net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 110
 111 #ifdef CONFIG_SYSCTL
 112 static void addrconf_sysctl_register(struct inet6_dev *idev);
 113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 114 #else
 115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 116 {
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 #ifdef CONFIG_IPV6_PRIVACY
 125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 127 static void ipv6_regen_rndid(unsigned long data);
 128 #endif
 129
 130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 131 static int ipv6_count_addresses(struct inet6_dev *idev);
 132
 133 /*
 134  *      Configured unicast address hash table
 135  */
 136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 137 static DEFINE_SPINLOCK(addrconf_hash_lock);
 138
 139 static void addrconf_verify(unsigned long);
 140
 141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 142 static DEFINE_SPINLOCK(addrconf_verify_lock);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 152 static void addrconf_dad_timer(unsigned long data);
 153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_run(struct inet6_dev *idev);
 155 static void addrconf_rs_timer(unsigned long data);
 156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158
 159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 160                                 struct prefix_info *pinfo);
 161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 162                                struct net_device *dev);
 163
 164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .dad_transmits          = 1,
 175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 178 #ifdef CONFIG_IPV6_PRIVACY
 179         .use_tempaddr           = 0,
 180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 182         .regen_max_retry        = REGEN_MAX_RETRY,
 183         .max_desync_factor      = MAX_DESYNC_FACTOR,
 184 #endif
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199 };
 200
 201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 202         .forwarding             = 0,
 203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 204         .mtu6                   = IPV6_MIN_MTU,
 205         .accept_ra              = 1,
 206         .accept_redirects       = 1,
 207         .autoconf               = 1,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212 #ifdef CONFIG_IPV6_PRIVACY
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218 #endif
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233 };
 234
 235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 240
 241 /* Check if a valid qdisc is available */
 242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 243 {
 244         return !qdisc_tx_is_noop(dev);
 245 }
 246
 247 /* Check if a route is valid prefix route */
 248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 249 {
 250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 251 }
 252
 253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 254 {
 255         if (del_timer(&ifp->timer))
 256                 __in6_ifa_put(ifp);
 257 }
 258
 259 enum addrconf_timer_t {
 260         AC_NONE,
 261         AC_DAD,
 262         AC_RS,
 263 };
 264
 265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 266                                enum addrconf_timer_t what,
 267                                unsigned long when)
 268 {
 269         if (!del_timer(&ifp->timer))
 270                 in6_ifa_hold(ifp);
 271
 272         switch (what) {
 273         case AC_DAD:
 274                 ifp->timer.function = addrconf_dad_timer;
 275                 break;
 276         case AC_RS:
 277                 ifp->timer.function = addrconf_rs_timer;
 278                 break;
 279         default:
 280                 break;
 281         }
 282         ifp->timer.expires = jiffies + when;
 283         add_timer(&ifp->timer);
 284 }
 285
 286 static int snmp6_alloc_dev(struct inet6_dev *idev)
 287 {
 288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 289                           sizeof(struct ipstats_mib),
 290                           __alignof__(struct ipstats_mib)) < 0)
 291                 goto err_ip;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 293                           sizeof(struct icmpv6_mib),
 294                           __alignof__(struct icmpv6_mib)) < 0)
 295                 goto err_icmp;
 296         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 297                           sizeof(struct icmpv6msg_mib),
 298                           __alignof__(struct icmpv6msg_mib)) < 0)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static void snmp6_free_dev(struct inet6_dev *idev)
 312 {
 313         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 314         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 321 {
 322         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 323         kfree(idev);
 324 }
 325
 326 void in6_dev_finish_destroy(struct inet6_dev *idev)
 327 {
 328         struct net_device *dev = idev->dev;
 329
 330         WARN_ON(!list_empty(&idev->addr_list));
 331         WARN_ON(idev->mc_list != NULL);
 332
 333 #ifdef NET_REFCNT_DEBUG
 334         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 335 #endif
 336         dev_put(dev);
 337         if (!idev->dead) {
 338                 pr_warning("Freeing alive inet6 device %p\n", idev);
 339                 return;
 340         }
 341         snmp6_free_dev(idev);
 342         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 343 }
 344
 345 EXPORT_SYMBOL(in6_dev_finish_destroy);
 346
 347 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 348 {
 349         struct inet6_dev *ndev;
 350
 351         ASSERT_RTNL();
 352
 353         if (dev->mtu < IPV6_MIN_MTU)
 354                 return NULL;
 355
 356         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 357
 358         if (ndev == NULL)
 359                 return NULL;
 360
 361         rwlock_init(&ndev->lock);
 362         ndev->dev = dev;
 363         INIT_LIST_HEAD(&ndev->addr_list);
 364
 365         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 366         ndev->cnf.mtu6 = dev->mtu;
 367         ndev->cnf.sysctl = NULL;
 368         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 369         if (ndev->nd_parms == NULL) {
 370                 kfree(ndev);
 371                 return NULL;
 372         }
 373         if (ndev->cnf.forwarding)
 374                 dev_disable_lro(dev);
 375         /* We refer to the device */
 376         dev_hold(dev);
 377
 378         if (snmp6_alloc_dev(ndev) < 0) {
 379                 ADBG((KERN_WARNING
 380                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 381                         __func__, dev->name));
 382                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 383                 ndev->dead = 1;
 384                 in6_dev_finish_destroy(ndev);
 385                 return NULL;
 386         }
 387
 388         if (snmp6_register_dev(ndev) < 0) {
 389                 ADBG((KERN_WARNING
 390                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 391                         __func__, dev->name));
 392                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 393                 ndev->dead = 1;
 394                 in6_dev_finish_destroy(ndev);
 395                 return NULL;
 396         }
 397
 398         /* One reference from device.  We must do this before
 399          * we invoke __ipv6_regen_rndid().
 400          */
 401         in6_dev_hold(ndev);
 402
 403         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 404                 ndev->cnf.accept_dad = -1;
 405
 406 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 407         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Multicast RS\n",
 410                        dev->name);
 411                 ndev->cnf.rtr_solicits = 0;
 412         }
 413 #endif
 414
 415 #ifdef CONFIG_IPV6_PRIVACY
 416         INIT_LIST_HEAD(&ndev->tempaddr_list);
 417         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 418         if ((dev->flags&IFF_LOOPBACK) ||
 419             dev->type == ARPHRD_TUNNEL ||
 420             dev->type == ARPHRD_TUNNEL6 ||
 421             dev->type == ARPHRD_SIT ||
 422             dev->type == ARPHRD_NONE) {
 423                 ndev->cnf.use_tempaddr = -1;
 424         } else {
 425                 in6_dev_hold(ndev);
 426                 ipv6_regen_rndid((unsigned long) ndev);
 427         }
 428 #endif
 429
 430         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 431                 ndev->if_flags |= IF_READY;
 432
 433         ipv6_mc_init_dev(ndev);
 434         ndev->tstamp = jiffies;
 435         addrconf_sysctl_register(ndev);
 436         /* protected by rtnl_lock */
 437         rcu_assign_pointer(dev->ip6_ptr, ndev);
 438
 439         /* Join all-node multicast group */
 440         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 441
 442         return ndev;
 443 }
 444
 445 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 446 {
 447         struct inet6_dev *idev;
 448
 449         ASSERT_RTNL();
 450
 451         idev = __in6_dev_get(dev);
 452         if (!idev) {
 453                 idev = ipv6_add_dev(dev);
 454                 if (!idev)
 455                         return NULL;
 456         }
 457
 458         if (dev->flags&IFF_UP)
 459                 ipv6_mc_up(idev);
 460         return idev;
 461 }
 462
 463 #ifdef CONFIG_SYSCTL
 464 static void dev_forward_change(struct inet6_dev *idev)
 465 {
 466         struct net_device *dev;
 467         struct inet6_ifaddr *ifa;
 468
 469         if (!idev)
 470                 return;
 471         dev = idev->dev;
 472         if (idev->cnf.forwarding)
 473                 dev_disable_lro(dev);
 474         if (dev && (dev->flags & IFF_MULTICAST)) {
 475                 if (idev->cnf.forwarding)
 476                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 477                 else
 478                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 479         }
 480
 481         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 482                 if (ifa->flags&IFA_F_TENTATIVE)
 483                         continue;
 484                 if (idev->cnf.forwarding)
 485                         addrconf_join_anycast(ifa);
 486                 else
 487                         addrconf_leave_anycast(ifa);
 488         }
 489 }
 490
 491
 492 static void addrconf_forward_change(struct net *net, __s32 newf)
 493 {
 494         struct net_device *dev;
 495         struct inet6_dev *idev;
 496
 497         rcu_read_lock();
 498         for_each_netdev_rcu(net, dev) {
 499                 idev = __in6_dev_get(dev);
 500                 if (idev) {
 501                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 502                         idev->cnf.forwarding = newf;
 503                         if (changed)
 504                                 dev_forward_change(idev);
 505                 }
 506         }
 507         rcu_read_unlock();
 508 }
 509
 510 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 511 {
 512         struct net *net;
 513
 514         net = (struct net *)table->extra2;
 515         if (p == &net->ipv6.devconf_dflt->forwarding)
 516                 return 0;
 517
 518         if (!rtnl_trylock()) {
 519                 /* Restore the original values before restarting */
 520                 *p = old;
 521                 return restart_syscall();
 522         }
 523
 524         if (p == &net->ipv6.devconf_all->forwarding) {
 525                 __s32 newf = net->ipv6.devconf_all->forwarding;
 526                 net->ipv6.devconf_dflt->forwarding = newf;
 527                 addrconf_forward_change(net, newf);
 528         } else if ((!*p) ^ (!old))
 529                 dev_forward_change((struct inet6_dev *)table->extra1);
 530         rtnl_unlock();
 531
 532         if (*p)
 533                 rt6_purge_dflt_routers(net);
 534         return 1;
 535 }
 536 #endif
 537
 538 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 539 {
 540         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 541         kfree(ifp);
 542 }
 543
 544 /* Nobody refers to this ifaddr, destroy it */
 545 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 546 {
 547         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 548
 549 #ifdef NET_REFCNT_DEBUG
 550         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 551 #endif
 552
 553         in6_dev_put(ifp->idev);
 554
 555         if (del_timer(&ifp->timer))
 556                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 557
 558         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 559                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 560                 return;
 561         }
 562         dst_release(&ifp->rt->dst);
 563
 564         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 565 }
 566
 567 static void
 568 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 569 {
 570         struct list_head *p;
 571         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 572
 573         /*
 574          * Each device address list is sorted in order of scope -
 575          * global before linklocal.
 576          */
 577         list_for_each(p, &idev->addr_list) {
 578                 struct inet6_ifaddr *ifa
 579                         = list_entry(p, struct inet6_ifaddr, if_list);
 580                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 581                         break;
 582         }
 583
 584         list_add_tail(&ifp->if_list, p);
 585 }
 586
 587 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 588 {
 589         /*
 590          * We perform the hash function over the last 64 bits of the address
 591          * This will include the IEEE address token on links that support it.
