net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_min_hop_limit= 1,
 189         .accept_ra_pinfo        = 1,
 190 #ifdef CONFIG_IPV6_ROUTER_PREF
 191         .accept_ra_rtr_pref     = 1,
 192         .rtr_probe_interval     = 60 * HZ,
 193 #ifdef CONFIG_IPV6_ROUTE_INFO
 194         .accept_ra_rt_info_max_plen = 0,
 195 #endif
 196 #endif
 197         .proxy_ndp              = 0,
 198         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 199         .disable_ipv6           = 0,
 200         .accept_dad             = 1,
 201 };
 202
 203 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 204         .forwarding             = 0,
 205         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 206         .mtu6                   = IPV6_MIN_MTU,
 207         .accept_ra              = 1,
 208         .accept_redirects       = 1,
 209         .autoconf               = 1,
 210         .dad_transmits          = 1,
 211         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 212         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 213         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 214 #ifdef CONFIG_IPV6_PRIVACY
 215         .use_tempaddr           = 0,
 216         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 217         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 218         .regen_max_retry        = REGEN_MAX_RETRY,
 219         .max_desync_factor      = MAX_DESYNC_FACTOR,
 220 #endif
 221         .max_addresses          = IPV6_MAX_ADDRESSES,
 222         .accept_ra_defrtr       = 1,
 223         .accept_ra_min_hop_limit= 1,
 224         .accept_ra_pinfo        = 1,
 225 #ifdef CONFIG_IPV6_ROUTER_PREF
 226         .accept_ra_rtr_pref     = 1,
 227         .rtr_probe_interval     = 60 * HZ,
 228 #ifdef CONFIG_IPV6_ROUTE_INFO
 229         .accept_ra_rt_info_max_plen = 0,
 230 #endif
 231 #endif
 232         .proxy_ndp              = 0,
 233         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 234         .disable_ipv6           = 0,
 235         .accept_dad             = 1,
 236 };
 237
 238 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 239 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 240 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 241 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 242 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 243
 244 /* Check if a valid qdisc is available */
 245 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 246 {
 247         return !qdisc_tx_is_noop(dev);
 248 }
 249
 250 /* Check if a route is valid prefix route */
 251 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 252 {
 253         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 254 }
 255
 256 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 257 {
 258         if (del_timer(&ifp->timer))
 259                 __in6_ifa_put(ifp);
 260 }
 261
 262 enum addrconf_timer_t {
 263         AC_NONE,
 264         AC_DAD,
 265         AC_RS,
 266 };
 267
 268 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 269                                enum addrconf_timer_t what,
 270                                unsigned long when)
 271 {
 272         if (!del_timer(&ifp->timer))
 273                 in6_ifa_hold(ifp);
 274
 275         switch (what) {
 276         case AC_DAD:
 277                 ifp->timer.function = addrconf_dad_timer;
 278                 break;
 279         case AC_RS:
 280                 ifp->timer.function = addrconf_rs_timer;
 281                 break;
 282         default:
 283                 break;
 284         }
 285         ifp->timer.expires = jiffies + when;
 286         add_timer(&ifp->timer);
 287 }
 288
 289 static int snmp6_alloc_dev(struct inet6_dev *idev)
 290 {
 291         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 292                           sizeof(struct ipstats_mib),
 293                           __alignof__(struct ipstats_mib)) < 0)
 294                 goto err_ip;
 295         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 296                                         GFP_KERNEL);
 297         if (!idev->stats.icmpv6dev)
 298                 goto err_icmp;
 299         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 300                                            GFP_KERNEL);
 301         if (!idev->stats.icmpv6msgdev)
 302                 goto err_icmpmsg;
 303
 304         return 0;
 305
 306 err_icmpmsg:
 307         kfree(idev->stats.icmpv6dev);
 308 err_icmp:
 309         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 310 err_ip:
 311         return -ENOMEM;
 312 }
 313
 314 static void snmp6_free_dev(struct inet6_dev *idev)
 315 {
 316         kfree(idev->stats.icmpv6msgdev);
 317         kfree(idev->stats.icmpv6dev);
 318         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 319 }
 320
 321 /* Nobody refers to this device, we may destroy it. */
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326
 327         WARN_ON(!list_empty(&idev->addr_list));
 328         WARN_ON(idev->mc_list != NULL);
 329
 330 #ifdef NET_REFCNT_DEBUG
 331         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 332 #endif
 333         dev_put(dev);
 334         if (!idev->dead) {
 335                 pr_warning("Freeing alive inet6 device %p\n", idev);
 336                 return;
 337         }
 338         snmp6_free_dev(idev);
 339         kfree_rcu(idev, rcu);
 340 }
 341
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361
 362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 363         ndev->cnf.mtu6 = dev->mtu;
 364         ndev->cnf.sysctl = NULL;
 365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 366         if (ndev->nd_parms == NULL) {
 367                 kfree(ndev);
 368                 return NULL;
 369         }
 370         if (ndev->cnf.forwarding)
 371                 dev_disable_lro(dev);
 372         /* We refer to the device */
 373         dev_hold(dev);
 374
 375         if (snmp6_alloc_dev(ndev) < 0) {
 376                 ADBG((KERN_WARNING
 377                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 378                         __func__, dev->name));
 379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 380                 dev_put(dev);
 381                 kfree(ndev);
 382                 return NULL;
 383         }
 384
 385         if (snmp6_register_dev(ndev) < 0) {
 386                 ADBG((KERN_WARNING
 387                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 388                         __func__, dev->name));
 389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 390                 ndev->dead = 1;
 391                 in6_dev_finish_destroy(ndev);
 392                 return NULL;
 393         }
 394
 395         /* One reference from device.  We must do this before
 396          * we invoke __ipv6_regen_rndid().
 397          */
 398         in6_dev_hold(ndev);
 399
 400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 401                 ndev->cnf.accept_dad = -1;
 402
 403 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 405                 printk(KERN_INFO
 406                        "%s: Disabled Multicast RS\n",
 407                        dev->name);
 408                 ndev->cnf.rtr_solicits = 0;
 409         }
 410 #endif
 411
 412 #ifdef CONFIG_IPV6_PRIVACY
 413         INIT_LIST_HEAD(&ndev->tempaddr_list);
 414         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 415         if ((dev->flags&IFF_LOOPBACK) ||
 416             dev->type == ARPHRD_TUNNEL ||
 417             dev->type == ARPHRD_TUNNEL6 ||
 418             dev->type == ARPHRD_SIT ||
 419             dev->type == ARPHRD_NONE) {
 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         /* Join all-router multicast group if forwarding is set */
 440         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 441                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 442
 443         return ndev;
 444 }
 445
 446 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 447 {
 448         struct inet6_dev *idev;
 449
 450         ASSERT_RTNL();
 451
 452         idev = __in6_dev_get(dev);
 453         if (!idev) {
 454                 idev = ipv6_add_dev(dev);
 455                 if (!idev)
 456                         return NULL;
 457         }
 458
 459         if (dev->flags&IFF_UP)
 460                 ipv6_mc_up(idev);
 461         return idev;
 462 }
 463
 464 #ifdef CONFIG_SYSCTL
 465 static void dev_forward_change(struct inet6_dev *idev)
 466 {
 467         struct net_device *dev;
 468         struct inet6_ifaddr *ifa;
 469
 470         if (!idev)
 471                 return;
 472         dev = idev->dev;
 473         if (idev->cnf.forwarding)
 474                 dev_disable_lro(dev);
 475         if (dev && (dev->flags & IFF_MULTICAST)) {
 476                 if (idev->cnf.forwarding)
 477                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 478                 else
 479                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 480         }
 481
 482         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 483                 if (ifa->flags&IFA_F_TENTATIVE)
 484                         continue;
 485                 if (idev->cnf.forwarding)
 486                         addrconf_join_anycast(ifa);
 487                 else
 488                         addrconf_leave_anycast(ifa);
 489         }
 490 }
 491
 492
 493 static void addrconf_forward_change(struct net *net, __s32 newf)
 494 {
 495         struct net_device *dev;
 496         struct inet6_dev *idev;
 497
 498         for_each_netdev(net, dev) {
 499                 idev = __in6_dev_get(dev);
 500                 if (idev) {
 501                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 502                         idev->cnf.forwarding = newf;
 503                         if (changed)
 504                                 dev_forward_change(idev);
 505                 }
 506         }
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 510 {
 511         struct net *net;
 512
 513         net = (struct net *)table->extra2;
 514         if (p == &net->ipv6.devconf_dflt->forwarding)
 515                 return 0;
 516
 517         if (!rtnl_trylock()) {
 518                 /* Restore the original values before restarting */
 519                 *p = old;
 520                 return restart_syscall();
 521         }
 522
 523         if (p == &net->ipv6.devconf_all->forwarding) {
 524                 __s32 newf = net->ipv6.devconf_all->forwarding;
 525                 net->ipv6.devconf_dflt->forwarding = newf;
 526                 addrconf_forward_change(net, newf);
 527         } else if ((!*p) ^ (!old))
 528                 dev_forward_change((struct inet6_dev *)table->extra1);
 529         rtnl_unlock();
 530
 531         if (*p)
 532                 rt6_purge_dflt_routers(net);
 533         return 1;
 534 }
 535 #endif
 536
 537 /* Nobody refers to this ifaddr, destroy it */
 538 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 539 {
 540         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 541
 542 #ifdef NET_REFCNT_DEBUG
 543         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 544 #endif
 545
 546         in6_dev_put(ifp->idev);
 547
 548         if (del_timer(&ifp->timer))
 549                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 550
 551         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 552                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 553                 return;
 554         }
 555         dst_release(&ifp->rt->dst);
 556
 557         kfree_rcu(ifp, rcu);
 558 }
 559
 560 static void
 561 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 562 {
 563         struct list_head *p;
 564         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 565
 566         /*
 567          * Each device address list is sorted in order of scope -
 568          * global before linklocal.
 569          */
 570         list_for_each(p, &idev->addr_list) {
 571                 struct inet6_ifaddr *ifa
 572                         = list_entry(p, struct inet6_ifaddr, if_list);
 573                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 574                         break;
 575         }
 576
 577         list_add_tail(&ifp->if_list, p);
 578 }
 579
 580 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 581 {
 582         /*
 583          * We perform the hash function over the last 64 bits of the address
 584          * This will include the IEEE address token on links that support it.
 585          */
 586         return jhash_2words((__force u32)addr->s6_addr32[2],
 587                             (__force u32)addr->s6_addr32[3], 0)
 588                 & (IN6_ADDR_HSIZE - 1);
 589 }
 590
 591 /* On success it returns ifp with increased reference count */
 592
 593 static struct inet6_ifaddr *
 594 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 595               int scope, u32 flags)
 596 {
 597         struct inet6_ifaddr *ifa = NULL;
 598         struct rt6_info *rt;
 599         unsigned int hash;
 600         int err = 0;
 601         int addr_type = ipv6_addr_type(addr);
 602
 603         if (addr_type == IPV6_ADDR_ANY ||
 604             addr_type & IPV6_ADDR_MULTICAST ||
 605             (!(idev->dev->flags & IFF_LOOPBACK) &&
 606              addr_type & IPV6_ADDR_LOOPBACK))
 607                 return ERR_PTR(-EADDRNOTAVAIL);
 608
 609         rcu_read_lock_bh();
 610         if (idev->dead) {
 611                 err = -ENODEV;                  /*XXX*/
 612                 goto out2;
 613         }
 614
 615         if (idev->cnf.disable_ipv6) {
 616                 err = -EACCES;
 617                 goto out2;
 618         }
 619
 620         spin_lock(&addrconf_hash_lock);
 621
 622         /* Ignore adding duplicate addresses on an interface */
 623         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 624                 ADBG(("ipv6_add_addr: already assigned\n"));
 625                 err = -EEXIST;
 626                 goto out;
 627         }
 628
 629         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 630
 631         if (ifa == NULL) {
 632                 ADBG(("ipv6_add_addr: malloc failed\n"));
 633                 err = -ENOBUFS;
 634                 goto out;
 635         }
 636
 637         rt = addrconf_dst_alloc(idev, addr, 0);
 638         if (IS_ERR(rt)) {
 639                 err = PTR_ERR(rt);
 640                 goto out;
 641         }
 642
 643         ipv6_addr_copy(&ifa->addr, addr);
 644
 645         spin_lock_init(&ifa->lock);
 646         spin_lock_init(&ifa->state_lock);
 647         init_timer(&ifa->timer);
 648         INIT_HLIST_NODE(&ifa->addr_lst);
 649         ifa->timer.data = (unsigned long) ifa;
 650         ifa->scope = scope;
 651         ifa->prefix_len = pfxlen;
 652         ifa->flags = flags | IFA_F_TENTATIVE;
 653         ifa->cstamp = ifa->tstamp = jiffies;
 654
 655         ifa->rt = rt;
 656
 657         /*
 658          * part one of RFC 4429, section 3.3
 659          * We should not configure an address as
 660          * optimistic if we do not yet know the link
 661          * layer address of our nexhop router
 662          */
 663
 664         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 665                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 666
 667         ifa->idev = idev;
 668         in6_dev_hold(idev);
 669         /* For caller */
 670         in6_ifa_hold(ifa);
 671
 672         /* Add to big hash table */
 673         hash = ipv6_addr_hash(addr);
 674
 675         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 676         spin_unlock(&addrconf_hash_lock);
 677
 678         write_lock(&idev->lock);
 679         /* Add to inet6_dev unicast addr list. */
 680         ipv6_link_dev_addr(idev, ifa);
 681
 682 #ifdef CONFIG_IPV6_PRIVACY
 683         if (ifa->flags&IFA_F_TEMPORARY) {
 684                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 685                 in6_ifa_hold(ifa);
 686         }
 687 #endif
 688
 689         in6_ifa_hold(ifa);
 690         write_unlock(&idev->lock);
 691 out2:
 692         rcu_read_unlock_bh();
 693
 694         if (likely(err == 0))
 695                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 696         else {
 697                 kfree(ifa);
 698                 ifa = ERR_PTR(err);
 699         }
 700
 701         return ifa;
 702 out:
 703         spin_unlock(&addrconf_hash_lock);
 704         goto out2;
 705 }
 706
 707 /* This function wants to get referenced ifp and releases it before return */
 708
 709 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 710 {
 711         struct inet6_ifaddr *ifa, *ifn;
 712         struct inet6_dev *idev = ifp->idev;
 713         int state;
 714         int deleted = 0, onlink = 0;
 715         unsigned long expires = jiffies;
 716
 717         spin_lock_bh(&ifp->state_lock);
 718         state = ifp->state;
 719         ifp->state = INET6_IFADDR_STATE_DEAD;
 720         spin_unlock_bh(&ifp->state_lock);
 721
 722         if (state == INET6_IFADDR_STATE_DEAD)
 723                 goto out;
 724
 725         spin_lock_bh(&addrconf_hash_lock);
 726         hlist_del_init_rcu(&ifp->addr_lst);
 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->dst);
 819         }
 820
 821         /* clean up prefsrc entries */
 822         rt6_remove_prefsrc(ifp);
 823 out:
 824         in6_ifa_put(ifp);
 825 }
 826
 827 #ifdef CONFIG_IPV6_PRIVACY
 828 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 829 {
 830         struct inet6_dev *idev = ifp->idev;
 831         struct in6_addr addr, *tmpaddr;
 832         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 833         unsigned long regen_advance;
 834         int tmp_plen;
 835         int ret = 0;
 836         int max_addresses;
 837         u32 addr_flags;
 838         unsigned long now = jiffies;
 839
 840         write_lock(&idev->lock);
 841         if (ift) {
 842                 spin_lock_bh(&ift->lock);
 843                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 844                 spin_unlock_bh(&ift->lock);
 845                 tmpaddr = &addr;
 846         } else {
 847                 tmpaddr = NULL;
 848         }
 849 retry:
 850         in6_dev_hold(idev);
 851         if (idev->cnf.use_tempaddr <= 0) {
 852                 write_unlock(&idev->lock);
 853                 printk(KERN_INFO
 854                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 855                 in6_dev_put(idev);
 856                 ret = -1;
 857                 goto out;
 858         }
 859         spin_lock_bh(&ifp->lock);
 860         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 861                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 862                 spin_unlock_bh(&ifp->lock);
 863                 write_unlock(&idev->lock);
 864                 printk(KERN_WARNING
 865                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 866                 in6_dev_put(idev);
 867                 ret = -1;
 868                 goto out;
 869         }
 870         in6_ifa_hold(ifp);
 871         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 872         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 873                 spin_unlock_bh(&ifp->lock);
 874                 write_unlock(&idev->lock);
 875                 printk(KERN_WARNING
 876                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 877                 in6_ifa_put(ifp);
 878                 in6_dev_put(idev);
 879                 ret = -1;
 880                 goto out;
 881         }
 882         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 883         age = (now - ifp->tstamp) / HZ;
 884         tmp_valid_lft = min_t(__u32,
 885                               ifp->valid_lft,
 886                               idev->cnf.temp_valid_lft + age);
 887         tmp_prefered_lft = min_t(__u32,
 888                                  ifp->prefered_lft,
 889                                  idev->cnf.temp_prefered_lft + age -
 890                                  idev->cnf.max_desync_factor);
 891         tmp_plen = ifp->prefix_len;
 892         max_addresses = idev->cnf.max_addresses;
 893         tmp_tstamp = ifp->tstamp;
 894         spin_unlock_bh(&ifp->lock);
 895
 896         regen_advance = idev->cnf.regen_max_retry *
 897                         idev->cnf.dad_transmits *
 898                         idev->nd_parms->retrans_time / HZ;
 899         write_unlock(&idev->lock);
 900
 901         /* A temporary address is created only if this calculated Preferred
 902          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 903          * an implementation must not create a temporary address with a zero
 904          * Preferred Lifetime.
