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 #define TIME_DELTA(a, b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 104 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 105 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 106
 107 #ifdef CONFIG_SYSCTL
 108 static void addrconf_sysctl_register(struct inet6_dev *idev);
 109 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 110 #else
 111 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 112 {
 113 }
 114
 115 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 116 {
 117 }
 118 #endif
 119
 120 #ifdef CONFIG_IPV6_PRIVACY
 121 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 122 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 123 static void ipv6_regen_rndid(unsigned long data);
 124 #endif
 125
 126 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 127 static int ipv6_count_addresses(struct inet6_dev *idev);
 128
 129 /*
 130  *      Configured unicast address hash table
 131  */
 132 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 133 static DEFINE_SPINLOCK(addrconf_hash_lock);
 134
 135 static void addrconf_verify(unsigned long);
 136
 137 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 138 static DEFINE_SPINLOCK(addrconf_verify_lock);
 139
 140 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 141 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 142
 143 static void addrconf_type_change(struct net_device *dev,
 144                                  unsigned long event);
 145 static int addrconf_ifdown(struct net_device *dev, int how);
 146
 147 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 148 static void addrconf_dad_timer(unsigned long data);
 149 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 150 static void addrconf_dad_run(struct inet6_dev *idev);
 151 static void addrconf_rs_timer(unsigned long data);
 152 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 153 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 154
 155 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 156                                 struct prefix_info *pinfo);
 157 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 158                                struct net_device *dev);
 159
 160 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 161
 162 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 163         .forwarding             = 0,
 164         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 165         .mtu6                   = IPV6_MIN_MTU,
 166         .accept_ra              = 1,
 167         .accept_redirects       = 1,
 168         .autoconf               = 1,
 169         .force_mld_version      = 0,
 170         .dad_transmits          = 1,
 171         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 172         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 173         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 174 #ifdef CONFIG_IPV6_PRIVACY
 175         .use_tempaddr           = 0,
 176         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 177         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 178         .regen_max_retry        = REGEN_MAX_RETRY,
 179         .max_desync_factor      = MAX_DESYNC_FACTOR,
 180 #endif
 181         .max_addresses          = IPV6_MAX_ADDRESSES,
 182         .accept_ra_defrtr       = 1,
 183         .accept_ra_pinfo        = 1,
 184 #ifdef CONFIG_IPV6_ROUTER_PREF
 185         .accept_ra_rtr_pref     = 1,
 186         .rtr_probe_interval     = 60 * HZ,
 187 #ifdef CONFIG_IPV6_ROUTE_INFO
 188         .accept_ra_rt_info_max_plen = 0,
 189 #endif
 190 #endif
 191         .proxy_ndp              = 0,
 192         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 193         .disable_ipv6           = 0,
 194         .accept_dad             = 1,
 195 };
 196
 197 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 198         .forwarding             = 0,
 199         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 200         .mtu6                   = IPV6_MIN_MTU,
 201         .accept_ra              = 1,
 202         .accept_redirects       = 1,
 203         .autoconf               = 1,
 204         .dad_transmits          = 1,
 205         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 206         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 207         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 208 #ifdef CONFIG_IPV6_PRIVACY
 209         .use_tempaddr           = 0,
 210         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 211         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 212         .regen_max_retry        = REGEN_MAX_RETRY,
 213         .max_desync_factor      = MAX_DESYNC_FACTOR,
 214 #endif
 215         .max_addresses          = IPV6_MAX_ADDRESSES,
 216         .accept_ra_defrtr       = 1,
 217         .accept_ra_pinfo        = 1,
 218 #ifdef CONFIG_IPV6_ROUTER_PREF
 219         .accept_ra_rtr_pref     = 1,
 220         .rtr_probe_interval     = 60 * HZ,
 221 #ifdef CONFIG_IPV6_ROUTE_INFO
 222         .accept_ra_rt_info_max_plen = 0,
 223 #endif
 224 #endif
 225         .proxy_ndp              = 0,
 226         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 227         .disable_ipv6           = 0,
 228         .accept_dad             = 1,
 229 };
 230
 231 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 232 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 233 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 234 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 235 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 236
 237 /* Check if a valid qdisc is available */
 238 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 239 {
 240         return !qdisc_tx_is_noop(dev);
 241 }
 242
 243 /* Check if a route is valid prefix route */
 244 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 245 {
 246         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 247 }
 248
 249 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 250 {
 251         if (del_timer(&ifp->timer))
 252                 __in6_ifa_put(ifp);
 253 }
 254
 255 enum addrconf_timer_t {
 256         AC_NONE,
 257         AC_DAD,
 258         AC_RS,
 259 };
 260
 261 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 262                                enum addrconf_timer_t what,
 263                                unsigned long when)
 264 {
 265         if (!del_timer(&ifp->timer))
 266                 in6_ifa_hold(ifp);
 267
 268         switch (what) {
 269         case AC_DAD:
 270                 ifp->timer.function = addrconf_dad_timer;
 271                 break;
 272         case AC_RS:
 273                 ifp->timer.function = addrconf_rs_timer;
 274                 break;
 275         default:
 276                 break;
 277         }
 278         ifp->timer.expires = jiffies + when;
 279         add_timer(&ifp->timer);
 280 }
 281
 282 static int snmp6_alloc_dev(struct inet6_dev *idev)
 283 {
 284         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 285                           sizeof(struct ipstats_mib),
 286                           __alignof__(struct ipstats_mib)) < 0)
 287                 goto err_ip;
 288         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 289                           sizeof(struct icmpv6_mib),
 290                           __alignof__(struct icmpv6_mib)) < 0)
 291                 goto err_icmp;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 293                           sizeof(struct icmpv6msg_mib),
 294                           __alignof__(struct icmpv6msg_mib)) < 0)
 295                 goto err_icmpmsg;
 296
 297         return 0;
 298
 299 err_icmpmsg:
 300         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 301 err_icmp:
 302         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 303 err_ip:
 304         return -ENOMEM;
 305 }
 306
 307 static void snmp6_free_dev(struct inet6_dev *idev)
 308 {
 309         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 310         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 311         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 312 }
 313
 314 /* Nobody refers to this device, we may destroy it. */
 315
 316 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 317 {
 318         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 319         kfree(idev);
 320 }
 321
 322 void in6_dev_finish_destroy(struct inet6_dev *idev)
 323 {
 324         struct net_device *dev = idev->dev;
 325
 326         WARN_ON(!list_empty(&idev->addr_list));
 327         WARN_ON(idev->mc_list != NULL);
 328
 329 #ifdef NET_REFCNT_DEBUG
 330         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 331 #endif
 332         dev_put(dev);
 333         if (!idev->dead) {
 334                 pr_warning("Freeing alive inet6 device %p\n", idev);
 335                 return;
 336         }
 337         snmp6_free_dev(idev);
 338         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 339 }
 340
 341 EXPORT_SYMBOL(in6_dev_finish_destroy);
 342
 343 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 344 {
 345         struct inet6_dev *ndev;
 346
 347         ASSERT_RTNL();
 348
 349         if (dev->mtu < IPV6_MIN_MTU)
 350                 return NULL;
 351
 352         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 353
 354         if (ndev == NULL)
 355                 return NULL;
 356
 357         rwlock_init(&ndev->lock);
 358         ndev->dev = dev;
 359         INIT_LIST_HEAD(&ndev->addr_list);
 360
 361         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 362         ndev->cnf.mtu6 = dev->mtu;
 363         ndev->cnf.sysctl = NULL;
 364         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 365         if (ndev->nd_parms == NULL) {
 366                 kfree(ndev);
 367                 return NULL;
 368         }
 369         if (ndev->cnf.forwarding)
 370                 dev_disable_lro(dev);
 371         /* We refer to the device */
 372         dev_hold(dev);
 373
 374         if (snmp6_alloc_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         if (snmp6_register_dev(ndev) < 0) {
 385                 ADBG((KERN_WARNING
 386                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 387                         __func__, dev->name));
 388                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 389                 ndev->dead = 1;
 390                 in6_dev_finish_destroy(ndev);
 391                 return NULL;
 392         }
 393
 394         /* One reference from device.  We must do this before
 395          * we invoke __ipv6_regen_rndid().
 396          */
 397         in6_dev_hold(ndev);
 398
 399         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 400                 ndev->cnf.accept_dad = -1;
 401
 402 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 403         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 404                 printk(KERN_INFO
 405                        "%s: Disabled Multicast RS\n",
 406                        dev->name);
 407                 ndev->cnf.rtr_solicits = 0;
 408         }
 409 #endif
 410
 411 #ifdef CONFIG_IPV6_PRIVACY
 412         INIT_LIST_HEAD(&ndev->tempaddr_list);
 413         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 414         if ((dev->flags&IFF_LOOPBACK) ||
 415             dev->type == ARPHRD_TUNNEL ||
 416             dev->type == ARPHRD_TUNNEL6 ||
 417             dev->type == ARPHRD_SIT ||
 418             dev->type == ARPHRD_NONE) {
 419                 printk(KERN_INFO
 420                        "%s: Disabled Privacy Extensions\n",
 421                        dev->name);
 422                 ndev->cnf.use_tempaddr = -1;
 423         } else {
 424                 in6_dev_hold(ndev);
 425                 ipv6_regen_rndid((unsigned long) ndev);
 426         }
 427 #endif
 428
 429         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 430                 ndev->if_flags |= IF_READY;
 431
 432         ipv6_mc_init_dev(ndev);
 433         ndev->tstamp = jiffies;
 434         addrconf_sysctl_register(ndev);
 435         /* protected by rtnl_lock */
 436         rcu_assign_pointer(dev->ip6_ptr, ndev);
 437
 438         /* Join all-node multicast group */
 439         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         rcu_read_lock();
 497         for_each_netdev_rcu(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506         rcu_read_unlock();
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 510 {
 511         struct net *net;
 512
 513         net = (struct net *)table->extra2;
 514         if (p == &net->ipv6.devconf_dflt->forwarding)
 515                 return 0;
 516
 517         if (!rtnl_trylock()) {
 518                 /* Restore the original values before restarting */
 519                 *p = old;
 520                 return restart_syscall();
 521         }
 522
 523         if (p == &net->ipv6.devconf_all->forwarding) {
 524                 __s32 newf = net->ipv6.devconf_all->forwarding;
 525                 net->ipv6.devconf_dflt->forwarding = newf;
 526                 addrconf_forward_change(net, newf);
 527         } else if ((!*p) ^ (!old))
 528                 dev_forward_change((struct inet6_dev *)table->extra1);
 529         rtnl_unlock();
 530
 531         if (*p)
 532                 rt6_purge_dflt_routers(net);
 533         return 1;
 534 }
 535 #endif
 536
 537 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 538 {
 539         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 540         kfree(ifp);
 541 }
 542
 543 /* Nobody refers to this ifaddr, destroy it */
 544 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 545 {
 546         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 547
 548 #ifdef NET_REFCNT_DEBUG
 549         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 550 #endif
 551
 552         in6_dev_put(ifp->idev);
 553
 554         if (del_timer(&ifp->timer))
 555                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 556
 557         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 558                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 559                 return;
 560         }
 561         dst_release(&ifp->rt->dst);
 562
 563         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 564 }
 565
 566 static void
 567 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 568 {
 569         struct list_head *p;
 570         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 571
 572         /*
 573          * Each device address list is sorted in order of scope -
 574          * global before linklocal.
