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 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 533 {
 534         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 535         kfree(ifp);
 536 }
 537
 538 /* Nobody refers to this ifaddr, destroy it */
 539 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 540 {
 541         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 542
 543 #ifdef NET_REFCNT_DEBUG
 544         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 545 #endif
 546
 547         in6_dev_put(ifp->idev);
 548
 549         if (del_timer(&ifp->timer))
 550                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 551
 552         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 553                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 554                 return;
 555         }
 556         dst_release(&ifp->rt->dst);
 557
 558         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 559 }
 560
 561 static void
 562 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 563 {
 564         struct list_head *p;
 565         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 566
 567         /*
 568          * Each device address list is sorted in order of scope -
 569          * global before linklocal.
 570          */
 571         list_for_each(p, &idev->addr_list) {
 572                 struct inet6_ifaddr *ifa
 573                         = list_entry(p, struct inet6_ifaddr, if_list);
 574                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 575                         break;
 576         }
 577
 578         list_add_tail(&ifp->if_list, p);
 579 }
 580
 581 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 582 {
 583         /*
 584          * We perform the hash function over the last 64 bits of the address
 585          * This will include the IEEE address token on links that support it.
 586          */
 587         return jhash_2words((__force u32)addr->s6_addr32[2],
 588                             (__force u32)addr->s6_addr32[3], 0)
 589                 & (IN6_ADDR_HSIZE - 1);
 590 }
 591
 592 /* On success it returns ifp with increased reference count */
 593
 594 static struct inet6_ifaddr *
 595 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 596               int scope, u32 flags)
 597 {
 598         struct inet6_ifaddr *ifa = NULL;
 599         struct rt6_info *rt;
 600         unsigned int hash;
 601         int err = 0;
 602         int addr_type = ipv6_addr_type(addr);
 603
 604         if (addr_type == IPV6_ADDR_ANY ||
 605             addr_type & IPV6_ADDR_MULTICAST ||
 606             (!(idev->dev->flags & IFF_LOOPBACK) &&
 607              addr_type & IPV6_ADDR_LOOPBACK))
 608                 return ERR_PTR(-EADDRNOTAVAIL);
 609
 610         rcu_read_lock_bh();
 611         if (idev->dead) {
 612                 err = -ENODEV;                  /*XXX*/
 613                 goto out2;
 614         }
 615
 616         if (idev->cnf.disable_ipv6) {
 617                 err = -EACCES;
 618                 goto out2;
 619         }
 620
 621         spin_lock(&addrconf_hash_lock);
 622
 623         /* Ignore adding duplicate addresses on an interface */
 624         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 625                 ADBG(("ipv6_add_addr: already assigned\n"));
 626                 err = -EEXIST;
 627                 goto out;
 628         }
 629
 630         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 631
 632         if (ifa == NULL) {
 633                 ADBG(("ipv6_add_addr: malloc failed\n"));
 634                 err = -ENOBUFS;
 635                 goto out;
 636         }
 637
 638         rt = addrconf_dst_alloc(idev, addr, 0);
 639         if (IS_ERR(rt)) {
 640                 err = PTR_ERR(rt);
 641                 goto out;
 642         }
 643
 644         ipv6_addr_copy(&ifa->addr, addr);
 645
 646         spin_lock_init(&ifa->lock);
 647         spin_lock_init(&ifa->state_lock);
 648         init_timer(&ifa->timer);
 649         INIT_HLIST_NODE(&ifa->addr_lst);
 650         ifa->timer.data = (unsigned long) ifa;
 651         ifa->scope = scope;
 652         ifa->prefix_len = pfxlen;
 653         ifa->flags = flags | IFA_F_TENTATIVE;
 654         ifa->cstamp = ifa->tstamp = jiffies;
 655
 656         ifa->rt = rt;
 657
 658         /*
 659          * part one of RFC 4429, section 3.3
 660          * We should not configure an address as
 661          * optimistic if we do not yet know the link
 662          * layer address of our nexhop router
 663          */
 664
 665         if (rt->rt6i_nexthop == NULL)
 666                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 667
 668         ifa->idev = idev;
 669         in6_dev_hold(idev);
 670         /* For caller */
 671         in6_ifa_hold(ifa);
 672
 673         /* Add to big hash table */
 674         hash = ipv6_addr_hash(addr);
 675
 676         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 677         spin_unlock(&addrconf_hash_lock);
 678
 679         write_lock(&idev->lock);
 680         /* Add to inet6_dev unicast addr list. */
 681         ipv6_link_dev_addr(idev, ifa);
 682
 683 #ifdef CONFIG_IPV6_PRIVACY
 684         if (ifa->flags&IFA_F_TEMPORARY) {
 685                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 686                 in6_ifa_hold(ifa);
 687         }
 688 #endif
 689
 690         in6_ifa_hold(ifa);
 691         write_unlock(&idev->lock);
 692 out2:
 693         rcu_read_unlock_bh();
 694
 695         if (likely(err == 0))
 696                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 697         else {
 698                 kfree(ifa);
 699                 ifa = ERR_PTR(err);
 700         }
 701
 702         return ifa;
 703 out:
 704         spin_unlock(&addrconf_hash_lock);
 705         goto out2;
 706 }
 707
 708 /* This function wants to get referenced ifp and releases it before return */
 709
 710 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 711 {
 712         struct inet6_ifaddr *ifa, *ifn;
 713         struct inet6_dev *idev = ifp->idev;
 714         int state;
 715         int deleted = 0, onlink = 0;
 716         unsigned long expires = jiffies;
 717
 718         spin_lock_bh(&ifp->state_lock);
 719         state = ifp->state;
 720         ifp->state = INET6_IFADDR_STATE_DEAD;
 721         spin_unlock_bh(&ifp->state_lock);
 722
 723         if (state == INET6_IFADDR_STATE_DEAD)
 724                 goto out;
 725
 726         spin_lock_bh(&addrconf_hash_lock);
 727         hlist_del_init_rcu(&ifp->addr_lst);
 728         spin_unlock_bh(&addrconf_hash_lock);
 729
 730         write_lock_bh(&idev->lock);
 731 #ifdef CONFIG_IPV6_PRIVACY
 732         if (ifp->flags&IFA_F_TEMPORARY) {
 733                 list_del(&ifp->tmp_list);
 734                 if (ifp->ifpub) {
 735                         in6_ifa_put(ifp->ifpub);
 736                         ifp->ifpub = NULL;
 737                 }
 738                 __in6_ifa_put(ifp);
 739         }
 740 #endif
 741
 742         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 743                 if (ifa == ifp) {
 744                         list_del_init(&ifp->if_list);
 745                         __in6_ifa_put(ifp);
 746
 747                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 748                                 break;
 749                         deleted = 1;
 750                         continue;
 751                 } else if (ifp->flags & IFA_F_PERMANENT) {
 752                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 753                                               ifp->prefix_len)) {
 754                                 if (ifa->flags & IFA_F_PERMANENT) {
 755                                         onlink = 1;
 756                                         if (deleted)
 757                                                 break;
 758                                 } else {
 759                                         unsigned long lifetime;
 760
 761                                         if (!onlink)
 762                                                 onlink = -1;
 763
 764                                         spin_lock(&ifa->lock);
 765
 766                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 767                                         /*
 768                                          * Note: Because this address is
 769                                          * not permanent, lifetime <
 770                                          * LONG_MAX / HZ here.
 771                                          */
 772                                         if (time_before(expires,
 773                                                         ifa->tstamp + lifetime * HZ))
 774                                                 expires = ifa->tstamp + lifetime * HZ;
 775                                         spin_unlock(&ifa->lock);
 776                                 }
 777                         }
 778                 }
 779         }
 780         write_unlock_bh(&idev->lock);
 781
 782         addrconf_del_timer(ifp);
 783
 784         ipv6_ifa_notify(RTM_DELADDR, ifp);
 785
 786         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 787
 788         /*
 789          * Purge or update corresponding prefix
 790          *
 791          * 1) we don't purge prefix here if address was not permanent.
 792          *    prefix is managed by its own lifetime.
 793          * 2) if there're no addresses, delete prefix.
 794          * 3) if there're still other permanent address(es),
 795          *    corresponding prefix is still permanent.
 796          * 4) otherwise, update prefix lifetime to the
 797          *    longest valid lifetime among the corresponding
 798          *    addresses on the device.
 799          *    Note: subsequent RA will update lifetime.
 800          *
 801          * --yoshfuji
 802          */
 803         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 804                 struct in6_addr prefix;
 805                 struct rt6_info *rt;
 806                 struct net *net = dev_net(ifp->idev->dev);
 807                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 808                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 809
 810                 if (rt && addrconf_is_prefix_route(rt)) {
 811                         if (onlink == 0) {
 812                                 ip6_del_rt(rt);
 813                                 rt = NULL;
 814                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 815                                 rt->rt6i_expires = expires;
 816                                 rt->rt6i_flags |= RTF_EXPIRES;
 817                         }
 818                 }
 819                 dst_release(&rt->dst);
 820         }
 821
 822 out:
 823         in6_ifa_put(ifp);
 824 }
 825
 826 #ifdef CONFIG_IPV6_PRIVACY
 827 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 828 {
 829         struct inet6_dev *idev = ifp->idev;
 830         struct in6_addr addr, *tmpaddr;
 831         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 832         unsigned long regen_advance;
 833         int tmp_plen;
 834         int ret = 0;
 835         int max_addresses;
 836         u32 addr_flags;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         age = (jiffies - ifp->tstamp) / HZ;
 882         tmp_valid_lft = min_t(__u32,
 883                               ifp->valid_lft,
 884                               idev->cnf.temp_valid_lft + age);
 885         tmp_prefered_lft = min_t(__u32,
 886                                  ifp->prefered_lft,
 887                                  idev->cnf.temp_prefered_lft + age -
 888                                  idev->cnf.max_desync_factor);
 889         tmp_plen = ifp->prefix_len;
 890         max_addresses = idev->cnf.max_addresses;
 891         tmp_cstamp = ifp->cstamp;
 892         tmp_tstamp = ifp->tstamp;
 893         spin_unlock_bh(&ifp->lock);
 894
 895         regen_advance = idev->cnf.regen_max_retry *
 896                         idev->cnf.dad_transmits *
 897                         idev->nd_parms->retrans_time / HZ;
 898         write_unlock(&idev->lock);
 899
 900         /* A temporary address is created only if this calculated Preferred
 901          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 902          * an implementation must not create a temporary address with a zero
 903          * Preferred Lifetime.
 904          */
 905         if (tmp_prefered_lft <= regen_advance) {
 906                 in6_ifa_put(ifp);
 907                 in6_dev_put(idev);
 908                 ret = -1;
 909                 goto out;
 910         }
 911
 912         addr_flags = IFA_F_TEMPORARY;
 913         /* set in addrconf_prefix_rcv() */
 914         if (ifp->flags & IFA_F_OPTIMISTIC)
 915                 addr_flags |= IFA_F_OPTIMISTIC;
 916
 917         ift = !max_addresses ||
 918               ipv6_count_addresses(idev) < max_addresses ?
 919                 ipv6_add_addr(idev, &addr, tmp_plen,
 920                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 921                               addr_flags) : NULL;
 922         if (!ift || IS_ERR(ift)) {
 923                 in6_ifa_put(ifp);
 924                 in6_dev_put(idev);
 925                 printk(KERN_INFO
 926                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 927                 tmpaddr = &addr;
 928                 write_lock(&idev->lock);
 929                 goto retry;
 930         }
 931
 932         spin_lock_bh(&ift->lock);
 933         ift->ifpub = ifp;
 934         ift->valid_lft = tmp_valid_lft;
 935         ift->prefered_lft = tmp_prefered_lft;
 936         ift->cstamp = tmp_cstamp;
 937         ift->tstamp = tmp_tstamp;
 938         spin_unlock_bh(&ift->lock);
 939
 940         addrconf_dad_start(ift, 0);
 941         in6_ifa_put(ift);
 942         in6_dev_put(idev);
 943 out:
 944         return ret;
 945 }
 946 #endif
 947
 948 /*
 949  *      Choose an appropriate source address (RFC3484)
 950  */
 951 enum {
 952         IPV6_SADDR_RULE_INIT = 0,
 953         IPV6_SADDR_RULE_LOCAL,
 954         IPV6_SADDR_RULE_SCOPE,
 955         IPV6_SADDR_RULE_PREFERRED,
 956 #ifdef CONFIG_IPV6_MIP6
 957         IPV6_SADDR_RULE_HOA,
 958 #endif
 959         IPV6_SADDR_RULE_OIF,
 960         IPV6_SADDR_RULE_LABEL,
 961 #ifdef CONFIG_IPV6_PRIVACY
 962         IPV6_SADDR_RULE_PRIVACY,
 963 #endif
 964         IPV6_SADDR_RULE_ORCHID,
 965         IPV6_SADDR_RULE_PREFIX,
 966         IPV6_SADDR_RULE_MAX
 967 };
 968
 969 struct ipv6_saddr_score {
 970         int                     rule;
 971         int                     addr_type;
 972         struct inet6_ifaddr     *ifa;
 973         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 974         int                     scopedist;
 975         int                     matchlen;
 976 };
 977
 978 struct ipv6_saddr_dst {
 979         const struct in6_addr *addr;
 980         int ifindex;
 981         int scope;
 982         int label;
 983         unsigned int prefs;
 984 };
 985
 986 static inline int ipv6_saddr_preferred(int type)
 987 {
 988         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 989                 return 1;
 990         return 0;
 991 }
 992
 993 static int ipv6_get_saddr_eval(struct net *net,
 994                                struct ipv6_saddr_score *score,
 995                                struct ipv6_saddr_dst *dst,
 996                                int i)
 997 {
 998         int ret;
 999
1000         if (i <= score->rule) {
1001                 switch (i) {
1002                 case IPV6_SADDR_RULE_SCOPE:
1003                         ret = score->scopedist;
1004                         break;
1005                 case IPV6_SADDR_RULE_PREFIX:
1006                         ret = score->matchlen;
1007                         break;
1008                 default:
1009                         ret = !!test_bit(i, score->scorebits);
1010                 }
1011                 goto out;
1012         }
1013
1014         switch (i) {
1015         case IPV6_SADDR_RULE_INIT:
1016                 /* Rule 0: remember if hiscore is not ready yet */
1017                 ret = !!score->ifa;
1018                 break;
1019         case IPV6_SADDR_RULE_LOCAL:
1020                 /* Rule 1: Prefer same address */
1021                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1022                 break;
1023         case IPV6_SADDR_RULE_SCOPE:
1024                 /* Rule 2: Prefer appropriate scope
1025                  *
1026                  *      ret
1027                  *       ^
1028                  *    -1 |  d 15
1029                  *    ---+--+-+---> scope
1030                  *       |
1031                  *       |             d is scope of the destination.
1032                  *  B-d  |  \
1033                  *       |   \      <- smaller scope is better if
1034                  *  B-15 |    \        if scope is enough for destinaion.
1035                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1036                  * d-C-1 | /
1037                  *       |/         <- greater is better
1038                  *   -C  /             if scope is not enough for destination.
1039                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1040                  *
1041                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1042                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1043                  * Assume B = 0 and we get C > 29.
