_parms->retrans_time / HZ;
3103 #endif
3104
3105                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3106                             age >= ifp->valid_lft) {
3107                                 spin_unlock(&ifp->lock);
3108                                 in6_ifa_hold(ifp);
3109                                 read_unlock(&addrconf_hash_lock);
3110                                 ipv6_del_addr(ifp);
3111                                 goto restart;
3112                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3113                                 spin_unlock(&ifp->lock);
3114                                 continue;
3115                         } else if (age >= ifp->prefered_lft) {
3116                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3117                                 int deprecate = 0;
3118
3119                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3120                                         deprecate = 1;
3121                                         ifp->flags |= IFA_F_DEPRECATED;
3122                                 }
3123
3124                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3125                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3126
3127                                 spin_unlock(&ifp->lock);
3128
3129                                 if (deprecate) {
3130                                         in6_ifa_hold(ifp);
3131                                         read_unlock(&addrconf_hash_lock);
3132
3133                                         ipv6_ifa_notify(0, ifp);
3134                                         in6_ifa_put(ifp);
3135                                         goto restart;
3136                                 }
3137 #ifdef CONFIG_IPV6_PRIVACY
3138                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3139                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3140                                 if (age >= ifp->prefered_lft - regen_advance) {
3141                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3142                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3143                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3144                                         if (!ifp->regen_count && ifpub) {
3145                                                 ifp->regen_count++;
3146                                                 in6_ifa_hold(ifp);
3147                                                 in6_ifa_hold(ifpub);
3148                                                 spin_unlock(&ifp->lock);
3149                                                 read_unlock(&addrconf_hash_lock);
3150                                                 spin_lock(&ifpub->lock);
3151                                                 ifpub->regen_count = 0;
3152                                                 spin_unlock(&ifpub->lock);
3153                                                 ipv6_create_tempaddr(ifpub, ifp);
3154                                                 in6_ifa_put(ifpub);
3155                                                 in6_ifa_put(ifp);
3156                                                 goto restart;
3157                                         }
3158                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3159                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3160                                 spin_unlock(&ifp->lock);
3161 #endif
3162                         } else {
3163                                 /* ifp->prefered_lft <= ifp->valid_lft */
3164                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3165                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3166                                 spin_unlock(&ifp->lock);
3167                         }
3168                 }
3169                 read_unlock(&addrconf_hash_lock);
3170         }
3171
3172         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3173         add_timer(&addr_chk_timer);
3174         spin_unlock_bh(&addrconf_verify_lock);
3175 }
3176
3177 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3178 {
3179         struct in6_addr *pfx = NULL;
3180
3181         if (addr)
3182                 pfx = nla_data(addr);
3183
3184         if (local) {
3185                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3186                         pfx = NULL;
3187                 else
3188                         pfx = nla_data(local);
3189         }
3190
3191         return pfx;
3192 }
3193
3194 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3195         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3196         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3197         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3198 };
3199
3200 static int
3201 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3202 {
3203         struct net *net = sock_net(skb->sk);
3204         struct ifaddrmsg *ifm;
3205         struct nlattr *tb[IFA_MAX+1];
3206         struct in6_addr *pfx;
3207         int err;
3208
3209         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3210         if (err < 0)
3211                 return err;
3212
3213         ifm = nlmsg_data(nlh);
3214         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3215         if (pfx == NULL)
3216                 return -EINVAL;
3217
3218         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3219 }
3220
3221 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3222                              u32 prefered_lft, u32 valid_lft)
3223 {
3224         u32 flags;
3225         clock_t expires;
3226         unsigned long timeout;
3227
3228         if (!valid_lft || (prefered_lft > valid_lft))
3229                 return -EINVAL;
3230
3231         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3232         if (addrconf_finite_timeout(timeout)) {
3233                 expires = jiffies_to_clock_t(timeout * HZ);
3234                 valid_lft = timeout;
3235                 flags = RTF_EXPIRES;
3236         } else {
3237                 expires = 0;
3238                 flags = 0;
3239                 ifa_flags |= IFA_F_PERMANENT;
3240         }
3241
3242         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3243         if (addrconf_finite_timeout(timeout)) {
3244                 if (timeout == 0)
3245                         ifa_flags |= IFA_F_DEPRECATED;
3246                 prefered_lft = timeout;
3247         }
3248
3249         spin_lock_bh(&ifp->lock);
3250         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3251         ifp->tstamp = jiffies;
3252         ifp->valid_lft = valid_lft;
3253         ifp->prefered_lft = prefered_lft;
3254
3255         spin_unlock_bh(&ifp->lock);
3256         if (!(ifp->flags&IFA_F_TENTATIVE))
3257                 ipv6_ifa_notify(0, ifp);
3258
3259         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3260                               expires, flags);
3261         addrconf_verify(0);
3262
3263         return 0;
3264 }
3265
3266 static int
3267 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3268 {
3269         struct net *net = sock_net(skb->sk);
3270         struct ifaddrmsg *ifm;
3271         struct nlattr *tb[IFA_MAX+1];
3272         struct in6_addr *pfx;
3273         struct inet6_ifaddr *ifa;
3274         struct net_device *dev;
3275         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3276         u8 ifa_flags;
3277         int err;
3278
3279         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3280         if (err < 0)
3281                 return err;
3282
3283         ifm = nlmsg_data(nlh);
3284         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3285         if (pfx == NULL)
3286                 return -EINVAL;
3287
3288         if (tb[IFA_CACHEINFO]) {
3289                 struct ifa_cacheinfo *ci;
3290
3291                 ci = nla_data(tb[IFA_CACHEINFO]);
3292                 valid_lft = ci->ifa_valid;
3293                 preferred_lft = ci->ifa_prefered;
3294         } else {
3295                 preferred_lft = INFINITY_LIFE_TIME;
3296                 valid_lft = INFINITY_LIFE_TIME;
3297         }
3298
3299         dev =  __dev_get_by_index(net, ifm->ifa_index);
3300         if (dev == NULL)
3301                 return -ENODEV;
3302
3303         /* We ignore other flags so far. */
3304         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3305
3306         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3307         if (ifa == NULL) {
3308                 /*
3309                  * It would be best to check for !NLM_F_CREATE here but
3310                  * userspace alreay relies on not having to provide this.