 575          */
 576         list_for_each(p, &idev->addr_list) {
 577                 struct inet6_ifaddr *ifa
 578                         = list_entry(p, struct inet6_ifaddr, if_list);
 579                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 580                         break;
 581         }
 582
 583         list_add_tail(&ifp->if_list, p);
 584 }
 585
 586 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 587 {
 588         /*
 589          * We perform the hash function over the last 64 bits of the address
 590          * This will include the IEEE address token on links that support it.
 591          */
 592         return jhash_2words((__force u32)addr->s6_addr32[2],
 593                             (__force u32)addr->s6_addr32[3], 0)
 594                 & (IN6_ADDR_HSIZE - 1);
 595 }
 596
 597 /* On success it returns ifp with increased reference count */
 598
 599 static struct inet6_ifaddr *
 600 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 601               int scope, u32 flags)
 602 {
 603         struct inet6_ifaddr *ifa = NULL;
 604         struct rt6_info *rt;
 605         unsigned int hash;
 606         int err = 0;
 607         int addr_type = ipv6_addr_type(addr);
 608
 609         if (addr_type == IPV6_ADDR_ANY ||
 610             addr_type & IPV6_ADDR_MULTICAST ||
 611             (!(idev->dev->flags & IFF_LOOPBACK) &&
 612              addr_type & IPV6_ADDR_LOOPBACK))
 613                 return ERR_PTR(-EADDRNOTAVAIL);
 614
 615         rcu_read_lock_bh();
 616         if (idev->dead) {
 617                 err = -ENODEV;                  /*XXX*/
 618                 goto out2;
 619         }
 620
 621         if (idev->cnf.disable_ipv6) {
 622                 err = -EACCES;
 623                 goto out2;
 624         }
 625
 626         spin_lock(&addrconf_hash_lock);
 627
 628         /* Ignore adding duplicate addresses on an interface */
 629         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 630                 ADBG(("ipv6_add_addr: already assigned\n"));
 631                 err = -EEXIST;
 632                 goto out;
 633         }
 634
 635         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 636
 637         if (ifa == NULL) {
 638                 ADBG(("ipv6_add_addr: malloc failed\n"));
 639                 err = -ENOBUFS;
 640                 goto out;
 641         }
 642
 643         rt = addrconf_dst_alloc(idev, addr, 0);
 644         if (IS_ERR(rt)) {
 645                 err = PTR_ERR(rt);
 646                 goto out;
 647         }
 648
 649         ipv6_addr_copy(&ifa->addr, addr);
 650
 651         spin_lock_init(&ifa->lock);
 652         spin_lock_init(&ifa->state_lock);
 653         init_timer(&ifa->timer);
 654         INIT_HLIST_NODE(&ifa->addr_lst);
 655         ifa->timer.data = (unsigned long) ifa;
 656         ifa->scope = scope;
 657         ifa->prefix_len = pfxlen;
 658         ifa->flags = flags | IFA_F_TENTATIVE;
 659         ifa->cstamp = ifa->tstamp = jiffies;
 660
 661         ifa->rt = rt;
 662
 663         /*
 664          * part one of RFC 4429, section 3.3
 665          * We should not configure an address as
 666          * optimistic if we do not yet know the link
 667          * layer address of our nexhop router
 668          */
 669
 670         if (rt->rt6i_nexthop == NULL)
 671                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 672
 673         ifa->idev = idev;
 674         in6_dev_hold(idev);
 675         /* For caller */
 676         in6_ifa_hold(ifa);
 677
 678         /* Add to big hash table */
 679         hash = ipv6_addr_hash(addr);
 680
 681         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 682         spin_unlock(&addrconf_hash_lock);
 683
 684         write_lock(&idev->lock);
 685         /* Add to inet6_dev unicast addr list. */
 686         ipv6_link_dev_addr(idev, ifa);
 687
 688 #ifdef CONFIG_IPV6_PRIVACY
 689         if (ifa->flags&IFA_F_TEMPORARY) {
 690                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 691                 in6_ifa_hold(ifa);
 692         }
 693 #endif
 694
 695         in6_ifa_hold(ifa);
 696         write_unlock(&idev->lock);
 697 out2:
 698         rcu_read_unlock_bh();
 699
 700         if (likely(err == 0))
 701                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 702         else {
 703                 kfree(ifa);
 704                 ifa = ERR_PTR(err);
 705         }
 706
 707         return ifa;
 708 out:
 709         spin_unlock(&addrconf_hash_lock);
 710         goto out2;
 711 }
 712
 713 /* This function wants to get referenced ifp and releases it before return */
 714
 715 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 716 {
 717         struct inet6_ifaddr *ifa, *ifn;
 718         struct inet6_dev *idev = ifp->idev;
 719         int state;
 720         int hash;
 721         int deleted = 0, onlink = 0;
 722         unsigned long expires = jiffies;
 723
 724         hash = ipv6_addr_hash(&ifp->addr);
 725
 726         spin_lock_bh(&ifp->state_lock);
 727         state = ifp->state;
 728         ifp->state = INET6_IFADDR_STATE_DEAD;
 729         spin_unlock_bh(&ifp->state_lock);
 730
 731         if (state == INET6_IFADDR_STATE_DEAD)
 732                 goto out;
 733
 734         spin_lock_bh(&addrconf_hash_lock);
 735         hlist_del_init_rcu(&ifp->addr_lst);
 736         spin_unlock_bh(&addrconf_hash_lock);
 737
 738         write_lock_bh(&idev->lock);
 739 #ifdef CONFIG_IPV6_PRIVACY
 740         if (ifp->flags&IFA_F_TEMPORARY) {
 741                 list_del(&ifp->tmp_list);
 742                 if (ifp->ifpub) {
 743                         in6_ifa_put(ifp->ifpub);
 744                         ifp->ifpub = NULL;
 745                 }
 746                 __in6_ifa_put(ifp);
 747         }
 748 #endif
 749
 750         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 751                 if (ifa == ifp) {
 752                         list_del_init(&ifp->if_list);
 753                         __in6_ifa_put(ifp);
 754
 755                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 756                                 break;
 757                         deleted = 1;
 758                         continue;
 759                 } else if (ifp->flags & IFA_F_PERMANENT) {
 760                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 761                                               ifp->prefix_len)) {
 762                                 if (ifa->flags & IFA_F_PERMANENT) {
 763                                         onlink = 1;
 764                                         if (deleted)
 765                                                 break;
 766                                 } else {
 767                                         unsigned long lifetime;
 768
 769                                         if (!onlink)
 770                                                 onlink = -1;
 771
 772                                         spin_lock(&ifa->lock);
 773
 774                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 775                                         /*
 776                                          * Note: Because this address is
 777                                          * not permanent, lifetime <
 778                                          * LONG_MAX / HZ here.
 779                                          */
 780                                         if (time_before(expires,
 781                                                         ifa->tstamp + lifetime * HZ))
 782                                                 expires = ifa->tstamp + lifetime * HZ;
 783                                         spin_unlock(&ifa->lock);
 784                                 }
 785                         }
 786                 }
 787         }
 788         write_unlock_bh(&idev->lock);
 789
 790         addrconf_del_timer(ifp);
 791
 792         ipv6_ifa_notify(RTM_DELADDR, ifp);
 793
 794         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 795
 796         /*
 797          * Purge or update corresponding prefix
 798          *
 799          * 1) we don't purge prefix here if address was not permanent.
 800          *    prefix is managed by its own lifetime.
 801          * 2) if there're no addresses, delete prefix.
 802          * 3) if there're still other permanent address(es),
 803          *    corresponding prefix is still permanent.
 804          * 4) otherwise, update prefix lifetime to the
 805          *    longest valid lifetime among the corresponding
 806          *    addresses on the device.
 807          *    Note: subsequent RA will update lifetime.
 808          *
 809          * --yoshfuji
 810          */
 811         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 812                 struct in6_addr prefix;
 813                 struct rt6_info *rt;
 814                 struct net *net = dev_net(ifp->idev->dev);
 815                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 816                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 817
 818                 if (rt && addrconf_is_prefix_route(rt)) {
 819                         if (onlink == 0) {
 820                                 ip6_del_rt(rt);
 821                                 rt = NULL;
 822                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 823                                 rt->rt6i_expires = expires;
 824                                 rt->rt6i_flags |= RTF_EXPIRES;
 825                         }
 826                 }
 827                 dst_release(&rt->dst);
 828         }
 829
 830 out:
 831         in6_ifa_put(ifp);
 832 }
 833
 834 #ifdef CONFIG_IPV6_PRIVACY
 835 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 836 {
 837         struct inet6_dev *idev = ifp->idev;
 838         struct in6_addr addr, *tmpaddr;
 839         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 840         unsigned long regen_advance;
 841         int tmp_plen;
 842         int ret = 0;
 843         int max_addresses;
 844         u32 addr_flags;
 845
 846         write_lock(&idev->lock);
 847         if (ift) {
 848                 spin_lock_bh(&ift->lock);
 849                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 850                 spin_unlock_bh(&ift->lock);
 851                 tmpaddr = &addr;
 852         } else {
 853                 tmpaddr = NULL;
 854         }
 855 retry:
 856         in6_dev_hold(idev);
 857         if (idev->cnf.use_tempaddr <= 0) {
 858                 write_unlock(&idev->lock);
 859                 printk(KERN_INFO
 860                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 861                 in6_dev_put(idev);
 862                 ret = -1;
 863                 goto out;
 864         }
 865         spin_lock_bh(&ifp->lock);
 866         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 867                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 868                 spin_unlock_bh(&ifp->lock);
 869                 write_unlock(&idev->lock);
 870                 printk(KERN_WARNING
 871                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         in6_ifa_hold(ifp);
 877         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 878         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 879                 spin_unlock_bh(&ifp->lock);
 880                 write_unlock(&idev->lock);
 881                 printk(KERN_WARNING
 882                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 883                 in6_ifa_put(ifp);
 884                 in6_dev_put(idev);
 885                 ret = -1;
 886                 goto out;
 887         }
 888         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 889         age = (jiffies - ifp->tstamp) / HZ;
 890         tmp_valid_lft = min_t(__u32,
 891                               ifp->valid_lft,
 892                               idev->cnf.temp_valid_lft + age);
 893         tmp_prefered_lft = min_t(__u32,
 894                                  ifp->prefered_lft,
 895                                  idev->cnf.temp_prefered_lft + age -
 896                                  idev->cnf.max_desync_factor);
 897         tmp_plen = ifp->prefix_len;
 898         max_addresses = idev->cnf.max_addresses;
 899         tmp_cstamp = ifp->cstamp;
 900         tmp_tstamp = ifp->tstamp;
 901         spin_unlock_bh(&ifp->lock);
 902
 903         regen_advance = idev->cnf.regen_max_retry *
 904                         idev->cnf.dad_transmits *
 905                         idev->nd_parms->retrans_time / HZ;
 906         write_unlock(&idev->lock);
 907
 908         /* A temporary address is created only if this calculated Preferred
 909          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 910          * an implementation must not create a temporary address with a zero
 911          * Preferred Lifetime.
 912          */
 913         if (tmp_prefered_lft <= regen_advance) {
 914                 in6_ifa_put(ifp);
 915                 in6_dev_put(idev);
 916                 ret = -1;
 917                 goto out;
 918         }
 919
 920         addr_flags = IFA_F_TEMPORARY;
 921         /* set in addrconf_prefix_rcv() */
 922         if (ifp->flags & IFA_F_OPTIMISTIC)
 923                 addr_flags |= IFA_F_OPTIMISTIC;
 924
 925         ift = !max_addresses ||
 926               ipv6_count_addresses(idev) < max_addresses ?