 905          * Use age calculation as in addrconf_verify to avoid unnecessary
 906          * temporary addresses being generated.
 907          */
 908         age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
 909         if (tmp_prefered_lft <= regen_advance + age) {
 910                 in6_ifa_put(ifp);
 911                 in6_dev_put(idev);
 912                 ret = -1;
 913                 goto out;
 914         }
 915
 916         addr_flags = IFA_F_TEMPORARY;
 917         /* set in addrconf_prefix_rcv() */
 918         if (ifp->flags & IFA_F_OPTIMISTIC)
 919                 addr_flags |= IFA_F_OPTIMISTIC;
 920
 921         ift = ipv6_add_addr(idev, &addr, tmp_plen,
 922                             ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 923                             addr_flags);
 924         if (IS_ERR(ift)) {
 925                 in6_ifa_put(ifp);
 926                 in6_dev_put(idev);
 927                 printk(KERN_INFO
 928                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 929                 tmpaddr = &addr;
 930                 write_lock(&idev->lock);
 931                 goto retry;
 932         }
 933
 934         spin_lock_bh(&ift->lock);
 935         ift->ifpub = ifp;
 936         ift->valid_lft = tmp_valid_lft;
 937         ift->prefered_lft = tmp_prefered_lft;
 938         ift->cstamp = now;
 939         ift->tstamp = tmp_tstamp;
 940         spin_unlock_bh(&ift->lock);
 941
 942         addrconf_dad_start(ift, 0);
 943         in6_ifa_put(ift);
 944         in6_dev_put(idev);
 945 out:
 946         return ret;
 947 }
 948 #endif
 949
 950 /*
 951  *      Choose an appropriate source address (RFC3484)
 952  */
 953 enum {
 954         IPV6_SADDR_RULE_INIT = 0,
 955         IPV6_SADDR_RULE_LOCAL,
 956         IPV6_SADDR_RULE_SCOPE,
 957         IPV6_SADDR_RULE_PREFERRED,
 958 #ifdef CONFIG_IPV6_MIP6
 959         IPV6_SADDR_RULE_HOA,
 960 #endif
 961         IPV6_SADDR_RULE_OIF,
 962         IPV6_SADDR_RULE_LABEL,
 963 #ifdef CONFIG_IPV6_PRIVACY
 964         IPV6_SADDR_RULE_PRIVACY,
 965 #endif
 966         IPV6_SADDR_RULE_ORCHID,
 967         IPV6_SADDR_RULE_PREFIX,
 968         IPV6_SADDR_RULE_MAX
 969 };
 970
 971 struct ipv6_saddr_score {
 972         int                     rule;
 973         int                     addr_type;
 974         struct inet6_ifaddr     *ifa;
 975         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 976         int                     scopedist;
 977         int                     matchlen;
 978 };
 979
 980 struct ipv6_saddr_dst {
 981         const struct in6_addr *addr;
 982         int ifindex;
 983         int scope;
 984         int label;
 985         unsigned int prefs;
 986 };
 987
 988 static inline int ipv6_saddr_preferred(int type)
 989 {
 990         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 991                 return 1;
 992         return 0;
 993 }
 994
 995 static int ipv6_get_saddr_eval(struct net *net,
 996                                struct ipv6_saddr_score *score,
 997                                struct ipv6_saddr_dst *dst,
 998                                int i)
 999 {
1000         int ret;
1001
1002         if (i <= score->rule) {
1003                 switch (i) {
1004                 case IPV6_SADDR_RULE_SCOPE:
1005                         ret = score->scopedist;
1006                         break;
1007                 case IPV6_SADDR_RULE_PREFIX:
1008                         ret = score->matchlen;
1009                         break;
1010                 default:
1011                         ret = !!test_bit(i, score->scorebits);
1012                 }
1013                 goto out;
1014         }
1015
1016         switch (i) {
1017         case IPV6_SADDR_RULE_INIT:
1018                 /* Rule 0: remember if hiscore is not ready yet */
1019                 ret = !!score->ifa;
1020                 break;
1021         case IPV6_SADDR_RULE_LOCAL:
1022                 /* Rule 1: Prefer same address */
1023                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1024                 break;
1025         case IPV6_SADDR_RULE_SCOPE:
1026                 /* Rule 2: Prefer appropriate scope
1027                  *
1028                  *      ret
1029                  *       ^
1030                  *    -1 |  d 15
1031                  *    ---+--+-+---> scope
1032                  *       |
1033                  *       |             d is scope of the destination.
1034                  *  B-d  |  \
1035                  *       |   \      <- smaller scope is better if
1036                  *  B-15 |    \        if scope is enough for destinaion.
1037                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1038                  * d-C-1 | /
1039                  *       |/         <- greater is better
1040                  *   -C  /             if scope is not enough for destination.
1041                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1042                  *
1043                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1044                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1045                  * Assume B = 0 and we get C > 29.
1046                  */
1047                 ret = __ipv6_addr_src_scope(score->addr_type);
1048                 if (ret >= dst->scope)
1049                         ret = -ret;
1050                 else
1051                         ret -= 128;     /* 30 is enough */
1052                 score->scopedist = ret;
1053                 break;
1054         case IPV6_SADDR_RULE_PREFERRED:
1055                 /* Rule 3: Avoid deprecated and optimistic addresses */
1056                 ret = ipv6_saddr_preferred(score->addr_type) ||
1057                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1058                 break;
1059 #ifdef CONFIG_IPV6_MIP6
1060         case IPV6_SADDR_RULE_HOA:
1061             {
1062                 /* Rule 4: Prefer home address */
1063                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1064                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1065                 break;
1066             }
1067 #endif
1068         case IPV6_SADDR_RULE_OIF:
1069                 /* Rule 5: Prefer outgoing interface */
1070                 ret = (!dst->ifindex ||
1071                        dst->ifindex == score->ifa->idev->dev->ifindex);
1072                 break;
1073         case IPV6_SADDR_RULE_LABEL:
1074                 /* Rule 6: Prefer matching label */
1075                 ret = ipv6_addr_label(net,
1076                                       &score->ifa->addr, score->addr_type,
1077                                       score->ifa->idev->dev->ifindex) == dst->label;
1078                 break;
1079 #ifdef CONFIG_IPV6_PRIVACY
1080         case IPV6_SADDR_RULE_PRIVACY:
1081             {
1082                 /* Rule 7: Prefer public address
1083                  * Note: prefer temporary address if use_tempaddr >= 2
1084                  */
1085                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1086                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1087                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1088                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1089                 break;
1090             }
1091 #endif
1092         case IPV6_SADDR_RULE_ORCHID:
1093                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1094                  *          non-ORCHID vs non-ORCHID
1095                  */
1096                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1097                         ipv6_addr_orchid(dst->addr));
1098                 break;
1099         case IPV6_SADDR_RULE_PREFIX:
1100                 /* Rule 8: Use longest matching prefix */
1101                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1102                                                        dst->addr);
1103                 break;
1104         default:
1105                 ret = 0;
1106         }
1107
1108         if (ret)
1109                 __set_bit(i, score->scorebits);
1110         score->rule = i;
1111 out:
1112         return ret;
1113 }
1114
1115 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1116                        const struct in6_addr *daddr, unsigned int prefs,
1117                        struct in6_addr *saddr)
1118 {
1119         struct ipv6_saddr_score scores[2],
1120                                 *score = &scores[0], *hiscore = &scores[1];
1121         struct ipv6_saddr_dst dst;
1122         struct net_device *dev;
1123         int dst_type;
1124
1125         dst_type = __ipv6_addr_type(daddr);
1126         dst.addr = daddr;
1127         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1128         dst.scope = __ipv6_addr_src_scope(dst_type);
1129         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1130         dst.prefs = prefs;
1131
1132         hiscore->rule = -1;
1133         hiscore->ifa = NULL;
1134
1135         rcu_read_lock();
1136
1137         for_each_netdev_rcu(net, dev) {
1138                 struct inet6_dev *idev;
1139
1140                 /* Candidate Source Address (section 4)
1141                  *  - multicast and link-local destination address,
1142                  *    the set of candidate source address MUST only
1143                  *    include addresses assigned to interfaces
1144                  *    belonging to the same link as the outgoing
1145                  *    interface.
1146                  * (- For site-local destination addresses, the
1147                  *    set of candidate source addresses MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same site as the outgoing
1150                  *    interface.)
1151                  */
1152                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1153                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1154                     dst.ifindex && dev->ifindex != dst.ifindex)
1155                         continue;
1156
1157                 idev = __in6_dev_get(dev);
1158                 if (!idev)
1159                         continue;
1160
1161                 read_lock_bh(&idev->lock);
1162                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1163                         int i;
1164
1165                         /*
1166                          * - Tentative Address (RFC2462 section 5.4)
1167                          *  - A tentative address is not considered
1168                          *    "assigned to an interface" in the traditional
1169                          *    sense, unless it is also flagged as optimistic.
1170                          * - Candidate Source Address (section 4)
1171                          *  - In any case, anycast addresses, multicast
1172                          *    addresses, and the unspecified address MUST
1173                          *    NOT be included in a candidate set.
1174                          */
1175                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1176                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1177                                 continue;
1178
1179                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1180
1181                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1182                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1183                                 LIMIT_NETDEBUG(KERN_DEBUG
1184                                                "ADDRCONF: unspecified / multicast address "
1185                                                "assigned as unicast address on %s",
1186                                                dev->name);
1187                                 continue;
1188                         }
1189
1190                         score->rule = -1;
1191                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1192
1193                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1194                                 int minihiscore, miniscore;
1195
1196                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1197                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1198
1199                                 if (minihiscore > miniscore) {
1200                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1201                                             score->scopedist > 0) {
1202                                                 /*
1203                                                  * special case:
1204                                                  * each remaining entry
1205                                                  * has too small (not enough)
1206                                                  * scope, because ifa entries
1207                                                  * are sorted by their scope
1208                                                  * values.
1209                                                  */
1210                                                 goto try_nextdev;
1211                                         }
1212                                         break;
1213                                 } else if (minihiscore < miniscore) {
1214                                         if (hiscore->ifa)
1215                                                 in6_ifa_put(hiscore->ifa);
1216
1217                                         in6_ifa_hold(score->ifa);
1218
1219                                         swap(hiscore, score);
1220
1221                                         /* restore our iterator */
1222                                         score->ifa = hiscore->ifa;
1223
1224                                         break;
1225                                 }
1226                         }
1227                 }
1228 try_nextdev:
1229                 read_unlock_bh(&idev->lock);
1230         }
1231         rcu_read_unlock();
1232
1233         if (!hiscore->ifa)
1234                 return -EADDRNOTAVAIL;
1235
1236         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1237         in6_ifa_put(hiscore->ifa);
1238         return 0;
1239 }
1240 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1241
1242 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1243                       unsigned char banned_flags)
1244 {
1245         struct inet6_ifaddr *ifp;
1246         int err = -EADDRNOTAVAIL;
1247
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         return err;
1257 }
1258
1259 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1260                     unsigned char banned_flags)
1261 {
1262         struct inet6_dev *idev;
1263         int err = -EADDRNOTAVAIL;
1264
1265         rcu_read_lock();
1266         idev = __in6_dev_get(dev);
1267         if (idev) {
1268                 read_lock_bh(&idev->lock);
1269                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1270                 read_unlock_bh(&idev->lock);
1271         }
1272         rcu_read_unlock();
1273         return err;
1274 }
1275
1276 static int ipv6_count_addresses(struct inet6_dev *idev)
1277 {
1278         int cnt = 0;
1279         struct inet6_ifaddr *ifp;
1280
1281         read_lock_bh(&idev->lock);
1282         list_for_each_entry(ifp, &idev->addr_list, if_list)
1283                 cnt++;
1284         read_unlock_bh(&idev->lock);
1285         return cnt;
1286 }
1287
1288 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1289                   struct net_device *dev, int strict)
1290 {
1291         struct inet6_ifaddr *ifp;
1292         struct hlist_node *node;
1293         unsigned int hash = ipv6_addr_hash(addr);
1294
1295         rcu_read_lock_bh();
1296         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1297                 if (!net_eq(dev_net(ifp->idev->dev), net))
1298                         continue;
1299                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1300                     !(ifp->flags&IFA_F_TENTATIVE) &&
1301                     (dev == NULL || ifp->idev->dev == dev ||
1302                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1303                         rcu_read_unlock_bh();
1304                         return 1;
1305                 }
1306         }
1307
1308         rcu_read_unlock_bh();
1309         return 0;
1310 }
1311 EXPORT_SYMBOL(ipv6_chk_addr);
1312
1313 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1314                                struct net_device *dev)
1315 {
1316         unsigned int hash = ipv6_addr_hash(addr);
1317         struct inet6_ifaddr *ifp;
1318         struct hlist_node *node;
1319
1320         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1321                 if (!net_eq(dev_net(ifp->idev->dev), net))
1322                         continue;
1323                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1324                         if (dev == NULL || ifp->idev->dev == dev)
1325                                 return true;
1326                 }
1327         }
1328         return false;
1329 }
1330
1331 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1332 {
1333         struct inet6_dev *idev;
1334         struct inet6_ifaddr *ifa;
1335         int     onlink;
1336
1337         onlink = 0;
1338         rcu_read_lock();
1339         idev = __in6_dev_get(dev);
1340         if (idev) {
1341                 read_lock_bh(&idev->lock);
1342                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1343                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1344                                                    ifa->prefix_len);
1345                         if (onlink)
1346                                 break;
1347                 }
1348                 read_unlock_bh(&idev->lock);
1349         }
1350         rcu_read_unlock();
1351         return onlink;
1352 }
1353
1354 EXPORT_SYMBOL(ipv6_chk_prefix);
1355
1356 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1357                                      struct net_device *dev, int strict)
1358 {
1359         struct inet6_ifaddr *ifp, *result = NULL;
1360         unsigned int hash = ipv6_addr_hash(addr);
1361         struct hlist_node *node;
1362
1363         rcu_read_lock_bh();
1364         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1365                 if (!net_eq(dev_net(ifp->idev->dev), net))
1366                         continue;
1367                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1368                         if (dev == NULL || ifp->idev->dev == dev ||
1369                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1370                                 result = ifp;
1371                                 in6_ifa_hold(ifp);
1372                                 break;
1373                         }
1374                 }
1375         }
1376         rcu_read_unlock_bh();
1377
1378         return result;
1379 }
1380
1381 /* Gets referenced address, destroys ifaddr */
1382
1383 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1384 {
1385         if (ifp->flags&IFA_F_PERMANENT) {
1386                 spin_lock_bh(&ifp->lock);
1387                 addrconf_del_timer(ifp);
1388                 ifp->flags |= IFA_F_TENTATIVE;
1389                 if (dad_failed)
1390                         ifp->flags |= IFA_F_DADFAILED;
1391                 spin_unlock_bh(&ifp->lock);
1392                 if (dad_failed)
1393                         ipv6_ifa_notify(0, ifp);
1394                 in6_ifa_put(ifp);
1395 #ifdef CONFIG_IPV6_PRIVACY
1396         } else if (ifp->flags&IFA_F_TEMPORARY) {
1397                 struct inet6_ifaddr *ifpub;
1398                 spin_lock_bh(&ifp->lock);
1399                 ifpub = ifp->ifpub;
1400                 if (ifpub) {
1401                         in6_ifa_hold(ifpub);
1402                         spin_unlock_bh(&ifp->lock);
1403                         ipv6_create_tempaddr(ifpub, ifp);
1404                         in6_ifa_put(ifpub);
1405                 } else {
1406                         spin_unlock_bh(&ifp->lock);
1407                 }
1408                 ipv6_del_addr(ifp);
1409 #endif
1410         } else
1411                 ipv6_del_addr(ifp);
1412 }
1413
1414 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1415 {
1416         int err = -ENOENT;
1417
1418         spin_lock(&ifp->state_lock);
1419         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1420                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1421                 err = 0;
1422         }
1423         spin_unlock(&ifp->state_lock);
1424
1425         return err;
1426 }
1427
1428 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1429 {
1430         struct inet6_dev *idev = ifp->idev;
1431
1432         if (addrconf_dad_end(ifp)) {
1433                 in6_ifa_put(ifp);
1434                 return;
1435         }
1436
1437         if (net_ratelimit())
1438                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1439                         ifp->idev->dev->name, &ifp->addr);
1440
1441         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1442                 struct in6_addr addr;
1443
1444                 addr.s6_addr32[0] = htonl(0xfe800000);
1445                 addr.s6_addr32[1] = 0;
1446
1447                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1448                     ipv6_addr_equal(&ifp->addr, &addr)) {
1449                         /* DAD failed for link-local based on MAC address */
1450                         idev->cnf.disable_ipv6 = 1;
1451
1452                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1453                                 ifp->idev->dev->name);
1454                 }
1455         }
1456
1457         addrconf_dad_stop(ifp, 1);
1458 }
1459
1460 /* Join to solicited addr multicast group. */
1461
1462 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1463 {
1464         struct in6_addr maddr;
1465
1466         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1467                 return;
1468
1469         addrconf_addr_solict_mult(addr, &maddr);
1470         ipv6_dev_mc_inc(dev, &maddr);
1471 }
1472
1473 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1474 {
1475         struct in6_addr maddr;
1476
1477         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1478                 return;
1479
1480         addrconf_addr_solict_mult(addr, &maddr);
1481         __ipv6_dev_mc_dec(idev, &maddr);
1482 }
1483
1484 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1485 {
1486         struct in6_addr addr;
1487         if (ifp->prefix_len == 127) /* RFC 6164 */
1488                 return;
1489         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490         if (ipv6_addr_any(&addr))
1491                 return;
1492         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1493 }
1494
1495 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1496 {
1497         struct in6_addr addr;
1498         if (ifp->prefix_len == 127) /* RFC 6164 */
1499                 return;
1500         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1501         if (ipv6_addr_any(&addr))
1502                 return;
1503         __ipv6_dev_ac_dec(ifp->idev, &addr);
1504 }
1505
1506 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1507 {
1508         if (dev->addr_len != ETH_ALEN)
1509                 return -1;
1510         memcpy(eui, dev->dev_addr, 3);
1511         memcpy(eui + 5, dev->dev_addr + 3, 3);
1512
1513         /*
1514          * The zSeries OSA network cards can be shared among various
1515          * OS instances, but the OSA cards have only one MAC address.
