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

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