 927                 ipv6_add_addr(idev, &addr, tmp_plen,
 928                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 929                               addr_flags) : NULL;
 930         if (!ift || IS_ERR(ift)) {
 931                 in6_ifa_put(ifp);
 932                 in6_dev_put(idev);
 933                 printk(KERN_INFO
 934                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 935                 tmpaddr = &addr;
 936                 write_lock(&idev->lock);
 937                 goto retry;
 938         }
 939
 940         spin_lock_bh(&ift->lock);
 941         ift->ifpub = ifp;
 942         ift->valid_lft = tmp_valid_lft;
 943         ift->prefered_lft = tmp_prefered_lft;
 944         ift->cstamp = tmp_cstamp;
 945         ift->tstamp = tmp_tstamp;
 946         spin_unlock_bh(&ift->lock);
 947
 948         addrconf_dad_start(ift, 0);
 949         in6_ifa_put(ift);
 950         in6_dev_put(idev);
 951 out:
 952         return ret;
 953 }
 954 #endif
 955
 956 /*
 957  *      Choose an appropriate source address (RFC3484)
 958  */
 959 enum {
 960         IPV6_SADDR_RULE_INIT = 0,
 961         IPV6_SADDR_RULE_LOCAL,
 962         IPV6_SADDR_RULE_SCOPE,
 963         IPV6_SADDR_RULE_PREFERRED,
 964 #ifdef CONFIG_IPV6_MIP6
 965         IPV6_SADDR_RULE_HOA,
 966 #endif
 967         IPV6_SADDR_RULE_OIF,
 968         IPV6_SADDR_RULE_LABEL,
 969 #ifdef CONFIG_IPV6_PRIVACY
 970         IPV6_SADDR_RULE_PRIVACY,
 971 #endif
 972         IPV6_SADDR_RULE_ORCHID,
 973         IPV6_SADDR_RULE_PREFIX,
 974         IPV6_SADDR_RULE_MAX
 975 };
 976
 977 struct ipv6_saddr_score {
 978         int                     rule;
 979         int                     addr_type;
 980         struct inet6_ifaddr     *ifa;
 981         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 982         int                     scopedist;
 983         int                     matchlen;
 984 };
 985
 986 struct ipv6_saddr_dst {
 987         const struct in6_addr *addr;
 988         int ifindex;
 989         int scope;
 990         int label;
 991         unsigned int prefs;
 992 };
 993
 994 static inline int ipv6_saddr_preferred(int type)
 995 {
 996         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 997                 return 1;
 998         return 0;
 999 }
1000
1001 static int ipv6_get_saddr_eval(struct net *net,
1002                                struct ipv6_saddr_score *score,
1003                                struct ipv6_saddr_dst *dst,
1004                                int i)
1005 {
1006         int ret;
1007
1008         if (i <= score->rule) {
1009                 switch (i) {
1010                 case IPV6_SADDR_RULE_SCOPE:
1011                         ret = score->scopedist;
1012                         break;
1013                 case IPV6_SADDR_RULE_PREFIX:
1014                         ret = score->matchlen;
1015                         break;
1016                 default:
1017                         ret = !!test_bit(i, score->scorebits);
1018                 }
1019                 goto out;
1020         }
1021
1022         switch (i) {
1023         case IPV6_SADDR_RULE_INIT:
1024                 /* Rule 0: remember if hiscore is not ready yet */
1025                 ret = !!score->ifa;
1026                 break;
1027         case IPV6_SADDR_RULE_LOCAL:
1028                 /* Rule 1: Prefer same address */
1029                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1030                 break;
1031         case IPV6_SADDR_RULE_SCOPE:
1032                 /* Rule 2: Prefer appropriate scope
1033                  *
1034                  *      ret
1035                  *       ^
1036                  *    -1 |  d 15
1037                  *    ---+--+-+---> scope
1038                  *       |
1039                  *       |             d is scope of the destination.
1040                  *  B-d  |  \
1041                  *       |   \      <- smaller scope is better if
1042                  *  B-15 |    \        if scope is enough for destinaion.
1043                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1044                  * d-C-1 | /
1045                  *       |/         <- greater is better
1046                  *   -C  /             if scope is not enough for destination.
1047                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1048                  *
1049                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1050                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1051                  * Assume B = 0 and we get C > 29.
1052                  */
1053                 ret = __ipv6_addr_src_scope(score->addr_type);
1054                 if (ret >= dst->scope)
1055                         ret = -ret;
1056                 else
1057                         ret -= 128;     /* 30 is enough */
1058                 score->scopedist = ret;
1059                 break;
1060         case IPV6_SADDR_RULE_PREFERRED:
1061                 /* Rule 3: Avoid deprecated and optimistic addresses */
1062                 ret = ipv6_saddr_preferred(score->addr_type) ||
1063                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1064                 break;
1065 #ifdef CONFIG_IPV6_MIP6
1066         case IPV6_SADDR_RULE_HOA:
1067             {
1068                 /* Rule 4: Prefer home address */
1069                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1070                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1071                 break;
1072             }
1073 #endif
1074         case IPV6_SADDR_RULE_OIF:
1075                 /* Rule 5: Prefer outgoing interface */
1076                 ret = (!dst->ifindex ||
1077                        dst->ifindex == score->ifa->idev->dev->ifindex);
1078                 break;
1079         case IPV6_SADDR_RULE_LABEL:
1080                 /* Rule 6: Prefer matching label */
1081                 ret = ipv6_addr_label(net,
1082                                       &score->ifa->addr, score->addr_type,
1083                                       score->ifa->idev->dev->ifindex) == dst->label;
1084                 break;
1085 #ifdef CONFIG_IPV6_PRIVACY
1086         case IPV6_SADDR_RULE_PRIVACY:
1087             {
1088                 /* Rule 7: Prefer public address
1089                  * Note: prefer temprary address if use_tempaddr >= 2
1090                  */
1091                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1092                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1093                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1094                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1095                 break;
1096             }
1097 #endif
1098         case IPV6_SADDR_RULE_ORCHID:
1099                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1100                  *          non-ORCHID vs non-ORCHID
1101                  */
1102                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1103                         ipv6_addr_orchid(dst->addr));
1104                 break;
1105         case IPV6_SADDR_RULE_PREFIX:
1106                 /* Rule 8: Use longest matching prefix */
1107                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1108                                                        dst->addr);
1109                 break;
1110         default:
1111                 ret = 0;
1112         }
1113
1114         if (ret)
1115                 __set_bit(i, score->scorebits);
1116         score->rule = i;
1117 out:
1118         return ret;
1119 }
1120
1121 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1122                        const struct in6_addr *daddr, unsigned int prefs,
1123                        struct in6_addr *saddr)
1124 {
1125         struct ipv6_saddr_score scores[2],
1126                                 *score = &scores[0], *hiscore = &scores[1];
1127         struct ipv6_saddr_dst dst;
1128         struct net_device *dev;
1129         int dst_type;
1130
1131         dst_type = __ipv6_addr_type(daddr);
1132         dst.addr = daddr;
1133         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1134         dst.scope = __ipv6_addr_src_scope(dst_type);
1135         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1136         dst.prefs = prefs;
1137
1138         hiscore->rule = -1;
1139         hiscore->ifa = NULL;
1140
1141         rcu_read_lock();
1142
1143         for_each_netdev_rcu(net, dev) {
1144                 struct inet6_dev *idev;
1145
1146                 /* Candidate Source Address (section 4)
1147                  *  - multicast and link-local destination address,
1148                  *    the set of candidate source address MUST only
1149                  *    include addresses assigned to interfaces
1150                  *    belonging to the same link as the outgoing
1151                  *    interface.
1152                  * (- For site-local destination addresses, the
1153                  *    set of candidate source addresses MUST only
1154                  *    include addresses assigned to interfaces
1155                  *    belonging to the same site as the outgoing
1156                  *    interface.)
1157                  */
1158                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1159                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1160                     dst.ifindex && dev->ifindex != dst.ifindex)
1161                         continue;
1162
1163                 idev = __in6_dev_get(dev);
1164                 if (!idev)
1165                         continue;
1166
1167                 read_lock_bh(&idev->lock);
1168                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1169                         int i;
1170
1171                         /*
1172                          * - Tentative Address (RFC2462 section 5.4)
1173                          *  - A tentative address is not considered
1174                          *    "assigned to an interface" in the traditional
1175                          *    sense, unless it is also flagged as optimistic.
1176                          * - Candidate Source Address (section 4)
1177                          *  - In any case, anycast addresses, multicast
1178                          *    addresses, and the unspecified address MUST
1179                          *    NOT be included in a candidate set.
1180                          */
1181                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1182                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1183                                 continue;
1184
1185                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1186
1187                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1188                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1189                                 LIMIT_NETDEBUG(KERN_DEBUG
1190                                                "ADDRCONF: unspecified / multicast address "
1191                                                "assigned as unicast address on %s",
1192                                                dev->name);
1193                                 continue;
1194                         }
1195
1196                         score->rule = -1;
1197                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1198
1199                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1200                                 int minihiscore, miniscore;
1201
1202                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1203                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1204
1205                                 if (minihiscore > miniscore) {
1206                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1207                                             score->scopedist > 0) {
1208                                                 /*
1209                                                  * special case:
1210                                                  * each remaining entry
1211                                                  * has too small (not enough)
1212                                                  * scope, because ifa entries
1213                                                  * are sorted by their scope
1214                                                  * values.