 592          */
 593         return jhash_2words((__force u32)addr->s6_addr32[2],
 594                             (__force u32)addr->s6_addr32[3], 0)
 595                 & (IN6_ADDR_HSIZE - 1);
 596 }
 597
 598 /* On success it returns ifp with increased reference count */
 599
 600 static struct inet6_ifaddr *
 601 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 602               int scope, u32 flags)
 603 {
 604         struct inet6_ifaddr *ifa = NULL;
 605         struct rt6_info *rt;
 606         unsigned int hash;
 607         int err = 0;
 608         int addr_type = ipv6_addr_type(addr);
 609
 610         if (addr_type == IPV6_ADDR_ANY ||
 611             addr_type & IPV6_ADDR_MULTICAST ||
 612             (!(idev->dev->flags & IFF_LOOPBACK) &&
 613              addr_type & IPV6_ADDR_LOOPBACK))
 614                 return ERR_PTR(-EADDRNOTAVAIL);
 615
 616         rcu_read_lock_bh();
 617         if (idev->dead) {
 618                 err = -ENODEV;                  /*XXX*/
 619                 goto out2;
 620         }
 621
 622         if (idev->cnf.disable_ipv6) {
 623                 err = -EACCES;
 624                 goto out2;
 625         }
 626
 627         spin_lock(&addrconf_hash_lock);
 628
 629         /* Ignore adding duplicate addresses on an interface */
 630         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 631                 ADBG(("ipv6_add_addr: already assigned\n"));
 632                 err = -EEXIST;
 633                 goto out;
 634         }
 635
 636         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 637
 638         if (ifa == NULL) {
 639                 ADBG(("ipv6_add_addr: malloc failed\n"));
 640                 err = -ENOBUFS;
 641                 goto out;
 642         }
 643
 644         rt = addrconf_dst_alloc(idev, addr, 0);
 645         if (IS_ERR(rt)) {
 646                 err = PTR_ERR(rt);
 647                 goto out;
 648         }
 649
 650         ipv6_addr_copy(&ifa->addr, addr);
 651
 652         spin_lock_init(&ifa->lock);
 653         spin_lock_init(&ifa->state_lock);
 654         init_timer(&ifa->timer);
 655         INIT_HLIST_NODE(&ifa->addr_lst);
 656         ifa->timer.data = (unsigned long) ifa;
 657         ifa->scope = scope;
 658         ifa->prefix_len = pfxlen;
 659         ifa->flags = flags | IFA_F_TENTATIVE;
 660         ifa->cstamp = ifa->tstamp = jiffies;
 661
 662         ifa->rt = rt;
 663
 664         /*
 665          * part one of RFC 4429, section 3.3
 666          * We should not configure an address as
 667          * optimistic if we do not yet know the link
 668          * layer address of our nexhop router
 669          */
 670
 671         if (rt->rt6i_nexthop == NULL)
 672                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 673
 674         ifa->idev = idev;
 675         in6_dev_hold(idev);
 676         /* For caller */
 677         in6_ifa_hold(ifa);
 678
 679         /* Add to big hash table */
 680         hash = ipv6_addr_hash(addr);
 681
 682         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 683         spin_unlock(&addrconf_hash_lock);
 684
 685         write_lock(&idev->lock);
 686         /* Add to inet6_dev unicast addr list. */
 687         ipv6_link_dev_addr(idev, ifa);
 688
 689 #ifdef CONFIG_IPV6_PRIVACY
 690         if (ifa->flags&IFA_F_TEMPORARY) {
 691                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 692                 in6_ifa_hold(ifa);
 693         }
 694 #endif
 695
 696         in6_ifa_hold(ifa);
 697         write_unlock(&idev->lock);
 698 out2:
 699         rcu_read_unlock_bh();
 700
 701         if (likely(err == 0))
 702                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 703         else {
 704                 kfree(ifa);
 705                 ifa = ERR_PTR(err);
 706         }
 707
 708         return ifa;
 709 out:
 710         spin_unlock(&addrconf_hash_lock);
 711         goto out2;
 712 }
 713
 714 /* This function wants to get referenced ifp and releases it before return */
 715
 716 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 717 {
 718         struct inet6_ifaddr *ifa, *ifn;
 719         struct inet6_dev *idev = ifp->idev;
 720         int state;
 721         int deleted = 0, onlink = 0;
 722         unsigned long expires = jiffies;
 723
 724         spin_lock_bh(&ifp->state_lock);
 725         state = ifp->state;
 726         ifp->state = INET6_IFADDR_STATE_DEAD;
 727         spin_unlock_bh(&ifp->state_lock);
 728
 729         if (state == INET6_IFADDR_STATE_DEAD)
 730                 goto out;
 731
 732         spin_lock_bh(&addrconf_hash_lock);
 733         hlist_del_init_rcu(&ifp->addr_lst);
 734         spin_unlock_bh(&addrconf_hash_lock);
 735
 736         write_lock_bh(&idev->lock);
 737 #ifdef CONFIG_IPV6_PRIVACY
 738         if (ifp->flags&IFA_F_TEMPORARY) {
 739                 list_del(&ifp->tmp_list);
 740                 if (ifp->ifpub) {
 741                         in6_ifa_put(ifp->ifpub);
 742                         ifp->ifpub = NULL;
 743                 }
 744                 __in6_ifa_put(ifp);
 745         }
 746 #endif
 747
 748         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 749                 if (ifa == ifp) {
 750                         list_del_init(&ifp->if_list);
 751                         __in6_ifa_put(ifp);
 752
 753                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 754                                 break;
 755                         deleted = 1;
 756                         continue;
 757                 } else if (ifp->flags & IFA_F_PERMANENT) {
 758                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 759                                               ifp->prefix_len)) {
 760                                 if (ifa->flags & IFA_F_PERMANENT) {
 761                                         onlink = 1;
 762                                         if (deleted)
 763                                                 break;
 764                                 } else {
 765                                         unsigned long lifetime;
 766
 767                                         if (!onlink)
 768                                                 onlink = -1;
 769
 770                                         spin_lock(&ifa->lock);
 771
 772                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 773                                         /*
 774                                          * Note: Because this address is
 775                                          * not permanent, lifetime <
 776                                          * LONG_MAX / HZ here.
 777                                          */
 778                                         if (time_before(expires,
 779                                                         ifa->tstamp + lifetime * HZ))
 780                                                 expires = ifa->tstamp + lifetime * HZ;
 781                                         spin_unlock(&ifa->lock);
 782                                 }
 783                         }
 784                 }
 785         }
 786         write_unlock_bh(&idev->lock);
 787
 788         addrconf_del_timer(ifp);
 789
 790         ipv6_ifa_notify(RTM_DELADDR, ifp);
 791
 792         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 793
 794         /*
 795          * Purge or update corresponding prefix
 796          *
 797          * 1) we don't purge prefix here if address was not permanent.
 798          *    prefix is managed by its own lifetime.
 799          * 2) if there're no addresses, delete prefix.
 800          * 3) if there're still other permanent address(es),
 801          *    corresponding prefix is still permanent.
 802          * 4) otherwise, update prefix lifetime to the
 803          *    longest valid lifetime among the corresponding
 804          *    addresses on the device.
 805          *    Note: subsequent RA will update lifetime.
 806          *
 807          * --yoshfuji
 808          */
 809         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 810                 struct in6_addr prefix;
 811                 struct rt6_info *rt;
 812                 struct net *net = dev_net(ifp->idev->dev);
 813                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 814                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 815
 816                 if (rt && addrconf_is_prefix_route(rt)) {
 817                         if (onlink == 0) {
 818                                 ip6_del_rt(rt);
 819                                 rt = NULL;
 820                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 821                                 rt->rt6i_expires = expires;
 822                                 rt->rt6i_flags |= RTF_EXPIRES;
 823                         }
 824                 }
 825                 dst_release(&rt->dst);
 826         }
 827
 828 out:
 829         in6_ifa_put(ifp);
 830 }
 831
 832 #ifdef CONFIG_IPV6_PRIVACY
 833 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 834 {
 835         struct inet6_dev *idev = ifp->idev;
 836         struct in6_addr addr, *tmpaddr;
 837         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 838         unsigned long regen_advance;
 839         int tmp_plen;
 840         int ret = 0;
 841         int max_addresses;
 842         u32 addr_flags;
 843
 844         write_lock(&idev->lock);
 845         if (ift) {
 846                 spin_lock_bh(&ift->lock);
 847                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 848                 spin_unlock_bh(&ift->lock);
 849                 tmpaddr = &addr;
 850         } else {
 851                 tmpaddr = NULL;
 852         }
 853 retry:
 854         in6_dev_hold(idev);
 855         if (idev->cnf.use_tempaddr <= 0) {
 856                 write_unlock(&idev->lock);
 857                 printk(KERN_INFO
 858                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 859                 in6_dev_put(idev);
 860                 ret = -1;
 861                 goto out;
 862         }
 863         spin_lock_bh(&ifp->lock);
 864         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 865                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 866                 spin_unlock_bh(&ifp->lock);
 867                 write_unlock(&idev->lock);
 868                 printk(KERN_WARNING
 869                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 870                 in6_dev_put(idev);
 871                 ret = -1;
 872                 goto out;
 873         }
 874         in6_ifa_hold(ifp);
 875         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 876         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 877                 spin_unlock_bh(&ifp->lock);
 878                 write_unlock(&idev->lock);
 879                 printk(KERN_WARNING
 880                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 881                 in6_ifa_put(ifp);
 882                 in6_dev_put(idev);
 883                 ret = -1;
 884                 goto out;
 885         }
 886         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 887         age = (jiffies - ifp->tstamp) / HZ;
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft + age);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft + age -
 894                                  idev->cnf.max_desync_factor);
 895         tmp_plen = ifp->prefix_len;
 896         max_addresses = idev->cnf.max_addresses;
 897         tmp_cstamp = ifp->cstamp;
 898         tmp_tstamp = ifp->tstamp;
 899         spin_unlock_bh(&ifp->lock);
 900
 901         regen_advance = idev->cnf.regen_max_retry *
 902                         idev->cnf.dad_transmits *
 903                         idev->nd_parms->retrans_time / HZ;
 904         write_unlock(&idev->lock);
 905
 906         /* A temporary address is created only if this calculated Preferred
 907          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 908          * an implementation must not create a temporary address with a zero
 909          * Preferred Lifetime.
 910          */
 911         if (tmp_prefered_lft <= regen_advance) {
 912                 in6_ifa_put(ifp);
 913                 in6_dev_put(idev);
 914                 ret = -1;
 915                 goto out;
 916         }
 917
 918         addr_flags = IFA_F_TEMPORARY;
 919         /* set in addrconf_prefix_rcv() */
 920         if (ifp->flags & IFA_F_OPTIMISTIC)
 921                 addr_flags |= IFA_F_OPTIMISTIC;
 922
 923         ift = !max_addresses ||
 924               ipv6_count_addresses(idev) < max_addresses ?
 925                 ipv6_add_addr(idev, &addr, tmp_plen,
 926                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 927                               addr_flags) : NULL;
 928         if (!ift || IS_ERR(ift)) {
 929                 in6_ifa_put(ifp);
 930                 in6_dev_put(idev);
 931                 printk(KERN_INFO
 932                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 933                 tmpaddr = &addr;
 934                 write_lock(&idev->lock);
 935                 goto retry;
 936         }
 937
 938         spin_lock_bh(&ift->lock);
 939         ift->ifpub = ifp;
 940         ift->valid_lft = tmp_valid_lft;
 941         ift->prefered_lft = tmp_prefered_lft;
 942         ift->cstamp = tmp_cstamp;
 943         ift->tstamp = tmp_tstamp;
 944         spin_unlock_bh(&ift->lock);
 945
 946         addrconf_dad_start(ift, 0);
 947         in6_ifa_put(ift);
 948         in6_dev_put(idev);
 949 out:
 950         return ret;
 951 }
 952 #endif
 953
 954 /*
 955  *      Choose an appropriate source address (RFC3484)
 956  */
 957 enum {
 958         IPV6_SADDR_RULE_INIT = 0,
 959         IPV6_SADDR_RULE_LOCAL,
 960         IPV6_SADDR_RULE_SCOPE,
 961         IPV6_SADDR_RULE_PREFERRED,
 962 #ifdef CONFIG_IPV6_MIP6
 963         IPV6_SADDR_RULE_HOA,
 964 #endif
 965         IPV6_SADDR_RULE_OIF,
 966         IPV6_SADDR_RULE_LABEL,
 967 #ifdef CONFIG_IPV6_PRIVACY
 968         IPV6_SADDR_RULE_PRIVACY,
 969 #endif
 970         IPV6_SADDR_RULE_ORCHID,
 971         IPV6_SADDR_RULE_PREFIX,
 972         IPV6_SADDR_RULE_MAX
 973 };
 974
 975 struct ipv6_saddr_score {
 976         int                     rule;
 977         int                     addr_type;
 978         struct inet6_ifaddr     *ifa;
 979         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 980         int                     scopedist;
 981         int                     matchlen;
 982 };
 983
 984 struct ipv6_saddr_dst {
 985         const struct in6_addr *addr;
 986         int ifindex;
 987         int scope;
 988         int label;
 989         unsigned int prefs;
 990 };
 991
 992 static inline int ipv6_saddr_preferred(int type)
 993 {
 994         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 995                 return 1;
 996         return 0;
 997 }
 998
 999 static int ipv6_get_saddr_eval(struct net *net,
1000                                struct ipv6_saddr_score *score,
1001                                struct ipv6_saddr_dst *dst,
1002                                int i)
1003 {
1004         int ret;
1005
1006         if (i <= score->rule) {
1007                 switch (i) {
1008                 case IPV6_SADDR_RULE_SCOPE:
1009                         ret = score->scopedist;
1010                         break;
1011                 case IPV6_SADDR_RULE_PREFIX:
1012                         ret = score->matchlen;
1013                         break;
1014                 default:
1015                         ret = !!test_bit(i, score->scorebits);
1016                 }
1017                 goto out;
1018         }
1019
1020         switch (i) {
1021         case IPV6_SADDR_RULE_INIT:
1022                 /* Rule 0: remember if hiscore is not ready yet */
1023                 ret = !!score->ifa;
1024                 break;
1025         case IPV6_SADDR_RULE_LOCAL:
1026                 /* Rule 1: Prefer same address */
1027                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1028                 break;
1029         case IPV6_SADDR_RULE_SCOPE:
1030                 /* Rule 2: Prefer appropriate scope
1031                  *
1032                  *      ret
1033                  *       ^
1034                  *    -1 |  d 15
1035                  *    ---+--+-+---> scope
1036                  *       |
1037                  *       |             d is scope of the destination.
