net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 110
 111 #ifdef CONFIG_SYSCTL
 112 static void addrconf_sysctl_register(struct inet6_dev *idev);
 113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 114 #else
 115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 116 {
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 #ifdef CONFIG_IPV6_PRIVACY
 125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 127 static void ipv6_regen_rndid(unsigned long data);
 128 #endif
 129
 130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 131 static int ipv6_count_addresses(struct inet6_dev *idev);
 132
 133 /*
 134  *      Configured unicast address hash table
 135  */
 136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 137 static DEFINE_SPINLOCK(addrconf_hash_lock);
 138
 139 static void addrconf_verify(unsigned long);
 140
 141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 142 static DEFINE_SPINLOCK(addrconf_verify_lock);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 152 static void addrconf_dad_timer(unsigned long data);
 153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_run(struct inet6_dev *idev);
 155 static void addrconf_rs_timer(unsigned long data);
 156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158
 159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 160                                 struct prefix_info *pinfo);
 161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 162                                struct net_device *dev);
 163
 164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .dad_transmits          = 1,
 175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 178 #ifdef CONFIG_IPV6_PRIVACY
 179         .use_tempaddr           = 0,
 180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 182         .regen_max_retry        = REGEN_MAX_RETRY,
 183         .max_desync_factor      = MAX_DESYNC_FACTOR,
 184 #endif
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199 };
 200
 201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 202         .forwarding             = 0,
 203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 204         .mtu6                   = IPV6_MIN_MTU,
 205         .accept_ra              = 1,
 206         .accept_redirects       = 1,
 207         .autoconf               = 1,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212 #ifdef CONFIG_IPV6_PRIVACY
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218 #endif
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233 };
 234
 235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 240
 241 /* Check if a valid qdisc is available */
 242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 243 {
 244         return !qdisc_tx_is_noop(dev);
 245 }
 246
 247 /* Check if a route is valid prefix route */
 248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 249 {
 250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 251 }
 252
 253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 254 {
 255         if (del_timer(&ifp->timer))
 256                 __in6_ifa_put(ifp);
 257 }
 258
 259 enum addrconf_timer_t {
 260         AC_NONE,
 261         AC_DAD,
 262         AC_RS,
 263 };
 264
 265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 266                                enum addrconf_timer_t what,
 267                                unsigned long when)
 268 {
 269         if (!del_timer(&ifp->timer))
 270                 in6_ifa_hold(ifp);
 271
 272         switch (what) {
 273         case AC_DAD:
 274                 ifp->timer.function = addrconf_dad_timer;
 275                 break;
 276         case AC_RS:
 277                 ifp->timer.function = addrconf_rs_timer;
 278                 break;
 279         default:
 280                 break;
 281         }
 282         ifp->timer.expires = jiffies + when;
 283         add_timer(&ifp->timer);
 284 }
 285
 286 static int snmp6_alloc_dev(struct inet6_dev *idev)
 287 {
 288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 289                           sizeof(struct ipstats_mib),
 290                           __alignof__(struct ipstats_mib)) < 0)
 291                 goto err_ip;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 293                           sizeof(struct icmpv6_mib),
 294                           __alignof__(struct icmpv6_mib)) < 0)
 295                 goto err_icmp;
 296         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 297                           sizeof(struct icmpv6msg_mib),
 298                           __alignof__(struct icmpv6msg_mib)) < 0)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static void snmp6_free_dev(struct inet6_dev *idev)
 312 {
 313         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 314         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 321 {
 322         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 323         kfree(idev);
 324 }
 325
 326 void in6_dev_finish_destroy(struct inet6_dev *idev)
 327 {
 328         struct net_device *dev = idev->dev;
 329
 330         WARN_ON(!list_empty(&idev->addr_list));
 331         WARN_ON(idev->mc_list != NULL);
 332
 333 #ifdef NET_REFCNT_DEBUG
 334         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 335 #endif
 336         dev_put(dev);
 337         if (!idev->dead) {
 338                 pr_warning("Freeing alive inet6 device %p\n", idev);
 339                 return;
 340         }
 341         snmp6_free_dev(idev);
 342         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 343 }
 344
 345 EXPORT_SYMBOL(in6_dev_finish_destroy);
 346
 347 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 348 {
 349         struct inet6_dev *ndev;
 350
 351         ASSERT_RTNL();
 352
 353         if (dev->mtu < IPV6_MIN_MTU)
 354                 return NULL;
 355
 356         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 357
 358         if (ndev == NULL)
 359                 return NULL;
 360
 361         rwlock_init(&ndev->lock);
 362         ndev->dev = dev;
 363         INIT_LIST_HEAD(&ndev->addr_list);
 364
 365         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 366         ndev->cnf.mtu6 = dev->mtu;
 367         ndev->cnf.sysctl = NULL;
 368         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 369         if (ndev->nd_parms == NULL) {
 370                 kfree(ndev);
 371                 return NULL;
 372         }
 373         if (ndev->cnf.forwarding)
 374                 dev_disable_lro(dev);
 375         /* We refer to the device */
 376         dev_hold(dev);
 377
 378         if (snmp6_alloc_dev(ndev) < 0) {
 379                 ADBG((KERN_WARNING
 380                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 381                         __func__, dev->name));
 382                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 383                 ndev->dead = 1;
 384                 in6_dev_finish_destroy(ndev);
 385                 return NULL;
 386         }
 387
 388         if (snmp6_register_dev(ndev) < 0) {
 389                 ADBG((KERN_WARNING
 390                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 391                         __func__, dev->name));
 392                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 393                 ndev->dead = 1;
 394                 in6_dev_finish_destroy(ndev);
 395                 return NULL;
 396         }
 397
 398         /* One reference from device.  We must do this before
 399          * we invoke __ipv6_regen_rndid().
 400          */
 401         in6_dev_hold(ndev);
 402
 403         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 404                 ndev->cnf.accept_dad = -1;
 405
 406 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 407         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Multicast RS\n",
 410                        dev->name);
 411                 ndev->cnf.rtr_solicits = 0;
 412         }
 413 #endif
 414
 415 #ifdef CONFIG_IPV6_PRIVACY
 416         INIT_LIST_HEAD(&ndev->tempaddr_list);
 417         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 418         if ((dev->flags&IFF_LOOPBACK) ||
 419             dev->type == ARPHRD_TUNNEL ||
 420             dev->type == ARPHRD_TUNNEL6 ||
 421             dev->type == ARPHRD_SIT ||
 422             dev->type == ARPHRD_NONE) {
 423                 printk(KERN_INFO
 424                        "%s: Disabled Privacy Extensions\n",
 425                        dev->name);
 426                 ndev->cnf.use_tempaddr = -1;
 427         } else {
 428                 in6_dev_hold(ndev);
 429                 ipv6_regen_rndid((unsigned long) ndev);
 430         }
 431 #endif
 432
 433         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 434                 ndev->if_flags |= IF_READY;
 435
 436         ipv6_mc_init_dev(ndev);
 437         ndev->tstamp = jiffies;
 438         addrconf_sysctl_register(ndev);
 439         /* protected by rtnl_lock */
 440         rcu_assign_pointer(dev->ip6_ptr, ndev);
 441
 442         /* Join all-node multicast group */
 443         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 444
 445         return ndev;
 446 }
 447
 448 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 449 {
 450         struct inet6_dev *idev;
 451
 452         ASSERT_RTNL();
 453
 454         idev = __in6_dev_get(dev);
 455         if (!idev) {
 456                 idev = ipv6_add_dev(dev);
 457                 if (!idev)
 458                         return NULL;
 459         }
 460
 461         if (dev->flags&IFF_UP)
 462                 ipv6_mc_up(idev);
 463         return idev;
 464 }
 465
 466 #ifdef CONFIG_SYSCTL
 467 static void dev_forward_change(struct inet6_dev *idev)
 468 {
 469         struct net_device *dev;
 470         struct inet6_ifaddr *ifa;
 471
 472         if (!idev)
 473                 return;
 474         dev = idev->dev;
 475         if (idev->cnf.forwarding)
 476                 dev_disable_lro(dev);
 477         if (dev && (dev->flags & IFF_MULTICAST)) {
 478                 if (idev->cnf.forwarding)
 479                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 480                 else
 481                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 482         }
 483
 484         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 485                 if (ifa->flags&IFA_F_TENTATIVE)
 486                         continue;
 487                 if (idev->cnf.forwarding)
 488                         addrconf_join_anycast(ifa);
 489                 else
 490                         addrconf_leave_anycast(ifa);
 491         }
 492 }
 493
 494
 495 static void addrconf_forward_change(struct net *net, __s32 newf)
 496 {
 497         struct net_device *dev;
 498         struct inet6_dev *idev;
 499
 500         rcu_read_lock();
 501         for_each_netdev_rcu(net, dev) {
 502                 idev = __in6_dev_get(dev);
 503                 if (idev) {
 504                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 505                         idev->cnf.forwarding = newf;
 506                         if (changed)
 507                                 dev_forward_change(idev);
 508                 }
 509         }
 510         rcu_read_unlock();
 511 }
 512
 513 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 514 {
 515         struct net *net;
 516
 517         net = (struct net *)table->extra2;
 518         if (p == &net->ipv6.devconf_dflt->forwarding)
 519                 return 0;
 520
 521         if (!rtnl_trylock()) {
 522                 /* Restore the original values before restarting */
 523                 *p = old;
 524                 return restart_syscall();
 525         }
 526
 527         if (p == &net->ipv6.devconf_all->forwarding) {
 528                 __s32 newf = net->ipv6.devconf_all->forwarding;
 529                 net->ipv6.devconf_dflt->forwarding = newf;
 530                 addrconf_forward_change(net, newf);
 531         } else if ((!*p) ^ (!old))
 532                 dev_forward_change((struct inet6_dev *)table->extra1);
 533         rtnl_unlock();
 534
 535         if (*p)
 536                 rt6_purge_dflt_routers(net);
 537         return 1;
 538 }
 539 #endif
 540
 541 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 542 {
 543         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 544         kfree(ifp);
 545 }
 546
 547 /* Nobody refers to this ifaddr, destroy it */
 548 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 549 {
 550         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 551
 552 #ifdef NET_REFCNT_DEBUG
 553         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 554 #endif
 555
 556         in6_dev_put(ifp->idev);
 557
 558         if (del_timer(&ifp->timer))
 559                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 560
 561         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 562                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 563                 return;
 564         }
 565         dst_release(&ifp->rt->dst);
 566
 567         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 568 }
 569
 570 static void
 571 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 572 {
 573         struct list_head *p;
 574         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 575
 576         /*
 577          * Each device address list is sorted in order of scope -
 578          * global before linklocal.
 579          */
 580         list_for_each(p, &idev->addr_list) {
 581                 struct inet6_ifaddr *ifa
 582                         = list_entry(p, struct inet6_ifaddr, if_list);
 583                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 584                         break;
 585         }
 586
 587         list_add_tail(&ifp->if_list, p);
 588 }
 589
 590 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 591 {
 592         /*
 593          * We perform the hash function over the last 64 bits of the address
 594          * This will include the IEEE address token on links that support it.
 595          */
 596         return jhash_2words((__force u32)addr->s6_addr32[2],
 597                             (__force u32)addr->s6_addr32[3], 0)
 598                 & (IN6_ADDR_HSIZE - 1);
 599 }
 600
 601 /* On success it returns ifp with increased reference count */
 602
 603 static struct inet6_ifaddr *
 604 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 605               int scope, u32 flags)
 606 {
 607         struct inet6_ifaddr *ifa = NULL;
 608         struct rt6_info *rt;
 609         unsigned int hash;
 610         int err = 0;
 611         int addr_type = ipv6_addr_type(addr);
 612
 613         if (addr_type == IPV6_ADDR_ANY ||
 614             addr_type & IPV6_ADDR_MULTICAST ||
 615             (!(idev->dev->flags & IFF_LOOPBACK) &&
 616              addr_type & IPV6_ADDR_LOOPBACK))
 617                 return ERR_PTR(-EADDRNOTAVAIL);
 618
 619         rcu_read_lock_bh();
 620         if (idev->dead) {
 621                 err = -ENODEV;                  /*XXX*/
 622                 goto out2;
 623         }
 624
 625         if (idev->cnf.disable_ipv6) {
 626                 err = -EACCES;
 627                 goto out2;
 628         }
 629
 630         spin_lock(&addrconf_hash_lock);
 631
 632         /* Ignore adding duplicate addresses on an interface */
 633         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 634                 ADBG(("ipv6_add_addr: already assigned\n"));
 635                 err = -EEXIST;
 636                 goto out;
 637         }
 638
 639         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 640
 641         if (ifa == NULL) {
 642                 ADBG(("ipv6_add_addr: malloc failed\n"));
 643                 err = -ENOBUFS;
 644                 goto out;
 645         }
 646
 647         rt = addrconf_dst_alloc(idev, addr, 0);
 648         if (IS_ERR(rt)) {
 649                 err = PTR_ERR(rt);
 650                 goto out;
 651         }
 652
 653         ipv6_addr_copy(&ifa->addr, addr);
 654
 655         spin_lock_init(&ifa->lock);
 656         spin_lock_init(&ifa->state_lock);
 657         init_timer(&ifa->timer);
 658         INIT_HLIST_NODE(&ifa->addr_lst);
 659         ifa->timer.data = (unsigned long) ifa;
 660         ifa->scope = scope;
 661         ifa->prefix_len = pfxlen;
 662         ifa->flags = flags | IFA_F_TENTATIVE;
 663         ifa->cstamp = ifa->tstamp = jiffies;
 664
 665         ifa->rt = rt;
 666
 667         /*
 668          * part one of RFC 4429, section 3.3
 669          * We should not configure an address as
 670          * optimistic if we do not yet know the link
 671          * layer address of our nexhop router
 672          */
 673
 674         if (rt->rt6i_nexthop == NULL)
 675                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 676
 677         ifa->idev = idev;
 678         in6_dev_hold(idev);
 679         /* For caller */
 680         in6_ifa_hold(ifa);
 681
 682         /* Add to big hash table */
 683         hash = ipv6_addr_hash(addr);
 684
 685         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 686         spin_unlock(&addrconf_hash_lock);
 687
 688         write_lock(&idev->lock);
 689         /* Add to inet6_dev unicast addr list. */
 690         ipv6_link_dev_addr(idev, ifa);
 691
 692 #ifdef CONFIG_IPV6_PRIVACY
 693         if (ifa->flags&IFA_F_TEMPORARY) {
 694                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 695                 in6_ifa_hold(ifa);
 696         }
 697 #endif
 698
 699         in6_ifa_hold(ifa);
 700         write_unlock(&idev->lock);
 701 out2:
 702         rcu_read_unlock_bh();
 703
 704         if (likely(err == 0))
 705                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 706         else {
 707                 kfree(ifa);
 708                 ifa = ERR_PTR(err);
 709         }
 710
 711         return ifa;
 712 out:
 713         spin_unlock(&addrconf_hash_lock);
 714         goto out2;
 715 }
 716
 717 /* This function wants to get referenced ifp and releases it before return */
 718
 719 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 720 {
 721         struct inet6_ifaddr *ifa, *ifn;
 722         struct inet6_dev *idev = ifp->idev;
 723         int state;
 724         int hash;
 725         int deleted = 0, onlink = 0;
 726         unsigned long expires = jiffies;
 727
 728         hash = ipv6_addr_hash(&ifp->addr);
 729
 730         spin_lock_bh(&ifp->state_lock);
 731         state = ifp->state;
 732         ifp->state = INET6_IFADDR_STATE_DEAD;
 733         spin_unlock_bh(&ifp->state_lock);
 734
 735         if (state == INET6_IFADDR_STATE_DEAD)
 736                 goto out;
 737
 738         spin_lock_bh(&addrconf_hash_lock);
 739         hlist_del_init_rcu(&ifp->addr_lst);
 740         spin_unlock_bh(&addrconf_hash_lock);
 741
 742         write_lock_bh(&idev->lock);
 743 #ifdef CONFIG_IPV6_PRIVACY
 744         if (ifp->flags&IFA_F_TEMPORARY) {
 745                 list_del(&ifp->tmp_list);
 746                 if (ifp->ifpub) {
 747                         in6_ifa_put(ifp->ifpub);
 748                         ifp->ifpub = NULL;
 749                 }
 750                 __in6_ifa_put(ifp);
 751         }
 752 #endif
 753
 754         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 755                 if (ifa == ifp) {
 756                         list_del_init(&ifp->if_list);
 757                         __in6_ifa_put(ifp);
 758
 759                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 760                                 break;
 761                         deleted = 1;
 762                         continue;
 763                 } else if (ifp->flags & IFA_F_PERMANENT) {
 764                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 765                                               ifp->prefix_len)) {
 766                                 if (ifa->flags & IFA_F_PERMANENT) {
 767                                         onlink = 1;
 768                                         if (deleted)
 769                                                 break;
 770                                 } else {
 771                                         unsigned long lifetime;
 772
 773                                         if (!onlink)
 774                                                 onlink = -1;
 775
 776                                         spin_lock(&ifa->lock);
 777
 778                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 779                                         /*
 780                                          * Note: Because this address is
 781                                          * not permanent, lifetime <
 782                                          * LONG_MAX / HZ here.