1215                                                  */
1216                                                 goto try_nextdev;
1217                                         }
1218                                         break;
1219                                 } else if (minihiscore < miniscore) {
1220                                         if (hiscore->ifa)
1221                                                 in6_ifa_put(hiscore->ifa);
1222
1223                                         in6_ifa_hold(score->ifa);
1224
1225                                         swap(hiscore, score);
1226
1227                                         /* restore our iterator */
1228                                         score->ifa = hiscore->ifa;
1229
1230                                         break;
1231                                 }
1232                         }
1233                 }
1234 try_nextdev:
1235                 read_unlock_bh(&idev->lock);
1236         }
1237         rcu_read_unlock();
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249                     unsigned char banned_flags)
1250 {
1251         struct inet6_dev *idev;
1252         int err = -EADDRNOTAVAIL;
1253
1254         rcu_read_lock();
1255         idev = __in6_dev_get(dev);
1256         if (idev) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1261                         if (ifp->scope == IFA_LINK &&
1262                             !(ifp->flags & banned_flags)) {
1263                                 ipv6_addr_copy(addr, &ifp->addr);
1264                                 err = 0;
1265                                 break;
1266                         }
1267                 }
1268                 read_unlock_bh(&idev->lock);
1269         }
1270         rcu_read_unlock();
1271         return err;
1272 }
1273
1274 static int ipv6_count_addresses(struct inet6_dev *idev)
1275 {
1276         int cnt = 0;
1277         struct inet6_ifaddr *ifp;
1278
1279         read_lock_bh(&idev->lock);
1280         list_for_each_entry(ifp, &idev->addr_list, if_list)
1281                 cnt++;
1282         read_unlock_bh(&idev->lock);
1283         return cnt;
1284 }
1285
1286 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1287                   struct net_device *dev, int strict)
1288 {
1289         struct inet6_ifaddr *ifp;
1290         struct hlist_node *node;
1291         unsigned int hash = ipv6_addr_hash(addr);
1292
1293         rcu_read_lock_bh();
1294         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1295                 if (!net_eq(dev_net(ifp->idev->dev), net))
1296                         continue;
1297                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1298                     !(ifp->flags&IFA_F_TENTATIVE) &&
1299                     (dev == NULL || ifp->idev->dev == dev ||
1300                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1301                         rcu_read_unlock_bh();
1302                         return 1;
1303                 }
1304         }
1305
1306         rcu_read_unlock_bh();
1307         return 0;
1308 }
1309 EXPORT_SYMBOL(ipv6_chk_addr);
1310
1311 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1312                                struct net_device *dev)
1313 {
1314         unsigned int hash = ipv6_addr_hash(addr);
1315         struct inet6_ifaddr *ifp;
1316         struct hlist_node *node;
1317
1318         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1319                 if (!net_eq(dev_net(ifp->idev->dev), net))
1320                         continue;
1321                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1322                         if (dev == NULL || ifp->idev->dev == dev)
1323                                 return true;
1324                 }
1325         }
1326         return false;
1327 }
1328
1329 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1330 {
1331         struct inet6_dev *idev;
1332         struct inet6_ifaddr *ifa;
1333         int     onlink;
1334
1335         onlink = 0;
1336         rcu_read_lock();
1337         idev = __in6_dev_get(dev);
1338         if (idev) {
1339                 read_lock_bh(&idev->lock);
1340                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1341                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1342                                                    ifa->prefix_len);
1343                         if (onlink)
1344                                 break;
1345                 }
1346                 read_unlock_bh(&idev->lock);
1347         }
1348         rcu_read_unlock();
1349         return onlink;
1350 }
1351
1352 EXPORT_SYMBOL(ipv6_chk_prefix);
1353
1354 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1355                                      struct net_device *dev, int strict)
1356 {
1357         struct inet6_ifaddr *ifp, *result = NULL;
1358         unsigned int hash = ipv6_addr_hash(addr);
1359         struct hlist_node *node;
1360
1361         rcu_read_lock_bh();
1362         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1363                 if (!net_eq(dev_net(ifp->idev->dev), net))
1364                         continue;
1365                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1366                         if (dev == NULL || ifp->idev->dev == dev ||
1367                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1368                                 result = ifp;
1369                                 in6_ifa_hold(ifp);
1370                                 break;
1371                         }
1372                 }
1373         }
1374         rcu_read_unlock_bh();
1375
1376         return result;
1377 }
1378
1379 /* Gets referenced address, destroys ifaddr */
1380
1381 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1382 {
1383         if (ifp->flags&IFA_F_PERMANENT) {
1384                 spin_lock_bh(&ifp->lock);
1385                 addrconf_del_timer(ifp);
1386                 ifp->flags |= IFA_F_TENTATIVE;
1387                 if (dad_failed)
1388                         ifp->flags |= IFA_F_DADFAILED;
1389                 spin_unlock_bh(&ifp->lock);
1390                 if (dad_failed)
1391                         ipv6_ifa_notify(0, ifp);
1392                 in6_ifa_put(ifp);
1393 #ifdef CONFIG_IPV6_PRIVACY
1394         } else if (ifp->flags&IFA_F_TEMPORARY) {
1395                 struct inet6_ifaddr *ifpub;
1396                 spin_lock_bh(&ifp->lock);
1397                 ifpub = ifp->ifpub;
1398                 if (ifpub) {
1399                         in6_ifa_hold(ifpub);
1400                         spin_unlock_bh(&ifp->lock);
1401                         ipv6_create_tempaddr(ifpub, ifp);
1402                         in6_ifa_put(ifpub);
1403                 } else {
1404                         spin_unlock_bh(&ifp->lock);
1405                 }
1406                 ipv6_del_addr(ifp);
1407 #endif
1408         } else
1409                 ipv6_del_addr(ifp);
1410 }
1411
1412 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1413 {
1414         int err = -ENOENT;
1415
1416         spin_lock(&ifp->state_lock);
1417         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1418                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1419                 err = 0;
1420         }
1421         spin_unlock(&ifp->state_lock);
1422
1423         return err;
1424 }
1425
1426 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1427 {
1428         struct inet6_dev *idev = ifp->idev;
1429
1430         if (addrconf_dad_end(ifp))
1431                 return;
1432
1433         if (net_ratelimit())
1434                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1435                         ifp->idev->dev->name, &ifp->addr);
1436
1437         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1438                 struct in6_addr addr;
1439
1440                 addr.s6_addr32[0] = htonl(0xfe800000);
1441                 addr.s6_addr32[1] = 0;
1442
1443                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1444                     ipv6_addr_equal(&ifp->addr, &addr)) {
1445                         /* DAD failed for link-local based on MAC address */
1446                         idev->cnf.disable_ipv6 = 1;
1447
1448                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1449                                 ifp->idev->dev->name);
1450                 }
1451         }
1452
1453         addrconf_dad_stop(ifp, 1);
1454 }
1455
1456 /* Join to solicited addr multicast group. */
1457
1458 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1459 {
1460         struct in6_addr maddr;
1461
1462         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1463                 return;
1464
1465         addrconf_addr_solict_mult(addr, &maddr);
1466         ipv6_dev_mc_inc(dev, &maddr);
1467 }
1468
1469 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1470 {
1471         struct in6_addr maddr;
1472
1473         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1474                 return;
1475
1476         addrconf_addr_solict_mult(addr, &maddr);
1477         __ipv6_dev_mc_dec(idev, &maddr);
1478 }
1479
1480 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1481 {
1482         struct in6_addr addr;
1483         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1484         if (ipv6_addr_any(&addr))
1485                 return;
1486         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1487 }
1488
1489 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1490 {
1491         struct in6_addr addr;
1492         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1493         if (ipv6_addr_any(&addr))
1494                 return;
1495         __ipv6_dev_ac_dec(ifp->idev, &addr);
1496 }
1497
1498 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1499 {
1500         if (dev->addr_len != ETH_ALEN)
1501                 return -1;
1502         memcpy(eui, dev->dev_addr, 3);
1503         memcpy(eui + 5, dev->dev_addr + 3, 3);
1504
1505         /*
1506          * The zSeries OSA network cards can be shared among various
1507          * OS instances, but the OSA cards have only one MAC address.
1508          * This leads to duplicate address conflicts in conjunction
1509          * with IPv6 if more than one instance uses the same card.
1510          *
1511          * The driver for these cards can deliver a unique 16-bit
1512          * identifier for each instance sharing the same card.  It is
1513          * placed instead of 0xFFFE in the interface identifier.  The
1514          * "u" bit of the interface identifier is not inverted in this
1515          * case.  Hence the resulting interface identifier has local
1516          * scope according to RFC2373.
1517          */
1518         if (dev->dev_id) {
1519                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1520                 eui[4] = dev->dev_id & 0xFF;
1521         } else {
1522                 eui[3] = 0xFF;
1523                 eui[4] = 0xFE;
1524                 eui[0] ^= 2;
1525         }
1526         return 0;
1527 }
1528
1529 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1530 {
1531         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1532         if (dev->addr_len != ARCNET_ALEN)
1533                 return -1;
1534         memset(eui, 0, 7);
1535         eui[7] = *(u8*)dev->dev_addr;
1536         return 0;
1537 }
1538
1539 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1540 {
1541         if (dev->addr_len != INFINIBAND_ALEN)
1542                 return -1;
1543         memcpy(eui, dev->dev_addr + 12, 8);
1544         eui[0] |= 2;
1545         return 0;
1546 }
1547
1548 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1549 {
1550         if (addr == 0)
1551                 return -1;
1552         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1553                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1554                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1555                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1556                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1557                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1558         eui[1] = 0;
1559         eui[2] = 0x5E;
1560         eui[3] = 0xFE;
1561         memcpy(eui + 4, &addr, 4);
1562         return 0;
1563 }
1564
1565 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1566 {
1567         if (dev->priv_flags & IFF_ISATAP)
1568                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569         return -1;
1570 }
1571
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574         switch (dev->type) {
1575         case ARPHRD_ETHER:
1576         case ARPHRD_FDDI:
1577         case ARPHRD_IEEE802_TR:
1578                 return addrconf_ifid_eui48(eui, dev);
1579         case ARPHRD_ARCNET:
1580                 return addrconf_ifid_arcnet(eui, dev);
1581         case ARPHRD_INFINIBAND:
1582                 return addrconf_ifid_infiniband(eui, dev);
1583         case ARPHRD_SIT:
1584                 return addrconf_ifid_sit(eui, dev);
1585         }
1586         return -1;
1587 }
1588
1589 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1590 {
1591         int err = -1;
1592         struct inet6_ifaddr *ifp;
1593
1594         read_lock_bh(&idev->lock);
1595         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1596                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1597                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1598                         err = 0;
1599                         break;
1600                 }
1601         }
1602         read_unlock_bh(&idev->lock);
1603         return err;
1604 }
1605
1606 #ifdef CONFIG_IPV6_PRIVACY
1607 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1608 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1609 {
1610 regen:
1611         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1612         idev->rndid[0] &= ~0x02;
1613
1614         /*
1615          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1616          * check if generated address is not inappropriate
1617          *
1618          *  - Reserved subnet anycast (RFC 2526)
1619          *      11111101 11....11 1xxxxxxx
1620          *  - ISATAP (RFC4214) 6.1
1621          *      00-00-5E-FE-xx-xx-xx-xx
1622          *  - value 0
1623          *  - XXX: already assigned to an address on the device
1624          */
1625         if (idev->rndid[0] == 0xfd &&
1626             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1627             (idev->rndid[7]&0x80))
1628                 goto regen;
1629         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1630                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1631                         goto regen;
1632                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1633                         goto regen;
1634         }
1635
1636         return 0;
1637 }
1638
1639 static void ipv6_regen_rndid(unsigned long data)
1640 {
1641         struct inet6_dev *idev = (struct inet6_dev *) data;
1642         unsigned long expires;
1643
1644         rcu_read_lock_bh();
1645         write_lock_bh(&idev->lock);
1646
1647         if (idev->dead)
1648                 goto out;
1649
1650         if (__ipv6_regen_rndid(idev) < 0)
1651                 goto out;
1652
1653         expires = jiffies +
1654                 idev->cnf.temp_prefered_lft * HZ -
1655                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1656                 idev->cnf.max_desync_factor * HZ;
1657         if (time_before(expires, jiffies)) {
1658                 printk(KERN_WARNING
1659                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1660                         idev->dev->name);
1661                 goto out;
1662         }
1663
1664         if (!mod_timer(&idev->regen_timer, expires))
1665                 in6_dev_hold(idev);
1666
1667 out:
1668         write_unlock_bh(&idev->lock);
1669         rcu_read_unlock_bh();
1670         in6_dev_put(idev);
1671 }
1672
1673 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1674         int ret = 0;
1675
1676         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1677                 ret = __ipv6_regen_rndid(idev);
1678         return ret;
1679 }
1680 #endif
1681
1682 /*
1683  *      Add prefix route.
1684  */
1685
1686 static void
1687 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1688                       unsigned long expires, u32 flags)
1689 {
1690         struct fib6_config cfg = {
1691                 .fc_table = RT6_TABLE_PREFIX,
1692                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1693                 .fc_ifindex = dev->ifindex,
1694                 .fc_expires = expires,
1695                 .fc_dst_len = plen,
1696                 .fc_flags = RTF_UP | flags,
1697                 .fc_nlinfo.nl_net = dev_net(dev),
1698                 .fc_protocol = RTPROT_KERNEL,
1699         };
1700
1701         ipv6_addr_copy(&cfg.fc_dst, pfx);
1702
1703         /* Prevent useless cloning on PtP SIT.
1704            This thing is done here expecting that the whole
1705            class of non-broadcast devices need not cloning.
