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         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 293                                         GFP_KERNEL);
 294         if (!idev->stats.icmpv6dev)
 295                 goto err_icmp;
 296         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 297                                            GFP_KERNEL);
 298         if (!idev->stats.icmpv6msgdev)
 299                 goto err_icmpmsg;
 300
 301         return 0;
 302
 303 err_icmpmsg:
 304         kfree(idev->stats.icmpv6dev);
 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         kfree(idev->stats.icmpv6msgdev);
 314         kfree(idev->stats.icmpv6dev);
 315         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 316 }
 317
 318 /* Nobody refers to this device, we may destroy it. */
 319
 320 void in6_dev_finish_destroy(struct inet6_dev *idev)
 321 {
 322         struct net_device *dev = idev->dev;
 323
 324         WARN_ON(!list_empty(&idev->addr_list));
 325         WARN_ON(idev->mc_list != NULL);
 326
 327 #ifdef NET_REFCNT_DEBUG
 328         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 329 #endif
 330         dev_put(dev);
 331         if (!idev->dead) {
 332                 pr_warning("Freeing alive inet6 device %p\n", idev);
 333                 return;
 334         }
 335         snmp6_free_dev(idev);
 336         kfree_rcu(idev, rcu);
 337 }
 338
 339 EXPORT_SYMBOL(in6_dev_finish_destroy);
 340
 341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 342 {
 343         struct inet6_dev *ndev;
 344
 345         ASSERT_RTNL();
 346
 347         if (dev->mtu < IPV6_MIN_MTU)
 348                 return NULL;
 349
 350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 351
 352         if (ndev == NULL)
 353                 return NULL;
 354
 355         rwlock_init(&ndev->lock);
 356         ndev->dev = dev;
 357         INIT_LIST_HEAD(&ndev->addr_list);
 358
 359         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 360         ndev->cnf.mtu6 = dev->mtu;
 361         ndev->cnf.sysctl = NULL;
 362         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 363         if (ndev->nd_parms == NULL) {
 364                 kfree(ndev);
 365                 return NULL;
 366         }
 367         if (ndev->cnf.forwarding)
 368                 dev_disable_lro(dev);
 369         /* We refer to the device */
 370         dev_hold(dev);
 371
 372         if (snmp6_alloc_dev(ndev) < 0) {
 373                 ADBG((KERN_WARNING
 374                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 375                         __func__, dev->name));
 376                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 377                 ndev->dead = 1;
 378                 in6_dev_finish_destroy(ndev);
 379                 return NULL;
 380         }
 381
 382         if (snmp6_register_dev(ndev) < 0) {
 383                 ADBG((KERN_WARNING
 384                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 385                         __func__, dev->name));
 386                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 387                 ndev->dead = 1;
 388                 in6_dev_finish_destroy(ndev);
 389                 return NULL;
 390         }
 391
 392         /* One reference from device.  We must do this before
 393          * we invoke __ipv6_regen_rndid().
 394          */
 395         in6_dev_hold(ndev);
 396
 397         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 398                 ndev->cnf.accept_dad = -1;
 399
 400 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 401         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 402                 printk(KERN_INFO
 403                        "%s: Disabled Multicast RS\n",
 404                        dev->name);
 405                 ndev->cnf.rtr_solicits = 0;
 406         }
 407 #endif
 408
 409 #ifdef CONFIG_IPV6_PRIVACY
 410         INIT_LIST_HEAD(&ndev->tempaddr_list);
 411         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 412         if ((dev->flags&IFF_LOOPBACK) ||
 413             dev->type == ARPHRD_TUNNEL ||
 414             dev->type == ARPHRD_TUNNEL6 ||
 415             dev->type == ARPHRD_SIT ||
 416             dev->type == ARPHRD_NONE) {
 417                 ndev->cnf.use_tempaddr = -1;
 418         } else {
 419                 in6_dev_hold(ndev);
 420                 ipv6_regen_rndid((unsigned long) ndev);
 421         }
 422 #endif
 423
 424         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 425                 ndev->if_flags |= IF_READY;
 426
 427         ipv6_mc_init_dev(ndev);
 428         ndev->tstamp = jiffies;
 429         addrconf_sysctl_register(ndev);
 430         /* protected by rtnl_lock */
 431         rcu_assign_pointer(dev->ip6_ptr, ndev);
 432
 433         /* Join all-node multicast group */
 434         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 435
 436         return ndev;
 437 }
 438
 439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 440 {
 441         struct inet6_dev *idev;
 442
 443         ASSERT_RTNL();
 444
 445         idev = __in6_dev_get(dev);
 446         if (!idev) {
 447                 idev = ipv6_add_dev(dev);
 448                 if (!idev)
 449                         return NULL;
 450         }
 451
 452         if (dev->flags&IFF_UP)
 453                 ipv6_mc_up(idev);
 454         return idev;
 455 }
 456
 457 #ifdef CONFIG_SYSCTL
 458 static void dev_forward_change(struct inet6_dev *idev)
 459 {
 460         struct net_device *dev;
 461         struct inet6_ifaddr *ifa;
 462
 463         if (!idev)
 464                 return;
 465         dev = idev->dev;
 466         if (idev->cnf.forwarding)
 467                 dev_disable_lro(dev);
 468         if (dev && (dev->flags & IFF_MULTICAST)) {
 469                 if (idev->cnf.forwarding)
 470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471                 else
 472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 473         }
 474
 475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 476                 if (ifa->flags&IFA_F_TENTATIVE)
 477                         continue;
 478                 if (idev->cnf.forwarding)
 479                         addrconf_join_anycast(ifa);
 480                 else
 481                         addrconf_leave_anycast(ifa);
 482         }
 483 }
 484
 485
 486 static void addrconf_forward_change(struct net *net, __s32 newf)
 487 {
 488         struct net_device *dev;
 489         struct inet6_dev *idev;
 490
 491         rcu_read_lock();
 492         for_each_netdev_rcu(net, dev) {
 493                 idev = __in6_dev_get(dev);
 494                 if (idev) {
 495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 496                         idev->cnf.forwarding = newf;
 497                         if (changed)
 498                                 dev_forward_change(idev);
 499                 }
 500         }
 501         rcu_read_unlock();
 502 }
 503
 504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 505 {
 506         struct net *net;
 507
 508         net = (struct net *)table->extra2;
 509         if (p == &net->ipv6.devconf_dflt->forwarding)
 510                 return 0;
 511
 512         if (!rtnl_trylock()) {
 513                 /* Restore the original values before restarting */
 514                 *p = old;
 515                 return restart_syscall();
 516         }
 517
 518         if (p == &net->ipv6.devconf_all->forwarding) {
 519                 __s32 newf = net->ipv6.devconf_all->forwarding;
 520                 net->ipv6.devconf_dflt->forwarding = newf;
 521                 addrconf_forward_change(net, newf);
 522         } else if ((!*p) ^ (!old))
 523                 dev_forward_change((struct inet6_dev *)table->extra1);
 524         rtnl_unlock();
 525
 526         if (*p)
 527                 rt6_purge_dflt_routers(net);
 528         return 1;
 529 }
 530 #endif
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 536
 537 #ifdef NET_REFCNT_DEBUG
 538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 539 #endif
 540
 541         in6_dev_put(ifp->idev);
 542
 543         if (del_timer(&ifp->timer))
 544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 545
 546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 548                 return;
 549         }
 550         dst_release(&ifp->rt->dst);
 551
 552         kfree_rcu(ifp, rcu);
 553 }
 554
 555 static void
 556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 557 {
 558         struct list_head *p;
 559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 560
 561         /*
 562          * Each device address list is sorted in order of scope -
 563          * global before linklocal.
 564          */
 565         list_for_each(p, &idev->addr_list) {
 566                 struct inet6_ifaddr *ifa
 567                         = list_entry(p, struct inet6_ifaddr, if_list);
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         list_add_tail(&ifp->if_list, p);
 573 }
 574
 575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 576 {
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581         return jhash_2words((__force u32)addr->s6_addr32[2],
 582                             (__force u32)addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         /*
 653          * part one of RFC 4429, section 3.3
 654          * We should not configure an address as
 655          * optimistic if we do not yet know the link
 656          * layer address of our nexhop router
 657          */
 658
 659         if (dst_get_neighbour(&rt->dst) == NULL)
 660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 661
 662         ifa->idev = idev;
 663         in6_dev_hold(idev);
 664         /* For caller */
 665         in6_ifa_hold(ifa);
 666
 667         /* Add to big hash table */
 668         hash = ipv6_addr_hash(addr);
 669
 670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, *ifn;
 707         struct inet6_dev *idev = ifp->idev;
 708         int state;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         spin_lock_bh(&ifp->state_lock);
 713         state = ifp->state;
 714         ifp->state = INET6_IFADDR_STATE_DEAD;
 715         spin_unlock_bh(&ifp->state_lock);
 716
 717         if (state == INET6_IFADDR_STATE_DEAD)
 718                 goto out;
 719
 720         spin_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init_rcu(&ifp->addr_lst);
 722         spin_unlock_bh(&addrconf_hash_lock);
 723
 724         write_lock_bh(&idev->lock);
 725 #ifdef CONFIG_IPV6_PRIVACY
 726         if (ifp->flags&IFA_F_TEMPORARY) {
 727                 list_del(&ifp->tmp_list);
 728                 if (ifp->ifpub) {
 729                         in6_ifa_put(ifp->ifpub);
 730                         ifp->ifpub = NULL;
 731                 }
 732                 __in6_ifa_put(ifp);
 733         }
 734 #endif
 735
 736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 737                 if (ifa == ifp) {
 738                         list_del_init(&ifp->if_list);
 739                         __in6_ifa_put(ifp);
 740
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773         }
 774         write_unlock_bh(&idev->lock);
 775
 776         addrconf_del_timer(ifp);
 777
 778         ipv6_ifa_notify(RTM_DELADDR, ifp);
 779
 780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 781
 782         /*
 783          * Purge or update corresponding prefix
 784          *
 785          * 1) we don't purge prefix here if address was not permanent.
 786          *    prefix is managed by its own lifetime.
 787          * 2) if there're no addresses, delete prefix.
 788          * 3) if there're still other permanent address(es),
 789          *    corresponding prefix is still permanent.
 790          * 4) otherwise, update prefix lifetime to the
 791          *    longest valid lifetime among the corresponding
 792          *    addresses on the device.
 793          *    Note: subsequent RA will update lifetime.
 794          *
 795          * --yoshfuji
 796          */
 797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 798                 struct in6_addr prefix;
 799                 struct rt6_info *rt;
 800                 struct net *net = dev_net(ifp->idev->dev);
 801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 803
 804                 if (rt && addrconf_is_prefix_route(rt)) {
 805                         if (onlink == 0) {
 806                                 ip6_del_rt(rt);
 807                                 rt = NULL;
 808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 809                                 rt->rt6i_expires = expires;
 810                                 rt->rt6i_flags |= RTF_EXPIRES;
 811                         }
 812                 }
 813                 dst_release(&rt->dst);
 814         }
 815
 816         /* clean up prefsrc entries */
 817         rt6_remove_prefsrc(ifp);
 818 out:
 819         in6_ifa_put(ifp);
 820 }
 821
 822 #ifdef CONFIG_IPV6_PRIVACY
 823 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 824 {
 825         struct inet6_dev *idev = ifp->idev;
 826         struct in6_addr addr, *tmpaddr;
 827         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 828         unsigned long regen_advance;
 829         int tmp_plen;
 830         int ret = 0;
 831         int max_addresses;
 832         u32 addr_flags;
 833
 834         write_lock(&idev->lock);
 835         if (ift) {
 836                 spin_lock_bh(&ift->lock);
 837                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 838                 spin_unlock_bh(&ift->lock);
 839                 tmpaddr = &addr;
 840         } else {
 841                 tmpaddr = NULL;
 842         }
 843 retry:
 844         in6_dev_hold(idev);
 845         if (idev->cnf.use_tempaddr <= 0) {
 846                 write_unlock(&idev->lock);
 847                 printk(KERN_INFO
 848                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         spin_lock_bh(&ifp->lock);
 854         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 855                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 856                 spin_unlock_bh(&ifp->lock);
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_WARNING
 859                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         in6_ifa_hold(ifp);
 865         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 866         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 871                 in6_ifa_put(ifp);
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 877         age = (jiffies - ifp->tstamp) / HZ;
 878         tmp_valid_lft = min_t(__u32,
 879                               ifp->valid_lft,
 880                               idev->cnf.temp_valid_lft + age);
 881         tmp_prefered_lft = min_t(__u32,
 882                                  ifp->prefered_lft,
 883                                  idev->cnf.temp_prefered_lft + age -
 884                                  idev->cnf.max_desync_factor);
 885         tmp_plen = ifp->prefix_len;
 886         max_addresses = idev->cnf.max_addresses;
 887         tmp_cstamp = ifp->cstamp;
 888         tmp_tstamp = ifp->tstamp;
 889         spin_unlock_bh(&ifp->lock);
 890
 891         regen_advance = idev->cnf.regen_max_retry *
 892                         idev->cnf.dad_transmits *
 893                         idev->nd_parms->retrans_time / HZ;
 894         write_unlock(&idev->lock);
 895
 896         /* A temporary address is created only if this calculated Preferred
 897          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 898          * an implementation must not create a temporary address with a zero
 899          * Preferred Lifetime.