 783                                          */
 784                                         if (time_before(expires,
 785                                                         ifa->tstamp + lifetime * HZ))
 786                                                 expires = ifa->tstamp + lifetime * HZ;
 787                                         spin_unlock(&ifa->lock);
 788                                 }
 789                         }
 790                 }
 791         }
 792         write_unlock_bh(&idev->lock);
 793
 794         addrconf_del_timer(ifp);
 795
 796         ipv6_ifa_notify(RTM_DELADDR, ifp);
 797
 798         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 799
 800         /*
 801          * Purge or update corresponding prefix
 802          *
 803          * 1) we don't purge prefix here if address was not permanent.
 804          *    prefix is managed by its own lifetime.
 805          * 2) if there're no addresses, delete prefix.
 806          * 3) if there're still other permanent address(es),
 807          *    corresponding prefix is still permanent.
 808          * 4) otherwise, update prefix lifetime to the
 809          *    longest valid lifetime among the corresponding
 810          *    addresses on the device.
 811          *    Note: subsequent RA will update lifetime.
 812          *
 813          * --yoshfuji
 814          */
 815         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 816                 struct in6_addr prefix;
 817                 struct rt6_info *rt;
 818                 struct net *net = dev_net(ifp->idev->dev);
 819                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 820                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 821
 822                 if (rt && addrconf_is_prefix_route(rt)) {
 823                         if (onlink == 0) {
 824                                 ip6_del_rt(rt);
 825                                 rt = NULL;
 826                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 827                                 rt->rt6i_expires = expires;
 828                                 rt->rt6i_flags |= RTF_EXPIRES;
 829                         }
 830                 }
 831                 dst_release(&rt->dst);
 832         }
 833
 834 out:
 835         in6_ifa_put(ifp);
 836 }
 837
 838 #ifdef CONFIG_IPV6_PRIVACY
 839 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 840 {
 841         struct inet6_dev *idev = ifp->idev;
 842         struct in6_addr addr, *tmpaddr;
 843         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 844         unsigned long regen_advance;
 845         int tmp_plen;
 846         int ret = 0;
 847         int max_addresses;
 848         u32 addr_flags;
 849
 850         write_lock(&idev->lock);
 851         if (ift) {
 852                 spin_lock_bh(&ift->lock);
 853                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 854                 spin_unlock_bh(&ift->lock);
 855                 tmpaddr = &addr;
 856         } else {
 857                 tmpaddr = NULL;
 858         }
 859 retry:
 860         in6_dev_hold(idev);
 861         if (idev->cnf.use_tempaddr <= 0) {
 862                 write_unlock(&idev->lock);
 863                 printk(KERN_INFO
 864                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 865                 in6_dev_put(idev);
 866                 ret = -1;
 867                 goto out;
 868         }
 869         spin_lock_bh(&ifp->lock);
 870         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 871                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 872                 spin_unlock_bh(&ifp->lock);
 873                 write_unlock(&idev->lock);
 874                 printk(KERN_WARNING
 875                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         in6_ifa_hold(ifp);
 881         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 882         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 883                 spin_unlock_bh(&ifp->lock);
 884                 write_unlock(&idev->lock);
 885                 printk(KERN_WARNING
 886                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 887                 in6_ifa_put(ifp);
 888                 in6_dev_put(idev);
 889                 ret = -1;
 890                 goto out;
 891         }
 892         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 893         age = (jiffies - ifp->tstamp) / HZ;
 894         tmp_valid_lft = min_t(__u32,
 895                               ifp->valid_lft,
 896                               idev->cnf.temp_valid_lft + age);
 897         tmp_prefered_lft = min_t(__u32,
 898                                  ifp->prefered_lft,
 899                                  idev->cnf.temp_prefered_lft + age -
 900                                  idev->cnf.max_desync_factor);
 901         tmp_plen = ifp->prefix_len;
 902         max_addresses = idev->cnf.max_addresses;
 903         tmp_cstamp = ifp->cstamp;
 904         tmp_tstamp = ifp->tstamp;
 905         spin_unlock_bh(&ifp->lock);
 906
 907         regen_advance = idev->cnf.regen_max_retry *
 908                         idev->cnf.dad_transmits *
 909                         idev->nd_parms->retrans_time / HZ;
 910         write_unlock(&idev->lock);
 911
 912         /* A temporary address is created only if this calculated Preferred
 913          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 914          * an implementation must not create a temporary address with a zero
 915          * Preferred Lifetime.
 916          */
 917         if (tmp_prefered_lft <= regen_advance) {
 918                 in6_ifa_put(ifp);
 919                 in6_dev_put(idev);
 920                 ret = -1;
 921                 goto out;
 922         }
 923
 924         addr_flags = IFA_F_TEMPORARY;
 925         /* set in addrconf_prefix_rcv() */
 926         if (ifp->flags & IFA_F_OPTIMISTIC)
 927                 addr_flags |= IFA_F_OPTIMISTIC;
 928
 929         ift = !max_addresses ||
 930               ipv6_count_addresses(idev) < max_addresses ?
 931                 ipv6_add_addr(idev, &addr, tmp_plen,
 932                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 933                               addr_flags) : NULL;
 934         if (!ift || IS_ERR(ift)) {
 935                 in6_ifa_put(ifp);
 936                 in6_dev_put(idev);
 937                 printk(KERN_INFO
 938                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 939                 tmpaddr = &addr;
 940                 write_lock(&idev->lock);
 941                 goto retry;
 942         }
 943
 944         spin_lock_bh(&ift->lock);
 945         ift->ifpub = ifp;
 946         ift->valid_lft = tmp_valid_lft;
 947         ift->prefered_lft = tmp_prefered_lft;
 948         ift->cstamp = tmp_cstamp;
 949         ift->tstamp = tmp_tstamp;
 950         spin_unlock_bh(&ift->lock);
 951
 952         addrconf_dad_start(ift, 0);
 953         in6_ifa_put(ift);
 954         in6_dev_put(idev);
 955 out:
 956         return ret;
 957 }
 958 #endif
 959
 960 /*
 961  *      Choose an appropriate source address (RFC3484)
 962  */
 963 enum {
 964         IPV6_SADDR_RULE_INIT = 0,
 965         IPV6_SADDR_RULE_LOCAL,
 966         IPV6_SADDR_RULE_SCOPE,
 967         IPV6_SADDR_RULE_PREFERRED,
 968 #ifdef CONFIG_IPV6_MIP6
 969         IPV6_SADDR_RULE_HOA,
 970 #endif
 971         IPV6_SADDR_RULE_OIF,
 972         IPV6_SADDR_RULE_LABEL,
 973 #ifdef CONFIG_IPV6_PRIVACY
 974         IPV6_SADDR_RULE_PRIVACY,
 975 #endif
 976         IPV6_SADDR_RULE_ORCHID,
 977         IPV6_SADDR_RULE_PREFIX,
 978         IPV6_SADDR_RULE_MAX
 979 };
 980
 981 struct ipv6_saddr_score {
 982         int                     rule;
 983         int                     addr_type;
 984         struct inet6_ifaddr     *ifa;
 985         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 986         int                     scopedist;
 987         int                     matchlen;
 988 };
 989
 990 struct ipv6_saddr_dst {
 991         const struct in6_addr *addr;
 992         int ifindex;
 993         int scope;
 994         int label;
 995         unsigned int prefs;
 996 };
 997
 998 static inline int ipv6_saddr_preferred(int type)
 999 {
1000         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
1001                 return 1;
1002         return 0;
1003 }
1004
1005 static int ipv6_get_saddr_eval(struct net *net,
1006                                struct ipv6_saddr_score *score,
1007                                struct ipv6_saddr_dst *dst,
1008                                int i)
1009 {
1010         int ret;
1011
1012         if (i <= score->rule) {
1013                 switch (i) {
1014                 case IPV6_SADDR_RULE_SCOPE:
1015                         ret = score->scopedist;
1016                         break;
1017                 case IPV6_SADDR_RULE_PREFIX:
1018                         ret = score->matchlen;
1019                         break;
1020                 default:
1021                         ret = !!test_bit(i, score->scorebits);
1022                 }
1023                 goto out;
1024         }
1025
1026         switch (i) {
1027         case IPV6_SADDR_RULE_INIT:
1028                 /* Rule 0: remember if hiscore is not ready yet */
1029                 ret = !!score->ifa;
1030                 break;
1031         case IPV6_SADDR_RULE_LOCAL:
1032                 /* Rule 1: Prefer same address */
1033                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1034                 break;
1035         case IPV6_SADDR_RULE_SCOPE:
1036                 /* Rule 2: Prefer appropriate scope
1037                  *
1038                  *      ret
1039                  *       ^
1040                  *    -1 |  d 15
1041                  *    ---+--+-+---> scope
1042                  *       |
1043                  *       |             d is scope of the destination.
1044                  *  B-d  |  \
1045                  *       |   \      <- smaller scope is better if
1046                  *  B-15 |    \        if scope is enough for destinaion.
1047                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1048                  * d-C-1 | /
1049                  *       |/         <- greater is better
1050                  *   -C  /             if scope is not enough for destination.
1051                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1052                  *
1053                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1054                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1055                  * Assume B = 0 and we get C > 29.
1056                  */
1057                 ret = __ipv6_addr_src_scope(score->addr_type);
1058                 if (ret >= dst->scope)
1059                         ret = -ret;
1060                 else
1061                         ret -= 128;     /* 30 is enough */
1062                 score->scopedist = ret;
1063                 break;
1064         case IPV6_SADDR_RULE_PREFERRED:
1065                 /* Rule 3: Avoid deprecated and optimistic addresses */
1066                 ret = ipv6_saddr_preferred(score->addr_type) ||
1067                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1068                 break;
1069 #ifdef CONFIG_IPV6_MIP6
1070         case IPV6_SADDR_RULE_HOA:
1071             {
1072                 /* Rule 4: Prefer home address */
1073                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1074                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1075                 break;
1076             }
1077 #endif
1078         case IPV6_SADDR_RULE_OIF:
1079                 /* Rule 5: Prefer outgoing interface */
1080                 ret = (!dst->ifindex ||
1081                        dst->ifindex == score->ifa->idev->dev->ifindex);
1082                 break;
1083         case IPV6_SADDR_RULE_LABEL:
1084                 /* Rule 6: Prefer matching label */
1085                 ret = ipv6_addr_label(net,
1086                                       &score->ifa->addr, score->addr_type,
1087                                       score->ifa->idev->dev->ifindex) == dst->label;
1088                 break;
1089 #ifdef CONFIG_IPV6_PRIVACY
1090         case IPV6_SADDR_RULE_PRIVACY:
1091             {
1092                 /* Rule 7: Prefer public address
1093                  * Note: prefer temprary address if use_tempaddr >= 2
1094                  */
1095                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1096                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1097                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1098                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1099                 break;
1100             }
1101 #endif
1102         case IPV6_SADDR_RULE_ORCHID:
1103                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1104                  *          non-ORCHID vs non-ORCHID
1105                  */
1106                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1107                         ipv6_addr_orchid(dst->addr));
1108                 break;
1109         case IPV6_SADDR_RULE_PREFIX:
1110                 /* Rule 8: Use longest matching prefix */
1111                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1112                                                        dst->addr);
1113                 break;
1114         default:
1115                 ret = 0;
1116         }
1117
1118         if (ret)
1119                 __set_bit(i, score->scorebits);
1120         score->rule = i;
1121 out:
1122         return ret;
1123 }
1124
1125 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1126                        const struct in6_addr *daddr, unsigned int prefs,
1127                        struct in6_addr *saddr)
1128 {
1129         struct ipv6_saddr_score scores[2],
1130                                 *score = &scores[0], *hiscore = &scores[1];
1131         struct ipv6_saddr_dst dst;
1132         struct net_device *dev;
1133         int dst_type;
1134
1135         dst_type = __ipv6_addr_type(daddr);
1136         dst.addr = daddr;
1137         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1138         dst.scope = __ipv6_addr_src_scope(dst_type);
1139         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1140         dst.prefs = prefs;
1141
1142         hiscore->rule = -1;
1143         hiscore->ifa = NULL;
1144
1145         rcu_read_lock();
1146
1147         for_each_netdev_rcu(net, dev) {
1148                 struct inet6_dev *idev;
1149
1150                 /* Candidate Source Address (section 4)
1151                  *  - multicast and link-local destination address,
1152                  *    the set of candidate source address MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same link as the outgoing
1155                  *    interface.
1156                  * (- For site-local destination addresses, the
1157                  *    set of candidate source addresses MUST only
1158                  *    include addresses assigned to interfaces
1159                  *    belonging to the same site as the outgoing
1160                  *    interface.)
1161                  */
1162                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1163                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1164                     dst.ifindex && dev->ifindex != dst.ifindex)
1165                         continue;
1166
1167                 idev = __in6_dev_get(dev);
1168                 if (!idev)
1169                         continue;
1170
1171                 read_lock_bh(&idev->lock);
1172                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1173                         int i;
1174
1175                         /*
1176                          * - Tentative Address (RFC2462 section 5.4)
1177                          *  - A tentative address is not considered
1178                          *    "assigned to an interface" in the traditional
1179                          *    sense, unless it is also flagged as optimistic.
1180                          * - Candidate Source Address (section 4)
1181                          *  - In any case, anycast addresses, multicast
1182                          *    addresses, and the unspecified address MUST
1183                          *    NOT be included in a candidate set.
1184                          */
1185                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1186                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1187                                 continue;
1188
1189                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1190
1191                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1192                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1193                                 LIMIT_NETDEBUG(KERN_DEBUG
1194                                                "ADDRCONF: unspecified / multicast address "
1195                                                "assigned as unicast address on %s",
1196                                                dev->name);
1197                                 continue;
1198                         }
1199
1200                         score->rule = -1;
1201                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1202
1203                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1204                                 int minihiscore, miniscore;
1205
1206                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1207                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1208
1209                                 if (minihiscore > miniscore) {
1210                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1211                                             score->scopedist > 0) {
1212                                                 /*
1213                                                  * special case:
1214                                                  * each remaining entry
1215                                                  * has too small (not enough)
1216                                                  * scope, because ifa entries
1217                                                  * are sorted by their scope
1218                                                  * values.