1706          */
1707 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1708         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1709                 cfg.fc_flags |= RTF_NONEXTHOP;
1710 #endif
1711
1712         ip6_route_add(&cfg);
1713 }
1714
1715 /* Create "default" multicast route to the interface */
1716
1717 static void addrconf_add_mroute(struct net_device *dev)
1718 {
1719         struct fib6_config cfg = {
1720                 .fc_table = RT6_TABLE_LOCAL,
1721                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1722                 .fc_ifindex = dev->ifindex,
1723                 .fc_dst_len = 8,
1724                 .fc_flags = RTF_UP,
1725                 .fc_nlinfo.nl_net = dev_net(dev),
1726         };
1727
1728         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1729
1730         ip6_route_add(&cfg);
1731 }
1732
1733 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1734 static void sit_route_add(struct net_device *dev)
1735 {
1736         struct fib6_config cfg = {
1737                 .fc_table = RT6_TABLE_MAIN,
1738                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1739                 .fc_ifindex = dev->ifindex,
1740                 .fc_dst_len = 96,
1741                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1742                 .fc_nlinfo.nl_net = dev_net(dev),
1743         };
1744
1745         /* prefix length - 96 bits "::d.d.d.d" */
1746         ip6_route_add(&cfg);
1747 }
1748 #endif
1749
1750 static void addrconf_add_lroute(struct net_device *dev)
1751 {
1752         struct in6_addr addr;
1753
1754         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1755         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1756 }
1757
1758 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1759 {
1760         struct inet6_dev *idev;
1761
1762         ASSERT_RTNL();
1763
1764         idev = ipv6_find_idev(dev);
1765         if (!idev)
1766                 return ERR_PTR(-ENOBUFS);
1767
1768         if (idev->cnf.disable_ipv6)
1769                 return ERR_PTR(-EACCES);
1770
1771         /* Add default multicast route */
1772         addrconf_add_mroute(dev);
1773
1774         /* Add link local route */
1775         addrconf_add_lroute(dev);
1776         return idev;
1777 }
1778
1779 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1780 {
1781         struct prefix_info *pinfo;
1782         __u32 valid_lft;
1783         __u32 prefered_lft;
1784         int addr_type;
1785         struct inet6_dev *in6_dev;
1786         struct net *net = dev_net(dev);
1787
1788         pinfo = (struct prefix_info *) opt;
1789
1790         if (len < sizeof(struct prefix_info)) {
1791                 ADBG(("addrconf: prefix option too short\n"));
1792                 return;
1793         }
1794
1795         /*
1796          *      Validation checks ([ADDRCONF], page 19)
1797          */
1798
1799         addr_type = ipv6_addr_type(&pinfo->prefix);
1800
1801         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1802                 return;
1803
1804         valid_lft = ntohl(pinfo->valid);
1805         prefered_lft = ntohl(pinfo->prefered);
1806
1807         if (prefered_lft > valid_lft) {
1808                 if (net_ratelimit())
1809                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1810                 return;
1811         }
1812
1813         in6_dev = in6_dev_get(dev);
1814
1815         if (in6_dev == NULL) {
1816                 if (net_ratelimit())
1817                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1818                 return;
1819         }
1820
1821         /*
1822          *      Two things going on here:
1823          *      1) Add routes for on-link prefixes
1824          *      2) Configure prefixes with the auto flag set
1825          */
1826
1827         if (pinfo->onlink) {
1828                 struct rt6_info *rt;
1829                 unsigned long rt_expires;
1830
1831                 /* Avoid arithmetic overflow. Really, we could
1832                  * save rt_expires in seconds, likely valid_lft,
1833                  * but it would require division in fib gc, that it
1834                  * not good.
1835                  */
1836                 if (HZ > USER_HZ)
1837                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1838                 else
1839                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1840
1841                 if (addrconf_finite_timeout(rt_expires))
1842                         rt_expires *= HZ;
1843
1844                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1845                                 dev->ifindex, 1);
1846
1847                 if (rt && addrconf_is_prefix_route(rt)) {
1848                         /* Autoconf prefix route */
1849                         if (valid_lft == 0) {
1850                                 ip6_del_rt(rt);
1851                                 rt = NULL;
1852                         } else if (addrconf_finite_timeout(rt_expires)) {
1853                                 /* not infinity */
1854                                 rt->rt6i_expires = jiffies + rt_expires;
1855                                 rt->rt6i_flags |= RTF_EXPIRES;
1856                         } else {
1857                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1858                                 rt->rt6i_expires = 0;
1859                         }
1860                 } else if (valid_lft) {
1861                         clock_t expires = 0;
1862                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1863                         if (addrconf_finite_timeout(rt_expires)) {
1864                                 /* not infinity */
1865                                 flags |= RTF_EXPIRES;
1866                                 expires = jiffies_to_clock_t(rt_expires);
1867                         }
1868                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1869                                               dev, expires, flags);
1870                 }
1871                 if (rt)
1872                         dst_release(&rt->dst);
1873         }
1874
1875         /* Try to figure out our local address for this prefix */
1876
1877         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1878                 struct inet6_ifaddr * ifp;
1879                 struct in6_addr addr;
1880                 int create = 0, update_lft = 0;
1881
1882                 if (pinfo->prefix_len == 64) {
1883                         memcpy(&addr, &pinfo->prefix, 8);
1884                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1885                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1886                                 in6_dev_put(in6_dev);
1887                                 return;
1888                         }
1889                         goto ok;
1890                 }
1891                 if (net_ratelimit())
1892                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1893                                pinfo->prefix_len);
1894                 in6_dev_put(in6_dev);
1895                 return;
1896
1897 ok:
1898
1899                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1900
1901                 if (ifp == NULL && valid_lft) {
1902                         int max_addresses = in6_dev->cnf.max_addresses;
1903                         u32 addr_flags = 0;
1904
1905 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1906                         if (in6_dev->cnf.optimistic_dad &&
1907                             !net->ipv6.devconf_all->forwarding)
1908                                 addr_flags = IFA_F_OPTIMISTIC;
1909 #endif
1910
1911                         /* Do not allow to create too much of autoconfigured
1912                          * addresses; this would be too easy way to crash kernel.
1913                          */
1914                         if (!max_addresses ||
1915                             ipv6_count_addresses(in6_dev) < max_addresses)
1916                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1917                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1918                                                     addr_flags);
1919
1920                         if (!ifp || IS_ERR(ifp)) {
1921                                 in6_dev_put(in6_dev);
1922                                 return;
1923                         }
1924
1925                         update_lft = create = 1;
1926                         ifp->cstamp = jiffies;
1927                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1928                 }
1929
1930                 if (ifp) {
1931                         int flags;
1932                         unsigned long now;
1933 #ifdef CONFIG_IPV6_PRIVACY
1934                         struct inet6_ifaddr *ift;
1935 #endif
1936                         u32 stored_lft;
1937
1938                         /* update lifetime (RFC2462 5.5.3 e) */
1939                         spin_lock(&ifp->lock);
1940                         now = jiffies;
1941                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1942                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1943                         else
1944                                 stored_lft = 0;
1945                         if (!update_lft && stored_lft) {
1946                                 if (valid_lft > MIN_VALID_LIFETIME ||
1947                                     valid_lft > stored_lft)
1948                                         update_lft = 1;
1949                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1950                                         /* valid_lft <= stored_lft is always true */
1951                                         /*
1952                                          * RFC 4862 Section 5.5.3e:
1953                                          * "Note that the preferred lifetime of
1954                                          *  the corresponding address is always
1955                                          *  reset to the Preferred Lifetime in
1956                                          *  the received Prefix Information
1957                                          *  option, regardless of whether the
1958                                          *  valid lifetime is also reset or
1959                                          *  ignored."
1960                                          *
1961                                          *  So if the preferred lifetime in
1962                                          *  this advertisement is different
1963                                          *  than what we have stored, but the
1964                                          *  valid lifetime is invalid, just
1965                                          *  reset prefered_lft.
1966                                          *
1967                                          *  We must set the valid lifetime
1968                                          *  to the stored lifetime since we'll
1969                                          *  be updating the timestamp below,
1970                                          *  else we'll set it back to the
1971                                          *  minumum.
1972                                          */
1973                                         if (prefered_lft != ifp->prefered_lft) {
1974                                                 valid_lft = stored_lft;
1975                                                 update_lft = 1;
1976                                         }
1977                                 } else {
1978                                         valid_lft = MIN_VALID_LIFETIME;
1979                                         if (valid_lft < prefered_lft)
1980                                                 prefered_lft = valid_lft;
1981                                         update_lft = 1;
1982                                 }
1983                         }
1984
1985                         if (update_lft) {
1986                                 ifp->valid_lft = valid_lft;
1987                                 ifp->prefered_lft = prefered_lft;
1988                                 ifp->tstamp = now;
1989                                 flags = ifp->flags;
1990                                 ifp->flags &= ~IFA_F_DEPRECATED;
1991                                 spin_unlock(&ifp->lock);
1992
1993                                 if (!(flags&IFA_F_TENTATIVE))
1994                                         ipv6_ifa_notify(0, ifp);
1995                         } else
1996                                 spin_unlock(&ifp->lock);
1997
1998 #ifdef CONFIG_IPV6_PRIVACY
1999                         read_lock_bh(&in6_dev->lock);
2000                         /* update all temporary addresses in the list */
2001                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2002                                 /*
2003                                  * When adjusting the lifetimes of an existing
2004                                  * temporary address, only lower the lifetimes.
2005                                  * Implementations must not increase the
2006                                  * lifetimes of an existing temporary address
2007                                  * when processing a Prefix Information Option.
2008                                  */
2009                                 if (ifp != ift->ifpub)
2010                                         continue;
2011
2012                                 spin_lock(&ift->lock);
2013                                 flags = ift->flags;
2014                                 if (ift->valid_lft > valid_lft &&
2015                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2016                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2017                                 if (ift->prefered_lft > prefered_lft &&
2018                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2019                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2020                                 spin_unlock(&ift->lock);
2021                                 if (!(flags&IFA_F_TENTATIVE))
2022                                         ipv6_ifa_notify(0, ift);
2023                         }
2024
2025                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2026                                 /*
2027                                  * When a new public address is created as described in [ADDRCONF],
2028                                  * also create a new temporary address. Also create a temporary
2029                                  * address if it's enabled but no temporary address currently exists.
2030                                  */
2031                                 read_unlock_bh(&in6_dev->lock);
2032                                 ipv6_create_tempaddr(ifp, NULL);
2033                         } else {
2034                                 read_unlock_bh(&in6_dev->lock);
2035                         }
2036 #endif
2037                         in6_ifa_put(ifp);
2038                         addrconf_verify(0);
2039                 }
2040         }
2041         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2042         in6_dev_put(in6_dev);
2043 }
2044
2045 /*
2046  *      Set destination address.
2047  *      Special case for SIT interfaces where we create a new "virtual"
2048  *      device.