 900          */
 901         if (tmp_prefered_lft <= regen_advance) {
 902                 in6_ifa_put(ifp);
 903                 in6_dev_put(idev);
 904                 ret = -1;
 905                 goto out;
 906         }
 907
 908         addr_flags = IFA_F_TEMPORARY;
 909         /* set in addrconf_prefix_rcv() */
 910         if (ifp->flags & IFA_F_OPTIMISTIC)
 911                 addr_flags |= IFA_F_OPTIMISTIC;
 912
 913         ift = !max_addresses ||
 914               ipv6_count_addresses(idev) < max_addresses ?
 915                 ipv6_add_addr(idev, &addr, tmp_plen,
 916                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 917                               addr_flags) : NULL;
 918         if (!ift || IS_ERR(ift)) {
 919                 in6_ifa_put(ifp);
 920                 in6_dev_put(idev);
 921                 printk(KERN_INFO
 922                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 923                 tmpaddr = &addr;
 924                 write_lock(&idev->lock);
 925                 goto retry;
 926         }
 927
 928         spin_lock_bh(&ift->lock);
 929         ift->ifpub = ifp;
 930         ift->valid_lft = tmp_valid_lft;
 931         ift->prefered_lft = tmp_prefered_lft;
 932         ift->cstamp = tmp_cstamp;
 933         ift->tstamp = tmp_tstamp;
 934         spin_unlock_bh(&ift->lock);
 935
 936         addrconf_dad_start(ift, 0);
 937         in6_ifa_put(ift);
 938         in6_dev_put(idev);
 939 out:
 940         return ret;
 941 }
 942 #endif
 943
 944 /*
 945  *      Choose an appropriate source address (RFC3484)
 946  */
 947 enum {
 948         IPV6_SADDR_RULE_INIT = 0,
 949         IPV6_SADDR_RULE_LOCAL,
 950         IPV6_SADDR_RULE_SCOPE,
 951         IPV6_SADDR_RULE_PREFERRED,
 952 #ifdef CONFIG_IPV6_MIP6
 953         IPV6_SADDR_RULE_HOA,
 954 #endif
 955         IPV6_SADDR_RULE_OIF,
 956         IPV6_SADDR_RULE_LABEL,
 957 #ifdef CONFIG_IPV6_PRIVACY
 958         IPV6_SADDR_RULE_PRIVACY,
 959 #endif
 960         IPV6_SADDR_RULE_ORCHID,
 961         IPV6_SADDR_RULE_PREFIX,
 962         IPV6_SADDR_RULE_MAX
 963 };
 964
 965 struct ipv6_saddr_score {
 966         int                     rule;
 967         int                     addr_type;
 968         struct inet6_ifaddr     *ifa;
 969         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 970         int                     scopedist;
 971         int                     matchlen;
 972 };
 973
 974 struct ipv6_saddr_dst {
 975         const struct in6_addr *addr;
 976         int ifindex;
 977         int scope;
 978         int label;
 979         unsigned int prefs;
 980 };
 981
 982 static inline int ipv6_saddr_preferred(int type)
 983 {
 984         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 985                 return 1;
 986         return 0;
 987 }
 988
 989 static int ipv6_get_saddr_eval(struct net *net,
 990                                struct ipv6_saddr_score *score,
 991                                struct ipv6_saddr_dst *dst,
 992                                int i)
 993 {
 994         int ret;
 995
 996         if (i <= score->rule) {
 997                 switch (i) {
 998                 case IPV6_SADDR_RULE_SCOPE:
 999                         ret = score->scopedist;
1000                         break;
1001                 case IPV6_SADDR_RULE_PREFIX:
1002                         ret = score->matchlen;
1003                         break;
1004                 default:
1005                         ret = !!test_bit(i, score->scorebits);
1006                 }
1007                 goto out;
1008         }
1009
1010         switch (i) {
1011         case IPV6_SADDR_RULE_INIT:
1012                 /* Rule 0: remember if hiscore is not ready yet */
1013                 ret = !!score->ifa;
1014                 break;
1015         case IPV6_SADDR_RULE_LOCAL:
1016                 /* Rule 1: Prefer same address */
1017                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018                 break;
1019         case IPV6_SADDR_RULE_SCOPE:
1020                 /* Rule 2: Prefer appropriate scope
1021                  *
1022                  *      ret
1023                  *       ^
1024                  *    -1 |  d 15
1025                  *    ---+--+-+---> scope
1026                  *       |
1027                  *       |             d is scope of the destination.
1028                  *  B-d  |  \
1029                  *       |   \      <- smaller scope is better if
1030                  *  B-15 |    \        if scope is enough for destinaion.
1031                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032                  * d-C-1 | /
1033                  *       |/         <- greater is better
1034                  *   -C  /             if scope is not enough for destination.
1035                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036                  *
1037                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039                  * Assume B = 0 and we get C > 29.
1040                  */
1041                 ret = __ipv6_addr_src_scope(score->addr_type);
1042                 if (ret >= dst->scope)
1043                         ret = -ret;
1044                 else
1045                         ret -= 128;     /* 30 is enough */
1046                 score->scopedist = ret;
1047                 break;
1048         case IPV6_SADDR_RULE_PREFERRED:
1049                 /* Rule 3: Avoid deprecated and optimistic addresses */
1050                 ret = ipv6_saddr_preferred(score->addr_type) ||
1051                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052                 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054         case IPV6_SADDR_RULE_HOA:
1055             {
1056                 /* Rule 4: Prefer home address */
1057                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059                 break;
1060             }
1061 #endif
1062         case IPV6_SADDR_RULE_OIF:
1063                 /* Rule 5: Prefer outgoing interface */
1064                 ret = (!dst->ifindex ||
1065                        dst->ifindex == score->ifa->idev->dev->ifindex);
1066                 break;
1067         case IPV6_SADDR_RULE_LABEL:
1068                 /* Rule 6: Prefer matching label */
1069                 ret = ipv6_addr_label(net,
1070                                       &score->ifa->addr, score->addr_type,
1071                                       score->ifa->idev->dev->ifindex) == dst->label;
1072                 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074         case IPV6_SADDR_RULE_PRIVACY:
1075             {
1076                 /* Rule 7: Prefer public address
1077                  * Note: prefer temporary address if use_tempaddr >= 2
1078                  */
1079                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1082                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083                 break;
1084             }
1085 #endif
1086         case IPV6_SADDR_RULE_ORCHID:
1087                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088                  *          non-ORCHID vs non-ORCHID
1089                  */
1090                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091                         ipv6_addr_orchid(dst->addr));
1092                 break;
1093         case IPV6_SADDR_RULE_PREFIX:
1094                 /* Rule 8: Use longest matching prefix */
1095                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096                                                        dst->addr);
1097                 break;
1098         default:
1099                 ret = 0;
1100         }
1101
1102         if (ret)
1103                 __set_bit(i, score->scorebits);
1104         score->rule = i;
1105 out:
1106         return ret;
1107 }
1108
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110                        const struct in6_addr *daddr, unsigned int prefs,
1111                        struct in6_addr *saddr)
1112 {
1113         struct ipv6_saddr_score scores[2],
1114                                 *score = &scores[0], *hiscore = &scores[1];
1115         struct ipv6_saddr_dst dst;
1116         struct net_device *dev;
1117         int dst_type;
1118
1119         dst_type = __ipv6_addr_type(daddr);
1120         dst.addr = daddr;
1121         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122         dst.scope = __ipv6_addr_src_scope(dst_type);
1123         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124         dst.prefs = prefs;
1125
1126         hiscore->rule = -1;
1127         hiscore->ifa = NULL;
1128
1129         rcu_read_lock();
1130
1131         for_each_netdev_rcu(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         if (hiscore->ifa)
1209                                                 in6_ifa_put(hiscore->ifa);
1210
1211                                         in6_ifa_hold(score->ifa);
1212
1213                                         swap(hiscore, score);
1214
1215                                         /* restore our iterator */
1216                                         score->ifa = hiscore->ifa;
1217
1218                                         break;
1219                                 }
1220                         }
1221                 }
1222 try_nextdev:
1223                 read_unlock_bh(&idev->lock);
1224         }
1225         rcu_read_unlock();
1226
1227         if (!hiscore->ifa)
1228                 return -EADDRNOTAVAIL;
1229
1230         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231         in6_ifa_put(hiscore->ifa);
1232         return 0;
1233 }
1234 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237                     unsigned char banned_flags)
1238 {
1239         struct inet6_dev *idev;
1240         int err = -EADDRNOTAVAIL;
1241
1242         rcu_read_lock();
1243         idev = __in6_dev_get(dev);
1244         if (idev) {
1245                 struct inet6_ifaddr *ifp;
1246
1247                 read_lock_bh(&idev->lock);
1248                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                         if (ifp->scope == IFA_LINK &&
1250                             !(ifp->flags & banned_flags)) {
1251                                 ipv6_addr_copy(addr, &ifp->addr);
1252                                 err = 0;
1253                                 break;
1254                         }
1255                 }
1256                 read_unlock_bh(&idev->lock);
1257         }
1258         rcu_read_unlock();
1259         return err;
1260 }
1261
1262 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 {
1264         int cnt = 0;
1265         struct inet6_ifaddr *ifp;
1266
1267         read_lock_bh(&idev->lock);
1268         list_for_each_entry(ifp, &idev->addr_list, if_list)
1269                 cnt++;
1270         read_unlock_bh(&idev->lock);
1271         return cnt;
1272 }
1273
1274 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275                   struct net_device *dev, int strict)
1276 {
1277         struct inet6_ifaddr *ifp;
1278         struct hlist_node *node;
1279         unsigned int hash = ipv6_addr_hash(addr);
1280
1281         rcu_read_lock_bh();
1282         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283                 if (!net_eq(dev_net(ifp->idev->dev), net))
1284                         continue;
1285                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286                     !(ifp->flags&IFA_F_TENTATIVE) &&
1287                     (dev == NULL || ifp->idev->dev == dev ||
1288                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289                         rcu_read_unlock_bh();
1290                         return 1;
1291                 }
1292         }
1293
1294         rcu_read_unlock_bh();
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300                                struct net_device *dev)
1301 {
1302         unsigned int hash = ipv6_addr_hash(addr);
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305
1306         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 return true;
1312                 }
1313         }
1314         return false;
1315 }
1316
1317 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr *ifp, *result = NULL;
1346         unsigned int hash = ipv6_addr_hash(addr);
1347         struct hlist_node *node;
1348
1349         rcu_read_lock_bh();
1350         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 result = ifp;
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return result;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401 {
1402         int err = -ENOENT;
1403
1404         spin_lock(&ifp->state_lock);
1405         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407                 err = 0;
1408         }
1409         spin_unlock(&ifp->state_lock);
1410
1411         return err;
1412 }
1413
1414 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415 {
1416         struct inet6_dev *idev = ifp->idev;
1417
1418         if (addrconf_dad_end(ifp)) {
1419                 in6_ifa_put(ifp);
1420                 return;
1421         }
1422
1423         if (net_ratelimit())
1424                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425                         ifp->idev->dev->name, &ifp->addr);
1426
1427         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428                 struct in6_addr addr;
1429
1430                 addr.s6_addr32[0] = htonl(0xfe800000);
1431                 addr.s6_addr32[1] = 0;
1432
1433                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434                     ipv6_addr_equal(&ifp->addr, &addr)) {
1435                         /* DAD failed for link-local based on MAC address */
1436                         idev->cnf.disable_ipv6 = 1;
1437
1438                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439                                 ifp->idev->dev->name);
1440                 }
1441         }
1442
1443         addrconf_dad_stop(ifp, 1);
1444 }
1445
1446 /* Join to solicited addr multicast group. */
1447
1448 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449 {
1450         struct in6_addr maddr;
1451
1452         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453                 return;
1454
1455         addrconf_addr_solict_mult(addr, &maddr);
1456         ipv6_dev_mc_inc(dev, &maddr);
1457 }
1458
1459 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         __ipv6_dev_mc_dec(idev, &maddr);
1468 }
1469
1470 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471 {
1472         struct in6_addr addr;
1473         if (ifp->prefix_len == 127) /* RFC 6164 */
1474                 return;
1475         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1476         if (ipv6_addr_any(&addr))
1477                 return;
1478         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1479 }
1480
1481 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1482 {
1483         struct in6_addr addr;
1484         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1485         if (ipv6_addr_any(&addr))
1486                 return;
1487         __ipv6_dev_ac_dec(ifp->idev, &addr);
1488 }
1489
1490 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1491 {
1492         if (dev->addr_len != ETH_ALEN)
1493                 return -1;
1494         memcpy(eui, dev->dev_addr, 3);
1495         memcpy(eui + 5, dev->dev_addr + 3, 3);
1496
1497         /*
1498          * The zSeries OSA network cards can be shared among various
1499          * OS instances, but the OSA cards have only one MAC address.