3311                  */
3312                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3313                                       ifm->ifa_prefixlen, ifa_flags,
3314                                       preferred_lft, valid_lft);
3315         }
3316
3317         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3318             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3319                 err = -EEXIST;
3320         else
3321                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3322
3323         in6_ifa_put(ifa);
3324
3325         return err;
3326 }
3327
3328 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3329                           u8 scope, int ifindex)
3330 {
3331         struct ifaddrmsg *ifm;
3332
3333         ifm = nlmsg_data(nlh);
3334         ifm->ifa_family = AF_INET6;
3335         ifm->ifa_prefixlen = prefixlen;
3336         ifm->ifa_flags = flags;
3337         ifm->ifa_scope = scope;
3338         ifm->ifa_index = ifindex;
3339 }
3340
3341 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3342                          unsigned long tstamp, u32 preferred, u32 valid)
3343 {
3344         struct ifa_cacheinfo ci;
3345
3346         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3347                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3348         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3349                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3350         ci.ifa_prefered = preferred;
3351         ci.ifa_valid = valid;
3352
3353         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3354 }
3355
3356 static inline int rt_scope(int ifa_scope)
3357 {
3358         if (ifa_scope & IFA_HOST)
3359                 return RT_SCOPE_HOST;
3360         else if (ifa_scope & IFA_LINK)
3361                 return RT_SCOPE_LINK;
3362         else if (ifa_scope & IFA_SITE)
3363                 return RT_SCOPE_SITE;
3364         else
3365                 return RT_SCOPE_UNIVERSE;
3366 }
3367
3368 static inline int inet6_ifaddr_msgsize(void)
3369 {
3370         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3371                + nla_total_size(16) /* IFA_ADDRESS */
3372                + nla_total_size(sizeof(struct ifa_cacheinfo));
3373 }
3374
3375 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3376                              u32 pid, u32 seq, int event, unsigned int flags)
3377 {
3378         struct nlmsghdr  *nlh;
3379         u32 preferred, valid;
3380
3381         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3382         if (nlh == NULL)
3383                 return -EMSGSIZE;
3384
3385         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3386                       ifa->idev->dev->ifindex);
3387
3388         if (!(ifa->flags&IFA_F_PERMANENT)) {
3389                 preferred = ifa->prefered_lft;
3390                 valid = ifa->valid_lft;
3391                 if (preferred != INFINITY_LIFE_TIME) {
3392                         long tval = (jiffies - ifa->tstamp)/HZ;
3393                         preferred -= tval;
3394                         if (valid != INFINITY_LIFE_TIME)
3395                                 valid -= tval;
3396                 }
3397         } else {
3398                 preferred = INFINITY_LIFE_TIME;
3399                 valid = INFINITY_LIFE_TIME;
3400         }
3401
3402         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3403             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3404                 nlmsg_cancel(skb, nlh);
3405                 return -EMSGSIZE;
3406         }
3407
3408         return nlmsg_end(skb, nlh);
3409 }
3410
3411 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3412                                 u32 pid, u32 seq, int event, u16 flags)
3413 {
3414         struct nlmsghdr  *nlh;
3415         u8 scope = RT_SCOPE_UNIVERSE;
3416         int ifindex = ifmca->idev->dev->ifindex;
3417
3418         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3419                 scope = RT_SCOPE_SITE;
3420
3421         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3422         if (nlh == NULL)
3423                 return -EMSGSIZE;
3424
3425         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3426         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3427             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3428                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3429                 nlmsg_cancel(skb, nlh);
3430                 return -EMSGSIZE;
3431         }
3432
3433         return nlmsg_end(skb, nlh);
3434 }
3435
3436 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3437                                 u32 pid, u32 seq, int event, unsigned int flags)
3438 {
3439         struct nlmsghdr  *nlh;
3440         u8 scope = RT_SCOPE_UNIVERSE;
3441         int ifindex = ifaca->aca_idev->dev->ifindex;
3442
3443         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3444                 scope = RT_SCOPE_SITE;
3445
3446         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3447         if (nlh == NULL)
3448                 return -EMSGSIZE;
3449
3450         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3451         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3452             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3453                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3454                 nlmsg_cancel(skb, nlh);
3455                 return -EMSGSIZE;
3456         }
3457
3458         return nlmsg_end(skb, nlh);
3459 }
3460
3461 enum addr_type_t
3462 {
3463         UNICAST_ADDR,
3464         MULTICAST_ADDR,
3465         ANYCAST_ADDR,
3466 };
3467
3468 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3469                            enum addr_type_t type)
3470 {
3471         int idx, ip_idx;
3472         int s_idx, s_ip_idx;
3473         int err = 1;
3474         struct net_device *dev;
3475         struct inet6_dev *idev = NULL;
3476         struct inet6_ifaddr *ifa;
3477         struct ifmcaddr6 *ifmca;
3478         struct ifacaddr6 *ifaca;
3479         struct net *net = sock_net(skb->sk);
3480
3481         s_idx = cb->args[0];
3482         s_ip_idx = ip_idx = cb->args[1];
3483
3484         idx = 0;
3485         for_each_netdev(net, dev) {
3486                 if (idx < s_idx)
3487                         goto cont;
3488                 if (idx > s_idx)
3489                         s_ip_idx = 0;
3490                 ip_idx = 0;
3491                 if ((idev = in6_dev_get(dev)) == NULL)
3492                         goto cont;
3493                 read_lock_bh(&idev->lock);
3494                 switch (type) {
3495                 case UNICAST_ADDR:
3496                         /* unicast address incl. temp addr */
3497                         for (ifa = idev->addr_list; ifa;
3498                              ifa = ifa->if_next, ip_idx++) {
3499                                 if (ip_idx < s_ip_idx)
3500                                         continue;
3501                                 err = inet6_fill_ifaddr(skb, ifa,
3502                                                         NETLINK_CB(cb->skb).pid,
3503                                                         cb->nlh->nlmsg_seq,
