net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101
 102 static inline u32 cstamp_delta(unsigned long cstamp)
 103 {
 104         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 105 }
 106
 107 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 108 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 109 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 110
 111 #ifdef CONFIG_SYSCTL
 112 static void addrconf_sysctl_register(struct inet6_dev *idev);
 113 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 114 #else
 115 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 116 {
 117 }
 118
 119 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 120 {
 121 }
 122 #endif
 123
 124 #ifdef CONFIG_IPV6_PRIVACY
 125 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 126 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 127 static void ipv6_regen_rndid(unsigned long data);
 128 #endif
 129
 130 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 131 static int ipv6_count_addresses(struct inet6_dev *idev);
 132
 133 /*
 134  *      Configured unicast address hash table
 135  */
 136 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 137 static DEFINE_SPINLOCK(addrconf_hash_lock);
 138
 139 static void addrconf_verify(unsigned long);
 140
 141 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 142 static DEFINE_SPINLOCK(addrconf_verify_lock);
 143
 144 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 145 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 146
 147 static void addrconf_type_change(struct net_device *dev,
 148                                  unsigned long event);
 149 static int addrconf_ifdown(struct net_device *dev, int how);
 150
 151 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 152 static void addrconf_dad_timer(unsigned long data);
 153 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 154 static void addrconf_dad_run(struct inet6_dev *idev);
 155 static void addrconf_rs_timer(unsigned long data);
 156 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 157 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158
 159 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 160                                 struct prefix_info *pinfo);
 161 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 162                                struct net_device *dev);
 163
 164 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 165
 166 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 167         .forwarding             = 0,
 168         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 169         .mtu6                   = IPV6_MIN_MTU,
 170         .accept_ra              = 1,
 171         .accept_redirects       = 1,
 172         .autoconf               = 1,
 173         .force_mld_version      = 0,
 174         .dad_transmits          = 1,
 175         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 176         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 177         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 178 #ifdef CONFIG_IPV6_PRIVACY
 179         .use_tempaddr           = 0,
 180         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 181         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 182         .regen_max_retry        = REGEN_MAX_RETRY,
 183         .max_desync_factor      = MAX_DESYNC_FACTOR,
 184 #endif
 185         .max_addresses          = IPV6_MAX_ADDRESSES,
 186         .accept_ra_defrtr       = 1,
 187         .accept_ra_pinfo        = 1,
 188 #ifdef CONFIG_IPV6_ROUTER_PREF
 189         .accept_ra_rtr_pref     = 1,
 190         .rtr_probe_interval     = 60 * HZ,
 191 #ifdef CONFIG_IPV6_ROUTE_INFO
 192         .accept_ra_rt_info_max_plen = 0,
 193 #endif
 194 #endif
 195         .proxy_ndp              = 0,
 196         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 197         .disable_ipv6           = 0,
 198         .accept_dad             = 1,
 199 };
 200
 201 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 202         .forwarding             = 0,
 203         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 204         .mtu6                   = IPV6_MIN_MTU,
 205         .accept_ra              = 1,
 206         .accept_redirects       = 1,
 207         .autoconf               = 1,
 208         .dad_transmits          = 1,
 209         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 210         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 211         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 212 #ifdef CONFIG_IPV6_PRIVACY
 213         .use_tempaddr           = 0,
 214         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 215         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 216         .regen_max_retry        = REGEN_MAX_RETRY,
 217         .max_desync_factor      = MAX_DESYNC_FACTOR,
 218 #endif
 219         .max_addresses          = IPV6_MAX_ADDRESSES,
 220         .accept_ra_defrtr       = 1,
 221         .accept_ra_pinfo        = 1,
 222 #ifdef CONFIG_IPV6_ROUTER_PREF
 223         .accept_ra_rtr_pref     = 1,
 224         .rtr_probe_interval     = 60 * HZ,
 225 #ifdef CONFIG_IPV6_ROUTE_INFO
 226         .accept_ra_rt_info_max_plen = 0,
 227 #endif
 228 #endif
 229         .proxy_ndp              = 0,
 230         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 231         .disable_ipv6           = 0,
 232         .accept_dad             = 1,
 233 };
 234
 235 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 236 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 237 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 238 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 239 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 240
 241 /* Check if a valid qdisc is available */
 242 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 243 {
 244         return !qdisc_tx_is_noop(dev);
 245 }
 246
 247 /* Check if a route is valid prefix route */
 248 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 249 {
 250         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 251 }
 252
 253 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 254 {
 255         if (del_timer(&ifp->timer))
 256                 __in6_ifa_put(ifp);
 257 }
 258
 259 enum addrconf_timer_t {
 260         AC_NONE,
 261         AC_DAD,
 262         AC_RS,
 263 };
 264
 265 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 266                                enum addrconf_timer_t what,
 267                                unsigned long when)
 268 {
 269         if (!del_timer(&ifp->timer))
 270                 in6_ifa_hold(ifp);
 271
 272         switch (what) {
 273         case AC_DAD:
 274                 ifp->timer.function = addrconf_dad_timer;
 275                 break;
 276         case AC_RS:
 277                 ifp->timer.function = addrconf_rs_timer;
 278                 break;
 279         default:
 280                 break;
 281         }
 282         ifp->timer.expires = jiffies + when;
 283         add_timer(&ifp->timer);
 284 }
 285
 286 static int snmp6_alloc_dev(struct inet6_dev *idev)
 287 {
 288         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 289                           sizeof(struct ipstats_mib),
 290                           __alignof__(struct ipstats_mib)) < 0)
 291                 goto err_ip;
 292         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 293                           sizeof(struct icmpv6_mib),
 294                           __alignof__(struct icmpv6_mib)) < 0)
 295                 goto err_icmp;
 296         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 297                           sizeof(struct icmpv6msg_mib),
 298                           __alignof__(struct icmpv6msg_mib)) < 0)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305 err_icmp:
 306         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 307 err_ip:
 308         return -ENOMEM;
 309 }
 310
 311 static void snmp6_free_dev(struct inet6_dev *idev)
 312 {
 313         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 314         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 void in6_dev_finish_destroy(struct inet6_dev *idev)
 321 {
 322         struct net_device *dev = idev->dev;
 323
 324         WARN_ON(!list_empty(&idev->addr_list));
 325         WARN_ON(idev->mc_list != NULL);
 326
 327 #ifdef NET_REFCNT_DEBUG
 328         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 329 #endif
 330         dev_put(dev);
 331         if (!idev->dead) {
 332                 pr_warning("Freeing alive inet6 device %p\n", idev);
 333                 return;
 334         }
 335         snmp6_free_dev(idev);
 336         kfree_rcu(idev, rcu);
 337 }
 338
 339 EXPORT_SYMBOL(in6_dev_finish_destroy);
 340
 341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 342 {
 343         struct inet6_dev *ndev;
 344
 345         ASSERT_RTNL();
 346
 347         if (dev->mtu < IPV6_MIN_MTU)
 348                 return NULL;
 349
 350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 351
 352         if (ndev == NULL)
 353                 return NULL;
 354
 355         rwlock_init(&ndev->lock);
 356         ndev->dev = dev;
 357         INIT_LIST_HEAD(&ndev->addr_list);
 358
 359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 360         ndev->cnf.mtu6 = dev->mtu;
 361         ndev->cnf.sysctl = NULL;
 362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 363         if (ndev->nd_parms == NULL) {
 364                 kfree(ndev);
 365                 return NULL;
 366         }
 367         if (ndev->cnf.forwarding)
 368                 dev_disable_lro(dev);
 369         /* We refer to the device */
 370         dev_hold(dev);
 371
 372         if (snmp6_alloc_dev(ndev) < 0) {
 373                 ADBG((KERN_WARNING
 374                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 375                         __func__, dev->name));
 376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 377                 ndev->dead = 1;
 378                 in6_dev_finish_destroy(ndev);
 379                 return NULL;
 380         }
 381
 382         if (snmp6_register_dev(ndev) < 0) {
 383                 ADBG((KERN_WARNING
 384                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 385                         __func__, dev->name));
 386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 387                 ndev->dead = 1;
 388                 in6_dev_finish_destroy(ndev);
 389                 return NULL;
 390         }
 391
 392         /* One reference from device.  We must do this before
 393          * we invoke __ipv6_regen_rndid().
 394          */
 395         in6_dev_hold(ndev);
 396
 397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 398                 ndev->cnf.accept_dad = -1;
 399
 400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 402                 printk(KERN_INFO
 403                        "%s: Disabled Multicast RS\n",
 404                        dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423
 424         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 425                 ndev->if_flags |= IF_READY;
 426
 427         ipv6_mc_init_dev(ndev);
 428         ndev->tstamp = jiffies;
 429         addrconf_sysctl_register(ndev);
 430         /* protected by rtnl_lock */
 431         rcu_assign_pointer(dev->ip6_ptr, ndev);
 432
 433         /* Join all-node multicast group */
 434         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 435
 436         return ndev;
 437 }
 438
 439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 440 {
 441         struct inet6_dev *idev;
 442
 443         ASSERT_RTNL();
 444
 445         idev = __in6_dev_get(dev);
 446         if (!idev) {
 447                 idev = ipv6_add_dev(dev);
 448                 if (!idev)
 449                         return NULL;
 450         }
 451
 452         if (dev->flags&IFF_UP)
 453                 ipv6_mc_up(idev);
 454         return idev;
 455 }
 456
 457 #ifdef CONFIG_SYSCTL
 458 static void dev_forward_change(struct inet6_dev *idev)
 459 {
 460         struct net_device *dev;
 461         struct inet6_ifaddr *ifa;
 462
 463         if (!idev)
 464                 return;
 465         dev = idev->dev;
 466         if (idev->cnf.forwarding)
 467                 dev_disable_lro(dev);
 468         if (dev && (dev->flags & IFF_MULTICAST)) {
 469                 if (idev->cnf.forwarding)
 470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471                 else
 472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 473         }
 474
 475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 476                 if (ifa->flags&IFA_F_TENTATIVE)
 477                         continue;
 478                 if (idev->cnf.forwarding)
 479                         addrconf_join_anycast(ifa);
 480                 else
 481                         addrconf_leave_anycast(ifa);
 482         }
 483 }
 484
 485
 486 static void addrconf_forward_change(struct net *net, __s32 newf)
 487 {
 488         struct net_device *dev;
 489         struct inet6_dev *idev;
 490
 491         rcu_read_lock();
 492         for_each_netdev_rcu(net, dev) {
 493                 idev = __in6_dev_get(dev);
 494                 if (idev) {
 495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 496                         idev->cnf.forwarding = newf;
 497                         if (changed)
 498                                 dev_forward_change(idev);
 499                 }
 500         }
 501         rcu_read_unlock();
 502 }
 503
 504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 505 {
 506         struct net *net;
 507
 508         net = (struct net *)table->extra2;
 509         if (p == &net->ipv6.devconf_dflt->forwarding)
 510                 return 0;
 511
 512         if (!rtnl_trylock()) {
 513                 /* Restore the original values before restarting */
 514                 *p = old;
 515                 return restart_syscall();
 516         }
 517
 518         if (p == &net->ipv6.devconf_all->forwarding) {
 519                 __s32 newf = net->ipv6.devconf_all->forwarding;
 520                 net->ipv6.devconf_dflt->forwarding = newf;
 521                 addrconf_forward_change(net, newf);
 522         } else if ((!*p) ^ (!old))
 523                 dev_forward_change((struct inet6_dev *)table->extra1);
 524         rtnl_unlock();
 525
 526         if (*p)
 527                 rt6_purge_dflt_routers(net);
 528         return 1;
 529 }
 530 #endif
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 536
 537 #ifdef NET_REFCNT_DEBUG
 538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 539 #endif
 540
 541         in6_dev_put(ifp->idev);
 542
 543         if (del_timer(&ifp->timer))
 544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 545
 546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 548                 return;
 549         }
 550         dst_release(&ifp->rt->dst);
 551
 552         kfree_rcu(ifp, rcu);
 553 }
 554
 555 static void
 556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 557 {
 558         struct list_head *p;
 559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 560
 561         /*
 562          * Each device address list is sorted in order of scope -
 563          * global before linklocal.
 564          */
 565         list_for_each(p, &idev->addr_list) {
 566                 struct inet6_ifaddr *ifa
 567                         = list_entry(p, struct inet6_ifaddr, if_list);
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         list_add_tail(&ifp->if_list, p);
 573 }
 574
 575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 576 {
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581         return jhash_2words((__force u32)addr->s6_addr32[2],
 582                             (__force u32)addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         /*
 653          * part one of RFC 4429, section 3.3
 654          * We should not configure an address as
 655          * optimistic if we do not yet know the link
 656          * layer address of our nexhop router
 657          */
 658
 659         if (rt->rt6i_nexthop == NULL)
 660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 661
 662         ifa->idev = idev;
 663         in6_dev_hold(idev);
 664         /* For caller */
 665         in6_ifa_hold(ifa);
 666
 667         /* Add to big hash table */
 668         hash = ipv6_addr_hash(addr);
 669
 670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, *ifn;
 707         struct inet6_dev *idev = ifp->idev;
 708         int state;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         spin_lock_bh(&ifp->state_lock);
 713         state = ifp->state;
 714         ifp->state = INET6_IFADDR_STATE_DEAD;
 715         spin_unlock_bh(&ifp->state_lock);
 716
 717         if (state == INET6_IFADDR_STATE_DEAD)
 718                 goto out;
 719
 720         spin_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init_rcu(&ifp->addr_lst);
 722         spin_unlock_bh(&addrconf_hash_lock);
 723
 724         write_lock_bh(&idev->lock);
 725 #ifdef CONFIG_IPV6_PRIVACY
 726         if (ifp->flags&IFA_F_TEMPORARY) {
 727                 list_del(&ifp->tmp_list);
 728                 if (ifp->ifpub) {
 729                         in6_ifa_put(ifp->ifpub);
 730                         ifp->ifpub = NULL;
 731                 }
 732                 __in6_ifa_put(ifp);
 733         }
 734 #endif
 735
 736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 737                 if (ifa == ifp) {
 738                         list_del_init(&ifp->if_list);
 739                         __in6_ifa_put(ifp);
 740
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773         }
 774         write_unlock_bh(&idev->lock);
 775
 776         addrconf_del_timer(ifp);
 777
 778         ipv6_ifa_notify(RTM_DELADDR, ifp);
 779
 780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 781
 782         /*
 783          * Purge or update corresponding prefix
 784          *
 785          * 1) we don't purge prefix here if address was not permanent.
 786          *    prefix is managed by its own lifetime.
 787          * 2) if there're no addresses, delete prefix.
 788          * 3) if there're still other permanent address(es),
 789          *    corresponding prefix is still permanent.
 790          * 4) otherwise, update prefix lifetime to the
 791          *    longest valid lifetime among the corresponding
 792          *    addresses on the device.
 793          *    Note: subsequent RA will update lifetime.