2049  */
2050 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2051 {
2052         struct in6_ifreq ireq;
2053         struct net_device *dev;
2054         int err = -EINVAL;
2055
2056         rtnl_lock();
2057
2058         err = -EFAULT;
2059         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2060                 goto err_exit;
2061
2062         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2063
2064         err = -ENODEV;
2065         if (dev == NULL)
2066                 goto err_exit;
2067
2068 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2069         if (dev->type == ARPHRD_SIT) {
2070                 const struct net_device_ops *ops = dev->netdev_ops;
2071                 struct ifreq ifr;
2072                 struct ip_tunnel_parm p;
2073
2074                 err = -EADDRNOTAVAIL;
2075                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2076                         goto err_exit;
2077
2078                 memset(&p, 0, sizeof(p));
2079                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2080                 p.iph.saddr = 0;
2081                 p.iph.version = 4;
2082                 p.iph.ihl = 5;
2083                 p.iph.protocol = IPPROTO_IPV6;
2084                 p.iph.ttl = 64;
2085                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2086
2087                 if (ops->ndo_do_ioctl) {
2088                         mm_segment_t oldfs = get_fs();
2089
2090                         set_fs(KERNEL_DS);
2091                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2092                         set_fs(oldfs);
2093                 } else
2094                         err = -EOPNOTSUPP;
2095
2096                 if (err == 0) {
2097                         err = -ENOBUFS;
2098                         dev = __dev_get_by_name(net, p.name);
2099                         if (!dev)
2100                                 goto err_exit;
2101                         err = dev_open(dev);
2102                 }
2103         }
2104 #endif
2105
2106 err_exit:
2107         rtnl_unlock();
2108         return err;
2109 }
2110
2111 /*
2112  *      Manual configuration of address on an interface
2113  */
2114 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2115                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2116                           __u32 valid_lft)
2117 {
2118         struct inet6_ifaddr *ifp;
2119         struct inet6_dev *idev;
2120         struct net_device *dev;
2121         int scope;
2122         u32 flags;
2123         clock_t expires;
2124         unsigned long timeout;
2125
2126         ASSERT_RTNL();
2127
2128         if (plen > 128)
2129                 return -EINVAL;
2130
2131         /* check the lifetime */
2132         if (!valid_lft || prefered_lft > valid_lft)
2133                 return -EINVAL;
2134
2135         dev = __dev_get_by_index(net, ifindex);
2136         if (!dev)
2137                 return -ENODEV;
2138
2139         idev = addrconf_add_dev(dev);
2140         if (IS_ERR(idev))
2141                 return PTR_ERR(idev);
2142
2143         scope = ipv6_addr_scope(pfx);
2144
2145         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2146         if (addrconf_finite_timeout(timeout)) {
2147                 expires = jiffies_to_clock_t(timeout * HZ);
2148                 valid_lft = timeout;
2149                 flags = RTF_EXPIRES;
2150         } else {
2151                 expires = 0;
2152                 flags = 0;
2153                 ifa_flags |= IFA_F_PERMANENT;
2154         }
2155
2156         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2157         if (addrconf_finite_timeout(timeout)) {
2158                 if (timeout == 0)
2159                         ifa_flags |= IFA_F_DEPRECATED;
2160                 prefered_lft = timeout;
2161         }
2162
2163         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2164
2165         if (!IS_ERR(ifp)) {
2166                 spin_lock_bh(&ifp->lock);
2167                 ifp->valid_lft = valid_lft;
2168                 ifp->prefered_lft = prefered_lft;
2169                 ifp->tstamp = jiffies;
2170                 spin_unlock_bh(&ifp->lock);
2171
2172                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2173                                       expires, flags);
2174                 /*
2175                  * Note that section 3.1 of RFC 4429 indicates
2176                  * that the Optimistic flag should not be set for
2177                  * manually configured addresses
2178                  */
2179                 addrconf_dad_start(ifp, 0);
2180                 in6_ifa_put(ifp);
2181                 addrconf_verify(0);
2182                 return 0;
2183         }
2184
2185         return PTR_ERR(ifp);
2186 }
2187
2188 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2189                           unsigned int plen)
2190 {
2191         struct inet6_ifaddr *ifp;
2192         struct inet6_dev *idev;
2193         struct net_device *dev;
2194
2195         if (plen > 128)
2196                 return -EINVAL;
2197
2198         dev = __dev_get_by_index(net, ifindex);
2199         if (!dev)
2200                 return -ENODEV;
2201
2202         if ((idev = __in6_dev_get(dev)) == NULL)
2203                 return -ENXIO;
2204
2205         read_lock_bh(&idev->lock);
2206         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2207                 if (ifp->prefix_len == plen &&
2208                     ipv6_addr_equal(pfx, &ifp->addr)) {
2209                         in6_ifa_hold(ifp);
2210                         read_unlock_bh(&idev->lock);
2211
2212                         ipv6_del_addr(ifp);
2213
2214                         /* If the last address is deleted administratively,
2215                            disable IPv6 on this interface.
2216                          */
2217                         if (list_empty(&idev->addr_list))
2218                                 addrconf_ifdown(idev->dev, 1);
2219                         return 0;
2220                 }
2221         }
2222         read_unlock_bh(&idev->lock);
2223         return -EADDRNOTAVAIL;
2224 }
2225
2226
2227 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2228 {
2229         struct in6_ifreq ireq;
2230         int err;
2231
2232         if (!capable(CAP_NET_ADMIN))
2233                 return -EPERM;
2234
2235         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2236                 return -EFAULT;
2237
2238         rtnl_lock();
2239         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2240                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2241                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2242         rtnl_unlock();
2243         return err;
2244 }
2245
2246 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2247 {
2248         struct in6_ifreq ireq;
2249         int err;
2250
2251         if (!capable(CAP_NET_ADMIN))
2252                 return -EPERM;
2253
2254         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2255                 return -EFAULT;
2256
2257         rtnl_lock();
2258         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2259                              ireq.ifr6_prefixlen);
2260         rtnl_unlock();
2261         return err;
2262 }
2263
2264 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2265                      int plen, int scope)
2266 {
2267         struct inet6_ifaddr *ifp;
2268
2269         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2270         if (!IS_ERR(ifp)) {
2271                 spin_lock_bh(&ifp->lock);
2272                 ifp->flags &= ~IFA_F_TENTATIVE;
2273                 spin_unlock_bh(&ifp->lock);
2274                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2275                 in6_ifa_put(ifp);
2276         }
2277 }
2278
2279 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2280 static void sit_add_v4_addrs(struct inet6_dev *idev)
2281 {
2282         struct in6_addr addr;
2283         struct net_device *dev;
2284         struct net *net = dev_net(idev->dev);
2285         int scope;
2286
2287         ASSERT_RTNL();
2288
2289         memset(&addr, 0, sizeof(struct in6_addr));
2290         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2291
2292         if (idev->dev->flags&IFF_POINTOPOINT) {
2293                 addr.s6_addr32[0] = htonl(0xfe800000);
2294                 scope = IFA_LINK;
2295         } else {
2296                 scope = IPV6_ADDR_COMPATv4;
2297         }
2298
2299         if (addr.s6_addr32[3]) {
2300                 add_addr(idev, &addr, 128, scope);
2301                 return;
2302         }
2303
2304         for_each_netdev(net, dev) {
2305                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2306                 if (in_dev && (dev->flags & IFF_UP)) {
2307                         struct in_ifaddr * ifa;
2308
2309                         int flag = scope;
2310
2311                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2312                                 int plen;
2313
2314                                 addr.s6_addr32[3] = ifa->ifa_local;
2315
2316                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2317                                         continue;
2318                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2319                                         if (idev->dev->flags&IFF_POINTOPOINT)
2320                                                 continue;
2321                                         flag |= IFA_HOST;
2322                                 }
2323                                 if (idev->dev->flags&IFF_POINTOPOINT)
2324                                         plen = 64;
2325                                 else
2326                                         plen = 96;
2327
2328                                 add_addr(idev, &addr, plen, flag);
2329                         }
2330                 }
2331         }
2332 }
2333 #endif
2334
2335 static void init_loopback(struct net_device *dev)
2336 {
2337         struct inet6_dev  *idev;
2338
2339         /* ::1 */
2340
2341         ASSERT_RTNL();
2342
2343         if ((idev = ipv6_find_idev(dev)) == NULL) {
2344                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2345                 return;
2346         }
2347
2348         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2349 }
2350
2351 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2352 {
2353         struct inet6_ifaddr * ifp;
2354         u32 addr_flags = IFA_F_PERMANENT;
2355
2356 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2357         if (idev->cnf.optimistic_dad &&
2358             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2359                 addr_flags |= IFA_F_OPTIMISTIC;
2360 #endif
2361
2362
2363         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2364         if (!IS_ERR(ifp)) {
2365                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2366                 addrconf_dad_start(ifp, 0);
2367                 in6_ifa_put(ifp);
2368         }
2369 }
2370
2371 static void addrconf_dev_config(struct net_device *dev)
2372 {
2373         struct in6_addr addr;
2374         struct inet6_dev    * idev;
2375
2376         ASSERT_RTNL();
2377
2378         if ((dev->type != ARPHRD_ETHER) &&
2379             (dev->type != ARPHRD_FDDI) &&
2380             (dev->type != ARPHRD_IEEE802_TR) &&
2381             (dev->type != ARPHRD_ARCNET) &&
2382             (dev->type != ARPHRD_INFINIBAND)) {
2383                 /* Alas, we support only Ethernet autoconfiguration. */
2384                 return;
2385         }
2386
2387         idev = addrconf_add_dev(dev);
2388         if (IS_ERR(idev))
2389                 return;
2390
2391         memset(&addr, 0, sizeof(struct in6_addr));
2392         addr.s6_addr32[0] = htonl(0xFE800000);
2393
2394         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2395                 addrconf_add_linklocal(idev, &addr);
2396 }
2397
2398 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2399 static void addrconf_sit_config(struct net_device *dev)
2400 {
2401         struct inet6_dev *idev;
2402
2403         ASSERT_RTNL();
2404
2405         /*
2406          * Configure the tunnel with one of our IPv4
2407          * addresses... we should configure all of
2408          * our v4 addrs in the tunnel
2409          */
2410
2411         if ((idev = ipv6_find_idev(dev)) == NULL) {
2412                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2413                 return;
2414         }
2415
2416         if (dev->priv_flags & IFF_ISATAP) {
2417                 struct in6_addr addr;
2418
2419                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2420                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2421                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2422                         addrconf_add_linklocal(idev, &addr);
2423                 return;
2424         }
2425
2426         sit_add_v4_addrs(idev);
2427
2428         if (dev->flags&IFF_POINTOPOINT) {
2429                 addrconf_add_mroute(dev);
2430                 addrconf_add_lroute(dev);
2431         } else
2432                 sit_route_add(dev);
2433 }
2434 #endif
2435
2436 static inline int
2437 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2438 {
2439         struct in6_addr lladdr;
2440
2441         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2442                 addrconf_add_linklocal(idev, &lladdr);
2443                 return 0;
2444         }
2445         return -1;
2446 }
2447
2448 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2449 {
2450         struct net_device *link_dev;
2451         struct net *net = dev_net(idev->dev);
2452
2453         /* first try to inherit the link-local address from the link device */
2454         if (idev->dev->iflink &&
2455             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2456                 if (!ipv6_inherit_linklocal(idev, link_dev))
2457                         return;
2458         }
2459         /* then try to inherit it from any device */
2460         for_each_netdev(net, link_dev) {
2461                 if (!ipv6_inherit_linklocal(idev, link_dev))
2462                         return;
2463         }
2464         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2465 }
2466
2467 /*
2468  * Autoconfigure tunnel with a link-local address so routing protocols,
2469  * DHCPv6, MLD etc. can be run over the virtual link
2470  */
2471
2472 static void addrconf_ip6_tnl_config(struct net_device *dev)
2473 {
2474         struct inet6_dev *idev;
2475
2476         ASSERT_RTNL();
2477
2478         idev = addrconf_add_dev(dev);
2479         if (IS_ERR(idev)) {
2480                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2481                 return;
2482         }
2483         ip6_tnl_add_linklocal(idev);
2484 }
2485
2486 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2487                            void * data)
2488 {
2489         struct net_device *dev = (struct net_device *) data;
2490         struct inet6_dev *idev = __in6_dev_get(dev);
2491         int run_pending = 0;
2492         int err;
2493
2494         switch (event) {
2495         case NETDEV_REGISTER:
2496                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2497                         idev = ipv6_add_dev(dev);
2498                         if (!idev)
2499                                 return notifier_from_errno(-ENOMEM);
2500                 }
2501                 break;
2502
2503         case NETDEV_UP:
2504         case NETDEV_CHANGE:
2505                 if (dev->flags & IFF_SLAVE)
2506                         break;
2507
2508                 if (event == NETDEV_UP) {
2509                         if (!addrconf_qdisc_ok(dev)) {
2510                                 /* device is not ready yet. */
2511                                 printk(KERN_INFO
2512                                         "ADDRCONF(NETDEV_UP): %s: "
2513                                         "link is not ready\n",
2514                                         dev->name);
2515                                 break;
2516                         }
2517
2518                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2519                                 idev = ipv6_add_dev(dev);
2520
2521                         if (idev) {
2522                                 idev->if_flags |= IF_READY;
2523                                 run_pending = 1;
2524                         }
2525                 } else {
2526                         if (!addrconf_qdisc_ok(dev)) {
2527                                 /* device is still not ready. */
2528                                 break;
2529                         }
2530
2531                         if (idev) {
2532                                 if (idev->if_flags & IF_READY)
2533                                         /* device is already configured. */
2534                                         break;
2535                                 idev->if_flags |= IF_READY;
2536                         }
2537
2538                         printk(KERN_INFO
2539                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2540                                         "link becomes ready\n",
2541                                         dev->name);
2542
2543                         run_pending = 1;
2544                 }
2545
2546                 switch (dev->type) {
2547 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2548                 case ARPHRD_SIT:
2549                         addrconf_sit_config(dev);
2550                         break;
2551 #endif
2552                 case ARPHRD_TUNNEL6:
2553                         addrconf_ip6_tnl_config(dev);
2554                         break;
2555                 case ARPHRD_LOOPBACK:
2556                         init_loopback(dev);
2557                         break;
2558
2559                 default:
2560                         addrconf_dev_config(dev);
2561                         break;
2562                 }
2563
2564                 if (idev) {
2565                         if (run_pending)
2566                                 addrconf_dad_run(idev);
2567
2568                         /*
2569                          * If the MTU changed during the interface down,
2570                          * when the interface up, the changed MTU must be
2571                          * reflected in the idev as well as routers.