3504                                                         RTM_NEWADDR,
3505                                                         NLM_F_MULTI);
3506                         }
3507                         break;
3508                 case MULTICAST_ADDR:
3509                         /* multicast address */
3510                         for (ifmca = idev->mc_list; ifmca;
3511                              ifmca = ifmca->next, ip_idx++) {
3512                                 if (ip_idx < s_ip_idx)
3513                                         continue;
3514                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3515                                                           NETLINK_CB(cb->skb).pid,
3516                                                           cb->nlh->nlmsg_seq,
3517                                                           RTM_GETMULTICAST,
3518                                                           NLM_F_MULTI);
3519                         }
3520                         break;
3521                 case ANYCAST_ADDR:
3522                         /* anycast address */
3523                         for (ifaca = idev->ac_list; ifaca;
3524                              ifaca = ifaca->aca_next, ip_idx++) {
3525                                 if (ip_idx < s_ip_idx)
3526                                         continue;
3527                                 err = inet6_fill_ifacaddr(skb, ifaca,
3528                                                           NETLINK_CB(cb->skb).pid,
3529                                                           cb->nlh->nlmsg_seq,
3530                                                           RTM_GETANYCAST,
3531                                                           NLM_F_MULTI);
3532                         }
3533                         break;
3534                 default:
3535                         break;
3536                 }
3537                 read_unlock_bh(&idev->lock);
3538                 in6_dev_put(idev);
3539
3540                 if (err <= 0)
3541                         break;
3542 cont:
3543                 idx++;
3544         }
3545         cb->args[0] = idx;
3546         cb->args[1] = ip_idx;
3547         return skb->len;
3548 }
3549
3550 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3551 {
3552         enum addr_type_t type = UNICAST_ADDR;
3553
3554         return inet6_dump_addr(skb, cb, type);
3555 }
3556
3557 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3558 {
3559         enum addr_type_t type = MULTICAST_ADDR;
3560
3561         return inet6_dump_addr(skb, cb, type);
3562 }
3563
3564
3565 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3566 {
3567         enum addr_type_t type = ANYCAST_ADDR;
3568
3569         return inet6_dump_addr(skb, cb, type);
3570 }
3571
3572 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3573                              void *arg)
3574 {
3575         struct net *net = sock_net(in_skb->sk);
3576         struct ifaddrmsg *ifm;
3577         struct nlattr *tb[IFA_MAX+1];
3578         struct in6_addr *addr = NULL;
3579         struct net_device *dev = NULL;
3580         struct inet6_ifaddr *ifa;
3581         struct sk_buff *skb;
3582         int err;
3583
3584         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3585         if (err < 0)
3586                 goto errout;
3587
3588         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3589         if (addr == NULL) {
3590                 err = -EINVAL;
3591                 goto errout;
3592         }
3593
3594         ifm = nlmsg_data(nlh);
3595         if (ifm->ifa_index)
3596                 dev = __dev_get_by_index(net, ifm->ifa_index);
3597
3598         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3599                 err = -EADDRNOTAVAIL;
3600                 goto errout;
3601         }
3602
3603         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3604                 err = -ENOBUFS;
3605                 goto errout_ifa;
3606         }
3607
3608         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3609                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3610         if (err < 0) {
3611                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3612                 WARN_ON(err == -EMSGSIZE);
3613                 kfree_skb(skb);
3614                 goto errout_ifa;
3615         }
3616         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3617 errout_ifa:
3618         in6_ifa_put(ifa);
3619 errout:
3620         return err;
3621 }
3622
3623 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3624 {
3625         struct sk_buff *skb;
3626         struct net *net = dev_net(ifa->idev->dev);
3627         int err = -ENOBUFS;
3628
3629         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3630         if (skb == NULL)
3631                 goto errout;
3632
3633         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3634         if (err < 0) {
3635                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3636                 WARN_ON(err == -EMSGSIZE);
3637                 kfree_skb(skb);
3638                 goto errout;
3639         }
3640         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3641 errout:
3642         if (err < 0)
3643                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3644 }
3645
3646 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3647                                 __s32 *array, int bytes)
3648 {
3649         BUG_ON(bytes < (DEVCONF_MAX * 4));
3650
3651         memset(array, 0, bytes);
3652         array[DEVCONF_FORWARDING] = cnf->forwarding;
3653         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3654         array[DEVCONF_MTU6] = cnf->mtu6;
3655         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3656         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3657         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3658         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3659         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3660         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3661         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3662         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3663 #ifdef CONFIG_IPV6_PRIVACY
3664         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3665         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3666         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3667         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3668         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3669 #endif
3670         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3671         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3672         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3673 #ifdef CONFIG_IPV6_ROUTER_PREF
3674         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3675         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3676 #ifdef CONFIG_IPV6_ROUTE_INFO
3677         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3678 #endif
3679 #endif