 794          *
 795          * --yoshfuji
 796          */
 797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 798                 struct in6_addr prefix;
 799                 struct rt6_info *rt;
 800                 struct net *net = dev_net(ifp->idev->dev);
 801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 803
 804                 if (rt && addrconf_is_prefix_route(rt)) {
 805                         if (onlink == 0) {
 806                                 ip6_del_rt(rt);
 807                                 rt = NULL;
 808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 809                                 rt->rt6i_expires = expires;
 810                                 rt->rt6i_flags |= RTF_EXPIRES;
 811                         }
 812                 }
 813                 dst_release(&rt->dst);
 814         }
 815
 816 out:
 817         in6_ifa_put(ifp);
 818 }
 819
 820 #ifdef CONFIG_IPV6_PRIVACY
 821 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 822 {
 823         struct inet6_dev *idev = ifp->idev;
 824         struct in6_addr addr, *tmpaddr;
 825         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 826         unsigned long regen_advance;
 827         int tmp_plen;
 828         int ret = 0;
 829         int max_addresses;
 830         u32 addr_flags;
 831
 832         write_lock(&idev->lock);
 833         if (ift) {
 834                 spin_lock_bh(&ift->lock);
 835                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 836                 spin_unlock_bh(&ift->lock);
 837                 tmpaddr = &addr;
 838         } else {
 839                 tmpaddr = NULL;
 840         }
 841 retry:
 842         in6_dev_hold(idev);
 843         if (idev->cnf.use_tempaddr <= 0) {
 844                 write_unlock(&idev->lock);
 845                 printk(KERN_INFO
 846                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 847                 in6_dev_put(idev);
 848                 ret = -1;
 849                 goto out;
 850         }
 851         spin_lock_bh(&ifp->lock);
 852         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 853                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 854                 spin_unlock_bh(&ifp->lock);
 855                 write_unlock(&idev->lock);
 856                 printk(KERN_WARNING
 857                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 858                 in6_dev_put(idev);
 859                 ret = -1;
 860                 goto out;
 861         }
 862         in6_ifa_hold(ifp);
 863         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 864         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 865                 spin_unlock_bh(&ifp->lock);
 866                 write_unlock(&idev->lock);
 867                 printk(KERN_WARNING
 868                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 869                 in6_ifa_put(ifp);
 870                 in6_dev_put(idev);
 871                 ret = -1;
 872                 goto out;
 873         }
 874         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 875         age = (jiffies - ifp->tstamp) / HZ;
 876         tmp_valid_lft = min_t(__u32,
 877                               ifp->valid_lft,
 878                               idev->cnf.temp_valid_lft + age);
 879         tmp_prefered_lft = min_t(__u32,
 880                                  ifp->prefered_lft,
 881                                  idev->cnf.temp_prefered_lft + age -
 882                                  idev->cnf.max_desync_factor);
 883         tmp_plen = ifp->prefix_len;
 884         max_addresses = idev->cnf.max_addresses;
 885         tmp_cstamp = ifp->cstamp;
 886         tmp_tstamp = ifp->tstamp;
 887         spin_unlock_bh(&ifp->lock);
 888
 889         regen_advance = idev->cnf.regen_max_retry *
 890                         idev->cnf.dad_transmits *
 891                         idev->nd_parms->retrans_time / HZ;
 892         write_unlock(&idev->lock);
 893
 894         /* A temporary address is created only if this calculated Preferred
 895          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 896          * an implementation must not create a temporary address with a zero
 897          * Preferred Lifetime.
 898          */
 899         if (tmp_prefered_lft <= regen_advance) {
 900                 in6_ifa_put(ifp);
 901                 in6_dev_put(idev);
 902                 ret = -1;
 903                 goto out;
 904         }
 905
 906         addr_flags = IFA_F_TEMPORARY;
 907         /* set in addrconf_prefix_rcv() */
 908         if (ifp->flags & IFA_F_OPTIMISTIC)
 909                 addr_flags |= IFA_F_OPTIMISTIC;
 910
 911         ift = !max_addresses ||
 912               ipv6_count_addresses(idev) < max_addresses ?
 913                 ipv6_add_addr(idev, &addr, tmp_plen,
 914                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 915                               addr_flags) : NULL;
 916         if (!ift || IS_ERR(ift)) {
 917                 in6_ifa_put(ifp);
 918                 in6_dev_put(idev);
 919                 printk(KERN_INFO
 920                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 921                 tmpaddr = &addr;
 922                 write_lock(&idev->lock);
 923                 goto retry;
 924         }
 925
 926         spin_lock_bh(&ift->lock);
 927         ift->ifpub = ifp;
 928         ift->valid_lft = tmp_valid_lft;
 929         ift->prefered_lft = tmp_prefered_lft;
 930         ift->cstamp = tmp_cstamp;
 931         ift->tstamp = tmp_tstamp;
 932         spin_unlock_bh(&ift->lock);
 933
 934         addrconf_dad_start(ift, 0);
 935         in6_ifa_put(ift);
 936         in6_dev_put(idev);
 937 out:
 938         return ret;
 939 }
 940 #endif
 941
 942 /*
 943  *      Choose an appropriate source address (RFC3484)
 944  */
 945 enum {
 946         IPV6_SADDR_RULE_INIT = 0,
 947         IPV6_SADDR_RULE_LOCAL,
 948         IPV6_SADDR_RULE_SCOPE,
 949         IPV6_SADDR_RULE_PREFERRED,
 950 #ifdef CONFIG_IPV6_MIP6
 951         IPV6_SADDR_RULE_HOA,
 952 #endif
 953         IPV6_SADDR_RULE_OIF,
 954         IPV6_SADDR_RULE_LABEL,
 955 #ifdef CONFIG_IPV6_PRIVACY
 956         IPV6_SADDR_RULE_PRIVACY,
 957 #endif
 958         IPV6_SADDR_RULE_ORCHID,
 959         IPV6_SADDR_RULE_PREFIX,
 960         IPV6_SADDR_RULE_MAX
 961 };
 962
 963 struct ipv6_saddr_score {
 964         int                     rule;
 965         int                     addr_type;
 966         struct inet6_ifaddr     *ifa;
 967         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 968         int                     scopedist;
 969         int                     matchlen;
 970 };
 971
 972 struct ipv6_saddr_dst {
 973         const struct in6_addr *addr;
 974         int ifindex;
 975         int scope;
 976         int label;
 977         unsigned int prefs;
 978 };
 979
 980 static inline int ipv6_saddr_preferred(int type)
 981 {
 982         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 983                 return 1;
 984         return 0;
 985 }
 986
 987 static int ipv6_get_saddr_eval(struct net *net,
 988                                struct ipv6_saddr_score *score,
 989                                struct ipv6_saddr_dst *dst,
 990                                int i)
 991 {
 992         int ret;
 993
 994         if (i <= score->rule) {
 995                 switch (i) {
 996                 case IPV6_SADDR_RULE_SCOPE:
 997                         ret = score->scopedist;
 998                         break;
 999                 case IPV6_SADDR_RULE_PREFIX:
1000                         ret = score->matchlen;
1001                         break;
1002                 default:
1003                         ret = !!test_bit(i, score->scorebits);
1004                 }
1005                 goto out;
1006         }
1007
1008         switch (i) {
1009         case IPV6_SADDR_RULE_INIT:
1010                 /* Rule 0: remember if hiscore is not ready yet */
1011                 ret = !!score->ifa;
1012                 break;
1013         case IPV6_SADDR_RULE_LOCAL:
1014                 /* Rule 1: Prefer same address */
1015                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1016                 break;
1017         case IPV6_SADDR_RULE_SCOPE:
1018                 /* Rule 2: Prefer appropriate scope
1019                  *
1020                  *      ret
1021                  *       ^
1022                  *    -1 |  d 15
1023                  *    ---+--+-+---> scope
1024                  *       |
1025                  *       |             d is scope of the destination.
1026                  *  B-d  |  \
1027                  *       |   \      <- smaller scope is better if
1028                  *  B-15 |    \        if scope is enough for destinaion.
1029                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1030                  * d-C-1 | /
1031                  *       |/         <- greater is better
1032                  *   -C  /             if scope is not enough for destination.
1033                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1034                  *
1035                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1036                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1037                  * Assume B = 0 and we get C > 29.
1038                  */
1039                 ret = __ipv6_addr_src_scope(score->addr_type);
1040                 if (ret >= dst->scope)
1041                         ret = -ret;
1042                 else
1043                         ret -= 128;     /* 30 is enough */
1044                 score->scopedist = ret;
1045                 break;
1046         case IPV6_SADDR_RULE_PREFERRED:
1047                 /* Rule 3: Avoid deprecated and optimistic addresses */
1048                 ret = ipv6_saddr_preferred(score->addr_type) ||
1049                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1050                 break;
1051 #ifdef CONFIG_IPV6_MIP6
1052         case IPV6_SADDR_RULE_HOA:
1053             {
1054                 /* Rule 4: Prefer home address */
1055                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1056                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1057                 break;
1058             }
1059 #endif
1060         case IPV6_SADDR_RULE_OIF:
1061                 /* Rule 5: Prefer outgoing interface */
1062                 ret = (!dst->ifindex ||
1063                        dst->ifindex == score->ifa->idev->dev->ifindex);
1064                 break;
1065         case IPV6_SADDR_RULE_LABEL:
1066                 /* Rule 6: Prefer matching label */
1067                 ret = ipv6_addr_label(net,
1068                                       &score->ifa->addr, score->addr_type,
1069                                       score->ifa->idev->dev->ifindex) == dst->label;
1070                 break;
1071 #ifdef CONFIG_IPV6_PRIVACY
1072         case IPV6_SADDR_RULE_PRIVACY:
1073             {
1074                 /* Rule 7: Prefer public address
1075                  * Note: prefer temporary address if use_tempaddr >= 2
1076                  */
1077                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1078                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1079                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1080                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1081                 break;
1082             }
1083 #endif
1084         case IPV6_SADDR_RULE_ORCHID:
1085                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1086                  *          non-ORCHID vs non-ORCHID
1087                  */
1088                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1089                         ipv6_addr_orchid(dst->addr));
1090                 break;
1091         case IPV6_SADDR_RULE_PREFIX:
1092                 /* Rule 8: Use longest matching prefix */
1093                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1094                                                        dst->addr);
1095                 break;
1096         default:
1097                 ret = 0;
1098         }
1099
1100         if (ret)
1101                 __set_bit(i, score->scorebits);
1102         score->rule = i;
1103 out:
1104         return ret;
1105 }
1106
1107 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1108                        const struct in6_addr *daddr, unsigned int prefs,
1109                        struct in6_addr *saddr)
1110 {
1111         struct ipv6_saddr_score scores[2],
1112                                 *score = &scores[0], *hiscore = &scores[1];
1113         struct ipv6_saddr_dst dst;
1114         struct net_device *dev;
1115         int dst_type;
1116
1117         dst_type = __ipv6_addr_type(daddr);
1118         dst.addr = daddr;
1119         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1120         dst.scope = __ipv6_addr_src_scope(dst_type);
1121         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1122         dst.prefs = prefs;
1123
1124         hiscore->rule = -1;
1125         hiscore->ifa = NULL;
1126
1127         rcu_read_lock();
1128
1129         for_each_netdev_rcu(net, dev) {
1130                 struct inet6_dev *idev;
1131
1132                 /* Candidate Source Address (section 4)
1133                  *  - multicast and link-local destination address,
1134                  *    the set of candidate source address MUST only
1135                  *    include addresses assigned to interfaces
1136                  *    belonging to the same link as the outgoing
1137                  *    interface.
1138                  * (- For site-local destination addresses, the
1139                  *    set of candidate source addresses MUST only
1140                  *    include addresses assigned to interfaces
1141                  *    belonging to the same site as the outgoing
1142                  *    interface.)
1143                  */
1144                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1145                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1146                     dst.ifindex && dev->ifindex != dst.ifindex)
1147                         continue;
1148
1149                 idev = __in6_dev_get(dev);
1150                 if (!idev)
1151                         continue;
1152
1153                 read_lock_bh(&idev->lock);
1154                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1155                         int i;
1156
1157                         /*
1158                          * - Tentative Address (RFC2462 section 5.4)
1159                          *  - A tentative address is not considered
1160                          *    "assigned to an interface" in the traditional
1161                          *    sense, unless it is also flagged as optimistic.
1162                          * - Candidate Source Address (section 4)
1163                          *  - In any case, anycast addresses, multicast
1164                          *    addresses, and the unspecified address MUST
1165                          *    NOT be included in a candidate set.
1166                          */
1167                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1168                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1169                                 continue;
1170
1171                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1172
1173                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1174                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1175                                 LIMIT_NETDEBUG(KERN_DEBUG
1176                                                "ADDRCONF: unspecified / multicast address "
1177                                                "assigned as unicast address on %s",
1178                                                dev->name);
1179                                 continue;
1180                         }
1181
1182                         score->rule = -1;
1183                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1184
1185                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1186                                 int minihiscore, miniscore;
1187
1188                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1189                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1190
1191                                 if (minihiscore > miniscore) {
1192                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1193                                             score->scopedist > 0) {
1194                                                 /*
1195                                                  * special case:
1196                                                  * each remaining entry
1197                                                  * has too small (not enough)
1198                                                  * scope, because ifa entries
1199                                                  * are sorted by their scope
1200                                                  * values.
1201                                                  */
1202                                                 goto try_nextdev;
1203                                         }
1204                                         break;
1205                                 } else if (minihiscore < miniscore) {
1206                                         if (hiscore->ifa)
1207                                                 in6_ifa_put(hiscore->ifa);
1208
1209                                         in6_ifa_hold(score->ifa);
1210
1211                                         swap(hiscore, score);
1212
1213                                         /* restore our iterator */
1214                                         score->ifa = hiscore->ifa;
1215
1216                                         break;
1217                                 }
1218                         }
1219                 }
1220 try_nextdev:
1221                 read_unlock_bh(&idev->lock);
1222         }
1223         rcu_read_unlock();
1224
1225         if (!hiscore->ifa)
1226                 return -EADDRNOTAVAIL;
1227
1228         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1229         in6_ifa_put(hiscore->ifa);
1230         return 0;
1231 }
1232 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1233
1234 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1235                     unsigned char banned_flags)
1236 {
1237         struct inet6_dev *idev;
1238         int err = -EADDRNOTAVAIL;
1239
1240         rcu_read_lock();
1241         idev = __in6_dev_get(dev);
1242         if (idev) {
1243                 struct inet6_ifaddr *ifp;
1244
1245                 read_lock_bh(&idev->lock);
1246                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1247                         if (ifp->scope == IFA_LINK &&
1248                             !(ifp->flags & banned_flags)) {
1249                                 ipv6_addr_copy(addr, &ifp->addr);
1250                                 err = 0;
1251                                 break;
1252                         }
1253                 }
1254                 read_unlock_bh(&idev->lock);
1255         }
1256         rcu_read_unlock();
1257         return err;
1258 }
1259
1260 static int ipv6_count_addresses(struct inet6_dev *idev)
1261 {
1262         int cnt = 0;
1263         struct inet6_ifaddr *ifp;
1264
1265         read_lock_bh(&idev->lock);
1266         list_for_each_entry(ifp, &idev->addr_list, if_list)
1267                 cnt++;
1268         read_unlock_bh(&idev->lock);
1269         return cnt;
1270 }
1271
1272 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1273                   struct net_device *dev, int strict)
1274 {
1275         struct inet6_ifaddr *ifp;
1276         struct hlist_node *node;
1277         unsigned int hash = ipv6_addr_hash(addr);
1278
1279         rcu_read_lock_bh();
1280         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1281                 if (!net_eq(dev_net(ifp->idev->dev), net))
1282                         continue;
1283                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1284                     !(ifp->flags&IFA_F_TENTATIVE) &&
1285                     (dev == NULL || ifp->idev->dev == dev ||
1286                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1287                         rcu_read_unlock_bh();
1288                         return 1;
1289                 }
1290         }
1291
1292         rcu_read_unlock_bh();
1293         return 0;
1294 }
1295 EXPORT_SYMBOL(ipv6_chk_addr);
1296
1297 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1298                                struct net_device *dev)
1299 {
1300         unsigned int hash = ipv6_addr_hash(addr);
1301         struct inet6_ifaddr *ifp;
1302         struct hlist_node *node;
1303
1304         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1305                 if (!net_eq(dev_net(ifp->idev->dev), net))
1306                         continue;
1307                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1308                         if (dev == NULL || ifp->idev->dev == dev)
1309                                 return true;
1310                 }
1311         }
1312         return false;
1313 }
1314
1315 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1316 {
1317         struct inet6_dev *idev;
1318         struct inet6_ifaddr *ifa;
1319         int     onlink;
1320
1321         onlink = 0;
1322         rcu_read_lock();
1323         idev = __in6_dev_get(dev);
1324         if (idev) {
1325                 read_lock_bh(&idev->lock);
1326                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1327                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1328                                                    ifa->prefix_len);
1329                         if (onlink)
1330                                 break;
1331                 }
1332                 read_unlock_bh(&idev->lock);
1333         }
1334         rcu_read_unlock();
1335         return onlink;
1336 }
1337
1338 EXPORT_SYMBOL(ipv6_chk_prefix);
1339
1340 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1341                                      struct net_device *dev, int strict)
1342 {
1343         struct inet6_ifaddr *ifp, *result = NULL;
1344         unsigned int hash = ipv6_addr_hash(addr);
1345         struct hlist_node *node;
1346
1347         rcu_read_lock_bh();
1348         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1349                 if (!net_eq(dev_net(ifp->idev->dev), net))
1350                         continue;
1351                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1352                         if (dev == NULL || ifp->idev->dev == dev ||
1353                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1354                                 result = ifp;
1355                                 in6_ifa_hold(ifp);
1356                                 break;
1357                         }
1358                 }
1359         }
1360         rcu_read_unlock_bh();
1361
1362         return result;
1363 }
1364
1365 /* Gets referenced address, destroys ifaddr */
1366
1367 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1368 {
1369         if (ifp->flags&IFA_F_PERMANENT) {
1370                 spin_lock_bh(&ifp->lock);
1371                 addrconf_del_timer(ifp);
1372                 ifp->flags |= IFA_F_TENTATIVE;
1373                 if (dad_failed)
1374                         ifp->flags |= IFA_F_DADFAILED;
1375                 spin_unlock_bh(&ifp->lock);
1376                 if (dad_failed)
1377                         ipv6_ifa_notify(0, ifp);
1378                 in6_ifa_put(ifp);
1379 #ifdef CONFIG_IPV6_PRIVACY
1380         } else if (ifp->flags&IFA_F_TEMPORARY) {
1381                 struct inet6_ifaddr *ifpub;
1382                 spin_lock_bh(&ifp->lock);
1383                 ifpub = ifp->ifpub;
1384                 if (ifpub) {
1385                         in6_ifa_hold(ifpub);
1386                         spin_unlock_bh(&ifp->lock);
1387                         ipv6_create_tempaddr(ifpub, ifp);
1388                         in6_ifa_put(ifpub);
1389                 } else {
1390                         spin_unlock_bh(&ifp->lock);
1391                 }
1392                 ipv6_del_addr(ifp);
1393 #endif
1394         } else
1395                 ipv6_del_addr(ifp);
1396 }
1397
1398 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1399 {
1400         int err = -ENOENT;
1401
1402         spin_lock(&ifp->state_lock);
1403         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1404                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1405                 err = 0;
1406         }
1407         spin_unlock(&ifp->state_lock);
1408
1409         return err;
1410 }
1411
1412 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1413 {
1414         struct inet6_dev *idev = ifp->idev;
1415
1416         if (addrconf_dad_end(ifp)) {
1417                 in6_ifa_put(ifp);
1418                 return;
1419         }
1420
1421         if (net_ratelimit())
1422                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1423                         ifp->idev->dev->name, &ifp->addr);
1424
1425         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1426                 struct in6_addr addr;
1427
1428                 addr.s6_addr32[0] = htonl(0xfe800000);
1429                 addr.s6_addr32[1] = 0;
1430
1431                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1432                     ipv6_addr_equal(&ifp->addr, &addr)) {
1433                         /* DAD failed for link-local based on MAC address */
1434                         idev->cnf.disable_ipv6 = 1;
1435
1436                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1437                                 ifp->idev->dev->name);
1438                 }
1439         }
1440
1441         addrconf_dad_stop(ifp, 1);
1442 }
1443
1444 /* Join to solicited addr multicast group. */
1445
1446 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1447 {
1448         struct in6_addr maddr;
1449
1450         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1451                 return;
1452
1453         addrconf_addr_solict_mult(addr, &maddr);
1454         ipv6_dev_mc_inc(dev, &maddr);
1455 }
1456
1457 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1458 {
1459         struct in6_addr maddr;
1460
1461         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1462                 return;
1463
1464         addrconf_addr_solict_mult(addr, &maddr);
1465         __ipv6_dev_mc_dec(idev, &maddr);
1466 }
1467
1468 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1469 {
1470         struct in6_addr addr;
1471         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1472         if (ipv6_addr_any(&addr))
1473                 return;
1474         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1475 }
1476
1477 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1478 {
1479         struct in6_addr addr;
1480         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1481         if (ipv6_addr_any(&addr))
1482                 return;
1483         __ipv6_dev_ac_dec(ifp->idev, &addr);
1484 }
1485
1486 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1487 {
1488         if (dev->addr_len != ETH_ALEN)
1489                 return -1;
1490         memcpy(eui, dev->dev_addr, 3);
1491         memcpy(eui + 5, dev->dev_addr + 3, 3);
1492
1493         /*
1494          * The zSeries OSA network cards can be shared among various
1495          * OS instances, but the OSA cards have only one MAC address.