1044                  */
1045                 ret = __ipv6_addr_src_scope(score->addr_type);
1046                 if (ret >= dst->scope)
1047                         ret = -ret;
1048                 else
1049                         ret -= 128;     /* 30 is enough */
1050                 score->scopedist = ret;
1051                 break;
1052         case IPV6_SADDR_RULE_PREFERRED:
1053                 /* Rule 3: Avoid deprecated and optimistic addresses */
1054                 ret = ipv6_saddr_preferred(score->addr_type) ||
1055                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1056                 break;
1057 #ifdef CONFIG_IPV6_MIP6
1058         case IPV6_SADDR_RULE_HOA:
1059             {
1060                 /* Rule 4: Prefer home address */
1061                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1062                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1063                 break;
1064             }
1065 #endif
1066         case IPV6_SADDR_RULE_OIF:
1067                 /* Rule 5: Prefer outgoing interface */
1068                 ret = (!dst->ifindex ||
1069                        dst->ifindex == score->ifa->idev->dev->ifindex);
1070                 break;
1071         case IPV6_SADDR_RULE_LABEL:
1072                 /* Rule 6: Prefer matching label */
1073                 ret = ipv6_addr_label(net,
1074                                       &score->ifa->addr, score->addr_type,
1075                                       score->ifa->idev->dev->ifindex) == dst->label;
1076                 break;
1077 #ifdef CONFIG_IPV6_PRIVACY
1078         case IPV6_SADDR_RULE_PRIVACY:
1079             {
1080                 /* Rule 7: Prefer public address
1081                  * Note: prefer temporary address if use_tempaddr >= 2
1082                  */
1083                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1084                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1085                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1086                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1087                 break;
1088             }
1089 #endif
1090         case IPV6_SADDR_RULE_ORCHID:
1091                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1092                  *          non-ORCHID vs non-ORCHID
1093                  */
1094                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1095                         ipv6_addr_orchid(dst->addr));
1096                 break;
1097         case IPV6_SADDR_RULE_PREFIX:
1098                 /* Rule 8: Use longest matching prefix */
1099                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1100                                                        dst->addr);
1101                 break;
1102         default:
1103                 ret = 0;
1104         }
1105
1106         if (ret)
1107                 __set_bit(i, score->scorebits);
1108         score->rule = i;
1109 out:
1110         return ret;
1111 }
1112
1113 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1114                        const struct in6_addr *daddr, unsigned int prefs,
1115                        struct in6_addr *saddr)
1116 {
1117         struct ipv6_saddr_score scores[2],
1118                                 *score = &scores[0], *hiscore = &scores[1];
1119         struct ipv6_saddr_dst dst;
1120         struct net_device *dev;
1121         int dst_type;
1122
1123         dst_type = __ipv6_addr_type(daddr);
1124         dst.addr = daddr;
1125         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1126         dst.scope = __ipv6_addr_src_scope(dst_type);
1127         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1128         dst.prefs = prefs;
1129
1130         hiscore->rule = -1;
1131         hiscore->ifa = NULL;
1132
1133         rcu_read_lock();
1134
1135         for_each_netdev_rcu(net, dev) {
1136                 struct inet6_dev *idev;
1137
1138                 /* Candidate Source Address (section 4)
1139                  *  - multicast and link-local destination address,
1140                  *    the set of candidate source address MUST only
1141                  *    include addresses assigned to interfaces
1142                  *    belonging to the same link as the outgoing
1143                  *    interface.
1144                  * (- For site-local destination addresses, the
1145                  *    set of candidate source addresses MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same site as the outgoing
1148                  *    interface.)
1149                  */
1150                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1151                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1152                     dst.ifindex && dev->ifindex != dst.ifindex)
1153                         continue;
1154
1155                 idev = __in6_dev_get(dev);
1156                 if (!idev)
1157                         continue;
1158
1159                 read_lock_bh(&idev->lock);
1160                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1161                         int i;
1162
1163                         /*
1164                          * - Tentative Address (RFC2462 section 5.4)
1165                          *  - A tentative address is not considered
1166                          *    "assigned to an interface" in the traditional
1167                          *    sense, unless it is also flagged as optimistic.
1168                          * - Candidate Source Address (section 4)
1169                          *  - In any case, anycast addresses, multicast
1170                          *    addresses, and the unspecified address MUST
1171                          *    NOT be included in a candidate set.
1172                          */
1173                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1174                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1175                                 continue;
1176
1177                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1178
1179                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1180                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1181                                 LIMIT_NETDEBUG(KERN_DEBUG
1182                                                "ADDRCONF: unspecified / multicast address "
1183                                                "assigned as unicast address on %s",
1184                                                dev->name);
1185                                 continue;
1186                         }
1187
1188                         score->rule = -1;
1189                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1190
1191                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1192                                 int minihiscore, miniscore;
1193
1194                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1195                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1196
1197                                 if (minihiscore > miniscore) {
1198                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1199                                             score->scopedist > 0) {
1200                                                 /*
1201                                                  * special case:
1202                                                  * each remaining entry
1203                                                  * has too small (not enough)
1204                                                  * scope, because ifa entries
1205                                                  * are sorted by their scope
1206                                                  * values.
1207                                                  */
1208                                                 goto try_nextdev;
1209                                         }
1210                                         break;
1211                                 } else if (minihiscore < miniscore) {
1212                                         if (hiscore->ifa)
1213                                                 in6_ifa_put(hiscore->ifa);
1214
1215                                         in6_ifa_hold(score->ifa);
1216
1217                                         swap(hiscore, score);
1218
1219                                         /* restore our iterator */
1220                                         score->ifa = hiscore->ifa;
1221
1222                                         break;
1223                                 }
1224                         }
1225                 }
1226 try_nextdev:
1227                 read_unlock_bh(&idev->lock);
1228         }
1229         rcu_read_unlock();
1230
1231         if (!hiscore->ifa)
1232                 return -EADDRNOTAVAIL;
1233
1234         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1235         in6_ifa_put(hiscore->ifa);
1236         return 0;
1237 }
1238 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1239
1240 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1241                     unsigned char banned_flags)
1242 {
1243         struct inet6_dev *idev;
1244         int err = -EADDRNOTAVAIL;
1245
1246         rcu_read_lock();
1247         idev = __in6_dev_get(dev);
1248         if (idev) {
1249                 struct inet6_ifaddr *ifp;
1250
1251                 read_lock_bh(&idev->lock);
1252                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1253                         if (ifp->scope == IFA_LINK &&
1254                             !(ifp->flags & banned_flags)) {
1255                                 ipv6_addr_copy(addr, &ifp->addr);
1256                                 err = 0;
1257                                 break;
1258                         }
1259                 }
1260                 read_unlock_bh(&idev->lock);
1261         }
1262         rcu_read_unlock();
1263         return err;
1264 }
1265
1266 static int ipv6_count_addresses(struct inet6_dev *idev)
1267 {
1268         int cnt = 0;
1269         struct inet6_ifaddr *ifp;
1270
1271         read_lock_bh(&idev->lock);
1272         list_for_each_entry(ifp, &idev->addr_list, if_list)
1273                 cnt++;
1274         read_unlock_bh(&idev->lock);
1275         return cnt;
1276 }
1277
1278 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1279                   struct net_device *dev, int strict)
1280 {
1281         struct inet6_ifaddr *ifp;
1282         struct hlist_node *node;
1283         unsigned int hash = ipv6_addr_hash(addr);
1284
1285         rcu_read_lock_bh();
1286         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1287                 if (!net_eq(dev_net(ifp->idev->dev), net))
1288                         continue;
1289                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1290                     !(ifp->flags&IFA_F_TENTATIVE) &&
1291                     (dev == NULL || ifp->idev->dev == dev ||
1292                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1293                         rcu_read_unlock_bh();
1294                         return 1;
1295                 }
1296         }
1297
1298         rcu_read_unlock_bh();
1299         return 0;
1300 }
1301 EXPORT_SYMBOL(ipv6_chk_addr);
1302
1303 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1304                                struct net_device *dev)
1305 {
1306         unsigned int hash = ipv6_addr_hash(addr);
1307         struct inet6_ifaddr *ifp;
1308         struct hlist_node *node;
1309
1310         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1311                 if (!net_eq(dev_net(ifp->idev->dev), net))
1312                         continue;
1313                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1314                         if (dev == NULL || ifp->idev->dev == dev)
1315                                 return true;
1316                 }
1317         }
1318         return false;
1319 }
1320
1321 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1322 {
1323         struct inet6_dev *idev;
1324         struct inet6_ifaddr *ifa;
1325         int     onlink;
1326
1327         onlink = 0;
1328         rcu_read_lock();
1329         idev = __in6_dev_get(dev);
1330         if (idev) {
1331                 read_lock_bh(&idev->lock);
1332                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1333                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1334                                                    ifa->prefix_len);
1335                         if (onlink)
1336                                 break;
1337                 }
1338                 read_unlock_bh(&idev->lock);
1339         }
1340         rcu_read_unlock();
1341         return onlink;
1342 }
1343
1344 EXPORT_SYMBOL(ipv6_chk_prefix);
1345
1346 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1347                                      struct net_device *dev, int strict)
1348 {
1349         struct inet6_ifaddr *ifp, *result = NULL;
1350         unsigned int hash = ipv6_addr_hash(addr);
1351         struct hlist_node *node;
1352
1353         rcu_read_lock_bh();
1354         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1355                 if (!net_eq(dev_net(ifp->idev->dev), net))
1356                         continue;
1357                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1358                         if (dev == NULL || ifp->idev->dev == dev ||
1359                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1360                                 result = ifp;
1361                                 in6_ifa_hold(ifp);
1362                                 break;
1363                         }
1364                 }
1365         }
1366         rcu_read_unlock_bh();
1367
1368         return result;
1369 }
1370
1371 /* Gets referenced address, destroys ifaddr */
1372
1373 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1374 {
1375         if (ifp->flags&IFA_F_PERMANENT) {
1376                 spin_lock_bh(&ifp->lock);
1377                 addrconf_del_timer(ifp);
1378                 ifp->flags |= IFA_F_TENTATIVE;
1379                 if (dad_failed)
1380                         ifp->flags |= IFA_F_DADFAILED;
1381                 spin_unlock_bh(&ifp->lock);
1382                 if (dad_failed)
1383                         ipv6_ifa_notify(0, ifp);
1384                 in6_ifa_put(ifp);
1385 #ifdef CONFIG_IPV6_PRIVACY
1386         } else if (ifp->flags&IFA_F_TEMPORARY) {
1387                 struct inet6_ifaddr *ifpub;
1388                 spin_lock_bh(&ifp->lock);
1389                 ifpub = ifp->ifpub;
1390                 if (ifpub) {
1391                         in6_ifa_hold(ifpub);
1392                         spin_unlock_bh(&ifp->lock);
1393                         ipv6_create_tempaddr(ifpub, ifp);
1394                         in6_ifa_put(ifpub);
1395                 } else {
1396                         spin_unlock_bh(&ifp->lock);
1397                 }
1398                 ipv6_del_addr(ifp);
1399 #endif
1400         } else
1401                 ipv6_del_addr(ifp);
1402 }
1403
1404 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1405 {
1406         int err = -ENOENT;
1407
1408         spin_lock(&ifp->state_lock);
1409         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1410                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1411                 err = 0;
1412         }
1413         spin_unlock(&ifp->state_lock);
1414
1415         return err;
1416 }
1417
1418 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1419 {
1420         struct inet6_dev *idev = ifp->idev;
1421
1422         if (addrconf_dad_end(ifp)) {
1423                 in6_ifa_put(ifp);
1424                 return;
1425         }
1426
1427         if (net_ratelimit())
1428                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1429                         ifp->idev->dev->name, &ifp->addr);
1430
1431         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1432                 struct in6_addr addr;
1433
1434                 addr.s6_addr32[0] = htonl(0xfe800000);
1435                 addr.s6_addr32[1] = 0;
1436
1437                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1438                     ipv6_addr_equal(&ifp->addr, &addr)) {
1439                         /* DAD failed for link-local based on MAC address */
1440                         idev->cnf.disable_ipv6 = 1;
1441
1442                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1443                                 ifp->idev->dev->name);
1444                 }
1445         }
1446
1447         addrconf_dad_stop(ifp, 1);
1448 }
1449
1450 /* Join to solicited addr multicast group. */
1451
1452 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1453 {
1454         struct in6_addr maddr;
1455
1456         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1457                 return;
1458
1459         addrconf_addr_solict_mult(addr, &maddr);
1460         ipv6_dev_mc_inc(dev, &maddr);
1461 }
1462
1463 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1464 {
1465         struct in6_addr maddr;
1466
1467         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1468                 return;
1469
1470         addrconf_addr_solict_mult(addr, &maddr);
1471         __ipv6_dev_mc_dec(idev, &maddr);
1472 }
1473
1474 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1475 {
1476         struct in6_addr addr;
1477         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1478         if (ipv6_addr_any(&addr))
1479                 return;
1480         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1481 }
1482
1483 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1484 {
1485         struct in6_addr addr;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         __ipv6_dev_ac_dec(ifp->idev, &addr);
1490 }
1491
1492 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1493 {
1494         if (dev->addr_len != ETH_ALEN)
1495                 return -1;
1496         memcpy(eui, dev->dev_addr, 3);
1497         memcpy(eui + 5, dev->dev_addr + 3, 3);
1498
1499         /*
1500          * The zSeries OSA network cards can be shared among various
1501          * OS instances, but the OSA cards have only one MAC address.
1502          * This leads to duplicate address conflicts in conjunction
1503          * with IPv6 if more than one instance uses the same card.
1504          *
1505          * The driver for these cards can deliver a unique 16-bit
1506          * identifier for each instance sharing the same card.  It is
1507          * placed instead of 0xFFFE in the interface identifier.  The
1508          * "u" bit of the interface identifier is not inverted in this
1509          * case.  Hence the resulting interface identifier has local
1510          * scope according to RFC2373.