2572                          */
2573                         if (idev->cnf.mtu6 != dev->mtu &&
2574                             dev->mtu >= IPV6_MIN_MTU) {
2575                                 rt6_mtu_change(dev, dev->mtu);
2576                                 idev->cnf.mtu6 = dev->mtu;
2577                         }
2578                         idev->tstamp = jiffies;
2579                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2580
2581                         /*
2582                          * If the changed mtu during down is lower than
2583                          * IPV6_MIN_MTU stop IPv6 on this interface.
2584                          */
2585                         if (dev->mtu < IPV6_MIN_MTU)
2586                                 addrconf_ifdown(dev, 1);
2587                 }
2588                 break;
2589
2590         case NETDEV_CHANGEMTU:
2591                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2592                         rt6_mtu_change(dev, dev->mtu);
2593                         idev->cnf.mtu6 = dev->mtu;
2594                         break;
2595                 }
2596
2597                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2598                         idev = ipv6_add_dev(dev);
2599                         if (idev)
2600                                 break;
2601                 }
2602
2603                 /*
2604                  * MTU falled under IPV6_MIN_MTU.
2605                  * Stop IPv6 on this interface.
2606                  */
2607
2608         case NETDEV_DOWN:
2609         case NETDEV_UNREGISTER:
2610                 /*
2611                  *      Remove all addresses from this interface.
2612                  */
2613                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2614                 break;
2615
2616         case NETDEV_CHANGENAME:
2617                 if (idev) {
2618                         snmp6_unregister_dev(idev);
2619                         addrconf_sysctl_unregister(idev);
2620                         addrconf_sysctl_register(idev);
2621                         err = snmp6_register_dev(idev);
2622                         if (err)
2623                                 return notifier_from_errno(err);
2624                 }
2625                 break;
2626
2627         case NETDEV_PRE_TYPE_CHANGE:
2628         case NETDEV_POST_TYPE_CHANGE:
2629                 addrconf_type_change(dev, event);
2630                 break;
2631         }
2632
2633         return NOTIFY_OK;
2634 }
2635
2636 /*
2637  *      addrconf module should be notified of a device going up
2638  */
2639 static struct notifier_block ipv6_dev_notf = {
2640         .notifier_call = addrconf_notify,
2641 };
2642
2643 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2644 {
2645         struct inet6_dev *idev;
2646         ASSERT_RTNL();
2647
2648         idev = __in6_dev_get(dev);
2649
2650         if (event == NETDEV_POST_TYPE_CHANGE)
2651                 ipv6_mc_remap(idev);
2652         else if (event == NETDEV_PRE_TYPE_CHANGE)
2653                 ipv6_mc_unmap(idev);
2654 }
2655
2656 static int addrconf_ifdown(struct net_device *dev, int how)
2657 {
2658         struct net *net = dev_net(dev);
2659         struct inet6_dev *idev;
2660         struct inet6_ifaddr *ifa;
2661         LIST_HEAD(keep_list);
2662         int state;
2663
2664         ASSERT_RTNL();
2665
2666         rt6_ifdown(net, dev);
2667         neigh_ifdown(&nd_tbl, dev);
2668
2669         idev = __in6_dev_get(dev);
2670         if (idev == NULL)
2671                 return -ENODEV;
2672
2673         /*
2674          * Step 1: remove reference to ipv6 device from parent device.
2675          *         Do not dev_put!
2676          */
2677         if (how) {
2678                 idev->dead = 1;
2679
2680                 /* protected by rtnl_lock */
2681                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2682
2683                 /* Step 1.5: remove snmp6 entry */
2684                 snmp6_unregister_dev(idev);
2685
2686         }
2687
2688         write_lock_bh(&idev->lock);
2689
2690         /* Step 2: clear flags for stateless addrconf */
2691         if (!how)
2692                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2693
2694 #ifdef CONFIG_IPV6_PRIVACY
2695         if (how && del_timer(&idev->regen_timer))
2696                 in6_dev_put(idev);
2697
2698         /* Step 3: clear tempaddr list */
2699         while (!list_empty(&idev->tempaddr_list)) {
2700                 ifa = list_first_entry(&idev->tempaddr_list,
2701                                        struct inet6_ifaddr, tmp_list);
2702                 list_del(&ifa->tmp_list);
2703                 write_unlock_bh(&idev->lock);
2704                 spin_lock_bh(&ifa->lock);
2705
2706                 if (ifa->ifpub) {
2707                         in6_ifa_put(ifa->ifpub);
2708                         ifa->ifpub = NULL;
2709                 }
2710                 spin_unlock_bh(&ifa->lock);
2711                 in6_ifa_put(ifa);
2712                 write_lock_bh(&idev->lock);
2713         }
2714 #endif
2715
2716         while (!list_empty(&idev->addr_list)) {
2717                 ifa = list_first_entry(&idev->addr_list,
2718                                        struct inet6_ifaddr, if_list);
2719                 addrconf_del_timer(ifa);
2720
2721                 /* If just doing link down, and address is permanent
2722                    and not link-local, then retain it. */
2723                 if (!how &&
2724                     (ifa->flags&IFA_F_PERMANENT) &&
2725                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2726                         list_move_tail(&ifa->if_list, &keep_list);
2727
2728                         /* If not doing DAD on this address, just keep it. */
2729                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2730                             idev->cnf.accept_dad <= 0 ||
2731                             (ifa->flags & IFA_F_NODAD))
2732                                 continue;
2733
2734                         /* If it was tentative already, no need to notify */
2735                         if (ifa->flags & IFA_F_TENTATIVE)
2736                                 continue;
2737
2738                         /* Flag it for later restoration when link comes up */
2739                         ifa->flags |= IFA_F_TENTATIVE;
2740                         ifa->state = INET6_IFADDR_STATE_DAD;
2741
2742                         write_unlock_bh(&idev->lock);
2743
2744                         in6_ifa_hold(ifa);
2745                 } else {
2746                         list_del(&ifa->if_list);
2747
2748                         /* clear hash table */
2749                         spin_lock_bh(&addrconf_hash_lock);
2750                         hlist_del_init_rcu(&ifa->addr_lst);
2751                         spin_unlock_bh(&addrconf_hash_lock);
2752
2753                         write_unlock_bh(&idev->lock);
2754                         spin_lock_bh(&ifa->state_lock);
2755                         state = ifa->state;
2756                         ifa->state = INET6_IFADDR_STATE_DEAD;
2757                         spin_unlock_bh(&ifa->state_lock);
2758
2759                         if (state == INET6_IFADDR_STATE_DEAD)
2760                                 goto put_ifa;
2761                 }
2762
2763                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2764                 if (ifa->state == INET6_IFADDR_STATE_DEAD)
2765                         atomic_notifier_call_chain(&inet6addr_chain,
2766                                                    NETDEV_DOWN, ifa);
2767
2768 put_ifa:
2769                 in6_ifa_put(ifa);
2770
2771                 write_lock_bh(&idev->lock);
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 = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3454                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3455         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3456                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3457         ci.ifa_prefered = preferred;
3458         ci.ifa_valid = valid;
3459
3460         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3461 }
3462
3463 static inline int rt_scope(int ifa_scope)
3464 {
3465         if (ifa_scope & IFA_HOST)
3466                 return RT_SCOPE_HOST;
3467         else if (ifa_scope & IFA_LINK)
3468                 return RT_SCOPE_LINK;
3469         else if (ifa_scope & IFA_SITE)
3470                 return RT_SCOPE_SITE;
3471         else
3472                 return RT_SCOPE_UNIVERSE;
3473 }
3474
3475 static inline int inet6_ifaddr_msgsize(void)
3476 {
3477         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3478                + nla_total_size(16) /* IFA_ADDRESS */
3479                + nla_total_size(sizeof(struct ifa_cacheinfo));
3480 }
3481
3482 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3483                              u32 pid, u32 seq, int event, unsigned int flags)
3484 {
3485         struct nlmsghdr  *nlh;
3486         u32 preferred, valid;
3487
3488         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3489         if (nlh == NULL)
3490                 return -EMSGSIZE;
3491
3492         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3493                       ifa->idev->dev->ifindex);
3494
3495         if (!(ifa->flags&IFA_F_PERMANENT)) {
3496                 preferred = ifa->prefered_lft;
3497                 valid = ifa->valid_lft;
3498                 if (preferred != INFINITY_LIFE_TIME) {
3499                         long tval = (jiffies - ifa->tstamp)/HZ;
3500                         if (preferred > tval)
3501                                 preferred -= tval;
3502                         else
3503                                 preferred = 0;
3504                         if (valid != INFINITY_LIFE_TIME) {
3505                                 if (valid > tval)
3506                                         valid -= tval;
3507                                 else
3508                                         valid = 0;
3509                         }
3510                 }
3511         } else {
3512                 preferred = INFINITY_LIFE_TIME;
3513                 valid = INFINITY_LIFE_TIME;
3514         }
3515
3516         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3517             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3518                 nlmsg_cancel(skb, nlh);
3519                 return -EMSGSIZE;
3520         }
3521
3522         return nlmsg_end(skb, nlh);
3523 }
3524
3525 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3526                                 u32 pid, u32 seq, int event, u16 flags)
3527 {
3528         struct nlmsghdr  *nlh;
3529         u8 scope = RT_SCOPE_UNIVERSE;
3530         int ifindex = ifmca->idev->dev->ifindex;
3531
3532         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3533                 scope = RT_SCOPE_SITE;
3534
3535         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3536         if (nlh == NULL)
3537                 return -EMSGSIZE;
3538
3539         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3540         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3541             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3542                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3543                 nlmsg_cancel(skb, nlh);
3544                 return -EMSGSIZE;
3545         }
3546
3547         return nlmsg_end(skb, nlh);
3548 }
3549
3550 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3551                                 u32 pid, u32 seq, int event, unsigned int flags)
3552 {
3553         struct nlmsghdr  *nlh;
3554         u8 scope = RT_SCOPE_UNIVERSE;
3555         int ifindex = ifaca->aca_idev->dev->ifindex;
3556
3557         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3558                 scope = RT_SCOPE_SITE;
3559
3560         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3561         if (nlh == NULL)
3562                 return -EMSGSIZE;
3563
3564         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3565         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3566             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3567                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3568                 nlmsg_cancel(skb, nlh);
3569                 return -EMSGSIZE;
3570         }
3571
3572         return nlmsg_end(skb, nlh);
3573 }
3574
3575 enum addr_type_t {
3576         UNICAST_ADDR,
3577         MULTICAST_ADDR,
3578         ANYCAST_ADDR,
3579 };
3580
3581 /* called with rcu_read_lock() */
3582 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3583                           struct netlink_callback *cb, enum addr_type_t type,
3584                           int s_ip_idx, int *p_ip_idx)
3585 {
3586         struct ifmcaddr6 *ifmca;
3587         struct ifacaddr6 *ifaca;
3588         int err = 1;
3589         int ip_idx = *p_ip_idx;
3590
3591         read_lock_bh(&idev->lock);
3592         switch (type) {
3593         case UNICAST_ADDR: {
3594                 struct inet6_ifaddr *ifa;
3595
3596                 /* unicast address incl. temp addr */