3680         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3681         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3682 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3683         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3684 #endif
3685 #ifdef CONFIG_IPV6_MROUTE
3686         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3687 #endif
3688         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3689         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3690 }
3691
3692 static inline size_t inet6_if_nlmsg_size(void)
3693 {
3694         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3695                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3696                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3697                + nla_total_size(4) /* IFLA_MTU */
3698                + nla_total_size(4) /* IFLA_LINK */
3699                + nla_total_size( /* IFLA_PROTINFO */
3700                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3701                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3702                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3703                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3704                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3705                  );
3706 }
3707
3708 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3709                                       int bytes)
3710 {
3711         int i;
3712         int pad = bytes - sizeof(u64) * items;
3713         BUG_ON(pad < 0);
3714
3715         /* Use put_unaligned() because stats may not be aligned for u64. */
3716         put_unaligned(items, &stats[0]);
3717         for (i = 1; i < items; i++)
3718                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3719
3720         memset(&stats[items], 0, pad);
3721 }
3722
3723 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3724                              int bytes)
3725 {
3726         switch(attrtype) {
3727         case IFLA_INET6_STATS:
3728                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3729                 break;
3730         case IFLA_INET6_ICMP6STATS:
3731                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3732                 break;
3733         }
3734 }
3735
3736 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3737                              u32 pid, u32 seq, int event, unsigned int flags)
3738 {
3739         struct net_device *dev = idev->dev;
3740         struct nlattr *nla;
3741         struct ifinfomsg *hdr;
3742         struct nlmsghdr *nlh;
3743         void *protoinfo;
3744         struct ifla_cacheinfo ci;
3745
3746         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3747         if (nlh == NULL)
3748                 return -EMSGSIZE;
3749
3750         hdr = nlmsg_data(nlh);
3751         hdr->ifi_family = AF_INET6;
3752         hdr->__ifi_pad = 0;
3753         hdr->ifi_type = dev->type;
3754         hdr->ifi_index = dev->ifindex;
3755         hdr->ifi_flags = dev_get_flags(dev);
3756         hdr->ifi_change = 0;
3757
3758         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3759
3760         if (dev->addr_len)
3761                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3762
3763         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3764         if (dev->ifindex != dev->iflink)
3765                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3766
3767         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3768         if (protoinfo == NULL)
3769                 goto nla_put_failure;
3770
3771         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3772
3773         ci.max_reasm_len = IPV6_MAXPLEN;
3774         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3775                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3776         ci.reachable_time = idev->nd_parms->reachable_time;
3777         ci.retrans_time = idev->nd_parms->retrans_time;
3778         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3779
3780         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3781         if (nla == NULL)
3782                 goto nla_put_failure;
3783         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3784
3785         /* XXX - MC not implemented */
3786
3787         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3788         if (nla == NULL)
3789                 goto nla_put_failure;
3790         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3791
3792         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3793         if (nla == NULL)
3794                 goto nla_put_failure;
3795         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3796
3797         nla_nest_end(skb, protoinfo);
3798         return nlmsg_end(skb, nlh);
3799
3800 nla_put_failure:
3801         nlmsg_cancel(skb, nlh);
3802         return -EMSGSIZE;
3803 }
3804
3805 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3806 {
3807         struct net *net = sock_net(skb->sk);
3808         int idx, err;
3809         int s_idx = cb->args[0];
3810         struct net_device *dev;
3811         struct inet6_dev *idev;
3812
3813         read_lock(&dev_base_lock);
3814         idx = 0;
3815         for_each_netdev(net, dev) {
3816                 if (idx < s_idx)
3817                         goto cont;
3818                 if ((idev = in6_dev_get(dev)) == NULL)
3819                         goto cont;
3820                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3821                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3822                 in6_dev_put(idev);
3823                 if (err <= 0)
3824                         break;
3825 cont:
3826                 idx++;
3827         }
3828         read_unlock(&dev_base_lock);
3829         cb->args[0] = idx;
3830
3831         return skb->len;
3832 }
3833
3834 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3835 {
3836         struct sk_buff *skb;
3837         struct net *net = dev_net(idev->dev);
3838         int err = -ENOBUFS;
3839
3840         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3841         if (skb == NULL)
3842                 goto errout;
3843
3844         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3845         if (err < 0) {
3846                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3847                 WARN_ON(err == -EMSGSIZE);
3848                 kfree_skb(skb);
3849                 goto errout;
3850         }
3851         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3852 errout:
3853         if (err < 0)
3854                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3855 }
3856
3857 static inline size_t inet6_prefix_nlmsg_size(void)
3858 {
3859         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3860                + nla_total_size(sizeof(struct in6_addr))
3861                + nla_total_size(sizeof(struct prefix_cacheinfo));
3862 }
3863
3864 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3865                              struct prefix_info *pinfo, u32 pid, u32 seq,
3866                              int event, unsigned int flags)
3867 {
3868         struct prefixmsg *pmsg;
3869         struct nlmsghdr *nlh;
3870         struct prefix_cacheinfo ci;
3871