1038                  *  B-d  |  \
1039                  *       |   \      <- smaller scope is better if
1040                  *  B-15 |    \        if scope is enough for destinaion.
1041                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1042                  * d-C-1 | /
1043                  *       |/         <- greater is better
1044                  *   -C  /             if scope is not enough for destination.
1045                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1046                  *
1047                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1048                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1049                  * Assume B = 0 and we get C > 29.
1050                  */
1051                 ret = __ipv6_addr_src_scope(score->addr_type);
1052                 if (ret >= dst->scope)
1053                         ret = -ret;
1054                 else
1055                         ret -= 128;     /* 30 is enough */
1056                 score->scopedist = ret;
1057                 break;
1058         case IPV6_SADDR_RULE_PREFERRED:
1059                 /* Rule 3: Avoid deprecated and optimistic addresses */
1060                 ret = ipv6_saddr_preferred(score->addr_type) ||
1061                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1062                 break;
1063 #ifdef CONFIG_IPV6_MIP6
1064         case IPV6_SADDR_RULE_HOA:
1065             {
1066                 /* Rule 4: Prefer home address */
1067                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1068                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1069                 break;
1070             }
1071 #endif
1072         case IPV6_SADDR_RULE_OIF:
1073                 /* Rule 5: Prefer outgoing interface */
1074                 ret = (!dst->ifindex ||
1075                        dst->ifindex == score->ifa->idev->dev->ifindex);
1076                 break;
1077         case IPV6_SADDR_RULE_LABEL:
1078                 /* Rule 6: Prefer matching label */
1079                 ret = ipv6_addr_label(net,
1080                                       &score->ifa->addr, score->addr_type,
1081                                       score->ifa->idev->dev->ifindex) == dst->label;
1082                 break;
1083 #ifdef CONFIG_IPV6_PRIVACY
1084         case IPV6_SADDR_RULE_PRIVACY:
1085             {
1086                 /* Rule 7: Prefer public address
1087                  * Note: prefer temporary address if use_tempaddr >= 2
1088                  */
1089                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1090                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1091                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1092                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1093                 break;
1094             }
1095 #endif
1096         case IPV6_SADDR_RULE_ORCHID:
1097                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1098                  *          non-ORCHID vs non-ORCHID
1099                  */
1100                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1101                         ipv6_addr_orchid(dst->addr));
1102                 break;
1103         case IPV6_SADDR_RULE_PREFIX:
1104                 /* Rule 8: Use longest matching prefix */
1105                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1106                                                        dst->addr);
1107                 break;
1108         default:
1109                 ret = 0;
1110         }
1111
1112         if (ret)
1113                 __set_bit(i, score->scorebits);
1114         score->rule = i;
1115 out:
1116         return ret;
1117 }
1118
1119 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1120                        const struct in6_addr *daddr, unsigned int prefs,
1121                        struct in6_addr *saddr)
1122 {
1123         struct ipv6_saddr_score scores[2],
1124                                 *score = &scores[0], *hiscore = &scores[1];
1125         struct ipv6_saddr_dst dst;
1126         struct net_device *dev;
1127         int dst_type;
1128
1129         dst_type = __ipv6_addr_type(daddr);
1130         dst.addr = daddr;
1131         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1132         dst.scope = __ipv6_addr_src_scope(dst_type);
1133         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1134         dst.prefs = prefs;
1135
1136         hiscore->rule = -1;
1137         hiscore->ifa = NULL;
1138
1139         rcu_read_lock();
1140
1141         for_each_netdev_rcu(net, dev) {
1142                 struct inet6_dev *idev;
1143
1144                 /* Candidate Source Address (section 4)
1145                  *  - multicast and link-local destination address,
1146                  *    the set of candidate source address MUST only
1147                  *    include addresses assigned to interfaces
1148                  *    belonging to the same link as the outgoing
1149                  *    interface.
1150                  * (- For site-local destination addresses, the
1151                  *    set of candidate source addresses MUST only
1152                  *    include addresses assigned to interfaces
1153                  *    belonging to the same site as the outgoing
1154                  *    interface.)
1155                  */
1156                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1157                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1158                     dst.ifindex && dev->ifindex != dst.ifindex)
1159                         continue;
1160
1161                 idev = __in6_dev_get(dev);
1162                 if (!idev)
1163                         continue;
1164
1165                 read_lock_bh(&idev->lock);
1166                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1167                         int i;
1168
1169                         /*
1170                          * - Tentative Address (RFC2462 section 5.4)
1171                          *  - A tentative address is not considered
1172                          *    "assigned to an interface" in the traditional
1173                          *    sense, unless it is also flagged as optimistic.
1174                          * - Candidate Source Address (section 4)
1175                          *  - In any case, anycast addresses, multicast
1176                          *    addresses, and the unspecified address MUST
1177                          *    NOT be included in a candidate set.
1178                          */
1179                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1180                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1181                                 continue;
1182
1183                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1184
1185                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1186                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1187                                 LIMIT_NETDEBUG(KERN_DEBUG
1188                                                "ADDRCONF: unspecified / multicast address "
1189                                                "assigned as unicast address on %s",
1190                                                dev->name);
1191                                 continue;
1192                         }
1193
1194                         score->rule = -1;
1195                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1196
1197                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1198                                 int minihiscore, miniscore;
1199
1200                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1201                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1202
1203                                 if (minihiscore > miniscore) {
1204                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1205                                             score->scopedist > 0) {
1206                                                 /*
1207                                                  * special case:
1208                                                  * each remaining entry
1209                                                  * has too small (not enough)
1210                                                  * scope, because ifa entries
1211                                                  * are sorted by their scope
1212                                                  * values.
1213                                                  */
1214                                                 goto try_nextdev;
1215                                         }
1216                                         break;
1217                                 } else if (minihiscore < miniscore) {
1218                                         if (hiscore->ifa)
1219                                                 in6_ifa_put(hiscore->ifa);
1220
1221                                         in6_ifa_hold(score->ifa);
1222
1223                                         swap(hiscore, score);
1224
1225                                         /* restore our iterator */
1226                                         score->ifa = hiscore->ifa;
1227
1228                                         break;
1229                                 }
1230                         }
1231                 }
1232 try_nextdev:
1233                 read_unlock_bh(&idev->lock);
1234         }
1235         rcu_read_unlock();
1236
1237         if (!hiscore->ifa)
1238                 return -EADDRNOTAVAIL;
1239
1240         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1241         in6_ifa_put(hiscore->ifa);
1242         return 0;
1243 }
1244 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1245
1246 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1247                     unsigned char banned_flags)
1248 {
1249         struct inet6_dev *idev;
1250         int err = -EADDRNOTAVAIL;
1251
1252         rcu_read_lock();
1253         idev = __in6_dev_get(dev);
1254         if (idev) {
1255                 struct inet6_ifaddr *ifp;
1256
1257                 read_lock_bh(&idev->lock);
1258                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1259                         if (ifp->scope == IFA_LINK &&
1260                             !(ifp->flags & banned_flags)) {
1261                                 ipv6_addr_copy(addr, &ifp->addr);
1262                                 err = 0;
1263                                 break;
1264                         }
1265                 }
1266                 read_unlock_bh(&idev->lock);
1267         }
1268         rcu_read_unlock();
1269         return err;
1270 }
1271
1272 static int ipv6_count_addresses(struct inet6_dev *idev)
1273 {
1274         int cnt = 0;
1275         struct inet6_ifaddr *ifp;
1276
1277         read_lock_bh(&idev->lock);
1278         list_for_each_entry(ifp, &idev->addr_list, if_list)
1279                 cnt++;
1280         read_unlock_bh(&idev->lock);
1281         return cnt;
1282 }
1283
1284 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1285                   struct net_device *dev, int strict)
1286 {
1287         struct inet6_ifaddr *ifp;
1288         struct hlist_node *node;
1289         unsigned int hash = ipv6_addr_hash(addr);
1290
1291         rcu_read_lock_bh();
1292         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE) &&
1297                     (dev == NULL || ifp->idev->dev == dev ||
1298                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1299                         rcu_read_unlock_bh();
1300                         return 1;
1301                 }
1302         }
1303
1304         rcu_read_unlock_bh();
1305         return 0;
1306 }
1307 EXPORT_SYMBOL(ipv6_chk_addr);
1308
1309 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1310                                struct net_device *dev)
1311 {
1312         unsigned int hash = ipv6_addr_hash(addr);
1313         struct inet6_ifaddr *ifp;
1314         struct hlist_node *node;
1315
1316         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1317                 if (!net_eq(dev_net(ifp->idev->dev), net))
1318                         continue;
1319                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1320                         if (dev == NULL || ifp->idev->dev == dev)
1321                                 return true;
1322                 }
1323         }
1324         return false;
1325 }
1326
1327 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1328 {
1329         struct inet6_dev *idev;
1330         struct inet6_ifaddr *ifa;
1331         int     onlink;
1332
1333         onlink = 0;
1334         rcu_read_lock();
1335         idev = __in6_dev_get(dev);
1336         if (idev) {
1337                 read_lock_bh(&idev->lock);
1338                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1339                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1340                                                    ifa->prefix_len);
1341                         if (onlink)
1342                                 break;
1343                 }
1344                 read_unlock_bh(&idev->lock);
1345         }
1346         rcu_read_unlock();
1347         return onlink;
1348 }
1349
1350 EXPORT_SYMBOL(ipv6_chk_prefix);
1351
1352 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1353                                      struct net_device *dev, int strict)
1354 {
1355         struct inet6_ifaddr *ifp, *result = NULL;
1356         unsigned int hash = ipv6_addr_hash(addr);
1357         struct hlist_node *node;
1358
1359         rcu_read_lock_bh();
1360         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1361                 if (!net_eq(dev_net(ifp->idev->dev), net))
1362                         continue;
1363                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1364                         if (dev == NULL || ifp->idev->dev == dev ||
1365                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1366                                 result = ifp;
1367                                 in6_ifa_hold(ifp);
1368                                 break;
1369                         }
1370                 }
1371         }
1372         rcu_read_unlock_bh();
1373
1374         return result;
1375 }
1376
1377 /* Gets referenced address, destroys ifaddr */
1378
1379 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1380 {
1381         if (ifp->flags&IFA_F_PERMANENT) {
1382                 spin_lock_bh(&ifp->lock);
1383                 addrconf_del_timer(ifp);
1384                 ifp->flags |= IFA_F_TENTATIVE;
1385                 if (dad_failed)
1386                         ifp->flags |= IFA_F_DADFAILED;
1387                 spin_unlock_bh(&ifp->lock);
1388                 if (dad_failed)
1389                         ipv6_ifa_notify(0, ifp);
1390                 in6_ifa_put(ifp);
1391 #ifdef CONFIG_IPV6_PRIVACY
1392         } else if (ifp->flags&IFA_F_TEMPORARY) {
1393                 struct inet6_ifaddr *ifpub;
1394                 spin_lock_bh(&ifp->lock);
1395                 ifpub = ifp->ifpub;
1396                 if (ifpub) {
1397                         in6_ifa_hold(ifpub);
1398                         spin_unlock_bh(&ifp->lock);
1399                         ipv6_create_tempaddr(ifpub, ifp);
1400                         in6_ifa_put(ifpub);
1401                 } else {
1402                         spin_unlock_bh(&ifp->lock);
1403                 }
1404                 ipv6_del_addr(ifp);
1405 #endif
1406         } else
1407                 ipv6_del_addr(ifp);
1408 }
1409
1410 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1411 {
1412         int err = -ENOENT;
1413
1414         spin_lock(&ifp->state_lock);
1415         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1416                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1417                 err = 0;
1418         }
1419         spin_unlock(&ifp->state_lock);
1420
1421         return err;
1422 }
1423
1424 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1425 {
1426         struct inet6_dev *idev = ifp->idev;
1427
1428         if (addrconf_dad_end(ifp)) {
1429                 in6_ifa_put(ifp);
1430                 return;
1431         }
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                                          *  minimum.