1511          */
1512         if (dev->dev_id) {
1513                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1514                 eui[4] = dev->dev_id & 0xFF;
1515         } else {
1516                 eui[3] = 0xFF;
1517                 eui[4] = 0xFE;
1518                 eui[0] ^= 2;
1519         }
1520         return 0;
1521 }
1522
1523 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 {
1525         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526         if (dev->addr_len != ARCNET_ALEN)
1527                 return -1;
1528         memset(eui, 0, 7);
1529         eui[7] = *(u8*)dev->dev_addr;
1530         return 0;
1531 }
1532
1533 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 {
1535         if (dev->addr_len != INFINIBAND_ALEN)
1536                 return -1;
1537         memcpy(eui, dev->dev_addr + 12, 8);
1538         eui[0] |= 2;
1539         return 0;
1540 }
1541
1542 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 {
1544         if (addr == 0)
1545                 return -1;
1546         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1547                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1548                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1549                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1550                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1551                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1552         eui[1] = 0;
1553         eui[2] = 0x5E;
1554         eui[3] = 0xFE;
1555         memcpy(eui + 4, &addr, 4);
1556         return 0;
1557 }
1558
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560 {
1561         if (dev->priv_flags & IFF_ISATAP)
1562                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563         return -1;
1564 }
1565
1566 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1567 {
1568         switch (dev->type) {
1569         case ARPHRD_ETHER:
1570         case ARPHRD_FDDI:
1571         case ARPHRD_IEEE802_TR:
1572                 return addrconf_ifid_eui48(eui, dev);
1573         case ARPHRD_ARCNET:
1574                 return addrconf_ifid_arcnet(eui, dev);
1575         case ARPHRD_INFINIBAND:
1576                 return addrconf_ifid_infiniband(eui, dev);
1577         case ARPHRD_SIT:
1578                 return addrconf_ifid_sit(eui, dev);
1579         }
1580         return -1;
1581 }
1582
1583 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1584 {
1585         int err = -1;
1586         struct inet6_ifaddr *ifp;
1587
1588         read_lock_bh(&idev->lock);
1589         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1590                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1591                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1592                         err = 0;
1593                         break;
1594                 }
1595         }
1596         read_unlock_bh(&idev->lock);
1597         return err;
1598 }
1599
1600 #ifdef CONFIG_IPV6_PRIVACY
1601 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1602 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1603 {
1604 regen:
1605         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1606         idev->rndid[0] &= ~0x02;
1607
1608         /*
1609          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1610          * check if generated address is not inappropriate
1611          *
1612          *  - Reserved subnet anycast (RFC 2526)
1613          *      11111101 11....11 1xxxxxxx
1614          *  - ISATAP (RFC4214) 6.1
1615          *      00-00-5E-FE-xx-xx-xx-xx
1616          *  - value 0
1617          *  - XXX: already assigned to an address on the device
1618          */
1619         if (idev->rndid[0] == 0xfd &&
1620             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1621             (idev->rndid[7]&0x80))
1622                 goto regen;
1623         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1624                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1625                         goto regen;
1626                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1627                         goto regen;
1628         }
1629
1630         return 0;
1631 }
1632
1633 static void ipv6_regen_rndid(unsigned long data)
1634 {
1635         struct inet6_dev *idev = (struct inet6_dev *) data;
1636         unsigned long expires;
1637
1638         rcu_read_lock_bh();
1639         write_lock_bh(&idev->lock);
1640
1641         if (idev->dead)
1642                 goto out;
1643
1644         if (__ipv6_regen_rndid(idev) < 0)
1645                 goto out;
1646
1647         expires = jiffies +
1648                 idev->cnf.temp_prefered_lft * HZ -
1649                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1650                 idev->cnf.max_desync_factor * HZ;
1651         if (time_before(expires, jiffies)) {
1652                 printk(KERN_WARNING
1653                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1654                         idev->dev->name);
1655                 goto out;
1656         }
1657
1658         if (!mod_timer(&idev->regen_timer, expires))
1659                 in6_dev_hold(idev);
1660
1661 out:
1662         write_unlock_bh(&idev->lock);
1663         rcu_read_unlock_bh();
1664         in6_dev_put(idev);
1665 }
1666
1667 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1668         int ret = 0;
1669
1670         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1671                 ret = __ipv6_regen_rndid(idev);
1672         return ret;
1673 }
1674 #endif
1675
1676 /*
1677  *      Add prefix route.
1678  */
1679
1680 static void
1681 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1682                       unsigned long expires, u32 flags)
1683 {
1684         struct fib6_config cfg = {
1685                 .fc_table = RT6_TABLE_PREFIX,
1686                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1687                 .fc_ifindex = dev->ifindex,
1688                 .fc_expires = expires,
1689                 .fc_dst_len = plen,
1690                 .fc_flags = RTF_UP | flags,
1691                 .fc_nlinfo.nl_net = dev_net(dev),
1692                 .fc_protocol = RTPROT_KERNEL,
1693         };
1694
1695         ipv6_addr_copy(&cfg.fc_dst, pfx);
1696
1697         /* Prevent useless cloning on PtP SIT.
1698            This thing is done here expecting that the whole
1699            class of non-broadcast devices need not cloning.
1700          */
1701 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1702         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1703                 cfg.fc_flags |= RTF_NONEXTHOP;
1704 #endif
1705
1706         ip6_route_add(&cfg);
1707 }
1708
1709 /* Create "default" multicast route to the interface */
1710
1711 static void addrconf_add_mroute(struct net_device *dev)
1712 {
1713         struct fib6_config cfg = {
1714                 .fc_table = RT6_TABLE_LOCAL,
1715                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1716                 .fc_ifindex = dev->ifindex,
1717                 .fc_dst_len = 8,
1718                 .fc_flags = RTF_UP,
1719                 .fc_nlinfo.nl_net = dev_net(dev),
1720         };
1721
1722         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1723
1724         ip6_route_add(&cfg);
1725 }
1726
1727 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1728 static void sit_route_add(struct net_device *dev)
1729 {
1730         struct fib6_config cfg = {
1731                 .fc_table = RT6_TABLE_MAIN,
1732                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1733                 .fc_ifindex = dev->ifindex,
1734                 .fc_dst_len = 96,
1735                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1736                 .fc_nlinfo.nl_net = dev_net(dev),
1737         };
1738
1739         /* prefix length - 96 bits "::d.d.d.d" */
1740         ip6_route_add(&cfg);
1741 }
1742 #endif
1743
1744 static void addrconf_add_lroute(struct net_device *dev)
1745 {
1746         struct in6_addr addr;
1747
1748         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1749         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1750 }
1751
1752 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1753 {
1754         struct inet6_dev *idev;
1755
1756         ASSERT_RTNL();
1757
1758         idev = ipv6_find_idev(dev);
1759         if (!idev)
1760                 return ERR_PTR(-ENOBUFS);
1761
1762         if (idev->cnf.disable_ipv6)
1763                 return ERR_PTR(-EACCES);
1764
1765         /* Add default multicast route */
1766         addrconf_add_mroute(dev);
1767
1768         /* Add link local route */
1769         addrconf_add_lroute(dev);
1770         return idev;
1771 }
1772
1773 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1774 {
1775         struct prefix_info *pinfo;
1776         __u32 valid_lft;
1777         __u32 prefered_lft;
1778         int addr_type;
1779         struct inet6_dev *in6_dev;
1780         struct net *net = dev_net(dev);
1781
1782         pinfo = (struct prefix_info *) opt;
1783
1784         if (len < sizeof(struct prefix_info)) {
1785                 ADBG(("addrconf: prefix option too short\n"));
1786                 return;
1787         }
1788
1789         /*
1790          *      Validation checks ([ADDRCONF], page 19)
1791          */
1792
1793         addr_type = ipv6_addr_type(&pinfo->prefix);
1794
1795         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1796                 return;
1797
1798         valid_lft = ntohl(pinfo->valid);
1799         prefered_lft = ntohl(pinfo->prefered);
1800
1801         if (prefered_lft > valid_lft) {
1802                 if (net_ratelimit())
1803                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1804                 return;
1805         }
1806
1807         in6_dev = in6_dev_get(dev);
1808
1809         if (in6_dev == NULL) {
1810                 if (net_ratelimit())
1811                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1812                 return;
1813         }
1814
1815         /*
1816          *      Two things going on here:
1817          *      1) Add routes for on-link prefixes
1818          *      2) Configure prefixes with the auto flag set
1819          */
1820
1821         if (pinfo->onlink) {
1822                 struct rt6_info *rt;
1823                 unsigned long rt_expires;
1824
1825                 /* Avoid arithmetic overflow. Really, we could
1826                  * save rt_expires in seconds, likely valid_lft,
1827                  * but it would require division in fib gc, that it
1828                  * not good.
1829                  */
1830                 if (HZ > USER_HZ)
1831                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1832                 else
1833                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1834
1835                 if (addrconf_finite_timeout(rt_expires))
1836                         rt_expires *= HZ;
1837
1838                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1839                                 dev->ifindex, 1);
1840
1841                 if (rt && addrconf_is_prefix_route(rt)) {
1842                         /* Autoconf prefix route */
1843                         if (valid_lft == 0) {
1844                                 ip6_del_rt(rt);
1845                                 rt = NULL;
1846                         } else if (addrconf_finite_timeout(rt_expires)) {
1847                                 /* not infinity */
1848                                 rt->rt6i_expires = jiffies + rt_expires;
1849                                 rt->rt6i_flags |= RTF_EXPIRES;
1850                         } else {
1851                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1852                                 rt->rt6i_expires = 0;
1853                         }
1854                 } else if (valid_lft) {
1855                         clock_t expires = 0;
1856                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1857                         if (addrconf_finite_timeout(rt_expires)) {
1858                                 /* not infinity */
1859                                 flags |= RTF_EXPIRES;
1860                                 expires = jiffies_to_clock_t(rt_expires);
1861                         }
1862                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1863                                               dev, expires, flags);
1864                 }
1865                 if (rt)
1866                         dst_release(&rt->dst);
1867         }
1868
1869         /* Try to figure out our local address for this prefix */
1870
1871         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1872                 struct inet6_ifaddr * ifp;
1873                 struct in6_addr addr;
1874                 int create = 0, update_lft = 0;
1875
1876                 if (pinfo->prefix_len == 64) {
1877                         memcpy(&addr, &pinfo->prefix, 8);
1878                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1879                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1880                                 in6_dev_put(in6_dev);
1881                                 return;
1882                         }
1883                         goto ok;
1884                 }
1885                 if (net_ratelimit())
1886                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1887                                pinfo->prefix_len);
1888                 in6_dev_put(in6_dev);
1889                 return;
1890
1891 ok:
1892
1893                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1894
1895                 if (ifp == NULL && valid_lft) {
1896                         int max_addresses = in6_dev->cnf.max_addresses;
1897                         u32 addr_flags = 0;
1898
1899 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1900                         if (in6_dev->cnf.optimistic_dad &&
1901                             !net->ipv6.devconf_all->forwarding)
1902                                 addr_flags = IFA_F_OPTIMISTIC;
1903 #endif
1904
1905                         /* Do not allow to create too much of autoconfigured
1906                          * addresses; this would be too easy way to crash kernel.
1907                          */
1908                         if (!max_addresses ||
1909                             ipv6_count_addresses(in6_dev) < max_addresses)
1910                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1911                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1912                                                     addr_flags);
1913
1914                         if (!ifp || IS_ERR(ifp)) {
1915                                 in6_dev_put(in6_dev);
1916                                 return;
1917                         }
1918
1919                         update_lft = create = 1;
1920                         ifp->cstamp = jiffies;
1921                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1922                 }
1923
1924                 if (ifp) {
1925                         int flags;
1926                         unsigned long now;
1927 #ifdef CONFIG_IPV6_PRIVACY
1928                         struct inet6_ifaddr *ift;
1929 #endif
1930                         u32 stored_lft;
1931
1932                         /* update lifetime (RFC2462 5.5.3 e) */
1933                         spin_lock(&ifp->lock);
1934                         now = jiffies;
1935                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1936                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1937                         else
1938                                 stored_lft = 0;
1939                         if (!update_lft && stored_lft) {
1940                                 if (valid_lft > MIN_VALID_LIFETIME ||
1941                                     valid_lft > stored_lft)
1942                                         update_lft = 1;
1943                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1944                                         /* valid_lft <= stored_lft is always true */
1945                                         /*
1946                                          * RFC 4862 Section 5.5.3e:
1947                                          * "Note that the preferred lifetime of
1948                                          *  the corresponding address is always
1949                                          *  reset to the Preferred Lifetime in
1950                                          *  the received Prefix Information
1951                                          *  option, regardless of whether the
1952                                          *  valid lifetime is also reset or
1953                                          *  ignored."
1954                                          *
1955                                          *  So if the preferred lifetime in
1956                                          *  this advertisement is different
1957                                          *  than what we have stored, but the
1958                                          *  valid lifetime is invalid, just
1959                                          *  reset prefered_lft.
1960                                          *
1961                                          *  We must set the valid lifetime
1962                                          *  to the stored lifetime since we'll
1963                                          *  be updating the timestamp below,
1964                                          *  else we'll set it back to the
1965                                          *  minimum.
1966                                          */
1967                                         if (prefered_lft != ifp->prefered_lft) {
1968                                                 valid_lft = stored_lft;
1969                                                 update_lft = 1;
1970                                         }
1971                                 } else {
1972                                         valid_lft = MIN_VALID_LIFETIME;
1973                                         if (valid_lft < prefered_lft)
1974                                                 prefered_lft = valid_lft;
1975                                         update_lft = 1;
1976                                 }
1977                         }
1978
1979                         if (update_lft) {
1980                                 ifp->valid_lft = valid_lft;
1981                                 ifp->prefered_lft = prefered_lft;
1982                                 ifp->tstamp = now;
1983                                 flags = ifp->flags;
1984                                 ifp->flags &= ~IFA_F_DEPRECATED;
1985                                 spin_unlock(&ifp->lock);
1986
1987                                 if (!(flags&IFA_F_TENTATIVE))
1988                                         ipv6_ifa_notify(0, ifp);
1989                         } else
1990                                 spin_unlock(&ifp->lock);
1991
1992 #ifdef CONFIG_IPV6_PRIVACY
1993                         read_lock_bh(&in6_dev->lock);
1994                         /* update all temporary addresses in the list */
1995                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1996                                 /*
1997                                  * When adjusting the lifetimes of an existing
1998                                  * temporary address, only lower the lifetimes.
1999                                  * Implementations must not increase the
2000                                  * lifetimes of an existing temporary address
2001                                  * when processing a Prefix Information Option.
2002                                  */
2003                                 if (ifp != ift->ifpub)
2004                                         continue;
2005
2006                                 spin_lock(&ift->lock);
2007                                 flags = ift->flags;
2008                                 if (ift->valid_lft > valid_lft &&
2009                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2010                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2011                                 if (ift->prefered_lft > prefered_lft &&
2012                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2013                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2014                                 spin_unlock(&ift->lock);
2015                                 if (!(flags&IFA_F_TENTATIVE))
2016                                         ipv6_ifa_notify(0, ift);
2017                         }
2018
2019                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2020                                 /*
2021                                  * When a new public address is created as described in [ADDRCONF],
2022                                  * also create a new temporary address. Also create a temporary
2023                                  * address if it's enabled but no temporary address currently exists.
2024                                  */
2025                                 read_unlock_bh(&in6_dev->lock);
2026                                 ipv6_create_tempaddr(ifp, NULL);
2027                         } else {
2028                                 read_unlock_bh(&in6_dev->lock);
2029                         }
2030 #endif
2031                         in6_ifa_put(ifp);
2032                         addrconf_verify(0);
2033                 }
2034         }
2035         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2036         in6_dev_put(in6_dev);
2037 }
2038
2039 /*
2040  *      Set destination address.
2041  *      Special case for SIT interfaces where we create a new "virtual"
2042  *      device.