1516          * This leads to duplicate address conflicts in conjunction
1517          * with IPv6 if more than one instance uses the same card.
1518          *
1519          * The driver for these cards can deliver a unique 16-bit
1520          * identifier for each instance sharing the same card.  It is
1521          * placed instead of 0xFFFE in the interface identifier.  The
1522          * "u" bit of the interface identifier is not inverted in this
1523          * case.  Hence the resulting interface identifier has local
1524          * scope according to RFC2373.
1525          */
1526         if (dev->dev_id) {
1527                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1528                 eui[4] = dev->dev_id & 0xFF;
1529         } else {
1530                 eui[3] = 0xFF;
1531                 eui[4] = 0xFE;
1532                 eui[0] ^= 2;
1533         }
1534         return 0;
1535 }
1536
1537 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1538 {
1539         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1540         if (dev->addr_len != ARCNET_ALEN)
1541                 return -1;
1542         memset(eui, 0, 7);
1543         eui[7] = *(u8*)dev->dev_addr;
1544         return 0;
1545 }
1546
1547 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1548 {
1549         if (dev->addr_len != INFINIBAND_ALEN)
1550                 return -1;
1551         memcpy(eui, dev->dev_addr + 12, 8);
1552         eui[0] |= 2;
1553         return 0;
1554 }
1555
1556 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1557 {
1558         if (addr == 0)
1559                 return -1;
1560         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1561                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1562                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1563                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1564                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1565                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1566         eui[1] = 0;
1567         eui[2] = 0x5E;
1568         eui[3] = 0xFE;
1569         memcpy(eui + 4, &addr, 4);
1570         return 0;
1571 }
1572
1573 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1574 {
1575         if (dev->priv_flags & IFF_ISATAP)
1576                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1577         return -1;
1578 }
1579
1580 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1581 {
1582         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1583 }
1584
1585 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1586 {
1587         switch (dev->type) {
1588         case ARPHRD_ETHER:
1589         case ARPHRD_FDDI:
1590         case ARPHRD_IEEE802_TR:
1591                 return addrconf_ifid_eui48(eui, dev);
1592         case ARPHRD_ARCNET:
1593                 return addrconf_ifid_arcnet(eui, dev);
1594         case ARPHRD_INFINIBAND:
1595                 return addrconf_ifid_infiniband(eui, dev);
1596         case ARPHRD_SIT:
1597                 return addrconf_ifid_sit(eui, dev);
1598         case ARPHRD_IPGRE:
1599                 return addrconf_ifid_gre(eui, dev);
1600         }
1601         return -1;
1602 }
1603
1604 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1605 {
1606         int err = -1;
1607         struct inet6_ifaddr *ifp;
1608
1609         read_lock_bh(&idev->lock);
1610         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1611                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1612                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1613                         err = 0;
1614                         break;
1615                 }
1616         }
1617         read_unlock_bh(&idev->lock);
1618         return err;
1619 }
1620
1621 #ifdef CONFIG_IPV6_PRIVACY
1622 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1623 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1624 {
1625 regen:
1626         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1627         idev->rndid[0] &= ~0x02;
1628
1629         /*
1630          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1631          * check if generated address is not inappropriate
1632          *
1633          *  - Reserved subnet anycast (RFC 2526)
1634          *      11111101 11....11 1xxxxxxx
1635          *  - ISATAP (RFC4214) 6.1
1636          *      00-00-5E-FE-xx-xx-xx-xx
1637          *  - value 0
1638          *  - XXX: already assigned to an address on the device
1639          */
1640         if (idev->rndid[0] == 0xfd &&
1641             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1642             (idev->rndid[7]&0x80))
1643                 goto regen;
1644         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1645                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1646                         goto regen;
1647                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1648                         goto regen;
1649         }
1650
1651         return 0;
1652 }
1653
1654 static void ipv6_regen_rndid(unsigned long data)
1655 {
1656         struct inet6_dev *idev = (struct inet6_dev *) data;
1657         unsigned long expires;
1658
1659         rcu_read_lock_bh();
1660         write_lock_bh(&idev->lock);
1661
1662         if (idev->dead)
1663                 goto out;
1664
1665         if (__ipv6_regen_rndid(idev) < 0)
1666                 goto out;
1667
1668         expires = jiffies +
1669                 idev->cnf.temp_prefered_lft * HZ -
1670                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1671                 idev->cnf.max_desync_factor * HZ;
1672         if (time_before(expires, jiffies)) {
1673                 printk(KERN_WARNING
1674                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1675                         idev->dev->name);
1676                 goto out;
1677         }
1678
1679         if (!mod_timer(&idev->regen_timer, expires))
1680                 in6_dev_hold(idev);
1681
1682 out:
1683         write_unlock_bh(&idev->lock);
1684         rcu_read_unlock_bh();
1685         in6_dev_put(idev);
1686 }
1687
1688 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1689         int ret = 0;
1690
1691         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1692                 ret = __ipv6_regen_rndid(idev);
1693         return ret;
1694 }
1695 #endif
1696
1697 /*
1698  *      Add prefix route.
1699  */
1700
1701 static void
1702 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1703                       unsigned long expires, u32 flags)
1704 {
1705         struct fib6_config cfg = {
1706                 .fc_table = RT6_TABLE_PREFIX,
1707                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1708                 .fc_ifindex = dev->ifindex,
1709                 .fc_expires = expires,
1710                 .fc_dst_len = plen,
1711                 .fc_flags = RTF_UP | flags,
1712                 .fc_nlinfo.nl_net = dev_net(dev),
1713                 .fc_protocol = RTPROT_KERNEL,
1714         };
1715
1716         ipv6_addr_copy(&cfg.fc_dst, pfx);
1717
1718         /* Prevent useless cloning on PtP SIT.
1719            This thing is done here expecting that the whole
1720            class of non-broadcast devices need not cloning.
1721          */
1722 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1723         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1724                 cfg.fc_flags |= RTF_NONEXTHOP;
1725 #endif
1726
1727         ip6_route_add(&cfg);
1728 }
1729
1730
1731 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1732                                                   int plen,
1733                                                   const struct net_device *dev,
1734                                                   u32 flags, u32 noflags)
1735 {
1736         struct fib6_node *fn;
1737         struct rt6_info *rt = NULL;
1738         struct fib6_table *table;
1739
1740         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1741         if (table == NULL)
1742                 return NULL;
1743
1744         write_lock_bh(&table->tb6_lock);
1745         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1746         if (!fn)
1747                 goto out;
1748         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1749                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1750                         continue;
1751                 if ((rt->rt6i_flags & flags) != flags)
1752                         continue;
1753                 if ((rt->rt6i_flags & noflags) != 0)
1754                         continue;
1755                 dst_hold(&rt->dst);
1756                 break;
1757         }
1758 out:
1759         write_unlock_bh(&table->tb6_lock);
1760         return rt;
1761 }
1762
1763
1764 /* Create "default" multicast route to the interface */
1765
1766 static void addrconf_add_mroute(struct net_device *dev)
1767 {
1768         struct fib6_config cfg = {
1769                 .fc_table = RT6_TABLE_LOCAL,
1770                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1771                 .fc_ifindex = dev->ifindex,
1772                 .fc_dst_len = 8,
1773                 .fc_flags = RTF_UP,
1774                 .fc_nlinfo.nl_net = dev_net(dev),
1775         };
1776
1777         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1778
1779         ip6_route_add(&cfg);
1780 }
1781
1782 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1783 static void sit_route_add(struct net_device *dev)
1784 {
1785         struct fib6_config cfg = {
1786                 .fc_table = RT6_TABLE_MAIN,
1787                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1788                 .fc_ifindex = dev->ifindex,
1789                 .fc_dst_len = 96,
1790                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1791                 .fc_nlinfo.nl_net = dev_net(dev),
1792         };
1793
1794         /* prefix length - 96 bits "::d.d.d.d" */
1795         ip6_route_add(&cfg);
1796 }
1797 #endif
1798
1799 static void addrconf_add_lroute(struct net_device *dev)
1800 {
1801         struct in6_addr addr;
1802
1803         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1804         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1805 }
1806
1807 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1808 {
1809         struct inet6_dev *idev;
1810
1811         ASSERT_RTNL();
1812
1813         idev = ipv6_find_idev(dev);
1814         if (!idev)
1815                 return ERR_PTR(-ENOBUFS);
1816
1817         if (idev->cnf.disable_ipv6)
1818                 return ERR_PTR(-EACCES);
1819
1820         /* Add default multicast route */
1821         if (!(dev->flags & IFF_LOOPBACK))
1822                 addrconf_add_mroute(dev);
1823
1824         /* Add link local route */
1825         addrconf_add_lroute(dev);
1826         return idev;
1827 }
1828
1829 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1830 {
1831         struct prefix_info *pinfo;
1832         __u32 valid_lft;
1833         __u32 prefered_lft;
1834         int addr_type;
1835         struct inet6_dev *in6_dev;
1836         struct net *net = dev_net(dev);
1837
1838         pinfo = (struct prefix_info *) opt;
1839
1840         if (len < sizeof(struct prefix_info)) {
1841                 ADBG(("addrconf: prefix option too short\n"));
1842                 return;
1843         }
1844
1845         /*
1846          *      Validation checks ([ADDRCONF], page 19)
1847          */
1848
1849         addr_type = ipv6_addr_type(&pinfo->prefix);
1850
1851         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1852                 return;
1853
1854         valid_lft = ntohl(pinfo->valid);
1855         prefered_lft = ntohl(pinfo->prefered);
1856
1857         if (prefered_lft > valid_lft) {
1858                 if (net_ratelimit())
1859                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1860                 return;
1861         }
1862
1863         in6_dev = in6_dev_get(dev);
1864
1865         if (in6_dev == NULL) {
1866                 if (net_ratelimit())
1867                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1868                 return;
1869         }
1870
1871         /*
1872          *      Two things going on here:
1873          *      1) Add routes for on-link prefixes
1874          *      2) Configure prefixes with the auto flag set
1875          */
1876
1877         if (pinfo->onlink) {
1878                 struct rt6_info *rt;
1879                 unsigned long rt_expires;
1880
1881                 /* Avoid arithmetic overflow. Really, we could
1882                  * save rt_expires in seconds, likely valid_lft,
1883                  * but it would require division in fib gc, that it
1884                  * not good.
1885                  */
1886                 if (HZ > USER_HZ)
1887                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1888                 else
1889                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1890
1891                 if (addrconf_finite_timeout(rt_expires))
1892                         rt_expires *= HZ;
1893
1894                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1895                                                pinfo->prefix_len,
1896                                                dev,
1897                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1898                                                RTF_GATEWAY | RTF_DEFAULT);
1899
1900                 if (rt) {
1901                         /* Autoconf prefix route */
1902                         if (valid_lft == 0) {
1903                                 ip6_del_rt(rt);
1904                                 rt = NULL;
1905                         } else if (addrconf_finite_timeout(rt_expires)) {
1906                                 /* not infinity */
1907                                 rt->rt6i_expires = jiffies + rt_expires;
1908                                 rt->rt6i_flags |= RTF_EXPIRES;
1909                         } else {
1910                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1911                                 rt->rt6i_expires = 0;
1912                         }
1913                 } else if (valid_lft) {
1914                         clock_t expires = 0;
1915                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1916                         if (addrconf_finite_timeout(rt_expires)) {
1917                                 /* not infinity */
1918                                 flags |= RTF_EXPIRES;
1919                                 expires = jiffies_to_clock_t(rt_expires);
1920                         }
1921                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1922                                               dev, expires, flags);
1923                 }
1924                 if (rt)
1925                         dst_release(&rt->dst);
1926         }
1927
1928         /* Try to figure out our local address for this prefix */
1929
1930         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1931                 struct inet6_ifaddr * ifp;
1932                 struct in6_addr addr;
1933                 int create = 0, update_lft = 0;
1934
1935                 if (pinfo->prefix_len == 64) {
1936                         memcpy(&addr, &pinfo->prefix, 8);
1937                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1938                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1939                                 in6_dev_put(in6_dev);
1940                                 return;
1941                         }
1942                         goto ok;
1943                 }
1944                 if (net_ratelimit())
1945                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1946                                pinfo->prefix_len);
1947                 in6_dev_put(in6_dev);
1948                 return;
1949
1950 ok:
1951
1952                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1953
1954                 if (ifp == NULL && valid_lft) {
1955                         int max_addresses = in6_dev->cnf.max_addresses;
1956                         u32 addr_flags = 0;
1957
1958 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1959                         if (in6_dev->cnf.optimistic_dad &&
1960                             !net->ipv6.devconf_all->forwarding)
1961                                 addr_flags = IFA_F_OPTIMISTIC;
1962 #endif
1963
1964                         /* Do not allow to create too much of autoconfigured
1965                          * addresses; this would be too easy way to crash kernel.
1966                          */
1967                         if (!max_addresses ||
1968                             ipv6_count_addresses(in6_dev) < max_addresses)
1969                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1970                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1971                                                     addr_flags);
1972
1973                         if (!ifp || IS_ERR(ifp)) {
1974                                 in6_dev_put(in6_dev);
1975                                 return;
1976                         }
1977
1978                         update_lft = create = 1;
1979                         ifp->cstamp = jiffies;
1980                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1981                 }
1982
1983                 if (ifp) {
1984                         int flags;
1985                         unsigned long now;
1986 #ifdef CONFIG_IPV6_PRIVACY
1987                         struct inet6_ifaddr *ift;
1988 #endif
1989                         u32 stored_lft;
1990
1991                         /* update lifetime (RFC2462 5.5.3 e) */
1992                         spin_lock(&ifp->lock);
1993                         now = jiffies;
1994                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1995                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1996                         else
1997                                 stored_lft = 0;
1998                         if (!update_lft && stored_lft) {
1999                                 if (valid_lft > MIN_VALID_LIFETIME ||
2000                                     valid_lft > stored_lft)
2001                                         update_lft = 1;
2002                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2003                                         /* valid_lft <= stored_lft is always true */
2004                                         /*
2005                                          * RFC 4862 Section 5.5.3e:
2006                                          * "Note that the preferred lifetime of
2007                                          *  the corresponding address is always
2008                                          *  reset to the Preferred Lifetime in
2009                                          *  the received Prefix Information
2010                                          *  option, regardless of whether the
2011                                          *  valid lifetime is also reset or
2012                                          *  ignored."
2013                                          *
2014                                          *  So if the preferred lifetime in
2015                                          *  this advertisement is different
2016                                          *  than what we have stored, but the
2017                                          *  valid lifetime is invalid, just
2018                                          *  reset prefered_lft.
2019                                          *
2020                                          *  We must set the valid lifetime
2021                                          *  to the stored lifetime since we'll
2022                                          *  be updating the timestamp below,
2023                                          *  else we'll set it back to the
2024                                          *  minimum.