3872         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3873         if (nlh == NULL)
3874                 return -EMSGSIZE;
3875
3876         pmsg = nlmsg_data(nlh);
3877         pmsg->prefix_family = AF_INET6;
3878         pmsg->prefix_pad1 = 0;
3879         pmsg->prefix_pad2 = 0;
3880         pmsg->prefix_ifindex = idev->dev->ifindex;
3881         pmsg->prefix_len = pinfo->prefix_len;
3882         pmsg->prefix_type = pinfo->type;
3883         pmsg->prefix_pad3 = 0;
3884         pmsg->prefix_flags = 0;
3885         if (pinfo->onlink)
3886                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3887         if (pinfo->autoconf)
3888                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3889
3890         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3891
3892         ci.preferred_time = ntohl(pinfo->prefered);
3893         ci.valid_time = ntohl(pinfo->valid);
3894         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3895
3896         return nlmsg_end(skb, nlh);
3897
3898 nla_put_failure:
3899         nlmsg_cancel(skb, nlh);
3900         return -EMSGSIZE;
3901 }
3902
3903 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3904                          struct prefix_info *pinfo)
3905 {
3906         struct sk_buff *skb;
3907         struct net *net = dev_net(idev->dev);
3908         int err = -ENOBUFS;
3909
3910         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3911         if (skb == NULL)
3912                 goto errout;
3913
3914         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3915         if (err < 0) {
3916                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3917                 WARN_ON(err == -EMSGSIZE);
3918                 kfree_skb(skb);
3919                 goto errout;
3920         }
3921         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3922 errout:
3923         if (err < 0)
3924                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3925 }
3926
3927 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3928 {
3929         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3930
3931         switch (event) {
3932         case RTM_NEWADDR:
3933                 /*
3934                  * If the address was optimistic
3935                  * we inserted the route at the start of
3936                  * our DAD process, so we don't need
3937                  * to do it again
3938                  */
3939                 if (!(ifp->rt->rt6i_node))
3940                         ip6_ins_rt(ifp->rt);
3941                 if (ifp->idev->cnf.forwarding)
3942                         addrconf_join_anycast(ifp);
3943                 break;
3944         case RTM_DELADDR:
3945                 if (ifp->idev->cnf.forwarding)
3946                         addrconf_leave_anycast(ifp);
3947                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3948                 dst_hold(&ifp->rt->u.dst);
3949                 if (ip6_del_rt(ifp->rt))
3950                         dst_free(&ifp->rt->u.dst);
3951                 break;
3952         }
3953 }
3954
3955 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3956 {
3957         rcu_read_lock_bh();
3958         if (likely(ifp->idev->dead == 0))
3959                 __ipv6_ifa_notify(event, ifp);
3960         rcu_read_unlock_bh();
3961 }
3962
3963 #ifdef CONFIG_SYSCTL
3964
3965 static
3966 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3967                            void __user *buffer, size_t *lenp, loff_t *ppos)
3968 {
3969         int *valp = ctl->data;
3970         int val = *valp;
3971         int ret;
3972
3973         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3974
3975         if (write)
3976                 addrconf_fixup_forwarding(ctl, valp, val);
3977         return ret;
3978 }
3979
3980 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3981                                             int __user *name, int nlen,
3982                                             void __user *oldval,
3983                                             size_t __user *oldlenp,
3984                                             void __user *newval, size_t newlen)
3985 {
3986         int *valp = table->data;
3987         int val = *valp;
3988         int new;
3989
3990         if (!newval || !newlen)
3991                 return 0;
3992         if (newlen != sizeof(int))
3993                 return -EINVAL;
3994         if (get_user(new, (int __user *)newval))
3995                 return -EFAULT;
3996         if (new == *valp)
3997                 return 0;
3998         if (oldval && oldlenp) {
3999                 size_t len;
4000                 if (get_user(len, oldlenp))
4001                         return -EFAULT;
4002                 if (len) {
4003                         if (len > table->maxlen)
4004                                 len = table->maxlen;
4005                         if (copy_to_user(oldval, valp, len))
4006                                 return -EFAULT;
4007                         if (put_user(len, oldlenp))
4008                                 return -EFAULT;
4009                 }
4010         }
4011
4012         *valp = new;
4013         addrconf_fixup_forwarding(table, valp, val);
4014         return 1;
4015 }
4016
4017 static struct addrconf_sysctl_table
4018 {
4019         struct ctl_table_header *sysctl_header;
4020         ctl_table addrconf_vars[DEVCONF_MAX+1];
4021         char *dev_name;
4022 } addrconf_sysctl __read_mostly = {
4023         .sysctl_header = NULL,
4024         .addrconf_vars = {
4025                 {
4026                         .ctl_name       =       NET_IPV6_FORWARDING,
4027                         .procname       =       "forwarding",
4028                         .data           =       &ipv6_devconf.forwarding,
4029                         .maxlen         =       sizeof(int),
4030                         .mode           =       0644,
4031                         .proc_handler   =       &addrconf_sysctl_forward,
4032                         .strategy       =       &addrconf_sysctl_forward_strategy,
4033                 },
4034                 {
4035                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4036                         .procname       =       "hop_limit",
4037                         .data           =       &ipv6_devconf.hop_limit,
4038                         .maxlen         =       sizeof(int),
4039                         .mode           =       0644,
4040                         .proc_handler   =       proc_dointvec,
4041                 },
4042                 {
4043                         .ctl_name       =       NET_IPV6_MTU,
4044                         .procname       =       "mtu",
4045                         .data           =       &ipv6_devconf.mtu6,
4046                         .maxlen         =       sizeof(int),
4047                         .mode           =       0644,
4048                         .proc_handler   =       &proc_dointvec,
4049                 },
4050                 {
4051                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4052                         .procname       =       "accept_ra",
4053                         .data           =       &ipv6_devconf.accept_ra,