2043  */
2044 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2045 {
2046         struct in6_ifreq ireq;
2047         struct net_device *dev;
2048         int err = -EINVAL;
2049
2050         rtnl_lock();
2051
2052         err = -EFAULT;
2053         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2054                 goto err_exit;
2055
2056         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2057
2058         err = -ENODEV;
2059         if (dev == NULL)
2060                 goto err_exit;
2061
2062 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2063         if (dev->type == ARPHRD_SIT) {
2064                 const struct net_device_ops *ops = dev->netdev_ops;
2065                 struct ifreq ifr;
2066                 struct ip_tunnel_parm p;
2067
2068                 err = -EADDRNOTAVAIL;
2069                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2070                         goto err_exit;
2071
2072                 memset(&p, 0, sizeof(p));
2073                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2074                 p.iph.saddr = 0;
2075                 p.iph.version = 4;
2076                 p.iph.ihl = 5;
2077                 p.iph.protocol = IPPROTO_IPV6;
2078                 p.iph.ttl = 64;
2079                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2080
2081                 if (ops->ndo_do_ioctl) {
2082                         mm_segment_t oldfs = get_fs();
2083
2084                         set_fs(KERNEL_DS);
2085                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2086                         set_fs(oldfs);
2087                 } else
2088                         err = -EOPNOTSUPP;
2089
2090                 if (err == 0) {
2091                         err = -ENOBUFS;
2092                         dev = __dev_get_by_name(net, p.name);
2093                         if (!dev)
2094                                 goto err_exit;
2095                         err = dev_open(dev);
2096                 }
2097         }
2098 #endif
2099
2100 err_exit:
2101         rtnl_unlock();
2102         return err;
2103 }
2104
2105 /*
2106  *      Manual configuration of address on an interface
2107  */
2108 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2109                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2110                           __u32 valid_lft)
2111 {
2112         struct inet6_ifaddr *ifp;
2113         struct inet6_dev *idev;
2114         struct net_device *dev;
2115         int scope;
2116         u32 flags;
2117         clock_t expires;
2118         unsigned long timeout;
2119
2120         ASSERT_RTNL();
2121
2122         if (plen > 128)
2123                 return -EINVAL;
2124
2125         /* check the lifetime */
2126         if (!valid_lft || prefered_lft > valid_lft)
2127                 return -EINVAL;
2128
2129         dev = __dev_get_by_index(net, ifindex);
2130         if (!dev)
2131                 return -ENODEV;
2132
2133         idev = addrconf_add_dev(dev);
2134         if (IS_ERR(idev))
2135                 return PTR_ERR(idev);
2136
2137         scope = ipv6_addr_scope(pfx);
2138
2139         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2140         if (addrconf_finite_timeout(timeout)) {
2141                 expires = jiffies_to_clock_t(timeout * HZ);
2142                 valid_lft = timeout;
2143                 flags = RTF_EXPIRES;
2144         } else {
2145                 expires = 0;
2146                 flags = 0;
2147                 ifa_flags |= IFA_F_PERMANENT;
2148         }
2149
2150         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2151         if (addrconf_finite_timeout(timeout)) {
2152                 if (timeout == 0)
2153                         ifa_flags |= IFA_F_DEPRECATED;
2154                 prefered_lft = timeout;
2155         }
2156
2157         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2158
2159         if (!IS_ERR(ifp)) {
2160                 spin_lock_bh(&ifp->lock);
2161                 ifp->valid_lft = valid_lft;
2162                 ifp->prefered_lft = prefered_lft;
2163                 ifp->tstamp = jiffies;
2164                 spin_unlock_bh(&ifp->lock);
2165
2166                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2167                                       expires, flags);
2168                 /*
2169                  * Note that section 3.1 of RFC 4429 indicates
2170                  * that the Optimistic flag should not be set for
2171                  * manually configured addresses
2172                  */
2173                 addrconf_dad_start(ifp, 0);
2174                 in6_ifa_put(ifp);
2175                 addrconf_verify(0);
2176                 return 0;
2177         }
2178
2179         return PTR_ERR(ifp);
2180 }
2181
2182 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2183                           unsigned int plen)
2184 {
2185         struct inet6_ifaddr *ifp;
2186         struct inet6_dev *idev;
2187         struct net_device *dev;
2188
2189         if (plen > 128)
2190                 return -EINVAL;
2191
2192         dev = __dev_get_by_index(net, ifindex);
2193         if (!dev)
2194                 return -ENODEV;
2195
2196         if ((idev = __in6_dev_get(dev)) == NULL)
2197                 return -ENXIO;
2198
2199         read_lock_bh(&idev->lock);
2200         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2201                 if (ifp->prefix_len == plen &&
2202                     ipv6_addr_equal(pfx, &ifp->addr)) {
2203                         in6_ifa_hold(ifp);
2204                         read_unlock_bh(&idev->lock);
2205
2206                         ipv6_del_addr(ifp);
2207
2208                         /* If the last address is deleted administratively,
2209                            disable IPv6 on this interface.
2210                          */
2211                         if (list_empty(&idev->addr_list))
2212                                 addrconf_ifdown(idev->dev, 1);
2213                         return 0;
2214                 }
2215         }
2216         read_unlock_bh(&idev->lock);
2217         return -EADDRNOTAVAIL;
2218 }
2219
2220
2221 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2222 {
2223         struct in6_ifreq ireq;
2224         int err;
2225
2226         if (!capable(CAP_NET_ADMIN))
2227                 return -EPERM;
2228
2229         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2230                 return -EFAULT;
2231
2232         rtnl_lock();
2233         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2234                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2235                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2236         rtnl_unlock();
2237         return err;
2238 }
2239
2240 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2241 {
2242         struct in6_ifreq ireq;
2243         int err;
2244
2245         if (!capable(CAP_NET_ADMIN))
2246                 return -EPERM;
2247
2248         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2249                 return -EFAULT;
2250
2251         rtnl_lock();
2252         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2253                              ireq.ifr6_prefixlen);
2254         rtnl_unlock();
2255         return err;
2256 }
2257
2258 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2259                      int plen, int scope)
2260 {
2261         struct inet6_ifaddr *ifp;
2262
2263         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2264         if (!IS_ERR(ifp)) {
2265                 spin_lock_bh(&ifp->lock);
2266                 ifp->flags &= ~IFA_F_TENTATIVE;
2267                 spin_unlock_bh(&ifp->lock);
2268                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2269                 in6_ifa_put(ifp);
2270         }
2271 }
2272
2273 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2274 static void sit_add_v4_addrs(struct inet6_dev *idev)
2275 {
2276         struct in6_addr addr;
2277         struct net_device *dev;
2278         struct net *net = dev_net(idev->dev);
2279         int scope;
2280
2281         ASSERT_RTNL();
2282
2283         memset(&addr, 0, sizeof(struct in6_addr));
2284         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2285
2286         if (idev->dev->flags&IFF_POINTOPOINT) {
2287                 addr.s6_addr32[0] = htonl(0xfe800000);
2288                 scope = IFA_LINK;
2289         } else {
2290                 scope = IPV6_ADDR_COMPATv4;
2291         }
2292
2293         if (addr.s6_addr32[3]) {
2294                 add_addr(idev, &addr, 128, scope);
2295                 return;
2296         }
2297
2298         for_each_netdev(net, dev) {
2299                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2300                 if (in_dev && (dev->flags & IFF_UP)) {
2301                         struct in_ifaddr * ifa;
2302
2303                         int flag = scope;
2304
2305                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2306                                 int plen;
2307
2308                                 addr.s6_addr32[3] = ifa->ifa_local;
2309
2310                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2311                                         continue;
2312                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2313                                         if (idev->dev->flags&IFF_POINTOPOINT)
2314                                                 continue;
2315                                         flag |= IFA_HOST;
2316                                 }
2317                                 if (idev->dev->flags&IFF_POINTOPOINT)
2318                                         plen = 64;
2319                                 else
2320                                         plen = 96;
2321
2322                                 add_addr(idev, &addr, plen, flag);
2323                         }
2324                 }
2325         }
2326 }
2327 #endif
2328
2329 static void init_loopback(struct net_device *dev)
2330 {
2331         struct inet6_dev  *idev;
2332
2333         /* ::1 */
2334
2335         ASSERT_RTNL();
2336
2337         if ((idev = ipv6_find_idev(dev)) == NULL) {
2338                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2339                 return;
2340         }
2341
2342         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2343 }
2344
2345 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2346 {
2347         struct inet6_ifaddr * ifp;
2348         u32 addr_flags = IFA_F_PERMANENT;
2349
2350 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2351         if (idev->cnf.optimistic_dad &&
2352             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2353                 addr_flags |= IFA_F_OPTIMISTIC;
2354 #endif
2355
2356
2357         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2358         if (!IS_ERR(ifp)) {
2359                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2360                 addrconf_dad_start(ifp, 0);
2361                 in6_ifa_put(ifp);
2362         }
2363 }
2364
2365 static void addrconf_dev_config(struct net_device *dev)
2366 {
2367         struct in6_addr addr;
2368         struct inet6_dev    * idev;
2369
2370         ASSERT_RTNL();
2371
2372         if ((dev->type != ARPHRD_ETHER) &&
2373             (dev->type != ARPHRD_FDDI) &&
2374             (dev->type != ARPHRD_IEEE802_TR) &&
2375             (dev->type != ARPHRD_ARCNET) &&
2376             (dev->type != ARPHRD_INFINIBAND)) {
2377                 /* Alas, we support only Ethernet autoconfiguration. */
2378                 return;
2379         }
2380
2381         idev = addrconf_add_dev(dev);
2382         if (IS_ERR(idev))
2383                 return;
2384
2385         memset(&addr, 0, sizeof(struct in6_addr));
2386         addr.s6_addr32[0] = htonl(0xFE800000);
2387
2388         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2389                 addrconf_add_linklocal(idev, &addr);
2390 }
2391
2392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2393 static void addrconf_sit_config(struct net_device *dev)
2394 {
2395         struct inet6_dev *idev;
2396
2397         ASSERT_RTNL();
2398
2399         /*
2400          * Configure the tunnel with one of our IPv4
2401          * addresses... we should configure all of
2402          * our v4 addrs in the tunnel
2403          */
2404
2405         if ((idev = ipv6_find_idev(dev)) == NULL) {
2406                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2407                 return;
2408         }
2409
2410         if (dev->priv_flags & IFF_ISATAP) {
2411                 struct in6_addr addr;
2412
2413                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2414                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2415                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2416                         addrconf_add_linklocal(idev, &addr);
2417                 return;
2418         }
2419
2420         sit_add_v4_addrs(idev);
2421
2422         if (dev->flags&IFF_POINTOPOINT) {
2423                 addrconf_add_mroute(dev);
2424                 addrconf_add_lroute(dev);
2425         } else
2426                 sit_route_add(dev);
2427 }
2428 #endif
2429
2430 static inline int
2431 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2432 {
2433         struct in6_addr lladdr;
2434
2435         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2436                 addrconf_add_linklocal(idev, &lladdr);
2437                 return 0;
2438         }
2439         return -1;
2440 }
2441
2442 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2443 {
2444         struct net_device *link_dev;
2445         struct net *net = dev_net(idev->dev);
2446
2447         /* first try to inherit the link-local address from the link device */
2448         if (idev->dev->iflink &&
2449             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2450                 if (!ipv6_inherit_linklocal(idev, link_dev))
2451                         return;
2452         }
2453         /* then try to inherit it from any device */
2454         for_each_netdev(net, link_dev) {
2455                 if (!ipv6_inherit_linklocal(idev, link_dev))
2456                         return;
2457         }
2458         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2459 }
2460
2461 /*
2462  * Autoconfigure tunnel with a link-local address so routing protocols,
2463  * DHCPv6, MLD etc. can be run over the virtual link
2464  */
2465
2466 static void addrconf_ip6_tnl_config(struct net_device *dev)
2467 {
2468         struct inet6_dev *idev;
2469
2470         ASSERT_RTNL();
2471
2472         idev = addrconf_add_dev(dev);
2473         if (IS_ERR(idev)) {
2474                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2475                 return;
2476         }
2477         ip6_tnl_add_linklocal(idev);
2478 }
2479
2480 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2481                            void * data)
2482 {
2483         struct net_device *dev = (struct net_device *) data;
2484         struct inet6_dev *idev = __in6_dev_get(dev);
2485         int run_pending = 0;
2486         int err;
2487
2488         switch (event) {
2489         case NETDEV_REGISTER:
2490                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2491                         idev = ipv6_add_dev(dev);
2492                         if (!idev)
2493                                 return notifier_from_errno(-ENOMEM);
2494                 }
2495                 break;
2496
2497         case NETDEV_UP:
2498         case NETDEV_CHANGE:
2499                 if (dev->flags & IFF_SLAVE)
2500                         break;
2501
2502                 if (event == NETDEV_UP) {
2503                         if (!addrconf_qdisc_ok(dev)) {
2504                                 /* device is not ready yet. */
2505                                 printk(KERN_INFO
2506                                         "ADDRCONF(NETDEV_UP): %s: "
2507                                         "link is not ready\n",
2508                                         dev->name);
2509                                 break;
2510                         }
2511
2512                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2513                                 idev = ipv6_add_dev(dev);
2514
2515                         if (idev) {
2516                                 idev->if_flags |= IF_READY;
2517                                 run_pending = 1;
2518                         }
2519                 } else {
2520                         if (!addrconf_qdisc_ok(dev)) {
2521                                 /* device is still not ready. */
2522                                 break;
2523                         }
2524
2525                         if (idev) {
2526                                 if (idev->if_flags & IF_READY)
2527                                         /* device is already configured. */
2528                                         break;
2529                                 idev->if_flags |= IF_READY;
2530                         }
2531
2532                         printk(KERN_INFO
2533                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2534                                         "link becomes ready\n",
2535                                         dev->name);
2536
2537                         run_pending = 1;
2538                 }
2539
2540                 switch (dev->type) {
2541 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2542                 case ARPHRD_SIT:
2543                         addrconf_sit_config(dev);
2544                         break;
2545 #endif
2546                 case ARPHRD_TUNNEL6:
2547                         addrconf_ip6_tnl_config(dev);
2548                         break;
2549                 case ARPHRD_LOOPBACK:
2550                         init_loopback(dev);
2551                         break;
2552
2553                 default:
2554                         addrconf_dev_config(dev);
2555                         break;
2556                 }
2557
2558                 if (idev) {
2559                         if (run_pending)
2560                                 addrconf_dad_run(idev);
2561
2562                         /*
2563                          * If the MTU changed during the interface down,
2564                          * when the interface up, the changed MTU must be
2565                          * reflected in the idev as well as routers.
2566                          */
2567                         if (idev->cnf.mtu6 != dev->mtu &&
2568                             dev->mtu >= IPV6_MIN_MTU) {
2569                                 rt6_mtu_change(dev, dev->mtu);
2570                                 idev->cnf.mtu6 = dev->mtu;
2571                         }
2572                         idev->tstamp = jiffies;
2573                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2574
2575                         /*
2576                          * If the changed mtu during down is lower than
2577                          * IPV6_MIN_MTU stop IPv6 on this interface.
2578                          */
2579                         if (dev->mtu < IPV6_MIN_MTU)
2580                                 addrconf_ifdown(dev, 1);
2581                 }
2582                 break;
2583
2584         case NETDEV_CHANGEMTU:
2585                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2586                         rt6_mtu_change(dev, dev->mtu);
2587                         idev->cnf.mtu6 = dev->mtu;
2588                         break;
2589                 }
2590
2591                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2592                         idev = ipv6_add_dev(dev);
2593                         if (idev)
2594                                 break;
2595                 }
2596
2597                 /*
2598                  * MTU falled under IPV6_MIN_MTU.
2599                  * Stop IPv6 on this interface.
2600                  */
2601
2602         case NETDEV_DOWN:
2603         case NETDEV_UNREGISTER:
2604                 /*
2605                  *      Remove all addresses from this interface.
2606                  */
2607                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2608                 break;
2609
2610         case NETDEV_CHANGENAME:
2611                 if (idev) {
2612                         snmp6_unregister_dev(idev);
2613                         addrconf_sysctl_unregister(idev);
2614                         addrconf_sysctl_register(idev);
2615                         err = snmp6_register_dev(idev);
2616                         if (err)
2617                                 return notifier_from_errno(err);
2618                 }
2619                 break;
2620
2621         case NETDEV_PRE_TYPE_CHANGE:
2622         case NETDEV_POST_TYPE_CHANGE:
2623                 addrconf_type_change(dev, event);
2624                 break;
2625         }
2626
2627         return NOTIFY_OK;
2628 }
2629
2630 /*
2631  *      addrconf module should be notified of a device going up
2632  */
2633 static struct notifier_block ipv6_dev_notf = {
2634         .notifier_call = addrconf_notify,
2635 };
2636
2637 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2638 {
2639         struct inet6_dev *idev;
2640         ASSERT_RTNL();
2641
2642         idev = __in6_dev_get(dev);
2643
2644         if (event == NETDEV_POST_TYPE_CHANGE)
2645                 ipv6_mc_remap(idev);
2646         else if (event == NETDEV_PRE_TYPE_CHANGE)
2647                 ipv6_mc_unmap(idev);
2648 }
2649
2650 static int addrconf_ifdown(struct net_device *dev, int how)
2651 {
2652         struct net *net = dev_net(dev);
2653         struct inet6_dev *idev;
2654         struct inet6_ifaddr *ifa;
2655         int state, i;
2656
2657         ASSERT_RTNL();
2658
2659         rt6_ifdown(net, dev);
2660         neigh_ifdown(&nd_tbl, dev);
2661
2662         idev = __in6_dev_get(dev);
2663         if (idev == NULL)
2664                 return -ENODEV;
2665
2666         /*
2667          * Step 1: remove reference to ipv6 device from parent device.