2025                                          */
2026                                         if (prefered_lft != ifp->prefered_lft) {
2027                                                 valid_lft = stored_lft;
2028                                                 update_lft = 1;
2029                                         }
2030                                 } else {
2031                                         valid_lft = MIN_VALID_LIFETIME;
2032                                         if (valid_lft < prefered_lft)
2033                                                 prefered_lft = valid_lft;
2034                                         update_lft = 1;
2035                                 }
2036                         }
2037
2038                         if (update_lft) {
2039                                 ifp->valid_lft = valid_lft;
2040                                 ifp->prefered_lft = prefered_lft;
2041                                 ifp->tstamp = now;
2042                                 flags = ifp->flags;
2043                                 ifp->flags &= ~IFA_F_DEPRECATED;
2044                                 spin_unlock(&ifp->lock);
2045
2046                                 if (!(flags&IFA_F_TENTATIVE))
2047                                         ipv6_ifa_notify(0, ifp);
2048                         } else
2049                                 spin_unlock(&ifp->lock);
2050
2051 #ifdef CONFIG_IPV6_PRIVACY
2052                         read_lock_bh(&in6_dev->lock);
2053                         /* update all temporary addresses in the list */
2054                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2055                                             tmp_list) {
2056                                 int age, max_valid, max_prefered;
2057
2058                                 if (ifp != ift->ifpub)
2059                                         continue;
2060
2061                                 /*
2062                                  * RFC 4941 section 3.3:
2063                                  * If a received option will extend the lifetime
2064                                  * of a public address, the lifetimes of
2065                                  * temporary addresses should be extended,
2066                                  * subject to the overall constraint that no
2067                                  * temporary addresses should ever remain
2068                                  * "valid" or "preferred" for a time longer than
2069                                  * (TEMP_VALID_LIFETIME) or
2070                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2071                                  * respectively.
2072                                  */
2073                                 age = (now - ift->cstamp) / HZ;
2074                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2075                                 if (max_valid < 0)
2076                                         max_valid = 0;
2077
2078                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2079                                                in6_dev->cnf.max_desync_factor -
2080                                                age;
2081                                 if (max_prefered < 0)
2082                                         max_prefered = 0;
2083
2084                                 if (valid_lft > max_valid)
2085                                         valid_lft = max_valid;
2086
2087                                 if (prefered_lft > max_prefered)
2088                                         prefered_lft = max_prefered;
2089
2090                                 spin_lock(&ift->lock);
2091                                 flags = ift->flags;
2092                                 ift->valid_lft = valid_lft;
2093                                 ift->prefered_lft = prefered_lft;
2094                                 ift->tstamp = now;
2095                                 if (prefered_lft > 0)
2096                                         ift->flags &= ~IFA_F_DEPRECATED;
2097
2098                                 spin_unlock(&ift->lock);
2099                                 if (!(flags&IFA_F_TENTATIVE))
2100                                         ipv6_ifa_notify(0, ift);
2101                         }
2102
2103                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2104                                 /*
2105                                  * When a new public address is created as
2106                                  * described in [ADDRCONF], also create a new
2107                                  * temporary address. Also create a temporary
2108                                  * address if it's enabled but no temporary
2109                                  * address currently exists.
2110                                  */
2111                                 read_unlock_bh(&in6_dev->lock);
2112                                 ipv6_create_tempaddr(ifp, NULL);
2113                         } else {
2114                                 read_unlock_bh(&in6_dev->lock);
2115                         }
2116 #endif
2117                         in6_ifa_put(ifp);
2118                         addrconf_verify(0);
2119                 }
2120         }
2121         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2122         in6_dev_put(in6_dev);
2123 }
2124
2125 /*
2126  *      Set destination address.
2127  *      Special case for SIT interfaces where we create a new "virtual"
2128  *      device.
2129  */
2130 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2131 {
2132         struct in6_ifreq ireq;
2133         struct net_device *dev;
2134         int err = -EINVAL;
2135
2136         rtnl_lock();
2137
2138         err = -EFAULT;
2139         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2140                 goto err_exit;
2141
2142         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2143
2144         err = -ENODEV;
2145         if (dev == NULL)
2146                 goto err_exit;
2147
2148 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2149         if (dev->type == ARPHRD_SIT) {
2150                 const struct net_device_ops *ops = dev->netdev_ops;
2151                 struct ifreq ifr;
2152                 struct ip_tunnel_parm p;
2153
2154                 err = -EADDRNOTAVAIL;
2155                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2156                         goto err_exit;
2157
2158                 memset(&p, 0, sizeof(p));
2159                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2160                 p.iph.saddr = 0;
2161                 p.iph.version = 4;
2162                 p.iph.ihl = 5;
2163                 p.iph.protocol = IPPROTO_IPV6;
2164                 p.iph.ttl = 64;
2165                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2166
2167                 if (ops->ndo_do_ioctl) {
2168                         mm_segment_t oldfs = get_fs();
2169
2170                         set_fs(KERNEL_DS);
2171                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2172                         set_fs(oldfs);
2173                 } else
2174                         err = -EOPNOTSUPP;
2175
2176                 if (err == 0) {
2177                         err = -ENOBUFS;
2178                         dev = __dev_get_by_name(net, p.name);
2179                         if (!dev)
2180                                 goto err_exit;
2181                         err = dev_open(dev);
2182                 }
2183         }
2184 #endif
2185
2186 err_exit:
2187         rtnl_unlock();
2188         return err;
2189 }
2190
2191 /*
2192  *      Manual configuration of address on an interface
2193  */
2194 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2195                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2196                           __u32 valid_lft)
2197 {
2198         struct inet6_ifaddr *ifp;
2199         struct inet6_dev *idev;
2200         struct net_device *dev;
2201         int scope;
2202         u32 flags;
2203         clock_t expires;
2204         unsigned long timeout;
2205
2206         ASSERT_RTNL();
2207
2208         if (plen > 128)
2209                 return -EINVAL;
2210
2211         /* check the lifetime */
2212         if (!valid_lft || prefered_lft > valid_lft)
2213                 return -EINVAL;
2214
2215         dev = __dev_get_by_index(net, ifindex);
2216         if (!dev)
2217                 return -ENODEV;
2218
2219         idev = addrconf_add_dev(dev);
2220         if (IS_ERR(idev))
2221                 return PTR_ERR(idev);
2222
2223         scope = ipv6_addr_scope(pfx);
2224
2225         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2226         if (addrconf_finite_timeout(timeout)) {
2227                 expires = jiffies_to_clock_t(timeout * HZ);
2228                 valid_lft = timeout;
2229                 flags = RTF_EXPIRES;
2230         } else {
2231                 expires = 0;
2232                 flags = 0;
2233                 ifa_flags |= IFA_F_PERMANENT;
2234         }
2235
2236         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2237         if (addrconf_finite_timeout(timeout)) {
2238                 if (timeout == 0)
2239                         ifa_flags |= IFA_F_DEPRECATED;
2240                 prefered_lft = timeout;
2241         }
2242
2243         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2244
2245         if (!IS_ERR(ifp)) {
2246                 spin_lock_bh(&ifp->lock);
2247                 ifp->valid_lft = valid_lft;
2248                 ifp->prefered_lft = prefered_lft;
2249                 ifp->tstamp = jiffies;
2250                 spin_unlock_bh(&ifp->lock);
2251
2252                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2253                                       expires, flags);
2254                 /*
2255                  * Note that section 3.1 of RFC 4429 indicates
2256                  * that the Optimistic flag should not be set for
2257                  * manually configured addresses
2258                  */
2259                 addrconf_dad_start(ifp, 0);
2260                 in6_ifa_put(ifp);
2261                 addrconf_verify(0);
2262                 return 0;
2263         }
2264
2265         return PTR_ERR(ifp);
2266 }
2267
2268 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2269                           unsigned int plen)
2270 {
2271         struct inet6_ifaddr *ifp;
2272         struct inet6_dev *idev;
2273         struct net_device *dev;
2274
2275         if (plen > 128)
2276                 return -EINVAL;
2277
2278         dev = __dev_get_by_index(net, ifindex);
2279         if (!dev)
2280                 return -ENODEV;
2281
2282         if ((idev = __in6_dev_get(dev)) == NULL)
2283                 return -ENXIO;
2284
2285         read_lock_bh(&idev->lock);
2286         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2287                 if (ifp->prefix_len == plen &&
2288                     ipv6_addr_equal(pfx, &ifp->addr)) {
2289                         in6_ifa_hold(ifp);
2290                         read_unlock_bh(&idev->lock);
2291
2292                         ipv6_del_addr(ifp);
2293
2294                         /* If the last address is deleted administratively,
2295                            disable IPv6 on this interface.
2296                          */
2297                         if (list_empty(&idev->addr_list))
2298                                 addrconf_ifdown(idev->dev, 1);
2299                         return 0;
2300                 }
2301         }
2302         read_unlock_bh(&idev->lock);
2303         return -EADDRNOTAVAIL;
2304 }
2305
2306
2307 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2308 {
2309         struct in6_ifreq ireq;
2310         int err;
2311
2312         if (!capable(CAP_NET_ADMIN))
2313                 return -EPERM;
2314
2315         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2316                 return -EFAULT;
2317
2318         rtnl_lock();
2319         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2320                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2321                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2322         rtnl_unlock();
2323         return err;
2324 }
2325
2326 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2327 {
2328         struct in6_ifreq ireq;
2329         int err;
2330
2331         if (!capable(CAP_NET_ADMIN))
2332                 return -EPERM;
2333
2334         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2335                 return -EFAULT;
2336
2337         rtnl_lock();
2338         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2339                              ireq.ifr6_prefixlen);
2340         rtnl_unlock();
2341         return err;
2342 }
2343
2344 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2345                      int plen, int scope)
2346 {
2347         struct inet6_ifaddr *ifp;
2348
2349         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2350         if (!IS_ERR(ifp)) {
2351                 spin_lock_bh(&ifp->lock);
2352                 ifp->flags &= ~IFA_F_TENTATIVE;
2353                 spin_unlock_bh(&ifp->lock);
2354                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2355                 in6_ifa_put(ifp);
2356         }
2357 }
2358
2359 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2360 static void sit_add_v4_addrs(struct inet6_dev *idev)
2361 {
2362         struct in6_addr addr;
2363         struct net_device *dev;
2364         struct net *net = dev_net(idev->dev);
2365         int scope;
2366
2367         ASSERT_RTNL();
2368
2369         memset(&addr, 0, sizeof(struct in6_addr));
2370         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2371
2372         if (idev->dev->flags&IFF_POINTOPOINT) {
2373                 addr.s6_addr32[0] = htonl(0xfe800000);
2374                 scope = IFA_LINK;
2375         } else {
2376                 scope = IPV6_ADDR_COMPATv4;
2377         }
2378
2379         if (addr.s6_addr32[3]) {
2380                 add_addr(idev, &addr, 128, scope);
2381                 return;
2382         }
2383
2384         for_each_netdev(net, dev) {
2385                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2386                 if (in_dev && (dev->flags & IFF_UP)) {
2387                         struct in_ifaddr * ifa;
2388
2389                         int flag = scope;
2390
2391                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2392                                 int plen;
2393
2394                                 addr.s6_addr32[3] = ifa->ifa_local;
2395
2396                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2397                                         continue;
2398                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2399                                         if (idev->dev->flags&IFF_POINTOPOINT)
2400                                                 continue;
2401                                         flag |= IFA_HOST;
2402                                 }
2403                                 if (idev->dev->flags&IFF_POINTOPOINT)
2404                                         plen = 64;
2405                                 else
2406                                         plen = 96;
2407
2408                                 add_addr(idev, &addr, plen, flag);
2409                         }
2410                 }
2411         }
2412 }
2413 #endif
2414
2415 static void init_loopback(struct net_device *dev)
2416 {
2417         struct inet6_dev  *idev;
2418         struct net_device *sp_dev;
2419         struct inet6_ifaddr *sp_ifa;
2420         struct rt6_info *sp_rt;
2421
2422         /* ::1 */
2423
2424         ASSERT_RTNL();
2425
2426         if ((idev = ipv6_find_idev(dev)) == NULL) {
2427                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2428                 return;
2429         }
2430
2431         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2432
2433         /* Add routes to other interface's IPv6 addresses */
2434         for_each_netdev(dev_net(dev), sp_dev) {
2435                 if (!strcmp(sp_dev->name, dev->name))
2436                         continue;
2437
2438                 idev = __in6_dev_get(sp_dev);
2439                 if (!idev)
2440                         continue;
2441
2442                 read_lock_bh(&idev->lock);
2443                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2444
2445                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2446                                 continue;
2447
2448                         if (sp_ifa->rt) {
2449                                 /* This dst has been added to garbage list when
2450                                  * lo device down, release this obsolete dst and
2451                                  * reallocate a new router for ifa.
2452                                  */
2453                                 if (sp_ifa->rt->dst.obsolete > 0) {
2454                                         dst_release(&sp_ifa->rt->dst);
2455                                         sp_ifa->rt = NULL;
2456                                 } else {
2457                                         continue;
2458                                 }
2459                         }
2460
2461                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2462
2463                         /* Failure cases are ignored */
2464                         if (!IS_ERR(sp_rt)) {
2465                                 sp_ifa->rt = sp_rt;
2466                                 ip6_ins_rt(sp_rt);
2467                         }
2468                 }
2469                 read_unlock_bh(&idev->lock);
2470         }
2471 }
2472
2473 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2474 {
2475         struct inet6_ifaddr * ifp;
2476         u32 addr_flags = IFA_F_PERMANENT;
2477
2478 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2479         if (idev->cnf.optimistic_dad &&
2480             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2481                 addr_flags |= IFA_F_OPTIMISTIC;
2482 #endif
2483
2484
2485         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2486         if (!IS_ERR(ifp)) {
2487                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2488                 addrconf_dad_start(ifp, 0);
2489                 in6_ifa_put(ifp);
2490         }
2491 }
2492
2493 static void addrconf_dev_config(struct net_device *dev)
2494 {
2495         struct in6_addr addr;
2496         struct inet6_dev    * idev;
2497
2498         ASSERT_RTNL();
2499
2500         if ((dev->type != ARPHRD_ETHER) &&
2501             (dev->type != ARPHRD_FDDI) &&
2502             (dev->type != ARPHRD_IEEE802_TR) &&
2503             (dev->type != ARPHRD_ARCNET) &&
2504             (dev->type != ARPHRD_INFINIBAND)) {
2505                 /* Alas, we support only Ethernet autoconfiguration. */
2506                 return;
2507         }
2508
2509         idev = addrconf_add_dev(dev);
2510         if (IS_ERR(idev))
2511                 return;
2512
2513         memset(&addr, 0, sizeof(struct in6_addr));
2514         addr.s6_addr32[0] = htonl(0xFE800000);
2515
2516         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2517                 addrconf_add_linklocal(idev, &addr);
2518 }
2519
2520 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2521 static void addrconf_sit_config(struct net_device *dev)
2522 {
2523         struct inet6_dev *idev;
2524
2525         ASSERT_RTNL();
2526
2527         /*
2528          * Configure the tunnel with one of our IPv4
2529          * addresses... we should configure all of
2530          * our v4 addrs in the tunnel
2531          */
2532
2533         if ((idev = ipv6_find_idev(dev)) == NULL) {
2534                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2535                 return;
2536         }
2537
2538         if (dev->priv_flags & IFF_ISATAP) {
2539                 struct in6_addr addr;
2540
2541                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2542                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2543                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2544                         addrconf_add_linklocal(idev, &addr);
2545                 return;
2546         }
2547
2548         sit_add_v4_addrs(idev);
2549
2550         if (dev->flags&IFF_POINTOPOINT) {
2551                 addrconf_add_mroute(dev);
2552                 addrconf_add_lroute(dev);
2553         } else
2554                 sit_route_add(dev);
2555 }
2556 #endif
2557
2558 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2559 static void addrconf_gre_config(struct net_device *dev)
2560 {
2561         struct inet6_dev *idev;
2562         struct in6_addr addr;
2563
2564         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2565
2566         ASSERT_RTNL();
2567
2568         if ((idev = ipv6_find_idev(dev)) == NULL) {
2569                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2570                 return;
2571         }
2572
2573         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2574         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2575
2576         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2577                 addrconf_add_linklocal(idev, &addr);
2578 }
2579 #endif
2580
2581 static inline int
2582 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2583 {
2584         struct in6_addr lladdr;
2585
2586         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2587                 addrconf_add_linklocal(idev, &lladdr);
2588                 return 0;
2589         }
2590         return -1;
2591 }
2592
2593 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2594 {
2595         struct net_device *link_dev;
2596         struct net *net = dev_net(idev->dev);
2597
2598         /* first try to inherit the link-local address from the link device */
2599         if (idev->dev->iflink &&
2600             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2601                 if (!ipv6_inherit_linklocal(idev, link_dev))
2602                         return;
2603         }
2604         /* then try to inherit it from any device */
2605         for_each_netdev(net, link_dev) {
2606                 if (!ipv6_inherit_linklocal(idev, link_dev))
2607                         return;
2608         }
2609         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2610 }
2611
2612 /*
2613  * Autoconfigure tunnel with a link-local address so routing protocols,
2614  * DHCPv6, MLD etc. can be run over the virtual link
2615  */
2616
2617 static void addrconf_ip6_tnl_config(struct net_device *dev)
2618 {
2619         struct inet6_dev *idev;
2620
2621         ASSERT_RTNL();
2622
2623         idev = addrconf_add_dev(dev);
2624         if (IS_ERR(idev)) {
2625                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2626                 return;
2627         }
2628         ip6_tnl_add_linklocal(idev);
2629 }
2630
2631 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2632                            void * data)
2633 {
2634         struct net_device *dev = (struct net_device *) data;
2635         struct inet6_dev *idev = __in6_dev_get(dev);
2636         int run_pending = 0;
2637         int err;
2638
2639         switch (event) {
2640         case NETDEV_REGISTER:
2641                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2642                         idev = ipv6_add_dev(dev);
2643                         if (!idev)
2644                                 return notifier_from_errno(-ENOMEM);
2645                 }
2646                 break;
2647
2648         case NETDEV_UP:
2649         case NETDEV_CHANGE:
2650                 if (dev->flags & IFF_SLAVE)
2651                         break;
2652
2653                 if (event == NETDEV_UP) {
2654                         if (!addrconf_qdisc_ok(dev)) {
2655                                 /* device is not ready yet. */
2656                                 printk(KERN_INFO
2657                                         "ADDRCONF(NETDEV_UP): %s: "
2658                                         "link is not ready\n",
2659                                         dev->name);
2660                                 break;
2661                         }
2662
2663                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2664                                 idev = ipv6_add_dev(dev);
2665
2666                         if (idev) {
2667                                 idev->if_flags |= IF_READY;
2668                                 run_pending = 1;
2669                         }
2670                 } else {
2671                         if (!addrconf_qdisc_ok(dev)) {
2672                                 /* device is still not ready. */
2673                                 break;
2674                         }
2675
2676                         if (idev) {
2677                                 if (idev->if_flags & IF_READY)
2678                                         /* device is already configured. */
2679                                         break;
2680                                 idev->if_flags |= IF_READY;
2681                         }
2682
2683                         printk(KERN_INFO
2684                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2685                                         "link becomes ready\n",
2686                                         dev->name);
2687
2688                         run_pending = 1;
2689                 }
2690
2691                 switch (dev->type) {
2692 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2693                 case ARPHRD_SIT:
2694                         addrconf_sit_config(dev);
2695                         break;
2696 #endif
2697 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2698                 case ARPHRD_IPGRE:
2699                         addrconf_gre_config(dev);
2700                         break;
2701 #endif
2702                 case ARPHRD_TUNNEL6:
2703                         addrconf_ip6_tnl_config(dev);
2704                         break;
2705                 case ARPHRD_LOOPBACK:
2706                         init_loopback(dev);
2707                         break;
2708
2709                 default:
2710                         addrconf_dev_config(dev);
2711                         break;
2712                 }
2713
2714                 if (idev) {
2715                         if (run_pending)
2716                                 addrconf_dad_run(idev);
2717
2718                         /*
2719                          * If the MTU changed during the interface down,
2720                          * when the interface up, the changed MTU must be
2721                          * reflected in the idev as well as routers.