1219                                                  */
1220                                                 goto try_nextdev;
1221                                         }
1222                                         break;
1223                                 } else if (minihiscore < miniscore) {
1224                                         if (hiscore->ifa)
1225                                                 in6_ifa_put(hiscore->ifa);
1226
1227                                         in6_ifa_hold(score->ifa);
1228
1229                                         swap(hiscore, score);
1230
1231                                         /* restore our iterator */
1232                                         score->ifa = hiscore->ifa;
1233
1234                                         break;
1235                                 }
1236                         }
1237                 }
1238 try_nextdev:
1239                 read_unlock_bh(&idev->lock);
1240         }
1241         rcu_read_unlock();
1242
1243         if (!hiscore->ifa)
1244                 return -EADDRNOTAVAIL;
1245
1246         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1247         in6_ifa_put(hiscore->ifa);
1248         return 0;
1249 }
1250 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1251
1252 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1253                     unsigned char banned_flags)
1254 {
1255         struct inet6_dev *idev;
1256         int err = -EADDRNOTAVAIL;
1257
1258         rcu_read_lock();
1259         idev = __in6_dev_get(dev);
1260         if (idev) {
1261                 struct inet6_ifaddr *ifp;
1262
1263                 read_lock_bh(&idev->lock);
1264                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1265                         if (ifp->scope == IFA_LINK &&
1266                             !(ifp->flags & banned_flags)) {
1267                                 ipv6_addr_copy(addr, &ifp->addr);
1268                                 err = 0;
1269                                 break;
1270                         }
1271                 }
1272                 read_unlock_bh(&idev->lock);
1273         }
1274         rcu_read_unlock();
1275         return err;
1276 }
1277
1278 static int ipv6_count_addresses(struct inet6_dev *idev)
1279 {
1280         int cnt = 0;
1281         struct inet6_ifaddr *ifp;
1282
1283         read_lock_bh(&idev->lock);
1284         list_for_each_entry(ifp, &idev->addr_list, if_list)
1285                 cnt++;
1286         read_unlock_bh(&idev->lock);
1287         return cnt;
1288 }
1289
1290 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1291                   struct net_device *dev, int strict)
1292 {
1293         struct inet6_ifaddr *ifp;
1294         struct hlist_node *node;
1295         unsigned int hash = ipv6_addr_hash(addr);
1296
1297         rcu_read_lock_bh();
1298         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1299                 if (!net_eq(dev_net(ifp->idev->dev), net))
1300                         continue;
1301                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1302                     !(ifp->flags&IFA_F_TENTATIVE) &&
1303                     (dev == NULL || ifp->idev->dev == dev ||
1304                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1305                         rcu_read_unlock_bh();
1306                         return 1;
1307                 }
1308         }
1309
1310         rcu_read_unlock_bh();
1311         return 0;
1312 }
1313 EXPORT_SYMBOL(ipv6_chk_addr);
1314
1315 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1316                                struct net_device *dev)
1317 {
1318         unsigned int hash = ipv6_addr_hash(addr);
1319         struct inet6_ifaddr *ifp;
1320         struct hlist_node *node;
1321
1322         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1323                 if (!net_eq(dev_net(ifp->idev->dev), net))
1324                         continue;
1325                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1326                         if (dev == NULL || ifp->idev->dev == dev)
1327                                 return true;
1328                 }
1329         }
1330         return false;
1331 }
1332
1333 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1334 {
1335         struct inet6_dev *idev;
1336         struct inet6_ifaddr *ifa;
1337         int     onlink;
1338
1339         onlink = 0;
1340         rcu_read_lock();
1341         idev = __in6_dev_get(dev);
1342         if (idev) {
1343                 read_lock_bh(&idev->lock);
1344                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1345                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1346                                                    ifa->prefix_len);
1347                         if (onlink)
1348                                 break;
1349                 }
1350                 read_unlock_bh(&idev->lock);
1351         }
1352         rcu_read_unlock();
1353         return onlink;
1354 }
1355
1356 EXPORT_SYMBOL(ipv6_chk_prefix);
1357
1358 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1359                                      struct net_device *dev, int strict)
1360 {
1361         struct inet6_ifaddr *ifp, *result = NULL;
1362         unsigned int hash = ipv6_addr_hash(addr);
1363         struct hlist_node *node;
1364
1365         rcu_read_lock_bh();
1366         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1367                 if (!net_eq(dev_net(ifp->idev->dev), net))
1368                         continue;
1369                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1370                         if (dev == NULL || ifp->idev->dev == dev ||
1371                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1372                                 result = ifp;
1373                                 in6_ifa_hold(ifp);
1374                                 break;
1375                         }
1376                 }
1377         }
1378         rcu_read_unlock_bh();
1379
1380         return result;
1381 }
1382
1383 /* Gets referenced address, destroys ifaddr */
1384
1385 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1386 {
1387         if (ifp->flags&IFA_F_PERMANENT) {
1388                 spin_lock_bh(&ifp->lock);
1389                 addrconf_del_timer(ifp);
1390                 ifp->flags |= IFA_F_TENTATIVE;
1391                 if (dad_failed)
1392                         ifp->flags |= IFA_F_DADFAILED;
1393                 spin_unlock_bh(&ifp->lock);
1394                 if (dad_failed)
1395                         ipv6_ifa_notify(0, ifp);
1396                 in6_ifa_put(ifp);
1397 #ifdef CONFIG_IPV6_PRIVACY
1398         } else if (ifp->flags&IFA_F_TEMPORARY) {
1399                 struct inet6_ifaddr *ifpub;
1400                 spin_lock_bh(&ifp->lock);
1401                 ifpub = ifp->ifpub;
1402                 if (ifpub) {
1403                         in6_ifa_hold(ifpub);
1404                         spin_unlock_bh(&ifp->lock);
1405                         ipv6_create_tempaddr(ifpub, ifp);
1406                         in6_ifa_put(ifpub);
1407                 } else {
1408                         spin_unlock_bh(&ifp->lock);
1409                 }
1410                 ipv6_del_addr(ifp);
1411 #endif
1412         } else
1413                 ipv6_del_addr(ifp);
1414 }
1415
1416 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1417 {
1418         int err = -ENOENT;
1419
1420         spin_lock(&ifp->state_lock);
1421         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1422                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1423                 err = 0;
1424         }
1425         spin_unlock(&ifp->state_lock);
1426
1427         return err;
1428 }
1429
1430 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1431 {
1432         struct inet6_dev *idev = ifp->idev;
1433
1434         if (addrconf_dad_end(ifp)) {
1435                 in6_ifa_put(ifp);
1436                 return;
1437         }
1438
1439         if (net_ratelimit())
1440                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1441                         ifp->idev->dev->name, &ifp->addr);
1442
1443         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1444                 struct in6_addr addr;
1445
1446                 addr.s6_addr32[0] = htonl(0xfe800000);
1447                 addr.s6_addr32[1] = 0;
1448
1449                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1450                     ipv6_addr_equal(&ifp->addr, &addr)) {
1451                         /* DAD failed for link-local based on MAC address */
1452                         idev->cnf.disable_ipv6 = 1;
1453
1454                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1455                                 ifp->idev->dev->name);
1456                 }
1457         }
1458
1459         addrconf_dad_stop(ifp, 1);
1460 }
1461
1462 /* Join to solicited addr multicast group. */
1463
1464 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1465 {
1466         struct in6_addr maddr;
1467
1468         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1469                 return;
1470
1471         addrconf_addr_solict_mult(addr, &maddr);
1472         ipv6_dev_mc_inc(dev, &maddr);
1473 }
1474
1475 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1476 {
1477         struct in6_addr maddr;
1478
1479         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1480                 return;
1481
1482         addrconf_addr_solict_mult(addr, &maddr);
1483         __ipv6_dev_mc_dec(idev, &maddr);
1484 }
1485
1486 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1487 {
1488         struct in6_addr addr;
1489         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490         if (ipv6_addr_any(&addr))
1491                 return;
1492         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1493 }
1494
1495 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1496 {
1497         struct in6_addr addr;
1498         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1499         if (ipv6_addr_any(&addr))
1500                 return;
1501         __ipv6_dev_ac_dec(ifp->idev, &addr);
1502 }
1503
1504 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1505 {
1506         if (dev->addr_len != ETH_ALEN)
1507                 return -1;
1508         memcpy(eui, dev->dev_addr, 3);
1509         memcpy(eui + 5, dev->dev_addr + 3, 3);
1510
1511         /*
1512          * The zSeries OSA network cards can be shared among various
1513          * OS instances, but the OSA cards have only one MAC address.
1514          * This leads to duplicate address conflicts in conjunction
1515          * with IPv6 if more than one instance uses the same card.
1516          *
1517          * The driver for these cards can deliver a unique 16-bit
1518          * identifier for each instance sharing the same card.  It is
1519          * placed instead of 0xFFFE in the interface identifier.  The
1520          * "u" bit of the interface identifier is not inverted in this
1521          * case.  Hence the resulting interface identifier has local
1522          * scope according to RFC2373.
1523          */
1524         if (dev->dev_id) {
1525                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1526                 eui[4] = dev->dev_id & 0xFF;
1527         } else {
1528                 eui[3] = 0xFF;
1529                 eui[4] = 0xFE;
1530                 eui[0] ^= 2;
1531         }
1532         return 0;
1533 }
1534
1535 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1536 {
1537         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1538         if (dev->addr_len != ARCNET_ALEN)
1539                 return -1;
1540         memset(eui, 0, 7);
1541         eui[7] = *(u8*)dev->dev_addr;
1542         return 0;
1543 }
1544
1545 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1546 {
1547         if (dev->addr_len != INFINIBAND_ALEN)
1548                 return -1;
1549         memcpy(eui, dev->dev_addr + 12, 8);
1550         eui[0] |= 2;
1551         return 0;
1552 }
1553
1554 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1555 {
1556         if (addr == 0)
1557                 return -1;
1558         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1559                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1560                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1561                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1562                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1563                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1564         eui[1] = 0;
1565         eui[2] = 0x5E;
1566         eui[3] = 0xFE;
1567         memcpy(eui + 4, &addr, 4);
1568         return 0;
1569 }
1570
1571 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1572 {
1573         if (dev->priv_flags & IFF_ISATAP)
1574                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1575         return -1;
1576 }
1577
1578 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1579 {
1580         switch (dev->type) {
1581         case ARPHRD_ETHER:
1582         case ARPHRD_FDDI:
1583         case ARPHRD_IEEE802_TR:
1584                 return addrconf_ifid_eui48(eui, dev);
1585         case ARPHRD_ARCNET:
1586                 return addrconf_ifid_arcnet(eui, dev);
1587         case ARPHRD_INFINIBAND:
1588                 return addrconf_ifid_infiniband(eui, dev);
1589         case ARPHRD_SIT:
1590                 return addrconf_ifid_sit(eui, dev);
1591         }
1592         return -1;
1593 }
1594
1595 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1596 {
1597         int err = -1;
1598         struct inet6_ifaddr *ifp;
1599
1600         read_lock_bh(&idev->lock);
1601         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1602                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1603                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1604                         err = 0;
1605                         break;
1606                 }
1607         }
1608         read_unlock_bh(&idev->lock);
1609         return err;
1610 }
1611
1612 #ifdef CONFIG_IPV6_PRIVACY
1613 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1614 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1615 {
1616 regen:
1617         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1618         idev->rndid[0] &= ~0x02;
1619
1620         /*
1621          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1622          * check if generated address is not inappropriate
1623          *
1624          *  - Reserved subnet anycast (RFC 2526)
1625          *      11111101 11....11 1xxxxxxx
1626          *  - ISATAP (RFC4214) 6.1
1627          *      00-00-5E-FE-xx-xx-xx-xx
1628          *  - value 0
1629          *  - XXX: already assigned to an address on the device
1630          */
1631         if (idev->rndid[0] == 0xfd &&
1632             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1633             (idev->rndid[7]&0x80))
1634                 goto regen;
1635         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1636                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1637                         goto regen;
1638                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1639                         goto regen;
1640         }
1641
1642         return 0;
1643 }
1644
1645 static void ipv6_regen_rndid(unsigned long data)
1646 {
1647         struct inet6_dev *idev = (struct inet6_dev *) data;
1648         unsigned long expires;
1649
1650         rcu_read_lock_bh();
1651         write_lock_bh(&idev->lock);
1652
1653         if (idev->dead)
1654                 goto out;
1655
1656         if (__ipv6_regen_rndid(idev) < 0)
1657                 goto out;
1658
1659         expires = jiffies +
1660                 idev->cnf.temp_prefered_lft * HZ -
1661                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1662                 idev->cnf.max_desync_factor * HZ;
1663         if (time_before(expires, jiffies)) {
1664                 printk(KERN_WARNING
1665                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1666                         idev->dev->name);
1667                 goto out;
1668         }
1669
1670         if (!mod_timer(&idev->regen_timer, expires))
1671                 in6_dev_hold(idev);
1672
1673 out:
1674         write_unlock_bh(&idev->lock);
1675         rcu_read_unlock_bh();
1676         in6_dev_put(idev);
1677 }
1678
1679 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1680         int ret = 0;
1681
1682         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1683                 ret = __ipv6_regen_rndid(idev);
1684         return ret;
1685 }
1686 #endif
1687
1688 /*
1689  *      Add prefix route.
1690  */
1691
1692 static void
1693 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1694                       unsigned long expires, u32 flags)
1695 {
1696         struct fib6_config cfg = {
1697                 .fc_table = RT6_TABLE_PREFIX,
1698                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1699                 .fc_ifindex = dev->ifindex,
1700                 .fc_expires = expires,
1701                 .fc_dst_len = plen,
1702                 .fc_flags = RTF_UP | flags,
1703                 .fc_nlinfo.nl_net = dev_net(dev),
1704                 .fc_protocol = RTPROT_KERNEL,
1705         };
1706
1707         ipv6_addr_copy(&cfg.fc_dst, pfx);
1708
1709         /* Prevent useless cloning on PtP SIT.
1710            This thing is done here expecting that the whole
1711            class of non-broadcast devices need not cloning.
1712          */
1713 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1714         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1715                 cfg.fc_flags |= RTF_NONEXTHOP;
1716 #endif
1717
1718         ip6_route_add(&cfg);
1719 }
1720
1721 /* Create "default" multicast route to the interface */
1722
1723 static void addrconf_add_mroute(struct net_device *dev)
1724 {
1725         struct fib6_config cfg = {
1726                 .fc_table = RT6_TABLE_LOCAL,
1727                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1728                 .fc_ifindex = dev->ifindex,
1729                 .fc_dst_len = 8,
1730                 .fc_flags = RTF_UP,
1731                 .fc_nlinfo.nl_net = dev_net(dev),
1732         };
1733
1734         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1735
1736         ip6_route_add(&cfg);
1737 }
1738
1739 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1740 static void sit_route_add(struct net_device *dev)
1741 {
1742         struct fib6_config cfg = {
1743                 .fc_table = RT6_TABLE_MAIN,
1744                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1745                 .fc_ifindex = dev->ifindex,
1746                 .fc_dst_len = 96,
1747                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1748                 .fc_nlinfo.nl_net = dev_net(dev),
1749         };
1750
1751         /* prefix length - 96 bits "::d.d.d.d" */
1752         ip6_route_add(&cfg);
1753 }
1754 #endif
1755
1756 static void addrconf_add_lroute(struct net_device *dev)
1757 {
1758         struct in6_addr addr;
1759
1760         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1761         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1762 }
1763
1764 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1765 {
1766         struct inet6_dev *idev;
1767
1768         ASSERT_RTNL();
1769
1770         idev = ipv6_find_idev(dev);
1771         if (!idev)
1772                 return ERR_PTR(-ENOBUFS);
1773
1774         if (idev->cnf.disable_ipv6)
1775                 return ERR_PTR(-EACCES);
1776
1777         /* Add default multicast route */
1778         addrconf_add_mroute(dev);
1779
1780         /* Add link local route */
1781         addrconf_add_lroute(dev);
1782         return idev;
1783 }
1784
1785 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1786 {
1787         struct prefix_info *pinfo;
1788         __u32 valid_lft;
1789         __u32 prefered_lft;
1790         int addr_type;
1791         struct inet6_dev *in6_dev;
1792         struct net *net = dev_net(dev);
1793
1794         pinfo = (struct prefix_info *) opt;
1795
1796         if (len < sizeof(struct prefix_info)) {
1797                 ADBG(("addrconf: prefix option too short\n"));
1798                 return;
1799         }
1800
1801         /*
1802          *      Validation checks ([ADDRCONF], page 19)
1803          */
1804
1805         addr_type = ipv6_addr_type(&pinfo->prefix);
1806
1807         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1808                 return;
1809
1810         valid_lft = ntohl(pinfo->valid);
1811         prefered_lft = ntohl(pinfo->prefered);
1812
1813         if (prefered_lft > valid_lft) {
1814                 if (net_ratelimit())
1815                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1816                 return;
1817         }
1818
1819         in6_dev = in6_dev_get(dev);
1820
1821         if (in6_dev == NULL) {
1822                 if (net_ratelimit())
1823                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1824                 return;
1825         }
1826
1827         /*
1828          *      Two things going on here:
1829          *      1) Add routes for on-link prefixes
1830          *      2) Configure prefixes with the auto flag set
1831          */
1832
1833         if (pinfo->onlink) {
1834                 struct rt6_info *rt;
1835                 unsigned long rt_expires;
1836
1837                 /* Avoid arithmetic overflow. Really, we could
1838                  * save rt_expires in seconds, likely valid_lft,
1839                  * but it would require division in fib gc, that it
1840                  * not good.
1841                  */
1842                 if (HZ > USER_HZ)
1843                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1844                 else
1845                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1846
1847                 if (addrconf_finite_timeout(rt_expires))
1848                         rt_expires *= HZ;
1849
1850                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1851                                 dev->ifindex, 1);
1852
1853                 if (rt && addrconf_is_prefix_route(rt)) {
1854                         /* Autoconf prefix route */
1855                         if (valid_lft == 0) {
1856                                 ip6_del_rt(rt);
1857                                 rt = NULL;
1858                         } else if (addrconf_finite_timeout(rt_expires)) {
1859                                 /* not infinity */
1860                                 rt->rt6i_expires = jiffies + rt_expires;
1861                                 rt->rt6i_flags |= RTF_EXPIRES;
1862                         } else {
1863                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1864                                 rt->rt6i_expires = 0;
1865                         }
1866                 } else if (valid_lft) {
1867                         clock_t expires = 0;
1868                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1869                         if (addrconf_finite_timeout(rt_expires)) {
1870                                 /* not infinity */
1871                                 flags |= RTF_EXPIRES;
1872                                 expires = jiffies_to_clock_t(rt_expires);
1873                         }
1874                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1875                                               dev, expires, flags);
1876                 }
1877                 if (rt)
1878                         dst_release(&rt->dst);
1879         }
1880
1881         /* Try to figure out our local address for this prefix */
1882
1883         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1884                 struct inet6_ifaddr * ifp;
1885                 struct in6_addr addr;
1886                 int create = 0, update_lft = 0;
1887
1888                 if (pinfo->prefix_len == 64) {
1889                         memcpy(&addr, &pinfo->prefix, 8);
1890                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1891                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1892                                 in6_dev_put(in6_dev);
1893                                 return;
1894                         }
1895                         goto ok;
1896                 }
1897                 if (net_ratelimit())
1898                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1899                                pinfo->prefix_len);
1900                 in6_dev_put(in6_dev);
1901                 return;
1902
1903 ok:
1904
1905                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1906
1907                 if (ifp == NULL && valid_lft) {
1908                         int max_addresses = in6_dev->cnf.max_addresses;
1909                         u32 addr_flags = 0;
1910
1911 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1912                         if (in6_dev->cnf.optimistic_dad &&
1913                             !net->ipv6.devconf_all->forwarding)
1914                                 addr_flags = IFA_F_OPTIMISTIC;
1915 #endif
1916
1917                         /* Do not allow to create too much of autoconfigured
1918                          * addresses; this would be too easy way to crash kernel.