4054                         .maxlen         =       sizeof(int),
4055                         .mode           =       0644,
4056                         .proc_handler   =       &proc_dointvec,
4057                 },
4058                 {
4059                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4060                         .procname       =       "accept_redirects",
4061                         .data           =       &ipv6_devconf.accept_redirects,
4062                         .maxlen         =       sizeof(int),
4063                         .mode           =       0644,
4064                         .proc_handler   =       &proc_dointvec,
4065                 },
4066                 {
4067                         .ctl_name       =       NET_IPV6_AUTOCONF,
4068                         .procname       =       "autoconf",
4069                         .data           =       &ipv6_devconf.autoconf,
4070                         .maxlen         =       sizeof(int),
4071                         .mode           =       0644,
4072                         .proc_handler   =       &proc_dointvec,
4073                 },
4074                 {
4075                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4076                         .procname       =       "dad_transmits",
4077                         .data           =       &ipv6_devconf.dad_transmits,
4078                         .maxlen         =       sizeof(int),
4079                         .mode           =       0644,
4080                         .proc_handler   =       &proc_dointvec,
4081                 },
4082                 {
4083                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4084                         .procname       =       "router_solicitations",
4085                         .data           =       &ipv6_devconf.rtr_solicits,
4086                         .maxlen         =       sizeof(int),
4087                         .mode           =       0644,
4088                         .proc_handler   =       &proc_dointvec,
4089                 },
4090                 {
4091                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4092                         .procname       =       "router_solicitation_interval",
4093                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4094                         .maxlen         =       sizeof(int),
4095                         .mode           =       0644,
4096                         .proc_handler   =       &proc_dointvec_jiffies,
4097                         .strategy       =       &sysctl_jiffies,
4098                 },
4099                 {
4100                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4101                         .procname       =       "router_solicitation_delay",
4102                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4103                         .maxlen         =       sizeof(int),
4104                         .mode           =       0644,
4105                         .proc_handler   =       &proc_dointvec_jiffies,
4106                         .strategy       =       &sysctl_jiffies,
4107                 },
4108                 {
4109                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4110                         .procname       =       "force_mld_version",
4111                         .data           =       &ipv6_devconf.force_mld_version,
4112                         .maxlen         =       sizeof(int),
4113                         .mode           =       0644,
4114                         .proc_handler   =       &proc_dointvec,
4115                 },
4116 #ifdef CONFIG_IPV6_PRIVACY
4117                 {
4118                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4119                         .procname       =       "use_tempaddr",
4120                         .data           =       &ipv6_devconf.use_tempaddr,
4121                         .maxlen         =       sizeof(int),
4122                         .mode           =       0644,
4123                         .proc_handler   =       &proc_dointvec,
4124                 },
4125                 {
4126                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4127                         .procname       =       "temp_valid_lft",
4128                         .data           =       &ipv6_devconf.temp_valid_lft,
4129                         .maxlen         =       sizeof(int),
4130                         .mode           =       0644,
4131                         .proc_handler   =       &proc_dointvec,
4132                 },
4133                 {
4134                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4135                         .procname       =       "temp_prefered_lft",
4136                         .data           =       &ipv6_devconf.temp_prefered_lft,
4137                         .maxlen         =       sizeof(int),
4138                         .mode           =       0644,
4139                         .proc_handler   =       &proc_dointvec,
4140                 },
4141                 {
4142                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4143                         .procname       =       "regen_max_retry",
4144                         .data           =       &ipv6_devconf.regen_max_retry,
4145                         .maxlen         =       sizeof(int),
4146                         .mode           =       0644,
4147                         .proc_handler   =       &proc_dointvec,
4148                 },
4149                 {
4150                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4151                         .procname       =       "max_desync_factor",
4152                         .data           =       &ipv6_devconf.max_desync_factor,
4153                         .maxlen         =       sizeof(int),
4154                         .mode           =       0644,
4155                         .proc_handler   =       &proc_dointvec,
4156                 },
4157 #endif
4158                 {
4159                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4160                         .procname       =       "max_addresses",
4161                         .data           =       &ipv6_devconf.max_addresses,
4162                         .maxlen         =       sizeof(int),
4163                         .mode           =       0644,
4164                         .proc_handler   =       &proc_dointvec,
4165                 },
4166                 {
4167                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4168                         .procname       =       "accept_ra_defrtr",
4169                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4170                         .maxlen         =       sizeof(int),
4171                         .mode           =       0644,
4172                         .proc_handler   =       &proc_dointvec,
4173                 },
4174                 {
4175                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4176                         .procname       =       "accept_ra_pinfo",
4177                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4178                         .maxlen         =       sizeof(int),