1500          * This leads to duplicate address conflicts in conjunction
1501          * with IPv6 if more than one instance uses the same card.
1502          *
1503          * The driver for these cards can deliver a unique 16-bit
1504          * identifier for each instance sharing the same card.  It is
1505          * placed instead of 0xFFFE in the interface identifier.  The
1506          * "u" bit of the interface identifier is not inverted in this
1507          * case.  Hence the resulting interface identifier has local
1508          * scope according to RFC2373.
1509          */
1510         if (dev->dev_id) {
1511                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1512                 eui[4] = dev->dev_id & 0xFF;
1513         } else {
1514                 eui[3] = 0xFF;
1515                 eui[4] = 0xFE;
1516                 eui[0] ^= 2;
1517         }
1518         return 0;
1519 }
1520
1521 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1522 {
1523         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1524         if (dev->addr_len != ARCNET_ALEN)
1525                 return -1;
1526         memset(eui, 0, 7);
1527         eui[7] = *(u8*)dev->dev_addr;
1528         return 0;
1529 }
1530
1531 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1532 {
1533         if (dev->addr_len != INFINIBAND_ALEN)
1534                 return -1;
1535         memcpy(eui, dev->dev_addr + 12, 8);
1536         eui[0] |= 2;
1537         return 0;
1538 }
1539
1540 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1541 {
1542         if (addr == 0)
1543                 return -1;
1544         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1545                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1546                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1547                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1548                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1549                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1550         eui[1] = 0;
1551         eui[2] = 0x5E;
1552         eui[3] = 0xFE;
1553         memcpy(eui + 4, &addr, 4);
1554         return 0;
1555 }
1556
1557 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1558 {
1559         if (dev->priv_flags & IFF_ISATAP)
1560                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1561         return -1;
1562 }
1563
1564 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1565 {
1566         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1567 }
1568
1569 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1570 {
1571         switch (dev->type) {
1572         case ARPHRD_ETHER:
1573         case ARPHRD_FDDI:
1574         case ARPHRD_IEEE802_TR:
1575                 return addrconf_ifid_eui48(eui, dev);
1576         case ARPHRD_ARCNET:
1577                 return addrconf_ifid_arcnet(eui, dev);
1578         case ARPHRD_INFINIBAND:
1579                 return addrconf_ifid_infiniband(eui, dev);
1580         case ARPHRD_SIT:
1581                 return addrconf_ifid_sit(eui, dev);
1582         case ARPHRD_IPGRE:
1583                 return addrconf_ifid_gre(eui, dev);
1584         }
1585         return -1;
1586 }
1587
1588 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1589 {
1590         int err = -1;
1591         struct inet6_ifaddr *ifp;
1592
1593         read_lock_bh(&idev->lock);
1594         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1595                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1596                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1597                         err = 0;
1598                         break;
1599                 }
1600         }
1601         read_unlock_bh(&idev->lock);
1602         return err;
1603 }
1604
1605 #ifdef CONFIG_IPV6_PRIVACY
1606 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1607 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1608 {
1609 regen:
1610         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1611         idev->rndid[0] &= ~0x02;
1612
1613         /*
1614          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1615          * check if generated address is not inappropriate
1616          *
1617          *  - Reserved subnet anycast (RFC 2526)
1618          *      11111101 11....11 1xxxxxxx
1619          *  - ISATAP (RFC4214) 6.1
1620          *      00-00-5E-FE-xx-xx-xx-xx
1621          *  - value 0
1622          *  - XXX: already assigned to an address on the device
1623          */
1624         if (idev->rndid[0] == 0xfd &&
1625             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1626             (idev->rndid[7]&0x80))
1627                 goto regen;
1628         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1629                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1630                         goto regen;
1631                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1632                         goto regen;
1633         }
1634
1635         return 0;
1636 }
1637
1638 static void ipv6_regen_rndid(unsigned long data)
1639 {
1640         struct inet6_dev *idev = (struct inet6_dev *) data;
1641         unsigned long expires;
1642
1643         rcu_read_lock_bh();
1644         write_lock_bh(&idev->lock);
1645
1646         if (idev->dead)
1647                 goto out;
1648
1649         if (__ipv6_regen_rndid(idev) < 0)
1650                 goto out;
1651
1652         expires = jiffies +
1653                 idev->cnf.temp_prefered_lft * HZ -
1654                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1655                 idev->cnf.max_desync_factor * HZ;
1656         if (time_before(expires, jiffies)) {
1657                 printk(KERN_WARNING
1658                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1659                         idev->dev->name);
1660                 goto out;
1661         }
1662
1663         if (!mod_timer(&idev->regen_timer, expires))
1664                 in6_dev_hold(idev);
1665
1666 out:
1667         write_unlock_bh(&idev->lock);
1668         rcu_read_unlock_bh();
1669         in6_dev_put(idev);
1670 }
1671
1672 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1673         int ret = 0;
1674
1675         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1676                 ret = __ipv6_regen_rndid(idev);
1677         return ret;
1678 }
1679 #endif
1680
1681 /*
1682  *      Add prefix route.
1683  */
1684
1685 static void
1686 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1687                       unsigned long expires, u32 flags)
1688 {
1689         struct fib6_config cfg = {
1690                 .fc_table = RT6_TABLE_PREFIX,
1691                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1692                 .fc_ifindex = dev->ifindex,
1693                 .fc_expires = expires,
1694                 .fc_dst_len = plen,
1695                 .fc_flags = RTF_UP | flags,
1696                 .fc_nlinfo.nl_net = dev_net(dev),
1697                 .fc_protocol = RTPROT_KERNEL,
1698         };
1699
1700         ipv6_addr_copy(&cfg.fc_dst, pfx);
1701
1702         /* Prevent useless cloning on PtP SIT.
1703            This thing is done here expecting that the whole
1704            class of non-broadcast devices need not cloning.
1705          */
1706 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1707         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1708                 cfg.fc_flags |= RTF_NONEXTHOP;
1709 #endif
1710
1711         ip6_route_add(&cfg);
1712 }
1713
1714 /* Create "default" multicast route to the interface */
1715
1716 static void addrconf_add_mroute(struct net_device *dev)
1717 {
1718         struct fib6_config cfg = {
1719                 .fc_table = RT6_TABLE_LOCAL,
1720                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1721                 .fc_ifindex = dev->ifindex,
1722                 .fc_dst_len = 8,
1723                 .fc_flags = RTF_UP,
1724                 .fc_nlinfo.nl_net = dev_net(dev),
1725         };
1726
1727         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1728
1729         ip6_route_add(&cfg);
1730 }
1731
1732 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1733 static void sit_route_add(struct net_device *dev)
1734 {
1735         struct fib6_config cfg = {
1736                 .fc_table = RT6_TABLE_MAIN,
1737                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1738                 .fc_ifindex = dev->ifindex,
1739                 .fc_dst_len = 96,
1740                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1741                 .fc_nlinfo.nl_net = dev_net(dev),
1742         };
1743
1744         /* prefix length - 96 bits "::d.d.d.d" */
1745         ip6_route_add(&cfg);
1746 }
1747 #endif
1748
1749 static void addrconf_add_lroute(struct net_device *dev)
1750 {
1751         struct in6_addr addr;
1752
1753         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1754         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1755 }
1756
1757 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1758 {
1759         struct inet6_dev *idev;
1760
1761         ASSERT_RTNL();
1762
1763         idev = ipv6_find_idev(dev);
1764         if (!idev)
1765                 return ERR_PTR(-ENOBUFS);
1766
1767         if (idev->cnf.disable_ipv6)
1768                 return ERR_PTR(-EACCES);
1769
1770         /* Add default multicast route */
1771         addrconf_add_mroute(dev);
1772
1773         /* Add link local route */
1774         addrconf_add_lroute(dev);
1775         return idev;
1776 }
1777
1778 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1779 {
1780         struct prefix_info *pinfo;
1781         __u32 valid_lft;
1782         __u32 prefered_lft;
1783         int addr_type;
1784         struct inet6_dev *in6_dev;
1785         struct net *net = dev_net(dev);
1786
1787         pinfo = (struct prefix_info *) opt;
1788
1789         if (len < sizeof(struct prefix_info)) {
1790                 ADBG(("addrconf: prefix option too short\n"));
1791                 return;
1792         }
1793
1794         /*
1795          *      Validation checks ([ADDRCONF], page 19)
1796          */
1797
1798         addr_type = ipv6_addr_type(&pinfo->prefix);
1799
1800         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1801                 return;
1802
1803         valid_lft = ntohl(pinfo->valid);
1804         prefered_lft = ntohl(pinfo->prefered);
1805
1806         if (prefered_lft > valid_lft) {
1807                 if (net_ratelimit())
1808                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1809                 return;
1810         }
1811
1812         in6_dev = in6_dev_get(dev);
1813
1814         if (in6_dev == NULL) {
1815                 if (net_ratelimit())
1816                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1817                 return;
1818         }
1819
1820         /*
1821          *      Two things going on here:
1822          *      1) Add routes for on-link prefixes
1823          *      2) Configure prefixes with the auto flag set
1824          */
1825
1826         if (pinfo->onlink) {
1827                 struct rt6_info *rt;
1828                 unsigned long rt_expires;
1829
1830                 /* Avoid arithmetic overflow. Really, we could
1831                  * save rt_expires in seconds, likely valid_lft,
1832                  * but it would require division in fib gc, that it
1833                  * not good.