1919                          */
1920                         if (!max_addresses ||
1921                             ipv6_count_addresses(in6_dev) < max_addresses)
1922                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1923                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1924                                                     addr_flags);
1925
1926                         if (!ifp || IS_ERR(ifp)) {
1927                                 in6_dev_put(in6_dev);
1928                                 return;
1929                         }
1930
1931                         update_lft = create = 1;
1932                         ifp->cstamp = jiffies;
1933                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1934                 }
1935
1936                 if (ifp) {
1937                         int flags;
1938                         unsigned long now;
1939 #ifdef CONFIG_IPV6_PRIVACY
1940                         struct inet6_ifaddr *ift;
1941 #endif
1942                         u32 stored_lft;
1943
1944                         /* update lifetime (RFC2462 5.5.3 e) */
1945                         spin_lock(&ifp->lock);
1946                         now = jiffies;
1947                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1948                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1949                         else
1950                                 stored_lft = 0;
1951                         if (!update_lft && stored_lft) {
1952                                 if (valid_lft > MIN_VALID_LIFETIME ||
1953                                     valid_lft > stored_lft)
1954                                         update_lft = 1;
1955                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1956                                         /* valid_lft <= stored_lft is always true */
1957                                         /*
1958                                          * RFC 4862 Section 5.5.3e:
1959                                          * "Note that the preferred lifetime of
1960                                          *  the corresponding address is always
1961                                          *  reset to the Preferred Lifetime in
1962                                          *  the received Prefix Information
1963                                          *  option, regardless of whether the
1964                                          *  valid lifetime is also reset or
1965                                          *  ignored."
1966                                          *
1967                                          *  So if the preferred lifetime in
1968                                          *  this advertisement is different
1969                                          *  than what we have stored, but the
1970                                          *  valid lifetime is invalid, just
1971                                          *  reset prefered_lft.
1972                                          *
1973                                          *  We must set the valid lifetime
1974                                          *  to the stored lifetime since we'll
1975                                          *  be updating the timestamp below,
1976                                          *  else we'll set it back to the
1977                                          *  minumum.
1978                                          */
1979                                         if (prefered_lft != ifp->prefered_lft) {
1980                                                 valid_lft = stored_lft;
1981                                                 update_lft = 1;
1982                                         }
1983                                 } else {
1984                                         valid_lft = MIN_VALID_LIFETIME;
1985                                         if (valid_lft < prefered_lft)
1986                                                 prefered_lft = valid_lft;
1987                                         update_lft = 1;
1988                                 }
1989                         }
1990
1991                         if (update_lft) {
1992                                 ifp->valid_lft = valid_lft;
1993                                 ifp->prefered_lft = prefered_lft;
1994                                 ifp->tstamp = now;
1995                                 flags = ifp->flags;
1996                                 ifp->flags &= ~IFA_F_DEPRECATED;
1997                                 spin_unlock(&ifp->lock);
1998
1999                                 if (!(flags&IFA_F_TENTATIVE))
2000                                         ipv6_ifa_notify(0, ifp);
2001                         } else
2002                                 spin_unlock(&ifp->lock);
2003
2004 #ifdef CONFIG_IPV6_PRIVACY
2005                         read_lock_bh(&in6_dev->lock);
2006                         /* update all temporary addresses in the list */
2007                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2008                                 /*
2009                                  * When adjusting the lifetimes of an existing
2010                                  * temporary address, only lower the lifetimes.
2011                                  * Implementations must not increase the
2012                                  * lifetimes of an existing temporary address
2013                                  * when processing a Prefix Information Option.
2014                                  */
2015                                 if (ifp != ift->ifpub)
2016                                         continue;
2017
2018                                 spin_lock(&ift->lock);
2019                                 flags = ift->flags;
2020                                 if (ift->valid_lft > valid_lft &&
2021                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2022                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2023                                 if (ift->prefered_lft > prefered_lft &&
2024                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2025                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2026                                 spin_unlock(&ift->lock);
2027                                 if (!(flags&IFA_F_TENTATIVE))
2028                                         ipv6_ifa_notify(0, ift);
2029                         }
2030
2031                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2032                                 /*
2033                                  * When a new public address is created as described in [ADDRCONF],
2034                                  * also create a new temporary address. Also create a temporary
2035                                  * address if it's enabled but no temporary address currently exists.
2036                                  */
2037                                 read_unlock_bh(&in6_dev->lock);
2038                                 ipv6_create_tempaddr(ifp, NULL);
2039                         } else {
2040                                 read_unlock_bh(&in6_dev->lock);
2041                         }
2042 #endif
2043                         in6_ifa_put(ifp);
2044                         addrconf_verify(0);
2045                 }
2046         }
2047         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2048         in6_dev_put(in6_dev);
2049 }
2050
2051 /*
2052  *      Set destination address.
2053  *      Special case for SIT interfaces where we create a new "virtual"
2054  *      device.
2055  */
2056 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2057 {
2058         struct in6_ifreq ireq;
2059         struct net_device *dev;
2060         int err = -EINVAL;
2061
2062         rtnl_lock();
2063
2064         err = -EFAULT;
2065         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2066                 goto err_exit;
2067
2068         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2069
2070         err = -ENODEV;
2071         if (dev == NULL)
2072                 goto err_exit;
2073
2074 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2075         if (dev->type == ARPHRD_SIT) {
2076                 const struct net_device_ops *ops = dev->netdev_ops;
2077                 struct ifreq ifr;
2078                 struct ip_tunnel_parm p;
2079
2080                 err = -EADDRNOTAVAIL;
2081                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2082                         goto err_exit;
2083
2084                 memset(&p, 0, sizeof(p));
2085                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2086                 p.iph.saddr = 0;
2087                 p.iph.version = 4;
2088                 p.iph.ihl = 5;
2089                 p.iph.protocol = IPPROTO_IPV6;
2090                 p.iph.ttl = 64;
2091                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2092
2093                 if (ops->ndo_do_ioctl) {
2094                         mm_segment_t oldfs = get_fs();
2095
2096                         set_fs(KERNEL_DS);
2097                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2098                         set_fs(oldfs);
2099                 } else
2100                         err = -EOPNOTSUPP;
2101
2102                 if (err == 0) {
2103                         err = -ENOBUFS;
2104                         dev = __dev_get_by_name(net, p.name);
2105                         if (!dev)
2106                                 goto err_exit;
2107                         err = dev_open(dev);
2108                 }
2109         }
2110 #endif
2111
2112 err_exit:
2113         rtnl_unlock();
2114         return err;
2115 }
2116
2117 /*
2118  *      Manual configuration of address on an interface
2119  */
2120 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2121                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2122                           __u32 valid_lft)
2123 {
2124         struct inet6_ifaddr *ifp;
2125         struct inet6_dev *idev;
2126         struct net_device *dev;
2127         int scope;
2128         u32 flags;
2129         clock_t expires;
2130         unsigned long timeout;
2131
2132         ASSERT_RTNL();
2133
2134         if (plen > 128)
2135                 return -EINVAL;
2136
2137         /* check the lifetime */
2138         if (!valid_lft || prefered_lft > valid_lft)
2139                 return -EINVAL;
2140
2141         dev = __dev_get_by_index(net, ifindex);
2142         if (!dev)
2143                 return -ENODEV;
2144
2145         idev = addrconf_add_dev(dev);
2146         if (IS_ERR(idev))
2147                 return PTR_ERR(idev);
2148
2149         scope = ipv6_addr_scope(pfx);
2150
2151         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2152         if (addrconf_finite_timeout(timeout)) {
2153                 expires = jiffies_to_clock_t(timeout * HZ);
2154                 valid_lft = timeout;
2155                 flags = RTF_EXPIRES;
2156         } else {
2157                 expires = 0;
2158                 flags = 0;
2159                 ifa_flags |= IFA_F_PERMANENT;
2160         }
2161
2162         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2163         if (addrconf_finite_timeout(timeout)) {
2164                 if (timeout == 0)
2165                         ifa_flags |= IFA_F_DEPRECATED;
2166                 prefered_lft = timeout;
2167         }
2168
2169         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2170
2171         if (!IS_ERR(ifp)) {
2172                 spin_lock_bh(&ifp->lock);
2173                 ifp->valid_lft = valid_lft;
2174                 ifp->prefered_lft = prefered_lft;
2175                 ifp->tstamp = jiffies;
2176                 spin_unlock_bh(&ifp->lock);
2177
2178                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2179                                       expires, flags);
2180                 /*
2181                  * Note that section 3.1 of RFC 4429 indicates
2182                  * that the Optimistic flag should not be set for
2183                  * manually configured addresses
2184                  */
2185                 addrconf_dad_start(ifp, 0);
2186                 in6_ifa_put(ifp);
2187                 addrconf_verify(0);
2188                 return 0;
2189         }
2190
2191         return PTR_ERR(ifp);
2192 }
2193
2194 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2195                           unsigned int plen)
2196 {
2197         struct inet6_ifaddr *ifp;
2198         struct inet6_dev *idev;
2199         struct net_device *dev;
2200
2201         if (plen > 128)
2202                 return -EINVAL;
2203
2204         dev = __dev_get_by_index(net, ifindex);
2205         if (!dev)
2206                 return -ENODEV;
2207
2208         if ((idev = __in6_dev_get(dev)) == NULL)
2209                 return -ENXIO;
2210
2211         read_lock_bh(&idev->lock);
2212         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2213                 if (ifp->prefix_len == plen &&
2214                     ipv6_addr_equal(pfx, &ifp->addr)) {
2215                         in6_ifa_hold(ifp);
2216                         read_unlock_bh(&idev->lock);
2217
2218                         ipv6_del_addr(ifp);
2219
2220                         /* If the last address is deleted administratively,
2221                            disable IPv6 on this interface.
2222                          */
2223                         if (list_empty(&idev->addr_list))
2224                                 addrconf_ifdown(idev->dev, 1);
2225                         return 0;
2226                 }
2227         }
2228         read_unlock_bh(&idev->lock);
2229         return -EADDRNOTAVAIL;
2230 }
2231
2232
2233 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2234 {
2235         struct in6_ifreq ireq;
2236         int err;
2237
2238         if (!capable(CAP_NET_ADMIN))
2239                 return -EPERM;
2240
2241         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2242                 return -EFAULT;
2243
2244         rtnl_lock();
2245         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2246                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2247                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2248         rtnl_unlock();
2249         return err;
2250 }
2251
2252 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2253 {
2254         struct in6_ifreq ireq;
2255         int err;
2256
2257         if (!capable(CAP_NET_ADMIN))
2258                 return -EPERM;
2259
2260         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2261                 return -EFAULT;
2262
2263         rtnl_lock();
2264         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2265                              ireq.ifr6_prefixlen);
2266         rtnl_unlock();
2267         return err;
2268 }
2269
2270 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2271                      int plen, int scope)
2272 {
2273         struct inet6_ifaddr *ifp;
2274
2275         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2276         if (!IS_ERR(ifp)) {
2277                 spin_lock_bh(&ifp->lock);
2278                 ifp->flags &= ~IFA_F_TENTATIVE;
2279                 spin_unlock_bh(&ifp->lock);
2280                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2281                 in6_ifa_put(ifp);
2282         }
2283 }
2284
2285 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2286 static void sit_add_v4_addrs(struct inet6_dev *idev)
2287 {
2288         struct in6_addr addr;
2289         struct net_device *dev;
2290         struct net *net = dev_net(idev->dev);
2291         int scope;
2292
2293         ASSERT_RTNL();
2294
2295         memset(&addr, 0, sizeof(struct in6_addr));
2296         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2297
2298         if (idev->dev->flags&IFF_POINTOPOINT) {
2299                 addr.s6_addr32[0] = htonl(0xfe800000);
2300                 scope = IFA_LINK;
2301         } else {
2302                 scope = IPV6_ADDR_COMPATv4;
2303         }
2304
2305         if (addr.s6_addr32[3]) {
2306                 add_addr(idev, &addr, 128, scope);
2307                 return;
2308         }
2309
2310         for_each_netdev(net, dev) {
2311                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2312                 if (in_dev && (dev->flags & IFF_UP)) {
2313                         struct in_ifaddr * ifa;
2314
2315                         int flag = scope;
2316
2317                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2318                                 int plen;
2319
2320                                 addr.s6_addr32[3] = ifa->ifa_local;
2321
2322                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2323                                         continue;
2324                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2325                                         if (idev->dev->flags&IFF_POINTOPOINT)
2326                                                 continue;
2327                                         flag |= IFA_HOST;
2328                                 }
2329                                 if (idev->dev->flags&IFF_POINTOPOINT)
2330                                         plen = 64;
2331                                 else
2332                                         plen = 96;
2333
2334                                 add_addr(idev, &addr, plen, flag);
2335                         }
2336                 }
2337         }
2338 }
2339 #endif
2340
2341 static void init_loopback(struct net_device *dev)
2342 {
2343         struct inet6_dev  *idev;
2344
2345         /* ::1 */
2346
2347         ASSERT_RTNL();
2348
2349         if ((idev = ipv6_find_idev(dev)) == NULL) {
2350                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2351                 return;
2352         }
2353
2354         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2355 }
2356
2357 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2358 {
2359         struct inet6_ifaddr * ifp;
2360         u32 addr_flags = IFA_F_PERMANENT;
2361
2362 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2363         if (idev->cnf.optimistic_dad &&
2364             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2365                 addr_flags |= IFA_F_OPTIMISTIC;
2366 #endif
2367
2368
2369         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2370         if (!IS_ERR(ifp)) {
2371                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2372                 addrconf_dad_start(ifp, 0);
2373                 in6_ifa_put(ifp);
2374         }
2375 }
2376
2377 static void addrconf_dev_config(struct net_device *dev)
2378 {
2379         struct in6_addr addr;
2380         struct inet6_dev    * idev;
2381
2382         ASSERT_RTNL();
2383
2384         if ((dev->type != ARPHRD_ETHER) &&
2385             (dev->type != ARPHRD_FDDI) &&
2386             (dev->type != ARPHRD_IEEE802_TR) &&
2387             (dev->type != ARPHRD_ARCNET) &&
2388             (dev->type != ARPHRD_INFINIBAND)) {
2389                 /* Alas, we support only Ethernet autoconfiguration. */
2390                 return;
2391         }
2392
2393         idev = addrconf_add_dev(dev);
2394         if (IS_ERR(idev))
2395                 return;
2396
2397         memset(&addr, 0, sizeof(struct in6_addr));
2398         addr.s6_addr32[0] = htonl(0xFE800000);
2399
2400         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2401                 addrconf_add_linklocal(idev, &addr);
2402 }
2403
2404 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2405 static void addrconf_sit_config(struct net_device *dev)
2406 {
2407         struct inet6_dev *idev;
2408
2409         ASSERT_RTNL();
2410
2411         /*
2412          * Configure the tunnel with one of our IPv4
2413          * addresses... we should configure all of
2414          * our v4 addrs in the tunnel
2415          */
2416
2417         if ((idev = ipv6_find_idev(dev)) == NULL) {
2418                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2419                 return;
2420         }
2421
2422         if (dev->priv_flags & IFF_ISATAP) {
2423                 struct in6_addr addr;
2424
2425                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2426                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2427                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2428                         addrconf_add_linklocal(idev, &addr);
2429                 return;
2430         }
2431
2432         sit_add_v4_addrs(idev);
2433
2434         if (dev->flags&IFF_POINTOPOINT) {
2435                 addrconf_add_mroute(dev);
2436                 addrconf_add_lroute(dev);
2437         } else
2438                 sit_route_add(dev);
2439 }
2440 #endif
2441
2442 static inline int
2443 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2444 {
2445         struct in6_addr lladdr;
2446
2447         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2448                 addrconf_add_linklocal(idev, &lladdr);
2449                 return 0;
2450         }
2451         return -1;
2452 }
2453
2454 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2455 {
2456         struct net_device *link_dev;
2457         struct net *net = dev_net(idev->dev);
2458
2459         /* first try to inherit the link-local address from the link device */
2460         if (idev->dev->iflink &&
2461             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2462                 if (!ipv6_inherit_linklocal(idev, link_dev))
2463                         return;
2464         }
2465         /* then try to inherit it from any device */
2466         for_each_netdev(net, link_dev) {
2467                 if (!ipv6_inherit_linklocal(idev, link_dev))
2468                         return;
2469         }
2470         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2471 }
2472
2473 /*
2474  * Autoconfigure tunnel with a link-local address so routing protocols,
2475  * DHCPv6, MLD etc. can be run over the virtual link
2476  */
2477
2478 static void addrconf_ip6_tnl_config(struct net_device *dev)
2479 {
2480         struct inet6_dev *idev;
2481
2482         ASSERT_RTNL();
2483
2484         idev = addrconf_add_dev(dev);
2485         if (IS_ERR(idev)) {
2486                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2487                 return;
2488         }
2489         ip6_tnl_add_linklocal(idev);
2490 }
2491
2492 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2493                            void * data)
2494 {
2495         struct net_device *dev = (struct net_device *) data;
2496         struct inet6_dev *idev = __in6_dev_get(dev);
2497         int run_pending = 0;
2498         int err;
2499
2500         switch (event) {
2501         case NETDEV_REGISTER:
2502                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2503                         idev = ipv6_add_dev(dev);
2504                         if (!idev)
2505                                 return notifier_from_errno(-ENOMEM);
2506                 }
2507                 break;
2508
2509         case NETDEV_UP:
2510         case NETDEV_CHANGE:
2511                 if (dev->flags & IFF_SLAVE)
2512                         break;
2513
2514                 if (event == NETDEV_UP) {
2515                         if (!addrconf_qdisc_ok(dev)) {
2516                                 /* device is not ready yet. */
2517                                 printk(KERN_INFO
2518                                         "ADDRCONF(NETDEV_UP): %s: "
2519                                         "link is not ready\n",
2520                                         dev->name);
2521                                 break;
2522                         }
2523
2524                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2525                                 idev = ipv6_add_dev(dev);
2526
2527                         if (idev) {
2528                                 idev->if_flags |= IF_READY;
2529                                 run_pending = 1;
2530                         }
2531                 } else {
2532                         if (!addrconf_qdisc_ok(dev)) {
2533                                 /* device is still not ready. */
2534                                 break;
2535                         }
2536
2537                         if (idev) {
2538                                 if (idev->if_flags & IF_READY)
2539                                         /* device is already configured. */
2540                                         break;
2541                                 idev->if_flags |= IF_READY;
2542                         }
2543
2544                         printk(KERN_INFO
2545                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2546                                         "link becomes ready\n",
2547                                         dev->name);
2548
2549                         run_pending = 1;
2550                 }
2551
2552                 switch (dev->type) {
2553 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2554                 case ARPHRD_SIT:
2555                         addrconf_sit_config(dev);
2556                         break;
2557 #endif
2558                 case ARPHRD_TUNNEL6:
2559                         addrconf_ip6_tnl_config(dev);
2560                         break;
2561                 case ARPHRD_LOOPBACK:
2562                         init_loopback(dev);
2563                         break;
2564
2565                 default:
2566                         addrconf_dev_config(dev);
2567                         break;
2568                 }
2569
2570                 if (idev) {
2571                         if (run_pending)
2572                                 addrconf_dad_run(idev);
2573
2574                         /*
2575                          * If the MTU changed during the interface down,
2576                          * when the interface up, the changed MTU must be
2577                          * reflected in the idev as well as routers.