1496          * This leads to duplicate address conflicts in conjunction
1497          * with IPv6 if more than one instance uses the same card.
1498          *
1499          * The driver for these cards can deliver a unique 16-bit
1500          * identifier for each instance sharing the same card.  It is
1501          * placed instead of 0xFFFE in the interface identifier.  The
1502          * "u" bit of the interface identifier is not inverted in this
1503          * case.  Hence the resulting interface identifier has local
1504          * scope according to RFC2373.
1505          */
1506         if (dev->dev_id) {
1507                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1508                 eui[4] = dev->dev_id & 0xFF;
1509         } else {
1510                 eui[3] = 0xFF;
1511                 eui[4] = 0xFE;
1512                 eui[0] ^= 2;
1513         }
1514         return 0;
1515 }
1516
1517 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1518 {
1519         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1520         if (dev->addr_len != ARCNET_ALEN)
1521                 return -1;
1522         memset(eui, 0, 7);
1523         eui[7] = *(u8*)dev->dev_addr;
1524         return 0;
1525 }
1526
1527 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1528 {
1529         if (dev->addr_len != INFINIBAND_ALEN)
1530                 return -1;
1531         memcpy(eui, dev->dev_addr + 12, 8);
1532         eui[0] |= 2;
1533         return 0;
1534 }
1535
1536 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1537 {
1538         if (addr == 0)
1539                 return -1;
1540         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1541                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1542                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1543                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1544                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1545                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1546         eui[1] = 0;
1547         eui[2] = 0x5E;
1548         eui[3] = 0xFE;
1549         memcpy(eui + 4, &addr, 4);
1550         return 0;
1551 }
1552
1553 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1554 {
1555         if (dev->priv_flags & IFF_ISATAP)
1556                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1557         return -1;
1558 }
1559
1560 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1561 {
1562         switch (dev->type) {
1563         case ARPHRD_ETHER:
1564         case ARPHRD_FDDI:
1565         case ARPHRD_IEEE802_TR:
1566                 return addrconf_ifid_eui48(eui, dev);
1567         case ARPHRD_ARCNET:
1568                 return addrconf_ifid_arcnet(eui, dev);
1569         case ARPHRD_INFINIBAND:
1570                 return addrconf_ifid_infiniband(eui, dev);
1571         case ARPHRD_SIT:
1572                 return addrconf_ifid_sit(eui, dev);
1573         }
1574         return -1;
1575 }
1576
1577 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1578 {
1579         int err = -1;
1580         struct inet6_ifaddr *ifp;
1581
1582         read_lock_bh(&idev->lock);
1583         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1584                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1585                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1586                         err = 0;
1587                         break;
1588                 }
1589         }
1590         read_unlock_bh(&idev->lock);
1591         return err;
1592 }
1593
1594 #ifdef CONFIG_IPV6_PRIVACY
1595 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1596 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1597 {
1598 regen:
1599         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1600         idev->rndid[0] &= ~0x02;
1601
1602         /*
1603          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1604          * check if generated address is not inappropriate
1605          *
1606          *  - Reserved subnet anycast (RFC 2526)
1607          *      11111101 11....11 1xxxxxxx
1608          *  - ISATAP (RFC4214) 6.1
1609          *      00-00-5E-FE-xx-xx-xx-xx
1610          *  - value 0
1611          *  - XXX: already assigned to an address on the device
1612          */
1613         if (idev->rndid[0] == 0xfd &&
1614             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1615             (idev->rndid[7]&0x80))
1616                 goto regen;
1617         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1618                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1619                         goto regen;
1620                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1621                         goto regen;
1622         }
1623
1624         return 0;
1625 }
1626
1627 static void ipv6_regen_rndid(unsigned long data)
1628 {
1629         struct inet6_dev *idev = (struct inet6_dev *) data;
1630         unsigned long expires;
1631
1632         rcu_read_lock_bh();
1633         write_lock_bh(&idev->lock);
1634
1635         if (idev->dead)
1636                 goto out;
1637
1638         if (__ipv6_regen_rndid(idev) < 0)
1639                 goto out;
1640
1641         expires = jiffies +
1642                 idev->cnf.temp_prefered_lft * HZ -
1643                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1644                 idev->cnf.max_desync_factor * HZ;
1645         if (time_before(expires, jiffies)) {
1646                 printk(KERN_WARNING
1647                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1648                         idev->dev->name);
1649                 goto out;
1650         }
1651
1652         if (!mod_timer(&idev->regen_timer, expires))
1653                 in6_dev_hold(idev);
1654
1655 out:
1656         write_unlock_bh(&idev->lock);
1657         rcu_read_unlock_bh();
1658         in6_dev_put(idev);
1659 }
1660
1661 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1662         int ret = 0;
1663
1664         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1665                 ret = __ipv6_regen_rndid(idev);
1666         return ret;
1667 }
1668 #endif
1669
1670 /*
1671  *      Add prefix route.
1672  */
1673
1674 static void
1675 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1676                       unsigned long expires, u32 flags)
1677 {
1678         struct fib6_config cfg = {
1679                 .fc_table = RT6_TABLE_PREFIX,
1680                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1681                 .fc_ifindex = dev->ifindex,
1682                 .fc_expires = expires,
1683                 .fc_dst_len = plen,
1684                 .fc_flags = RTF_UP | flags,
1685                 .fc_nlinfo.nl_net = dev_net(dev),
1686                 .fc_protocol = RTPROT_KERNEL,
1687         };
1688
1689         ipv6_addr_copy(&cfg.fc_dst, pfx);
1690
1691         /* Prevent useless cloning on PtP SIT.
1692            This thing is done here expecting that the whole
1693            class of non-broadcast devices need not cloning.
1694          */
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1697                 cfg.fc_flags |= RTF_NONEXTHOP;
1698 #endif
1699
1700         ip6_route_add(&cfg);
1701 }
1702
1703 /* Create "default" multicast route to the interface */
1704
1705 static void addrconf_add_mroute(struct net_device *dev)
1706 {
1707         struct fib6_config cfg = {
1708                 .fc_table = RT6_TABLE_LOCAL,
1709                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1710                 .fc_ifindex = dev->ifindex,
1711                 .fc_dst_len = 8,
1712                 .fc_flags = RTF_UP,
1713                 .fc_nlinfo.nl_net = dev_net(dev),
1714         };
1715
1716         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1717
1718         ip6_route_add(&cfg);
1719 }
1720
1721 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1722 static void sit_route_add(struct net_device *dev)
1723 {
1724         struct fib6_config cfg = {
1725                 .fc_table = RT6_TABLE_MAIN,
1726                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1727                 .fc_ifindex = dev->ifindex,
1728                 .fc_dst_len = 96,
1729                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1730                 .fc_nlinfo.nl_net = dev_net(dev),
1731         };
1732
1733         /* prefix length - 96 bits "::d.d.d.d" */
1734         ip6_route_add(&cfg);
1735 }
1736 #endif
1737
1738 static void addrconf_add_lroute(struct net_device *dev)
1739 {
1740         struct in6_addr addr;
1741
1742         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1743         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1744 }
1745
1746 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1747 {
1748         struct inet6_dev *idev;
1749
1750         ASSERT_RTNL();
1751
1752         idev = ipv6_find_idev(dev);
1753         if (!idev)
1754                 return ERR_PTR(-ENOBUFS);
1755
1756         if (idev->cnf.disable_ipv6)
1757                 return ERR_PTR(-EACCES);
1758
1759         /* Add default multicast route */
1760         addrconf_add_mroute(dev);
1761
1762         /* Add link local route */
1763         addrconf_add_lroute(dev);
1764         return idev;
1765 }
1766
1767 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1768 {
1769         struct prefix_info *pinfo;
1770         __u32 valid_lft;
1771         __u32 prefered_lft;
1772         int addr_type;
1773         struct inet6_dev *in6_dev;
1774         struct net *net = dev_net(dev);
1775
1776         pinfo = (struct prefix_info *) opt;
1777
1778         if (len < sizeof(struct prefix_info)) {
1779                 ADBG(("addrconf: prefix option too short\n"));
1780                 return;
1781         }
1782
1783         /*
1784          *      Validation checks ([ADDRCONF], page 19)
1785          */
1786
1787         addr_type = ipv6_addr_type(&pinfo->prefix);
1788
1789         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1790                 return;
1791
1792         valid_lft = ntohl(pinfo->valid);
1793         prefered_lft = ntohl(pinfo->prefered);
1794
1795         if (prefered_lft > valid_lft) {
1796                 if (net_ratelimit())
1797                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1798                 return;
1799         }
1800
1801         in6_dev = in6_dev_get(dev);
1802
1803         if (in6_dev == NULL) {
1804                 if (net_ratelimit())
1805                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1806                 return;
1807         }
1808
1809         /*
1810          *      Two things going on here:
1811          *      1) Add routes for on-link prefixes
1812          *      2) Configure prefixes with the auto flag set
1813          */
1814
1815         if (pinfo->onlink) {
1816                 struct rt6_info *rt;
1817                 unsigned long rt_expires;
1818
1819                 /* Avoid arithmetic overflow. Really, we could
1820                  * save rt_expires in seconds, likely valid_lft,
1821                  * but it would require division in fib gc, that it
1822                  * not good.
1823                  */
1824                 if (HZ > USER_HZ)
1825                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1826                 else
1827                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1828
1829                 if (addrconf_finite_timeout(rt_expires))
1830                         rt_expires *= HZ;
1831
1832                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1833                                 dev->ifindex, 1);
1834
1835                 if (rt && addrconf_is_prefix_route(rt)) {
1836                         /* Autoconf prefix route */
1837                         if (valid_lft == 0) {
1838                                 ip6_del_rt(rt);
1839                                 rt = NULL;
1840                         } else if (addrconf_finite_timeout(rt_expires)) {
1841                                 /* not infinity */
1842                                 rt->rt6i_expires = jiffies + rt_expires;
1843                                 rt->rt6i_flags |= RTF_EXPIRES;
1844                         } else {
1845                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1846                                 rt->rt6i_expires = 0;
1847                         }
1848                 } else if (valid_lft) {
1849                         clock_t expires = 0;
1850                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1851                         if (addrconf_finite_timeout(rt_expires)) {
1852                                 /* not infinity */
1853                                 flags |= RTF_EXPIRES;
1854                                 expires = jiffies_to_clock_t(rt_expires);
1855                         }
1856                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1857                                               dev, expires, flags);
1858                 }
1859                 if (rt)
1860                         dst_release(&rt->dst);
1861         }
1862
1863         /* Try to figure out our local address for this prefix */
1864
1865         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1866                 struct inet6_ifaddr * ifp;
1867                 struct in6_addr addr;
1868                 int create = 0, update_lft = 0;
1869
1870                 if (pinfo->prefix_len == 64) {
1871                         memcpy(&addr, &pinfo->prefix, 8);
1872                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1873                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1874                                 in6_dev_put(in6_dev);
1875                                 return;
1876                         }
1877                         goto ok;
1878                 }
1879                 if (net_ratelimit())
1880                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1881                                pinfo->prefix_len);
1882                 in6_dev_put(in6_dev);
1883                 return;
1884
1885 ok:
1886
1887                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1888
1889                 if (ifp == NULL && valid_lft) {
1890                         int max_addresses = in6_dev->cnf.max_addresses;
1891                         u32 addr_flags = 0;
1892
1893 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1894                         if (in6_dev->cnf.optimistic_dad &&
1895                             !net->ipv6.devconf_all->forwarding)
1896                                 addr_flags = IFA_F_OPTIMISTIC;
1897 #endif
1898
1899                         /* Do not allow to create too much of autoconfigured
1900                          * addresses; this would be too easy way to crash kernel.
1901                          */
1902                         if (!max_addresses ||
1903                             ipv6_count_addresses(in6_dev) < max_addresses)
1904                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1905                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1906                                                     addr_flags);
1907
1908                         if (!ifp || IS_ERR(ifp)) {
1909                                 in6_dev_put(in6_dev);
1910                                 return;
1911                         }
1912
1913                         update_lft = create = 1;
1914                         ifp->cstamp = jiffies;
1915                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1916                 }
1917
1918                 if (ifp) {
1919                         int flags;
1920                         unsigned long now;
1921 #ifdef CONFIG_IPV6_PRIVACY
1922                         struct inet6_ifaddr *ift;
1923 #endif
1924                         u32 stored_lft;
1925
1926                         /* update lifetime (RFC2462 5.5.3 e) */
1927                         spin_lock(&ifp->lock);
1928                         now = jiffies;
1929                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1930                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1931                         else
1932                                 stored_lft = 0;
1933                         if (!update_lft && stored_lft) {
1934                                 if (valid_lft > MIN_VALID_LIFETIME ||
1935                                     valid_lft > stored_lft)
1936                                         update_lft = 1;
1937                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1938                                         /* valid_lft <= stored_lft is always true */
1939                                         /*
1940                                          * RFC 4862 Section 5.5.3e:
1941                                          * "Note that the preferred lifetime of
1942                                          *  the corresponding address is always
1943                                          *  reset to the Preferred Lifetime in
1944                                          *  the received Prefix Information
1945                                          *  option, regardless of whether the
1946                                          *  valid lifetime is also reset or
1947                                          *  ignored."