2668          *         Do not dev_put!
2669          */
2670         if (how) {
2671                 idev->dead = 1;
2672
2673                 /* protected by rtnl_lock */
2674                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2675
2676                 /* Step 1.5: remove snmp6 entry */
2677                 snmp6_unregister_dev(idev);
2678
2679         }
2680
2681         /* Step 2: clear hash table */
2682         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2683                 struct hlist_head *h = &inet6_addr_lst[i];
2684                 struct hlist_node *n;
2685
2686                 spin_lock_bh(&addrconf_hash_lock);
2687         restart:
2688                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2689                         if (ifa->idev == idev) {
2690                                 hlist_del_init_rcu(&ifa->addr_lst);
2691                                 addrconf_del_timer(ifa);
2692                                 goto restart;
2693                         }
2694                 }
2695                 spin_unlock_bh(&addrconf_hash_lock);
2696         }
2697
2698         write_lock_bh(&idev->lock);
2699
2700         /* Step 2: clear flags for stateless addrconf */
2701         if (!how)
2702                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2703
2704 #ifdef CONFIG_IPV6_PRIVACY
2705         if (how && del_timer(&idev->regen_timer))
2706                 in6_dev_put(idev);
2707
2708         /* Step 3: clear tempaddr list */
2709         while (!list_empty(&idev->tempaddr_list)) {
2710                 ifa = list_first_entry(&idev->tempaddr_list,
2711                                        struct inet6_ifaddr, tmp_list);
2712                 list_del(&ifa->tmp_list);
2713                 write_unlock_bh(&idev->lock);
2714                 spin_lock_bh(&ifa->lock);
2715
2716                 if (ifa->ifpub) {
2717                         in6_ifa_put(ifa->ifpub);
2718                         ifa->ifpub = NULL;
2719                 }
2720                 spin_unlock_bh(&ifa->lock);
2721                 in6_ifa_put(ifa);
2722                 write_lock_bh(&idev->lock);
2723         }
2724 #endif
2725
2726         while (!list_empty(&idev->addr_list)) {
2727                 ifa = list_first_entry(&idev->addr_list,
2728                                        struct inet6_ifaddr, if_list);
2729                 addrconf_del_timer(ifa);
2730
2731                 list_del(&ifa->if_list);
2732
2733                 write_unlock_bh(&idev->lock);
2734
2735                 spin_lock_bh(&ifa->state_lock);
2736                 state = ifa->state;
2737                 ifa->state = INET6_IFADDR_STATE_DEAD;
2738                 spin_unlock_bh(&ifa->state_lock);
2739
2740                 if (state != INET6_IFADDR_STATE_DEAD) {
2741                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2742                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2743                 }
2744                 in6_ifa_put(ifa);
2745
2746                 write_lock_bh(&idev->lock);
2747         }
2748
2749         write_unlock_bh(&idev->lock);
2750
2751         /* Step 5: Discard multicast list */
2752         if (how)
2753                 ipv6_mc_destroy_dev(idev);
2754         else
2755                 ipv6_mc_down(idev);
2756
2757         idev->tstamp = jiffies;
2758
2759         /* Last: Shot the device (if unregistered) */
2760         if (how) {
2761                 addrconf_sysctl_unregister(idev);
2762                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2763                 neigh_ifdown(&nd_tbl, dev);
2764                 in6_dev_put(idev);
2765         }
2766         return 0;
2767 }
2768
2769 static void addrconf_rs_timer(unsigned long data)
2770 {
2771         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2772         struct inet6_dev *idev = ifp->idev;
2773
2774         read_lock(&idev->lock);
2775         if (idev->dead || !(idev->if_flags & IF_READY))
2776                 goto out;
2777
2778         if (idev->cnf.forwarding)
2779                 goto out;
2780
2781         /* Announcement received after solicitation was sent */
2782         if (idev->if_flags & IF_RA_RCVD)
2783                 goto out;
2784
2785         spin_lock(&ifp->lock);
2786         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2787                 /* The wait after the last probe can be shorter */
2788                 addrconf_mod_timer(ifp, AC_RS,
2789                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2790                                    idev->cnf.rtr_solicit_delay :
2791                                    idev->cnf.rtr_solicit_interval);
2792                 spin_unlock(&ifp->lock);
2793
2794                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2795         } else {
2796                 spin_unlock(&ifp->lock);
2797                 /*
2798                  * Note: we do not support deprecated "all on-link"
2799                  * assumption any longer.
2800                  */
2801                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2802                        idev->dev->name);
2803         }
2804
2805 out:
2806         read_unlock(&idev->lock);
2807         in6_ifa_put(ifp);
2808 }
2809
2810 /*
2811  *      Duplicate Address Detection
2812  */
2813 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2814 {
2815         unsigned long rand_num;
2816         struct inet6_dev *idev = ifp->idev;
2817
2818         if (ifp->flags & IFA_F_OPTIMISTIC)
2819                 rand_num = 0;
2820         else
2821                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2822
2823         ifp->probes = idev->cnf.dad_transmits;
2824         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2825 }
2826
2827 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2828 {
2829         struct inet6_dev *idev = ifp->idev;
2830         struct net_device *dev = idev->dev;
2831
2832         addrconf_join_solict(dev, &ifp->addr);
2833
2834         net_srandom(ifp->addr.s6_addr32[3]);
2835
2836         read_lock_bh(&idev->lock);
2837         spin_lock(&ifp->lock);
2838         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2839                 goto out;
2840
2841         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2842             idev->cnf.accept_dad < 1 ||
2843             !(ifp->flags&IFA_F_TENTATIVE) ||
2844             ifp->flags & IFA_F_NODAD) {
2845                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2846                 spin_unlock(&ifp->lock);
2847                 read_unlock_bh(&idev->lock);
2848
2849                 addrconf_dad_completed(ifp);
2850                 return;
2851         }
2852
2853         if (!(idev->if_flags & IF_READY)) {
2854                 spin_unlock(&ifp->lock);
2855                 read_unlock_bh(&idev->lock);
2856                 /*
2857                  * If the device is not ready:
2858                  * - keep it tentative if it is a permanent address.
2859                  * - otherwise, kill it.
2860                  */
2861                 in6_ifa_hold(ifp);
2862                 addrconf_dad_stop(ifp, 0);
2863                 return;
2864         }
2865
2866         /*
2867          * Optimistic nodes can start receiving
2868          * Frames right away
2869          */
2870         if (ifp->flags & IFA_F_OPTIMISTIC)
2871                 ip6_ins_rt(ifp->rt);
2872
2873         addrconf_dad_kick(ifp);
2874 out:
2875         spin_unlock(&ifp->lock);
2876         read_unlock_bh(&idev->lock);
2877 }
2878
2879 static void addrconf_dad_timer(unsigned long data)
2880 {
2881         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2882         struct inet6_dev *idev = ifp->idev;
2883         struct in6_addr mcaddr;
2884
2885         if (!ifp->probes && addrconf_dad_end(ifp))
2886                 goto out;
2887
2888         read_lock(&idev->lock);
2889         if (idev->dead || !(idev->if_flags & IF_READY)) {
2890                 read_unlock(&idev->lock);
2891                 goto out;
2892         }
2893
2894         spin_lock(&ifp->lock);
2895         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2896                 spin_unlock(&ifp->lock);
2897                 read_unlock(&idev->lock);
2898                 goto out;
2899         }
2900
2901         if (ifp->probes == 0) {
2902                 /*
2903                  * DAD was successful
2904                  */
2905
2906                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2907                 spin_unlock(&ifp->lock);
2908                 read_unlock(&idev->lock);
2909
2910                 addrconf_dad_completed(ifp);
2911
2912                 goto out;
2913         }
2914
2915         ifp->probes--;
2916         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2917         spin_unlock(&ifp->lock);
2918         read_unlock(&idev->lock);
2919
2920         /* send a neighbour solicitation for our addr */
2921         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2922         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2923 out:
2924         in6_ifa_put(ifp);
2925 }
2926
2927 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2928 {
2929         struct net_device *dev = ifp->idev->dev;
2930
2931         /*
2932          *      Configure the address for reception. Now it is valid.
2933          */
2934
2935         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2936
2937         /* If added prefix is link local and forwarding is off,
2938            start sending router solicitations.
2939          */
2940
2941         if ((ifp->idev->cnf.forwarding == 0 ||
2942              ifp->idev->cnf.forwarding == 2) &&
2943             ifp->idev->cnf.rtr_solicits > 0 &&
2944             (dev->flags&IFF_LOOPBACK) == 0 &&
2945             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2946                 /*
2947                  *      If a host as already performed a random delay
2948                  *      [...] as part of DAD [...] there is no need
2949                  *      to delay again before sending the first RS
2950                  */
2951                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2952
2953                 spin_lock_bh(&ifp->lock);
2954                 ifp->probes = 1;
2955                 ifp->idev->if_flags |= IF_RS_SENT;
2956                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2957                 spin_unlock_bh(&ifp->lock);
2958         }
2959 }
2960
2961 static void addrconf_dad_run(struct inet6_dev *idev)
2962 {
2963         struct inet6_ifaddr *ifp;
2964
2965         read_lock_bh(&idev->lock);
2966         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2967                 spin_lock(&ifp->lock);
2968                 if (ifp->flags & IFA_F_TENTATIVE &&
2969                     ifp->state == INET6_IFADDR_STATE_DAD)
2970                         addrconf_dad_kick(ifp);
2971                 spin_unlock(&ifp->lock);
2972         }
2973         read_unlock_bh(&idev->lock);
2974 }
2975
2976 #ifdef CONFIG_PROC_FS
2977 struct if6_iter_state {
2978         struct seq_net_private p;
2979         int bucket;
2980 };
2981
2982 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2983 {
2984         struct inet6_ifaddr *ifa = NULL;
2985         struct if6_iter_state *state = seq->private;
2986         struct net *net = seq_file_net(seq);
2987
2988         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2989                 struct hlist_node *n;
2990                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
2991                                          addr_lst)
2992                         if (net_eq(dev_net(ifa->idev->dev), net))
2993                                 return ifa;
2994         }
2995         return NULL;
2996 }
2997
2998 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2999                                          struct inet6_ifaddr *ifa)
3000 {
3001         struct if6_iter_state *state = seq->private;
3002         struct net *net = seq_file_net(seq);
3003         struct hlist_node *n = &ifa->addr_lst;
3004
3005         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3006                 if (net_eq(dev_net(ifa->idev->dev), net))
3007                         return ifa;
3008
3009         while (++state->bucket < IN6_ADDR_HSIZE) {
3010                 hlist_for_each_entry_rcu_bh(ifa, n,
3011                                      &inet6_addr_lst[state->bucket], addr_lst) {
3012                         if (net_eq(dev_net(ifa->idev->dev), net))
3013                                 return ifa;
3014                 }
3015         }
3016
3017         return NULL;
3018 }
3019
3020 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3021 {
3022         struct inet6_ifaddr *ifa = if6_get_first(seq);
3023
3024         if (ifa)
3025                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3026                         --pos;
3027         return pos ? NULL : ifa;
3028 }
3029
3030 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3031         __acquires(rcu_bh)
3032 {
3033         rcu_read_lock_bh();
3034         return if6_get_idx(seq, *pos);
3035 }
3036
3037 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3038 {
3039         struct inet6_ifaddr *ifa;
3040
3041         ifa = if6_get_next(seq, v);
3042         ++*pos;
3043         return ifa;
3044 }
3045
3046 static void if6_seq_stop(struct seq_file *seq, void *v)
3047         __releases(rcu_bh)
3048 {
3049         rcu_read_unlock_bh();
3050 }
3051
3052 static int if6_seq_show(struct seq_file *seq, void *v)
3053 {
3054         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3055         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3056                    &ifp->addr,
3057                    ifp->idev->dev->ifindex,
3058                    ifp->prefix_len,
3059                    ifp->scope,
3060                    ifp->flags,
3061                    ifp->idev->dev->name);
3062         return 0;
3063 }
3064
3065 static const struct seq_operations if6_seq_ops = {
3066         .start  = if6_seq_start,
3067         .next   = if6_seq_next,
3068         .show   = if6_seq_show,
3069         .stop   = if6_seq_stop,
3070 };
3071
3072 static int if6_seq_open(struct inode *inode, struct file *file)
3073 {
3074         return seq_open_net(inode, file, &if6_seq_ops,
3075                             sizeof(struct if6_iter_state));
3076 }
3077
3078 static const struct file_operations if6_fops = {
3079         .owner          = THIS_MODULE,
3080         .open           = if6_seq_open,
3081         .read           = seq_read,
3082         .llseek         = seq_lseek,
3083         .release        = seq_release_net,
3084 };
3085
3086 static int __net_init if6_proc_net_init(struct net *net)
3087 {
3088         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3089                 return -ENOMEM;
3090         return 0;
3091 }
3092
3093 static void __net_exit if6_proc_net_exit(struct net *net)
3094 {
3095        proc_net_remove(net, "if_inet6");
3096 }
3097
3098 static struct pernet_operations if6_proc_net_ops = {
3099        .init = if6_proc_net_init,
3100        .exit = if6_proc_net_exit,
3101 };
3102
3103 int __init if6_proc_init(void)
3104 {
3105         return register_pernet_subsys(&if6_proc_net_ops);
3106 }
3107
3108 void if6_proc_exit(void)
3109 {
3110         unregister_pernet_subsys(&if6_proc_net_ops);
3111 }
3112 #endif  /* CONFIG_PROC_FS */
3113
3114 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3115 /* Check if address is a home address configured on any interface. */
3116 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3117 {
3118         int ret = 0;
3119         struct inet6_ifaddr *ifp = NULL;
3120         struct hlist_node *n;
3121         unsigned int hash = ipv6_addr_hash(addr);
3122
3123         rcu_read_lock_bh();
3124         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3125                 if (!net_eq(dev_net(ifp->idev->dev), net))
3126                         continue;
3127                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3128                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3129                         ret = 1;
3130                         break;
3131                 }
3132         }
3133         rcu_read_unlock_bh();
3134         return ret;
3135 }
3136 #endif
3137
3138 /*
3139  *      Periodic address status verification
3140  */
3141
3142 static void addrconf_verify(unsigned long foo)
3143 {
3144         unsigned long now, next, next_sec, next_sched;
3145         struct inet6_ifaddr *ifp;
3146         struct hlist_node *node;
3147         int i;
3148
3149         rcu_read_lock_bh();
3150         spin_lock(&addrconf_verify_lock);
3151         now = jiffies;
3152         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3153
3154         del_timer(&addr_chk_timer);
3155
3156         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3157 restart:
3158                 hlist_for_each_entry_rcu_bh(ifp, node,
3159                                          &inet6_addr_lst[i], addr_lst) {
3160                         unsigned long age;
3161
3162                         if (ifp->flags & IFA_F_PERMANENT)
3163                                 continue;
3164
3165                         spin_lock(&ifp->lock);
3166                         /* We try to batch several events at once. */
3167                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3168
3169                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3170                             age >= ifp->valid_lft) {
3171                                 spin_unlock(&ifp->lock);
3172                                 in6_ifa_hold(ifp);
3173                                 ipv6_del_addr(ifp);
3174                                 goto restart;
3175                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3176                                 spin_unlock(&ifp->lock);
3177                                 continue;
3178                         } else if (age >= ifp->prefered_lft) {
3179                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3180                                 int deprecate = 0;
3181
3182                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3183                                         deprecate = 1;
3184                                         ifp->flags |= IFA_F_DEPRECATED;
3185                                 }
3186
3187                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3188                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3189
3190                                 spin_unlock(&ifp->lock);
3191
3192                                 if (deprecate) {
3193                                         in6_ifa_hold(ifp);
3194
3195                                         ipv6_ifa_notify(0, ifp);
3196                                         in6_ifa_put(ifp);
3197                                         goto restart;
3198                                 }
3199 #ifdef CONFIG_IPV6_PRIVACY
3200                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3201                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3202                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3203                                         ifp->idev->cnf.dad_transmits *
3204                                         ifp->idev->nd_parms->retrans_time / HZ;
3205
3206                                 if (age >= ifp->prefered_lft - regen_advance) {