2578                          */
2579                         if (idev->cnf.mtu6 != dev->mtu &&
2580                             dev->mtu >= IPV6_MIN_MTU) {
2581                                 rt6_mtu_change(dev, dev->mtu);
2582                                 idev->cnf.mtu6 = dev->mtu;
2583                         }
2584                         idev->tstamp = jiffies;
2585                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2586
2587                         /*
2588                          * If the changed mtu during down is lower than
2589                          * IPV6_MIN_MTU stop IPv6 on this interface.
2590                          */
2591                         if (dev->mtu < IPV6_MIN_MTU)
2592                                 addrconf_ifdown(dev, 1);
2593                 }
2594                 break;
2595
2596         case NETDEV_CHANGEMTU:
2597                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2598                         rt6_mtu_change(dev, dev->mtu);
2599                         idev->cnf.mtu6 = dev->mtu;
2600                         break;
2601                 }
2602
2603                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2604                         idev = ipv6_add_dev(dev);
2605                         if (idev)
2606                                 break;
2607                 }
2608
2609                 /*
2610                  * MTU falled under IPV6_MIN_MTU.
2611                  * Stop IPv6 on this interface.
2612                  */
2613
2614         case NETDEV_DOWN:
2615         case NETDEV_UNREGISTER:
2616                 /*
2617                  *      Remove all addresses from this interface.
2618                  */
2619                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2620                 break;
2621
2622         case NETDEV_CHANGENAME:
2623                 if (idev) {
2624                         snmp6_unregister_dev(idev);
2625                         addrconf_sysctl_unregister(idev);
2626                         addrconf_sysctl_register(idev);
2627                         err = snmp6_register_dev(idev);
2628                         if (err)
2629                                 return notifier_from_errno(err);
2630                 }
2631                 break;
2632
2633         case NETDEV_PRE_TYPE_CHANGE:
2634         case NETDEV_POST_TYPE_CHANGE:
2635                 addrconf_type_change(dev, event);
2636                 break;
2637         }
2638
2639         return NOTIFY_OK;
2640 }
2641
2642 /*
2643  *      addrconf module should be notified of a device going up
2644  */
2645 static struct notifier_block ipv6_dev_notf = {
2646         .notifier_call = addrconf_notify,
2647 };
2648
2649 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2650 {
2651         struct inet6_dev *idev;
2652         ASSERT_RTNL();
2653
2654         idev = __in6_dev_get(dev);
2655
2656         if (event == NETDEV_POST_TYPE_CHANGE)
2657                 ipv6_mc_remap(idev);
2658         else if (event == NETDEV_PRE_TYPE_CHANGE)
2659                 ipv6_mc_unmap(idev);
2660 }
2661
2662 static int addrconf_ifdown(struct net_device *dev, int how)
2663 {
2664         struct net *net = dev_net(dev);
2665         struct inet6_dev *idev;
2666         struct inet6_ifaddr *ifa;
2667         LIST_HEAD(keep_list);
2668         int state;
2669
2670         ASSERT_RTNL();
2671
2672         /* Flush routes if device is being removed or it is not loopback */
2673         if (how || !(dev->flags & IFF_LOOPBACK))
2674                 rt6_ifdown(net, dev);
2675         neigh_ifdown(&nd_tbl, dev);
2676
2677         idev = __in6_dev_get(dev);
2678         if (idev == NULL)
2679                 return -ENODEV;
2680
2681         /*
2682          * Step 1: remove reference to ipv6 device from parent device.
2683          *         Do not dev_put!
2684          */
2685         if (how) {
2686                 idev->dead = 1;
2687
2688                 /* protected by rtnl_lock */
2689                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2690
2691                 /* Step 1.5: remove snmp6 entry */
2692                 snmp6_unregister_dev(idev);
2693
2694         }
2695
2696         write_lock_bh(&idev->lock);
2697
2698         /* Step 2: clear flags for stateless addrconf */
2699         if (!how)
2700                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2701
2702 #ifdef CONFIG_IPV6_PRIVACY
2703         if (how && del_timer(&idev->regen_timer))
2704                 in6_dev_put(idev);
2705
2706         /* Step 3: clear tempaddr list */
2707         while (!list_empty(&idev->tempaddr_list)) {
2708                 ifa = list_first_entry(&idev->tempaddr_list,
2709                                        struct inet6_ifaddr, tmp_list);
2710                 list_del(&ifa->tmp_list);
2711                 write_unlock_bh(&idev->lock);
2712                 spin_lock_bh(&ifa->lock);
2713
2714                 if (ifa->ifpub) {
2715                         in6_ifa_put(ifa->ifpub);
2716                         ifa->ifpub = NULL;
2717                 }
2718                 spin_unlock_bh(&ifa->lock);
2719                 in6_ifa_put(ifa);
2720                 write_lock_bh(&idev->lock);
2721         }
2722 #endif
2723
2724         while (!list_empty(&idev->addr_list)) {
2725                 ifa = list_first_entry(&idev->addr_list,
2726                                        struct inet6_ifaddr, if_list);
2727                 addrconf_del_timer(ifa);
2728
2729                 /* If just doing link down, and address is permanent
2730                    and not link-local, then retain it. */
2731                 if (!how &&
2732                     (ifa->flags&IFA_F_PERMANENT) &&
2733                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2734                         list_move_tail(&ifa->if_list, &keep_list);
2735
2736                         /* If not doing DAD on this address, just keep it. */
2737                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2738                             idev->cnf.accept_dad <= 0 ||
2739                             (ifa->flags & IFA_F_NODAD))
2740                                 continue;
2741
2742                         /* If it was tentative already, no need to notify */
2743                         if (ifa->flags & IFA_F_TENTATIVE)
2744                                 continue;
2745
2746                         /* Flag it for later restoration when link comes up */
2747                         ifa->flags |= IFA_F_TENTATIVE;
2748                         ifa->state = INET6_IFADDR_STATE_DAD;
2749                 } else {
2750                         list_del(&ifa->if_list);
2751
2752                         /* clear hash table */
2753                         spin_lock_bh(&addrconf_hash_lock);
2754                         hlist_del_init_rcu(&ifa->addr_lst);
2755                         spin_unlock_bh(&addrconf_hash_lock);
2756
2757                         write_unlock_bh(&idev->lock);
2758                         spin_lock_bh(&ifa->state_lock);
2759                         state = ifa->state;
2760                         ifa->state = INET6_IFADDR_STATE_DEAD;
2761                         spin_unlock_bh(&ifa->state_lock);
2762
2763                         if (state != INET6_IFADDR_STATE_DEAD) {
2764                                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2765                                 atomic_notifier_call_chain(&inet6addr_chain,
2766                                                            NETDEV_DOWN, ifa);
2767                         }
2768
2769                         in6_ifa_put(ifa);
2770                         write_lock_bh(&idev->lock);
2771                 }
2772         }
2773
2774         list_splice(&keep_list, &idev->addr_list);
2775
2776         write_unlock_bh(&idev->lock);
2777
2778         /* Step 5: Discard multicast list */
2779         if (how)
2780                 ipv6_mc_destroy_dev(idev);
2781         else
2782                 ipv6_mc_down(idev);
2783
2784         idev->tstamp = jiffies;
2785
2786         /* Last: Shot the device (if unregistered) */
2787         if (how) {
2788                 addrconf_sysctl_unregister(idev);
2789                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2790                 neigh_ifdown(&nd_tbl, dev);
2791                 in6_dev_put(idev);
2792         }
2793         return 0;
2794 }
2795
2796 static void addrconf_rs_timer(unsigned long data)
2797 {
2798         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2799         struct inet6_dev *idev = ifp->idev;
2800
2801         read_lock(&idev->lock);
2802         if (idev->dead || !(idev->if_flags & IF_READY))
2803                 goto out;
2804
2805         if (idev->cnf.forwarding)
2806                 goto out;
2807
2808         /* Announcement received after solicitation was sent */
2809         if (idev->if_flags & IF_RA_RCVD)
2810                 goto out;
2811
2812         spin_lock(&ifp->lock);
2813         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2814                 /* The wait after the last probe can be shorter */
2815                 addrconf_mod_timer(ifp, AC_RS,
2816                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2817                                    idev->cnf.rtr_solicit_delay :
2818                                    idev->cnf.rtr_solicit_interval);
2819                 spin_unlock(&ifp->lock);
2820
2821                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2822         } else {
2823                 spin_unlock(&ifp->lock);
2824                 /*
2825                  * Note: we do not support deprecated "all on-link"
2826                  * assumption any longer.
2827                  */
2828                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2829                        idev->dev->name);
2830         }
2831
2832 out:
2833         read_unlock(&idev->lock);
2834         in6_ifa_put(ifp);
2835 }
2836
2837 /*
2838  *      Duplicate Address Detection
2839  */
2840 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2841 {
2842         unsigned long rand_num;
2843         struct inet6_dev *idev = ifp->idev;
2844
2845         if (ifp->flags & IFA_F_OPTIMISTIC)
2846                 rand_num = 0;
2847         else
2848                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2849
2850         ifp->probes = idev->cnf.dad_transmits;
2851         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2852 }
2853
2854 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2855 {
2856         struct inet6_dev *idev = ifp->idev;
2857         struct net_device *dev = idev->dev;
2858
2859         addrconf_join_solict(dev, &ifp->addr);
2860
2861         net_srandom(ifp->addr.s6_addr32[3]);
2862
2863         read_lock_bh(&idev->lock);
2864         spin_lock(&ifp->lock);
2865         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2866                 goto out;
2867
2868         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2869             idev->cnf.accept_dad < 1 ||
2870             !(ifp->flags&IFA_F_TENTATIVE) ||
2871             ifp->flags & IFA_F_NODAD) {
2872                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2873                 spin_unlock(&ifp->lock);
2874                 read_unlock_bh(&idev->lock);
2875
2876                 addrconf_dad_completed(ifp);
2877                 return;
2878         }
2879
2880         if (!(idev->if_flags & IF_READY)) {
2881                 spin_unlock(&ifp->lock);
2882                 read_unlock_bh(&idev->lock);
2883                 /*
2884                  * If the device is not ready:
2885                  * - keep it tentative if it is a permanent address.
2886                  * - otherwise, kill it.
2887                  */
2888                 in6_ifa_hold(ifp);
2889                 addrconf_dad_stop(ifp, 0);
2890                 return;
2891         }
2892
2893         /*
2894          * Optimistic nodes can start receiving
2895          * Frames right away
2896          */
2897         if (ifp->flags & IFA_F_OPTIMISTIC)
2898                 ip6_ins_rt(ifp->rt);
2899
2900         addrconf_dad_kick(ifp);
2901 out:
2902         spin_unlock(&ifp->lock);
2903         read_unlock_bh(&idev->lock);
2904 }
2905
2906 static void addrconf_dad_timer(unsigned long data)
2907 {
2908         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2909         struct inet6_dev *idev = ifp->idev;
2910         struct in6_addr mcaddr;
2911
2912         if (!ifp->probes && addrconf_dad_end(ifp))
2913                 goto out;
2914
2915         read_lock(&idev->lock);
2916         if (idev->dead || !(idev->if_flags & IF_READY)) {
2917                 read_unlock(&idev->lock);
2918                 goto out;
2919         }
2920
2921         spin_lock(&ifp->lock);
2922         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2923                 spin_unlock(&ifp->lock);
2924                 read_unlock(&idev->lock);
2925                 goto out;
2926         }
2927
2928         if (ifp->probes == 0) {
2929                 /*
2930                  * DAD was successful
2931                  */
2932
2933                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2934                 spin_unlock(&ifp->lock);
2935                 read_unlock(&idev->lock);
2936
2937                 addrconf_dad_completed(ifp);
2938
2939                 goto out;
2940         }
2941
2942         ifp->probes--;
2943         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2944         spin_unlock(&ifp->lock);
2945         read_unlock(&idev->lock);
2946
2947         /* send a neighbour solicitation for our addr */
2948         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2949         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2950 out:
2951         in6_ifa_put(ifp);
2952 }
2953
2954 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2955 {
2956         struct net_device *dev = ifp->idev->dev;
2957
2958         /*
2959          *      Configure the address for reception. Now it is valid.
2960          */
2961
2962         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2963
2964         /* If added prefix is link local and forwarding is off,
2965            start sending router solicitations.