1948                                          *
1949                                          *  So if the preferred lifetime in
1950                                          *  this advertisement is different
1951                                          *  than what we have stored, but the
1952                                          *  valid lifetime is invalid, just
1953                                          *  reset prefered_lft.
1954                                          *
1955                                          *  We must set the valid lifetime
1956                                          *  to the stored lifetime since we'll
1957                                          *  be updating the timestamp below,
1958                                          *  else we'll set it back to the
1959                                          *  minimum.
1960                                          */
1961                                         if (prefered_lft != ifp->prefered_lft) {
1962                                                 valid_lft = stored_lft;
1963                                                 update_lft = 1;
1964                                         }
1965                                 } else {
1966                                         valid_lft = MIN_VALID_LIFETIME;
1967                                         if (valid_lft < prefered_lft)
1968                                                 prefered_lft = valid_lft;
1969                                         update_lft = 1;
1970                                 }
1971                         }
1972
1973                         if (update_lft) {
1974                                 ifp->valid_lft = valid_lft;
1975                                 ifp->prefered_lft = prefered_lft;
1976                                 ifp->tstamp = now;
1977                                 flags = ifp->flags;
1978                                 ifp->flags &= ~IFA_F_DEPRECATED;
1979                                 spin_unlock(&ifp->lock);
1980
1981                                 if (!(flags&IFA_F_TENTATIVE))
1982                                         ipv6_ifa_notify(0, ifp);
1983                         } else
1984                                 spin_unlock(&ifp->lock);
1985
1986 #ifdef CONFIG_IPV6_PRIVACY
1987                         read_lock_bh(&in6_dev->lock);
1988                         /* update all temporary addresses in the list */
1989                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1990                                 /*
1991                                  * When adjusting the lifetimes of an existing
1992                                  * temporary address, only lower the lifetimes.
1993                                  * Implementations must not increase the
1994                                  * lifetimes of an existing temporary address
1995                                  * when processing a Prefix Information Option.
1996                                  */
1997                                 if (ifp != ift->ifpub)
1998                                         continue;
1999
2000                                 spin_lock(&ift->lock);
2001                                 flags = ift->flags;
2002                                 if (ift->valid_lft > valid_lft &&
2003                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2004                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2005                                 if (ift->prefered_lft > prefered_lft &&
2006                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2007                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2008                                 spin_unlock(&ift->lock);
2009                                 if (!(flags&IFA_F_TENTATIVE))
2010                                         ipv6_ifa_notify(0, ift);
2011                         }
2012
2013                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2014                                 /*
2015                                  * When a new public address is created as described in [ADDRCONF],
2016                                  * also create a new temporary address. Also create a temporary
2017                                  * address if it's enabled but no temporary address currently exists.
2018                                  */
2019                                 read_unlock_bh(&in6_dev->lock);
2020                                 ipv6_create_tempaddr(ifp, NULL);
2021                         } else {
2022                                 read_unlock_bh(&in6_dev->lock);
2023                         }
2024 #endif
2025                         in6_ifa_put(ifp);
2026                         addrconf_verify(0);
2027                 }
2028         }
2029         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2030         in6_dev_put(in6_dev);
2031 }
2032
2033 /*
2034  *      Set destination address.
2035  *      Special case for SIT interfaces where we create a new "virtual"
2036  *      device.
2037  */
2038 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2039 {
2040         struct in6_ifreq ireq;
2041         struct net_device *dev;
2042         int err = -EINVAL;
2043
2044         rtnl_lock();
2045
2046         err = -EFAULT;
2047         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2048                 goto err_exit;
2049
2050         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2051
2052         err = -ENODEV;
2053         if (dev == NULL)
2054                 goto err_exit;
2055
2056 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2057         if (dev->type == ARPHRD_SIT) {
2058                 const struct net_device_ops *ops = dev->netdev_ops;
2059                 struct ifreq ifr;
2060                 struct ip_tunnel_parm p;
2061
2062                 err = -EADDRNOTAVAIL;
2063                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2064                         goto err_exit;
2065
2066                 memset(&p, 0, sizeof(p));
2067                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2068                 p.iph.saddr = 0;
2069                 p.iph.version = 4;
2070                 p.iph.ihl = 5;
2071                 p.iph.protocol = IPPROTO_IPV6;
2072                 p.iph.ttl = 64;
2073                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2074
2075                 if (ops->ndo_do_ioctl) {
2076                         mm_segment_t oldfs = get_fs();
2077
2078                         set_fs(KERNEL_DS);
2079                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2080                         set_fs(oldfs);
2081                 } else
2082                         err = -EOPNOTSUPP;
2083
2084                 if (err == 0) {
2085                         err = -ENOBUFS;
2086                         dev = __dev_get_by_name(net, p.name);
2087                         if (!dev)
2088                                 goto err_exit;
2089                         err = dev_open(dev);
2090                 }
2091         }
2092 #endif
2093
2094 err_exit:
2095         rtnl_unlock();
2096         return err;
2097 }
2098
2099 /*
2100  *      Manual configuration of address on an interface
2101  */
2102 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2103                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2104                           __u32 valid_lft)
2105 {
2106         struct inet6_ifaddr *ifp;
2107         struct inet6_dev *idev;
2108         struct net_device *dev;
2109         int scope;
2110         u32 flags;
2111         clock_t expires;
2112         unsigned long timeout;
2113
2114         ASSERT_RTNL();
2115
2116         if (plen > 128)
2117                 return -EINVAL;
2118
2119         /* check the lifetime */
2120         if (!valid_lft || prefered_lft > valid_lft)
2121                 return -EINVAL;
2122
2123         dev = __dev_get_by_index(net, ifindex);
2124         if (!dev)
2125                 return -ENODEV;
2126
2127         idev = addrconf_add_dev(dev);
2128         if (IS_ERR(idev))
2129                 return PTR_ERR(idev);
2130
2131         scope = ipv6_addr_scope(pfx);
2132
2133         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2134         if (addrconf_finite_timeout(timeout)) {
2135                 expires = jiffies_to_clock_t(timeout * HZ);
2136                 valid_lft = timeout;
2137                 flags = RTF_EXPIRES;
2138         } else {
2139                 expires = 0;
2140                 flags = 0;
2141                 ifa_flags |= IFA_F_PERMANENT;
2142         }
2143
2144         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2145         if (addrconf_finite_timeout(timeout)) {
2146                 if (timeout == 0)
2147                         ifa_flags |= IFA_F_DEPRECATED;
2148                 prefered_lft = timeout;
2149         }
2150
2151         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2152
2153         if (!IS_ERR(ifp)) {
2154                 spin_lock_bh(&ifp->lock);
2155                 ifp->valid_lft = valid_lft;
2156                 ifp->prefered_lft = prefered_lft;
2157                 ifp->tstamp = jiffies;
2158                 spin_unlock_bh(&ifp->lock);
2159
2160                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2161                                       expires, flags);
2162                 /*
2163                  * Note that section 3.1 of RFC 4429 indicates
2164                  * that the Optimistic flag should not be set for
2165                  * manually configured addresses
2166                  */
2167                 addrconf_dad_start(ifp, 0);
2168                 in6_ifa_put(ifp);
2169                 addrconf_verify(0);
2170                 return 0;
2171         }
2172
2173         return PTR_ERR(ifp);
2174 }
2175
2176 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2177                           unsigned int plen)
2178 {
2179         struct inet6_ifaddr *ifp;
2180         struct inet6_dev *idev;
2181         struct net_device *dev;
2182
2183         if (plen > 128)
2184                 return -EINVAL;
2185
2186         dev = __dev_get_by_index(net, ifindex);
2187         if (!dev)
2188                 return -ENODEV;
2189
2190         if ((idev = __in6_dev_get(dev)) == NULL)
2191                 return -ENXIO;
2192
2193         read_lock_bh(&idev->lock);
2194         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2195                 if (ifp->prefix_len == plen &&
2196                     ipv6_addr_equal(pfx, &ifp->addr)) {
2197                         in6_ifa_hold(ifp);
2198                         read_unlock_bh(&idev->lock);
2199
2200                         ipv6_del_addr(ifp);
2201
2202                         /* If the last address is deleted administratively,
2203                            disable IPv6 on this interface.
2204                          */
2205                         if (list_empty(&idev->addr_list))
2206                                 addrconf_ifdown(idev->dev, 1);
2207                         return 0;
2208                 }
2209         }
2210         read_unlock_bh(&idev->lock);
2211         return -EADDRNOTAVAIL;
2212 }
2213
2214
2215 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2216 {
2217         struct in6_ifreq ireq;
2218         int err;
2219
2220         if (!capable(CAP_NET_ADMIN))
2221                 return -EPERM;
2222
2223         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2224                 return -EFAULT;
2225
2226         rtnl_lock();
2227         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2228                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2229                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2230         rtnl_unlock();
2231         return err;
2232 }
2233
2234 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2235 {
2236         struct in6_ifreq ireq;
2237         int err;
2238
2239         if (!capable(CAP_NET_ADMIN))
2240                 return -EPERM;
2241
2242         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2243                 return -EFAULT;
2244
2245         rtnl_lock();
2246         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2247                              ireq.ifr6_prefixlen);
2248         rtnl_unlock();
2249         return err;
2250 }
2251
2252 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2253                      int plen, int scope)
2254 {
2255         struct inet6_ifaddr *ifp;
2256
2257         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2258         if (!IS_ERR(ifp)) {
2259                 spin_lock_bh(&ifp->lock);
2260                 ifp->flags &= ~IFA_F_TENTATIVE;
2261                 spin_unlock_bh(&ifp->lock);
2262                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2263                 in6_ifa_put(ifp);
2264         }
2265 }
2266
2267 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2268 static void sit_add_v4_addrs(struct inet6_dev *idev)
2269 {
2270         struct in6_addr addr;
2271         struct net_device *dev;
2272         struct net *net = dev_net(idev->dev);
2273         int scope;
2274
2275         ASSERT_RTNL();
2276
2277         memset(&addr, 0, sizeof(struct in6_addr));
2278         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2279
2280         if (idev->dev->flags&IFF_POINTOPOINT) {
2281                 addr.s6_addr32[0] = htonl(0xfe800000);
2282                 scope = IFA_LINK;
2283         } else {
2284                 scope = IPV6_ADDR_COMPATv4;
2285         }
2286
2287         if (addr.s6_addr32[3]) {
2288                 add_addr(idev, &addr, 128, scope);
2289                 return;
2290         }
2291
2292         for_each_netdev(net, dev) {
2293                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2294                 if (in_dev && (dev->flags & IFF_UP)) {
2295                         struct in_ifaddr * ifa;
2296
2297                         int flag = scope;
2298
2299                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2300                                 int plen;
2301
2302                                 addr.s6_addr32[3] = ifa->ifa_local;
2303
2304                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2305                                         continue;
2306                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2307                                         if (idev->dev->flags&IFF_POINTOPOINT)
2308                                                 continue;
2309                                         flag |= IFA_HOST;
2310                                 }
2311                                 if (idev->dev->flags&IFF_POINTOPOINT)
2312                                         plen = 64;
2313                                 else
2314                                         plen = 96;
2315
2316                                 add_addr(idev, &addr, plen, flag);
2317                         }
2318                 }
2319         }
2320 }
2321 #endif
2322
2323 static void init_loopback(struct net_device *dev)
2324 {
2325         struct inet6_dev  *idev;
2326
2327         /* ::1 */
2328
2329         ASSERT_RTNL();
2330
2331         if ((idev = ipv6_find_idev(dev)) == NULL) {
2332                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2333                 return;
2334         }
2335
2336         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2337 }
2338
2339 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2340 {
2341         struct inet6_ifaddr * ifp;
2342         u32 addr_flags = IFA_F_PERMANENT;
2343
2344 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2345         if (idev->cnf.optimistic_dad &&
2346             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2347                 addr_flags |= IFA_F_OPTIMISTIC;
2348 #endif
2349
2350
2351         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2352         if (!IS_ERR(ifp)) {
2353                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2354                 addrconf_dad_start(ifp, 0);
2355                 in6_ifa_put(ifp);
2356         }
2357 }
2358
2359 static void addrconf_dev_config(struct net_device *dev)
2360 {
2361         struct in6_addr addr;
2362         struct inet6_dev    * idev;
2363
2364         ASSERT_RTNL();
2365
2366         if ((dev->type != ARPHRD_ETHER) &&
2367             (dev->type != ARPHRD_FDDI) &&
2368             (dev->type != ARPHRD_IEEE802_TR) &&
2369             (dev->type != ARPHRD_ARCNET) &&
2370             (dev->type != ARPHRD_INFINIBAND)) {
2371                 /* Alas, we support only Ethernet autoconfiguration. */
2372                 return;
2373         }
2374
2375         idev = addrconf_add_dev(dev);
2376         if (IS_ERR(idev))
2377                 return;
2378
2379         memset(&addr, 0, sizeof(struct in6_addr));
2380         addr.s6_addr32[0] = htonl(0xFE800000);
2381
2382         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2383                 addrconf_add_linklocal(idev, &addr);
2384 }
2385
2386 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2387 static void addrconf_sit_config(struct net_device *dev)
2388 {
2389         struct inet6_dev *idev;
2390
2391         ASSERT_RTNL();
2392
2393         /*
2394          * Configure the tunnel with one of our IPv4
2395          * addresses... we should configure all of
2396          * our v4 addrs in the tunnel
2397          */
2398
2399         if ((idev = ipv6_find_idev(dev)) == NULL) {
2400                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2401                 return;
2402         }
2403
2404         if (dev->priv_flags & IFF_ISATAP) {
2405                 struct in6_addr addr;
2406
2407                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2408                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2409                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2410                         addrconf_add_linklocal(idev, &addr);
2411                 return;
2412         }
2413
2414         sit_add_v4_addrs(idev);
2415
2416         if (dev->flags&IFF_POINTOPOINT) {
2417                 addrconf_add_mroute(dev);
2418                 addrconf_add_lroute(dev);
2419         } else
2420                 sit_route_add(dev);
2421 }
2422 #endif
2423
2424 static inline int
2425 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2426 {
2427         struct in6_addr lladdr;
2428
2429         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2430                 addrconf_add_linklocal(idev, &lladdr);
2431                 return 0;
2432         }
2433         return -1;
2434 }
2435
2436 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2437 {
2438         struct net_device *link_dev;
2439         struct net *net = dev_net(idev->dev);
2440
2441         /* first try to inherit the link-local address from the link device */
2442         if (idev->dev->iflink &&
2443             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2444                 if (!ipv6_inherit_linklocal(idev, link_dev))
2445                         return;
2446         }
2447         /* then try to inherit it from any device */
2448         for_each_netdev(net, link_dev) {
2449                 if (!ipv6_inherit_linklocal(idev, link_dev))
2450                         return;
2451         }
2452         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2453 }
2454
2455 /*
2456  * Autoconfigure tunnel with a link-local address so routing protocols,
2457  * DHCPv6, MLD etc. can be run over the virtual link
2458  */
2459
2460 static void addrconf_ip6_tnl_config(struct net_device *dev)
2461 {
2462         struct inet6_dev *idev;
2463
2464         ASSERT_RTNL();
2465
2466         idev = addrconf_add_dev(dev);
2467         if (IS_ERR(idev)) {
2468                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2469                 return;
2470         }
2471         ip6_tnl_add_linklocal(idev);
2472 }
2473
2474 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2475                            void * data)
2476 {
2477         struct net_device *dev = (struct net_device *) data;
2478         struct inet6_dev *idev = __in6_dev_get(dev);
2479         int run_pending = 0;
2480         int err;
2481
2482         switch (event) {
2483         case NETDEV_REGISTER:
2484                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2485                         idev = ipv6_add_dev(dev);
2486                         if (!idev)
2487                                 return notifier_from_errno(-ENOMEM);
2488                 }
2489                 break;
2490
2491         case NETDEV_UP:
2492         case NETDEV_CHANGE:
2493                 if (dev->flags & IFF_SLAVE)
2494                         break;
2495
2496                 if (event == NETDEV_UP) {
2497                         if (!addrconf_qdisc_ok(dev)) {
2498                                 /* device is not ready yet. */
2499                                 printk(KERN_INFO
2500                                         "ADDRCONF(NETDEV_UP): %s: "
2501                                         "link is not ready\n",
2502                                         dev->name);
2503                                 break;
2504                         }
2505
2506                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2507                                 idev = ipv6_add_dev(dev);
2508
2509                         if (idev) {
2510                                 idev->if_flags |= IF_READY;
2511                                 run_pending = 1;
2512                         }
2513                 } else {
2514                         if (!addrconf_qdisc_ok(dev)) {
2515                                 /* device is still not ready. */
2516                                 break;
2517                         }
2518
2519                         if (idev) {
2520                                 if (idev->if_flags & IF_READY)
2521                                         /* device is already configured. */
2522                                         break;
2523                                 idev->if_flags |= IF_READY;
2524                         }
2525
2526                         printk(KERN_INFO
2527                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2528                                         "link becomes ready\n",
2529                                         dev->name);
2530
2531                         run_pending = 1;
2532                 }
2533
2534                 switch (dev->type) {
2535 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2536                 case ARPHRD_SIT:
2537                         addrconf_sit_config(dev);
2538                         break;
2539 #endif
2540                 case ARPHRD_TUNNEL6:
2541                         addrconf_ip6_tnl_config(dev);
2542                         break;
2543                 case ARPHRD_LOOPBACK:
2544                         init_loopback(dev);
2545                         break;
2546
2547                 default:
2548                         addrconf_dev_config(dev);
2549                         break;
2550                 }
2551
2552                 if (idev) {
2553                         if (run_pending)
2554                                 addrconf_dad_run(idev);
2555
2556                         /*
2557                          * If the MTU changed during the interface down,
2558                          * when the interface up, the changed MTU must be
2559                          * reflected in the idev as well as routers.