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         int state, i;
2662
2663         ASSERT_RTNL();
2664
2665         rt6_ifdown(net, dev);
2666         neigh_ifdown(&nd_tbl, dev);
2667
2668         idev = __in6_dev_get(dev);
2669         if (idev == NULL)
2670                 return -ENODEV;
2671
2672         /*
2673          * Step 1: remove reference to ipv6 device from parent device.
2674          *         Do not dev_put!
2675          */
2676         if (how) {
2677                 idev->dead = 1;
2678
2679                 /* protected by rtnl_lock */
2680                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2681
2682                 /* Step 1.5: remove snmp6 entry */
2683                 snmp6_unregister_dev(idev);
2684
2685         }
2686
2687         /* Step 2: clear hash table */
2688         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2689                 struct hlist_head *h = &inet6_addr_lst[i];
2690                 struct hlist_node *n;
2691
2692                 spin_lock_bh(&addrconf_hash_lock);
2693         restart:
2694                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2695                         if (ifa->idev == idev) {
2696                                 hlist_del_init_rcu(&ifa->addr_lst);
2697                                 addrconf_del_timer(ifa);
2698                                 goto restart;
2699                         }
2700                 }
2701                 spin_unlock_bh(&addrconf_hash_lock);
2702         }
2703
2704         write_lock_bh(&idev->lock);
2705
2706         /* Step 2: clear flags for stateless addrconf */
2707         if (!how)
2708                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2709
2710 #ifdef CONFIG_IPV6_PRIVACY
2711         if (how && del_timer(&idev->regen_timer))
2712                 in6_dev_put(idev);
2713
2714         /* Step 3: clear tempaddr list */
2715         while (!list_empty(&idev->tempaddr_list)) {
2716                 ifa = list_first_entry(&idev->tempaddr_list,
2717                                        struct inet6_ifaddr, tmp_list);
2718                 list_del(&ifa->tmp_list);
2719                 write_unlock_bh(&idev->lock);
2720                 spin_lock_bh(&ifa->lock);
2721
2722                 if (ifa->ifpub) {
2723                         in6_ifa_put(ifa->ifpub);
2724                         ifa->ifpub = NULL;
2725                 }
2726                 spin_unlock_bh(&ifa->lock);
2727                 in6_ifa_put(ifa);
2728                 write_lock_bh(&idev->lock);
2729         }
2730 #endif
2731
2732         while (!list_empty(&idev->addr_list)) {
2733                 ifa = list_first_entry(&idev->addr_list,
2734                                        struct inet6_ifaddr, if_list);
2735                 addrconf_del_timer(ifa);
2736
2737                 list_del(&ifa->if_list);
2738
2739                 write_unlock_bh(&idev->lock);
2740
2741                 spin_lock_bh(&ifa->state_lock);
2742                 state = ifa->state;
2743                 ifa->state = INET6_IFADDR_STATE_DEAD;
2744                 spin_unlock_bh(&ifa->state_lock);
2745
2746                 if (state != INET6_IFADDR_STATE_DEAD) {
2747                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2748                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2749                 }
2750                 in6_ifa_put(ifa);
2751
2752                 write_lock_bh(&idev->lock);
2753         }
2754
2755         write_unlock_bh(&idev->lock);
2756
2757         /* Step 5: Discard multicast list */
2758         if (how)
2759                 ipv6_mc_destroy_dev(idev);
2760         else
2761                 ipv6_mc_down(idev);
2762
2763         idev->tstamp = jiffies;
2764
2765         /* Last: Shot the device (if unregistered) */
2766         if (how) {
2767                 addrconf_sysctl_unregister(idev);
2768                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2769                 neigh_ifdown(&nd_tbl, dev);
2770                 in6_dev_put(idev);
2771         }
2772         return 0;
2773 }
2774
2775 static void addrconf_rs_timer(unsigned long data)
2776 {
2777         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2778         struct inet6_dev *idev = ifp->idev;
2779
2780         read_lock(&idev->lock);
2781         if (idev->dead || !(idev->if_flags & IF_READY))
2782                 goto out;
2783
2784         if (idev->cnf.forwarding)
2785                 goto out;
2786
2787         /* Announcement received after solicitation was sent */
2788         if (idev->if_flags & IF_RA_RCVD)
2789                 goto out;
2790
2791         spin_lock(&ifp->lock);
2792         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2793                 /* The wait after the last probe can be shorter */
2794                 addrconf_mod_timer(ifp, AC_RS,
2795                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2796                                    idev->cnf.rtr_solicit_delay :
2797                                    idev->cnf.rtr_solicit_interval);
2798                 spin_unlock(&ifp->lock);
2799
2800                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2801         } else {
2802                 spin_unlock(&ifp->lock);
2803                 /*
2804                  * Note: we do not support deprecated "all on-link"
2805                  * assumption any longer.
2806                  */
2807                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2808                        idev->dev->name);
2809         }
2810
2811 out:
2812         read_unlock(&idev->lock);
2813         in6_ifa_put(ifp);
2814 }
2815
2816 /*
2817  *      Duplicate Address Detection
2818  */
2819 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2820 {
2821         unsigned long rand_num;
2822         struct inet6_dev *idev = ifp->idev;
2823
2824         if (ifp->flags & IFA_F_OPTIMISTIC)
2825                 rand_num = 0;
2826         else
2827                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2828
2829         ifp->probes = idev->cnf.dad_transmits;
2830         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2831 }
2832
2833 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2834 {
2835         struct inet6_dev *idev = ifp->idev;
2836         struct net_device *dev = idev->dev;
2837
2838         addrconf_join_solict(dev, &ifp->addr);
2839
2840         net_srandom(ifp->addr.s6_addr32[3]);
2841
2842         read_lock_bh(&idev->lock);
2843         spin_lock(&ifp->lock);
2844         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2845                 goto out;
2846
2847         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2848             idev->cnf.accept_dad < 1 ||
2849             !(ifp->flags&IFA_F_TENTATIVE) ||
2850             ifp->flags & IFA_F_NODAD) {
2851                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2852                 spin_unlock(&ifp->lock);
2853                 read_unlock_bh(&idev->lock);
2854
2855                 addrconf_dad_completed(ifp);
2856                 return;
2857         }
2858
2859         if (!(idev->if_flags & IF_READY)) {
2860                 spin_unlock(&ifp->lock);
2861                 read_unlock_bh(&idev->lock);
2862                 /*
2863                  * If the device is not ready:
2864                  * - keep it tentative if it is a permanent address.
2865                  * - otherwise, kill it.
2866                  */
2867                 in6_ifa_hold(ifp);
2868                 addrconf_dad_stop(ifp, 0);
2869                 return;
2870         }
2871
2872         /*
2873          * Optimistic nodes can start receiving
2874          * Frames right away
2875          */
2876         if (ifp->flags & IFA_F_OPTIMISTIC)
2877                 ip6_ins_rt(ifp->rt);
2878
2879         addrconf_dad_kick(ifp);
2880 out:
2881         spin_unlock(&ifp->lock);
2882         read_unlock_bh(&idev->lock);
2883 }
2884
2885 static void addrconf_dad_timer(unsigned long data)
2886 {
2887         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2888         struct inet6_dev *idev = ifp->idev;
2889         struct in6_addr mcaddr;
2890
2891         if (!ifp->probes && addrconf_dad_end(ifp))
2892                 goto out;
2893
2894         read_lock(&idev->lock);
2895         if (idev->dead || !(idev->if_flags & IF_READY)) {
2896                 read_unlock(&idev->lock);
2897                 goto out;
2898         }
2899
2900         spin_lock(&ifp->lock);
2901         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2902                 spin_unlock(&ifp->lock);
2903                 read_unlock(&idev->lock);
2904                 goto out;
2905         }
2906
2907         if (ifp->probes == 0) {
2908                 /*
2909                  * DAD was successful
2910                  */
2911
2912                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2913                 spin_unlock(&ifp->lock);
2914                 read_unlock(&idev->lock);
2915
2916                 addrconf_dad_completed(ifp);
2917
2918                 goto out;
2919         }
2920
2921         ifp->probes--;
2922         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2923         spin_unlock(&ifp->lock);
2924         read_unlock(&idev->lock);
2925
2926         /* send a neighbour solicitation for our addr */
2927         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2928         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2929 out:
2930         in6_ifa_put(ifp);
2931 }
2932
2933 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2934 {
2935         struct net_device *dev = ifp->idev->dev;
2936
2937         /*
2938          *      Configure the address for reception. Now it is valid.
2939          */
2940
2941         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2942
2943         /* If added prefix is link local and forwarding is off,
2944            start sending router solicitations.