1834                  */
1835                 if (HZ > USER_HZ)
1836                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1837                 else
1838                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1839
1840                 if (addrconf_finite_timeout(rt_expires))
1841                         rt_expires *= HZ;
1842
1843                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1844                                 dev->ifindex, 1);
1845
1846                 if (rt && addrconf_is_prefix_route(rt)) {
1847                         /* Autoconf prefix route */
1848                         if (valid_lft == 0) {
1849                                 ip6_del_rt(rt);
1850                                 rt = NULL;
1851                         } else if (addrconf_finite_timeout(rt_expires)) {
1852                                 /* not infinity */
1853                                 rt->rt6i_expires = jiffies + rt_expires;
1854                                 rt->rt6i_flags |= RTF_EXPIRES;
1855                         } else {
1856                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1857                                 rt->rt6i_expires = 0;
1858                         }
1859                 } else if (valid_lft) {
1860                         clock_t expires = 0;
1861                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1862                         if (addrconf_finite_timeout(rt_expires)) {
1863                                 /* not infinity */
1864                                 flags |= RTF_EXPIRES;
1865                                 expires = jiffies_to_clock_t(rt_expires);
1866                         }
1867                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1868                                               dev, expires, flags);
1869                 }
1870                 if (rt)
1871                         dst_release(&rt->dst);
1872         }
1873
1874         /* Try to figure out our local address for this prefix */
1875
1876         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1877                 struct inet6_ifaddr * ifp;
1878                 struct in6_addr addr;
1879                 int create = 0, update_lft = 0;
1880
1881                 if (pinfo->prefix_len == 64) {
1882                         memcpy(&addr, &pinfo->prefix, 8);
1883                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1884                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1885                                 in6_dev_put(in6_dev);
1886                                 return;
1887                         }
1888                         goto ok;
1889                 }
1890                 if (net_ratelimit())
1891                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1892                                pinfo->prefix_len);
1893                 in6_dev_put(in6_dev);
1894                 return;
1895
1896 ok:
1897
1898                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1899
1900                 if (ifp == NULL && valid_lft) {
1901                         int max_addresses = in6_dev->cnf.max_addresses;
1902                         u32 addr_flags = 0;
1903
1904 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1905                         if (in6_dev->cnf.optimistic_dad &&
1906                             !net->ipv6.devconf_all->forwarding)
1907                                 addr_flags = IFA_F_OPTIMISTIC;
1908 #endif
1909
1910                         /* Do not allow to create too much of autoconfigured
1911                          * addresses; this would be too easy way to crash kernel.
1912                          */
1913                         if (!max_addresses ||
1914                             ipv6_count_addresses(in6_dev) < max_addresses)
1915                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1916                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1917                                                     addr_flags);
1918
1919                         if (!ifp || IS_ERR(ifp)) {
1920                                 in6_dev_put(in6_dev);
1921                                 return;
1922                         }
1923
1924                         update_lft = create = 1;
1925                         ifp->cstamp = jiffies;
1926                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1927                 }
1928
1929                 if (ifp) {
1930                         int flags;
1931                         unsigned long now;
1932 #ifdef CONFIG_IPV6_PRIVACY
1933                         struct inet6_ifaddr *ift;
1934 #endif
1935                         u32 stored_lft;
1936
1937                         /* update lifetime (RFC2462 5.5.3 e) */
1938                         spin_lock(&ifp->lock);
1939                         now = jiffies;
1940                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1941                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1942                         else
1943                                 stored_lft = 0;
1944                         if (!update_lft && stored_lft) {
1945                                 if (valid_lft > MIN_VALID_LIFETIME ||
1946                                     valid_lft > stored_lft)
1947                                         update_lft = 1;
1948                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1949                                         /* valid_lft <= stored_lft is always true */
1950                                         /*
1951                                          * RFC 4862 Section 5.5.3e:
1952                                          * "Note that the preferred lifetime of
1953                                          *  the corresponding address is always
1954                                          *  reset to the Preferred Lifetime in
1955                                          *  the received Prefix Information
1956                                          *  option, regardless of whether the
1957                                          *  valid lifetime is also reset or
1958                                          *  ignored."
1959                                          *
1960                                          *  So if the preferred lifetime in
1961                                          *  this advertisement is different
1962                                          *  than what we have stored, but the
1963                                          *  valid lifetime is invalid, just
1964                                          *  reset prefered_lft.
1965                                          *
1966                                          *  We must set the valid lifetime
1967                                          *  to the stored lifetime since we'll
1968                                          *  be updating the timestamp below,
1969                                          *  else we'll set it back to the
1970                                          *  minimum.
1971                                          */
1972                                         if (prefered_lft != ifp->prefered_lft) {
1973                                                 valid_lft = stored_lft;
1974                                                 update_lft = 1;
1975                                         }
1976                                 } else {
1977                                         valid_lft = MIN_VALID_LIFETIME;
1978                                         if (valid_lft < prefered_lft)
1979                                                 prefered_lft = valid_lft;
1980                                         update_lft = 1;
1981                                 }
1982                         }
1983
1984                         if (update_lft) {
1985                                 ifp->valid_lft = valid_lft;
1986                                 ifp->prefered_lft = prefered_lft;
1987                                 ifp->tstamp = now;
1988                                 flags = ifp->flags;
1989                                 ifp->flags &= ~IFA_F_DEPRECATED;
1990                                 spin_unlock(&ifp->lock);
1991
1992                                 if (!(flags&IFA_F_TENTATIVE))
1993                                         ipv6_ifa_notify(0, ifp);
1994                         } else
1995                                 spin_unlock(&ifp->lock);
1996
1997 #ifdef CONFIG_IPV6_PRIVACY
1998                         read_lock_bh(&in6_dev->lock);
1999                         /* update all temporary addresses in the list */
2000                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
2001                                 /*
2002                                  * When adjusting the lifetimes of an existing
2003                                  * temporary address, only lower the lifetimes.
2004                                  * Implementations must not increase the
2005                                  * lifetimes of an existing temporary address
2006                                  * when processing a Prefix Information Option.
2007                                  */
2008                                 if (ifp != ift->ifpub)
2009                                         continue;
2010
2011                                 spin_lock(&ift->lock);
2012                                 flags = ift->flags;
2013                                 if (ift->valid_lft > valid_lft &&
2014                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2015                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2016                                 if (ift->prefered_lft > prefered_lft &&
2017                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2018                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2019                                 spin_unlock(&ift->lock);
2020                                 if (!(flags&IFA_F_TENTATIVE))
2021                                         ipv6_ifa_notify(0, ift);
2022                         }
2023
2024                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2025                                 /*
2026                                  * When a new public address is created as described in [ADDRCONF],
2027                                  * also create a new temporary address. Also create a temporary
2028                                  * address if it's enabled but no temporary address currently exists.
2029                                  */
2030                                 read_unlock_bh(&in6_dev->lock);
2031                                 ipv6_create_tempaddr(ifp, NULL);
2032                         } else {
2033                                 read_unlock_bh(&in6_dev->lock);
2034                         }
2035 #endif
2036                         in6_ifa_put(ifp);
2037                         addrconf_verify(0);
2038                 }
2039         }
2040         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2041         in6_dev_put(in6_dev);
2042 }
2043
2044 /*
2045  *      Set destination address.
2046  *      Special case for SIT interfaces where we create a new "virtual"
2047  *      device.
2048  */
2049 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2050 {
2051         struct in6_ifreq ireq;
2052         struct net_device *dev;
2053         int err = -EINVAL;
2054
2055         rtnl_lock();
2056
2057         err = -EFAULT;
2058         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2059                 goto err_exit;
2060
2061         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2062
2063         err = -ENODEV;
2064         if (dev == NULL)
2065                 goto err_exit;
2066
2067 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2068         if (dev->type == ARPHRD_SIT) {
2069                 const struct net_device_ops *ops = dev->netdev_ops;
2070                 struct ifreq ifr;
2071                 struct ip_tunnel_parm p;
2072
2073                 err = -EADDRNOTAVAIL;
2074                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2075                         goto err_exit;
2076
2077                 memset(&p, 0, sizeof(p));
2078                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2079                 p.iph.saddr = 0;
2080                 p.iph.version = 4;
2081                 p.iph.ihl = 5;
2082                 p.iph.protocol = IPPROTO_IPV6;
2083                 p.iph.ttl = 64;
2084                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2085
2086                 if (ops->ndo_do_ioctl) {
2087                         mm_segment_t oldfs = get_fs();
2088
2089                         set_fs(KERNEL_DS);
2090                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2091                         set_fs(oldfs);
2092                 } else
2093                         err = -EOPNOTSUPP;
2094
2095                 if (err == 0) {
2096                         err = -ENOBUFS;
2097                         dev = __dev_get_by_name(net, p.name);
2098                         if (!dev)
2099                                 goto err_exit;
2100                         err = dev_open(dev);
2101                 }
2102         }
2103 #endif
2104
2105 err_exit:
2106         rtnl_unlock();
2107         return err;
2108 }
2109
2110 /*
2111  *      Manual configuration of address on an interface
2112  */
2113 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2114                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2115                           __u32 valid_lft)
2116 {
2117         struct inet6_ifaddr *ifp;
2118         struct inet6_dev *idev;
2119         struct net_device *dev;
2120         int scope;
2121         u32 flags;
2122         clock_t expires;
2123         unsigned long timeout;
2124
2125         ASSERT_RTNL();
2126
2127         if (plen > 128)
2128                 return -EINVAL;
2129
2130         /* check the lifetime */
2131         if (!valid_lft || prefered_lft > valid_lft)
2132                 return -EINVAL;
2133
2134         dev = __dev_get_by_index(net, ifindex);
2135         if (!dev)
2136                 return -ENODEV;
2137
2138         idev = addrconf_add_dev(dev);
2139         if (IS_ERR(idev))
2140                 return PTR_ERR(idev);
2141
2142         scope = ipv6_addr_scope(pfx);
2143
2144         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2145         if (addrconf_finite_timeout(timeout)) {
2146                 expires = jiffies_to_clock_t(timeout * HZ);
2147                 valid_lft = timeout;
2148                 flags = RTF_EXPIRES;
2149         } else {
2150                 expires = 0;
2151                 flags = 0;
2152                 ifa_flags |= IFA_F_PERMANENT;
2153         }
2154
2155         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2156         if (addrconf_finite_timeout(timeout)) {
2157                 if (timeout == 0)
2158                         ifa_flags |= IFA_F_DEPRECATED;
2159                 prefered_lft = timeout;
2160         }
2161
2162         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2163
2164         if (!IS_ERR(ifp)) {
2165                 spin_lock_bh(&ifp->lock);
2166                 ifp->valid_lft = valid_lft;
2167                 ifp->prefered_lft = prefered_lft;
2168                 ifp->tstamp = jiffies;
2169                 spin_unlock_bh(&ifp->lock);
2170
2171                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2172                                       expires, flags);
2173                 /*
2174                  * Note that section 3.1 of RFC 4429 indicates
2175                  * that the Optimistic flag should not be set for
2176                  * manually configured addresses
2177                  */
2178                 addrconf_dad_start(ifp, 0);
2179                 in6_ifa_put(ifp);
2180                 addrconf_verify(0);
2181                 return 0;
2182         }
2183
2184         return PTR_ERR(ifp);
2185 }
2186
2187 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2188                           unsigned int plen)
2189 {
2190         struct inet6_ifaddr *ifp;
2191         struct inet6_dev *idev;
2192         struct net_device *dev;
2193
2194         if (plen > 128)
2195                 return -EINVAL;
2196
2197         dev = __dev_get_by_index(net, ifindex);
2198         if (!dev)
2199                 return -ENODEV;
2200
2201         if ((idev = __in6_dev_get(dev)) == NULL)
2202                 return -ENXIO;
2203
2204         read_lock_bh(&idev->lock);
2205         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2206                 if (ifp->prefix_len == plen &&
2207                     ipv6_addr_equal(pfx, &ifp->addr)) {
2208                         in6_ifa_hold(ifp);
2209                         read_unlock_bh(&idev->lock);
2210
2211                         ipv6_del_addr(ifp);
2212
2213                         /* If the last address is deleted administratively,
2214                            disable IPv6 on this interface.