2560                          */
2561                         if (idev->cnf.mtu6 != dev->mtu &&
2562                             dev->mtu >= IPV6_MIN_MTU) {
2563                                 rt6_mtu_change(dev, dev->mtu);
2564                                 idev->cnf.mtu6 = dev->mtu;
2565                         }
2566                         idev->tstamp = jiffies;
2567                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2568
2569                         /*
2570                          * If the changed mtu during down is lower than
2571                          * IPV6_MIN_MTU stop IPv6 on this interface.
2572                          */
2573                         if (dev->mtu < IPV6_MIN_MTU)
2574                                 addrconf_ifdown(dev, 1);
2575                 }
2576                 break;
2577
2578         case NETDEV_CHANGEMTU:
2579                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2580                         rt6_mtu_change(dev, dev->mtu);
2581                         idev->cnf.mtu6 = dev->mtu;
2582                         break;
2583                 }
2584
2585                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2586                         idev = ipv6_add_dev(dev);
2587                         if (idev)
2588                                 break;
2589                 }
2590
2591                 /*
2592                  * MTU falled under IPV6_MIN_MTU.
2593                  * Stop IPv6 on this interface.
2594                  */
2595
2596         case NETDEV_DOWN:
2597         case NETDEV_UNREGISTER:
2598                 /*
2599                  *      Remove all addresses from this interface.
2600                  */
2601                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2602                 break;
2603
2604         case NETDEV_CHANGENAME:
2605                 if (idev) {
2606                         snmp6_unregister_dev(idev);
2607                         addrconf_sysctl_unregister(idev);
2608                         addrconf_sysctl_register(idev);
2609                         err = snmp6_register_dev(idev);
2610                         if (err)
2611                                 return notifier_from_errno(err);
2612                 }
2613                 break;
2614
2615         case NETDEV_PRE_TYPE_CHANGE:
2616         case NETDEV_POST_TYPE_CHANGE:
2617                 addrconf_type_change(dev, event);
2618                 break;
2619         }
2620
2621         return NOTIFY_OK;
2622 }
2623
2624 /*
2625  *      addrconf module should be notified of a device going up
2626  */
2627 static struct notifier_block ipv6_dev_notf = {
2628         .notifier_call = addrconf_notify,
2629 };
2630
2631 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2632 {
2633         struct inet6_dev *idev;
2634         ASSERT_RTNL();
2635
2636         idev = __in6_dev_get(dev);
2637
2638         if (event == NETDEV_POST_TYPE_CHANGE)
2639                 ipv6_mc_remap(idev);
2640         else if (event == NETDEV_PRE_TYPE_CHANGE)
2641                 ipv6_mc_unmap(idev);
2642 }
2643
2644 static int addrconf_ifdown(struct net_device *dev, int how)
2645 {
2646         struct net *net = dev_net(dev);
2647         struct inet6_dev *idev;
2648         struct inet6_ifaddr *ifa;
2649         int state, i;
2650
2651         ASSERT_RTNL();
2652
2653         rt6_ifdown(net, dev);
2654         neigh_ifdown(&nd_tbl, dev);
2655
2656         idev = __in6_dev_get(dev);
2657         if (idev == NULL)
2658                 return -ENODEV;
2659
2660         /*
2661          * Step 1: remove reference to ipv6 device from parent device.
2662          *         Do not dev_put!
2663          */
2664         if (how) {
2665                 idev->dead = 1;
2666
2667                 /* protected by rtnl_lock */
2668                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2669
2670                 /* Step 1.5: remove snmp6 entry */
2671                 snmp6_unregister_dev(idev);
2672
2673         }
2674
2675         /* Step 2: clear hash table */
2676         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2677                 struct hlist_head *h = &inet6_addr_lst[i];
2678                 struct hlist_node *n;
2679
2680                 spin_lock_bh(&addrconf_hash_lock);
2681         restart:
2682                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2683                         if (ifa->idev == idev) {
2684                                 hlist_del_init_rcu(&ifa->addr_lst);
2685                                 addrconf_del_timer(ifa);
2686                                 goto restart;
2687                         }
2688                 }
2689                 spin_unlock_bh(&addrconf_hash_lock);
2690         }
2691
2692         write_lock_bh(&idev->lock);
2693
2694         /* Step 2: clear flags for stateless addrconf */
2695         if (!how)
2696                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2697
2698 #ifdef CONFIG_IPV6_PRIVACY
2699         if (how && del_timer(&idev->regen_timer))
2700                 in6_dev_put(idev);
2701
2702         /* Step 3: clear tempaddr list */
2703         while (!list_empty(&idev->tempaddr_list)) {
2704                 ifa = list_first_entry(&idev->tempaddr_list,
2705                                        struct inet6_ifaddr, tmp_list);
2706                 list_del(&ifa->tmp_list);
2707                 write_unlock_bh(&idev->lock);
2708                 spin_lock_bh(&ifa->lock);
2709
2710                 if (ifa->ifpub) {
2711                         in6_ifa_put(ifa->ifpub);
2712                         ifa->ifpub = NULL;
2713                 }
2714                 spin_unlock_bh(&ifa->lock);
2715                 in6_ifa_put(ifa);
2716                 write_lock_bh(&idev->lock);
2717         }
2718 #endif
2719
2720         while (!list_empty(&idev->addr_list)) {
2721                 ifa = list_first_entry(&idev->addr_list,
2722                                        struct inet6_ifaddr, if_list);
2723                 addrconf_del_timer(ifa);
2724
2725                 list_del(&ifa->if_list);
2726
2727                 write_unlock_bh(&idev->lock);
2728
2729                 spin_lock_bh(&ifa->state_lock);
2730                 state = ifa->state;
2731                 ifa->state = INET6_IFADDR_STATE_DEAD;
2732                 spin_unlock_bh(&ifa->state_lock);
2733
2734                 if (state != INET6_IFADDR_STATE_DEAD) {
2735                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2736                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2737                 }
2738                 in6_ifa_put(ifa);
2739
2740                 write_lock_bh(&idev->lock);
2741         }
2742
2743         write_unlock_bh(&idev->lock);
2744
2745         /* Step 5: Discard multicast list */
2746         if (how)
2747                 ipv6_mc_destroy_dev(idev);
2748         else
2749                 ipv6_mc_down(idev);
2750
2751         idev->tstamp = jiffies;
2752
2753         /* Last: Shot the device (if unregistered) */
2754         if (how) {
2755                 addrconf_sysctl_unregister(idev);
2756                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2757                 neigh_ifdown(&nd_tbl, dev);
2758                 in6_dev_put(idev);
2759         }
2760         return 0;
2761 }
2762
2763 static void addrconf_rs_timer(unsigned long data)
2764 {
2765         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2766         struct inet6_dev *idev = ifp->idev;
2767
2768         read_lock(&idev->lock);
2769         if (idev->dead || !(idev->if_flags & IF_READY))
2770                 goto out;
2771
2772         if (idev->cnf.forwarding)
2773                 goto out;
2774
2775         /* Announcement received after solicitation was sent */
2776         if (idev->if_flags & IF_RA_RCVD)
2777                 goto out;
2778
2779         spin_lock(&ifp->lock);
2780         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2781                 /* The wait after the last probe can be shorter */
2782                 addrconf_mod_timer(ifp, AC_RS,
2783                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2784                                    idev->cnf.rtr_solicit_delay :
2785                                    idev->cnf.rtr_solicit_interval);
2786                 spin_unlock(&ifp->lock);
2787
2788                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2789         } else {
2790                 spin_unlock(&ifp->lock);
2791                 /*
2792                  * Note: we do not support deprecated "all on-link"
2793                  * assumption any longer.
2794                  */
2795                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2796                        idev->dev->name);
2797         }
2798
2799 out:
2800         read_unlock(&idev->lock);
2801         in6_ifa_put(ifp);
2802 }
2803
2804 /*
2805  *      Duplicate Address Detection
2806  */
2807 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2808 {
2809         unsigned long rand_num;
2810         struct inet6_dev *idev = ifp->idev;
2811
2812         if (ifp->flags & IFA_F_OPTIMISTIC)
2813                 rand_num = 0;
2814         else
2815                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2816
2817         ifp->probes = idev->cnf.dad_transmits;
2818         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2819 }
2820
2821 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2822 {
2823         struct inet6_dev *idev = ifp->idev;
2824         struct net_device *dev = idev->dev;
2825
2826         addrconf_join_solict(dev, &ifp->addr);
2827
2828         net_srandom(ifp->addr.s6_addr32[3]);
2829
2830         read_lock_bh(&idev->lock);
2831         spin_lock(&ifp->lock);
2832         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2833                 goto out;
2834
2835         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2836             idev->cnf.accept_dad < 1 ||
2837             !(ifp->flags&IFA_F_TENTATIVE) ||
2838             ifp->flags & IFA_F_NODAD) {
2839                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2840                 spin_unlock(&ifp->lock);
2841                 read_unlock_bh(&idev->lock);
2842
2843                 addrconf_dad_completed(ifp);
2844                 return;
2845         }
2846
2847         if (!(idev->if_flags & IF_READY)) {
2848                 spin_unlock(&ifp->lock);
2849                 read_unlock_bh(&idev->lock);
2850                 /*
2851                  * If the device is not ready:
2852                  * - keep it tentative if it is a permanent address.
2853                  * - otherwise, kill it.
2854                  */
2855                 in6_ifa_hold(ifp);
2856                 addrconf_dad_stop(ifp, 0);
2857                 return;
2858         }
2859
2860         /*
2861          * Optimistic nodes can start receiving
2862          * Frames right away
2863          */
2864         if (ifp->flags & IFA_F_OPTIMISTIC)
2865                 ip6_ins_rt(ifp->rt);
2866
2867         addrconf_dad_kick(ifp);
2868 out:
2869         spin_unlock(&ifp->lock);
2870         read_unlock_bh(&idev->lock);
2871 }
2872
2873 static void addrconf_dad_timer(unsigned long data)
2874 {
2875         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2876         struct inet6_dev *idev = ifp->idev;
2877         struct in6_addr mcaddr;
2878
2879         if (!ifp->probes && addrconf_dad_end(ifp))
2880                 goto out;
2881
2882         read_lock(&idev->lock);
2883         if (idev->dead || !(idev->if_flags & IF_READY)) {
2884                 read_unlock(&idev->lock);
2885                 goto out;
2886         }
2887
2888         spin_lock(&ifp->lock);
2889         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2890                 spin_unlock(&ifp->lock);
2891                 read_unlock(&idev->lock);
2892                 goto out;
2893         }
2894
2895         if (ifp->probes == 0) {
2896                 /*
2897                  * DAD was successful
2898                  */
2899
2900                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2901                 spin_unlock(&ifp->lock);
2902                 read_unlock(&idev->lock);
2903
2904                 addrconf_dad_completed(ifp);
2905
2906                 goto out;
2907         }
2908
2909         ifp->probes--;
2910         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2911         spin_unlock(&ifp->lock);
2912         read_unlock(&idev->lock);
2913
2914         /* send a neighbour solicitation for our addr */
2915         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2916         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2917 out:
2918         in6_ifa_put(ifp);
2919 }
2920
2921 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2922 {
2923         struct net_device *dev = ifp->idev->dev;
2924
2925         /*
2926          *      Configure the address for reception. Now it is valid.
2927          */
2928
2929         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2930
2931         /* If added prefix is link local and forwarding is off,
2932            start sending router solicitations.