2722                          */
2723                         if (idev->cnf.mtu6 != dev->mtu &&
2724                             dev->mtu >= IPV6_MIN_MTU) {
2725                                 rt6_mtu_change(dev, dev->mtu);
2726                                 idev->cnf.mtu6 = dev->mtu;
2727                         }
2728                         idev->tstamp = jiffies;
2729                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2730
2731                         /*
2732                          * If the changed mtu during down is lower than
2733                          * IPV6_MIN_MTU stop IPv6 on this interface.
2734                          */
2735                         if (dev->mtu < IPV6_MIN_MTU)
2736                                 addrconf_ifdown(dev, 1);
2737                 }
2738                 break;
2739
2740         case NETDEV_CHANGEMTU:
2741                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2742                         rt6_mtu_change(dev, dev->mtu);
2743                         idev->cnf.mtu6 = dev->mtu;
2744                         break;
2745                 }
2746
2747                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2748                         idev = ipv6_add_dev(dev);
2749                         if (idev)
2750                                 break;
2751                 }
2752
2753                 /*
2754                  * MTU falled under IPV6_MIN_MTU.
2755                  * Stop IPv6 on this interface.
2756                  */
2757
2758         case NETDEV_DOWN:
2759         case NETDEV_UNREGISTER:
2760                 /*
2761                  *      Remove all addresses from this interface.
2762                  */
2763                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2764                 break;
2765
2766         case NETDEV_CHANGENAME:
2767                 if (idev) {
2768                         snmp6_unregister_dev(idev);
2769                         addrconf_sysctl_unregister(idev);
2770                         addrconf_sysctl_register(idev);
2771                         err = snmp6_register_dev(idev);
2772                         if (err)
2773                                 return notifier_from_errno(err);
2774                 }
2775                 break;
2776
2777         case NETDEV_PRE_TYPE_CHANGE:
2778         case NETDEV_POST_TYPE_CHANGE:
2779                 addrconf_type_change(dev, event);
2780                 break;
2781         }
2782
2783         return NOTIFY_OK;
2784 }
2785
2786 /*
2787  *      addrconf module should be notified of a device going up
2788  */
2789 static struct notifier_block ipv6_dev_notf = {
2790         .notifier_call = addrconf_notify,
2791 };
2792
2793 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2794 {
2795         struct inet6_dev *idev;
2796         ASSERT_RTNL();
2797
2798         idev = __in6_dev_get(dev);
2799
2800         if (event == NETDEV_POST_TYPE_CHANGE)
2801                 ipv6_mc_remap(idev);
2802         else if (event == NETDEV_PRE_TYPE_CHANGE)
2803                 ipv6_mc_unmap(idev);
2804 }
2805
2806 static int addrconf_ifdown(struct net_device *dev, int how)
2807 {
2808         struct net *net = dev_net(dev);
2809         struct inet6_dev *idev;
2810         struct inet6_ifaddr *ifa;
2811         int state, i;
2812
2813         ASSERT_RTNL();
2814
2815         rt6_ifdown(net, dev);
2816         neigh_ifdown(&nd_tbl, dev);
2817
2818         idev = __in6_dev_get(dev);
2819         if (idev == NULL)
2820                 return -ENODEV;
2821
2822         /*
2823          * Step 1: remove reference to ipv6 device from parent device.
2824          *         Do not dev_put!
2825          */
2826         if (how) {
2827                 idev->dead = 1;
2828
2829                 /* protected by rtnl_lock */
2830                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2831
2832                 /* Step 1.5: remove snmp6 entry */
2833                 snmp6_unregister_dev(idev);
2834
2835         }
2836
2837         /* Step 2: clear hash table */
2838         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2839                 struct hlist_head *h = &inet6_addr_lst[i];
2840                 struct hlist_node *n;
2841
2842                 spin_lock_bh(&addrconf_hash_lock);
2843         restart:
2844                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2845                         if (ifa->idev == idev) {
2846                                 hlist_del_init_rcu(&ifa->addr_lst);
2847                                 addrconf_del_timer(ifa);
2848                                 goto restart;
2849                         }
2850                 }
2851                 spin_unlock_bh(&addrconf_hash_lock);
2852         }
2853
2854         write_lock_bh(&idev->lock);
2855
2856         /* Step 2: clear flags for stateless addrconf */
2857         if (!how)
2858                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2859
2860 #ifdef CONFIG_IPV6_PRIVACY
2861         if (how && del_timer(&idev->regen_timer))
2862                 in6_dev_put(idev);
2863
2864         /* Step 3: clear tempaddr list */
2865         while (!list_empty(&idev->tempaddr_list)) {
2866                 ifa = list_first_entry(&idev->tempaddr_list,
2867                                        struct inet6_ifaddr, tmp_list);
2868                 list_del(&ifa->tmp_list);
2869                 write_unlock_bh(&idev->lock);
2870                 spin_lock_bh(&ifa->lock);
2871
2872                 if (ifa->ifpub) {
2873                         in6_ifa_put(ifa->ifpub);
2874                         ifa->ifpub = NULL;
2875                 }
2876                 spin_unlock_bh(&ifa->lock);
2877                 in6_ifa_put(ifa);
2878                 write_lock_bh(&idev->lock);
2879         }
2880 #endif
2881
2882         while (!list_empty(&idev->addr_list)) {
2883                 ifa = list_first_entry(&idev->addr_list,
2884                                        struct inet6_ifaddr, if_list);
2885                 addrconf_del_timer(ifa);
2886
2887                 list_del(&ifa->if_list);
2888
2889                 write_unlock_bh(&idev->lock);
2890
2891                 spin_lock_bh(&ifa->state_lock);
2892                 state = ifa->state;
2893                 ifa->state = INET6_IFADDR_STATE_DEAD;
2894                 spin_unlock_bh(&ifa->state_lock);
2895
2896                 if (state != INET6_IFADDR_STATE_DEAD) {
2897                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2898                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2899                 }
2900                 in6_ifa_put(ifa);
2901
2902                 write_lock_bh(&idev->lock);
2903         }
2904
2905         write_unlock_bh(&idev->lock);
2906
2907         /* Step 5: Discard multicast list */
2908         if (how)
2909                 ipv6_mc_destroy_dev(idev);
2910         else
2911                 ipv6_mc_down(idev);
2912
2913         idev->tstamp = jiffies;
2914
2915         /* Last: Shot the device (if unregistered) */
2916         if (how) {
2917                 addrconf_sysctl_unregister(idev);
2918                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2919                 neigh_ifdown(&nd_tbl, dev);
2920                 in6_dev_put(idev);
2921         }
2922         return 0;
2923 }
2924
2925 static void addrconf_rs_timer(unsigned long data)
2926 {
2927         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2928         struct inet6_dev *idev = ifp->idev;
2929
2930         read_lock(&idev->lock);
2931         if (idev->dead || !(idev->if_flags & IF_READY))
2932                 goto out;
2933
2934         if (idev->cnf.forwarding)
2935                 goto out;
2936
2937         /* Announcement received after solicitation was sent */
2938         if (idev->if_flags & IF_RA_RCVD)
2939                 goto out;
2940
2941         spin_lock(&ifp->lock);
2942         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2943                 /* The wait after the last probe can be shorter */
2944                 addrconf_mod_timer(ifp, AC_RS,
2945                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2946                                    idev->cnf.rtr_solicit_delay :
2947                                    idev->cnf.rtr_solicit_interval);
2948                 spin_unlock(&ifp->lock);
2949
2950                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2951         } else {
2952                 spin_unlock(&ifp->lock);
2953                 /*
2954                  * Note: we do not support deprecated "all on-link"
2955                  * assumption any longer.
2956                  */
2957                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2958                        idev->dev->name);
2959         }
2960
2961 out:
2962         read_unlock(&idev->lock);
2963         in6_ifa_put(ifp);
2964 }
2965
2966 /*
2967  *      Duplicate Address Detection
2968  */
2969 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2970 {
2971         unsigned long rand_num;
2972         struct inet6_dev *idev = ifp->idev;
2973
2974         if (ifp->flags & IFA_F_OPTIMISTIC)
2975                 rand_num = 0;
2976         else
2977                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2978
2979         ifp->probes = idev->cnf.dad_transmits;
2980         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2981 }
2982
2983 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2984 {
2985         struct inet6_dev *idev = ifp->idev;
2986         struct net_device *dev = idev->dev;
2987
2988         addrconf_join_solict(dev, &ifp->addr);
2989
2990         net_srandom(ifp->addr.s6_addr32[3]);
2991
2992         read_lock_bh(&idev->lock);
2993         spin_lock(&ifp->lock);
2994         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2995                 goto out;
2996
2997         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2998             idev->cnf.accept_dad < 1 ||
2999             !(ifp->flags&IFA_F_TENTATIVE) ||
3000             ifp->flags & IFA_F_NODAD) {
3001                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3002                 spin_unlock(&ifp->lock);
3003                 read_unlock_bh(&idev->lock);
3004
3005                 addrconf_dad_completed(ifp);
3006                 return;
3007         }
3008
3009         if (!(idev->if_flags & IF_READY)) {
3010                 spin_unlock(&ifp->lock);
3011                 read_unlock_bh(&idev->lock);
3012                 /*
3013                  * If the device is not ready:
3014                  * - keep it tentative if it is a permanent address.
3015                  * - otherwise, kill it.
3016                  */
3017                 in6_ifa_hold(ifp);
3018                 addrconf_dad_stop(ifp, 0);
3019                 return;
3020         }
3021
3022         /*
3023          * Optimistic nodes can start receiving
3024          * Frames right away
3025          */
3026         if (ifp->flags & IFA_F_OPTIMISTIC)
3027                 ip6_ins_rt(ifp->rt);
3028
3029         addrconf_dad_kick(ifp);
3030 out:
3031         spin_unlock(&ifp->lock);
3032         read_unlock_bh(&idev->lock);
3033 }
3034
3035 static void addrconf_dad_timer(unsigned long data)
3036 {
3037         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3038         struct inet6_dev *idev = ifp->idev;
3039         struct in6_addr mcaddr;
3040
3041         if (!ifp->probes && addrconf_dad_end(ifp))
3042                 goto out;
3043
3044         read_lock(&idev->lock);
3045         if (idev->dead || !(idev->if_flags & IF_READY)) {
3046                 read_unlock(&idev->lock);
3047                 goto out;
3048         }
3049
3050         spin_lock(&ifp->lock);
3051         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3052                 spin_unlock(&ifp->lock);
3053                 read_unlock(&idev->lock);
3054                 goto out;
3055         }
3056
3057         if (ifp->probes == 0) {
3058                 /*
3059                  * DAD was successful
3060                  */
3061
3062                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3063                 spin_unlock(&ifp->lock);
3064                 read_unlock(&idev->lock);
3065
3066                 addrconf_dad_completed(ifp);
3067
3068                 goto out;
3069         }
3070
3071         ifp->probes--;
3072         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3073         spin_unlock(&ifp->lock);
3074         read_unlock(&idev->lock);
3075
3076         /* send a neighbour solicitation for our addr */
3077         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3078         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3079 out:
3080         in6_ifa_put(ifp);
3081 }
3082
3083 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3084 {
3085         struct net_device *dev = ifp->idev->dev;
3086
3087         /*
3088          *      Configure the address for reception. Now it is valid.
3089          */
3090
3091         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3092
3093         /* If added prefix is link local and we are prepared to process
3094            router advertisements, start sending router solicitations.