2215                          */
2216                         if (list_empty(&idev->addr_list))
2217                                 addrconf_ifdown(idev->dev, 1);
2218                         return 0;
2219                 }
2220         }
2221         read_unlock_bh(&idev->lock);
2222         return -EADDRNOTAVAIL;
2223 }
2224
2225
2226 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2227 {
2228         struct in6_ifreq ireq;
2229         int err;
2230
2231         if (!capable(CAP_NET_ADMIN))
2232                 return -EPERM;
2233
2234         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2235                 return -EFAULT;
2236
2237         rtnl_lock();
2238         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2239                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2240                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2241         rtnl_unlock();
2242         return err;
2243 }
2244
2245 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2246 {
2247         struct in6_ifreq ireq;
2248         int err;
2249
2250         if (!capable(CAP_NET_ADMIN))
2251                 return -EPERM;
2252
2253         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2254                 return -EFAULT;
2255
2256         rtnl_lock();
2257         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2258                              ireq.ifr6_prefixlen);
2259         rtnl_unlock();
2260         return err;
2261 }
2262
2263 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2264                      int plen, int scope)
2265 {
2266         struct inet6_ifaddr *ifp;
2267
2268         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2269         if (!IS_ERR(ifp)) {
2270                 spin_lock_bh(&ifp->lock);
2271                 ifp->flags &= ~IFA_F_TENTATIVE;
2272                 spin_unlock_bh(&ifp->lock);
2273                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2274                 in6_ifa_put(ifp);
2275         }
2276 }
2277
2278 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2279 static void sit_add_v4_addrs(struct inet6_dev *idev)
2280 {
2281         struct in6_addr addr;
2282         struct net_device *dev;
2283         struct net *net = dev_net(idev->dev);
2284         int scope;
2285
2286         ASSERT_RTNL();
2287
2288         memset(&addr, 0, sizeof(struct in6_addr));
2289         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2290
2291         if (idev->dev->flags&IFF_POINTOPOINT) {
2292                 addr.s6_addr32[0] = htonl(0xfe800000);
2293                 scope = IFA_LINK;
2294         } else {
2295                 scope = IPV6_ADDR_COMPATv4;
2296         }
2297
2298         if (addr.s6_addr32[3]) {
2299                 add_addr(idev, &addr, 128, scope);
2300                 return;
2301         }
2302
2303         for_each_netdev(net, dev) {
2304                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2305                 if (in_dev && (dev->flags & IFF_UP)) {
2306                         struct in_ifaddr * ifa;
2307
2308                         int flag = scope;
2309
2310                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2311                                 int plen;
2312
2313                                 addr.s6_addr32[3] = ifa->ifa_local;
2314
2315                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2316                                         continue;
2317                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2318                                         if (idev->dev->flags&IFF_POINTOPOINT)
2319                                                 continue;
2320                                         flag |= IFA_HOST;
2321                                 }
2322                                 if (idev->dev->flags&IFF_POINTOPOINT)
2323                                         plen = 64;
2324                                 else
2325                                         plen = 96;
2326
2327                                 add_addr(idev, &addr, plen, flag);
2328                         }
2329                 }
2330         }
2331 }
2332 #endif
2333
2334 static void init_loopback(struct net_device *dev)
2335 {
2336         struct inet6_dev  *idev;
2337
2338         /* ::1 */
2339
2340         ASSERT_RTNL();
2341
2342         if ((idev = ipv6_find_idev(dev)) == NULL) {
2343                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2344                 return;
2345         }
2346
2347         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2348 }
2349
2350 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2351 {
2352         struct inet6_ifaddr * ifp;
2353         u32 addr_flags = IFA_F_PERMANENT;
2354
2355 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2356         if (idev->cnf.optimistic_dad &&
2357             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2358                 addr_flags |= IFA_F_OPTIMISTIC;
2359 #endif
2360
2361
2362         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2363         if (!IS_ERR(ifp)) {
2364                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2365                 addrconf_dad_start(ifp, 0);
2366                 in6_ifa_put(ifp);
2367         }
2368 }
2369
2370 static void addrconf_dev_config(struct net_device *dev)
2371 {
2372         struct in6_addr addr;
2373         struct inet6_dev    * idev;
2374
2375         ASSERT_RTNL();
2376
2377         if ((dev->type != ARPHRD_ETHER) &&
2378             (dev->type != ARPHRD_FDDI) &&
2379             (dev->type != ARPHRD_IEEE802_TR) &&
2380             (dev->type != ARPHRD_ARCNET) &&
2381             (dev->type != ARPHRD_INFINIBAND)) {
2382                 /* Alas, we support only Ethernet autoconfiguration. */
2383                 return;
2384         }
2385
2386         idev = addrconf_add_dev(dev);
2387         if (IS_ERR(idev))
2388                 return;
2389
2390         memset(&addr, 0, sizeof(struct in6_addr));
2391         addr.s6_addr32[0] = htonl(0xFE800000);
2392
2393         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2394                 addrconf_add_linklocal(idev, &addr);
2395 }
2396
2397 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2398 static void addrconf_sit_config(struct net_device *dev)
2399 {
2400         struct inet6_dev *idev;
2401
2402         ASSERT_RTNL();
2403
2404         /*
2405          * Configure the tunnel with one of our IPv4
2406          * addresses... we should configure all of
2407          * our v4 addrs in the tunnel
2408          */
2409
2410         if ((idev = ipv6_find_idev(dev)) == NULL) {
2411                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2412                 return;
2413         }
2414
2415         if (dev->priv_flags & IFF_ISATAP) {
2416                 struct in6_addr addr;
2417
2418                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2419                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2420                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2421                         addrconf_add_linklocal(idev, &addr);
2422                 return;
2423         }
2424
2425         sit_add_v4_addrs(idev);
2426
2427         if (dev->flags&IFF_POINTOPOINT) {
2428                 addrconf_add_mroute(dev);
2429                 addrconf_add_lroute(dev);
2430         } else
2431                 sit_route_add(dev);
2432 }
2433 #endif
2434
2435 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2436 static void addrconf_gre_config(struct net_device *dev)
2437 {
2438         struct inet6_dev *idev;
2439         struct in6_addr addr;
2440
2441         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2442
2443         ASSERT_RTNL();
2444
2445         if ((idev = ipv6_find_idev(dev)) == NULL) {
2446                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2447                 return;
2448         }
2449
2450         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2451         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2452
2453         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2454                 addrconf_add_linklocal(idev, &addr);
2455 }
2456 #endif
2457
2458 static inline int
2459 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2460 {
2461         struct in6_addr lladdr;
2462
2463         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2464                 addrconf_add_linklocal(idev, &lladdr);
2465                 return 0;
2466         }
2467         return -1;
2468 }
2469
2470 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2471 {
2472         struct net_device *link_dev;
2473         struct net *net = dev_net(idev->dev);
2474
2475         /* first try to inherit the link-local address from the link device */
2476         if (idev->dev->iflink &&
2477             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2478                 if (!ipv6_inherit_linklocal(idev, link_dev))
2479                         return;
2480         }
2481         /* then try to inherit it from any device */
2482         for_each_netdev(net, link_dev) {
2483                 if (!ipv6_inherit_linklocal(idev, link_dev))
2484                         return;
2485         }
2486         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2487 }
2488
2489 /*
2490  * Autoconfigure tunnel with a link-local address so routing protocols,
2491  * DHCPv6, MLD etc. can be run over the virtual link
2492  */
2493
2494 static void addrconf_ip6_tnl_config(struct net_device *dev)
2495 {
2496         struct inet6_dev *idev;
2497
2498         ASSERT_RTNL();
2499
2500         idev = addrconf_add_dev(dev);
2501         if (IS_ERR(idev)) {
2502                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2503                 return;
2504         }
2505         ip6_tnl_add_linklocal(idev);
2506 }
2507
2508 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2509                            void * data)
2510 {
2511         struct net_device *dev = (struct net_device *) data;
2512         struct inet6_dev *idev = __in6_dev_get(dev);
2513         int run_pending = 0;
2514         int err;
2515
2516         switch (event) {
2517         case NETDEV_REGISTER:
2518                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2519                         idev = ipv6_add_dev(dev);
2520                         if (!idev)
2521                                 return notifier_from_errno(-ENOMEM);
2522                 }
2523                 break;
2524
2525         case NETDEV_UP:
2526         case NETDEV_CHANGE:
2527                 if (dev->flags & IFF_SLAVE)
2528                         break;
2529
2530                 if (event == NETDEV_UP) {
2531                         if (!addrconf_qdisc_ok(dev)) {
2532                                 /* device is not ready yet. */
2533                                 printk(KERN_INFO
2534                                         "ADDRCONF(NETDEV_UP): %s: "
2535                                         "link is not ready\n",
2536                                         dev->name);
2537                                 break;
2538                         }
2539
2540                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2541                                 idev = ipv6_add_dev(dev);
2542
2543                         if (idev) {
2544                                 idev->if_flags |= IF_READY;
2545                                 run_pending = 1;
2546                         }
2547                 } else {
2548                         if (!addrconf_qdisc_ok(dev)) {
2549                                 /* device is still not ready. */
2550                                 break;
2551                         }
2552
2553                         if (idev) {
2554                                 if (idev->if_flags & IF_READY)
2555                                         /* device is already configured. */
2556                                         break;
2557                                 idev->if_flags |= IF_READY;
2558                         }
2559
2560                         printk(KERN_INFO
2561                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2562                                         "link becomes ready\n",
2563                                         dev->name);
2564
2565                         run_pending = 1;
2566                 }
2567
2568                 switch (dev->type) {
2569 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2570                 case ARPHRD_SIT:
2571                         addrconf_sit_config(dev);
2572                         break;
2573 #endif
2574 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2575                 case ARPHRD_IPGRE:
2576                         addrconf_gre_config(dev);
2577                         break;
2578 #endif
2579                 case ARPHRD_TUNNEL6:
2580                         addrconf_ip6_tnl_config(dev);
2581                         break;
2582                 case ARPHRD_LOOPBACK:
2583                         init_loopback(dev);
2584                         break;
2585
2586                 default:
2587                         addrconf_dev_config(dev);
2588                         break;
2589                 }
2590
2591                 if (idev) {
2592                         if (run_pending)
2593                                 addrconf_dad_run(idev);
2594
2595                         /*
2596                          * If the MTU changed during the interface down,
2597                          * when the interface up, the changed MTU must be
2598                          * reflected in the idev as well as routers.
2599                          */
2600                         if (idev->cnf.mtu6 != dev->mtu &&
2601                             dev->mtu >= IPV6_MIN_MTU) {
2602                                 rt6_mtu_change(dev, dev->mtu);
2603                                 idev->cnf.mtu6 = dev->mtu;
2604                         }
2605                         idev->tstamp = jiffies;
2606                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2607
2608                         /*
2609                          * If the changed mtu during down is lower than
2610                          * IPV6_MIN_MTU stop IPv6 on this interface.
2611                          */
2612                         if (dev->mtu < IPV6_MIN_MTU)
2613                                 addrconf_ifdown(dev, 1);
2614                 }
2615                 break;
2616
2617         case NETDEV_CHANGEMTU:
2618                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2619                         rt6_mtu_change(dev, dev->mtu);
2620                         idev->cnf.mtu6 = dev->mtu;
2621                         break;
2622                 }
2623
2624                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2625                         idev = ipv6_add_dev(dev);
2626                         if (idev)
2627                                 break;
2628                 }
2629
2630                 /*
2631                  * MTU falled under IPV6_MIN_MTU.
2632                  * Stop IPv6 on this interface.
2633                  */
2634
2635         case NETDEV_DOWN:
2636         case NETDEV_UNREGISTER:
2637                 /*
2638                  *      Remove all addresses from this interface.