2933          */
2934
2935         if ((ifp->idev->cnf.forwarding == 0 ||
2936              ifp->idev->cnf.forwarding == 2) &&
2937             ifp->idev->cnf.rtr_solicits > 0 &&
2938             (dev->flags&IFF_LOOPBACK) == 0 &&
2939             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2940                 /*
2941                  *      If a host as already performed a random delay
2942                  *      [...] as part of DAD [...] there is no need
2943                  *      to delay again before sending the first RS
2944                  */
2945                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2946
2947                 spin_lock_bh(&ifp->lock);
2948                 ifp->probes = 1;
2949                 ifp->idev->if_flags |= IF_RS_SENT;
2950                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2951                 spin_unlock_bh(&ifp->lock);
2952         }
2953 }
2954
2955 static void addrconf_dad_run(struct inet6_dev *idev)
2956 {
2957         struct inet6_ifaddr *ifp;
2958
2959         read_lock_bh(&idev->lock);
2960         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2961                 spin_lock(&ifp->lock);
2962                 if (ifp->flags & IFA_F_TENTATIVE &&
2963                     ifp->state == INET6_IFADDR_STATE_DAD)
2964                         addrconf_dad_kick(ifp);
2965                 spin_unlock(&ifp->lock);
2966         }
2967         read_unlock_bh(&idev->lock);
2968 }
2969
2970 #ifdef CONFIG_PROC_FS
2971 struct if6_iter_state {
2972         struct seq_net_private p;
2973         int bucket;
2974 };
2975
2976 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2977 {
2978         struct inet6_ifaddr *ifa = NULL;
2979         struct if6_iter_state *state = seq->private;
2980         struct net *net = seq_file_net(seq);
2981
2982         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2983                 struct hlist_node *n;
2984                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2985                                          addr_lst)
2986                         if (net_eq(dev_net(ifa->idev->dev), net))
2987                                 return ifa;
2988         }
2989         return NULL;
2990 }
2991
2992 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2993                                          struct inet6_ifaddr *ifa)
2994 {
2995         struct if6_iter_state *state = seq->private;
2996         struct net *net = seq_file_net(seq);
2997         struct hlist_node *n = &ifa->addr_lst;
2998
2999         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3000                 if (net_eq(dev_net(ifa->idev->dev), net))
3001                         return ifa;
3002
3003         while (++state->bucket < IN6_ADDR_HSIZE) {
3004                 hlist_for_each_entry_rcu_bh(ifa, n,
3005                                      &inet6_addr_lst[state->bucket], addr_lst) {
3006                         if (net_eq(dev_net(ifa->idev->dev), net))
3007                                 return ifa;
3008                 }
3009         }
3010
3011         return NULL;
3012 }
3013
3014 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3015 {
3016         struct inet6_ifaddr *ifa = if6_get_first(seq);
3017
3018         if (ifa)
3019                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3020                         --pos;
3021         return pos ? NULL : ifa;
3022 }
3023
3024 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3025         __acquires(rcu_bh)
3026 {
3027         rcu_read_lock_bh();
3028         return if6_get_idx(seq, *pos);
3029 }
3030
3031 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3032 {
3033         struct inet6_ifaddr *ifa;
3034
3035         ifa = if6_get_next(seq, v);
3036         ++*pos;
3037         return ifa;
3038 }
3039
3040 static void if6_seq_stop(struct seq_file *seq, void *v)
3041         __releases(rcu_bh)
3042 {
3043         rcu_read_unlock_bh();
3044 }
3045
3046 static int if6_seq_show(struct seq_file *seq, void *v)
3047 {
3048         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3049         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3050                    &ifp->addr,
3051                    ifp->idev->dev->ifindex,
3052                    ifp->prefix_len,
3053                    ifp->scope,
3054                    ifp->flags,
3055                    ifp->idev->dev->name);
3056         return 0;
3057 }
3058
3059 static const struct seq_operations if6_seq_ops = {
3060         .start  = if6_seq_start,
3061         .next   = if6_seq_next,
3062         .show   = if6_seq_show,
3063         .stop   = if6_seq_stop,
3064 };
3065
3066 static int if6_seq_open(struct inode *inode, struct file *file)
3067 {
3068         return seq_open_net(inode, file, &if6_seq_ops,
3069                             sizeof(struct if6_iter_state));
3070 }
3071
3072 static const struct file_operations if6_fops = {
3073         .owner          = THIS_MODULE,
3074         .open           = if6_seq_open,
3075         .read           = seq_read,
3076         .llseek         = seq_lseek,
3077         .release        = seq_release_net,
3078 };
3079
3080 static int __net_init if6_proc_net_init(struct net *net)
3081 {
3082         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3083                 return -ENOMEM;
3084         return 0;
3085 }
3086
3087 static void __net_exit if6_proc_net_exit(struct net *net)
3088 {
3089        proc_net_remove(net, "if_inet6");
3090 }
3091
3092 static struct pernet_operations if6_proc_net_ops = {
3093        .init = if6_proc_net_init,
3094        .exit = if6_proc_net_exit,
3095 };
3096
3097 int __init if6_proc_init(void)
3098 {
3099         return register_pernet_subsys(&if6_proc_net_ops);
3100 }
3101
3102 void if6_proc_exit(void)
3103 {
3104         unregister_pernet_subsys(&if6_proc_net_ops);
3105 }
3106 #endif  /* CONFIG_PROC_FS */
3107
3108 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3109 /* Check if address is a home address configured on any interface. */
3110 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3111 {
3112         int ret = 0;
3113         struct inet6_ifaddr *ifp = NULL;
3114         struct hlist_node *n;
3115         unsigned int hash = ipv6_addr_hash(addr);
3116
3117         rcu_read_lock_bh();
3118         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3119                 if (!net_eq(dev_net(ifp->idev->dev), net))
3120                         continue;
3121                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3122                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3123                         ret = 1;
3124                         break;
3125                 }
3126         }
3127         rcu_read_unlock_bh();
3128         return ret;
3129 }
3130 #endif
3131
3132 /*
3133  *      Periodic address status verification
3134  */
3135
3136 static void addrconf_verify(unsigned long foo)
3137 {
3138         unsigned long now, next, next_sec, next_sched;
3139         struct inet6_ifaddr *ifp;
3140         struct hlist_node *node;
3141         int i;
3142
3143         rcu_read_lock_bh();
3144         spin_lock(&addrconf_verify_lock);
3145         now = jiffies;
3146         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3147
3148         del_timer(&addr_chk_timer);
3149
3150         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3151 restart:
3152                 hlist_for_each_entry_rcu_bh(ifp, node,
3153                                          &inet6_addr_lst[i], addr_lst) {
3154                         unsigned long age;
3155
3156                         if (ifp->flags & IFA_F_PERMANENT)
3157                                 continue;
3158
3159                         spin_lock(&ifp->lock);
3160                         /* We try to batch several events at once. */
3161                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3162
3163                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3164                             age >= ifp->valid_lft) {
3165                                 spin_unlock(&ifp->lock);
3166                                 in6_ifa_hold(ifp);
3167                                 ipv6_del_addr(ifp);
3168                                 goto restart;
3169                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3170                                 spin_unlock(&ifp->lock);
3171                                 continue;
3172                         } else if (age >= ifp->prefered_lft) {
3173                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3174                                 int deprecate = 0;
3175
3176                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3177                                         deprecate = 1;
3178                                         ifp->flags |= IFA_F_DEPRECATED;
3179                                 }
3180
3181                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3182                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3183
3184                                 spin_unlock(&ifp->lock);
3185
3186                                 if (deprecate) {
3187                                         in6_ifa_hold(ifp);
3188
3189                                         ipv6_ifa_notify(0, ifp);
3190                                         in6_ifa_put(ifp);
3191                                         goto restart;
3192                                 }
3193 #ifdef CONFIG_IPV6_PRIVACY
3194                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3195                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3196                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3197                                         ifp->idev->cnf.dad_transmits *
3198                                         ifp->idev->nd_parms->retrans_time / HZ;
3199
3200                                 if (age >= ifp->prefered_lft - regen_advance) {
3201                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3202                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3203                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3204                                         if (!ifp->regen_count && ifpub) {
3205                                                 ifp->regen_count++;
3206                                                 in6_ifa_hold(ifp);
3207                                                 in6_ifa_hold(ifpub);
3208                                                 spin_unlock(&ifp->lock);
3209
3210                                                 spin_lock(&ifpub->lock);
3211                                                 ifpub->regen_count = 0;
3212                                                 spin_unlock(&ifpub->lock);
3213                                                 ipv6_create_tempaddr(ifpub, ifp);
3214                                                 in6_ifa_put(ifpub);
3215                                                 in6_ifa_put(ifp);
3216                                                 goto restart;
3217                                         }
3218                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3219                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3220                                 spin_unlock(&ifp->lock);
3221 #endif
3222                         } else {
3223                                 /* ifp->prefered_lft <= ifp->valid_lft */
3224                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3225                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3226                                 spin_unlock(&ifp->lock);
3227                         }
3228                 }
3229         }
3230
3231         next_sec = round_jiffies_up(next);
3232         next_sched = next;
3233
3234         /* If rounded timeout is accurate enough, accept it. */
3235         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3236                 next_sched = next_sec;
3237
3238         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3239         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3240                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3241
3242         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3243               now, next, next_sec, next_sched));
3244
3245         addr_chk_timer.expires = next_sched;
3246         add_timer(&addr_chk_timer);
3247         spin_unlock(&addrconf_verify_lock);
3248         rcu_read_unlock_bh();
3249 }
3250
3251 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3252 {
3253         struct in6_addr *pfx = NULL;
3254
3255         if (addr)
3256                 pfx = nla_data(addr);
3257
3258         if (local) {
3259                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3260                         pfx = NULL;
3261                 else
3262                         pfx = nla_data(local);
3263         }
3264
3265         return pfx;
3266 }
3267
3268 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3269         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3270         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3271         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3272 };
3273
3274 static int
3275 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3276 {
3277         struct net *net = sock_net(skb->sk);
3278         struct ifaddrmsg *ifm;
3279         struct nlattr *tb[IFA_MAX+1];
3280         struct in6_addr *pfx;
3281         int err;
3282
3283         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3284         if (err < 0)
3285                 return err;
3286
3287         ifm = nlmsg_data(nlh);
3288         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3289         if (pfx == NULL)
3290                 return -EINVAL;
3291
3292         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3293 }
3294
3295 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3296                              u32 prefered_lft, u32 valid_lft)
3297 {
3298         u32 flags;
3299         clock_t expires;
3300         unsigned long timeout;
3301
3302         if (!valid_lft || (prefered_lft > valid_lft))
3303                 return -EINVAL;
3304
3305         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3306         if (addrconf_finite_timeout(timeout)) {
3307                 expires = jiffies_to_clock_t(timeout * HZ);
3308                 valid_lft = timeout;
3309                 flags = RTF_EXPIRES;
3310         } else {
3311                 expires = 0;
3312                 flags = 0;
3313                 ifa_flags |= IFA_F_PERMANENT;
3314         }
3315
3316         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3317         if (addrconf_finite_timeout(timeout)) {
3318                 if (timeout == 0)
3319                         ifa_flags |= IFA_F_DEPRECATED;
3320                 prefered_lft = timeout;
3321         }
3322
3323         spin_lock_bh(&ifp->lock);
3324         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3325         ifp->tstamp = jiffies;
3326         ifp->valid_lft = valid_lft;
3327         ifp->prefered_lft = prefered_lft;
3328
3329         spin_unlock_bh(&ifp->lock);
3330         if (!(ifp->flags&IFA_F_TENTATIVE))
3331                 ipv6_ifa_notify(0, ifp);
3332
3333         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3334                               expires, flags);
3335         addrconf_verify(0);
3336
3337         return 0;
3338 }
3339
3340 static int
3341 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3342 {
3343         struct net *net = sock_net(skb->sk);
3344         struct ifaddrmsg *ifm;
3345         struct nlattr *tb[IFA_MAX+1];
3346         struct in6_addr *pfx;
3347         struct inet6_ifaddr *ifa;
3348         struct net_device *dev;
3349         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3350         u8 ifa_flags;
3351         int err;
3352
3353         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3354         if (err < 0)
3355                 return err;
3356
3357         ifm = nlmsg_data(nlh);
3358         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3359         if (pfx == NULL)
3360                 return -EINVAL;
3361
3362         if (tb[IFA_CACHEINFO]) {
3363                 struct ifa_cacheinfo *ci;
3364
3365                 ci = nla_data(tb[IFA_CACHEINFO]);
3366                 valid_lft = ci->ifa_valid;
3367                 preferred_lft = ci->ifa_prefered;
3368         } else {
3369                 preferred_lft = INFINITY_LIFE_TIME;
3370                 valid_lft = INFINITY_LIFE_TIME;
3371         }
3372
3373         dev =  __dev_get_by_index(net, ifm->ifa_index);
3374         if (dev == NULL)
3375                 return -ENODEV;
3376
3377         /* We ignore other flags so far. */
3378         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3379
3380         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3381         if (ifa == NULL) {
3382                 /*
3383                  * It would be best to check for !NLM_F_CREATE here but
3384                  * userspace alreay relies on not having to provide this.
3385                  */
3386                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3387                                       ifm->ifa_prefixlen, ifa_flags,
3388                                       preferred_lft, valid_lft);
3389         }
3390
3391         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3392             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3393                 err = -EEXIST;
3394         else
3395                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3396
3397         in6_ifa_put(ifa);
3398
3399         return err;
3400 }
3401
3402 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3403                           u8 scope, int ifindex)
3404 {
3405         struct ifaddrmsg *ifm;
3406
3407         ifm = nlmsg_data(nlh);
3408         ifm->ifa_family = AF_INET6;
3409         ifm->ifa_prefixlen = prefixlen;
3410         ifm->ifa_flags = flags;
3411         ifm->ifa_scope = scope;
3412         ifm->ifa_index = ifindex;
3413 }
3414
3415 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3416                          unsigned long tstamp, u32 preferred, u32 valid)
3417 {
3418         struct ifa_cacheinfo ci;
3419
3420         ci.cstamp = cstamp_delta(cstamp);
3421         ci.tstamp = cstamp_delta(tstamp);
3422         ci.ifa_prefered = preferred;
3423         ci.ifa_valid = valid;
3424
3425         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3426 }
3427
3428 static inline int rt_scope(int ifa_scope)
3429 {
3430         if (ifa_scope & IFA_HOST)
3431                 return RT_SCOPE_HOST;
3432         else if (ifa_scope & IFA_LINK)
3433                 return RT_SCOPE_LINK;
3434         else if (ifa_scope & IFA_SITE)
3435                 return RT_SCOPE_SITE;
3436         else
3437                 return RT_SCOPE_UNIVERSE;
3438 }
3439
3440 static inline int inet6_ifaddr_msgsize(void)
3441 {
3442         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3443                + nla_total_size(16) /* IFA_ADDRESS */
3444                + nla_total_size(sizeof(struct ifa_cacheinfo));
3445 }
3446
3447 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3448                              u32 pid, u32 seq, int event, unsigned int flags)
3449 {
3450         struct nlmsghdr  *nlh;
3451         u32 preferred, valid;
3452
3453         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3454         if (nlh == NULL)
3455                 return -EMSGSIZE;
3456
3457         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3458                       ifa->idev->dev->ifindex);
3459
3460         if (!(ifa->flags&IFA_F_PERMANENT)) {
3461                 preferred = ifa->prefered_lft;
3462                 valid = ifa->valid_lft;
3463                 if (preferred != INFINITY_LIFE_TIME) {
3464                         long tval = (jiffies - ifa->tstamp)/HZ;
3465                         if (preferred > tval)
3466                                 preferred -= tval;
3467                         else
3468                                 preferred = 0;
3469                         if (valid != INFINITY_LIFE_TIME) {
3470                                 if (valid > tval)
3471                                         valid -= tval;
3472                                 else
3473                                         valid = 0;
3474                         }
3475                 }
3476         } else {
3477                 preferred = INFINITY_LIFE_TIME;
3478                 valid = INFINITY_LIFE_TIME;
3479         }
3480
3481         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3482             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3483                 nlmsg_cancel(skb, nlh);
3484                 return -EMSGSIZE;
3485         }
3486
3487         return nlmsg_end(skb, nlh);
3488 }
3489
3490 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3491                                 u32 pid, u32 seq, int event, u16 flags)
3492 {
3493         struct nlmsghdr  *nlh;
3494         u8 scope = RT_SCOPE_UNIVERSE;
3495         int ifindex = ifmca->idev->dev->ifindex;
3496
3497         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3498                 scope = RT_SCOPE_SITE;
3499
3500         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3501         if (nlh == NULL)
3502                 return -EMSGSIZE;
3503
3504         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3505         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3506             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3507                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3508                 nlmsg_cancel(skb, nlh);
3509                 return -EMSGSIZE;
3510         }
3511
3512         return nlmsg_end(skb, nlh);
3513 }
3514
3515 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3516                                 u32 pid, u32 seq, int event, unsigned int flags)
3517 {
3518         struct nlmsghdr  *nlh;
3519         u8 scope = RT_SCOPE_UNIVERSE;
3520         int ifindex = ifaca->aca_idev->dev->ifindex;
3521
3522         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3523                 scope = RT_SCOPE_SITE;
3524
3525         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3526         if (nlh == NULL)
3527                 return -EMSGSIZE;
3528
3529         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3530         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3531             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3532                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3533                 nlmsg_cancel(skb, nlh);
3534                 return -EMSGSIZE;
3535         }
3536
3537         return nlmsg_end(skb, nlh);