4179                         .mode           =       0644,
4180                         .proc_handler   =       &proc_dointvec,
4181                 },
4182 #ifdef CONFIG_IPV6_ROUTER_PREF
4183                 {
4184                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4185                         .procname       =       "accept_ra_rtr_pref",
4186                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4187                         .maxlen         =       sizeof(int),
4188                         .mode           =       0644,
4189                         .proc_handler   =       &proc_dointvec,
4190                 },
4191                 {
4192                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4193                         .procname       =       "router_probe_interval",
4194                         .data           =       &ipv6_devconf.rtr_probe_interval,
4195                         .maxlen         =       sizeof(int),
4196                         .mode           =       0644,
4197                         .proc_handler   =       &proc_dointvec_jiffies,
4198                         .strategy       =       &sysctl_jiffies,
4199                 },
4200 #ifdef CONFIG_IPV6_ROUTE_INFO
4201                 {
4202                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4203                         .procname       =       "accept_ra_rt_info_max_plen",
4204                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4205                         .maxlen         =       sizeof(int),
4206                         .mode           =       0644,
4207                         .proc_handler   =       &proc_dointvec,
4208                 },
4209 #endif
4210 #endif
4211                 {
4212                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4213                         .procname       =       "proxy_ndp",
4214                         .data           =       &ipv6_devconf.proxy_ndp,
4215                         .maxlen         =       sizeof(int),
4216                         .mode           =       0644,
4217                         .proc_handler   =       &proc_dointvec,
4218                 },
4219                 {
4220                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4221                         .procname       =       "accept_source_route",
4222                         .data           =       &ipv6_devconf.accept_source_route,
4223                         .maxlen         =       sizeof(int),
4224                         .mode           =       0644,
4225                         .proc_handler   =       &proc_dointvec,
4226                 },
4227 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4228                 {
4229                         .ctl_name       =       CTL_UNNUMBERED,
4230                         .procname       =       "optimistic_dad",
4231                         .data           =       &ipv6_devconf.optimistic_dad,
4232                         .maxlen         =       sizeof(int),
4233                         .mode           =       0644,
4234                         .proc_handler   =       &proc_dointvec,
4235
4236                 },
4237 #endif
4238 #ifdef CONFIG_IPV6_MROUTE
4239                 {
4240                         .ctl_name       =       CTL_UNNUMBERED,
4241                         .procname       =       "mc_forwarding",
4242                         .data           =       &ipv6_devconf.mc_forwarding,
4243                         .maxlen         =       sizeof(int),
4244                         .mode           =       0644,
4245                         .proc_handler   =       &proc_dointvec,
4246                 },
4247 #endif
4248                 {
4249                         .ctl_name       =       CTL_UNNUMBERED,
4250                         .procname       =       "disable_ipv6",
4251                         .data           =       &ipv6_devconf.disable_ipv6,
4252                         .maxlen         =       sizeof(int),
4253                         .mode           =       0644,
4254                         .proc_handler   =       &proc_dointvec,
4255                 },
4256                 {
4257                         .ctl_name       =       CTL_UNNUMBERED,
4258                         .procname       =       "accept_dad",
4259                         .data           =       &ipv6_devconf.accept_dad,
4260                         .maxlen         =       sizeof(int),
4261                         .mode           =       0644,
4262                         .proc_handler   =       &proc_dointvec,
4263                 },
4264                 {
4265                         .ctl_name       =       0,      /* sentinel */
4266                 }
4267         },
4268 };
4269
4270 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4271                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4272 {
4273         int i;
4274         struct addrconf_sysctl_table *t;
4275
4276 #define ADDRCONF_CTL_PATH_DEV   3
4277
4278         struct ctl_path addrconf_ctl_path[] = {
4279                 { .procname = "net", .ctl_name = CTL_NET, },
4280                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4281                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4282                 { /* to be set */ },
4283                 { },
4284         };
4285
4286
4287         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4288         if (t == NULL)
4289                 goto out;
4290
4291         for (i=0; t->addrconf_vars[i].data; i++) {
4292                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4293                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4294                 t->addrconf_vars[i].extra2 = net;
4295         }
4296
4297         /*
4298          * Make a copy of dev_name, because '.procname' is regarded as const
4299          * by sysctl and we wouldn't want anyone to change it under our feet
4300          * (see SIOCSIFNAME).
4301          */
4302         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4303         if (!t->dev_name)
4304                 goto free;
4305
4306         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4307         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4308
4309         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4310                         t->addrconf_vars);
4311         if (t->sysctl_header == NULL)
4312                 goto free_procname;
4313
4314         p->sysctl = t;
4315         return 0;
4316
4317 free_procname:
4318         kfree(t->dev_name);
4319 free:
4320         kfree(t);
4321 out:
4322         return -ENOBUFS;
4323 }
4324
4325 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4326 {
4327         struct addrconf_sysctl_table *t;
4328
4329         if (p->sysctl == NULL)
4330                 return;
4331
4332         t = p->sysctl;
4333         p->sysctl = NULL;
4334         unregister_sysctl_table(t->sysctl_header);
4335         kfree(t->dev_name);
4336         kfree(t);
4337 }
4338
4339 static void addrconf_sysctl_register(struct inet6_dev *idev)
4340 {
4341         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4342                               NET_IPV6_NEIGH, "ipv6",
4343                               &ndisc_ifinfo_sysctl_change,