2639                  */
2640                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2641                 break;
2642
2643         case NETDEV_CHANGENAME:
2644                 if (idev) {
2645                         snmp6_unregister_dev(idev);
2646                         addrconf_sysctl_unregister(idev);
2647                         addrconf_sysctl_register(idev);
2648                         err = snmp6_register_dev(idev);
2649                         if (err)
2650                                 return notifier_from_errno(err);
2651                 }
2652                 break;
2653
2654         case NETDEV_PRE_TYPE_CHANGE:
2655         case NETDEV_POST_TYPE_CHANGE:
2656                 addrconf_type_change(dev, event);
2657                 break;
2658         }
2659
2660         return NOTIFY_OK;
2661 }
2662
2663 /*
2664  *      addrconf module should be notified of a device going up
2665  */
2666 static struct notifier_block ipv6_dev_notf = {
2667         .notifier_call = addrconf_notify,
2668 };
2669
2670 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2671 {
2672         struct inet6_dev *idev;
2673         ASSERT_RTNL();
2674
2675         idev = __in6_dev_get(dev);
2676
2677         if (event == NETDEV_POST_TYPE_CHANGE)
2678                 ipv6_mc_remap(idev);
2679         else if (event == NETDEV_PRE_TYPE_CHANGE)
2680                 ipv6_mc_unmap(idev);
2681 }
2682
2683 static int addrconf_ifdown(struct net_device *dev, int how)
2684 {
2685         struct net *net = dev_net(dev);
2686         struct inet6_dev *idev;
2687         struct inet6_ifaddr *ifa;
2688         int state, i;
2689
2690         ASSERT_RTNL();
2691
2692         rt6_ifdown(net, dev);
2693         neigh_ifdown(&nd_tbl, dev);
2694
2695         idev = __in6_dev_get(dev);
2696         if (idev == NULL)
2697                 return -ENODEV;
2698
2699         /*
2700          * Step 1: remove reference to ipv6 device from parent device.
2701          *         Do not dev_put!
2702          */
2703         if (how) {
2704                 idev->dead = 1;
2705
2706                 /* protected by rtnl_lock */
2707                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2708
2709                 /* Step 1.5: remove snmp6 entry */
2710                 snmp6_unregister_dev(idev);
2711
2712         }
2713
2714         /* Step 2: clear hash table */
2715         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2716                 struct hlist_head *h = &inet6_addr_lst[i];
2717                 struct hlist_node *n;
2718
2719                 spin_lock_bh(&addrconf_hash_lock);
2720         restart:
2721                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2722                         if (ifa->idev == idev) {
2723                                 hlist_del_init_rcu(&ifa->addr_lst);
2724                                 addrconf_del_timer(ifa);
2725                                 goto restart;
2726                         }
2727                 }
2728                 spin_unlock_bh(&addrconf_hash_lock);
2729         }
2730
2731         write_lock_bh(&idev->lock);
2732
2733         /* Step 2: clear flags for stateless addrconf */
2734         if (!how)
2735                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2736
2737 #ifdef CONFIG_IPV6_PRIVACY
2738         if (how && del_timer(&idev->regen_timer))
2739                 in6_dev_put(idev);
2740
2741         /* Step 3: clear tempaddr list */
2742         while (!list_empty(&idev->tempaddr_list)) {
2743                 ifa = list_first_entry(&idev->tempaddr_list,
2744                                        struct inet6_ifaddr, tmp_list);
2745                 list_del(&ifa->tmp_list);
2746                 write_unlock_bh(&idev->lock);
2747                 spin_lock_bh(&ifa->lock);
2748
2749                 if (ifa->ifpub) {
2750                         in6_ifa_put(ifa->ifpub);
2751                         ifa->ifpub = NULL;
2752                 }
2753                 spin_unlock_bh(&ifa->lock);
2754                 in6_ifa_put(ifa);
2755                 write_lock_bh(&idev->lock);
2756         }
2757 #endif
2758
2759         while (!list_empty(&idev->addr_list)) {
2760                 ifa = list_first_entry(&idev->addr_list,
2761                                        struct inet6_ifaddr, if_list);
2762                 addrconf_del_timer(ifa);
2763
2764                 list_del(&ifa->if_list);
2765
2766                 write_unlock_bh(&idev->lock);
2767
2768                 spin_lock_bh(&ifa->state_lock);
2769                 state = ifa->state;
2770                 ifa->state = INET6_IFADDR_STATE_DEAD;
2771                 spin_unlock_bh(&ifa->state_lock);
2772
2773                 if (state != INET6_IFADDR_STATE_DEAD) {
2774                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2775                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2776                 }
2777                 in6_ifa_put(ifa);
2778
2779                 write_lock_bh(&idev->lock);
2780         }
2781
2782         write_unlock_bh(&idev->lock);
2783
2784         /* Step 5: Discard multicast list */
2785         if (how)
2786                 ipv6_mc_destroy_dev(idev);
2787         else
2788                 ipv6_mc_down(idev);
2789
2790         idev->tstamp = jiffies;
2791
2792         /* Last: Shot the device (if unregistered) */
2793         if (how) {
2794                 addrconf_sysctl_unregister(idev);
2795                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2796                 neigh_ifdown(&nd_tbl, dev);
2797                 in6_dev_put(idev);
2798         }
2799         return 0;
2800 }
2801
2802 static void addrconf_rs_timer(unsigned long data)
2803 {
2804         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2805         struct inet6_dev *idev = ifp->idev;
2806
2807         read_lock(&idev->lock);
2808         if (idev->dead || !(idev->if_flags & IF_READY))
2809                 goto out;
2810
2811         if (idev->cnf.forwarding)
2812                 goto out;
2813
2814         /* Announcement received after solicitation was sent */
2815         if (idev->if_flags & IF_RA_RCVD)
2816                 goto out;
2817
2818         spin_lock(&ifp->lock);
2819         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2820                 /* The wait after the last probe can be shorter */
2821                 addrconf_mod_timer(ifp, AC_RS,
2822                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2823                                    idev->cnf.rtr_solicit_delay :
2824                                    idev->cnf.rtr_solicit_interval);
2825                 spin_unlock(&ifp->lock);
2826
2827                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2828         } else {
2829                 spin_unlock(&ifp->lock);
2830                 /*
2831                  * Note: we do not support deprecated "all on-link"
2832                  * assumption any longer.
2833                  */
2834                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2835                        idev->dev->name);
2836         }
2837
2838 out:
2839         read_unlock(&idev->lock);
2840         in6_ifa_put(ifp);
2841 }
2842
2843 /*
2844  *      Duplicate Address Detection
2845  */
2846 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2847 {
2848         unsigned long rand_num;
2849         struct inet6_dev *idev = ifp->idev;
2850
2851         if (ifp->flags & IFA_F_OPTIMISTIC)
2852                 rand_num = 0;
2853         else
2854                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2855
2856         ifp->probes = idev->cnf.dad_transmits;
2857         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2858 }
2859
2860 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2861 {
2862         struct inet6_dev *idev = ifp->idev;
2863         struct net_device *dev = idev->dev;
2864
2865         addrconf_join_solict(dev, &ifp->addr);
2866
2867         net_srandom(ifp->addr.s6_addr32[3]);
2868
2869         read_lock_bh(&idev->lock);
2870         spin_lock(&ifp->lock);
2871         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2872                 goto out;
2873
2874         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2875             idev->cnf.accept_dad < 1 ||
2876             !(ifp->flags&IFA_F_TENTATIVE) ||
2877             ifp->flags & IFA_F_NODAD) {
2878                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2879                 spin_unlock(&ifp->lock);
2880                 read_unlock_bh(&idev->lock);
2881
2882                 addrconf_dad_completed(ifp);
2883                 return;
2884         }
2885
2886         if (!(idev->if_flags & IF_READY)) {
2887                 spin_unlock(&ifp->lock);
2888                 read_unlock_bh(&idev->lock);
2889                 /*
2890                  * If the device is not ready:
2891                  * - keep it tentative if it is a permanent address.
2892                  * - otherwise, kill it.
2893                  */
2894                 in6_ifa_hold(ifp);
2895                 addrconf_dad_stop(ifp, 0);
2896                 return;
2897         }
2898
2899         /*
2900          * Optimistic nodes can start receiving
2901          * Frames right away
2902          */
2903         if (ifp->flags & IFA_F_OPTIMISTIC)
2904                 ip6_ins_rt(ifp->rt);
2905
2906         addrconf_dad_kick(ifp);
2907 out:
2908         spin_unlock(&ifp->lock);
2909         read_unlock_bh(&idev->lock);
2910 }
2911
2912 static void addrconf_dad_timer(unsigned long data)
2913 {
2914         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2915         struct inet6_dev *idev = ifp->idev;
2916         struct in6_addr mcaddr;
2917
2918         if (!ifp->probes && addrconf_dad_end(ifp))
2919                 goto out;
2920
2921         read_lock(&idev->lock);
2922         if (idev->dead || !(idev->if_flags & IF_READY)) {
2923                 read_unlock(&idev->lock);
2924                 goto out;
2925         }
2926
2927         spin_lock(&ifp->lock);
2928         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2929                 spin_unlock(&ifp->lock);
2930                 read_unlock(&idev->lock);
2931                 goto out;
2932         }
2933
2934         if (ifp->probes == 0) {
2935                 /*
2936                  * DAD was successful
2937                  */
2938
2939                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2940                 spin_unlock(&ifp->lock);
2941                 read_unlock(&idev->lock);
2942
2943                 addrconf_dad_completed(ifp);
2944
2945                 goto out;
2946         }
2947
2948         ifp->probes--;
2949         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2950         spin_unlock(&ifp->lock);
2951         read_unlock(&idev->lock);
2952
2953         /* send a neighbour solicitation for our addr */
2954         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2955         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2956 out:
2957         in6_ifa_put(ifp);
2958 }
2959
2960 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2961 {
2962         struct net_device *dev = ifp->idev->dev;
2963
2964         /*
2965          *      Configure the address for reception. Now it is valid.
2966          */
2967
2968         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2969
2970         /* If added prefix is link local and forwarding is off,
2971            start sending router solicitations.