2945          */
2946
2947         if ((ifp->idev->cnf.forwarding == 0 ||
2948              ifp->idev->cnf.forwarding == 2) &&
2949             ifp->idev->cnf.rtr_solicits > 0 &&
2950             (dev->flags&IFF_LOOPBACK) == 0 &&
2951             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2952                 /*
2953                  *      If a host as already performed a random delay
2954                  *      [...] as part of DAD [...] there is no need
2955                  *      to delay again before sending the first RS
2956                  */
2957                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2958
2959                 spin_lock_bh(&ifp->lock);
2960                 ifp->probes = 1;
2961                 ifp->idev->if_flags |= IF_RS_SENT;
2962                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2963                 spin_unlock_bh(&ifp->lock);
2964         }
2965 }
2966
2967 static void addrconf_dad_run(struct inet6_dev *idev)
2968 {
2969         struct inet6_ifaddr *ifp;
2970
2971         read_lock_bh(&idev->lock);
2972         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2973                 spin_lock(&ifp->lock);
2974                 if (ifp->flags & IFA_F_TENTATIVE &&
2975                     ifp->state == INET6_IFADDR_STATE_DAD)
2976                         addrconf_dad_kick(ifp);
2977                 spin_unlock(&ifp->lock);
2978         }
2979         read_unlock_bh(&idev->lock);
2980 }
2981
2982 #ifdef CONFIG_PROC_FS
2983 struct if6_iter_state {
2984         struct seq_net_private p;
2985         int bucket;
2986 };
2987
2988 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2989 {
2990         struct inet6_ifaddr *ifa = NULL;
2991         struct if6_iter_state *state = seq->private;
2992         struct net *net = seq_file_net(seq);
2993
2994         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2995                 struct hlist_node *n;
2996                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2997                                          addr_lst)
2998                         if (net_eq(dev_net(ifa->idev->dev), net))
2999                                 return ifa;
3000         }
3001         return NULL;
3002 }
3003
3004 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3005                                          struct inet6_ifaddr *ifa)
3006 {
3007         struct if6_iter_state *state = seq->private;
3008         struct net *net = seq_file_net(seq);
3009         struct hlist_node *n = &ifa->addr_lst;
3010
3011         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3012                 if (net_eq(dev_net(ifa->idev->dev), net))
3013                         return ifa;
3014
3015         while (++state->bucket < IN6_ADDR_HSIZE) {
3016                 hlist_for_each_entry_rcu_bh(ifa, n,
3017                                      &inet6_addr_lst[state->bucket], addr_lst) {
3018                         if (net_eq(dev_net(ifa->idev->dev), net))
3019                                 return ifa;
3020                 }
3021         }
3022
3023         return NULL;
3024 }
3025
3026 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3027 {
3028         struct inet6_ifaddr *ifa = if6_get_first(seq);
3029
3030         if (ifa)
3031                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3032                         --pos;
3033         return pos ? NULL : ifa;
3034 }
3035
3036 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3037         __acquires(rcu_bh)
3038 {
3039         rcu_read_lock_bh();
3040         return if6_get_idx(seq, *pos);
3041 }
3042
3043 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3044 {
3045         struct inet6_ifaddr *ifa;
3046
3047         ifa = if6_get_next(seq, v);
3048         ++*pos;
3049         return ifa;
3050 }
3051
3052 static void if6_seq_stop(struct seq_file *seq, void *v)
3053         __releases(rcu_bh)
3054 {
3055         rcu_read_unlock_bh();
3056 }
3057
3058 static int if6_seq_show(struct seq_file *seq, void *v)
3059 {
3060         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3061         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3062                    &ifp->addr,
3063                    ifp->idev->dev->ifindex,
3064                    ifp->prefix_len,
3065                    ifp->scope,
3066                    ifp->flags,
3067                    ifp->idev->dev->name);
3068         return 0;
3069 }
3070
3071 static const struct seq_operations if6_seq_ops = {
3072         .start  = if6_seq_start,
3073         .next   = if6_seq_next,
3074         .show   = if6_seq_show,
3075         .stop   = if6_seq_stop,
3076 };
3077
3078 static int if6_seq_open(struct inode *inode, struct file *file)
3079 {
3080         return seq_open_net(inode, file, &if6_seq_ops,
3081                             sizeof(struct if6_iter_state));
3082 }
3083
3084 static const struct file_operations if6_fops = {
3085         .owner          = THIS_MODULE,
3086         .open           = if6_seq_open,
3087         .read           = seq_read,
3088         .llseek         = seq_lseek,
3089         .release        = seq_release_net,
3090 };
3091
3092 static int __net_init if6_proc_net_init(struct net *net)
3093 {
3094         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3095                 return -ENOMEM;
3096         return 0;
3097 }
3098
3099 static void __net_exit if6_proc_net_exit(struct net *net)
3100 {
3101        proc_net_remove(net, "if_inet6");
3102 }
3103
3104 static struct pernet_operations if6_proc_net_ops = {
3105        .init = if6_proc_net_init,
3106        .exit = if6_proc_net_exit,
3107 };
3108
3109 int __init if6_proc_init(void)
3110 {
3111         return register_pernet_subsys(&if6_proc_net_ops);
3112 }
3113
3114 void if6_proc_exit(void)
3115 {
3116         unregister_pernet_subsys(&if6_proc_net_ops);
3117 }
3118 #endif  /* CONFIG_PROC_FS */
3119
3120 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3121 /* Check if address is a home address configured on any interface. */
3122 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3123 {
3124         int ret = 0;
3125         struct inet6_ifaddr *ifp = NULL;
3126         struct hlist_node *n;
3127         unsigned int hash = ipv6_addr_hash(addr);
3128
3129         rcu_read_lock_bh();
3130         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3131                 if (!net_eq(dev_net(ifp->idev->dev), net))
3132                         continue;
3133                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3134                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3135                         ret = 1;
3136                         break;
3137                 }
3138         }
3139         rcu_read_unlock_bh();
3140         return ret;
3141 }
3142 #endif
3143
3144 /*
3145  *      Periodic address status verification
3146  */
3147
3148 static void addrconf_verify(unsigned long foo)
3149 {
3150         unsigned long now, next, next_sec, next_sched;
3151         struct inet6_ifaddr *ifp;
3152         struct hlist_node *node;
3153         int i;
3154
3155         rcu_read_lock_bh();
3156         spin_lock(&addrconf_verify_lock);
3157         now = jiffies;
3158         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3159
3160         del_timer(&addr_chk_timer);
3161
3162         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3163 restart:
3164                 hlist_for_each_entry_rcu_bh(ifp, node,
3165                                          &inet6_addr_lst[i], addr_lst) {
3166                         unsigned long age;
3167
3168                         if (ifp->flags & IFA_F_PERMANENT)
3169                                 continue;
3170
3171                         spin_lock(&ifp->lock);
3172                         /* We try to batch several events at once. */
3173                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3174
3175                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3176                             age >= ifp->valid_lft) {
3177                                 spin_unlock(&ifp->lock);
3178                                 in6_ifa_hold(ifp);
3179                                 ipv6_del_addr(ifp);
3180                                 goto restart;
3181                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3182                                 spin_unlock(&ifp->lock);
3183                                 continue;
3184                         } else if (age >= ifp->prefered_lft) {
3185                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3186                                 int deprecate = 0;
3187
3188                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3189                                         deprecate = 1;
3190                                         ifp->flags |= IFA_F_DEPRECATED;
3191                                 }
3192
3193                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3194                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3195
3196                                 spin_unlock(&ifp->lock);
3197
3198                                 if (deprecate) {
3199                                         in6_ifa_hold(ifp);
3200
3201                                         ipv6_ifa_notify(0, ifp);
3202                                         in6_ifa_put(ifp);
3203                                         goto restart;
3204                                 }
3205 #ifdef CONFIG_IPV6_PRIVACY
3206                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3207                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3208                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3209                                         ifp->idev->cnf.dad_transmits *
3210                                         ifp->idev->nd_parms->retrans_time / HZ;
3211
3212                                 if (age >= ifp->prefered_lft - regen_advance) {
3213                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3214                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3215                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3216                                         if (!ifp->regen_count && ifpub) {
3217                                                 ifp->regen_count++;
3218                                                 in6_ifa_hold(ifp);
3219                                                 in6_ifa_hold(ifpub);
3220                                                 spin_unlock(&ifp->lock);
3221
3222                                                 spin_lock(&ifpub->lock);
3223                                                 ifpub->regen_count = 0;
3224                                                 spin_unlock(&ifpub->lock);
3225                                                 ipv6_create_tempaddr(ifpub, ifp);
3226                                                 in6_ifa_put(ifpub);
3227                                                 in6_ifa_put(ifp);
3228                                                 goto restart;
3229                                         }
3230                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3231                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3232                                 spin_unlock(&ifp->lock);
3233 #endif
3234                         } else {
3235                                 /* ifp->prefered_lft <= ifp->valid_lft */
3236                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3237                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3238                                 spin_unlock(&ifp->lock);
3239                         }
3240                 }
3241         }
3242
3243         next_sec = round_jiffies_up(next);
3244         next_sched = next;
3245
3246         /* If rounded timeout is accurate enough, accept it. */
3247         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3248                 next_sched = next_sec;
3249
3250         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3251         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3252                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3253
3254         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3255               now, next, next_sec, next_sched));
3256
3257         addr_chk_timer.expires = next_sched;
3258         add_timer(&addr_chk_timer);
3259         spin_unlock(&addrconf_verify_lock);
3260         rcu_read_unlock_bh();
3261 }
3262
3263 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3264 {
3265         struct in6_addr *pfx = NULL;
3266
3267         if (addr)
3268                 pfx = nla_data(addr);
3269
3270         if (local) {
3271                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3272                         pfx = NULL;
3273                 else
3274                         pfx = nla_data(local);
3275         }
3276
3277         return pfx;
3278 }
3279
3280 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3281         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3282         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3283         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3284 };
3285
3286 static int
3287 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3288 {
3289         struct net *net = sock_net(skb->sk);
3290         struct ifaddrmsg *ifm;
3291         struct nlattr *tb[IFA_MAX+1];
3292         struct in6_addr *pfx;
3293         int err;
3294
3295         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3296         if (err < 0)
3297                 return err;
3298
3299         ifm = nlmsg_data(nlh);
3300         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3301         if (pfx == NULL)
3302                 return -EINVAL;
3303
3304         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3305 }
3306
3307 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3308                              u32 prefered_lft, u32 valid_lft)
3309 {
3310         u32 flags;
3311         clock_t expires;
3312         unsigned long timeout;
3313
3314         if (!valid_lft || (prefered_lft > valid_lft))
3315                 return -EINVAL;
3316
3317         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3318         if (addrconf_finite_timeout(timeout)) {
3319                 expires = jiffies_to_clock_t(timeout * HZ);
3320                 valid_lft = timeout;
3321                 flags = RTF_EXPIRES;
3322         } else {
3323                 expires = 0;
3324                 flags = 0;
3325                 ifa_flags |= IFA_F_PERMANENT;
3326         }
3327
3328         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3329         if (addrconf_finite_timeout(timeout)) {
3330                 if (timeout == 0)
3331                         ifa_flags |= IFA_F_DEPRECATED;
3332                 prefered_lft = timeout;
3333         }
3334
3335         spin_lock_bh(&ifp->lock);
3336         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3337         ifp->tstamp = jiffies;
3338         ifp->valid_lft = valid_lft;
3339         ifp->prefered_lft = prefered_lft;
3340
3341         spin_unlock_bh(&ifp->lock);
3342         if (!(ifp->flags&IFA_F_TENTATIVE))
3343                 ipv6_ifa_notify(0, ifp);
3344
3345         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3346                               expires, flags);
3347         addrconf_verify(0);
3348
3349         return 0;
3350 }
3351
3352 static int
3353 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3354 {
3355         struct net *net = sock_net(skb->sk);
3356         struct ifaddrmsg *ifm;
3357         struct nlattr *tb[IFA_MAX+1];
3358         struct in6_addr *pfx;
3359         struct inet6_ifaddr *ifa;
3360         struct net_device *dev;
3361         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3362         u8 ifa_flags;
3363         int err;
3364
3365         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3366         if (err < 0)
3367                 return err;
3368
3369         ifm = nlmsg_data(nlh);
3370         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3371         if (pfx == NULL)
3372                 return -EINVAL;
3373
3374         if (tb[IFA_CACHEINFO]) {
3375                 struct ifa_cacheinfo *ci;
3376
3377                 ci = nla_data(tb[IFA_CACHEINFO]);
3378                 valid_lft = ci->ifa_valid;
3379                 preferred_lft = ci->ifa_prefered;
3380         } else {
3381                 preferred_lft = INFINITY_LIFE_TIME;
3382                 valid_lft = INFINITY_LIFE_TIME;
3383         }
3384
3385         dev =  __dev_get_by_index(net, ifm->ifa_index);
3386         if (dev == NULL)
3387                 return -ENODEV;
3388
3389         /* We ignore other flags so far. */
3390         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3391
3392         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3393         if (ifa == NULL) {
3394                 /*
3395                  * It would be best to check for !NLM_F_CREATE here but
3396                  * userspace alreay relies on not having to provide this.