4344                               ndisc_ifinfo_sysctl_strategy);
4345         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4346                         idev->dev->ifindex, idev, &idev->cnf);
4347 }
4348
4349 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4350 {
4351         __addrconf_sysctl_unregister(&idev->cnf);
4352         neigh_sysctl_unregister(idev->nd_parms);
4353 }
4354
4355
4356 #endif
4357
4358 static int addrconf_init_net(struct net *net)
4359 {
4360         int err;
4361         struct ipv6_devconf *all, *dflt;
4362
4363         err = -ENOMEM;
4364         all = &ipv6_devconf;
4365         dflt = &ipv6_devconf_dflt;
4366
4367         if (net != &init_net) {
4368                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4369                 if (all == NULL)
4370                         goto err_alloc_all;
4371
4372                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4373                 if (dflt == NULL)
4374                         goto err_alloc_dflt;
4375         }
4376
4377         net->ipv6.devconf_all = all;
4378         net->ipv6.devconf_dflt = dflt;
4379
4380 #ifdef CONFIG_SYSCTL
4381         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4382                         NULL, all);
4383         if (err < 0)
4384                 goto err_reg_all;
4385
4386         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4387                         NULL, dflt);
4388         if (err < 0)
4389                 goto err_reg_dflt;
4390 #endif
4391         return 0;
4392
4393 #ifdef CONFIG_SYSCTL
4394 err_reg_dflt:
4395         __addrconf_sysctl_unregister(all);
4396 err_reg_all:
4397         kfree(dflt);
4398 #endif
4399 err_alloc_dflt:
4400         kfree(all);
4401 err_alloc_all:
4402         return err;
4403 }
4404
4405 static void addrconf_exit_net(struct net *net)
4406 {
4407 #ifdef CONFIG_SYSCTL
4408         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4409         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4410 #endif
4411         if (net != &init_net) {
4412                 kfree(net->ipv6.devconf_dflt);
4413                 kfree(net->ipv6.devconf_all);
4414         }
4415 }
4416
4417 static struct pernet_operations addrconf_ops = {
4418         .init = addrconf_init_net,
4419         .exit = addrconf_exit_net,
4420 };
4421
4422 /*
4423  *      Device notifier
4424  */
4425
4426 int register_inet6addr_notifier(struct notifier_block *nb)
4427 {
4428         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4429 }
4430
4431 EXPORT_SYMBOL(register_inet6addr_notifier);
4432
4433 int unregister_inet6addr_notifier(struct notifier_block *nb)
4434 {
4435         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4436 }
4437
4438 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4439
4440 static void addrconf_net_exit(struct net *net)
4441 {
4442         struct net_device *dev;
4443
4444         rtnl_lock();
4445         /* clean dev list */
4446         for_each_netdev(net, dev) {
4447                 if (__in6_dev_get(dev) == NULL)
4448                         continue;
4449                 addrconf_ifdown(dev, 1);
4450         }
4451         addrconf_ifdown(net->loopback_dev, 2);
4452         rtnl_unlock();
4453 }
4454
4455 static struct pernet_operations addrconf_net_ops = {
4456         .exit = addrconf_net_exit,
4457 };
4458
4459 /*
4460  *      Init / cleanup code
4461  */
4462
4463 int __init addrconf_init(void)
4464 {
4465         int err;
4466
4467         if ((err = ipv6_addr_label_init()) < 0) {
4468                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4469                         err);
4470                 return err;
4471         }
4472
4473         register_pernet_subsys(&addrconf_ops);
4474
4475         /* The addrconf netdev notifier requires that loopback_dev
4476          * has it's ipv6 private information allocated and setup
4477          * before it can bring up and give link-local addresses
4478          * to other devices which are up.
4479          *
4480          * Unfortunately, loopback_dev is not necessarily the first
4481          * entry in the global dev_base list of net devices.  In fact,
4482          * it is likely to be the very last entry on that list.
4483          * So this causes the notifier registry below to try and
4484          * give link-local addresses to all devices besides loopback_dev
4485          * first, then loopback_dev, which cases all the non-loopback_dev
4486          * devices to fail to get a link-local address.
4487          *
4488          * So, as a temporary fix, allocate the ipv6 structure for
4489          * loopback_dev first by hand.
4490          * Longer term, all of the dependencies ipv6 has upon the loopback
4491          * device and it being up should be removed.
4492          */
4493         rtnl_lock();
4494         if (!ipv6_add_dev(init_net.loopback_dev))
4495                 err = -ENOMEM;
4496         rtnl_unlock();
4497         if (err)
4498                 goto errlo;
4499
4500         err = register_pernet_device(&addrconf_net_ops);
4501         if (err)
4502                 return err;
4503
4504         register_netdevice_notifier(&ipv6_dev_notf);
4505
4506         addrconf_verify(0);
4507
4508         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4509         if (err < 0)
4510                 goto errout;
4511
4512         /* Only the first call to __rtnl_register can fail */
4513         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4514         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4515         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4516         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4517         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4518
4519         ipv6_addr_label_rtnl_register();
4520
4521         return 0;
4522 errout:
4523         unregister_netdevice_notifier(&ipv6_dev_notf);
4524 errlo:
4525         unregister_pernet_subsys(&addrconf_ops);
4526
4527         return err;
4528 }
4529
4530 void addrconf_cleanup(void)
4531 {
4532         struct inet6_ifaddr *ifa;
4533         int i;
4534
4535         unregister_netdevice_notifier(&ipv6_dev_notf);
4536         unregister_pernet_device(&addrconf_net_ops);
4537
4538         unregister_pernet_subsys(&addrconf_ops);
4539
4540         rtnl_lock();
4541
4542         /*
4543          *      Check hash table.
4544          */
4545         write_lock_bh(&addrconf_hash_lock);
4546         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4547                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4548                         struct inet6_ifaddr *bifa;
4549
4550                         bifa = ifa;
4551                         ifa = ifa->lst_next;
4552                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4553                         /* Do not free it; something is wrong.
4554                            Now we can investigate it with debugger.
4555                          */
4556                 }
4557         }
4558         write_unlock_bh(&addrconf_hash_lock);
4559
4560         del_timer(&addr_chk_timer);
4561         rtnl_unlock();
4562
4563         unregister_pernet_subsys(&addrconf_net_ops);
4564 }