2966          */
2967
2968         if ((ifp->idev->cnf.forwarding == 0 ||
2969              ifp->idev->cnf.forwarding == 2) &&
2970             ifp->idev->cnf.rtr_solicits > 0 &&
2971             (dev->flags&IFF_LOOPBACK) == 0 &&
2972             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2973                 /*
2974                  *      If a host as already performed a random delay
2975                  *      [...] as part of DAD [...] there is no need
2976                  *      to delay again before sending the first RS
2977                  */
2978                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2979
2980                 spin_lock_bh(&ifp->lock);
2981                 ifp->probes = 1;
2982                 ifp->idev->if_flags |= IF_RS_SENT;
2983                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2984                 spin_unlock_bh(&ifp->lock);
2985         }
2986 }
2987
2988 static void addrconf_dad_run(struct inet6_dev *idev)
2989 {
2990         struct inet6_ifaddr *ifp;
2991
2992         read_lock_bh(&idev->lock);
2993         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2994                 spin_lock(&ifp->lock);
2995                 if (ifp->flags & IFA_F_TENTATIVE &&
2996                     ifp->state == INET6_IFADDR_STATE_DAD)
2997                         addrconf_dad_kick(ifp);
2998                 spin_unlock(&ifp->lock);
2999         }
3000         read_unlock_bh(&idev->lock);
3001 }
3002
3003 #ifdef CONFIG_PROC_FS
3004 struct if6_iter_state {
3005         struct seq_net_private p;
3006         int bucket;
3007 };
3008
3009 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3010 {
3011         struct inet6_ifaddr *ifa = NULL;
3012         struct if6_iter_state *state = seq->private;
3013         struct net *net = seq_file_net(seq);
3014
3015         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3016                 struct hlist_node *n;
3017                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3018                                          addr_lst)
3019                         if (net_eq(dev_net(ifa->idev->dev), net))
3020                                 return ifa;
3021         }
3022         return NULL;
3023 }
3024
3025 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3026                                          struct inet6_ifaddr *ifa)
3027 {
3028         struct if6_iter_state *state = seq->private;
3029         struct net *net = seq_file_net(seq);
3030         struct hlist_node *n = &ifa->addr_lst;
3031
3032         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3033                 if (net_eq(dev_net(ifa->idev->dev), net))
3034                         return ifa;
3035
3036         while (++state->bucket < IN6_ADDR_HSIZE) {
3037                 hlist_for_each_entry_rcu_bh(ifa, n,
3038                                      &inet6_addr_lst[state->bucket], addr_lst) {
3039                         if (net_eq(dev_net(ifa->idev->dev), net))
3040                                 return ifa;
3041                 }
3042         }
3043
3044         return NULL;
3045 }
3046
3047 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3048 {
3049         struct inet6_ifaddr *ifa = if6_get_first(seq);
3050
3051         if (ifa)
3052                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3053                         --pos;
3054         return pos ? NULL : ifa;
3055 }
3056
3057 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3058         __acquires(rcu_bh)
3059 {
3060         rcu_read_lock_bh();
3061         return if6_get_idx(seq, *pos);
3062 }
3063
3064 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3065 {
3066         struct inet6_ifaddr *ifa;
3067
3068         ifa = if6_get_next(seq, v);
3069         ++*pos;
3070         return ifa;
3071 }
3072
3073 static void if6_seq_stop(struct seq_file *seq, void *v)
3074         __releases(rcu_bh)
3075 {
3076         rcu_read_unlock_bh();
3077 }
3078
3079 static int if6_seq_show(struct seq_file *seq, void *v)
3080 {
3081         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3082         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3083                    &ifp->addr,
3084                    ifp->idev->dev->ifindex,
3085                    ifp->prefix_len,
3086                    ifp->scope,
3087                    ifp->flags,
3088                    ifp->idev->dev->name);
3089         return 0;
3090 }
3091
3092 static const struct seq_operations if6_seq_ops = {
3093         .start  = if6_seq_start,
3094         .next   = if6_seq_next,
3095         .show   = if6_seq_show,
3096         .stop   = if6_seq_stop,
3097 };
3098
3099 static int if6_seq_open(struct inode *inode, struct file *file)
3100 {
3101         return seq_open_net(inode, file, &if6_seq_ops,
3102                             sizeof(struct if6_iter_state));
3103 }
3104
3105 static const struct file_operations if6_fops = {
3106         .owner          = THIS_MODULE,
3107         .open           = if6_seq_open,
3108         .read           = seq_read,
3109         .llseek         = seq_lseek,
3110         .release        = seq_release_net,
3111 };
3112
3113 static int __net_init if6_proc_net_init(struct net *net)
3114 {
3115         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3116                 return -ENOMEM;
3117         return 0;
3118 }
3119
3120 static void __net_exit if6_proc_net_exit(struct net *net)
3121 {
3122        proc_net_remove(net, "if_inet6");
3123 }
3124
3125 static struct pernet_operations if6_proc_net_ops = {
3126        .init = if6_proc_net_init,
3127        .exit = if6_proc_net_exit,
3128 };
3129
3130 int __init if6_proc_init(void)
3131 {
3132         return register_pernet_subsys(&if6_proc_net_ops);
3133 }
3134
3135 void if6_proc_exit(void)
3136 {
3137         unregister_pernet_subsys(&if6_proc_net_ops);
3138 }
3139 #endif  /* CONFIG_PROC_FS */
3140
3141 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3142 /* Check if address is a home address configured on any interface. */
3143 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3144 {
3145         int ret = 0;
3146         struct inet6_ifaddr *ifp = NULL;
3147         struct hlist_node *n;
3148         unsigned int hash = ipv6_addr_hash(addr);
3149
3150         rcu_read_lock_bh();
3151         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3152                 if (!net_eq(dev_net(ifp->idev->dev), net))
3153                         continue;
3154                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3155                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3156                         ret = 1;
3157                         break;
3158                 }
3159         }
3160         rcu_read_unlock_bh();
3161         return ret;
3162 }
3163 #endif
3164
3165 /*
3166  *      Periodic address status verification
3167  */
3168
3169 static void addrconf_verify(unsigned long foo)
3170 {
3171         unsigned long now, next, next_sec, next_sched;
3172         struct inet6_ifaddr *ifp;
3173         struct hlist_node *node;
3174         int i;
3175
3176         rcu_read_lock_bh();
3177         spin_lock(&addrconf_verify_lock);
3178         now = jiffies;
3179         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3180
3181         del_timer(&addr_chk_timer);
3182
3183         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3184 restart:
3185                 hlist_for_each_entry_rcu_bh(ifp, node,
3186                                          &inet6_addr_lst[i], addr_lst) {
3187                         unsigned long age;
3188
3189                         if (ifp->flags & IFA_F_PERMANENT)
3190                                 continue;
3191
3192                         spin_lock(&ifp->lock);
3193                         /* We try to batch several events at once. */
3194                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3195
3196                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3197                             age >= ifp->valid_lft) {
3198                                 spin_unlock(&ifp->lock);
3199                                 in6_ifa_hold(ifp);
3200                                 ipv6_del_addr(ifp);
3201                                 goto restart;
3202                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3203                                 spin_unlock(&ifp->lock);
3204                                 continue;
3205                         } else if (age >= ifp->prefered_lft) {
3206                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3207                                 int deprecate = 0;
3208
3209                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3210                                         deprecate = 1;
3211                                         ifp->flags |= IFA_F_DEPRECATED;
3212                                 }
3213
3214                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3215                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3216
3217                                 spin_unlock(&ifp->lock);
3218
3219                                 if (deprecate) {
3220                                         in6_ifa_hold(ifp);
3221
3222                                         ipv6_ifa_notify(0, ifp);
3223                                         in6_ifa_put(ifp);
3224                                         goto restart;
3225                                 }
3226 #ifdef CONFIG_IPV6_PRIVACY
3227                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3228                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3229                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3230                                         ifp->idev->cnf.dad_transmits *
3231                                         ifp->idev->nd_parms->retrans_time / HZ;
3232
3233                                 if (age >= ifp->prefered_lft - regen_advance) {
3234                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3235                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3236                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3237                                         if (!ifp->regen_count && ifpub) {
3238                                                 ifp->regen_count++;
3239                                                 in6_ifa_hold(ifp);
3240                                                 in6_ifa_hold(ifpub);
3241                                                 spin_unlock(&ifp->lock);
3242
3243                                                 spin_lock(&ifpub->lock);
3244                                                 ifpub->regen_count = 0;
3245                                                 spin_unlock(&ifpub->lock);
3246                                                 ipv6_create_tempaddr(ifpub, ifp);
3247                                                 in6_ifa_put(ifpub);
3248                                                 in6_ifa_put(ifp);
3249                                                 goto restart;
3250                                         }
3251                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3252                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3253                                 spin_unlock(&ifp->lock);
3254 #endif
3255                         } else {
3256                                 /* ifp->prefered_lft <= ifp->valid_lft */
3257                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3258                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3259                                 spin_unlock(&ifp->lock);
3260                         }
3261                 }
3262         }
3263
3264         next_sec = round_jiffies_up(next);
3265         next_sched = next;
3266
3267         /* If rounded timeout is accurate enough, accept it. */
3268         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3269                 next_sched = next_sec;
3270
3271         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3272         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3273                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3274
3275         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3276               now, next, next_sec, next_sched));
3277
3278         addr_chk_timer.expires = next_sched;
3279         add_timer(&addr_chk_timer);
3280         spin_unlock(&addrconf_verify_lock);
3281         rcu_read_unlock_bh();
3282 }
3283
3284 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3285 {
3286         struct in6_addr *pfx = NULL;
3287
3288         if (addr)
3289                 pfx = nla_data(addr);
3290
3291         if (local) {
3292                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3293                         pfx = NULL;
3294                 else
3295                         pfx = nla_data(local);
3296         }
3297
3298         return pfx;
3299 }
3300
3301 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3302         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3303         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3304         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3305 };
3306
3307 static int
3308 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3309 {
3310         struct net *net = sock_net(skb->sk);
3311         struct ifaddrmsg *ifm;
3312         struct nlattr *tb[IFA_MAX+1];
3313         struct in6_addr *pfx;
3314         int err;
3315
3316         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3317         if (err < 0)
3318                 return err;
3319
3320         ifm = nlmsg_data(nlh);
3321         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3322         if (pfx == NULL)
3323                 return -EINVAL;
3324
3325         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3326 }
3327
3328 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3329                              u32 prefered_lft, u32 valid_lft)
3330 {
3331         u32 flags;
3332         clock_t expires;
3333         unsigned long timeout;
3334
3335         if (!valid_lft || (prefered_lft > valid_lft))
3336                 return -EINVAL;
3337
3338         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3339         if (addrconf_finite_timeout(timeout)) {
3340                 expires = jiffies_to_clock_t(timeout * HZ);
3341                 valid_lft = timeout;
3342                 flags = RTF_EXPIRES;
3343         } else {
3344                 expires = 0;
3345                 flags = 0;
3346                 ifa_flags |= IFA_F_PERMANENT;
3347         }
3348
3349         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3350         if (addrconf_finite_timeout(timeout)) {
3351                 if (timeout == 0)
3352                         ifa_flags |= IFA_F_DEPRECATED;
3353                 prefered_lft = timeout;
3354         }
3355
3356         spin_lock_bh(&ifp->lock);
3357         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3358         ifp->tstamp = jiffies;
3359         ifp->valid_lft = valid_lft;
3360         ifp->prefered_lft = prefered_lft;
3361
3362         spin_unlock_bh(&ifp->lock);
3363         if (!(ifp->flags&IFA_F_TENTATIVE))
3364                 ipv6_ifa_notify(0, ifp);
3365
3366         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3367                               expires, flags);
3368         addrconf_verify(0);
3369
3370         return 0;
3371 }
3372
3373 static int
3374 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3375 {
3376         struct net *net = sock_net(skb->sk);
3377         struct ifaddrmsg *ifm;
3378         struct nlattr *tb[IFA_MAX+1];
3379         struct in6_addr *pfx;
3380         struct inet6_ifaddr *ifa;
3381         struct net_device *dev;
3382         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3383         u8 ifa_flags;
3384         int err;
3385
3386         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3387         if (err < 0)
3388                 return err;
3389
3390         ifm = nlmsg_data(nlh);
3391         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3392         if (pfx == NULL)
3393                 return -EINVAL;
3394
3395         if (tb[IFA_CACHEINFO]) {
3396                 struct ifa_cacheinfo *ci;
3397
3398                 ci = nla_data(tb[IFA_CACHEINFO]);
3399                 valid_lft = ci->ifa_valid;
3400                 preferred_lft = ci->ifa_prefered;
3401         } else {
3402                 preferred_lft = INFINITY_LIFE_TIME;
3403                 valid_lft = INFINITY_LIFE_TIME;
3404         }
3405
3406         dev =  __dev_get_by_index(net, ifm->ifa_index);
3407         if (dev == NULL)
3408                 return -ENODEV;
3409
3410         /* We ignore other flags so far. */
3411         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3412
3413         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3414         if (ifa == NULL) {
3415                 /*
3416                  * It would be best to check for !NLM_F_CREATE here but
3417                  * userspace alreay relies on not having to provide this.
3418                  */
3419                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3420                                       ifm->ifa_prefixlen, ifa_flags,
3421                                       preferred_lft, valid_lft);
3422         }
3423
3424         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3425             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3426                 err = -EEXIST;
3427         else
3428                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3429
3430         in6_ifa_put(ifa);
3431
3432         return err;
3433 }
3434
3435 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3436                           u8 scope, int ifindex)
3437 {
3438         struct ifaddrmsg *ifm;
3439
3440         ifm = nlmsg_data(nlh);
3441         ifm->ifa_family = AF_INET6;
3442         ifm->ifa_prefixlen = prefixlen;
3443         ifm->ifa_flags = flags;
3444         ifm->ifa_scope = scope;
3445         ifm->ifa_index = ifindex;
3446 }
3447
3448 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3449                          unsigned long tstamp, u32 preferred, u32 valid)
3450 {
3451         struct ifa_cacheinfo ci;
3452
3453         ci.cstamp = cstamp_delta(cstamp);
3454         ci.tstamp = cstamp_delta(tstamp);
3455         ci.ifa_prefered = preferred;
3456         ci.ifa_valid = valid;
3457
3458         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3459 }
3460
3461 static inline int rt_scope(int ifa_scope)
3462 {
3463         if (ifa_scope & IFA_HOST)
3464                 return RT_SCOPE_HOST;
3465         else if (ifa_scope & IFA_LINK)
3466                 return RT_SCOPE_LINK;
3467         else if (ifa_scope & IFA_SITE)
3468                 return RT_SCOPE_SITE;
3469         else
3470                 return RT_SCOPE_UNIVERSE;
3471 }
3472
3473 static inline int inet6_ifaddr_msgsize(void)
3474 {
3475         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3476                + nla_total_size(16) /* IFA_ADDRESS */
3477                + nla_total_size(sizeof(struct ifa_cacheinfo));
3478 }
3479
3480 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3481                              u32 pid, u32 seq, int event, unsigned int flags)
3482 {
3483         struct nlmsghdr  *nlh;
3484         u32 preferred, valid;
3485
3486         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3487         if (nlh == NULL)
3488                 return -EMSGSIZE;
3489
3490         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3491                       ifa->idev->dev->ifindex);
3492
3493         if (!(ifa->flags&IFA_F_PERMANENT)) {
3494                 preferred = ifa->prefered_lft;
3495                 valid = ifa->valid_lft;
3496                 if (preferred != INFINITY_LIFE_TIME) {
3497                         long tval = (jiffies - ifa->tstamp)/HZ;
3498                         if (preferred > tval)
3499                                 preferred -= tval;
3500                         else
3501                                 preferred = 0;
3502                         if (valid != INFINITY_LIFE_TIME) {
3503                                 if (valid > tval)
3504                                         valid -= tval;
3505                                 else
3506                                         valid = 0;
3507                         }
3508                 }
3509         } else {
3510                 preferred = INFINITY_LIFE_TIME;
3511                 valid = INFINITY_LIFE_TIME;
3512         }
3513
3514         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3515             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3516                 nlmsg_cancel(skb, nlh);
3517                 return -EMSGSIZE;
3518         }
3519
3520         return nlmsg_end(skb, nlh);
3521 }
3522
3523 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3524                                 u32 pid, u32 seq, int event, u16 flags)
3525 {
3526         struct nlmsghdr  *nlh;
3527         u8 scope = RT_SCOPE_UNIVERSE;
3528         int ifindex = ifmca->idev->dev->ifindex;
3529
3530         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3531                 scope = RT_SCOPE_SITE;
3532
3533         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3534         if (nlh == NULL)
3535                 return -EMSGSIZE;
3536
3537         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3538         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3539             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3540                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3541                 nlmsg_cancel(skb, nlh);
3542                 return -EMSGSIZE;
3543         }
3544
3545         return nlmsg_end(skb, nlh);
3546 }
3547
3548 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3549                                 u32 pid, u32 seq, int event, unsigned int flags)
3550 {
3551         struct nlmsghdr  *nlh;
3552         u8 scope = RT_SCOPE_UNIVERSE;
3553         int ifindex = ifaca->aca_idev->dev->ifindex;
3554
3555         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3556                 scope = RT_SCOPE_SITE;
3557
3558         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3559         if (nlh == NULL)
3560                 return -EMSGSIZE;
3561
3562         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3563         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3564             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3565                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3566                 nlmsg_cancel(skb, nlh);