2972          */
2973
2974         if ((ifp->idev->cnf.forwarding == 0 ||
2975              ifp->idev->cnf.forwarding == 2) &&
2976             ifp->idev->cnf.rtr_solicits > 0 &&
2977             (dev->flags&IFF_LOOPBACK) == 0 &&
2978             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2979                 /*
2980                  *      If a host as already performed a random delay
2981                  *      [...] as part of DAD [...] there is no need
2982                  *      to delay again before sending the first RS
2983                  */
2984                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2985
2986                 spin_lock_bh(&ifp->lock);
2987                 ifp->probes = 1;
2988                 ifp->idev->if_flags |= IF_RS_SENT;
2989                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2990                 spin_unlock_bh(&ifp->lock);
2991         }
2992 }
2993
2994 static void addrconf_dad_run(struct inet6_dev *idev)
2995 {
2996         struct inet6_ifaddr *ifp;
2997
2998         read_lock_bh(&idev->lock);
2999         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3000                 spin_lock(&ifp->lock);
3001                 if (ifp->flags & IFA_F_TENTATIVE &&
3002                     ifp->state == INET6_IFADDR_STATE_DAD)
3003                         addrconf_dad_kick(ifp);
3004                 spin_unlock(&ifp->lock);
3005         }
3006         read_unlock_bh(&idev->lock);
3007 }
3008
3009 #ifdef CONFIG_PROC_FS
3010 struct if6_iter_state {
3011         struct seq_net_private p;
3012         int bucket;
3013 };
3014
3015 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3016 {
3017         struct inet6_ifaddr *ifa = NULL;
3018         struct if6_iter_state *state = seq->private;
3019         struct net *net = seq_file_net(seq);
3020
3021         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3022                 struct hlist_node *n;
3023                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3024                                          addr_lst)
3025                         if (net_eq(dev_net(ifa->idev->dev), net))
3026                                 return ifa;
3027         }
3028         return NULL;
3029 }
3030
3031 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3032                                          struct inet6_ifaddr *ifa)
3033 {
3034         struct if6_iter_state *state = seq->private;
3035         struct net *net = seq_file_net(seq);
3036         struct hlist_node *n = &ifa->addr_lst;
3037
3038         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3039                 if (net_eq(dev_net(ifa->idev->dev), net))
3040                         return ifa;
3041
3042         while (++state->bucket < IN6_ADDR_HSIZE) {
3043                 hlist_for_each_entry_rcu_bh(ifa, n,
3044                                      &inet6_addr_lst[state->bucket], addr_lst) {
3045                         if (net_eq(dev_net(ifa->idev->dev), net))
3046                                 return ifa;
3047                 }
3048         }
3049
3050         return NULL;
3051 }
3052
3053 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3054 {
3055         struct inet6_ifaddr *ifa = if6_get_first(seq);
3056
3057         if (ifa)
3058                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3059                         --pos;
3060         return pos ? NULL : ifa;
3061 }
3062
3063 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3064         __acquires(rcu_bh)
3065 {
3066         rcu_read_lock_bh();
3067         return if6_get_idx(seq, *pos);
3068 }
3069
3070 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3071 {
3072         struct inet6_ifaddr *ifa;
3073
3074         ifa = if6_get_next(seq, v);
3075         ++*pos;
3076         return ifa;
3077 }
3078
3079 static void if6_seq_stop(struct seq_file *seq, void *v)
3080         __releases(rcu_bh)
3081 {
3082         rcu_read_unlock_bh();
3083 }
3084
3085 static int if6_seq_show(struct seq_file *seq, void *v)
3086 {
3087         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3088         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3089                    &ifp->addr,
3090                    ifp->idev->dev->ifindex,
3091                    ifp->prefix_len,
3092                    ifp->scope,
3093                    ifp->flags,
3094                    ifp->idev->dev->name);
3095         return 0;
3096 }
3097
3098 static const struct seq_operations if6_seq_ops = {
3099         .start  = if6_seq_start,
3100         .next   = if6_seq_next,
3101         .show   = if6_seq_show,
3102         .stop   = if6_seq_stop,
3103 };
3104
3105 static int if6_seq_open(struct inode *inode, struct file *file)
3106 {
3107         return seq_open_net(inode, file, &if6_seq_ops,
3108                             sizeof(struct if6_iter_state));
3109 }
3110
3111 static const struct file_operations if6_fops = {
3112         .owner          = THIS_MODULE,
3113         .open           = if6_seq_open,
3114         .read           = seq_read,
3115         .llseek         = seq_lseek,
3116         .release        = seq_release_net,
3117 };
3118
3119 static int __net_init if6_proc_net_init(struct net *net)
3120 {
3121         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3122                 return -ENOMEM;
3123         return 0;
3124 }
3125
3126 static void __net_exit if6_proc_net_exit(struct net *net)
3127 {
3128        proc_net_remove(net, "if_inet6");
3129 }
3130
3131 static struct pernet_operations if6_proc_net_ops = {
3132        .init = if6_proc_net_init,
3133        .exit = if6_proc_net_exit,
3134 };
3135
3136 int __init if6_proc_init(void)
3137 {
3138         return register_pernet_subsys(&if6_proc_net_ops);
3139 }
3140
3141 void if6_proc_exit(void)
3142 {
3143         unregister_pernet_subsys(&if6_proc_net_ops);
3144 }
3145 #endif  /* CONFIG_PROC_FS */
3146
3147 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3148 /* Check if address is a home address configured on any interface. */
3149 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3150 {
3151         int ret = 0;
3152         struct inet6_ifaddr *ifp = NULL;
3153         struct hlist_node *n;
3154         unsigned int hash = ipv6_addr_hash(addr);
3155
3156         rcu_read_lock_bh();
3157         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3158                 if (!net_eq(dev_net(ifp->idev->dev), net))
3159                         continue;
3160                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3161                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3162                         ret = 1;
3163                         break;
3164                 }
3165         }
3166         rcu_read_unlock_bh();
3167         return ret;
3168 }
3169 #endif
3170
3171 /*
3172  *      Periodic address status verification
3173  */
3174
3175 static void addrconf_verify(unsigned long foo)
3176 {
3177         unsigned long now, next, next_sec, next_sched;
3178         struct inet6_ifaddr *ifp;
3179         struct hlist_node *node;
3180         int i;
3181
3182         rcu_read_lock_bh();
3183         spin_lock(&addrconf_verify_lock);
3184         now = jiffies;
3185         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3186
3187         del_timer(&addr_chk_timer);
3188
3189         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3190 restart:
3191                 hlist_for_each_entry_rcu_bh(ifp, node,
3192                                          &inet6_addr_lst[i], addr_lst) {
3193                         unsigned long age;
3194
3195                         if (ifp->flags & IFA_F_PERMANENT)
3196                                 continue;
3197
3198                         spin_lock(&ifp->lock);
3199                         /* We try to batch several events at once. */
3200                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3201
3202                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3203                             age >= ifp->valid_lft) {
3204                                 spin_unlock(&ifp->lock);
3205                                 in6_ifa_hold(ifp);
3206                                 ipv6_del_addr(ifp);
3207                                 goto restart;
3208                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3209                                 spin_unlock(&ifp->lock);
3210                                 continue;
3211                         } else if (age >= ifp->prefered_lft) {
3212                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3213                                 int deprecate = 0;
3214
3215                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3216                                         deprecate = 1;
3217                                         ifp->flags |= IFA_F_DEPRECATED;
3218                                 }
3219
3220                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3221                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3222
3223                                 spin_unlock(&ifp->lock);
3224
3225                                 if (deprecate) {
3226                                         in6_ifa_hold(ifp);
3227
3228                                         ipv6_ifa_notify(0, ifp);
3229                                         in6_ifa_put(ifp);
3230                                         goto restart;
3231                                 }
3232 #ifdef CONFIG_IPV6_PRIVACY
3233                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3234                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3235                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3236                                         ifp->idev->cnf.dad_transmits *
3237                                         ifp->idev->nd_parms->retrans_time / HZ;
3238
3239                                 if (age >= ifp->prefered_lft - regen_advance) {
3240                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3241                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3242                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3243                                         if (!ifp->regen_count && ifpub) {
3244                                                 ifp->regen_count++;
3245                                                 in6_ifa_hold(ifp);
3246                                                 in6_ifa_hold(ifpub);
3247                                                 spin_unlock(&ifp->lock);
3248
3249                                                 spin_lock(&ifpub->lock);
3250                                                 ifpub->regen_count = 0;
3251                                                 spin_unlock(&ifpub->lock);
3252                                                 ipv6_create_tempaddr(ifpub, ifp);
3253                                                 in6_ifa_put(ifpub);
3254                                                 in6_ifa_put(ifp);
3255                                                 goto restart;
3256                                         }
3257                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3258                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3259                                 spin_unlock(&ifp->lock);
3260 #endif
3261                         } else {
3262                                 /* ifp->prefered_lft <= ifp->valid_lft */
3263                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3264                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3265                                 spin_unlock(&ifp->lock);
3266                         }
3267                 }
3268         }
3269
3270         next_sec = round_jiffies_up(next);
3271         next_sched = next;
3272
3273         /* If rounded timeout is accurate enough, accept it. */
3274         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3275                 next_sched = next_sec;
3276
3277         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3278         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3279                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3280
3281         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3282               now, next, next_sec, next_sched));
3283
3284         addr_chk_timer.expires = next_sched;
3285         add_timer(&addr_chk_timer);
3286         spin_unlock(&addrconf_verify_lock);
3287         rcu_read_unlock_bh();
3288 }
3289
3290 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3291 {
3292         struct in6_addr *pfx = NULL;
3293
3294         if (addr)
3295                 pfx = nla_data(addr);
3296
3297         if (local) {
3298                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3299                         pfx = NULL;
3300                 else
3301                         pfx = nla_data(local);
3302         }
3303
3304         return pfx;
3305 }
3306
3307 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3308         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3309         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3310         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3311 };
3312
3313 static int
3314 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3315 {
3316         struct net *net = sock_net(skb->sk);
3317         struct ifaddrmsg *ifm;
3318         struct nlattr *tb[IFA_MAX+1];
3319         struct in6_addr *pfx;
3320         int err;
3321
3322         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3323         if (err < 0)
3324                 return err;
3325
3326         ifm = nlmsg_data(nlh);
3327         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3328         if (pfx == NULL)
3329                 return -EINVAL;
3330
3331         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3332 }
3333
3334 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3335                              u32 prefered_lft, u32 valid_lft)
3336 {
3337         u32 flags;
3338         clock_t expires;
3339         unsigned long timeout;
3340
3341         if (!valid_lft || (prefered_lft > valid_lft))
3342                 return -EINVAL;
3343
3344         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3345         if (addrconf_finite_timeout(timeout)) {
3346                 expires = jiffies_to_clock_t(timeout * HZ);
3347                 valid_lft = timeout;
3348                 flags = RTF_EXPIRES;
3349         } else {
3350                 expires = 0;
3351                 flags = 0;
3352                 ifa_flags |= IFA_F_PERMANENT;
3353         }
3354
3355         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3356         if (addrconf_finite_timeout(timeout)) {
3357                 if (timeout == 0)
3358                         ifa_flags |= IFA_F_DEPRECATED;
3359                 prefered_lft = timeout;
3360         }
3361
3362         spin_lock_bh(&ifp->lock);
3363         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3364         ifp->tstamp = jiffies;
3365         ifp->valid_lft = valid_lft;
3366         ifp->prefered_lft = prefered_lft;
3367
3368         spin_unlock_bh(&ifp->lock);
3369         if (!(ifp->flags&IFA_F_TENTATIVE))
3370                 ipv6_ifa_notify(0, ifp);
3371
3372         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3373                               expires, flags);
3374         addrconf_verify(0);
3375
3376         return 0;
3377 }
3378
3379 static int
3380 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3381 {
3382         struct net *net = sock_net(skb->sk);
3383         struct ifaddrmsg *ifm;
3384         struct nlattr *tb[IFA_MAX+1];
3385         struct in6_addr *pfx;
3386         struct inet6_ifaddr *ifa;
3387         struct net_device *dev;
3388         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3389         u8 ifa_flags;
3390         int err;
3391
3392         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3393         if (err < 0)
3394                 return err;
3395
3396         ifm = nlmsg_data(nlh);
3397         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3398         if (pfx == NULL)
3399                 return -EINVAL;
3400
3401         if (tb[IFA_CACHEINFO]) {
3402                 struct ifa_cacheinfo *ci;
3403
3404                 ci = nla_data(tb[IFA_CACHEINFO]);
3405                 valid_lft = ci->ifa_valid;
3406                 preferred_lft = ci->ifa_prefered;
3407         } else {
3408                 preferred_lft = INFINITY_LIFE_TIME;
3409                 valid_lft = INFINITY_LIFE_TIME;
3410         }
3411
3412         dev =  __dev_get_by_index(net, ifm->ifa_index);
3413         if (dev == NULL)
3414                 return -ENODEV;
3415
3416         /* We ignore other flags so far. */
3417         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3418
3419         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3420         if (ifa == NULL) {
3421                 /*
3422                  * It would be best to check for !NLM_F_CREATE here but
3423                  * userspace alreay relies on not having to provide this.