3095          */
3096
3097         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3098              ifp->idev->cnf.accept_ra == 2) &&
3099             ifp->idev->cnf.rtr_solicits > 0 &&
3100             (dev->flags&IFF_LOOPBACK) == 0 &&
3101             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3102                 /*
3103                  *      If a host as already performed a random delay
3104                  *      [...] as part of DAD [...] there is no need
3105                  *      to delay again before sending the first RS
3106                  */
3107                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3108
3109                 spin_lock_bh(&ifp->lock);
3110                 ifp->probes = 1;
3111                 ifp->idev->if_flags |= IF_RS_SENT;
3112                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3113                 spin_unlock_bh(&ifp->lock);
3114         }
3115 }
3116
3117 static void addrconf_dad_run(struct inet6_dev *idev)
3118 {
3119         struct inet6_ifaddr *ifp;
3120
3121         read_lock_bh(&idev->lock);
3122         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3123                 spin_lock(&ifp->lock);
3124                 if (ifp->flags & IFA_F_TENTATIVE &&
3125                     ifp->state == INET6_IFADDR_STATE_DAD)
3126                         addrconf_dad_kick(ifp);
3127                 spin_unlock(&ifp->lock);
3128         }
3129         read_unlock_bh(&idev->lock);
3130 }
3131
3132 #ifdef CONFIG_PROC_FS
3133 struct if6_iter_state {
3134         struct seq_net_private p;
3135         int bucket;
3136 };
3137
3138 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3139 {
3140         struct inet6_ifaddr *ifa = NULL;
3141         struct if6_iter_state *state = seq->private;
3142         struct net *net = seq_file_net(seq);
3143
3144         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3145                 struct hlist_node *n;
3146                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3147                                          addr_lst)
3148                         if (net_eq(dev_net(ifa->idev->dev), net))
3149                                 return ifa;
3150         }
3151         return NULL;
3152 }
3153
3154 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3155                                          struct inet6_ifaddr *ifa)
3156 {
3157         struct if6_iter_state *state = seq->private;
3158         struct net *net = seq_file_net(seq);
3159         struct hlist_node *n = &ifa->addr_lst;
3160
3161         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3162                 if (net_eq(dev_net(ifa->idev->dev), net))
3163                         return ifa;
3164
3165         while (++state->bucket < IN6_ADDR_HSIZE) {
3166                 hlist_for_each_entry_rcu_bh(ifa, n,
3167                                      &inet6_addr_lst[state->bucket], addr_lst) {
3168                         if (net_eq(dev_net(ifa->idev->dev), net))
3169                                 return ifa;
3170                 }
3171         }
3172
3173         return NULL;
3174 }
3175
3176 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3177 {
3178         struct inet6_ifaddr *ifa = if6_get_first(seq);
3179
3180         if (ifa)
3181                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3182                         --pos;
3183         return pos ? NULL : ifa;
3184 }
3185
3186 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3187         __acquires(rcu_bh)
3188 {
3189         rcu_read_lock_bh();
3190         return if6_get_idx(seq, *pos);
3191 }
3192
3193 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3194 {
3195         struct inet6_ifaddr *ifa;
3196
3197         ifa = if6_get_next(seq, v);
3198         ++*pos;
3199         return ifa;
3200 }
3201
3202 static void if6_seq_stop(struct seq_file *seq, void *v)
3203         __releases(rcu_bh)
3204 {
3205         rcu_read_unlock_bh();
3206 }
3207
3208 static int if6_seq_show(struct seq_file *seq, void *v)
3209 {
3210         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3211         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3212                    &ifp->addr,
3213                    ifp->idev->dev->ifindex,
3214                    ifp->prefix_len,
3215                    ifp->scope,
3216                    ifp->flags,
3217                    ifp->idev->dev->name);
3218         return 0;
3219 }
3220
3221 static const struct seq_operations if6_seq_ops = {
3222         .start  = if6_seq_start,
3223         .next   = if6_seq_next,
3224         .show   = if6_seq_show,
3225         .stop   = if6_seq_stop,
3226 };
3227
3228 static int if6_seq_open(struct inode *inode, struct file *file)
3229 {
3230         return seq_open_net(inode, file, &if6_seq_ops,
3231                             sizeof(struct if6_iter_state));
3232 }
3233
3234 static const struct file_operations if6_fops = {
3235         .owner          = THIS_MODULE,
3236         .open           = if6_seq_open,
3237         .read           = seq_read,
3238         .llseek         = seq_lseek,
3239         .release        = seq_release_net,
3240 };
3241
3242 static int __net_init if6_proc_net_init(struct net *net)
3243 {
3244         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3245                 return -ENOMEM;
3246         return 0;
3247 }
3248
3249 static void __net_exit if6_proc_net_exit(struct net *net)
3250 {
3251        proc_net_remove(net, "if_inet6");
3252 }
3253
3254 static struct pernet_operations if6_proc_net_ops = {
3255        .init = if6_proc_net_init,
3256        .exit = if6_proc_net_exit,
3257 };
3258
3259 int __init if6_proc_init(void)
3260 {
3261         return register_pernet_subsys(&if6_proc_net_ops);
3262 }
3263
3264 void if6_proc_exit(void)
3265 {
3266         unregister_pernet_subsys(&if6_proc_net_ops);
3267 }
3268 #endif  /* CONFIG_PROC_FS */
3269
3270 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3271 /* Check if address is a home address configured on any interface. */
3272 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3273 {
3274         int ret = 0;
3275         struct inet6_ifaddr *ifp = NULL;
3276         struct hlist_node *n;
3277         unsigned int hash = ipv6_addr_hash(addr);
3278
3279         rcu_read_lock_bh();
3280         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3281                 if (!net_eq(dev_net(ifp->idev->dev), net))
3282                         continue;
3283                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3284                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3285                         ret = 1;
3286                         break;
3287                 }
3288         }
3289         rcu_read_unlock_bh();
3290         return ret;
3291 }
3292 #endif
3293
3294 /*
3295  *      Periodic address status verification
3296  */
3297
3298 static void addrconf_verify(unsigned long foo)
3299 {
3300         unsigned long now, next, next_sec, next_sched;
3301         struct inet6_ifaddr *ifp;
3302         struct hlist_node *node;
3303         int i;
3304
3305         rcu_read_lock_bh();
3306         spin_lock(&addrconf_verify_lock);
3307         now = jiffies;
3308         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3309
3310         del_timer(&addr_chk_timer);
3311
3312         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3313 restart:
3314                 hlist_for_each_entry_rcu_bh(ifp, node,
3315                                          &inet6_addr_lst[i], addr_lst) {
3316                         unsigned long age;
3317
3318                         if (ifp->flags & IFA_F_PERMANENT)
3319                                 continue;
3320
3321                         spin_lock(&ifp->lock);
3322                         /* We try to batch several events at once. */
3323                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3324
3325                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3326                             age >= ifp->valid_lft) {
3327                                 spin_unlock(&ifp->lock);
3328                                 in6_ifa_hold(ifp);
3329                                 ipv6_del_addr(ifp);
3330                                 goto restart;
3331                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3332                                 spin_unlock(&ifp->lock);
3333                                 continue;
3334                         } else if (age >= ifp->prefered_lft) {
3335                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3336                                 int deprecate = 0;
3337
3338                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3339                                         deprecate = 1;
3340                                         ifp->flags |= IFA_F_DEPRECATED;
3341                                 }
3342
3343                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3344                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3345
3346                                 spin_unlock(&ifp->lock);
3347
3348                                 if (deprecate) {
3349                                         in6_ifa_hold(ifp);
3350
3351                                         ipv6_ifa_notify(0, ifp);
3352                                         in6_ifa_put(ifp);
3353                                         goto restart;
3354                                 }
3355 #ifdef CONFIG_IPV6_PRIVACY
3356                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3357                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3358                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3359                                         ifp->idev->cnf.dad_transmits *
3360                                         ifp->idev->nd_parms->retrans_time / HZ;
3361
3362                                 if (age >= ifp->prefered_lft - regen_advance) {
3363                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3364                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3365                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3366                                         if (!ifp->regen_count && ifpub) {
3367                                                 ifp->regen_count++;
3368                                                 in6_ifa_hold(ifp);
3369                                                 in6_ifa_hold(ifpub);
3370                                                 spin_unlock(&ifp->lock);
3371
3372                                                 spin_lock(&ifpub->lock);
3373                                                 ifpub->regen_count = 0;
3374                                                 spin_unlock(&ifpub->lock);
3375                                                 ipv6_create_tempaddr(ifpub, ifp);
3376                                                 in6_ifa_put(ifpub);
3377                                                 in6_ifa_put(ifp);
3378                                                 goto restart;
3379                                         }
3380                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3381                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3382                                 spin_unlock(&ifp->lock);
3383 #endif
3384                         } else {
3385                                 /* ifp->prefered_lft <= ifp->valid_lft */
3386                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3387                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3388                                 spin_unlock(&ifp->lock);
3389                         }
3390                 }
3391         }
3392
3393         next_sec = round_jiffies_up(next);
3394         next_sched = next;
3395
3396         /* If rounded timeout is accurate enough, accept it. */
3397         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3398                 next_sched = next_sec;
3399
3400         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3401         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3402                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3403
3404         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3405               now, next, next_sec, next_sched));
3406
3407         addr_chk_timer.expires = next_sched;
3408         add_timer(&addr_chk_timer);
3409         spin_unlock(&addrconf_verify_lock);
3410         rcu_read_unlock_bh();
3411 }
3412
3413 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3414 {
3415         struct in6_addr *pfx = NULL;
3416
3417         if (addr)
3418                 pfx = nla_data(addr);
3419
3420         if (local) {
3421                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3422                         pfx = NULL;
3423                 else
3424                         pfx = nla_data(local);
3425         }
3426
3427         return pfx;
3428 }
3429
3430 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3431         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3432         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3433         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3434 };
3435
3436 static int
3437 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3438 {
3439         struct net *net = sock_net(skb->sk);
3440         struct ifaddrmsg *ifm;
3441         struct nlattr *tb[IFA_MAX+1];
3442         struct in6_addr *pfx;
3443         int err;
3444
3445         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3446         if (err < 0)
3447                 return err;
3448
3449         ifm = nlmsg_data(nlh);
3450         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3451         if (pfx == NULL)
3452                 return -EINVAL;
3453
3454         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3455 }
3456
3457 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3458                              u32 prefered_lft, u32 valid_lft)
3459 {
3460         u32 flags;
3461         clock_t expires;
3462         unsigned long timeout;
3463
3464         if (!valid_lft || (prefered_lft > valid_lft))
3465                 return -EINVAL;
3466
3467         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3468         if (addrconf_finite_timeout(timeout)) {
3469                 expires = jiffies_to_clock_t(timeout * HZ);
3470                 valid_lft = timeout;
3471                 flags = RTF_EXPIRES;
3472         } else {
3473                 expires = 0;
3474                 flags = 0;
3475                 ifa_flags |= IFA_F_PERMANENT;
3476         }
3477
3478         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3479         if (addrconf_finite_timeout(timeout)) {
3480                 if (timeout == 0)
3481                         ifa_flags |= IFA_F_DEPRECATED;
3482                 prefered_lft = timeout;
3483         }
3484
3485         spin_lock_bh(&ifp->lock);
3486         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3487         ifp->tstamp = jiffies;
3488         ifp->valid_lft = valid_lft;
3489         ifp->prefered_lft = prefered_lft;
3490
3491         spin_unlock_bh(&ifp->lock);
3492         if (!(ifp->flags&IFA_F_TENTATIVE))
3493                 ipv6_ifa_notify(0, ifp);
3494
3495         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3496                               expires, flags);
3497         addrconf_verify(0);
3498
3499         return 0;
3500 }
3501
3502 static int
3503 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3504 {
3505         struct net *net = sock_net(skb->sk);
3506         struct ifaddrmsg *ifm;
3507         struct nlattr *tb[IFA_MAX+1];
3508         struct in6_addr *pfx;
3509         struct inet6_ifaddr *ifa;
3510         struct net_device *dev;
3511         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3512         u8 ifa_flags;
3513         int err;
3514
3515         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3516         if (err < 0)
3517                 return err;
3518
3519         ifm = nlmsg_data(nlh);
3520         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3521         if (pfx == NULL)
3522                 return -EINVAL;
3523
3524         if (tb[IFA_CACHEINFO]) {
3525                 struct ifa_cacheinfo *ci;
3526
3527                 ci = nla_data(tb[IFA_CACHEINFO]);
3528                 valid_lft = ci->ifa_valid;
3529                 preferred_lft = ci->ifa_prefered;
3530         } else {
3531                 preferred_lft = INFINITY_LIFE_TIME;
3532                 valid_lft = INFINITY_LIFE_TIME;
3533         }
3534
3535         dev =  __dev_get_by_index(net, ifm->ifa_index);
3536         if (dev == NULL)
3537                 return -ENODEV;
3538
3539         /* We ignore other flags so far. */
3540         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3541
3542         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3543         if (ifa == NULL) {
3544                 /*
3545                  * It would be best to check for !NLM_F_CREATE here but
3546                  * userspace alreay relies on not having to provide this.
3547                  */
3548                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3549                                       ifm->ifa_prefixlen, ifa_flags,
3550                                       preferred_lft, valid_lft);
3551         }
3552
3553         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3554             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3555                 err = -EEXIST;
3556         else
3557                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3558
3559         in6_ifa_put(ifa);
3560
3561         return err;
3562 }
3563
3564 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3565                           u8 scope, int ifindex)
3566 {
3567         struct ifaddrmsg *ifm;
3568
3569         ifm = nlmsg_data(nlh);
3570         ifm->ifa_family = AF_INET6;
3571         ifm->ifa_prefixlen = prefixlen;
3572         ifm->ifa_flags = flags;
3573         ifm->ifa_scope = scope;
3574         ifm->ifa_index = ifindex;
3575 }
3576
3577 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3578                          unsigned long tstamp, u32 preferred, u32 valid)
3579 {
3580         struct ifa_cacheinfo ci;
3581
3582         ci.cstamp = cstamp_delta(cstamp);
3583         ci.tstamp = cstamp_delta(tstamp);
3584         ci.ifa_prefered = preferred;
3585         ci.ifa_valid = valid;
3586
3587         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3588 }
3589
3590 static inline int rt_scope(int ifa_scope)
3591 {
3592         if (ifa_scope & IFA_HOST)
3593                 return RT_SCOPE_HOST;
3594         else if (ifa_scope & IFA_LINK)
3595                 return RT_SCOPE_LINK;
3596         else if (ifa_scope & IFA_SITE)
3597                 return RT_SCOPE_SITE;
3598         else
3599                 return RT_SCOPE_UNIVERSE;
3600 }
3601
3602 static inline int inet6_ifaddr_msgsize(void)
3603 {
3604         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3605                + nla_total_size(16) /* IFA_ADDRESS */
3606                + nla_total_size(sizeof(struct ifa_cacheinfo));
3607 }
3608
3609 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3610                              u32 pid, u32 seq, int event, unsigned int flags)
3611 {
3612         struct nlmsghdr  *nlh;
3613         u32 preferred, valid;
3614
3615         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3616         if (nlh == NULL)
3617                 return -EMSGSIZE;
3618
3619         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3620                       ifa->idev->dev->ifindex);
3621
3622         if (!(ifa->flags&IFA_F_PERMANENT)) {
3623                 preferred = ifa->prefered_lft;
3624                 valid = ifa->valid_lft;
3625                 if (preferred != INFINITY_LIFE_TIME) {
3626                         long tval = (jiffies - ifa->tstamp)/HZ;
3627                         if (preferred > tval)
3628                                 preferred -= tval;
3629                         else
3630                                 preferred = 0;
3631                         if (valid != INFINITY_LIFE_TIME) {
3632                                 if (valid > tval)
3633                                         valid -= tval;
3634                                 else
3635                                         valid = 0;
3636                         }
3637                 }
3638         } else {
3639                 preferred = INFINITY_LIFE_TIME;
3640                 valid = INFINITY_LIFE_TIME;
3641         }
3642
3643         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3644             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3645                 nlmsg_cancel(skb, nlh);
3646                 return -EMSGSIZE;
3647         }
3648
3649         return nlmsg_end(skb, nlh);
3650 }
3651
3652 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3653                                 u32 pid, u32 seq, int event, u16 flags)
3654 {
3655         struct nlmsghdr  *nlh;
3656         u8 scope = RT_SCOPE_UNIVERSE;
3657         int ifindex = ifmca->idev->dev->ifindex;
3658
3659         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3660                 scope = RT_SCOPE_SITE;
3661
3662         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3663         if (nlh == NULL)
3664                 return -EMSGSIZE;
3665
3666         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3667         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3668             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3669                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3670                 nlmsg_cancel(skb, nlh);
3671                 return -EMSGSIZE;
3672         }
3673
3674         return nlmsg_end(skb, nlh);
3675 }
3676
3677 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3678                                 u32 pid, u32 seq, int event, unsigned int flags)
3679 {
3680         struct nlmsghdr  *nlh;
3681         u8 scope = RT_SCOPE_UNIVERSE;
3682         int ifindex = ifaca->aca_idev->dev->ifindex;
3683
3684         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3685                 scope = RT_SCOPE_SITE;
3686
3687         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3688         if (nlh == NULL)
3689                 return -EMSGSIZE;
3690
3691         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3692         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3693             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3694                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3695                 nlmsg_cancel(skb, nlh);
3696                 return -EMSGSIZE;
3697         }
3698
3699         return nlmsg_end(skb, nlh);
3700 }
3701
3702 enum addr_type_t {
3703         UNICAST_ADDR,
3704         MULTICAST_ADDR,
3705         ANYCAST_ADDR,
3706 };
3707
3708 /* called with rcu_read_lock() */
3709 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3710                           struct netlink_callback *cb, enum addr_type_t type,
3711                           int s_ip_idx, int *p_ip_idx)
3712 {
3713         struct ifmcaddr6 *ifmca;
3714         struct ifacaddr6 *ifaca;
3715         int err = 1;