3538 }
3539
3540 enum addr_type_t {
3541         UNICAST_ADDR,
3542         MULTICAST_ADDR,
3543         ANYCAST_ADDR,
3544 };
3545
3546 /* called with rcu_read_lock() */
3547 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3548                           struct netlink_callback *cb, enum addr_type_t type,
3549                           int s_ip_idx, int *p_ip_idx)
3550 {
3551         struct ifmcaddr6 *ifmca;
3552         struct ifacaddr6 *ifaca;
3553         int err = 1;
3554         int ip_idx = *p_ip_idx;
3555
3556         read_lock_bh(&idev->lock);
3557         switch (type) {
3558         case UNICAST_ADDR: {
3559                 struct inet6_ifaddr *ifa;
3560
3561                 /* unicast address incl. temp addr */
3562                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3563                         if (++ip_idx < s_ip_idx)
3564                                 continue;
3565                         err = inet6_fill_ifaddr(skb, ifa,
3566                                                 NETLINK_CB(cb->skb).pid,
3567                                                 cb->nlh->nlmsg_seq,
3568                                                 RTM_NEWADDR,
3569                                                 NLM_F_MULTI);
3570                         if (err <= 0)
3571                                 break;
3572                 }
3573                 break;
3574         }
3575         case MULTICAST_ADDR:
3576                 /* multicast address */
3577                 for (ifmca = idev->mc_list; ifmca;
3578                      ifmca = ifmca->next, ip_idx++) {
3579                         if (ip_idx < s_ip_idx)
3580                                 continue;
3581                         err = inet6_fill_ifmcaddr(skb, ifmca,
3582                                                   NETLINK_CB(cb->skb).pid,
3583                                                   cb->nlh->nlmsg_seq,
3584                                                   RTM_GETMULTICAST,
3585                                                   NLM_F_MULTI);
3586                         if (err <= 0)
3587                                 break;
3588                 }
3589                 break;
3590         case ANYCAST_ADDR:
3591                 /* anycast address */
3592                 for (ifaca = idev->ac_list; ifaca;
3593                      ifaca = ifaca->aca_next, ip_idx++) {
3594                         if (ip_idx < s_ip_idx)
3595                                 continue;
3596                         err = inet6_fill_ifacaddr(skb, ifaca,
3597                                                   NETLINK_CB(cb->skb).pid,
3598                                                   cb->nlh->nlmsg_seq,
3599                                                   RTM_GETANYCAST,
3600                                                   NLM_F_MULTI);
3601                         if (err <= 0)
3602                                 break;
3603                 }
3604                 break;
3605         default:
3606                 break;
3607         }
3608         read_unlock_bh(&idev->lock);
3609         *p_ip_idx = ip_idx;
3610         return err;
3611 }
3612
3613 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3614                            enum addr_type_t type)
3615 {
3616         struct net *net = sock_net(skb->sk);
3617         int h, s_h;
3618         int idx, ip_idx;
3619         int s_idx, s_ip_idx;
3620         struct net_device *dev;
3621         struct inet6_dev *idev;
3622         struct hlist_head *head;
3623         struct hlist_node *node;
3624
3625         s_h = cb->args[0];
3626         s_idx = idx = cb->args[1];
3627         s_ip_idx = ip_idx = cb->args[2];
3628
3629         rcu_read_lock();
3630         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3631                 idx = 0;
3632                 head = &net->dev_index_head[h];
3633                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3634                         if (idx < s_idx)
3635                                 goto cont;
3636                         if (h > s_h || idx > s_idx)
3637                                 s_ip_idx = 0;
3638                         ip_idx = 0;
3639                         idev = __in6_dev_get(dev);
3640                         if (!idev)
3641                                 goto cont;
3642
3643                         if (in6_dump_addrs(idev, skb, cb, type,
3644                                            s_ip_idx, &ip_idx) <= 0)
3645                                 goto done;
3646 cont:
3647                         idx++;
3648                 }
3649         }
3650 done:
3651         rcu_read_unlock();
3652         cb->args[0] = h;
3653         cb->args[1] = idx;
3654         cb->args[2] = ip_idx;
3655
3656         return skb->len;
3657 }
3658
3659 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3660 {
3661         enum addr_type_t type = UNICAST_ADDR;
3662
3663         return inet6_dump_addr(skb, cb, type);
3664 }
3665
3666 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3667 {
3668         enum addr_type_t type = MULTICAST_ADDR;
3669
3670         return inet6_dump_addr(skb, cb, type);
3671 }
3672
3673
3674 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3675 {
3676         enum addr_type_t type = ANYCAST_ADDR;
3677
3678         return inet6_dump_addr(skb, cb, type);
3679 }
3680
3681 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3682                              void *arg)
3683 {
3684         struct net *net = sock_net(in_skb->sk);
3685         struct ifaddrmsg *ifm;
3686         struct nlattr *tb[IFA_MAX+1];
3687         struct in6_addr *addr = NULL;
3688         struct net_device *dev = NULL;
3689         struct inet6_ifaddr *ifa;
3690         struct sk_buff *skb;
3691         int err;
3692
3693         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3694         if (err < 0)
3695                 goto errout;
3696
3697         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3698         if (addr == NULL) {
3699                 err = -EINVAL;
3700                 goto errout;
3701         }
3702
3703         ifm = nlmsg_data(nlh);
3704         if (ifm->ifa_index)
3705                 dev = __dev_get_by_index(net, ifm->ifa_index);
3706
3707         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3708         if (!ifa) {
3709                 err = -EADDRNOTAVAIL;
3710                 goto errout;
3711         }
3712
3713         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3714         if (!skb) {
3715                 err = -ENOBUFS;
3716                 goto errout_ifa;
3717         }
3718
3719         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3720                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3721         if (err < 0) {
3722                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3723                 WARN_ON(err == -EMSGSIZE);
3724                 kfree_skb(skb);
3725                 goto errout_ifa;
3726         }
3727         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3728 errout_ifa:
3729         in6_ifa_put(ifa);
3730 errout:
3731         return err;
3732 }
3733
3734 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3735 {
3736         struct sk_buff *skb;
3737         struct net *net = dev_net(ifa->idev->dev);
3738         int err = -ENOBUFS;
3739
3740         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3741         if (skb == NULL)
3742                 goto errout;
3743
3744         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3745         if (err < 0) {
3746                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3747                 WARN_ON(err == -EMSGSIZE);
3748                 kfree_skb(skb);
3749                 goto errout;
3750         }
3751         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3752         return;
3753 errout:
3754         if (err < 0)
3755                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3756 }
3757
3758 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3759                                 __s32 *array, int bytes)
3760 {
3761         BUG_ON(bytes < (DEVCONF_MAX * 4));
3762
3763         memset(array, 0, bytes);
3764         array[DEVCONF_FORWARDING] = cnf->forwarding;
3765         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3766         array[DEVCONF_MTU6] = cnf->mtu6;
3767         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3768         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3769         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3770         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3771         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3772         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3773                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3774         array[DEVCONF_RTR_SOLICIT_DELAY] =
3775                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3776         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3777 #ifdef CONFIG_IPV6_PRIVACY
3778         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3779         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3780         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3781         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3782         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3783 #endif
3784         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3785         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3786         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3787 #ifdef CONFIG_IPV6_ROUTER_PREF
3788         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3789         array[DEVCONF_RTR_PROBE_INTERVAL] =
3790                 jiffies_to_msecs(cnf->rtr_probe_interval);
3791 #ifdef CONFIG_IPV6_ROUTE_INFO
3792         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3793 #endif
3794 #endif
3795         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3796         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3797 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3798         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3799 #endif
3800 #ifdef CONFIG_IPV6_MROUTE
3801         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3802 #endif
3803         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3804         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3805         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3806 }
3807
3808 static inline size_t inet6_ifla6_size(void)
3809 {
3810         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3811              + nla_total_size(sizeof(struct ifla_cacheinfo))
3812              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3813              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3814              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3815 }
3816
3817 static inline size_t inet6_if_nlmsg_size(void)
3818 {
3819         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3820                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3821                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3822                + nla_total_size(4) /* IFLA_MTU */
3823                + nla_total_size(4) /* IFLA_LINK */
3824                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3825 }
3826
3827 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3828                                       int items, int bytes)
3829 {
3830         int i;
3831         int pad = bytes - sizeof(u64) * items;
3832         BUG_ON(pad < 0);
3833
3834         /* Use put_unaligned() because stats may not be aligned for u64. */
3835         put_unaligned(items, &stats[0]);
3836         for (i = 1; i < items; i++)
3837                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3838
3839         memset(&stats[items], 0, pad);
3840 }
3841
3842 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3843                                       int items, int bytes, size_t syncpoff)
3844 {
3845         int i;
3846         int pad = bytes - sizeof(u64) * items;
3847         BUG_ON(pad < 0);
3848
3849         /* Use put_unaligned() because stats may not be aligned for u64. */
3850         put_unaligned(items, &stats[0]);
3851         for (i = 1; i < items; i++)
3852                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3853
3854         memset(&stats[items], 0, pad);
3855 }
3856
3857 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3858                              int bytes)
3859 {
3860         switch (attrtype) {
3861         case IFLA_INET6_STATS:
3862                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3863                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3864                 break;
3865         case IFLA_INET6_ICMP6STATS:
3866                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3867                 break;
3868         }
3869 }
3870
3871 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3872 {
3873         struct nlattr *nla;
3874         struct ifla_cacheinfo ci;
3875
3876         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3877
3878         ci.max_reasm_len = IPV6_MAXPLEN;
3879         ci.tstamp = cstamp_delta(idev->tstamp);
3880         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3881         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3882         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3883
3884         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3885         if (nla == NULL)
3886                 goto nla_put_failure;
3887         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3888
3889         /* XXX - MC not implemented */
3890
3891         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3892         if (nla == NULL)
3893                 goto nla_put_failure;
3894         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3895
3896         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3897         if (nla == NULL)
3898                 goto nla_put_failure;
3899         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3900
3901         return 0;
3902
3903 nla_put_failure:
3904         return -EMSGSIZE;
3905 }
3906
3907 static size_t inet6_get_link_af_size(const struct net_device *dev)
3908 {
3909         if (!__in6_dev_get(dev))
3910                 return 0;
3911
3912         return inet6_ifla6_size();
3913 }
3914
3915 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3916 {
3917         struct inet6_dev *idev = __in6_dev_get(dev);
3918
3919         if (!idev)
3920                 return -ENODATA;
3921
3922         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3923                 return -EMSGSIZE;
3924
3925         return 0;
3926 }
3927
3928 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3929                              u32 pid, u32 seq, int event, unsigned int flags)
3930 {
3931         struct net_device *dev = idev->dev;
3932         struct ifinfomsg *hdr;
3933         struct nlmsghdr *nlh;
3934         void *protoinfo;
3935
3936         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3937         if (nlh == NULL)
3938                 return -EMSGSIZE;
3939
3940         hdr = nlmsg_data(nlh);
3941         hdr->ifi_family = AF_INET6;
3942         hdr->__ifi_pad = 0;
3943         hdr->ifi_type = dev->type;
3944         hdr->ifi_index = dev->ifindex;
3945         hdr->ifi_flags = dev_get_flags(dev);
3946         hdr->ifi_change = 0;
3947
3948         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3949
3950         if (dev->addr_len)
3951                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3952
3953         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3954         if (dev->ifindex != dev->iflink)
3955                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3956
3957         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3958         if (protoinfo == NULL)
3959                 goto nla_put_failure;
3960
3961         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3962                 goto nla_put_failure;
3963
3964         nla_nest_end(skb, protoinfo);
3965         return nlmsg_end(skb, nlh);
3966
3967 nla_put_failure:
3968         nlmsg_cancel(skb, nlh);
3969         return -EMSGSIZE;
3970 }
3971
3972 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3973 {
3974         struct net *net = sock_net(skb->sk);
3975         int h, s_h;
3976         int idx = 0, s_idx;
3977         struct net_device *dev;
3978         struct inet6_dev *idev;
3979         struct hlist_head *head;
3980         struct hlist_node *node;
3981
3982         s_h = cb->args[0];
3983         s_idx = cb->args[1];
3984
3985         rcu_read_lock();
3986         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3987                 idx = 0;
3988                 head = &net->dev_index_head[h];
3989                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3990                         if (idx < s_idx)
3991                                 goto cont;
3992                         idev = __in6_dev_get(dev);
3993                         if (!idev)
3994                                 goto cont;
3995                         if (inet6_fill_ifinfo(skb, idev,
3996                                               NETLINK_CB(cb->skb).pid,
3997                                               cb->nlh->nlmsg_seq,
3998                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3999                                 goto out;
4000 cont:
4001                         idx++;
4002                 }
4003         }
4004 out:
4005         rcu_read_unlock();
4006         cb->args[1] = idx;
4007         cb->args[0] = h;
4008
4009         return skb->len;
4010 }
4011
4012 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4013 {
4014         struct sk_buff *skb;
4015         struct net *net = dev_net(idev->dev);
4016         int err = -ENOBUFS;
4017
4018         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4019         if (skb == NULL)
4020                 goto errout;
4021
4022         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4023         if (err < 0) {
4024                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4025                 WARN_ON(err == -EMSGSIZE);
4026                 kfree_skb(skb);
4027                 goto errout;
4028         }
4029         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4030         return;
4031 errout:
4032         if (err < 0)
4033                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4034 }
4035
4036 static inline size_t inet6_prefix_nlmsg_size(void)
4037 {
4038         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4039                + nla_total_size(sizeof(struct in6_addr))
4040                + nla_total_size(sizeof(struct prefix_cacheinfo));
4041 }
4042
4043 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4044                              struct prefix_info *pinfo, u32 pid, u32 seq,
4045                              int event, unsigned int flags)
4046 {
4047         struct prefixmsg *pmsg;
4048         struct nlmsghdr *nlh;
4049         struct prefix_cacheinfo ci;
4050
4051         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4052         if (nlh == NULL)
4053                 return -EMSGSIZE;
4054
4055         pmsg = nlmsg_data(nlh);
4056         pmsg->prefix_family = AF_INET6;
4057         pmsg->prefix_pad1 = 0;
4058         pmsg->prefix_pad2 = 0;
4059         pmsg->prefix_ifindex = idev->dev->ifindex;
4060         pmsg->prefix_len = pinfo->prefix_len;
4061         pmsg->prefix_type = pinfo->type;
4062         pmsg->prefix_pad3 = 0;
4063         pmsg->prefix_flags = 0;
4064         if (pinfo->onlink)
4065                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4066         if (pinfo->autoconf)
4067                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4068
4069         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4070
4071         ci.preferred_time = ntohl(pinfo->prefered);
4072         ci.valid_time = ntohl(pinfo->valid);
4073         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4074
4075         return nlmsg_end(skb, nlh);
4076
4077 nla_put_failure:
4078         nlmsg_cancel(skb, nlh);
4079         return -EMSGSIZE;
4080 }
4081
4082 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4083                          struct prefix_info *pinfo)
4084 {
4085         struct sk_buff *skb;
4086         struct net *net = dev_net(idev->dev);
4087         int err = -ENOBUFS;
4088
4089         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4090         if (skb == NULL)
4091                 goto errout;
4092
4093         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4094         if (err < 0) {
4095                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4096                 WARN_ON(err == -EMSGSIZE);
4097                 kfree_skb(skb);
4098                 goto errout;
4099         }
4100         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4101         return;
4102 errout:
4103         if (err < 0)
4104                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4105 }
4106
4107 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4108 {
4109         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4110
4111         switch (event) {
4112         case RTM_NEWADDR:
4113                 /*
4114                  * If the address was optimistic
4115                  * we inserted the route at the start of
4116                  * our DAD process, so we don't need
4117                  * to do it again
4118                  */
4119                 if (!(ifp->rt->rt6i_node))
4120                         ip6_ins_rt(ifp->rt);
4121                 if (ifp->idev->cnf.forwarding)
4122                         addrconf_join_anycast(ifp);
4123                 break;
4124         case RTM_DELADDR:
4125                 if (ifp->idev->cnf.forwarding)
4126                         addrconf_leave_anycast(ifp);
4127                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4128                 dst_hold(&ifp->rt->dst);
4129
4130                 if (ip6_del_rt(ifp->rt))
4131                         dst_free(&ifp->rt->dst);
4132                 break;
4133         }
4134 }
4135
4136 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4137 {
4138         rcu_read_lock_bh();
4139         if (likely(ifp->idev->dead == 0))
4140                 __ipv6_ifa_notify(event, ifp);
4141         rcu_read_unlock_bh();
4142 }
4143
4144 #ifdef CONFIG_SYSCTL
4145
4146 static
4147 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4148                            void __user *buffer, size_t *lenp, loff_t *ppos)
4149 {
4150         int *valp = ctl->data;
4151         int val = *valp;