3597                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3598                         if (++ip_idx < s_ip_idx)
3599                                 continue;
3600                         err = inet6_fill_ifaddr(skb, ifa,
3601                                                 NETLINK_CB(cb->skb).pid,
3602                                                 cb->nlh->nlmsg_seq,
3603                                                 RTM_NEWADDR,
3604                                                 NLM_F_MULTI);
3605                         if (err <= 0)
3606                                 break;
3607                 }
3608                 break;
3609         }
3610         case MULTICAST_ADDR:
3611                 /* multicast address */
3612                 for (ifmca = idev->mc_list; ifmca;
3613                      ifmca = ifmca->next, ip_idx++) {
3614                         if (ip_idx < s_ip_idx)
3615                                 continue;
3616                         err = inet6_fill_ifmcaddr(skb, ifmca,
3617                                                   NETLINK_CB(cb->skb).pid,
3618                                                   cb->nlh->nlmsg_seq,
3619                                                   RTM_GETMULTICAST,
3620                                                   NLM_F_MULTI);
3621                         if (err <= 0)
3622                                 break;
3623                 }
3624                 break;
3625         case ANYCAST_ADDR:
3626                 /* anycast address */
3627                 for (ifaca = idev->ac_list; ifaca;
3628                      ifaca = ifaca->aca_next, ip_idx++) {
3629                         if (ip_idx < s_ip_idx)
3630                                 continue;
3631                         err = inet6_fill_ifacaddr(skb, ifaca,
3632                                                   NETLINK_CB(cb->skb).pid,
3633                                                   cb->nlh->nlmsg_seq,
3634                                                   RTM_GETANYCAST,
3635                                                   NLM_F_MULTI);
3636                         if (err <= 0)
3637                                 break;
3638                 }
3639                 break;
3640         default:
3641                 break;
3642         }
3643         read_unlock_bh(&idev->lock);
3644         *p_ip_idx = ip_idx;
3645         return err;
3646 }
3647
3648 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3649                            enum addr_type_t type)
3650 {
3651         struct net *net = sock_net(skb->sk);
3652         int h, s_h;
3653         int idx, ip_idx;
3654         int s_idx, s_ip_idx;
3655         struct net_device *dev;
3656         struct inet6_dev *idev;
3657         struct hlist_head *head;
3658         struct hlist_node *node;
3659
3660         s_h = cb->args[0];
3661         s_idx = idx = cb->args[1];
3662         s_ip_idx = ip_idx = cb->args[2];
3663
3664         rcu_read_lock();
3665         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3666                 idx = 0;
3667                 head = &net->dev_index_head[h];
3668                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3669                         if (idx < s_idx)
3670                                 goto cont;
3671                         if (h > s_h || idx > s_idx)
3672                                 s_ip_idx = 0;
3673                         ip_idx = 0;
3674                         idev = __in6_dev_get(dev);
3675                         if (!idev)
3676                                 goto cont;
3677
3678                         if (in6_dump_addrs(idev, skb, cb, type,
3679                                            s_ip_idx, &ip_idx) <= 0)
3680                                 goto done;
3681 cont:
3682                         idx++;
3683                 }
3684         }
3685 done:
3686         rcu_read_unlock();
3687         cb->args[0] = h;
3688         cb->args[1] = idx;
3689         cb->args[2] = ip_idx;
3690
3691         return skb->len;
3692 }
3693
3694 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3695 {
3696         enum addr_type_t type = UNICAST_ADDR;
3697
3698         return inet6_dump_addr(skb, cb, type);
3699 }
3700
3701 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3702 {
3703         enum addr_type_t type = MULTICAST_ADDR;
3704
3705         return inet6_dump_addr(skb, cb, type);
3706 }
3707
3708
3709 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3710 {
3711         enum addr_type_t type = ANYCAST_ADDR;
3712
3713         return inet6_dump_addr(skb, cb, type);
3714 }
3715
3716 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3717                              void *arg)
3718 {
3719         struct net *net = sock_net(in_skb->sk);
3720         struct ifaddrmsg *ifm;
3721         struct nlattr *tb[IFA_MAX+1];
3722         struct in6_addr *addr = NULL;
3723         struct net_device *dev = NULL;
3724         struct inet6_ifaddr *ifa;
3725         struct sk_buff *skb;
3726         int err;
3727
3728         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3729         if (err < 0)
3730                 goto errout;
3731
3732         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3733         if (addr == NULL) {
3734                 err = -EINVAL;
3735                 goto errout;
3736         }
3737
3738         ifm = nlmsg_data(nlh);
3739         if (ifm->ifa_index)
3740                 dev = __dev_get_by_index(net, ifm->ifa_index);
3741
3742         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3743         if (!ifa) {
3744                 err = -EADDRNOTAVAIL;
3745                 goto errout;
3746         }
3747
3748         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3749         if (!skb) {
3750                 err = -ENOBUFS;
3751                 goto errout_ifa;
3752         }
3753
3754         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3755                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3756         if (err < 0) {
3757                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3758                 WARN_ON(err == -EMSGSIZE);
3759                 kfree_skb(skb);
3760                 goto errout_ifa;
3761         }
3762         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3763 errout_ifa:
3764         in6_ifa_put(ifa);
3765 errout:
3766         return err;
3767 }
3768
3769 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3770 {
3771         struct sk_buff *skb;
3772         struct net *net = dev_net(ifa->idev->dev);
3773         int err = -ENOBUFS;
3774
3775         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3776         if (skb == NULL)
3777                 goto errout;
3778
3779         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3780         if (err < 0) {
3781                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3782                 WARN_ON(err == -EMSGSIZE);
3783                 kfree_skb(skb);
3784                 goto errout;
3785         }
3786         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3787         return;
3788 errout:
3789         if (err < 0)
3790                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3791 }
3792
3793 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3794                                 __s32 *array, int bytes)
3795 {
3796         BUG_ON(bytes < (DEVCONF_MAX * 4));
3797
3798         memset(array, 0, bytes);
3799         array[DEVCONF_FORWARDING] = cnf->forwarding;
3800         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3801         array[DEVCONF_MTU6] = cnf->mtu6;
3802         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3803         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3804         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3805         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3806         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3807         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3808         array[DEVCONF_RTR_SOLICIT_DELAY] = 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] = cnf->rtr_probe_interval;
3823 #ifdef CONFIG_IPV6_ROUTE_INFO
3824         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3825 #endif
3826 #endif
3827         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3828         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3829 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3830         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3831 #endif
3832 #ifdef CONFIG_IPV6_MROUTE
3833         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3834 #endif
3835         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3836         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3837         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3838 }
3839
3840 static inline size_t inet6_if_nlmsg_size(void)
3841 {
3842         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3843                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3844                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3845                + nla_total_size(4) /* IFLA_MTU */
3846                + nla_total_size(4) /* IFLA_LINK */
3847                + nla_total_size( /* IFLA_PROTINFO */
3848                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3849                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3850                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3851                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3852                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3853                  );
3854 }
3855
3856 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3857                                       int items, int bytes)
3858 {
3859         int i;
3860         int pad = bytes - sizeof(u64) * items;
3861         BUG_ON(pad < 0);
3862
3863         /* Use put_unaligned() because stats may not be aligned for u64. */
3864         put_unaligned(items, &stats[0]);
3865         for (i = 1; i < items; i++)
3866                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3867
3868         memset(&stats[items], 0, pad);
3869 }
3870
3871 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3872                                       int items, int bytes, size_t syncpoff)
3873 {
3874         int i;
3875         int pad = bytes - sizeof(u64) * items;
3876         BUG_ON(pad < 0);
3877
3878         /* Use put_unaligned() because stats may not be aligned for u64. */
3879         put_unaligned(items, &stats[0]);
3880         for (i = 1; i < items; i++)
3881                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3882
3883         memset(&stats[items], 0, pad);
3884 }
3885
3886 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3887                              int bytes)
3888 {
3889         switch (attrtype) {
3890         case IFLA_INET6_STATS:
3891                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3892                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3893                 break;
3894         case IFLA_INET6_ICMP6STATS:
3895                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3896                 break;
3897         }
3898 }
3899
3900 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3901                              u32 pid, u32 seq, int event, unsigned int flags)
3902 {
3903         struct net_device *dev = idev->dev;
3904         struct nlattr *nla;
3905         struct ifinfomsg *hdr;
3906         struct nlmsghdr *nlh;
3907         void *protoinfo;
3908         struct ifla_cacheinfo ci;
3909
3910         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3911         if (nlh == NULL)
3912                 return -EMSGSIZE;
3913
3914         hdr = nlmsg_data(nlh);
3915         hdr->ifi_family = AF_INET6;
3916         hdr->__ifi_pad = 0;
3917         hdr->ifi_type = dev->type;
3918         hdr->ifi_index = dev->ifindex;
3919         hdr->ifi_flags = dev_get_flags(dev);
3920         hdr->ifi_change = 0;
3921
3922         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3923
3924         if (dev->addr_len)
3925                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3926
3927         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3928         if (dev->ifindex != dev->iflink)
3929                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3930
3931         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3932         if (protoinfo == NULL)
3933                 goto nla_put_failure;
3934
3935         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3936
3937         ci.max_reasm_len = IPV6_MAXPLEN;
3938         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3939                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3940         ci.reachable_time = idev->nd_parms->reachable_time;
3941         ci.retrans_time = 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_IFADDR, NULL, GFP_ATOMIC);
4027         return;
4028 errout:
4029         if (err < 0)
4030                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, 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 }