3207                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3208                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3209                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3210                                         if (!ifp->regen_count && ifpub) {
3211                                                 ifp->regen_count++;
3212                                                 in6_ifa_hold(ifp);
3213                                                 in6_ifa_hold(ifpub);
3214                                                 spin_unlock(&ifp->lock);
3215
3216                                                 spin_lock(&ifpub->lock);
3217                                                 ifpub->regen_count = 0;
3218                                                 spin_unlock(&ifpub->lock);
3219                                                 ipv6_create_tempaddr(ifpub, ifp);
3220                                                 in6_ifa_put(ifpub);
3221                                                 in6_ifa_put(ifp);
3222                                                 goto restart;
3223                                         }
3224                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3225                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3226                                 spin_unlock(&ifp->lock);
3227 #endif
3228                         } else {
3229                                 /* ifp->prefered_lft <= ifp->valid_lft */
3230                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3231                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3232                                 spin_unlock(&ifp->lock);
3233                         }
3234                 }
3235         }
3236
3237         next_sec = round_jiffies_up(next);
3238         next_sched = next;
3239
3240         /* If rounded timeout is accurate enough, accept it. */
3241         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3242                 next_sched = next_sec;
3243
3244         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3245         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3246                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3247
3248         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3249               now, next, next_sec, next_sched));
3250
3251         addr_chk_timer.expires = next_sched;
3252         add_timer(&addr_chk_timer);
3253         spin_unlock(&addrconf_verify_lock);
3254         rcu_read_unlock_bh();
3255 }
3256
3257 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3258 {
3259         struct in6_addr *pfx = NULL;
3260
3261         if (addr)
3262                 pfx = nla_data(addr);
3263
3264         if (local) {
3265                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3266                         pfx = NULL;
3267                 else
3268                         pfx = nla_data(local);
3269         }
3270
3271         return pfx;
3272 }
3273
3274 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3275         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3276         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3277         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3278 };
3279
3280 static int
3281 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3282 {
3283         struct net *net = sock_net(skb->sk);
3284         struct ifaddrmsg *ifm;
3285         struct nlattr *tb[IFA_MAX+1];
3286         struct in6_addr *pfx;
3287         int err;
3288
3289         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3290         if (err < 0)
3291                 return err;
3292
3293         ifm = nlmsg_data(nlh);
3294         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3295         if (pfx == NULL)
3296                 return -EINVAL;
3297
3298         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3299 }
3300
3301 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3302                              u32 prefered_lft, u32 valid_lft)
3303 {
3304         u32 flags;
3305         clock_t expires;
3306         unsigned long timeout;
3307
3308         if (!valid_lft || (prefered_lft > valid_lft))
3309                 return -EINVAL;
3310
3311         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3312         if (addrconf_finite_timeout(timeout)) {
3313                 expires = jiffies_to_clock_t(timeout * HZ);
3314                 valid_lft = timeout;
3315                 flags = RTF_EXPIRES;
3316         } else {
3317                 expires = 0;
3318                 flags = 0;
3319                 ifa_flags |= IFA_F_PERMANENT;
3320         }
3321
3322         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3323         if (addrconf_finite_timeout(timeout)) {
3324                 if (timeout == 0)
3325                         ifa_flags |= IFA_F_DEPRECATED;
3326                 prefered_lft = timeout;
3327         }
3328
3329         spin_lock_bh(&ifp->lock);
3330         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3331         ifp->tstamp = jiffies;
3332         ifp->valid_lft = valid_lft;
3333         ifp->prefered_lft = prefered_lft;
3334
3335         spin_unlock_bh(&ifp->lock);
3336         if (!(ifp->flags&IFA_F_TENTATIVE))
3337                 ipv6_ifa_notify(0, ifp);
3338
3339         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3340                               expires, flags);
3341         addrconf_verify(0);
3342
3343         return 0;
3344 }
3345
3346 static int
3347 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3348 {
3349         struct net *net = sock_net(skb->sk);
3350         struct ifaddrmsg *ifm;
3351         struct nlattr *tb[IFA_MAX+1];
3352         struct in6_addr *pfx;
3353         struct inet6_ifaddr *ifa;
3354         struct net_device *dev;
3355         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3356         u8 ifa_flags;
3357         int err;
3358
3359         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3360         if (err < 0)
3361                 return err;
3362
3363         ifm = nlmsg_data(nlh);
3364         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3365         if (pfx == NULL)
3366                 return -EINVAL;
3367
3368         if (tb[IFA_CACHEINFO]) {
3369                 struct ifa_cacheinfo *ci;
3370
3371                 ci = nla_data(tb[IFA_CACHEINFO]);
3372                 valid_lft = ci->ifa_valid;
3373                 preferred_lft = ci->ifa_prefered;
3374         } else {
3375                 preferred_lft = INFINITY_LIFE_TIME;
3376                 valid_lft = INFINITY_LIFE_TIME;
3377         }
3378
3379         dev =  __dev_get_by_index(net, ifm->ifa_index);
3380         if (dev == NULL)
3381                 return -ENODEV;
3382
3383         /* We ignore other flags so far. */
3384         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3385
3386         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3387         if (ifa == NULL) {
3388                 /*
3389                  * It would be best to check for !NLM_F_CREATE here but
3390                  * userspace alreay relies on not having to provide this.
3391                  */
3392                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3393                                       ifm->ifa_prefixlen, ifa_flags,
3394                                       preferred_lft, valid_lft);
3395         }
3396
3397         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3398             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3399                 err = -EEXIST;
3400         else
3401                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3402
3403         in6_ifa_put(ifa);
3404
3405         return err;
3406 }
3407
3408 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3409                           u8 scope, int ifindex)
3410 {
3411         struct ifaddrmsg *ifm;
3412
3413         ifm = nlmsg_data(nlh);
3414         ifm->ifa_family = AF_INET6;
3415         ifm->ifa_prefixlen = prefixlen;
3416         ifm->ifa_flags = flags;
3417         ifm->ifa_scope = scope;
3418         ifm->ifa_index = ifindex;
3419 }
3420
3421 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3422                          unsigned long tstamp, u32 preferred, u32 valid)
3423 {
3424         struct ifa_cacheinfo ci;
3425
3426         ci.cstamp = cstamp_delta(cstamp);
3427         ci.tstamp = cstamp_delta(tstamp);
3428         ci.ifa_prefered = preferred;
3429         ci.ifa_valid = valid;
3430
3431         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3432 }
3433
3434 static inline int rt_scope(int ifa_scope)
3435 {
3436         if (ifa_scope & IFA_HOST)
3437                 return RT_SCOPE_HOST;
3438         else if (ifa_scope & IFA_LINK)
3439                 return RT_SCOPE_LINK;
3440         else if (ifa_scope & IFA_SITE)
3441                 return RT_SCOPE_SITE;
3442         else
3443                 return RT_SCOPE_UNIVERSE;
3444 }
3445
3446 static inline int inet6_ifaddr_msgsize(void)
3447 {
3448         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3449                + nla_total_size(16) /* IFA_ADDRESS */
3450                + nla_total_size(sizeof(struct ifa_cacheinfo));
3451 }
3452
3453 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3454                              u32 pid, u32 seq, int event, unsigned int flags)
3455 {
3456         struct nlmsghdr  *nlh;
3457         u32 preferred, valid;
3458
3459         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3460         if (nlh == NULL)
3461                 return -EMSGSIZE;
3462
3463         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3464                       ifa->idev->dev->ifindex);
3465
3466         if (!(ifa->flags&IFA_F_PERMANENT)) {
3467                 preferred = ifa->prefered_lft;
3468                 valid = ifa->valid_lft;
3469                 if (preferred != INFINITY_LIFE_TIME) {
3470                         long tval = (jiffies - ifa->tstamp)/HZ;
3471                         if (preferred > tval)
3472                                 preferred -= tval;
3473                         else
3474                                 preferred = 0;
3475                         if (valid != INFINITY_LIFE_TIME) {
3476                                 if (valid > tval)
3477                                         valid -= tval;
3478                                 else
3479                                         valid = 0;
3480                         }
3481                 }
3482         } else {
3483                 preferred = INFINITY_LIFE_TIME;
3484                 valid = INFINITY_LIFE_TIME;
3485         }
3486
3487         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3488             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3489                 nlmsg_cancel(skb, nlh);
3490                 return -EMSGSIZE;
3491         }
3492
3493         return nlmsg_end(skb, nlh);
3494 }
3495
3496 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3497                                 u32 pid, u32 seq, int event, u16 flags)
3498 {
3499         struct nlmsghdr  *nlh;
3500         u8 scope = RT_SCOPE_UNIVERSE;
3501         int ifindex = ifmca->idev->dev->ifindex;
3502
3503         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3504                 scope = RT_SCOPE_SITE;
3505
3506         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3507         if (nlh == NULL)
3508                 return -EMSGSIZE;
3509
3510         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3511         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3512             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3513                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3514                 nlmsg_cancel(skb, nlh);
3515                 return -EMSGSIZE;
3516         }
3517
3518         return nlmsg_end(skb, nlh);
3519 }
3520
3521 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3522                                 u32 pid, u32 seq, int event, unsigned int flags)
3523 {
3524         struct nlmsghdr  *nlh;
3525         u8 scope = RT_SCOPE_UNIVERSE;
3526         int ifindex = ifaca->aca_idev->dev->ifindex;
3527
3528         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3529                 scope = RT_SCOPE_SITE;
3530
3531         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3532         if (nlh == NULL)
3533                 return -EMSGSIZE;
3534
3535         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3536         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3537             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3538                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3539                 nlmsg_cancel(skb, nlh);
3540                 return -EMSGSIZE;
3541         }
3542
3543         return nlmsg_end(skb, nlh);
3544 }
3545
3546 enum addr_type_t {
3547         UNICAST_ADDR,
3548         MULTICAST_ADDR,
3549         ANYCAST_ADDR,
3550 };
3551
3552 /* called with rcu_read_lock() */
3553 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3554                           struct netlink_callback *cb, enum addr_type_t type,
3555                           int s_ip_idx, int *p_ip_idx)
3556 {
3557         struct ifmcaddr6 *ifmca;
3558         struct ifacaddr6 *ifaca;
3559         int err = 1;
3560         int ip_idx = *p_ip_idx;
3561
3562         read_lock_bh(&idev->lock);
3563         switch (type) {
3564         case UNICAST_ADDR: {
3565                 struct inet6_ifaddr *ifa;
3566
3567                 /* unicast address incl. temp addr */
3568                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3569                         if (++ip_idx < s_ip_idx)
3570                                 continue;
3571                         err = inet6_fill_ifaddr(skb, ifa,
3572                                                 NETLINK_CB(cb->skb).pid,
3573                                                 cb->nlh->nlmsg_seq,
3574                                                 RTM_NEWADDR,
3575                                                 NLM_F_MULTI);
3576                         if (err <= 0)
3577                                 break;
3578                 }
3579                 break;
3580         }
3581         case MULTICAST_ADDR:
3582                 /* multicast address */
3583                 for (ifmca = idev->mc_list; ifmca;
3584                      ifmca = ifmca->next, ip_idx++) {
3585                         if (ip_idx < s_ip_idx)
3586                                 continue;
3587                         err = inet6_fill_ifmcaddr(skb, ifmca,
3588                                                   NETLINK_CB(cb->skb).pid,
3589                                                   cb->nlh->nlmsg_seq,
3590                                                   RTM_GETMULTICAST,
3591                                                   NLM_F_MULTI);
3592                         if (err <= 0)
3593                                 break;
3594                 }
3595                 break;
3596         case ANYCAST_ADDR:
3597                 /* anycast address */
3598                 for (ifaca = idev->ac_list; ifaca;
3599                      ifaca = ifaca->aca_next, ip_idx++) {
3600                         if (ip_idx < s_ip_idx)
3601                                 continue;
3602                         err = inet6_fill_ifacaddr(skb, ifaca,
3603                                                   NETLINK_CB(cb->skb).pid,
3604                                                   cb->nlh->nlmsg_seq,
3605                                                   RTM_GETANYCAST,
3606                                                   NLM_F_MULTI);
3607                         if (err <= 0)
3608                                 break;
3609                 }
3610                 break;
3611         default:
3612                 break;
3613         }
3614         read_unlock_bh(&idev->lock);
3615         *p_ip_idx = ip_idx;
3616         return err;
3617 }
3618
3619 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3620                            enum addr_type_t type)
3621 {
3622         struct net *net = sock_net(skb->sk);
3623         int h, s_h;
3624         int idx, ip_idx;
3625         int s_idx, s_ip_idx;
3626         struct net_device *dev;
3627         struct inet6_dev *idev;
3628         struct hlist_head *head;
3629         struct hlist_node *node;
3630
3631         s_h = cb->args[0];
3632         s_idx = idx = cb->args[1];
3633         s_ip_idx = ip_idx = cb->args[2];
3634
3635         rcu_read_lock();
3636         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3637                 idx = 0;
3638                 head = &net->dev_index_head[h];
3639                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3640                         if (idx < s_idx)
3641                                 goto cont;
3642                         if (h > s_h || idx > s_idx)
3643                                 s_ip_idx = 0;
3644                         ip_idx = 0;
3645                         idev = __in6_dev_get(dev);
3646                         if (!idev)
3647                                 goto cont;
3648
3649                         if (in6_dump_addrs(idev, skb, cb, type,
3650                                            s_ip_idx, &ip_idx) <= 0)
3651                                 goto done;
3652 cont:
3653                         idx++;
3654                 }
3655         }
3656 done:
3657         rcu_read_unlock();
3658         cb->args[0] = h;
3659         cb->args[1] = idx;
3660         cb->args[2] = ip_idx;
3661
3662         return skb->len;
3663 }
3664
3665 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3666 {
3667         enum addr_type_t type = UNICAST_ADDR;
3668
3669         return inet6_dump_addr(skb, cb, type);
3670 }
3671
3672 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3673 {
3674         enum addr_type_t type = MULTICAST_ADDR;
3675
3676         return inet6_dump_addr(skb, cb, type);
3677 }
3678
3679
3680 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3681 {
3682         enum addr_type_t type = ANYCAST_ADDR;
3683
3684         return inet6_dump_addr(skb, cb, type);
3685 }
3686
3687 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3688                              void *arg)
3689 {
3690         struct net *net = sock_net(in_skb->sk);
3691         struct ifaddrmsg *ifm;
3692         struct nlattr *tb[IFA_MAX+1];
3693         struct in6_addr *addr = NULL;
3694         struct net_device *dev = NULL;
3695         struct inet6_ifaddr *ifa;
3696         struct sk_buff *skb;
3697         int err;
3698
3699         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3700         if (err < 0)
3701                 goto errout;
3702
3703         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3704         if (addr == NULL) {
3705                 err = -EINVAL;
3706                 goto errout;
3707         }
3708
3709         ifm = nlmsg_data(nlh);
3710         if (ifm->ifa_index)
3711                 dev = __dev_get_by_index(net, ifm->ifa_index);
3712
3713         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3714         if (!ifa) {
3715                 err = -EADDRNOTAVAIL;
3716                 goto errout;
3717         }
3718
3719         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3720         if (!skb) {
3721                 err = -ENOBUFS;
3722                 goto errout_ifa;
3723         }
3724
3725         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3726                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3727         if (err < 0) {
3728                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3729                 WARN_ON(err == -EMSGSIZE);