3567                 return -EMSGSIZE;
3568         }
3569
3570         return nlmsg_end(skb, nlh);
3571 }
3572
3573 enum addr_type_t {
3574         UNICAST_ADDR,
3575         MULTICAST_ADDR,
3576         ANYCAST_ADDR,
3577 };
3578
3579 /* called with rcu_read_lock() */
3580 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3581                           struct netlink_callback *cb, enum addr_type_t type,
3582                           int s_ip_idx, int *p_ip_idx)
3583 {
3584         struct ifmcaddr6 *ifmca;
3585         struct ifacaddr6 *ifaca;
3586         int err = 1;
3587         int ip_idx = *p_ip_idx;
3588
3589         read_lock_bh(&idev->lock);
3590         switch (type) {
3591         case UNICAST_ADDR: {
3592                 struct inet6_ifaddr *ifa;
3593
3594                 /* unicast address incl. temp addr */
3595                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3596                         if (++ip_idx < s_ip_idx)
3597                                 continue;
3598                         err = inet6_fill_ifaddr(skb, ifa,
3599                                                 NETLINK_CB(cb->skb).pid,
3600                                                 cb->nlh->nlmsg_seq,
3601                                                 RTM_NEWADDR,
3602                                                 NLM_F_MULTI);
3603                         if (err <= 0)
3604                                 break;
3605                 }
3606                 break;
3607         }
3608         case MULTICAST_ADDR:
3609                 /* multicast address */
3610                 for (ifmca = idev->mc_list; ifmca;
3611                      ifmca = ifmca->next, ip_idx++) {
3612                         if (ip_idx < s_ip_idx)
3613                                 continue;
3614                         err = inet6_fill_ifmcaddr(skb, ifmca,
3615                                                   NETLINK_CB(cb->skb).pid,
3616                                                   cb->nlh->nlmsg_seq,
3617                                                   RTM_GETMULTICAST,
3618                                                   NLM_F_MULTI);
3619                         if (err <= 0)
3620                                 break;
3621                 }
3622                 break;
3623         case ANYCAST_ADDR:
3624                 /* anycast address */
3625                 for (ifaca = idev->ac_list; ifaca;
3626                      ifaca = ifaca->aca_next, ip_idx++) {
3627                         if (ip_idx < s_ip_idx)
3628                                 continue;
3629                         err = inet6_fill_ifacaddr(skb, ifaca,
3630                                                   NETLINK_CB(cb->skb).pid,
3631                                                   cb->nlh->nlmsg_seq,
3632                                                   RTM_GETANYCAST,
3633                                                   NLM_F_MULTI);
3634                         if (err <= 0)
3635                                 break;
3636                 }
3637                 break;
3638         default:
3639                 break;
3640         }
3641         read_unlock_bh(&idev->lock);
3642         *p_ip_idx = ip_idx;
3643         return err;
3644 }
3645
3646 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3647                            enum addr_type_t type)
3648 {
3649         struct net *net = sock_net(skb->sk);
3650         int h, s_h;
3651         int idx, ip_idx;
3652         int s_idx, s_ip_idx;
3653         struct net_device *dev;
3654         struct inet6_dev *idev;
3655         struct hlist_head *head;
3656         struct hlist_node *node;
3657
3658         s_h = cb->args[0];
3659         s_idx = idx = cb->args[1];
3660         s_ip_idx = ip_idx = cb->args[2];
3661
3662         rcu_read_lock();
3663         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3664                 idx = 0;
3665                 head = &net->dev_index_head[h];
3666                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3667                         if (idx < s_idx)
3668                                 goto cont;
3669                         if (h > s_h || idx > s_idx)
3670                                 s_ip_idx = 0;
3671                         ip_idx = 0;
3672                         idev = __in6_dev_get(dev);
3673                         if (!idev)
3674                                 goto cont;
3675
3676                         if (in6_dump_addrs(idev, skb, cb, type,
3677                                            s_ip_idx, &ip_idx) <= 0)
3678                                 goto done;
3679 cont:
3680                         idx++;
3681                 }
3682         }
3683 done:
3684         rcu_read_unlock();
3685         cb->args[0] = h;
3686         cb->args[1] = idx;
3687         cb->args[2] = ip_idx;
3688
3689         return skb->len;
3690 }
3691
3692 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3693 {
3694         enum addr_type_t type = UNICAST_ADDR;
3695
3696         return inet6_dump_addr(skb, cb, type);
3697 }
3698
3699 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3700 {
3701         enum addr_type_t type = MULTICAST_ADDR;
3702
3703         return inet6_dump_addr(skb, cb, type);
3704 }
3705
3706
3707 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3708 {
3709         enum addr_type_t type = ANYCAST_ADDR;
3710
3711         return inet6_dump_addr(skb, cb, type);
3712 }
3713
3714 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3715                              void *arg)
3716 {
3717         struct net *net = sock_net(in_skb->sk);
3718         struct ifaddrmsg *ifm;
3719         struct nlattr *tb[IFA_MAX+1];
3720         struct in6_addr *addr = NULL;
3721         struct net_device *dev = NULL;
3722         struct inet6_ifaddr *ifa;
3723         struct sk_buff *skb;
3724         int err;
3725
3726         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3727         if (err < 0)
3728                 goto errout;
3729
3730         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3731         if (addr == NULL) {
3732                 err = -EINVAL;
3733                 goto errout;
3734         }
3735
3736         ifm = nlmsg_data(nlh);
3737         if (ifm->ifa_index)
3738                 dev = __dev_get_by_index(net, ifm->ifa_index);
3739
3740         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3741         if (!ifa) {
3742                 err = -EADDRNOTAVAIL;
3743                 goto errout;
3744         }
3745
3746         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3747         if (!skb) {
3748                 err = -ENOBUFS;
3749                 goto errout_ifa;
3750         }
3751
3752         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3753                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3754         if (err < 0) {
3755                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3756                 WARN_ON(err == -EMSGSIZE);
3757                 kfree_skb(skb);
3758                 goto errout_ifa;
3759         }
3760         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3761 errout_ifa:
3762         in6_ifa_put(ifa);
3763 errout:
3764         return err;
3765 }
3766
3767 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3768 {
3769         struct sk_buff *skb;
3770         struct net *net = dev_net(ifa->idev->dev);
3771         int err = -ENOBUFS;
3772
3773         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3774         if (skb == NULL)
3775                 goto errout;
3776
3777         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3778         if (err < 0) {
3779                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3780                 WARN_ON(err == -EMSGSIZE);
3781                 kfree_skb(skb);
3782                 goto errout;
3783         }
3784         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3785         return;
3786 errout:
3787         if (err < 0)
3788                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3789 }
3790
3791 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3792                                 __s32 *array, int bytes)
3793 {
3794         BUG_ON(bytes < (DEVCONF_MAX * 4));
3795
3796         memset(array, 0, bytes);
3797         array[DEVCONF_FORWARDING] = cnf->forwarding;
3798         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3799         array[DEVCONF_MTU6] = cnf->mtu6;
3800         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3801         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3802         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3803         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3804         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3805         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3806                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3807         array[DEVCONF_RTR_SOLICIT_DELAY] =
3808                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3809         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3810 #ifdef CONFIG_IPV6_PRIVACY
3811         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3812         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3813         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3814         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3815         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3816 #endif
3817         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3818         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3819         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3820 #ifdef CONFIG_IPV6_ROUTER_PREF
3821         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3822         array[DEVCONF_RTR_PROBE_INTERVAL] =
3823                 jiffies_to_msecs(cnf->rtr_probe_interval);
3824 #ifdef CONFIG_IPV6_ROUTE_INFO
3825         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3826 #endif
3827 #endif
3828         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3829         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3830 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3831         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3832 #endif
3833 #ifdef CONFIG_IPV6_MROUTE
3834         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3835 #endif
3836         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3837         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3838         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3839 }
3840
3841 static inline size_t inet6_if_nlmsg_size(void)
3842 {
3843         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3844                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3845                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3846                + nla_total_size(4) /* IFLA_MTU */
3847                + nla_total_size(4) /* IFLA_LINK */
3848                + nla_total_size( /* IFLA_PROTINFO */
3849                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3850                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3851                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3852                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3853                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3854                  );
3855 }
3856
3857 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3858                                       int items, int bytes)
3859 {
3860         int i;
3861         int pad = bytes - sizeof(u64) * items;
3862         BUG_ON(pad < 0);
3863
3864         /* Use put_unaligned() because stats may not be aligned for u64. */
3865         put_unaligned(items, &stats[0]);
3866         for (i = 1; i < items; i++)
3867                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3868
3869         memset(&stats[items], 0, pad);
3870 }
3871
3872 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3873                                       int items, int bytes, size_t syncpoff)
3874 {
3875         int i;
3876         int pad = bytes - sizeof(u64) * items;
3877         BUG_ON(pad < 0);
3878
3879         /* Use put_unaligned() because stats may not be aligned for u64. */
3880         put_unaligned(items, &stats[0]);
3881         for (i = 1; i < items; i++)
3882                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3883
3884         memset(&stats[items], 0, pad);
3885 }
3886
3887 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3888                              int bytes)
3889 {
3890         switch (attrtype) {
3891         case IFLA_INET6_STATS:
3892                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3893                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3894                 break;
3895         case IFLA_INET6_ICMP6STATS:
3896                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3897                 break;
3898         }
3899 }
3900
3901 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3902                              u32 pid, u32 seq, int event, unsigned int flags)
3903 {
3904         struct net_device *dev = idev->dev;
3905         struct nlattr *nla;
3906         struct ifinfomsg *hdr;
3907         struct nlmsghdr *nlh;
3908         void *protoinfo;
3909         struct ifla_cacheinfo ci;
3910
3911         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3912         if (nlh == NULL)
3913                 return -EMSGSIZE;
3914
3915         hdr = nlmsg_data(nlh);
3916         hdr->ifi_family = AF_INET6;
3917         hdr->__ifi_pad = 0;
3918         hdr->ifi_type = dev->type;
3919         hdr->ifi_index = dev->ifindex;
3920         hdr->ifi_flags = dev_get_flags(dev);
3921         hdr->ifi_change = 0;
3922
3923         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3924
3925         if (dev->addr_len)
3926                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3927
3928         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3929         if (dev->ifindex != dev->iflink)
3930                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3931
3932         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3933         if (protoinfo == NULL)
3934                 goto nla_put_failure;
3935
3936         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3937
3938         ci.max_reasm_len = IPV6_MAXPLEN;
3939         ci.tstamp = cstamp_delta(idev->tstamp);
3940         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3941         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3942         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3943
3944         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3945         if (nla == NULL)
3946                 goto nla_put_failure;
3947         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3948
3949         /* XXX - MC not implemented */
3950
3951         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3952         if (nla == NULL)
3953                 goto nla_put_failure;
3954         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3955
3956         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3957         if (nla == NULL)
3958                 goto nla_put_failure;
3959         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3960
3961         nla_nest_end(skb, protoinfo);
3962         return nlmsg_end(skb, nlh);
3963
3964 nla_put_failure:
3965         nlmsg_cancel(skb, nlh);
3966         return -EMSGSIZE;
3967 }
3968
3969 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3970 {
3971         struct net *net = sock_net(skb->sk);
3972         int h, s_h;
3973         int idx = 0, s_idx;
3974         struct net_device *dev;
3975         struct inet6_dev *idev;
3976         struct hlist_head *head;
3977         struct hlist_node *node;
3978
3979         s_h = cb->args[0];
3980         s_idx = cb->args[1];
3981
3982         rcu_read_lock();
3983         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3984                 idx = 0;
3985                 head = &net->dev_index_head[h];
3986                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3987                         if (idx < s_idx)
3988                                 goto cont;
3989                         idev = __in6_dev_get(dev);
3990                         if (!idev)
3991                                 goto cont;
3992                         if (inet6_fill_ifinfo(skb, idev,
3993                                               NETLINK_CB(cb->skb).pid,
3994                                               cb->nlh->nlmsg_seq,
3995                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3996                                 goto out;
3997 cont:
3998                         idx++;
3999                 }
4000         }
4001 out:
4002         rcu_read_unlock();
4003         cb->args[1] = idx;
4004         cb->args[0] = h;
4005
4006         return skb->len;
4007 }
4008
4009 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4010 {
4011         struct sk_buff *skb;
4012         struct net *net = dev_net(idev->dev);
4013         int err = -ENOBUFS;
4014
4015         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4016         if (skb == NULL)
4017                 goto errout;
4018
4019         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4020         if (err < 0) {
4021                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4022                 WARN_ON(err == -EMSGSIZE);
4023                 kfree_skb(skb);
4024                 goto errout;
4025         }
4026         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4027         return;
4028 errout:
4029         if (err < 0)
4030                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4031 }
4032
4033 static inline size_t inet6_prefix_nlmsg_size(void)
4034 {
4035         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4036                + nla_total_size(sizeof(struct in6_addr))
4037                + nla_total_size(sizeof(struct prefix_cacheinfo));
4038 }
4039
4040 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4041                              struct prefix_info *pinfo, u32 pid, u32 seq,
4042                              int event, unsigned int flags)
4043 {
4044         struct prefixmsg *pmsg;
4045         struct nlmsghdr *nlh;
4046         struct prefix_cacheinfo ci;
4047
4048         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4049         if (nlh == NULL)
4050                 return -EMSGSIZE;
4051
4052         pmsg = nlmsg_data(nlh);
4053         pmsg->prefix_family = AF_INET6;
4054         pmsg->prefix_pad1 = 0;
4055         pmsg->prefix_pad2 = 0;
4056         pmsg->prefix_ifindex = idev->dev->ifindex;
4057         pmsg->prefix_len = pinfo->prefix_len;
4058         pmsg->prefix_type = pinfo->type;
4059         pmsg->prefix_pad3 = 0;
4060         pmsg->prefix_flags = 0;
4061         if (pinfo->onlink)
4062                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4063         if (pinfo->autoconf)
4064                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4065
4066         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4067
4068         ci.preferred_time = ntohl(pinfo->prefered);
4069         ci.valid_time = ntohl(pinfo->valid);
4070         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4071
4072         return nlmsg_end(skb, nlh);
4073
4074 nla_put_failure:
4075         nlmsg_cancel(skb, nlh);
4076         return -EMSGSIZE;
4077 }
4078
4079 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4080                          struct prefix_info *pinfo)
4081 {
4082         struct sk_buff *skb;
4083         struct net *net = dev_net(idev->dev);
4084         int err = -ENOBUFS;
4085
4086         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4087         if (skb == NULL)
4088                 goto errout;
4089
4090         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4091         if (err < 0) {
4092                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4093                 WARN_ON(err == -EMSGSIZE);
4094                 kfree_skb(skb);
4095                 goto errout;
4096         }
4097         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4098         return;
4099 errout:
4100         if (err < 0)
4101                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4102 }
4103
4104 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4105 {
4106         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4107
4108         switch (event) {
4109         case RTM_NEWADDR:
4110                 /*
4111                  * If the address was optimistic
4112                  * we inserted the route at the start of
4113                  * our DAD process, so we don't need
4114                  * to do it again
4115                  */
4116                 if (!(ifp->rt->rt6i_node))
4117                         ip6_ins_rt(ifp->rt);
4118                 if (ifp->idev->cnf.forwarding)
4119                         addrconf_join_anycast(ifp);
4120                 break;
4121         case RTM_DELADDR:
4122                 if (ifp->idev->cnf.forwarding)
4123                         addrconf_leave_anycast(ifp);
4124                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4125                 dst_hold(&ifp->rt->dst);
4126
4127                 if (ifp->state == INET6_IFADDR_STATE_DEAD &&
4128                     ip6_del_rt(ifp->rt))
4129                         dst_free(&ifp->rt->dst);
4130                 break;
4131         }
4132 }
4133
4134 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4135 {
4136         rcu_read_lock_bh();
4137         if (likely(ifp->idev->dead == 0))
4138                 __ipv6_ifa_notify(event, ifp);
4139         rcu_read_unlock_bh();
4140 }
4141
4142 #ifdef CONFIG_SYSCTL
4143
4144 static
4145 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4146                            void __user *buffer, size_t *lenp, loff_t *ppos)
4147 {
4148         int *valp = ctl->data;
4149         int val = *valp;
4150         loff_t pos = *ppos;
4151         int ret;
4152
4153         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4154
4155         if (write)
4156                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4157         if (ret)
4158                 *ppos = pos;
4159         return ret;
4160 }
4161
4162 static void dev_disable_change(struct inet6_dev *idev)