3716         int ip_idx = *p_ip_idx;
3717
3718         read_lock_bh(&idev->lock);
3719         switch (type) {
3720         case UNICAST_ADDR: {
3721                 struct inet6_ifaddr *ifa;
3722
3723                 /* unicast address incl. temp addr */
3724                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3725                         if (++ip_idx < s_ip_idx)
3726                                 continue;
3727                         err = inet6_fill_ifaddr(skb, ifa,
3728                                                 NETLINK_CB(cb->skb).pid,
3729                                                 cb->nlh->nlmsg_seq,
3730                                                 RTM_NEWADDR,
3731                                                 NLM_F_MULTI);
3732                         if (err <= 0)
3733                                 break;
3734                 }
3735                 break;
3736         }
3737         case MULTICAST_ADDR:
3738                 /* multicast address */
3739                 for (ifmca = idev->mc_list; ifmca;
3740                      ifmca = ifmca->next, ip_idx++) {
3741                         if (ip_idx < s_ip_idx)
3742                                 continue;
3743                         err = inet6_fill_ifmcaddr(skb, ifmca,
3744                                                   NETLINK_CB(cb->skb).pid,
3745                                                   cb->nlh->nlmsg_seq,
3746                                                   RTM_GETMULTICAST,
3747                                                   NLM_F_MULTI);
3748                         if (err <= 0)
3749                                 break;
3750                 }
3751                 break;
3752         case ANYCAST_ADDR:
3753                 /* anycast address */
3754                 for (ifaca = idev->ac_list; ifaca;
3755                      ifaca = ifaca->aca_next, ip_idx++) {
3756                         if (ip_idx < s_ip_idx)
3757                                 continue;
3758                         err = inet6_fill_ifacaddr(skb, ifaca,
3759                                                   NETLINK_CB(cb->skb).pid,
3760                                                   cb->nlh->nlmsg_seq,
3761                                                   RTM_GETANYCAST,
3762                                                   NLM_F_MULTI);
3763                         if (err <= 0)
3764                                 break;
3765                 }
3766                 break;
3767         default:
3768                 break;
3769         }
3770         read_unlock_bh(&idev->lock);
3771         *p_ip_idx = ip_idx;
3772         return err;
3773 }
3774
3775 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3776                            enum addr_type_t type)
3777 {
3778         struct net *net = sock_net(skb->sk);
3779         int h, s_h;
3780         int idx, ip_idx;
3781         int s_idx, s_ip_idx;
3782         struct net_device *dev;
3783         struct inet6_dev *idev;
3784         struct hlist_head *head;
3785         struct hlist_node *node;
3786
3787         s_h = cb->args[0];
3788         s_idx = idx = cb->args[1];
3789         s_ip_idx = ip_idx = cb->args[2];
3790
3791         rcu_read_lock();
3792         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3793                 idx = 0;
3794                 head = &net->dev_index_head[h];
3795                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3796                         if (idx < s_idx)
3797                                 goto cont;
3798                         if (h > s_h || idx > s_idx)
3799                                 s_ip_idx = 0;
3800                         ip_idx = 0;
3801                         idev = __in6_dev_get(dev);
3802                         if (!idev)
3803                                 goto cont;
3804
3805                         if (in6_dump_addrs(idev, skb, cb, type,
3806                                            s_ip_idx, &ip_idx) <= 0)
3807                                 goto done;
3808 cont:
3809                         idx++;
3810                 }
3811         }
3812 done:
3813         rcu_read_unlock();
3814         cb->args[0] = h;
3815         cb->args[1] = idx;
3816         cb->args[2] = ip_idx;
3817
3818         return skb->len;
3819 }
3820
3821 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3822 {
3823         enum addr_type_t type = UNICAST_ADDR;
3824
3825         return inet6_dump_addr(skb, cb, type);
3826 }
3827
3828 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3829 {
3830         enum addr_type_t type = MULTICAST_ADDR;
3831
3832         return inet6_dump_addr(skb, cb, type);
3833 }
3834
3835
3836 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3837 {
3838         enum addr_type_t type = ANYCAST_ADDR;
3839
3840         return inet6_dump_addr(skb, cb, type);
3841 }
3842
3843 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3844                              void *arg)
3845 {
3846         struct net *net = sock_net(in_skb->sk);
3847         struct ifaddrmsg *ifm;
3848         struct nlattr *tb[IFA_MAX+1];
3849         struct in6_addr *addr = NULL;
3850         struct net_device *dev = NULL;
3851         struct inet6_ifaddr *ifa;
3852         struct sk_buff *skb;
3853         int err;
3854
3855         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3856         if (err < 0)
3857                 goto errout;
3858
3859         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3860         if (addr == NULL) {
3861                 err = -EINVAL;
3862                 goto errout;
3863         }
3864
3865         ifm = nlmsg_data(nlh);
3866         if (ifm->ifa_index)
3867                 dev = __dev_get_by_index(net, ifm->ifa_index);
3868
3869         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3870         if (!ifa) {
3871                 err = -EADDRNOTAVAIL;
3872                 goto errout;
3873         }
3874
3875         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3876         if (!skb) {
3877                 err = -ENOBUFS;
3878                 goto errout_ifa;
3879         }
3880
3881         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3882                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3883         if (err < 0) {
3884                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3885                 WARN_ON(err == -EMSGSIZE);
3886                 kfree_skb(skb);
3887                 goto errout_ifa;
3888         }
3889         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3890 errout_ifa:
3891         in6_ifa_put(ifa);
3892 errout:
3893         return err;
3894 }
3895
3896 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3897 {
3898         struct sk_buff *skb;
3899         struct net *net = dev_net(ifa->idev->dev);
3900         int err = -ENOBUFS;
3901
3902         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3903         if (skb == NULL)
3904                 goto errout;
3905
3906         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3907         if (err < 0) {
3908                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3909                 WARN_ON(err == -EMSGSIZE);
3910                 kfree_skb(skb);
3911                 goto errout;
3912         }
3913         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3914         return;
3915 errout:
3916         if (err < 0)
3917                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3918 }
3919
3920 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3921                                 __s32 *array, int bytes)
3922 {
3923         BUG_ON(bytes < (DEVCONF_MAX * 4));
3924
3925         memset(array, 0, bytes);
3926         array[DEVCONF_FORWARDING] = cnf->forwarding;
3927         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3928         array[DEVCONF_MTU6] = cnf->mtu6;
3929         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3930         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3931         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3932         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3933         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3934         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3935                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3936         array[DEVCONF_RTR_SOLICIT_DELAY] =
3937                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3938         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3939 #ifdef CONFIG_IPV6_PRIVACY
3940         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3941         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3942         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3943         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3944         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3945 #endif
3946         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3947         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3948         array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit;
3949         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3950 #ifdef CONFIG_IPV6_ROUTER_PREF
3951         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3952         array[DEVCONF_RTR_PROBE_INTERVAL] =
3953                 jiffies_to_msecs(cnf->rtr_probe_interval);
3954 #ifdef CONFIG_IPV6_ROUTE_INFO
3955         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3956 #endif
3957 #endif
3958         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3959         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3960 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3961         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3962 #endif
3963 #ifdef CONFIG_IPV6_MROUTE
3964         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3965 #endif
3966         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3967         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3968         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3969 }
3970
3971 static inline size_t inet6_ifla6_size(void)
3972 {
3973         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3974              + nla_total_size(sizeof(struct ifla_cacheinfo))
3975              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3976              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3977              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3978 }
3979
3980 static inline size_t inet6_if_nlmsg_size(void)
3981 {
3982         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3983                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3984                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3985                + nla_total_size(4) /* IFLA_MTU */
3986                + nla_total_size(4) /* IFLA_LINK */
3987                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3988 }
3989
3990 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3991                                       int items, int bytes)
3992 {
3993         int i;
3994         int pad = bytes - sizeof(u64) * items;
3995         BUG_ON(pad < 0);
3996
3997         /* Use put_unaligned() because stats may not be aligned for u64. */
3998         put_unaligned(items, &stats[0]);
3999         for (i = 1; i < items; i++)
4000                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
4001
4002         memset(&stats[items], 0, pad);
4003 }
4004
4005 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
4006                                       int items, int bytes, size_t syncpoff)
4007 {
4008         int i;
4009         int pad = bytes - sizeof(u64) * items;
4010         BUG_ON(pad < 0);
4011
4012         /* Use put_unaligned() because stats may not be aligned for u64. */
4013         put_unaligned(items, &stats[0]);
4014         for (i = 1; i < items; i++)
4015                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
4016
4017         memset(&stats[items], 0, pad);
4018 }
4019
4020 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
4021                              int bytes)
4022 {
4023         switch (attrtype) {
4024         case IFLA_INET6_STATS:
4025                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
4026                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
4027                 break;
4028         case IFLA_INET6_ICMP6STATS:
4029                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
4030                 break;
4031         }
4032 }
4033
4034 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
4035 {
4036         struct nlattr *nla;
4037         struct ifla_cacheinfo ci;
4038
4039         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
4040
4041         ci.max_reasm_len = IPV6_MAXPLEN;
4042         ci.tstamp = cstamp_delta(idev->tstamp);
4043         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
4044         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
4045         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
4046
4047         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
4048         if (nla == NULL)
4049                 goto nla_put_failure;
4050         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
4051
4052         /* XXX - MC not implemented */
4053
4054         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4055         if (nla == NULL)
4056                 goto nla_put_failure;
4057         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4058
4059         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4060         if (nla == NULL)
4061                 goto nla_put_failure;
4062         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4063
4064         return 0;
4065
4066 nla_put_failure:
4067         return -EMSGSIZE;
4068 }
4069
4070 static size_t inet6_get_link_af_size(const struct net_device *dev)
4071 {
4072         if (!__in6_dev_get(dev))
4073                 return 0;
4074
4075         return inet6_ifla6_size();
4076 }
4077
4078 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4079 {
4080         struct inet6_dev *idev = __in6_dev_get(dev);
4081
4082         if (!idev)
4083                 return -ENODATA;
4084
4085         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4086                 return -EMSGSIZE;
4087
4088         return 0;
4089 }
4090
4091 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4092                              u32 pid, u32 seq, int event, unsigned int flags)
4093 {
4094         struct net_device *dev = idev->dev;
4095         struct ifinfomsg *hdr;
4096         struct nlmsghdr *nlh;
4097         void *protoinfo;
4098
4099         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4100         if (nlh == NULL)
4101                 return -EMSGSIZE;
4102
4103         hdr = nlmsg_data(nlh);
4104         hdr->ifi_family = AF_INET6;
4105         hdr->__ifi_pad = 0;
4106         hdr->ifi_type = dev->type;
4107         hdr->ifi_index = dev->ifindex;
4108         hdr->ifi_flags = dev_get_flags(dev);
4109         hdr->ifi_change = 0;
4110
4111         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4112
4113         if (dev->addr_len)
4114                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4115
4116         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4117         if (dev->ifindex != dev->iflink)
4118                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4119
4120         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4121         if (protoinfo == NULL)
4122                 goto nla_put_failure;
4123
4124         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4125                 goto nla_put_failure;
4126
4127         nla_nest_end(skb, protoinfo);
4128         return nlmsg_end(skb, nlh);
4129
4130 nla_put_failure:
4131         nlmsg_cancel(skb, nlh);
4132         return -EMSGSIZE;
4133 }
4134
4135 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4136 {
4137         struct net *net = sock_net(skb->sk);
4138         int h, s_h;
4139         int idx = 0, s_idx;
4140         struct net_device *dev;
4141         struct inet6_dev *idev;
4142         struct hlist_head *head;
4143         struct hlist_node *node;
4144
4145         s_h = cb->args[0];
4146         s_idx = cb->args[1];
4147
4148         rcu_read_lock();
4149         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4150                 idx = 0;
4151                 head = &net->dev_index_head[h];
4152                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4153                         if (idx < s_idx)
4154                                 goto cont;
4155                         idev = __in6_dev_get(dev);
4156                         if (!idev)
4157                                 goto cont;
4158                         if (inet6_fill_ifinfo(skb, idev,
4159                                               NETLINK_CB(cb->skb).pid,
4160                                               cb->nlh->nlmsg_seq,
4161                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4162                                 goto out;
4163 cont:
4164                         idx++;
4165                 }
4166         }
4167 out:
4168         rcu_read_unlock();
4169         cb->args[1] = idx;
4170         cb->args[0] = h;
4171
4172         return skb->len;
4173 }
4174
4175 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4176 {
4177         struct sk_buff *skb;
4178         struct net *net = dev_net(idev->dev);
4179         int err = -ENOBUFS;
4180
4181         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4182         if (skb == NULL)
4183                 goto errout;
4184
4185         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4186         if (err < 0) {
4187                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4188                 WARN_ON(err == -EMSGSIZE);
4189                 kfree_skb(skb);
4190                 goto errout;
4191         }
4192         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4193         return;
4194 errout:
4195         if (err < 0)
4196                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4197 }
4198
4199 static inline size_t inet6_prefix_nlmsg_size(void)
4200 {
4201         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4202                + nla_total_size(sizeof(struct in6_addr))
4203                + nla_total_size(sizeof(struct prefix_cacheinfo));
4204 }
4205
4206 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4207                              struct prefix_info *pinfo, u32 pid, u32 seq,
4208                              int event, unsigned int flags)
4209 {
4210         struct prefixmsg *pmsg;
4211         struct nlmsghdr *nlh;
4212         struct prefix_cacheinfo ci;
4213
4214         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4215         if (nlh == NULL)
4216                 return -EMSGSIZE;
4217
4218         pmsg = nlmsg_data(nlh);
4219         pmsg->prefix_family = AF_INET6;
4220         pmsg->prefix_pad1 = 0;
4221         pmsg->prefix_pad2 = 0;
4222         pmsg->prefix_ifindex = idev->dev->ifindex;
4223         pmsg->prefix_len = pinfo->prefix_len;
4224         pmsg->prefix_type = pinfo->type;
4225         pmsg->prefix_pad3 = 0;
4226         pmsg->prefix_flags = 0;
4227         if (pinfo->onlink)
4228                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4229         if (pinfo->autoconf)
4230                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4231
4232         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4233
4234         ci.preferred_time = ntohl(pinfo->prefered);
4235         ci.valid_time = ntohl(pinfo->valid);
4236         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4237
4238         return nlmsg_end(skb, nlh);
4239
4240 nla_put_failure:
4241         nlmsg_cancel(skb, nlh);
4242         return -EMSGSIZE;
4243 }
4244
4245 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4246                          struct prefix_info *pinfo)
4247 {
4248         struct sk_buff *skb;
4249         struct net *net = dev_net(idev->dev);
4250         int err = -ENOBUFS;
4251
4252         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4253         if (skb == NULL)
4254                 goto errout;
4255
4256         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4257         if (err < 0) {
4258                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4259                 WARN_ON(err == -EMSGSIZE);
4260                 kfree_skb(skb);
4261                 goto errout;
4262         }
4263         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4264         return;
4265 errout:
4266         if (err < 0)
4267                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4268 }
4269
4270 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4271 {
4272         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4273
4274         switch (event) {
4275         case RTM_NEWADDR:
4276                 /*
4277                  * If the address was optimistic
4278                  * we inserted the route at the start of
4279                  * our DAD process, so we don't need
4280                  * to do it again
4281                  */
4282                 if (!(ifp->rt->rt6i_node))
4283                         ip6_ins_rt(ifp->rt);
4284                 if (ifp->idev->cnf.forwarding)
4285                         addrconf_join_anycast(ifp);
4286                 break;
4287         case RTM_DELADDR:
4288                 if (ifp->idev->cnf.forwarding)
4289                         addrconf_leave_anycast(ifp);
4290                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4291                 dst_hold(&ifp->rt->dst);
4292
4293                 if (ip6_del_rt(ifp->rt))
4294                         dst_free(&ifp->rt->dst);
4295                 break;
4296         }
4297 }
4298
4299 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4300 {
4301         rcu_read_lock_bh();
4302         if (likely(ifp->idev->dead == 0))
4303                 __ipv6_ifa_notify(event, ifp);
4304         rcu_read_unlock_bh();
4305 }
4306
4307 #ifdef CONFIG_SYSCTL
4308
4309 static
4310 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4311                            void __user *buffer, size_t *lenp, loff_t *ppos)
4312 {
4313         int *valp = ctl->data;
4314         int val = *valp;
4315         loff_t pos = *ppos;
4316         int ret;
4317
4318         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4319
4320         if (write)
4321                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4322         if (ret)
4323                 *ppos = pos;
4324         return ret;
4325 }
4326
4327 static
4328 int addrconf_sysctl_mtu(struct ctl_table *ctl, int write,
4329                         void __user *buffer, size_t *lenp, loff_t *ppos)
4330 {
4331         struct inet6_dev *idev = ctl->extra1;
4332         int min_mtu = IPV6_MIN_MTU;
4333         struct ctl_table lctl;
4334
4335         lctl = *ctl;
4336         lctl.extra1 = &min_mtu;
4337         lctl.extra2 = idev ? &idev->dev->mtu : NULL;
4338
4339         return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos);
4340 }
4341
4342 static void dev_disable_change(struct inet6_dev *idev)
4343 {