3397                  */
3398                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3399                                       ifm->ifa_prefixlen, ifa_flags,
3400                                       preferred_lft, valid_lft);
3401         }
3402
3403         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3404             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3405                 err = -EEXIST;
3406         else
3407                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3408
3409         in6_ifa_put(ifa);
3410
3411         return err;
3412 }
3413
3414 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3415                           u8 scope, int ifindex)
3416 {
3417         struct ifaddrmsg *ifm;
3418
3419         ifm = nlmsg_data(nlh);
3420         ifm->ifa_family = AF_INET6;
3421         ifm->ifa_prefixlen = prefixlen;
3422         ifm->ifa_flags = flags;
3423         ifm->ifa_scope = scope;
3424         ifm->ifa_index = ifindex;
3425 }
3426
3427 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3428                          unsigned long tstamp, u32 preferred, u32 valid)
3429 {
3430         struct ifa_cacheinfo ci;
3431
3432         ci.cstamp = cstamp_delta(cstamp);
3433         ci.tstamp = cstamp_delta(tstamp);
3434         ci.ifa_prefered = preferred;
3435         ci.ifa_valid = valid;
3436
3437         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3438 }
3439
3440 static inline int rt_scope(int ifa_scope)
3441 {
3442         if (ifa_scope & IFA_HOST)
3443                 return RT_SCOPE_HOST;
3444         else if (ifa_scope & IFA_LINK)
3445                 return RT_SCOPE_LINK;
3446         else if (ifa_scope & IFA_SITE)
3447                 return RT_SCOPE_SITE;
3448         else
3449                 return RT_SCOPE_UNIVERSE;
3450 }
3451
3452 static inline int inet6_ifaddr_msgsize(void)
3453 {
3454         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3455                + nla_total_size(16) /* IFA_ADDRESS */
3456                + nla_total_size(sizeof(struct ifa_cacheinfo));
3457 }
3458
3459 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3460                              u32 pid, u32 seq, int event, unsigned int flags)
3461 {
3462         struct nlmsghdr  *nlh;
3463         u32 preferred, valid;
3464
3465         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3466         if (nlh == NULL)
3467                 return -EMSGSIZE;
3468
3469         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3470                       ifa->idev->dev->ifindex);
3471
3472         if (!(ifa->flags&IFA_F_PERMANENT)) {
3473                 preferred = ifa->prefered_lft;
3474                 valid = ifa->valid_lft;
3475                 if (preferred != INFINITY_LIFE_TIME) {
3476                         long tval = (jiffies - ifa->tstamp)/HZ;
3477                         if (preferred > tval)
3478                                 preferred -= tval;
3479                         else
3480                                 preferred = 0;
3481                         if (valid != INFINITY_LIFE_TIME) {
3482                                 if (valid > tval)
3483                                         valid -= tval;
3484                                 else
3485                                         valid = 0;
3486                         }
3487                 }
3488         } else {
3489                 preferred = INFINITY_LIFE_TIME;
3490                 valid = INFINITY_LIFE_TIME;
3491         }
3492
3493         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3494             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3495                 nlmsg_cancel(skb, nlh);
3496                 return -EMSGSIZE;
3497         }
3498
3499         return nlmsg_end(skb, nlh);
3500 }
3501
3502 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3503                                 u32 pid, u32 seq, int event, u16 flags)
3504 {
3505         struct nlmsghdr  *nlh;
3506         u8 scope = RT_SCOPE_UNIVERSE;
3507         int ifindex = ifmca->idev->dev->ifindex;
3508
3509         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3510                 scope = RT_SCOPE_SITE;
3511
3512         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3513         if (nlh == NULL)
3514                 return -EMSGSIZE;
3515
3516         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3517         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3518             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3519                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3520                 nlmsg_cancel(skb, nlh);
3521                 return -EMSGSIZE;
3522         }
3523
3524         return nlmsg_end(skb, nlh);
3525 }
3526
3527 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3528                                 u32 pid, u32 seq, int event, unsigned int flags)
3529 {
3530         struct nlmsghdr  *nlh;
3531         u8 scope = RT_SCOPE_UNIVERSE;
3532         int ifindex = ifaca->aca_idev->dev->ifindex;
3533
3534         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3535                 scope = RT_SCOPE_SITE;
3536
3537         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3538         if (nlh == NULL)
3539                 return -EMSGSIZE;
3540
3541         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3542         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3543             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3544                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3545                 nlmsg_cancel(skb, nlh);
3546                 return -EMSGSIZE;
3547         }
3548
3549         return nlmsg_end(skb, nlh);
3550 }
3551
3552 enum addr_type_t {
3553         UNICAST_ADDR,
3554         MULTICAST_ADDR,
3555         ANYCAST_ADDR,
3556 };
3557
3558 /* called with rcu_read_lock() */
3559 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3560                           struct netlink_callback *cb, enum addr_type_t type,
3561                           int s_ip_idx, int *p_ip_idx)
3562 {
3563         struct ifmcaddr6 *ifmca;
3564         struct ifacaddr6 *ifaca;
3565         int err = 1;
3566         int ip_idx = *p_ip_idx;
3567
3568         read_lock_bh(&idev->lock);
3569         switch (type) {
3570         case UNICAST_ADDR: {
3571                 struct inet6_ifaddr *ifa;
3572
3573                 /* unicast address incl. temp addr */
3574                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3575                         if (++ip_idx < s_ip_idx)
3576                                 continue;
3577                         err = inet6_fill_ifaddr(skb, ifa,
3578                                                 NETLINK_CB(cb->skb).pid,
3579                                                 cb->nlh->nlmsg_seq,
3580                                                 RTM_NEWADDR,
3581                                                 NLM_F_MULTI);
3582                         if (err <= 0)
3583                                 break;
3584                 }
3585                 break;
3586         }
3587         case MULTICAST_ADDR:
3588                 /* multicast address */
3589                 for (ifmca = idev->mc_list; ifmca;
3590                      ifmca = ifmca->next, ip_idx++) {
3591                         if (ip_idx < s_ip_idx)
3592                                 continue;
3593                         err = inet6_fill_ifmcaddr(skb, ifmca,
3594                                                   NETLINK_CB(cb->skb).pid,
3595                                                   cb->nlh->nlmsg_seq,
3596                                                   RTM_GETMULTICAST,
3597                                                   NLM_F_MULTI);
3598                         if (err <= 0)
3599                                 break;
3600                 }
3601                 break;
3602         case ANYCAST_ADDR:
3603                 /* anycast address */
3604                 for (ifaca = idev->ac_list; ifaca;
3605                      ifaca = ifaca->aca_next, ip_idx++) {
3606                         if (ip_idx < s_ip_idx)
3607                                 continue;
3608                         err = inet6_fill_ifacaddr(skb, ifaca,
3609                                                   NETLINK_CB(cb->skb).pid,
3610                                                   cb->nlh->nlmsg_seq,
3611                                                   RTM_GETANYCAST,
3612                                                   NLM_F_MULTI);
3613                         if (err <= 0)
3614                                 break;
3615                 }
3616                 break;
3617         default:
3618                 break;
3619         }
3620         read_unlock_bh(&idev->lock);
3621         *p_ip_idx = ip_idx;
3622         return err;
3623 }
3624
3625 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3626                            enum addr_type_t type)
3627 {
3628         struct net *net = sock_net(skb->sk);
3629         int h, s_h;
3630         int idx, ip_idx;
3631         int s_idx, s_ip_idx;
3632         struct net_device *dev;
3633         struct inet6_dev *idev;
3634         struct hlist_head *head;
3635         struct hlist_node *node;
3636
3637         s_h = cb->args[0];
3638         s_idx = idx = cb->args[1];
3639         s_ip_idx = ip_idx = cb->args[2];
3640
3641         rcu_read_lock();
3642         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3643                 idx = 0;
3644                 head = &net->dev_index_head[h];
3645                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3646                         if (idx < s_idx)
3647                                 goto cont;
3648                         if (h > s_h || idx > s_idx)
3649                                 s_ip_idx = 0;
3650                         ip_idx = 0;
3651                         idev = __in6_dev_get(dev);
3652                         if (!idev)
3653                                 goto cont;
3654
3655                         if (in6_dump_addrs(idev, skb, cb, type,
3656                                            s_ip_idx, &ip_idx) <= 0)
3657                                 goto done;
3658 cont:
3659                         idx++;
3660                 }
3661         }
3662 done:
3663         rcu_read_unlock();
3664         cb->args[0] = h;
3665         cb->args[1] = idx;
3666         cb->args[2] = ip_idx;
3667
3668         return skb->len;
3669 }
3670
3671 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3672 {
3673         enum addr_type_t type = UNICAST_ADDR;
3674
3675         return inet6_dump_addr(skb, cb, type);
3676 }
3677
3678 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3679 {
3680         enum addr_type_t type = MULTICAST_ADDR;
3681
3682         return inet6_dump_addr(skb, cb, type);
3683 }
3684
3685
3686 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3687 {
3688         enum addr_type_t type = ANYCAST_ADDR;
3689
3690         return inet6_dump_addr(skb, cb, type);
3691 }
3692
3693 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3694                              void *arg)
3695 {
3696         struct net *net = sock_net(in_skb->sk);
3697         struct ifaddrmsg *ifm;
3698         struct nlattr *tb[IFA_MAX+1];
3699         struct in6_addr *addr = NULL;
3700         struct net_device *dev = NULL;
3701         struct inet6_ifaddr *ifa;
3702         struct sk_buff *skb;
3703         int err;
3704
3705         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3706         if (err < 0)
3707                 goto errout;
3708
3709         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3710         if (addr == NULL) {
3711                 err = -EINVAL;
3712                 goto errout;
3713         }
3714
3715         ifm = nlmsg_data(nlh);
3716         if (ifm->ifa_index)
3717                 dev = __dev_get_by_index(net, ifm->ifa_index);
3718
3719         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3720         if (!ifa) {
3721                 err = -EADDRNOTAVAIL;
3722                 goto errout;
3723         }
3724
3725         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3726         if (!skb) {
3727                 err = -ENOBUFS;
3728                 goto errout_ifa;
3729         }
3730
3731         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3732                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3733         if (err < 0) {
3734                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3735                 WARN_ON(err == -EMSGSIZE);
3736                 kfree_skb(skb);
3737                 goto errout_ifa;
3738         }
3739         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3740 errout_ifa:
3741         in6_ifa_put(ifa);
3742 errout:
3743         return err;
3744 }
3745
3746 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3747 {
3748         struct sk_buff *skb;
3749         struct net *net = dev_net(ifa->idev->dev);
3750         int err = -ENOBUFS;
3751
3752         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3753         if (skb == NULL)
3754                 goto errout;
3755
3756         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3757         if (err < 0) {
3758                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3759                 WARN_ON(err == -EMSGSIZE);
3760                 kfree_skb(skb);
3761                 goto errout;
3762         }
3763         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3764         return;
3765 errout:
3766         if (err < 0)
3767                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3768 }
3769
3770 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3771                                 __s32 *array, int bytes)
3772 {
3773         BUG_ON(bytes < (DEVCONF_MAX * 4));
3774
3775         memset(array, 0, bytes);
3776         array[DEVCONF_FORWARDING] = cnf->forwarding;
3777         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3778         array[DEVCONF_MTU6] = cnf->mtu6;
3779         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3780         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3781         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3782         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3783         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3784         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3785                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3786         array[DEVCONF_RTR_SOLICIT_DELAY] =
3787                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3788         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3789 #ifdef CONFIG_IPV6_PRIVACY
3790         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3791         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3792         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3793         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3794         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3795 #endif
3796         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3797         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3798         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3799 #ifdef CONFIG_IPV6_ROUTER_PREF
3800         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3801         array[DEVCONF_RTR_PROBE_INTERVAL] =
3802                 jiffies_to_msecs(cnf->rtr_probe_interval);
3803 #ifdef CONFIG_IPV6_ROUTE_INFO
3804         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3805 #endif
3806 #endif
3807         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3808         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3809 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3810         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3811 #endif
3812 #ifdef CONFIG_IPV6_MROUTE
3813         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3814 #endif
3815         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3816         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3817         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3818 }
3819
3820 static inline size_t inet6_ifla6_size(void)
3821 {
3822         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3823              + nla_total_size(sizeof(struct ifla_cacheinfo))
3824              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3825              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3826              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3827 }
3828
3829 static inline size_t inet6_if_nlmsg_size(void)
3830 {
3831         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3832                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3833                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3834                + nla_total_size(4) /* IFLA_MTU */
3835                + nla_total_size(4) /* IFLA_LINK */
3836                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3837 }
3838
3839 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3840                                       int items, int bytes)
3841 {
3842         int i;
3843         int pad = bytes - sizeof(u64) * items;
3844         BUG_ON(pad < 0);
3845