4163 {
4164         if (!idev || !idev->dev)
4165                 return;
4166
4167         if (idev->cnf.disable_ipv6)
4168                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4169         else
4170                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4171 }
4172
4173 static void addrconf_disable_change(struct net *net, __s32 newf)
4174 {
4175         struct net_device *dev;
4176         struct inet6_dev *idev;
4177
4178         rcu_read_lock();
4179         for_each_netdev_rcu(net, dev) {
4180                 idev = __in6_dev_get(dev);
4181                 if (idev) {
4182                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4183                         idev->cnf.disable_ipv6 = newf;
4184                         if (changed)
4185                                 dev_disable_change(idev);
4186                 }
4187         }
4188         rcu_read_unlock();
4189 }
4190
4191 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4192 {
4193         struct net *net;
4194
4195         net = (struct net *)table->extra2;
4196
4197         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4198                 return 0;
4199
4200         if (!rtnl_trylock()) {
4201                 /* Restore the original values before restarting */
4202                 *p = old;
4203                 return restart_syscall();
4204         }
4205
4206         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4207                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4208                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4209                 addrconf_disable_change(net, newf);
4210         } else if ((!*p) ^ (!old))
4211                 dev_disable_change((struct inet6_dev *)table->extra1);
4212
4213         rtnl_unlock();
4214         return 0;
4215 }
4216
4217 static
4218 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4219                             void __user *buffer, size_t *lenp, loff_t *ppos)
4220 {
4221         int *valp = ctl->data;
4222         int val = *valp;
4223         loff_t pos = *ppos;
4224         int ret;
4225
4226         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4227
4228         if (write)
4229                 ret = addrconf_disable_ipv6(ctl, valp, val);
4230         if (ret)
4231                 *ppos = pos;
4232         return ret;
4233 }
4234
4235 static struct addrconf_sysctl_table
4236 {
4237         struct ctl_table_header *sysctl_header;
4238         ctl_table addrconf_vars[DEVCONF_MAX+1];
4239         char *dev_name;
4240 } addrconf_sysctl __read_mostly = {
4241         .sysctl_header = NULL,
4242         .addrconf_vars = {
4243                 {
4244                         .procname       = "forwarding",
4245                         .data           = &ipv6_devconf.forwarding,
4246                         .maxlen         = sizeof(int),
4247                         .mode           = 0644,
4248                         .proc_handler   = addrconf_sysctl_forward,
4249                 },
4250                 {
4251                         .procname       = "hop_limit",
4252                         .data           = &ipv6_devconf.hop_limit,
4253                         .maxlen         = sizeof(int),
4254                         .mode           = 0644,
4255                         .proc_handler   = proc_dointvec,
4256                 },
4257                 {
4258                         .procname       = "mtu",
4259                         .data           = &ipv6_devconf.mtu6,
4260                         .maxlen         = sizeof(int),
4261                         .mode           = 0644,
4262                         .proc_handler   = proc_dointvec,
4263                 },
4264                 {
4265                         .procname       = "accept_ra",
4266                         .data           = &ipv6_devconf.accept_ra,
4267                         .maxlen         = sizeof(int),
4268                         .mode           = 0644,
4269                         .proc_handler   = proc_dointvec,
4270                 },
4271                 {
4272                         .procname       = "accept_redirects",
4273                         .data           = &ipv6_devconf.accept_redirects,
4274                         .maxlen         = sizeof(int),
4275                         .mode           = 0644,
4276                         .proc_handler   = proc_dointvec,
4277                 },
4278                 {
4279                         .procname       = "autoconf",
4280                         .data           = &ipv6_devconf.autoconf,
4281                         .maxlen         = sizeof(int),
4282                         .mode           = 0644,
4283                         .proc_handler   = proc_dointvec,
4284                 },
4285                 {
4286                         .procname       = "dad_transmits",
4287                         .data           = &ipv6_devconf.dad_transmits,
4288                         .maxlen         = sizeof(int),
4289                         .mode           = 0644,
4290                         .proc_handler   = proc_dointvec,
4291                 },
4292                 {
4293                         .procname       = "router_solicitations",
4294                         .data           = &ipv6_devconf.rtr_solicits,
4295                         .maxlen         = sizeof(int),
4296                         .mode           = 0644,
4297                         .proc_handler   = proc_dointvec,
4298                 },
4299                 {
4300                         .procname       = "router_solicitation_interval",
4301                         .data           = &ipv6_devconf.rtr_solicit_interval,
4302                         .maxlen         = sizeof(int),
4303                         .mode           = 0644,
4304                         .proc_handler   = proc_dointvec_jiffies,
4305                 },
4306                 {
4307                         .procname       = "router_solicitation_delay",
4308                         .data           = &ipv6_devconf.rtr_solicit_delay,
4309                         .maxlen         = sizeof(int),
4310                         .mode           = 0644,
4311                         .proc_handler   = proc_dointvec_jiffies,
4312                 },
4313                 {
4314                         .procname       = "force_mld_version",
4315                         .data           = &ipv6_devconf.force_mld_version,
4316                         .maxlen         = sizeof(int),
4317                         .mode           = 0644,
4318                         .proc_handler   = proc_dointvec,
4319                 },
4320 #ifdef CONFIG_IPV6_PRIVACY
4321                 {
4322                         .procname       = "use_tempaddr",
4323                         .data           = &ipv6_devconf.use_tempaddr,
4324                         .maxlen         = sizeof(int),
4325                         .mode           = 0644,
4326                         .proc_handler   = proc_dointvec,
4327                 },
4328                 {
4329                         .procname       = "temp_valid_lft",
4330                         .data           = &ipv6_devconf.temp_valid_lft,
4331                         .maxlen         = sizeof(int),
4332                         .mode           = 0644,
4333                         .proc_handler   = proc_dointvec,
4334                 },
4335                 {
4336                         .procname       = "temp_prefered_lft",
4337                         .data           = &ipv6_devconf.temp_prefered_lft,
4338                         .maxlen         = sizeof(int),
4339                         .mode           = 0644,
4340                         .proc_handler   = proc_dointvec,
4341                 },
4342                 {
4343                         .procname       = "regen_max_retry",
4344                         .data           = &ipv6_devconf.regen_max_retry,
4345                         .maxlen         = sizeof(int),
4346                         .mode           = 0644,
4347                         .proc_handler   = proc_dointvec,
4348                 },
4349                 {
4350                         .procname       = "max_desync_factor",
4351                         .data           = &ipv6_devconf.max_desync_factor,
4352                         .maxlen         = sizeof(int),
4353                         .mode           = 0644,
4354                         .proc_handler   = proc_dointvec,
4355                 },
4356 #endif
4357                 {
4358                         .procname       = "max_addresses",
4359                         .data           = &ipv6_devconf.max_addresses,
4360                         .maxlen         = sizeof(int),
4361                         .mode           = 0644,
4362                         .proc_handler   = proc_dointvec,
4363                 },
4364                 {
4365                         .procname       = "accept_ra_defrtr",
4366                         .data           = &ipv6_devconf.accept_ra_defrtr,
4367                         .maxlen         = sizeof(int),
4368                         .mode           = 0644,
4369                         .proc_handler   = proc_dointvec,
4370                 },
4371                 {
4372                         .procname       = "accept_ra_pinfo",
4373                         .data           = &ipv6_devconf.accept_ra_pinfo,
4374                         .maxlen         = sizeof(int),
4375                         .mode           = 0644,
4376                         .proc_handler   = proc_dointvec,
4377                 },
4378 #ifdef CONFIG_IPV6_ROUTER_PREF
4379                 {
4380                         .procname       = "accept_ra_rtr_pref",
4381                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4382                         .maxlen         = sizeof(int),
4383                         .mode           = 0644,
4384                         .proc_handler   = proc_dointvec,
4385                 },
4386                 {
4387                         .procname       = "router_probe_interval",
4388                         .data           = &ipv6_devconf.rtr_probe_interval,
4389                         .maxlen         = sizeof(int),
4390                         .mode           = 0644,
4391                         .proc_handler   = proc_dointvec_jiffies,
4392                 },
4393 #ifdef CONFIG_IPV6_ROUTE_INFO
4394                 {
4395                         .procname       = "accept_ra_rt_info_max_plen",
4396                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4397                         .maxlen         = sizeof(int),
4398                         .mode           = 0644,
4399                         .proc_handler   = proc_dointvec,
4400                 },
4401 #endif
4402 #endif
4403                 {
4404                         .procname       = "proxy_ndp",
4405                         .data           = &ipv6_devconf.proxy_ndp,
4406                         .maxlen         = sizeof(int),
4407                         .mode           = 0644,
4408                         .proc_handler   = proc_dointvec,
4409                 },
4410                 {
4411                         .procname       = "accept_source_route",
4412                         .data           = &ipv6_devconf.accept_source_route,
4413                         .maxlen         = sizeof(int),
4414                         .mode           = 0644,
4415                         .proc_handler   = proc_dointvec,
4416                 },
4417 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4418                 {
4419                         .procname       = "optimistic_dad",
4420                         .data           = &ipv6_devconf.optimistic_dad,
4421                         .maxlen         = sizeof(int),
4422                         .mode           = 0644,
4423                         .proc_handler   = proc_dointvec,
4424
4425                 },
4426 #endif
4427 #ifdef CONFIG_IPV6_MROUTE
4428                 {
4429                         .procname       = "mc_forwarding",
4430                         .data           = &ipv6_devconf.mc_forwarding,
4431                         .maxlen         = sizeof(int),
4432                         .mode           = 0444,
4433                         .proc_handler   = proc_dointvec,
4434                 },
4435 #endif
4436                 {
4437                         .procname       = "disable_ipv6",
4438                         .data           = &ipv6_devconf.disable_ipv6,
4439                         .maxlen         = sizeof(int),
4440                         .mode           = 0644,
4441                         .proc_handler   = addrconf_sysctl_disable,
4442                 },
4443                 {
4444                         .procname       = "accept_dad",
4445                         .data           = &ipv6_devconf.accept_dad,
4446                         .maxlen         = sizeof(int),
4447                         .mode           = 0644,
4448                         .proc_handler   = proc_dointvec,
4449                 },
4450                 {
4451                         .procname       = "force_tllao",
4452                         .data           = &ipv6_devconf.force_tllao,
4453                         .maxlen         = sizeof(int),
4454                         .mode           = 0644,
4455                         .proc_handler   = proc_dointvec
4456                 },
4457                 {
4458                         /* sentinel */
4459                 }
4460         },
4461 };
4462
4463 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4464                 struct inet6_dev *idev, struct ipv6_devconf *p)
4465 {
4466         int i;
4467         struct addrconf_sysctl_table *t;
4468
4469 #define ADDRCONF_CTL_PATH_DEV   3
4470
4471         struct ctl_path addrconf_ctl_path[] = {
4472                 { .procname = "net", },
4473                 { .procname = "ipv6", },
4474                 { .procname = "conf", },
4475                 { /* to be set */ },
4476                 { },
4477         };
4478
4479
4480         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4481         if (t == NULL)
4482                 goto out;
4483
4484         for (i = 0; t->addrconf_vars[i].data; i++) {
4485                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4486                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4487                 t->addrconf_vars[i].extra2 = net;
4488         }
4489
4490         /*
4491          * Make a copy of dev_name, because '.procname' is regarded as const
4492          * by sysctl and we wouldn't want anyone to change it under our feet
4493          * (see SIOCSIFNAME).
4494          */
4495         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4496         if (!t->dev_name)
4497                 goto free;
4498
4499         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4500
4501         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4502                         t->addrconf_vars);
4503         if (t->sysctl_header == NULL)
4504                 goto free_procname;
4505
4506         p->sysctl = t;
4507         return 0;
4508
4509 free_procname:
4510         kfree(t->dev_name);
4511 free:
4512         kfree(t);
4513 out:
4514         return -ENOBUFS;
4515 }
4516
4517 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4518 {
4519         struct addrconf_sysctl_table *t;
4520
4521         if (p->sysctl == NULL)
4522                 return;
4523
4524         t = p->sysctl;
4525         p->sysctl = NULL;
4526         unregister_sysctl_table(t->sysctl_header);
4527         kfree(t->dev_name);
4528         kfree(t);
4529 }
4530
4531 static void addrconf_sysctl_register(struct inet6_dev *idev)
4532 {
4533         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4534                               &ndisc_ifinfo_sysctl_change);
4535         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4536                                         idev, &idev->cnf);
4537 }
4538
4539 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4540 {
4541         __addrconf_sysctl_unregister(&idev->cnf);
4542         neigh_sysctl_unregister(idev->nd_parms);
4543 }
4544
4545
4546 #endif
4547
4548 static int __net_init addrconf_init_net(struct net *net)
4549 {
4550         int err;
4551         struct ipv6_devconf *all, *dflt;
4552
4553         err = -ENOMEM;
4554         all = &ipv6_devconf;
4555         dflt = &ipv6_devconf_dflt;
4556
4557         if (!net_eq(net, &init_net)) {
4558                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4559                 if (all == NULL)
4560                         goto err_alloc_all;
4561
4562                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4563                 if (dflt == NULL)
4564                         goto err_alloc_dflt;
4565         } else {
4566                 /* these will be inherited by all namespaces */
4567                 dflt->autoconf = ipv6_defaults.autoconf;
4568                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4569         }
4570
4571         net->ipv6.devconf_all = all;
4572         net->ipv6.devconf_dflt = dflt;
4573
4574 #ifdef CONFIG_SYSCTL
4575         err = __addrconf_sysctl_register(net, "all", NULL, all);
4576         if (err < 0)
4577                 goto err_reg_all;
4578
4579         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4580         if (err < 0)
4581                 goto err_reg_dflt;
4582 #endif
4583         return 0;
4584
4585 #ifdef CONFIG_SYSCTL
4586 err_reg_dflt:
4587         __addrconf_sysctl_unregister(all);
4588 err_reg_all:
4589         kfree(dflt);
4590 #endif
4591 err_alloc_dflt:
4592         kfree(all);
4593 err_alloc_all:
4594         return err;
4595 }
4596
4597 static void __net_exit addrconf_exit_net(struct net *net)
4598 {
4599 #ifdef CONFIG_SYSCTL
4600         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4601         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4602 #endif
4603         if (!net_eq(net, &init_net)) {
4604                 kfree(net->ipv6.devconf_dflt);
4605                 kfree(net->ipv6.devconf_all);
4606         }
4607 }
4608
4609 static struct pernet_operations addrconf_ops = {
4610         .init = addrconf_init_net,
4611         .exit = addrconf_exit_net,
4612 };
4613
4614 /*
4615  *      Device notifier
4616  */
4617
4618 int register_inet6addr_notifier(struct notifier_block *nb)
4619 {
4620         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4621 }
4622 EXPORT_SYMBOL(register_inet6addr_notifier);
4623
4624 int unregister_inet6addr_notifier(struct notifier_block *nb)
4625 {
4626         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4627 }
4628 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4629
4630 /*
4631  *      Init / cleanup code
4632  */
4633
4634 int __init addrconf_init(void)
4635 {
4636         int i, err;
4637
4638         err = ipv6_addr_label_init();
4639         if (err < 0) {
4640                 printk(KERN_CRIT "IPv6 Addrconf:"
4641                        " cannot initialize default policy table: %d.\n", err);
4642                 goto out;
4643         }
4644
4645         err = register_pernet_subsys(&addrconf_ops);
4646         if (err < 0)
4647                 goto out_addrlabel;
4648
4649         /* The addrconf netdev notifier requires that loopback_dev
4650          * has it's ipv6 private information allocated and setup
4651          * before it can bring up and give link-local addresses
4652          * to other devices which are up.
4653          *
4654          * Unfortunately, loopback_dev is not necessarily the first
4655          * entry in the global dev_base list of net devices.  In fact,
4656          * it is likely to be the very last entry on that list.
4657          * So this causes the notifier registry below to try and
4658          * give link-local addresses to all devices besides loopback_dev
4659          * first, then loopback_dev, which cases all the non-loopback_dev
4660          * devices to fail to get a link-local address.
4661          *
4662          * So, as a temporary fix, allocate the ipv6 structure for
4663          * loopback_dev first by hand.
4664          * Longer term, all of the dependencies ipv6 has upon the loopback
4665          * device and it being up should be removed.
4666          */
4667         rtnl_lock();
4668         if (!ipv6_add_dev(init_net.loopback_dev))
4669                 err = -ENOMEM;
4670         rtnl_unlock();
4671         if (err)
4672                 goto errlo;
4673
4674         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4675                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4676
4677         register_netdevice_notifier(&ipv6_dev_notf);
4678
4679         addrconf_verify(0);
4680
4681         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4682         if (err < 0)
4683                 goto errout;
4684
4685         /* Only the first call to __rtnl_register can fail */
4686         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4687         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4688         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4689         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4690         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4691
4692         ipv6_addr_label_rtnl_register();
4693
4694         return 0;
4695 errout:
4696         unregister_netdevice_notifier(&ipv6_dev_notf);
4697 errlo:
4698         unregister_pernet_subsys(&addrconf_ops);
4699 out_addrlabel:
4700         ipv6_addr_label_cleanup();
4701 out:
4702         return err;
4703 }
4704
4705 void addrconf_cleanup(void)
4706 {
4707         struct net_device *dev;
4708         int i;
4709
4710         unregister_netdevice_notifier(&ipv6_dev_notf);
4711         unregister_pernet_subsys(&addrconf_ops);
4712         ipv6_addr_label_cleanup();
4713
4714         rtnl_lock();
4715
4716         /* clean dev list */
4717         for_each_netdev(&init_net, dev) {
4718                 if (__in6_dev_get(dev) == NULL)
4719                         continue;
4720                 addrconf_ifdown(dev, 1);
4721         }
4722         addrconf_ifdown(init_net.loopback_dev, 2);
4723
4724         /*
4725          *      Check hash table.
4726          */
4727         spin_lock_bh(&addrconf_hash_lock);
4728         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4729                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4730         spin_unlock_bh(&addrconf_hash_lock);
4731
4732         del_timer(&addr_chk_timer);
4733         rtnl_unlock();
4734 }