3730                 kfree_skb(skb);
3731                 goto errout_ifa;
3732         }
3733         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3734 errout_ifa:
3735         in6_ifa_put(ifa);
3736 errout:
3737         return err;
3738 }
3739
3740 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3741 {
3742         struct sk_buff *skb;
3743         struct net *net = dev_net(ifa->idev->dev);
3744         int err = -ENOBUFS;
3745
3746         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3747         if (skb == NULL)
3748                 goto errout;
3749
3750         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3751         if (err < 0) {
3752                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3753                 WARN_ON(err == -EMSGSIZE);
3754                 kfree_skb(skb);
3755                 goto errout;
3756         }
3757         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3758         return;
3759 errout:
3760         if (err < 0)
3761                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3762 }
3763
3764 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3765                                 __s32 *array, int bytes)
3766 {
3767         BUG_ON(bytes < (DEVCONF_MAX * 4));
3768
3769         memset(array, 0, bytes);
3770         array[DEVCONF_FORWARDING] = cnf->forwarding;
3771         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3772         array[DEVCONF_MTU6] = cnf->mtu6;
3773         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3774         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3775         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3776         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3777         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3778         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3779                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3780         array[DEVCONF_RTR_SOLICIT_DELAY] =
3781                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3782         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3783 #ifdef CONFIG_IPV6_PRIVACY
3784         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3785         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3786         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3787         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3788         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3789 #endif
3790         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3791         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3792         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3793 #ifdef CONFIG_IPV6_ROUTER_PREF
3794         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3795         array[DEVCONF_RTR_PROBE_INTERVAL] =
3796                 jiffies_to_msecs(cnf->rtr_probe_interval);
3797 #ifdef CONFIG_IPV6_ROUTE_INFO
3798         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3799 #endif
3800 #endif
3801         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3802         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3803 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3804         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3805 #endif
3806 #ifdef CONFIG_IPV6_MROUTE
3807         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3808 #endif
3809         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3810         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3811         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3812 }
3813
3814 static inline size_t inet6_ifla6_size(void)
3815 {
3816         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3817              + nla_total_size(sizeof(struct ifla_cacheinfo))
3818              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3819              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3820              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3821 }
3822
3823 static inline size_t inet6_if_nlmsg_size(void)
3824 {
3825         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3826                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3827                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3828                + nla_total_size(4) /* IFLA_MTU */
3829                + nla_total_size(4) /* IFLA_LINK */
3830                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3831 }
3832
3833 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3834                                       int items, int bytes)
3835 {
3836         int i;
3837         int pad = bytes - sizeof(u64) * items;
3838         BUG_ON(pad < 0);
3839
3840         /* Use put_unaligned() because stats may not be aligned for u64. */
3841         put_unaligned(items, &stats[0]);
3842         for (i = 1; i < items; i++)
3843                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3844
3845         memset(&stats[items], 0, pad);
3846 }
3847
3848 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3849                                       int items, int bytes, size_t syncpoff)
3850 {
3851         int i;
3852         int pad = bytes - sizeof(u64) * items;
3853         BUG_ON(pad < 0);
3854
3855         /* Use put_unaligned() because stats may not be aligned for u64. */
3856         put_unaligned(items, &stats[0]);
3857         for (i = 1; i < items; i++)
3858                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3859
3860         memset(&stats[items], 0, pad);
3861 }
3862
3863 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3864                              int bytes)
3865 {
3866         switch (attrtype) {
3867         case IFLA_INET6_STATS:
3868                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3869                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3870                 break;
3871         case IFLA_INET6_ICMP6STATS:
3872                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3873                 break;
3874         }
3875 }
3876
3877 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3878 {
3879         struct nlattr *nla;
3880         struct ifla_cacheinfo ci;
3881
3882         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3883
3884         ci.max_reasm_len = IPV6_MAXPLEN;
3885         ci.tstamp = cstamp_delta(idev->tstamp);
3886         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3887         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3888         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3889
3890         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3891         if (nla == NULL)
3892                 goto nla_put_failure;
3893         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3894
3895         /* XXX - MC not implemented */
3896
3897         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3898         if (nla == NULL)
3899                 goto nla_put_failure;
3900         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3901
3902         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3903         if (nla == NULL)
3904                 goto nla_put_failure;
3905         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3906
3907         return 0;
3908
3909 nla_put_failure:
3910         return -EMSGSIZE;
3911 }
3912
3913 static size_t inet6_get_link_af_size(const struct net_device *dev)
3914 {
3915         if (!__in6_dev_get(dev))
3916                 return 0;
3917
3918         return inet6_ifla6_size();
3919 }
3920
3921 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3922 {
3923         struct inet6_dev *idev = __in6_dev_get(dev);
3924
3925         if (!idev)
3926                 return -ENODATA;
3927
3928         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3929                 return -EMSGSIZE;
3930
3931         return 0;
3932 }
3933
3934 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3935                              u32 pid, u32 seq, int event, unsigned int flags)
3936 {
3937         struct net_device *dev = idev->dev;
3938         struct ifinfomsg *hdr;
3939         struct nlmsghdr *nlh;
3940         void *protoinfo;
3941
3942         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3943         if (nlh == NULL)
3944                 return -EMSGSIZE;
3945
3946         hdr = nlmsg_data(nlh);
3947         hdr->ifi_family = AF_INET6;
3948         hdr->__ifi_pad = 0;
3949         hdr->ifi_type = dev->type;
3950         hdr->ifi_index = dev->ifindex;
3951         hdr->ifi_flags = dev_get_flags(dev);
3952         hdr->ifi_change = 0;
3953
3954         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3955
3956         if (dev->addr_len)
3957                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3958
3959         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3960         if (dev->ifindex != dev->iflink)
3961                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3962
3963         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3964         if (protoinfo == NULL)
3965                 goto nla_put_failure;
3966
3967         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3968                 goto nla_put_failure;
3969
3970         nla_nest_end(skb, protoinfo);
3971         return nlmsg_end(skb, nlh);
3972
3973 nla_put_failure:
3974         nlmsg_cancel(skb, nlh);
3975         return -EMSGSIZE;
3976 }
3977
3978 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3979 {
3980         struct net *net = sock_net(skb->sk);
3981         int h, s_h;
3982         int idx = 0, s_idx;
3983         struct net_device *dev;
3984         struct inet6_dev *idev;
3985         struct hlist_head *head;
3986         struct hlist_node *node;
3987
3988         s_h = cb->args[0];
3989         s_idx = cb->args[1];
3990
3991         rcu_read_lock();
3992         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3993                 idx = 0;
3994                 head = &net->dev_index_head[h];
3995                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3996                         if (idx < s_idx)
3997                                 goto cont;
3998                         idev = __in6_dev_get(dev);
3999                         if (!idev)
4000                                 goto cont;
4001                         if (inet6_fill_ifinfo(skb, idev,
4002                                               NETLINK_CB(cb->skb).pid,
4003                                               cb->nlh->nlmsg_seq,
4004                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4005                                 goto out;
4006 cont:
4007                         idx++;
4008                 }
4009         }
4010 out:
4011         rcu_read_unlock();
4012         cb->args[1] = idx;
4013         cb->args[0] = h;
4014
4015         return skb->len;
4016 }
4017
4018 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4019 {
4020         struct sk_buff *skb;
4021         struct net *net = dev_net(idev->dev);
4022         int err = -ENOBUFS;
4023
4024         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4025         if (skb == NULL)
4026                 goto errout;
4027
4028         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4029         if (err < 0) {
4030                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4031                 WARN_ON(err == -EMSGSIZE);
4032                 kfree_skb(skb);
4033                 goto errout;
4034         }
4035         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4036         return;
4037 errout:
4038         if (err < 0)
4039                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4040 }
4041
4042 static inline size_t inet6_prefix_nlmsg_size(void)
4043 {
4044         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4045                + nla_total_size(sizeof(struct in6_addr))
4046                + nla_total_size(sizeof(struct prefix_cacheinfo));
4047 }
4048
4049 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4050                              struct prefix_info *pinfo, u32 pid, u32 seq,
4051                              int event, unsigned int flags)
4052 {
4053         struct prefixmsg *pmsg;
4054         struct nlmsghdr *nlh;
4055         struct prefix_cacheinfo ci;
4056
4057         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4058         if (nlh == NULL)
4059                 return -EMSGSIZE;
4060
4061         pmsg = nlmsg_data(nlh);
4062         pmsg->prefix_family = AF_INET6;
4063         pmsg->prefix_pad1 = 0;
4064         pmsg->prefix_pad2 = 0;
4065         pmsg->prefix_ifindex = idev->dev->ifindex;
4066         pmsg->prefix_len = pinfo->prefix_len;
4067         pmsg->prefix_type = pinfo->type;
4068         pmsg->prefix_pad3 = 0;
4069         pmsg->prefix_flags = 0;
4070         if (pinfo->onlink)
4071                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4072         if (pinfo->autoconf)
4073                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4074
4075         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4076
4077         ci.preferred_time = ntohl(pinfo->prefered);
4078         ci.valid_time = ntohl(pinfo->valid);
4079         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4080
4081         return nlmsg_end(skb, nlh);
4082
4083 nla_put_failure:
4084         nlmsg_cancel(skb, nlh);
4085         return -EMSGSIZE;
4086 }
4087
4088 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4089                          struct prefix_info *pinfo)
4090 {
4091         struct sk_buff *skb;
4092         struct net *net = dev_net(idev->dev);
4093         int err = -ENOBUFS;
4094
4095         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4096         if (skb == NULL)
4097                 goto errout;
4098
4099         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4100         if (err < 0) {
4101                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4102                 WARN_ON(err == -EMSGSIZE);
4103                 kfree_skb(skb);
4104                 goto errout;
4105         }
4106         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4107         return;
4108 errout:
4109         if (err < 0)
4110                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4111 }
4112
4113 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4114 {
4115         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4116
4117         switch (event) {
4118         case RTM_NEWADDR:
4119                 /*
4120                  * If the address was optimistic
4121                  * we inserted the route at the start of
4122                  * our DAD process, so we don't need
4123                  * to do it again
4124                  */
4125                 if (!(ifp->rt->rt6i_node))
4126                         ip6_ins_rt(ifp->rt);
4127                 if (ifp->idev->cnf.forwarding)
4128                         addrconf_join_anycast(ifp);
4129                 break;
4130         case RTM_DELADDR:
4131                 if (ifp->idev->cnf.forwarding)
4132                         addrconf_leave_anycast(ifp);
4133                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4134                 dst_hold(&ifp->rt->dst);
4135
4136                 if (ip6_del_rt(ifp->rt))
4137                         dst_free(&ifp->rt->dst);
4138                 break;
4139         }
4140 }
4141
4142 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4143 {
4144         rcu_read_lock_bh();
4145         if (likely(ifp->idev->dead == 0))
4146                 __ipv6_ifa_notify(event, ifp);
4147         rcu_read_unlock_bh();
4148 }
4149
4150 #ifdef CONFIG_SYSCTL
4151
4152 static
4153 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4154                            void __user *buffer, size_t *lenp, loff_t *ppos)
4155 {
4156         int *valp = ctl->data;
4157         int val = *valp;
4158         loff_t pos = *ppos;
4159         int ret;
4160
4161         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4162
4163         if (write)
4164                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4165         if (ret)
4166                 *ppos = pos;
4167         return ret;
4168 }
4169
4170 static void dev_disable_change(struct inet6_dev *idev)
4171 {
4172         if (!idev || !idev->dev)
4173                 return;
4174
4175         if (idev->cnf.disable_ipv6)
4176                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4177         else
4178                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4179 }
4180
4181 static void addrconf_disable_change(struct net *net, __s32 newf)
4182 {
4183         struct net_device *dev;
4184         struct inet6_dev *idev;
4185
4186         rcu_read_lock();
4187         for_each_netdev_rcu(net, dev) {
4188                 idev = __in6_dev_get(dev);
4189                 if (idev) {
4190                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4191                         idev->cnf.disable_ipv6 = newf;
4192                         if (changed)
4193                                 dev_disable_change(idev);
4194                 }
4195         }
4196         rcu_read_unlock();
4197 }
4198
4199 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4200 {
4201         struct net *net;
4202
4203         net = (struct net *)table->extra2;
4204
4205         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4206                 return 0;
4207
4208         if (!rtnl_trylock()) {
4209                 /* Restore the original values before restarting */
4210                 *p = old;
4211                 return restart_syscall();
4212         }
4213
4214         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4215                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4216                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4217                 addrconf_disable_change(net, newf);
4218         } else if ((!*p) ^ (!old))
4219                 dev_disable_change((struct inet6_dev *)table->extra1);
4220
4221         rtnl_unlock();
4222         return 0;
4223 }
4224
4225 static
4226 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4227                             void __user *buffer, size_t *lenp, loff_t *ppos)
4228 {
4229         int *valp = ctl->data;
4230         int val = *valp;
4231         loff_t pos = *ppos;
4232         int ret;
4233
4234         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4235
4236         if (write)
4237                 ret = addrconf_disable_ipv6(ctl, valp, val);
4238         if (ret)
4239                 *ppos = pos;
4240         return ret;
4241 }
4242
4243 static struct addrconf_sysctl_table
4244 {
4245         struct ctl_table_header *sysctl_header;
4246         ctl_table addrconf_vars[DEVCONF_MAX+1];
4247         char *dev_name;
4248 } addrconf_sysctl __read_mostly = {
4249         .sysctl_header = NULL,
4250         .addrconf_vars = {
4251                 {
4252                         .procname       = "forwarding",
4253                         .data           = &ipv6_devconf.forwarding,
4254                         .maxlen         = sizeof(int),
4255                         .mode           = 0644,
4256                         .proc_handler   = addrconf_sysctl_forward,
4257                 },
4258                 {
4259                         .procname       = "hop_limit",
4260                         .data           = &ipv6_devconf.hop_limit,
4261                         .maxlen         = sizeof(int),
4262                         .mode           = 0644,