3846         /* Use put_unaligned() because stats may not be aligned for u64. */
3847         put_unaligned(items, &stats[0]);
3848         for (i = 1; i < items; i++)
3849                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3850
3851         memset(&stats[items], 0, pad);
3852 }
3853
3854 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3855                                       int items, int bytes, size_t syncpoff)
3856 {
3857         int i;
3858         int pad = bytes - sizeof(u64) * items;
3859         BUG_ON(pad < 0);
3860
3861         /* Use put_unaligned() because stats may not be aligned for u64. */
3862         put_unaligned(items, &stats[0]);
3863         for (i = 1; i < items; i++)
3864                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3865
3866         memset(&stats[items], 0, pad);
3867 }
3868
3869 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3870                              int bytes)
3871 {
3872         switch (attrtype) {
3873         case IFLA_INET6_STATS:
3874                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3875                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3876                 break;
3877         case IFLA_INET6_ICMP6STATS:
3878                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3879                 break;
3880         }
3881 }
3882
3883 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3884 {
3885         struct nlattr *nla;
3886         struct ifla_cacheinfo ci;
3887
3888         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3889
3890         ci.max_reasm_len = IPV6_MAXPLEN;
3891         ci.tstamp = cstamp_delta(idev->tstamp);
3892         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3893         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3894         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3895
3896         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3897         if (nla == NULL)
3898                 goto nla_put_failure;
3899         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3900
3901         /* XXX - MC not implemented */
3902
3903         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3904         if (nla == NULL)
3905                 goto nla_put_failure;
3906         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3907
3908         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3909         if (nla == NULL)
3910                 goto nla_put_failure;
3911         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3912
3913         return 0;
3914
3915 nla_put_failure:
3916         return -EMSGSIZE;
3917 }
3918
3919 static size_t inet6_get_link_af_size(const struct net_device *dev)
3920 {
3921         if (!__in6_dev_get(dev))
3922                 return 0;
3923
3924         return inet6_ifla6_size();
3925 }
3926
3927 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3928 {
3929         struct inet6_dev *idev = __in6_dev_get(dev);
3930
3931         if (!idev)
3932                 return -ENODATA;
3933
3934         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3935                 return -EMSGSIZE;
3936
3937         return 0;
3938 }
3939
3940 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3941                              u32 pid, u32 seq, int event, unsigned int flags)
3942 {
3943         struct net_device *dev = idev->dev;
3944         struct ifinfomsg *hdr;
3945         struct nlmsghdr *nlh;
3946         void *protoinfo;
3947
3948         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3949         if (nlh == NULL)
3950                 return -EMSGSIZE;
3951
3952         hdr = nlmsg_data(nlh);
3953         hdr->ifi_family = AF_INET6;
3954         hdr->__ifi_pad = 0;
3955         hdr->ifi_type = dev->type;
3956         hdr->ifi_index = dev->ifindex;
3957         hdr->ifi_flags = dev_get_flags(dev);
3958         hdr->ifi_change = 0;
3959
3960         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3961
3962         if (dev->addr_len)
3963                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3964
3965         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3966         if (dev->ifindex != dev->iflink)
3967                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3968
3969         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3970         if (protoinfo == NULL)
3971                 goto nla_put_failure;
3972
3973         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3974                 goto nla_put_failure;
3975
3976         nla_nest_end(skb, protoinfo);
3977         return nlmsg_end(skb, nlh);
3978
3979 nla_put_failure:
3980         nlmsg_cancel(skb, nlh);
3981         return -EMSGSIZE;
3982 }
3983
3984 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3985 {
3986         struct net *net = sock_net(skb->sk);
3987         int h, s_h;
3988         int idx = 0, s_idx;
3989         struct net_device *dev;
3990         struct inet6_dev *idev;
3991         struct hlist_head *head;
3992         struct hlist_node *node;
3993
3994         s_h = cb->args[0];
3995         s_idx = cb->args[1];
3996
3997         rcu_read_lock();
3998         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3999                 idx = 0;
4000                 head = &net->dev_index_head[h];
4001                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4002                         if (idx < s_idx)
4003                                 goto cont;
4004                         idev = __in6_dev_get(dev);
4005                         if (!idev)
4006                                 goto cont;
4007                         if (inet6_fill_ifinfo(skb, idev,
4008                                               NETLINK_CB(cb->skb).pid,
4009                                               cb->nlh->nlmsg_seq,
4010                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4011                                 goto out;
4012 cont:
4013                         idx++;
4014                 }
4015         }
4016 out:
4017         rcu_read_unlock();
4018         cb->args[1] = idx;
4019         cb->args[0] = h;
4020
4021         return skb->len;
4022 }
4023
4024 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4025 {
4026         struct sk_buff *skb;
4027         struct net *net = dev_net(idev->dev);
4028         int err = -ENOBUFS;
4029
4030         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4031         if (skb == NULL)
4032                 goto errout;
4033
4034         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4035         if (err < 0) {
4036                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4037                 WARN_ON(err == -EMSGSIZE);
4038                 kfree_skb(skb);
4039                 goto errout;
4040         }
4041         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4042         return;
4043 errout:
4044         if (err < 0)
4045                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4046 }
4047
4048 static inline size_t inet6_prefix_nlmsg_size(void)
4049 {
4050         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4051                + nla_total_size(sizeof(struct in6_addr))
4052                + nla_total_size(sizeof(struct prefix_cacheinfo));
4053 }
4054
4055 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4056                              struct prefix_info *pinfo, u32 pid, u32 seq,
4057                              int event, unsigned int flags)
4058 {
4059         struct prefixmsg *pmsg;
4060         struct nlmsghdr *nlh;
4061         struct prefix_cacheinfo ci;
4062
4063         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4064         if (nlh == NULL)
4065                 return -EMSGSIZE;
4066
4067         pmsg = nlmsg_data(nlh);
4068         pmsg->prefix_family = AF_INET6;
4069         pmsg->prefix_pad1 = 0;
4070         pmsg->prefix_pad2 = 0;
4071         pmsg->prefix_ifindex = idev->dev->ifindex;
4072         pmsg->prefix_len = pinfo->prefix_len;
4073         pmsg->prefix_type = pinfo->type;
4074         pmsg->prefix_pad3 = 0;
4075         pmsg->prefix_flags = 0;
4076         if (pinfo->onlink)
4077                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4078         if (pinfo->autoconf)
4079                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4080
4081         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4082
4083         ci.preferred_time = ntohl(pinfo->prefered);
4084         ci.valid_time = ntohl(pinfo->valid);
4085         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4086
4087         return nlmsg_end(skb, nlh);
4088
4089 nla_put_failure:
4090         nlmsg_cancel(skb, nlh);
4091         return -EMSGSIZE;
4092 }
4093
4094 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4095                          struct prefix_info *pinfo)
4096 {
4097         struct sk_buff *skb;
4098         struct net *net = dev_net(idev->dev);
4099         int err = -ENOBUFS;
4100
4101         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4102         if (skb == NULL)
4103                 goto errout;
4104
4105         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4106         if (err < 0) {
4107                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4108                 WARN_ON(err == -EMSGSIZE);
4109                 kfree_skb(skb);
4110                 goto errout;
4111         }
4112         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4113         return;
4114 errout:
4115         if (err < 0)
4116                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4117 }
4118
4119 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4120 {
4121         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4122
4123         switch (event) {
4124         case RTM_NEWADDR:
4125                 /*
4126                  * If the address was optimistic
4127                  * we inserted the route at the start of
4128                  * our DAD process, so we don't need
4129                  * to do it again
4130                  */
4131                 if (!(ifp->rt->rt6i_node))
4132                         ip6_ins_rt(ifp->rt);
4133                 if (ifp->idev->cnf.forwarding)
4134                         addrconf_join_anycast(ifp);
4135                 break;
4136         case RTM_DELADDR:
4137                 if (ifp->idev->cnf.forwarding)
4138                         addrconf_leave_anycast(ifp);
4139                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4140                 dst_hold(&ifp->rt->dst);
4141
4142                 if (ip6_del_rt(ifp->rt))
4143                         dst_free(&ifp->rt->dst);
4144                 break;
4145         }
4146 }
4147
4148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4149 {
4150         rcu_read_lock_bh();
4151         if (likely(ifp->idev->dead == 0))
4152                 __ipv6_ifa_notify(event, ifp);
4153         rcu_read_unlock_bh();
4154 }
4155
4156 #ifdef CONFIG_SYSCTL
4157
4158 static
4159 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4160                            void __user *buffer, size_t *lenp, loff_t *ppos)
4161 {
4162         int *valp = ctl->data;
4163         int val = *valp;
4164         loff_t pos = *ppos;
4165         int ret;
4166
4167         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4168
4169         if (write)
4170                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4171         if (ret)
4172                 *ppos = pos;
4173         return ret;
4174 }
4175
4176 static void dev_disable_change(struct inet6_dev *idev)
4177 {
4178         if (!idev || !idev->dev)
4179                 return;
4180
4181         if (idev->cnf.disable_ipv6)
4182                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4183         else
4184                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4185 }
4186
4187 static void addrconf_disable_change(struct net *net, __s32 newf)
4188 {
4189         struct net_device *dev;
4190         struct inet6_dev *idev;
4191
4192         rcu_read_lock();
4193         for_each_netdev_rcu(net, dev) {
4194                 idev = __in6_dev_get(dev);
4195                 if (idev) {
4196                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4197                         idev->cnf.disable_ipv6 = newf;
4198                         if (changed)
4199                                 dev_disable_change(idev);
4200                 }
4201         }
4202         rcu_read_unlock();
4203 }
4204
4205 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4206 {
4207         struct net *net;
4208
4209         net = (struct net *)table->extra2;
4210
4211         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4212                 return 0;
4213
4214         if (!rtnl_trylock()) {
4215                 /* Restore the original values before restarting */
4216                 *p = old;
4217                 return restart_syscall();
4218         }
4219
4220         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4221                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4222                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4223                 addrconf_disable_change(net, newf);
4224         } else if ((!*p) ^ (!old))
4225                 dev_disable_change((struct inet6_dev *)table->extra1);
4226
4227         rtnl_unlock();
4228         return 0;
4229 }
4230
4231 static
4232 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4233                             void __user *buffer, size_t *lenp, loff_t *ppos)
4234 {
4235         int *valp = ctl->data;
4236         int val = *valp;
4237         loff_t pos = *ppos;
4238         int ret;
4239
4240         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4241
4242         if (write)
4243                 ret = addrconf_disable_ipv6(ctl, valp, val);
4244         if (ret)
4245                 *ppos = pos;
4246         return ret;
4247 }
4248
4249 static struct addrconf_sysctl_table
4250 {
4251         struct ctl_table_header *sysctl_header;
4252         ctl_table addrconf_vars[DEVCONF_MAX+1];
4253         char *dev_name;
4254 } addrconf_sysctl __read_mostly = {
4255         .sysctl_header = NULL,
4256         .addrconf_vars = {
4257                 {
4258                         .procname       = "forwarding",
4259                         .data           = &ipv6_devconf.forwarding,
4260                         .maxlen         = sizeof(int),
4261                         .mode           = 0644,
4262                         .proc_handler   = addrconf_sysctl_forward,
4263                 },
4264                 {
4265                         .procname       = "hop_limit",
4266                         .data           = &ipv6_devconf.hop_limit,
4267                         .maxlen         = sizeof(int),
4268                         .mode           = 0644,
4269                         .proc_handler   = proc_dointvec,
4270                 },
4271                 {
4272                         .procname       = "mtu",
4273                         .data           = &ipv6_devconf.mtu6,
4274                         .maxlen         = sizeof(int),
4275                         .mode           = 0644,
4276                         .proc_handler   = proc_dointvec,
4277                 },
4278                 {
4279                         .procname       = "accept_ra",
4280                         .data           = &ipv6_devconf.accept_ra,
4281                         .maxlen         = sizeof(int),
4282                         .mode           = 0644,
4283                         .proc_handler   = proc_dointvec,
4284                 },
4285                 {
4286                         .procname       = "accept_redirects",
4287                         .data           = &ipv6_devconf.accept_redirects,
4288                         .maxlen         = sizeof(int),
4289                         .mode           = 0644,
4290                         .proc_handler   = proc_dointvec,
4291                 },
4292                 {
4293                         .procname       = "autoconf",
4294                         .data           = &ipv6_devconf.autoconf,
4295                         .maxlen         = sizeof(int),
4296                         .mode           = 0644,
4297                         .proc_handler   = proc_dointvec,
4298                 },
4299                 {
4300                         .procname       = "dad_transmits",
4301                         .data           = &ipv6_devconf.dad_transmits,
4302                         .maxlen         = sizeof(int),
4303                         .mode           = 0644,
4304                         .proc_handler   = proc_dointvec,
4305                 },
4306                 {
4307                         .procname       = "router_solicitations",
4308                         .data           = &ipv6_devconf.rtr_solicits,
4309                         .maxlen         = sizeof(int),
4310                         .mode           = 0644,
4311                         .proc_handler   = proc_dointvec,
4312                 },
4313                 {
4314                         .procname       = "router_solicitation_interval",
4315                         .data           = &ipv6_devconf.rtr_solicit_interval,
4316                         .maxlen         = sizeof(int),