4344         if (!idev || !idev->dev)
4345                 return;
4346
4347         if (idev->cnf.disable_ipv6)
4348                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4349         else
4350                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4351 }
4352
4353 static void addrconf_disable_change(struct net *net, __s32 newf)
4354 {
4355         struct net_device *dev;
4356         struct inet6_dev *idev;
4357
4358         for_each_netdev(net, dev) {
4359                 idev = __in6_dev_get(dev);
4360                 if (idev) {
4361                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4362                         idev->cnf.disable_ipv6 = newf;
4363                         if (changed)
4364                                 dev_disable_change(idev);
4365                 }
4366         }
4367 }
4368
4369 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4370 {
4371         struct net *net;
4372
4373         net = (struct net *)table->extra2;
4374
4375         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4376                 return 0;
4377
4378         if (!rtnl_trylock()) {
4379                 /* Restore the original values before restarting */
4380                 *p = old;
4381                 return restart_syscall();
4382         }
4383
4384         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4385                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4386                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4387                 addrconf_disable_change(net, newf);
4388         } else if ((!*p) ^ (!old))
4389                 dev_disable_change((struct inet6_dev *)table->extra1);
4390
4391         rtnl_unlock();
4392         return 0;
4393 }
4394
4395 static
4396 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4397                             void __user *buffer, size_t *lenp, loff_t *ppos)
4398 {
4399         int *valp = ctl->data;
4400         int val = *valp;
4401         loff_t pos = *ppos;
4402         int ret;
4403
4404         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4405
4406         if (write)
4407                 ret = addrconf_disable_ipv6(ctl, valp, val);
4408         if (ret)
4409                 *ppos = pos;
4410         return ret;
4411 }
4412
4413 static struct addrconf_sysctl_table
4414 {
4415         struct ctl_table_header *sysctl_header;
4416         ctl_table addrconf_vars[DEVCONF_MAX+1];
4417         char *dev_name;
4418 } addrconf_sysctl __read_mostly = {
4419         .sysctl_header = NULL,
4420         .addrconf_vars = {
4421                 {
4422                         .procname       = "forwarding",
4423                         .data           = &ipv6_devconf.forwarding,
4424                         .maxlen         = sizeof(int),
4425                         .mode           = 0644,
4426                         .proc_handler   = addrconf_sysctl_forward,
4427                 },
4428                 {
4429                         .procname       = "hop_limit",
4430                         .data           = &ipv6_devconf.hop_limit,
4431                         .maxlen         = sizeof(int),
4432                         .mode           = 0644,
4433                         .proc_handler   = proc_dointvec,
4434                 },
4435                 {
4436                         .procname       = "mtu",
4437                         .data           = &ipv6_devconf.mtu6,
4438                         .maxlen         = sizeof(int),
4439                         .mode           = 0644,
4440                         .proc_handler   = addrconf_sysctl_mtu,
4441                 },
4442                 {
4443                         .procname       = "accept_ra",
4444                         .data           = &ipv6_devconf.accept_ra,
4445                         .maxlen         = sizeof(int),
4446                         .mode           = 0644,
4447                         .proc_handler   = proc_dointvec,
4448                 },
4449                 {
4450                         .procname       = "accept_redirects",
4451                         .data           = &ipv6_devconf.accept_redirects,
4452                         .maxlen         = sizeof(int),
4453                         .mode           = 0644,
4454                         .proc_handler   = proc_dointvec,
4455                 },
4456                 {
4457                         .procname       = "autoconf",
4458                         .data           = &ipv6_devconf.autoconf,
4459                         .maxlen         = sizeof(int),
4460                         .mode           = 0644,
4461                         .proc_handler   = proc_dointvec,
4462                 },
4463                 {
4464                         .procname       = "dad_transmits",
4465                         .data           = &ipv6_devconf.dad_transmits,
4466                         .maxlen         = sizeof(int),
4467                         .mode           = 0644,
4468                         .proc_handler   = proc_dointvec,
4469                 },
4470                 {
4471                         .procname       = "router_solicitations",
4472                         .data           = &ipv6_devconf.rtr_solicits,
4473                         .maxlen         = sizeof(int),
4474                         .mode           = 0644,
4475                         .proc_handler   = proc_dointvec,
4476                 },
4477                 {
4478                         .procname       = "router_solicitation_interval",
4479                         .data           = &ipv6_devconf.rtr_solicit_interval,
4480                         .maxlen         = sizeof(int),
4481                         .mode           = 0644,
4482                         .proc_handler   = proc_dointvec_jiffies,
4483                 },
4484                 {
4485                         .procname       = "router_solicitation_delay",
4486                         .data           = &ipv6_devconf.rtr_solicit_delay,
4487                         .maxlen         = sizeof(int),
4488                         .mode           = 0644,
4489                         .proc_handler   = proc_dointvec_jiffies,
4490                 },
4491                 {
4492                         .procname       = "force_mld_version",
4493                         .data           = &ipv6_devconf.force_mld_version,
4494                         .maxlen         = sizeof(int),
4495                         .mode           = 0644,
4496                         .proc_handler   = proc_dointvec,
4497                 },
4498 #ifdef CONFIG_IPV6_PRIVACY
4499                 {
4500                         .procname       = "use_tempaddr",
4501                         .data           = &ipv6_devconf.use_tempaddr,
4502                         .maxlen         = sizeof(int),
4503                         .mode           = 0644,
4504                         .proc_handler   = proc_dointvec,
4505                 },
4506                 {
4507                         .procname       = "temp_valid_lft",
4508                         .data           = &ipv6_devconf.temp_valid_lft,
4509                         .maxlen         = sizeof(int),
4510                         .mode           = 0644,
4511                         .proc_handler   = proc_dointvec,
4512                 },
4513                 {
4514                         .procname       = "temp_prefered_lft",
4515                         .data           = &ipv6_devconf.temp_prefered_lft,
4516                         .maxlen         = sizeof(int),
4517                         .mode           = 0644,
4518                         .proc_handler   = proc_dointvec,
4519                 },
4520                 {
4521                         .procname       = "regen_max_retry",
4522                         .data           = &ipv6_devconf.regen_max_retry,
4523                         .maxlen         = sizeof(int),
4524                         .mode           = 0644,
4525                         .proc_handler   = proc_dointvec,
4526                 },
4527                 {
4528                         .procname       = "max_desync_factor",
4529                         .data           = &ipv6_devconf.max_desync_factor,
4530                         .maxlen         = sizeof(int),
4531                         .mode           = 0644,
4532                         .proc_handler   = proc_dointvec,
4533                 },
4534 #endif
4535                 {
4536                         .procname       = "max_addresses",
4537                         .data           = &ipv6_devconf.max_addresses,
4538                         .maxlen         = sizeof(int),
4539                         .mode           = 0644,
4540                         .proc_handler   = proc_dointvec,
4541                 },
4542                 {
4543                         .procname       = "accept_ra_defrtr",
4544                         .data           = &ipv6_devconf.accept_ra_defrtr,
4545                         .maxlen         = sizeof(int),
4546                         .mode           = 0644,
4547                         .proc_handler   = proc_dointvec,
4548                 },
4549                 {
4550                         .procname       = "accept_ra_min_hop_limit",
4551                         .data           = &ipv6_devconf.accept_ra_min_hop_limit,
4552                         .maxlen         = sizeof(int),
4553                         .mode           = 0644,
4554                         .proc_handler   = proc_dointvec,
4555                 },
4556                 {
4557                         .procname       = "accept_ra_pinfo",
4558                         .data           = &ipv6_devconf.accept_ra_pinfo,
4559                         .maxlen         = sizeof(int),
4560                         .mode           = 0644,
4561                         .proc_handler   = proc_dointvec,
4562                 },
4563 #ifdef CONFIG_IPV6_ROUTER_PREF
4564                 {
4565                         .procname       = "accept_ra_rtr_pref",
4566                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4567                         .maxlen         = sizeof(int),
4568                         .mode           = 0644,
4569                         .proc_handler   = proc_dointvec,
4570                 },
4571                 {
4572                         .procname       = "router_probe_interval",
4573                         .data           = &ipv6_devconf.rtr_probe_interval,
4574                         .maxlen         = sizeof(int),
4575                         .mode           = 0644,
4576                         .proc_handler   = proc_dointvec_jiffies,
4577                 },
4578 #ifdef CONFIG_IPV6_ROUTE_INFO
4579                 {
4580                         .procname       = "accept_ra_rt_info_max_plen",
4581                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4582                         .maxlen         = sizeof(int),
4583                         .mode           = 0644,
4584                         .proc_handler   = proc_dointvec,
4585                 },
4586 #endif
4587 #endif
4588                 {
4589                         .procname       = "proxy_ndp",
4590                         .data           = &ipv6_devconf.proxy_ndp,
4591                         .maxlen         = sizeof(int),
4592                         .mode           = 0644,
4593                         .proc_handler   = proc_dointvec,
4594                 },
4595                 {
4596                         .procname       = "accept_source_route",
4597                         .data           = &ipv6_devconf.accept_source_route,
4598                         .maxlen         = sizeof(int),
4599                         .mode           = 0644,
4600                         .proc_handler   = proc_dointvec,
4601                 },
4602 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4603                 {
4604                         .procname       = "optimistic_dad",
4605                         .data           = &ipv6_devconf.optimistic_dad,
4606                         .maxlen         = sizeof(int),
4607                         .mode           = 0644,
4608                         .proc_handler   = proc_dointvec,
4609
4610                 },
4611 #endif
4612 #ifdef CONFIG_IPV6_MROUTE
4613                 {
4614                         .procname       = "mc_forwarding",
4615                         .data           = &ipv6_devconf.mc_forwarding,
4616                         .maxlen         = sizeof(int),
4617                         .mode           = 0444,
4618                         .proc_handler   = proc_dointvec,
4619                 },
4620 #endif
4621                 {
4622                         .procname       = "disable_ipv6",
4623                         .data           = &ipv6_devconf.disable_ipv6,
4624                         .maxlen         = sizeof(int),
4625                         .mode           = 0644,
4626                         .proc_handler   = addrconf_sysctl_disable,
4627                 },
4628                 {
4629                         .procname       = "accept_dad",
4630                         .data           = &ipv6_devconf.accept_dad,
4631                         .maxlen         = sizeof(int),
4632                         .mode           = 0644,
4633                         .proc_handler   = proc_dointvec,
4634                 },
4635                 {
4636                         .procname       = "force_tllao",
4637                         .data           = &ipv6_devconf.force_tllao,
4638                         .maxlen         = sizeof(int),
4639                         .mode           = 0644,
4640                         .proc_handler   = proc_dointvec
4641                 },
4642                 {
4643                         /* sentinel */
4644                 }
4645         },
4646 };
4647
4648 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4649                 struct inet6_dev *idev, struct ipv6_devconf *p)
4650 {
4651         int i;
4652         struct addrconf_sysctl_table *t;
4653
4654 #define ADDRCONF_CTL_PATH_DEV   3
4655
4656         struct ctl_path addrconf_ctl_path[] = {
4657                 { .procname = "net", },
4658                 { .procname = "ipv6", },
4659                 { .procname = "conf", },
4660                 { /* to be set */ },
4661                 { },
4662         };
4663
4664
4665         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4666         if (t == NULL)
4667                 goto out;
4668
4669         for (i = 0; t->addrconf_vars[i].data; i++) {
4670                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4671                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4672                 t->addrconf_vars[i].extra2 = net;
4673         }
4674
4675         /*
4676          * Make a copy of dev_name, because '.procname' is regarded as const
4677          * by sysctl and we wouldn't want anyone to change it under our feet
4678          * (see SIOCSIFNAME).
4679          */
4680         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4681         if (!t->dev_name)
4682                 goto free;
4683
4684         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4685
4686         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4687                         t->addrconf_vars);
4688         if (t->sysctl_header == NULL)
4689                 goto free_procname;
4690
4691         p->sysctl = t;
4692         return 0;
4693
4694 free_procname:
4695         kfree(t->dev_name);
4696 free:
4697         kfree(t);
4698 out:
4699         return -ENOBUFS;
4700 }
4701
4702 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4703 {
4704         struct addrconf_sysctl_table *t;
4705
4706         if (p->sysctl == NULL)
4707                 return;
4708
4709         t = p->sysctl;
4710         p->sysctl = NULL;
4711         unregister_net_sysctl_table(t->sysctl_header);
4712         kfree(t->dev_name);
4713         kfree(t);
4714 }
4715
4716 static void addrconf_sysctl_register(struct inet6_dev *idev)
4717 {
4718         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4719                               &ndisc_ifinfo_sysctl_change);
4720         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4721                                         idev, &idev->cnf);
4722 }
4723
4724 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4725 {
4726         __addrconf_sysctl_unregister(&idev->cnf);
4727         neigh_sysctl_unregister(idev->nd_parms);
4728 }
4729
4730
4731 #endif
4732
4733 static int __net_init addrconf_init_net(struct net *net)
4734 {
4735         int err = -ENOMEM;
4736         struct ipv6_devconf *all, *dflt;
4737
4738         all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL);
4739         if (all == NULL)
4740                 goto err_alloc_all;
4741
4742         dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4743         if (dflt == NULL)
4744                 goto err_alloc_dflt;
4745
4746         /* these will be inherited by all namespaces */
4747         dflt->autoconf = ipv6_defaults.autoconf;
4748         dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4749
4750         net->ipv6.devconf_all = all;
4751         net->ipv6.devconf_dflt = dflt;
4752
4753 #ifdef CONFIG_SYSCTL
4754         err = __addrconf_sysctl_register(net, "all", NULL, all);
4755         if (err < 0)
4756                 goto err_reg_all;
4757
4758         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4759         if (err < 0)
4760                 goto err_reg_dflt;
4761 #endif
4762         return 0;
4763
4764 #ifdef CONFIG_SYSCTL
4765 err_reg_dflt:
4766         __addrconf_sysctl_unregister(all);
4767 err_reg_all:
4768         kfree(dflt);
4769 #endif
4770 err_alloc_dflt:
4771         kfree(all);
4772 err_alloc_all:
4773         return err;
4774 }
4775
4776 static void __net_exit addrconf_exit_net(struct net *net)
4777 {
4778 #ifdef CONFIG_SYSCTL
4779         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4780         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4781 #endif
4782         if (!net_eq(net, &init_net)) {
4783                 kfree(net->ipv6.devconf_dflt);
4784                 kfree(net->ipv6.devconf_all);
4785         }
4786 }
4787
4788 static struct pernet_operations addrconf_ops = {
4789         .init = addrconf_init_net,
4790         .exit = addrconf_exit_net,
4791 };
4792
4793 /*
4794  *      Device notifier
4795  */
4796
4797 int register_inet6addr_notifier(struct notifier_block *nb)
4798 {
4799         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4800 }
4801 EXPORT_SYMBOL(register_inet6addr_notifier);
4802
4803 int unregister_inet6addr_notifier(struct notifier_block *nb)
4804 {
4805         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4806 }
4807 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4808
4809 static struct rtnl_af_ops inet6_ops = {
4810         .family           = AF_INET6,
4811         .fill_link_af     = inet6_fill_link_af,
4812         .get_link_af_size = inet6_get_link_af_size,
4813 };
4814
4815 /*
4816  *      Init / cleanup code
4817  */
4818
4819 int __init addrconf_init(void)
4820 {
4821         int i, err;
4822
4823         err = ipv6_addr_label_init();
4824         if (err < 0) {
4825                 printk(KERN_CRIT "IPv6 Addrconf:"
4826                        " cannot initialize default policy table: %d.\n", err);
4827                 goto out;
4828         }
4829
4830         err = register_pernet_subsys(&addrconf_ops);
4831         if (err < 0)
4832                 goto out_addrlabel;
4833
4834         /* The addrconf netdev notifier requires that loopback_dev
4835          * has it's ipv6 private information allocated and setup
4836          * before it can bring up and give link-local addresses
4837          * to other devices which are up.
4838          *
4839          * Unfortunately, loopback_dev is not necessarily the first
4840          * entry in the global dev_base list of net devices.  In fact,
4841          * it is likely to be the very last entry on that list.
4842          * So this causes the notifier registry below to try and
4843          * give link-local addresses to all devices besides loopback_dev
4844          * first, then loopback_dev, which cases all the non-loopback_dev
4845          * devices to fail to get a link-local address.
4846          *
4847          * So, as a temporary fix, allocate the ipv6 structure for
4848          * loopback_dev first by hand.
4849          * Longer term, all of the dependencies ipv6 has upon the loopback
4850          * device and it being up should be removed.
4851          */
4852         rtnl_lock();
4853         if (!ipv6_add_dev(init_net.loopback_dev))
4854                 err = -ENOMEM;
4855         rtnl_unlock();
4856         if (err)
4857                 goto errlo;
4858
4859         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4860                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4861
4862         register_netdevice_notifier(&ipv6_dev_notf);
4863
4864         addrconf_verify(0);
4865
4866         err = rtnl_af_register(&inet6_ops);
4867         if (err < 0)
4868                 goto errout_af;
4869
4870         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4871                               NULL);
4872         if (err < 0)
4873                 goto errout;
4874
4875         /* Only the first call to __rtnl_register can fail */
4876         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4877         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4878         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4879                         inet6_dump_ifaddr, NULL);
4880         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4881                         inet6_dump_ifmcaddr, NULL);
4882         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4883                         inet6_dump_ifacaddr, NULL);
4884
4885         ipv6_addr_label_rtnl_register();
4886
4887         return 0;
4888 errout:
4889         rtnl_af_unregister(&inet6_ops);
4890 errout_af:
4891         unregister_netdevice_notifier(&ipv6_dev_notf);
4892 errlo:
4893         unregister_pernet_subsys(&addrconf_ops);
4894 out_addrlabel:
4895         ipv6_addr_label_cleanup();
4896 out:
4897         return err;
4898 }
4899
4900 void addrconf_cleanup(void)
4901 {
4902         struct net_device *dev;
4903         int i;
4904
4905         unregister_netdevice_notifier(&ipv6_dev_notf);
4906         unregister_pernet_subsys(&addrconf_ops);
4907         ipv6_addr_label_cleanup();
4908
4909         rtnl_lock();
4910
4911         __rtnl_af_unregister(&inet6_ops);
4912
4913         /* clean dev list */
4914         for_each_netdev(&init_net, dev) {
4915                 if (__in6_dev_get(dev) == NULL)
4916                         continue;
4917                 addrconf_ifdown(dev, 1);
4918         }
4919         addrconf_ifdown(init_net.loopback_dev, 2);
4920
4921         /*
4922          *      Check hash table.
4923          */
4924         spin_lock_bh(&addrconf_hash_lock);
4925         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4926                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4927         spin_unlock_bh(&addrconf_hash_lock);
4928
4929         del_timer(&addr_chk_timer);
4930         rtnl_unlock();
4931 }