4263                         .proc_handler   = proc_dointvec,
4264                 },
4265                 {
4266                         .procname       = "mtu",
4267                         .data           = &ipv6_devconf.mtu6,
4268                         .maxlen         = sizeof(int),
4269                         .mode           = 0644,
4270                         .proc_handler   = proc_dointvec,
4271                 },
4272                 {
4273                         .procname       = "accept_ra",
4274                         .data           = &ipv6_devconf.accept_ra,
4275                         .maxlen         = sizeof(int),
4276                         .mode           = 0644,
4277                         .proc_handler   = proc_dointvec,
4278                 },
4279                 {
4280                         .procname       = "accept_redirects",
4281                         .data           = &ipv6_devconf.accept_redirects,
4282                         .maxlen         = sizeof(int),
4283                         .mode           = 0644,
4284                         .proc_handler   = proc_dointvec,
4285                 },
4286                 {
4287                         .procname       = "autoconf",
4288                         .data           = &ipv6_devconf.autoconf,
4289                         .maxlen         = sizeof(int),
4290                         .mode           = 0644,
4291                         .proc_handler   = proc_dointvec,
4292                 },
4293                 {
4294                         .procname       = "dad_transmits",
4295                         .data           = &ipv6_devconf.dad_transmits,
4296                         .maxlen         = sizeof(int),
4297                         .mode           = 0644,
4298                         .proc_handler   = proc_dointvec,
4299                 },
4300                 {
4301                         .procname       = "router_solicitations",
4302                         .data           = &ipv6_devconf.rtr_solicits,
4303                         .maxlen         = sizeof(int),
4304                         .mode           = 0644,
4305                         .proc_handler   = proc_dointvec,
4306                 },
4307                 {
4308                         .procname       = "router_solicitation_interval",
4309                         .data           = &ipv6_devconf.rtr_solicit_interval,
4310                         .maxlen         = sizeof(int),
4311                         .mode           = 0644,
4312                         .proc_handler   = proc_dointvec_jiffies,
4313                 },
4314                 {
4315                         .procname       = "router_solicitation_delay",
4316                         .data           = &ipv6_devconf.rtr_solicit_delay,
4317                         .maxlen         = sizeof(int),
4318                         .mode           = 0644,
4319                         .proc_handler   = proc_dointvec_jiffies,
4320                 },
4321                 {
4322                         .procname       = "force_mld_version",
4323                         .data           = &ipv6_devconf.force_mld_version,
4324                         .maxlen         = sizeof(int),
4325                         .mode           = 0644,
4326                         .proc_handler   = proc_dointvec,
4327                 },
4328 #ifdef CONFIG_IPV6_PRIVACY
4329                 {
4330                         .procname       = "use_tempaddr",
4331                         .data           = &ipv6_devconf.use_tempaddr,
4332                         .maxlen         = sizeof(int),
4333                         .mode           = 0644,
4334                         .proc_handler   = proc_dointvec,
4335                 },
4336                 {
4337                         .procname       = "temp_valid_lft",
4338                         .data           = &ipv6_devconf.temp_valid_lft,
4339                         .maxlen         = sizeof(int),
4340                         .mode           = 0644,
4341                         .proc_handler   = proc_dointvec,
4342                 },
4343                 {
4344                         .procname       = "temp_prefered_lft",
4345                         .data           = &ipv6_devconf.temp_prefered_lft,
4346                         .maxlen         = sizeof(int),
4347                         .mode           = 0644,
4348                         .proc_handler   = proc_dointvec,
4349                 },
4350                 {
4351                         .procname       = "regen_max_retry",
4352                         .data           = &ipv6_devconf.regen_max_retry,
4353                         .maxlen         = sizeof(int),
4354                         .mode           = 0644,
4355                         .proc_handler   = proc_dointvec,
4356                 },
4357                 {
4358                         .procname       = "max_desync_factor",
4359                         .data           = &ipv6_devconf.max_desync_factor,
4360                         .maxlen         = sizeof(int),
4361                         .mode           = 0644,
4362                         .proc_handler   = proc_dointvec,
4363                 },
4364 #endif
4365                 {
4366                         .procname       = "max_addresses",
4367                         .data           = &ipv6_devconf.max_addresses,
4368                         .maxlen         = sizeof(int),
4369                         .mode           = 0644,
4370                         .proc_handler   = proc_dointvec,
4371                 },
4372                 {
4373                         .procname       = "accept_ra_defrtr",
4374                         .data           = &ipv6_devconf.accept_ra_defrtr,
4375                         .maxlen         = sizeof(int),
4376                         .mode           = 0644,
4377                         .proc_handler   = proc_dointvec,
4378                 },
4379                 {
4380                         .procname       = "accept_ra_pinfo",
4381                         .data           = &ipv6_devconf.accept_ra_pinfo,
4382                         .maxlen         = sizeof(int),
4383                         .mode           = 0644,
4384                         .proc_handler   = proc_dointvec,
4385                 },
4386 #ifdef CONFIG_IPV6_ROUTER_PREF
4387                 {
4388                         .procname       = "accept_ra_rtr_pref",
4389                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4390                         .maxlen         = sizeof(int),
4391                         .mode           = 0644,
4392                         .proc_handler   = proc_dointvec,
4393                 },
4394                 {
4395                         .procname       = "router_probe_interval",
4396                         .data           = &ipv6_devconf.rtr_probe_interval,
4397                         .maxlen         = sizeof(int),
4398                         .mode           = 0644,
4399                         .proc_handler   = proc_dointvec_jiffies,
4400                 },
4401 #ifdef CONFIG_IPV6_ROUTE_INFO
4402                 {
4403                         .procname       = "accept_ra_rt_info_max_plen",
4404                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4405                         .maxlen         = sizeof(int),
4406                         .mode           = 0644,
4407                         .proc_handler   = proc_dointvec,
4408                 },
4409 #endif
4410 #endif
4411                 {
4412                         .procname       = "proxy_ndp",
4413                         .data           = &ipv6_devconf.proxy_ndp,
4414                         .maxlen         = sizeof(int),
4415                         .mode           = 0644,
4416                         .proc_handler   = proc_dointvec,
4417                 },
4418                 {
4419                         .procname       = "accept_source_route",
4420                         .data           = &ipv6_devconf.accept_source_route,
4421                         .maxlen         = sizeof(int),
4422                         .mode           = 0644,
4423                         .proc_handler   = proc_dointvec,
4424                 },
4425 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4426                 {
4427                         .procname       = "optimistic_dad",
4428                         .data           = &ipv6_devconf.optimistic_dad,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = proc_dointvec,
4432
4433                 },
4434 #endif
4435 #ifdef CONFIG_IPV6_MROUTE
4436                 {
4437                         .procname       = "mc_forwarding",
4438                         .data           = &ipv6_devconf.mc_forwarding,
4439                         .maxlen         = sizeof(int),
4440                         .mode           = 0444,
4441                         .proc_handler   = proc_dointvec,
4442                 },
4443 #endif
4444                 {
4445                         .procname       = "disable_ipv6",
4446                         .data           = &ipv6_devconf.disable_ipv6,
4447                         .maxlen         = sizeof(int),
4448                         .mode           = 0644,
4449                         .proc_handler   = addrconf_sysctl_disable,
4450                 },
4451                 {
4452                         .procname       = "accept_dad",
4453                         .data           = &ipv6_devconf.accept_dad,
4454                         .maxlen         = sizeof(int),
4455                         .mode           = 0644,
4456                         .proc_handler   = proc_dointvec,
4457                 },
4458                 {
4459                         .procname       = "force_tllao",
4460                         .data           = &ipv6_devconf.force_tllao,
4461                         .maxlen         = sizeof(int),
4462                         .mode           = 0644,
4463                         .proc_handler   = proc_dointvec
4464                 },
4465                 {
4466                         /* sentinel */
4467                 }
4468         },
4469 };
4470
4471 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4472                 struct inet6_dev *idev, struct ipv6_devconf *p)
4473 {
4474         int i;
4475         struct addrconf_sysctl_table *t;
4476
4477 #define ADDRCONF_CTL_PATH_DEV   3
4478
4479         struct ctl_path addrconf_ctl_path[] = {
4480                 { .procname = "net", },
4481                 { .procname = "ipv6", },
4482                 { .procname = "conf", },
4483                 { /* to be set */ },
4484                 { },
4485         };
4486
4487
4488         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4489         if (t == NULL)
4490                 goto out;
4491
4492         for (i = 0; t->addrconf_vars[i].data; i++) {
4493                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4494                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4495                 t->addrconf_vars[i].extra2 = net;
4496         }
4497
4498         /*
4499          * Make a copy of dev_name, because '.procname' is regarded as const
4500          * by sysctl and we wouldn't want anyone to change it under our feet
4501          * (see SIOCSIFNAME).
4502          */
4503         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4504         if (!t->dev_name)
4505                 goto free;
4506
4507         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4508
4509         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4510                         t->addrconf_vars);
4511         if (t->sysctl_header == NULL)
4512                 goto free_procname;
4513
4514         p->sysctl = t;
4515         return 0;
4516
4517 free_procname:
4518         kfree(t->dev_name);
4519 free:
4520         kfree(t);
4521 out:
4522         return -ENOBUFS;
4523 }
4524
4525 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4526 {
4527         struct addrconf_sysctl_table *t;
4528
4529         if (p->sysctl == NULL)
4530                 return;
4531
4532         t = p->sysctl;
4533         p->sysctl = NULL;
4534         unregister_net_sysctl_table(t->sysctl_header);
4535         kfree(t->dev_name);
4536         kfree(t);
4537 }
4538
4539 static void addrconf_sysctl_register(struct inet6_dev *idev)
4540 {
4541         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4542                               &ndisc_ifinfo_sysctl_change);
4543         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4544                                         idev, &idev->cnf);
4545 }
4546
4547 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4548 {
4549         __addrconf_sysctl_unregister(&idev->cnf);
4550         neigh_sysctl_unregister(idev->nd_parms);
4551 }
4552
4553
4554 #endif
4555
4556 static int __net_init addrconf_init_net(struct net *net)
4557 {
4558         int err;
4559         struct ipv6_devconf *all, *dflt;
4560
4561         err = -ENOMEM;
4562         all = &ipv6_devconf;
4563         dflt = &ipv6_devconf_dflt;
4564
4565         if (!net_eq(net, &init_net)) {
4566                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4567                 if (all == NULL)
4568                         goto err_alloc_all;
4569
4570                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4571                 if (dflt == NULL)
4572                         goto err_alloc_dflt;
4573         } else {
4574                 /* these will be inherited by all namespaces */
4575                 dflt->autoconf = ipv6_defaults.autoconf;
4576                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4577         }
4578
4579         net->ipv6.devconf_all = all;
4580         net->ipv6.devconf_dflt = dflt;
4581
4582 #ifdef CONFIG_SYSCTL
4583         err = __addrconf_sysctl_register(net, "all", NULL, all);
4584         if (err < 0)
4585                 goto err_reg_all;
4586
4587         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4588         if (err < 0)
4589                 goto err_reg_dflt;
4590 #endif
4591         return 0;
4592
4593 #ifdef CONFIG_SYSCTL
4594 err_reg_dflt:
4595         __addrconf_sysctl_unregister(all);
4596 err_reg_all:
4597         kfree(dflt);
4598 #endif
4599 err_alloc_dflt:
4600         kfree(all);
4601 err_alloc_all:
4602         return err;
4603 }
4604
4605 static void __net_exit addrconf_exit_net(struct net *net)
4606 {
4607 #ifdef CONFIG_SYSCTL
4608         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4609         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4610 #endif
4611         if (!net_eq(net, &init_net)) {
4612                 kfree(net->ipv6.devconf_dflt);
4613                 kfree(net->ipv6.devconf_all);
4614         }
4615 }
4616
4617 static struct pernet_operations addrconf_ops = {
4618         .init = addrconf_init_net,
4619         .exit = addrconf_exit_net,
4620 };
4621
4622 /*
4623  *      Device notifier
4624  */
4625
4626 int register_inet6addr_notifier(struct notifier_block *nb)
4627 {
4628         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4629 }
4630 EXPORT_SYMBOL(register_inet6addr_notifier);
4631
4632 int unregister_inet6addr_notifier(struct notifier_block *nb)
4633 {
4634         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4635 }
4636 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4637
4638 static struct rtnl_af_ops inet6_ops = {
4639         .family           = AF_INET6,
4640         .fill_link_af     = inet6_fill_link_af,
4641         .get_link_af_size = inet6_get_link_af_size,
4642 };
4643
4644 /*
4645  *      Init / cleanup code
4646  */
4647
4648 int __init addrconf_init(void)
4649 {
4650         int i, err;
4651
4652         err = ipv6_addr_label_init();
4653         if (err < 0) {
4654                 printk(KERN_CRIT "IPv6 Addrconf:"
4655                        " cannot initialize default policy table: %d.\n", err);
4656                 goto out;
4657         }
4658
4659         err = register_pernet_subsys(&addrconf_ops);
4660         if (err < 0)
4661                 goto out_addrlabel;
4662
4663         /* The addrconf netdev notifier requires that loopback_dev
4664          * has it's ipv6 private information allocated and setup
4665          * before it can bring up and give link-local addresses
4666          * to other devices which are up.
4667          *
4668          * Unfortunately, loopback_dev is not necessarily the first
4669          * entry in the global dev_base list of net devices.  In fact,
4670          * it is likely to be the very last entry on that list.
4671          * So this causes the notifier registry below to try and
4672          * give link-local addresses to all devices besides loopback_dev
4673          * first, then loopback_dev, which cases all the non-loopback_dev
4674          * devices to fail to get a link-local address.
4675          *
4676          * So, as a temporary fix, allocate the ipv6 structure for
4677          * loopback_dev first by hand.
4678          * Longer term, all of the dependencies ipv6 has upon the loopback
4679          * device and it being up should be removed.
4680          */
4681         rtnl_lock();
4682         if (!ipv6_add_dev(init_net.loopback_dev))
4683                 err = -ENOMEM;
4684         rtnl_unlock();
4685         if (err)
4686                 goto errlo;
4687
4688         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4689                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4690
4691         register_netdevice_notifier(&ipv6_dev_notf);
4692
4693         addrconf_verify(0);
4694
4695         err = rtnl_af_register(&inet6_ops);
4696         if (err < 0)
4697                 goto errout_af;
4698
4699         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4700         if (err < 0)
4701                 goto errout;
4702
4703         /* Only the first call to __rtnl_register can fail */
4704         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4705         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4706         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4707         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4708         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4709
4710         ipv6_addr_label_rtnl_register();
4711
4712         return 0;
4713 errout:
4714         rtnl_af_unregister(&inet6_ops);
4715 errout_af:
4716         unregister_netdevice_notifier(&ipv6_dev_notf);
4717 errlo:
4718         unregister_pernet_subsys(&addrconf_ops);
4719 out_addrlabel:
4720         ipv6_addr_label_cleanup();
4721 out:
4722         return err;
4723 }
4724
4725 void addrconf_cleanup(void)
4726 {
4727         struct net_device *dev;
4728         int i;
4729
4730         unregister_netdevice_notifier(&ipv6_dev_notf);
4731         unregister_pernet_subsys(&addrconf_ops);
4732         ipv6_addr_label_cleanup();
4733
4734         rtnl_lock();
4735
4736         __rtnl_af_unregister(&inet6_ops);
4737
4738         /* clean dev list */
4739         for_each_netdev(&init_net, dev) {
4740                 if (__in6_dev_get(dev) == NULL)
4741                         continue;
4742                 addrconf_ifdown(dev, 1);
4743         }
4744         addrconf_ifdown(init_net.loopback_dev, 2);
4745
4746         /*
4747          *      Check hash table.
4748          */
4749         spin_lock_bh(&addrconf_hash_lock);
4750         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4751                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4752         spin_unlock_bh(&addrconf_hash_lock);
4753
4754         del_timer(&addr_chk_timer);
4755         rtnl_unlock();
4756 }