4317                         .mode           = 0644,
4318                         .proc_handler   = proc_dointvec_jiffies,
4319                 },
4320                 {
4321                         .procname       = "router_solicitation_delay",
4322                         .data           = &ipv6_devconf.rtr_solicit_delay,
4323                         .maxlen         = sizeof(int),
4324                         .mode           = 0644,
4325                         .proc_handler   = proc_dointvec_jiffies,
4326                 },
4327                 {
4328                         .procname       = "force_mld_version",
4329                         .data           = &ipv6_devconf.force_mld_version,
4330                         .maxlen         = sizeof(int),
4331                         .mode           = 0644,
4332                         .proc_handler   = proc_dointvec,
4333                 },
4334 #ifdef CONFIG_IPV6_PRIVACY
4335                 {
4336                         .procname       = "use_tempaddr",
4337                         .data           = &ipv6_devconf.use_tempaddr,
4338                         .maxlen         = sizeof(int),
4339                         .mode           = 0644,
4340                         .proc_handler   = proc_dointvec,
4341                 },
4342                 {
4343                         .procname       = "temp_valid_lft",
4344                         .data           = &ipv6_devconf.temp_valid_lft,
4345                         .maxlen         = sizeof(int),
4346                         .mode           = 0644,
4347                         .proc_handler   = proc_dointvec,
4348                 },
4349                 {
4350                         .procname       = "temp_prefered_lft",
4351                         .data           = &ipv6_devconf.temp_prefered_lft,
4352                         .maxlen         = sizeof(int),
4353                         .mode           = 0644,
4354                         .proc_handler   = proc_dointvec,
4355                 },
4356                 {
4357                         .procname       = "regen_max_retry",
4358                         .data           = &ipv6_devconf.regen_max_retry,
4359                         .maxlen         = sizeof(int),
4360                         .mode           = 0644,
4361                         .proc_handler   = proc_dointvec,
4362                 },
4363                 {
4364                         .procname       = "max_desync_factor",
4365                         .data           = &ipv6_devconf.max_desync_factor,
4366                         .maxlen         = sizeof(int),
4367                         .mode           = 0644,
4368                         .proc_handler   = proc_dointvec,
4369                 },
4370 #endif
4371                 {
4372                         .procname       = "max_addresses",
4373                         .data           = &ipv6_devconf.max_addresses,
4374                         .maxlen         = sizeof(int),
4375                         .mode           = 0644,
4376                         .proc_handler   = proc_dointvec,
4377                 },
4378                 {
4379                         .procname       = "accept_ra_defrtr",
4380                         .data           = &ipv6_devconf.accept_ra_defrtr,
4381                         .maxlen         = sizeof(int),
4382                         .mode           = 0644,
4383                         .proc_handler   = proc_dointvec,
4384                 },
4385                 {
4386                         .procname       = "accept_ra_pinfo",
4387                         .data           = &ipv6_devconf.accept_ra_pinfo,
4388                         .maxlen         = sizeof(int),
4389                         .mode           = 0644,
4390                         .proc_handler   = proc_dointvec,
4391                 },
4392 #ifdef CONFIG_IPV6_ROUTER_PREF
4393                 {
4394                         .procname       = "accept_ra_rtr_pref",
4395                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4396                         .maxlen         = sizeof(int),
4397                         .mode           = 0644,
4398                         .proc_handler   = proc_dointvec,
4399                 },
4400                 {
4401                         .procname       = "router_probe_interval",
4402                         .data           = &ipv6_devconf.rtr_probe_interval,
4403                         .maxlen         = sizeof(int),
4404                         .mode           = 0644,
4405                         .proc_handler   = proc_dointvec_jiffies,
4406                 },
4407 #ifdef CONFIG_IPV6_ROUTE_INFO
4408                 {
4409                         .procname       = "accept_ra_rt_info_max_plen",
4410                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4411                         .maxlen         = sizeof(int),
4412                         .mode           = 0644,
4413                         .proc_handler   = proc_dointvec,
4414                 },
4415 #endif
4416 #endif
4417                 {
4418                         .procname       = "proxy_ndp",
4419                         .data           = &ipv6_devconf.proxy_ndp,
4420                         .maxlen         = sizeof(int),
4421                         .mode           = 0644,
4422                         .proc_handler   = proc_dointvec,
4423                 },
4424                 {
4425                         .procname       = "accept_source_route",
4426                         .data           = &ipv6_devconf.accept_source_route,
4427                         .maxlen         = sizeof(int),
4428                         .mode           = 0644,
4429                         .proc_handler   = proc_dointvec,
4430                 },
4431 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4432                 {
4433                         .procname       = "optimistic_dad",
4434                         .data           = &ipv6_devconf.optimistic_dad,
4435                         .maxlen         = sizeof(int),
4436                         .mode           = 0644,
4437                         .proc_handler   = proc_dointvec,
4438
4439                 },
4440 #endif
4441 #ifdef CONFIG_IPV6_MROUTE
4442                 {
4443                         .procname       = "mc_forwarding",
4444                         .data           = &ipv6_devconf.mc_forwarding,
4445                         .maxlen         = sizeof(int),
4446                         .mode           = 0444,
4447                         .proc_handler   = proc_dointvec,
4448                 },
4449 #endif
4450                 {
4451                         .procname       = "disable_ipv6",
4452                         .data           = &ipv6_devconf.disable_ipv6,
4453                         .maxlen         = sizeof(int),
4454                         .mode           = 0644,
4455                         .proc_handler   = addrconf_sysctl_disable,
4456                 },
4457                 {
4458                         .procname       = "accept_dad",
4459                         .data           = &ipv6_devconf.accept_dad,
4460                         .maxlen         = sizeof(int),
4461                         .mode           = 0644,
4462                         .proc_handler   = proc_dointvec,
4463                 },
4464                 {
4465                         .procname       = "force_tllao",
4466                         .data           = &ipv6_devconf.force_tllao,
4467                         .maxlen         = sizeof(int),
4468                         .mode           = 0644,
4469                         .proc_handler   = proc_dointvec
4470                 },
4471                 {
4472                         /* sentinel */
4473                 }
4474         },
4475 };
4476
4477 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4478                 struct inet6_dev *idev, struct ipv6_devconf *p)
4479 {
4480         int i;
4481         struct addrconf_sysctl_table *t;
4482
4483 #define ADDRCONF_CTL_PATH_DEV   3
4484
4485         struct ctl_path addrconf_ctl_path[] = {
4486                 { .procname = "net", },
4487                 { .procname = "ipv6", },
4488                 { .procname = "conf", },
4489                 { /* to be set */ },
4490                 { },
4491         };
4492
4493
4494         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4495         if (t == NULL)
4496                 goto out;
4497
4498         for (i = 0; t->addrconf_vars[i].data; i++) {
4499                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4500                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4501                 t->addrconf_vars[i].extra2 = net;
4502         }
4503
4504         /*
4505          * Make a copy of dev_name, because '.procname' is regarded as const
4506          * by sysctl and we wouldn't want anyone to change it under our feet
4507          * (see SIOCSIFNAME).
4508          */
4509         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4510         if (!t->dev_name)
4511                 goto free;
4512
4513         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4514
4515         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4516                         t->addrconf_vars);
4517         if (t->sysctl_header == NULL)
4518                 goto free_procname;
4519
4520         p->sysctl = t;
4521         return 0;
4522
4523 free_procname:
4524         kfree(t->dev_name);
4525 free:
4526         kfree(t);
4527 out:
4528         return -ENOBUFS;
4529 }
4530
4531 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4532 {
4533         struct addrconf_sysctl_table *t;
4534
4535         if (p->sysctl == NULL)
4536                 return;
4537
4538         t = p->sysctl;
4539         p->sysctl = NULL;
4540         unregister_net_sysctl_table(t->sysctl_header);
4541         kfree(t->dev_name);
4542         kfree(t);
4543 }
4544
4545 static void addrconf_sysctl_register(struct inet6_dev *idev)
4546 {
4547         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4548                               &ndisc_ifinfo_sysctl_change);
4549         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4550                                         idev, &idev->cnf);
4551 }
4552
4553 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4554 {
4555         __addrconf_sysctl_unregister(&idev->cnf);
4556         neigh_sysctl_unregister(idev->nd_parms);
4557 }
4558
4559
4560 #endif
4561
4562 static int __net_init addrconf_init_net(struct net *net)
4563 {
4564         int err;
4565         struct ipv6_devconf *all, *dflt;
4566
4567         err = -ENOMEM;
4568         all = &ipv6_devconf;
4569         dflt = &ipv6_devconf_dflt;
4570
4571         if (!net_eq(net, &init_net)) {
4572                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4573                 if (all == NULL)
4574                         goto err_alloc_all;
4575
4576                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4577                 if (dflt == NULL)
4578                         goto err_alloc_dflt;
4579         } else {
4580                 /* these will be inherited by all namespaces */
4581                 dflt->autoconf = ipv6_defaults.autoconf;
4582                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4583         }
4584
4585         net->ipv6.devconf_all = all;
4586         net->ipv6.devconf_dflt = dflt;
4587
4588 #ifdef CONFIG_SYSCTL
4589         err = __addrconf_sysctl_register(net, "all", NULL, all);
4590         if (err < 0)
4591                 goto err_reg_all;
4592
4593         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4594         if (err < 0)
4595                 goto err_reg_dflt;
4596 #endif
4597         return 0;
4598
4599 #ifdef CONFIG_SYSCTL
4600 err_reg_dflt:
4601         __addrconf_sysctl_unregister(all);
4602 err_reg_all:
4603         kfree(dflt);
4604 #endif
4605 err_alloc_dflt:
4606         kfree(all);
4607 err_alloc_all:
4608         return err;
4609 }
4610
4611 static void __net_exit addrconf_exit_net(struct net *net)
4612 {
4613 #ifdef CONFIG_SYSCTL
4614         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4615         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4616 #endif
4617         if (!net_eq(net, &init_net)) {
4618                 kfree(net->ipv6.devconf_dflt);
4619                 kfree(net->ipv6.devconf_all);
4620         }
4621 }
4622
4623 static struct pernet_operations addrconf_ops = {
4624         .init = addrconf_init_net,
4625         .exit = addrconf_exit_net,
4626 };
4627
4628 /*
4629  *      Device notifier
4630  */
4631
4632 int register_inet6addr_notifier(struct notifier_block *nb)
4633 {
4634         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4635 }
4636 EXPORT_SYMBOL(register_inet6addr_notifier);
4637
4638 int unregister_inet6addr_notifier(struct notifier_block *nb)
4639 {
4640         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4641 }
4642 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4643
4644 static struct rtnl_af_ops inet6_ops = {
4645         .family           = AF_INET6,
4646         .fill_link_af     = inet6_fill_link_af,
4647         .get_link_af_size = inet6_get_link_af_size,
4648 };
4649
4650 /*
4651  *      Init / cleanup code
4652  */
4653
4654 int __init addrconf_init(void)
4655 {
4656         int i, err;
4657
4658         err = ipv6_addr_label_init();
4659         if (err < 0) {
4660                 printk(KERN_CRIT "IPv6 Addrconf:"
4661                        " cannot initialize default policy table: %d.\n", err);
4662                 goto out;
4663         }
4664
4665         err = register_pernet_subsys(&addrconf_ops);
4666         if (err < 0)
4667                 goto out_addrlabel;
4668
4669         /* The addrconf netdev notifier requires that loopback_dev
4670          * has it's ipv6 private information allocated and setup
4671          * before it can bring up and give link-local addresses
4672          * to other devices which are up.
4673          *
4674          * Unfortunately, loopback_dev is not necessarily the first
4675          * entry in the global dev_base list of net devices.  In fact,
4676          * it is likely to be the very last entry on that list.
4677          * So this causes the notifier registry below to try and
4678          * give link-local addresses to all devices besides loopback_dev
4679          * first, then loopback_dev, which cases all the non-loopback_dev
4680          * devices to fail to get a link-local address.
4681          *
4682          * So, as a temporary fix, allocate the ipv6 structure for
4683          * loopback_dev first by hand.
4684          * Longer term, all of the dependencies ipv6 has upon the loopback
4685          * device and it being up should be removed.
4686          */
4687         rtnl_lock();
4688         if (!ipv6_add_dev(init_net.loopback_dev))
4689                 err = -ENOMEM;
4690         rtnl_unlock();
4691         if (err)
4692                 goto errlo;
4693
4694         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4695                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4696
4697         register_netdevice_notifier(&ipv6_dev_notf);
4698
4699         addrconf_verify(0);
4700
4701         err = rtnl_af_register(&inet6_ops);
4702         if (err < 0)
4703                 goto errout_af;
4704
4705         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4706         if (err < 0)
4707                 goto errout;
4708
4709         /* Only the first call to __rtnl_register can fail */
4710         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4711         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4712         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4713         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4714         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4715
4716         ipv6_addr_label_rtnl_register();
4717
4718         return 0;
4719 errout:
4720         rtnl_af_unregister(&inet6_ops);
4721 errout_af:
4722         unregister_netdevice_notifier(&ipv6_dev_notf);
4723 errlo:
4724         unregister_pernet_subsys(&addrconf_ops);
4725 out_addrlabel:
4726         ipv6_addr_label_cleanup();
4727 out:
4728         return err;
4729 }
4730
4731 void addrconf_cleanup(void)
4732 {
4733         struct net_device *dev;
4734         int i;
4735
4736         unregister_netdevice_notifier(&ipv6_dev_notf);
4737         unregister_pernet_subsys(&addrconf_ops);
4738         ipv6_addr_label_cleanup();
4739
4740         rtnl_lock();
4741
4742         __rtnl_af_unregister(&inet6_ops);
4743
4744         /* clean dev list */
4745         for_each_netdev(&init_net, dev) {
4746                 if (__in6_dev_get(dev) == NULL)
4747                         continue;
4748                 addrconf_ifdown(dev, 1);
4749         }
4750         addrconf_ifdown(init_net.loopback_dev, 2);
4751
4752         /*
4753          *      Check hash table.
4754          */
4755         spin_lock_bh(&addrconf_hash_lock);
4756         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4757                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4758         spin_unlock_bh(&addrconf_hash_lock);
4759
4760         del_timer(&addr_chk_timer);
4761         rtnl_unlock();
4762 }