4152         loff_t pos = *ppos;
4153         int ret;
4154
4155         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4156
4157         if (write)
4158                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4159         if (ret)
4160                 *ppos = pos;
4161         return ret;
4162 }
4163
4164 static void dev_disable_change(struct inet6_dev *idev)
4165 {
4166         if (!idev || !idev->dev)
4167                 return;
4168
4169         if (idev->cnf.disable_ipv6)
4170                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4171         else
4172                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4173 }
4174
4175 static void addrconf_disable_change(struct net *net, __s32 newf)
4176 {
4177         struct net_device *dev;
4178         struct inet6_dev *idev;
4179
4180         rcu_read_lock();
4181         for_each_netdev_rcu(net, dev) {
4182                 idev = __in6_dev_get(dev);
4183                 if (idev) {
4184                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4185                         idev->cnf.disable_ipv6 = newf;
4186                         if (changed)
4187                                 dev_disable_change(idev);
4188                 }
4189         }
4190         rcu_read_unlock();
4191 }
4192
4193 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4194 {
4195         struct net *net;
4196
4197         net = (struct net *)table->extra2;
4198
4199         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4200                 return 0;
4201
4202         if (!rtnl_trylock()) {
4203                 /* Restore the original values before restarting */
4204                 *p = old;
4205                 return restart_syscall();
4206         }
4207
4208         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4209                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4210                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4211                 addrconf_disable_change(net, newf);
4212         } else if ((!*p) ^ (!old))
4213                 dev_disable_change((struct inet6_dev *)table->extra1);
4214
4215         rtnl_unlock();
4216         return 0;
4217 }
4218
4219 static
4220 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4221                             void __user *buffer, size_t *lenp, loff_t *ppos)
4222 {
4223         int *valp = ctl->data;
4224         int val = *valp;
4225         loff_t pos = *ppos;
4226         int ret;
4227
4228         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4229
4230         if (write)
4231                 ret = addrconf_disable_ipv6(ctl, valp, val);
4232         if (ret)
4233                 *ppos = pos;
4234         return ret;
4235 }
4236
4237 static struct addrconf_sysctl_table
4238 {
4239         struct ctl_table_header *sysctl_header;
4240         ctl_table addrconf_vars[DEVCONF_MAX+1];
4241         char *dev_name;
4242 } addrconf_sysctl __read_mostly = {
4243         .sysctl_header = NULL,
4244         .addrconf_vars = {
4245                 {
4246                         .procname       = "forwarding",
4247                         .data           = &ipv6_devconf.forwarding,
4248                         .maxlen         = sizeof(int),
4249                         .mode           = 0644,
4250                         .proc_handler   = addrconf_sysctl_forward,
4251                 },
4252                 {
4253                         .procname       = "hop_limit",
4254                         .data           = &ipv6_devconf.hop_limit,
4255                         .maxlen         = sizeof(int),
4256                         .mode           = 0644,
4257                         .proc_handler   = proc_dointvec,
4258                 },
4259                 {
4260                         .procname       = "mtu",
4261                         .data           = &ipv6_devconf.mtu6,
4262                         .maxlen         = sizeof(int),
4263                         .mode           = 0644,
4264                         .proc_handler   = proc_dointvec,
4265                 },
4266                 {
4267                         .procname       = "accept_ra",
4268                         .data           = &ipv6_devconf.accept_ra,
4269                         .maxlen         = sizeof(int),
4270                         .mode           = 0644,
4271                         .proc_handler   = proc_dointvec,
4272                 },
4273                 {
4274                         .procname       = "accept_redirects",
4275                         .data           = &ipv6_devconf.accept_redirects,
4276                         .maxlen         = sizeof(int),
4277                         .mode           = 0644,
4278                         .proc_handler   = proc_dointvec,
4279                 },
4280                 {
4281                         .procname       = "autoconf",
4282                         .data           = &ipv6_devconf.autoconf,
4283                         .maxlen         = sizeof(int),
4284                         .mode           = 0644,
4285                         .proc_handler   = proc_dointvec,
4286                 },
4287                 {
4288                         .procname       = "dad_transmits",
4289                         .data           = &ipv6_devconf.dad_transmits,
4290                         .maxlen         = sizeof(int),
4291                         .mode           = 0644,
4292                         .proc_handler   = proc_dointvec,
4293                 },
4294                 {
4295                         .procname       = "router_solicitations",
4296                         .data           = &ipv6_devconf.rtr_solicits,
4297                         .maxlen         = sizeof(int),
4298                         .mode           = 0644,
4299                         .proc_handler   = proc_dointvec,
4300                 },
4301                 {
4302                         .procname       = "router_solicitation_interval",
4303                         .data           = &ipv6_devconf.rtr_solicit_interval,
4304                         .maxlen         = sizeof(int),
4305                         .mode           = 0644,
4306                         .proc_handler   = proc_dointvec_jiffies,
4307                 },
4308                 {
4309                         .procname       = "router_solicitation_delay",
4310                         .data           = &ipv6_devconf.rtr_solicit_delay,
4311                         .maxlen         = sizeof(int),
4312                         .mode           = 0644,
4313                         .proc_handler   = proc_dointvec_jiffies,
4314                 },
4315                 {
4316                         .procname       = "force_mld_version",
4317                         .data           = &ipv6_devconf.force_mld_version,
4318                         .maxlen         = sizeof(int),
4319                         .mode           = 0644,
4320                         .proc_handler   = proc_dointvec,
4321                 },
4322 #ifdef CONFIG_IPV6_PRIVACY
4323                 {
4324                         .procname       = "use_tempaddr",
4325                         .data           = &ipv6_devconf.use_tempaddr,
4326                         .maxlen         = sizeof(int),
4327                         .mode           = 0644,
4328                         .proc_handler   = proc_dointvec,
4329                 },
4330                 {
4331                         .procname       = "temp_valid_lft",
4332                         .data           = &ipv6_devconf.temp_valid_lft,
4333                         .maxlen         = sizeof(int),
4334                         .mode           = 0644,
4335                         .proc_handler   = proc_dointvec,
4336                 },
4337                 {
4338                         .procname       = "temp_prefered_lft",
4339                         .data           = &ipv6_devconf.temp_prefered_lft,
4340                         .maxlen         = sizeof(int),
4341                         .mode           = 0644,
4342                         .proc_handler   = proc_dointvec,
4343                 },
4344                 {
4345                         .procname       = "regen_max_retry",
4346                         .data           = &ipv6_devconf.regen_max_retry,
4347                         .maxlen         = sizeof(int),
4348                         .mode           = 0644,
4349                         .proc_handler   = proc_dointvec,
4350                 },
4351                 {
4352                         .procname       = "max_desync_factor",
4353                         .data           = &ipv6_devconf.max_desync_factor,
4354                         .maxlen         = sizeof(int),
4355                         .mode           = 0644,
4356                         .proc_handler   = proc_dointvec,
4357                 },
4358 #endif
4359                 {
4360                         .procname       = "max_addresses",
4361                         .data           = &ipv6_devconf.max_addresses,
4362                         .maxlen         = sizeof(int),
4363                         .mode           = 0644,
4364                         .proc_handler   = proc_dointvec,
4365                 },
4366                 {
4367                         .procname       = "accept_ra_defrtr",
4368                         .data           = &ipv6_devconf.accept_ra_defrtr,
4369                         .maxlen         = sizeof(int),
4370                         .mode           = 0644,
4371                         .proc_handler   = proc_dointvec,
4372                 },
4373                 {
4374                         .procname       = "accept_ra_pinfo",
4375                         .data           = &ipv6_devconf.accept_ra_pinfo,
4376                         .maxlen         = sizeof(int),
4377                         .mode           = 0644,
4378                         .proc_handler   = proc_dointvec,
4379                 },
4380 #ifdef CONFIG_IPV6_ROUTER_PREF
4381                 {
4382                         .procname       = "accept_ra_rtr_pref",
4383                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4384                         .maxlen         = sizeof(int),
4385                         .mode           = 0644,
4386                         .proc_handler   = proc_dointvec,
4387                 },
4388                 {
4389                         .procname       = "router_probe_interval",
4390                         .data           = &ipv6_devconf.rtr_probe_interval,
4391                         .maxlen         = sizeof(int),
4392                         .mode           = 0644,
4393                         .proc_handler   = proc_dointvec_jiffies,
4394                 },
4395 #ifdef CONFIG_IPV6_ROUTE_INFO
4396                 {
4397                         .procname       = "accept_ra_rt_info_max_plen",
4398                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec,
4402                 },
4403 #endif
4404 #endif
4405                 {
4406                         .procname       = "proxy_ndp",
4407                         .data           = &ipv6_devconf.proxy_ndp,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "accept_source_route",
4414                         .data           = &ipv6_devconf.accept_source_route,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4420                 {
4421                         .procname       = "optimistic_dad",
4422                         .data           = &ipv6_devconf.optimistic_dad,
4423                         .maxlen         = sizeof(int),
4424                         .mode           = 0644,
4425                         .proc_handler   = proc_dointvec,
4426
4427                 },
4428 #endif
4429 #ifdef CONFIG_IPV6_MROUTE
4430                 {
4431                         .procname       = "mc_forwarding",
4432                         .data           = &ipv6_devconf.mc_forwarding,
4433                         .maxlen         = sizeof(int),
4434                         .mode           = 0444,
4435                         .proc_handler   = proc_dointvec,
4436                 },
4437 #endif
4438                 {
4439                         .procname       = "disable_ipv6",
4440                         .data           = &ipv6_devconf.disable_ipv6,
4441                         .maxlen         = sizeof(int),
4442                         .mode           = 0644,
4443                         .proc_handler   = addrconf_sysctl_disable,
4444                 },
4445                 {
4446                         .procname       = "accept_dad",
4447                         .data           = &ipv6_devconf.accept_dad,
4448                         .maxlen         = sizeof(int),
4449                         .mode           = 0644,
4450                         .proc_handler   = proc_dointvec,
4451                 },
4452                 {
4453                         .procname       = "force_tllao",
4454                         .data           = &ipv6_devconf.force_tllao,
4455                         .maxlen         = sizeof(int),
4456                         .mode           = 0644,
4457                         .proc_handler   = proc_dointvec
4458                 },
4459                 {
4460                         /* sentinel */
4461                 }
4462         },
4463 };
4464
4465 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4466                 struct inet6_dev *idev, struct ipv6_devconf *p)
4467 {
4468         int i;
4469         struct addrconf_sysctl_table *t;
4470
4471 #define ADDRCONF_CTL_PATH_DEV   3
4472
4473         struct ctl_path addrconf_ctl_path[] = {
4474                 { .procname = "net", },
4475                 { .procname = "ipv6", },
4476                 { .procname = "conf", },
4477                 { /* to be set */ },
4478                 { },
4479         };
4480
4481
4482         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4483         if (t == NULL)
4484                 goto out;
4485
4486         for (i = 0; t->addrconf_vars[i].data; i++) {
4487                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4488                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4489                 t->addrconf_vars[i].extra2 = net;
4490         }
4491
4492         /*
4493          * Make a copy of dev_name, because '.procname' is regarded as const
4494          * by sysctl and we wouldn't want anyone to change it under our feet
4495          * (see SIOCSIFNAME).
4496          */
4497         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4498         if (!t->dev_name)
4499                 goto free;
4500
4501         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4502
4503         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4504                         t->addrconf_vars);
4505         if (t->sysctl_header == NULL)
4506                 goto free_procname;
4507
4508         p->sysctl = t;
4509         return 0;
4510
4511 free_procname:
4512         kfree(t->dev_name);
4513 free:
4514         kfree(t);
4515 out:
4516         return -ENOBUFS;
4517 }
4518
4519 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4520 {
4521         struct addrconf_sysctl_table *t;
4522
4523         if (p->sysctl == NULL)
4524                 return;
4525
4526         t = p->sysctl;
4527         p->sysctl = NULL;
4528         unregister_net_sysctl_table(t->sysctl_header);
4529         kfree(t->dev_name);
4530         kfree(t);
4531 }
4532
4533 static void addrconf_sysctl_register(struct inet6_dev *idev)
4534 {
4535         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4536                               &ndisc_ifinfo_sysctl_change);
4537         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4538                                         idev, &idev->cnf);
4539 }
4540
4541 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4542 {
4543         __addrconf_sysctl_unregister(&idev->cnf);
4544         neigh_sysctl_unregister(idev->nd_parms);
4545 }
4546
4547
4548 #endif
4549
4550 static int __net_init addrconf_init_net(struct net *net)
4551 {
4552         int err;
4553         struct ipv6_devconf *all, *dflt;
4554
4555         err = -ENOMEM;
4556         all = &ipv6_devconf;
4557         dflt = &ipv6_devconf_dflt;
4558
4559         if (!net_eq(net, &init_net)) {
4560                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4561                 if (all == NULL)
4562                         goto err_alloc_all;
4563
4564                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4565                 if (dflt == NULL)
4566                         goto err_alloc_dflt;
4567         } else {
4568                 /* these will be inherited by all namespaces */
4569                 dflt->autoconf = ipv6_defaults.autoconf;
4570                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4571         }
4572
4573         net->ipv6.devconf_all = all;
4574         net->ipv6.devconf_dflt = dflt;
4575
4576 #ifdef CONFIG_SYSCTL
4577         err = __addrconf_sysctl_register(net, "all", NULL, all);
4578         if (err < 0)
4579                 goto err_reg_all;
4580
4581         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4582         if (err < 0)
4583                 goto err_reg_dflt;
4584 #endif
4585         return 0;
4586
4587 #ifdef CONFIG_SYSCTL
4588 err_reg_dflt:
4589         __addrconf_sysctl_unregister(all);
4590 err_reg_all:
4591         kfree(dflt);
4592 #endif
4593 err_alloc_dflt:
4594         kfree(all);
4595 err_alloc_all:
4596         return err;
4597 }
4598
4599 static void __net_exit addrconf_exit_net(struct net *net)
4600 {
4601 #ifdef CONFIG_SYSCTL
4602         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4603         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4604 #endif
4605         if (!net_eq(net, &init_net)) {
4606                 kfree(net->ipv6.devconf_dflt);
4607                 kfree(net->ipv6.devconf_all);
4608         }
4609 }
4610
4611 static struct pernet_operations addrconf_ops = {
4612         .init = addrconf_init_net,
4613         .exit = addrconf_exit_net,
4614 };
4615
4616 /*
4617  *      Device notifier
4618  */
4619
4620 int register_inet6addr_notifier(struct notifier_block *nb)
4621 {
4622         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4623 }
4624 EXPORT_SYMBOL(register_inet6addr_notifier);
4625
4626 int unregister_inet6addr_notifier(struct notifier_block *nb)
4627 {
4628         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4629 }
4630 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4631
4632 static struct rtnl_af_ops inet6_ops = {
4633         .family           = AF_INET6,
4634         .fill_link_af     = inet6_fill_link_af,
4635         .get_link_af_size = inet6_get_link_af_size,
4636 };
4637
4638 /*
4639  *      Init / cleanup code
4640  */
4641
4642 int __init addrconf_init(void)
4643 {
4644         int i, err;
4645
4646         err = ipv6_addr_label_init();
4647         if (err < 0) {
4648                 printk(KERN_CRIT "IPv6 Addrconf:"
4649                        " cannot initialize default policy table: %d.\n", err);
4650                 goto out;
4651         }
4652
4653         err = register_pernet_subsys(&addrconf_ops);
4654         if (err < 0)
4655                 goto out_addrlabel;
4656
4657         /* The addrconf netdev notifier requires that loopback_dev
4658          * has it's ipv6 private information allocated and setup
4659          * before it can bring up and give link-local addresses
4660          * to other devices which are up.
4661          *
4662          * Unfortunately, loopback_dev is not necessarily the first
4663          * entry in the global dev_base list of net devices.  In fact,
4664          * it is likely to be the very last entry on that list.
4665          * So this causes the notifier registry below to try and
4666          * give link-local addresses to all devices besides loopback_dev
4667          * first, then loopback_dev, which cases all the non-loopback_dev
4668          * devices to fail to get a link-local address.
4669          *
4670          * So, as a temporary fix, allocate the ipv6 structure for
4671          * loopback_dev first by hand.
4672          * Longer term, all of the dependencies ipv6 has upon the loopback
4673          * device and it being up should be removed.
4674          */
4675         rtnl_lock();
4676         if (!ipv6_add_dev(init_net.loopback_dev))
4677                 err = -ENOMEM;
4678         rtnl_unlock();
4679         if (err)
4680                 goto errlo;
4681
4682         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4683                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4684
4685         register_netdevice_notifier(&ipv6_dev_notf);
4686
4687         addrconf_verify(0);
4688
4689         err = rtnl_af_register(&inet6_ops);
4690         if (err < 0)
4691                 goto errout_af;
4692
4693         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4694         if (err < 0)
4695                 goto errout;
4696
4697         /* Only the first call to __rtnl_register can fail */
4698         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4699         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4700         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4701         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4702         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4703
4704         ipv6_addr_label_rtnl_register();
4705
4706         return 0;
4707 errout:
4708         rtnl_af_unregister(&inet6_ops);
4709 errout_af:
4710         unregister_netdevice_notifier(&ipv6_dev_notf);
4711 errlo:
4712         unregister_pernet_subsys(&addrconf_ops);
4713 out_addrlabel:
4714         ipv6_addr_label_cleanup();
4715 out:
4716         return err;
4717 }
4718
4719 void addrconf_cleanup(void)
4720 {
4721         struct net_device *dev;
4722         int i;
4723
4724         unregister_netdevice_notifier(&ipv6_dev_notf);
4725         unregister_pernet_subsys(&addrconf_ops);
4726         ipv6_addr_label_cleanup();
4727
4728         rtnl_lock();
4729
4730         __rtnl_af_unregister(&inet6_ops);
4731
4732         /* clean dev list */
4733         for_each_netdev(&init_net, dev) {
4734                 if (__in6_dev_get(dev) == NULL)
4735                         continue;
4736                 addrconf_ifdown(dev, 1);
4737         }
4738         addrconf_ifdown(init_net.loopback_dev, 2);
4739
4740         /*
4741          *      Check hash table.
4742          */
4743         spin_lock_bh(&addrconf_hash_lock);
4744         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4745                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4746         spin_unlock_bh(&addrconf_hash_lock);
4747
4748         del_timer(&addr_chk_timer);
4749         rtnl_unlock();
4750 }