3424                  */
3425                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3426                                       ifm->ifa_prefixlen, ifa_flags,
3427                                       preferred_lft, valid_lft);
3428         }
3429
3430         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3431             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3432                 err = -EEXIST;
3433         else
3434                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3435
3436         in6_ifa_put(ifa);
3437
3438         return err;
3439 }
3440
3441 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3442                           u8 scope, int ifindex)
3443 {
3444         struct ifaddrmsg *ifm;
3445
3446         ifm = nlmsg_data(nlh);
3447         ifm->ifa_family = AF_INET6;
3448         ifm->ifa_prefixlen = prefixlen;
3449         ifm->ifa_flags = flags;
3450         ifm->ifa_scope = scope;
3451         ifm->ifa_index = ifindex;
3452 }
3453
3454 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3455                          unsigned long tstamp, u32 preferred, u32 valid)
3456 {
3457         struct ifa_cacheinfo ci;
3458
3459         ci.cstamp = cstamp_delta(cstamp);
3460         ci.tstamp = cstamp_delta(tstamp);
3461         ci.ifa_prefered = preferred;
3462         ci.ifa_valid = valid;
3463
3464         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3465 }
3466
3467 static inline int rt_scope(int ifa_scope)
3468 {
3469         if (ifa_scope & IFA_HOST)
3470                 return RT_SCOPE_HOST;
3471         else if (ifa_scope & IFA_LINK)
3472                 return RT_SCOPE_LINK;
3473         else if (ifa_scope & IFA_SITE)
3474                 return RT_SCOPE_SITE;
3475         else
3476                 return RT_SCOPE_UNIVERSE;
3477 }
3478
3479 static inline int inet6_ifaddr_msgsize(void)
3480 {
3481         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3482                + nla_total_size(16) /* IFA_ADDRESS */
3483                + nla_total_size(sizeof(struct ifa_cacheinfo));
3484 }
3485
3486 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3487                              u32 pid, u32 seq, int event, unsigned int flags)
3488 {
3489         struct nlmsghdr  *nlh;
3490         u32 preferred, valid;
3491
3492         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3493         if (nlh == NULL)
3494                 return -EMSGSIZE;
3495
3496         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3497                       ifa->idev->dev->ifindex);
3498
3499         if (!(ifa->flags&IFA_F_PERMANENT)) {
3500                 preferred = ifa->prefered_lft;
3501                 valid = ifa->valid_lft;
3502                 if (preferred != INFINITY_LIFE_TIME) {
3503                         long tval = (jiffies - ifa->tstamp)/HZ;
3504                         if (preferred > tval)
3505                                 preferred -= tval;
3506                         else
3507                                 preferred = 0;
3508                         if (valid != INFINITY_LIFE_TIME) {
3509                                 if (valid > tval)
3510                                         valid -= tval;
3511                                 else
3512                                         valid = 0;
3513                         }
3514                 }
3515         } else {
3516                 preferred = INFINITY_LIFE_TIME;
3517                 valid = INFINITY_LIFE_TIME;
3518         }
3519
3520         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3521             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3522                 nlmsg_cancel(skb, nlh);
3523                 return -EMSGSIZE;
3524         }
3525
3526         return nlmsg_end(skb, nlh);
3527 }
3528
3529 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3530                                 u32 pid, u32 seq, int event, u16 flags)
3531 {
3532         struct nlmsghdr  *nlh;
3533         u8 scope = RT_SCOPE_UNIVERSE;
3534         int ifindex = ifmca->idev->dev->ifindex;
3535
3536         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3537                 scope = RT_SCOPE_SITE;
3538
3539         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3540         if (nlh == NULL)
3541                 return -EMSGSIZE;
3542
3543         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3544         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3545             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3546                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3547                 nlmsg_cancel(skb, nlh);
3548                 return -EMSGSIZE;
3549         }
3550
3551         return nlmsg_end(skb, nlh);
3552 }
3553
3554 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3555                                 u32 pid, u32 seq, int event, unsigned int flags)
3556 {
3557         struct nlmsghdr  *nlh;
3558         u8 scope = RT_SCOPE_UNIVERSE;
3559         int ifindex = ifaca->aca_idev->dev->ifindex;
3560
3561         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3562                 scope = RT_SCOPE_SITE;
3563
3564         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3565         if (nlh == NULL)
3566                 return -EMSGSIZE;
3567
3568         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3569         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3570             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3571                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3572                 nlmsg_cancel(skb, nlh);
3573                 return -EMSGSIZE;
3574         }
3575
3576         return nlmsg_end(skb, nlh);
3577 }
3578
3579 enum addr_type_t {
3580         UNICAST_ADDR,
3581         MULTICAST_ADDR,
3582         ANYCAST_ADDR,
3583 };
3584
3585 /* called with rcu_read_lock() */
3586 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3587                           struct netlink_callback *cb, enum addr_type_t type,
3588                           int s_ip_idx, int *p_ip_idx)
3589 {
3590         struct ifmcaddr6 *ifmca;
3591         struct ifacaddr6 *ifaca;
3592         int err = 1;
3593         int ip_idx = *p_ip_idx;
3594
3595         read_lock_bh(&idev->lock);
3596         switch (type) {
3597         case UNICAST_ADDR: {
3598                 struct inet6_ifaddr *ifa;
3599
3600                 /* unicast address incl. temp addr */
3601                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3602                         if (++ip_idx < s_ip_idx)
3603                                 continue;
3604                         err = inet6_fill_ifaddr(skb, ifa,
3605                                                 NETLINK_CB(cb->skb).pid,
3606                                                 cb->nlh->nlmsg_seq,
3607                                                 RTM_NEWADDR,
3608                                                 NLM_F_MULTI);
3609                         if (err <= 0)
3610                                 break;
3611                 }
3612                 break;
3613         }
3614         case MULTICAST_ADDR:
3615                 /* multicast address */
3616                 for (ifmca = idev->mc_list; ifmca;
3617                      ifmca = ifmca->next, ip_idx++) {
3618                         if (ip_idx < s_ip_idx)
3619                                 continue;
3620                         err = inet6_fill_ifmcaddr(skb, ifmca,
3621                                                   NETLINK_CB(cb->skb).pid,
3622                                                   cb->nlh->nlmsg_seq,
3623                                                   RTM_GETMULTICAST,
3624                                                   NLM_F_MULTI);
3625                         if (err <= 0)
3626                                 break;
3627                 }
3628                 break;
3629         case ANYCAST_ADDR:
3630                 /* anycast address */
3631                 for (ifaca = idev->ac_list; ifaca;
3632                      ifaca = ifaca->aca_next, ip_idx++) {
3633                         if (ip_idx < s_ip_idx)
3634                                 continue;
3635                         err = inet6_fill_ifacaddr(skb, ifaca,
3636                                                   NETLINK_CB(cb->skb).pid,
3637                                                   cb->nlh->nlmsg_seq,
3638                                                   RTM_GETANYCAST,
3639                                                   NLM_F_MULTI);
3640                         if (err <= 0)
3641                                 break;
3642                 }
3643                 break;
3644         default:
3645                 break;
3646         }
3647         read_unlock_bh(&idev->lock);
3648         *p_ip_idx = ip_idx;
3649         return err;
3650 }
3651
3652 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3653                            enum addr_type_t type)
3654 {
3655         struct net *net = sock_net(skb->sk);
3656         int h, s_h;
3657         int idx, ip_idx;
3658         int s_idx, s_ip_idx;
3659         struct net_device *dev;
3660         struct inet6_dev *idev;
3661         struct hlist_head *head;
3662         struct hlist_node *node;
3663
3664         s_h = cb->args[0];
3665         s_idx = idx = cb->args[1];
3666         s_ip_idx = ip_idx = cb->args[2];
3667
3668         rcu_read_lock();
3669         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3670                 idx = 0;
3671                 head = &net->dev_index_head[h];
3672                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3673                         if (idx < s_idx)
3674                                 goto cont;
3675                         if (h > s_h || idx > s_idx)
3676                                 s_ip_idx = 0;
3677                         ip_idx = 0;
3678                         idev = __in6_dev_get(dev);
3679                         if (!idev)
3680                                 goto cont;
3681
3682                         if (in6_dump_addrs(idev, skb, cb, type,
3683                                            s_ip_idx, &ip_idx) <= 0)
3684                                 goto done;
3685 cont:
3686                         idx++;
3687                 }
3688         }
3689 done:
3690         rcu_read_unlock();
3691         cb->args[0] = h;
3692         cb->args[1] = idx;
3693         cb->args[2] = ip_idx;
3694
3695         return skb->len;
3696 }
3697
3698 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3699 {
3700         enum addr_type_t type = UNICAST_ADDR;
3701
3702         return inet6_dump_addr(skb, cb, type);
3703 }
3704
3705 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3706 {
3707         enum addr_type_t type = MULTICAST_ADDR;
3708
3709         return inet6_dump_addr(skb, cb, type);
3710 }
3711
3712
3713 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3714 {
3715         enum addr_type_t type = ANYCAST_ADDR;
3716
3717         return inet6_dump_addr(skb, cb, type);
3718 }
3719
3720 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3721                              void *arg)
3722 {
3723         struct net *net = sock_net(in_skb->sk);
3724         struct ifaddrmsg *ifm;
3725         struct nlattr *tb[IFA_MAX+1];
3726         struct in6_addr *addr = NULL;
3727         struct net_device *dev = NULL;
3728         struct inet6_ifaddr *ifa;
3729         struct sk_buff *skb;
3730         int err;
3731
3732         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3733         if (err < 0)
3734                 goto errout;
3735
3736         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3737         if (addr == NULL) {
3738                 err = -EINVAL;
3739                 goto errout;
3740         }
3741
3742         ifm = nlmsg_data(nlh);
3743         if (ifm->ifa_index)
3744                 dev = __dev_get_by_index(net, ifm->ifa_index);
3745
3746         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3747         if (!ifa) {
3748                 err = -EADDRNOTAVAIL;
3749                 goto errout;
3750         }
3751
3752         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3753         if (!skb) {
3754                 err = -ENOBUFS;
3755                 goto errout_ifa;
3756         }
3757
3758         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3759                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3760         if (err < 0) {
3761                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3762                 WARN_ON(err == -EMSGSIZE);
3763                 kfree_skb(skb);
3764                 goto errout_ifa;
3765         }
3766         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3767 errout_ifa:
3768         in6_ifa_put(ifa);
3769 errout:
3770         return err;
3771 }
3772
3773 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3774 {
3775         struct sk_buff *skb;
3776         struct net *net = dev_net(ifa->idev->dev);
3777         int err = -ENOBUFS;
3778
3779         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3780         if (skb == NULL)
3781                 goto errout;
3782
3783         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3784         if (err < 0) {
3785                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3786                 WARN_ON(err == -EMSGSIZE);
3787                 kfree_skb(skb);
3788                 goto errout;
3789         }
3790         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3791         return;
3792 errout:
3793         if (err < 0)
3794                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3795 }
3796
3797 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3798                                 __s32 *array, int bytes)
3799 {
3800         BUG_ON(bytes < (DEVCONF_MAX * 4));
3801
3802         memset(array, 0, bytes);
3803         array[DEVCONF_FORWARDING] = cnf->forwarding;
3804         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3805         array[DEVCONF_MTU6] = cnf->mtu6;
3806         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3807         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3808         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3809         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3810         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3811         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3812                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3813         array[DEVCONF_RTR_SOLICIT_DELAY] =
3814                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3815         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3816 #ifdef CONFIG_IPV6_PRIVACY
3817         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3818         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3819         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3820         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3821         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3822 #endif
3823         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3824         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3825         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3826 #ifdef CONFIG_IPV6_ROUTER_PREF
3827         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3828         array[DEVCONF_RTR_PROBE_INTERVAL] =
3829                 jiffies_to_msecs(cnf->rtr_probe_interval);
3830 #ifdef CONFIG_IPV6_ROUTE_INFO
3831         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3832 #endif
3833 #endif
3834         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3835         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3836 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3837         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3838 #endif
3839 #ifdef CONFIG_IPV6_MROUTE
3840         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3841 #endif
3842         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3843         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3844         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3845 }
3846
3847 static inline size_t inet6_ifla6_size(void)
3848 {
3849         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3850              + nla_total_size(sizeof(struct ifla_cacheinfo))
3851              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3852              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3853              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3854 }
3855
3856 static inline size_t inet6_if_nlmsg_size(void)
3857 {
3858         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3859                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3860                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3861                + nla_total_size(4) /* IFLA_MTU */
3862                + nla_total_size(4) /* IFLA_LINK */
3863                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3864 }
3865
3866 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3867                                       int items, int bytes)
3868 {
3869         int i;
3870         int pad = bytes - sizeof(u64) * items;
3871         BUG_ON(pad < 0);
3872
3873         /* Use put_unaligned() because stats may not be aligned for u64. */
3874         put_unaligned(items, &stats[0]);
3875         for (i = 1; i < items; i++)
3876                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3877
3878         memset(&stats[items], 0, pad);
3879 }
3880
3881 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3882                                       int items, int bytes, size_t syncpoff)
3883 {
3884         int i;
3885         int pad = bytes - sizeof(u64) * items;
3886         BUG_ON(pad < 0);
3887
3888         /* Use put_unaligned() because stats may not be aligned for u64. */
3889         put_unaligned(items, &stats[0]);
3890         for (i = 1; i < items; i++)
3891                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3892
3893         memset(&stats[items], 0, pad);
3894 }
3895
3896 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3897                              int bytes)
3898 {
3899         switch (attrtype) {
3900         case IFLA_INET6_STATS:
3901                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3902                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3903                 break;
3904         case IFLA_INET6_ICMP6STATS:
3905                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3906                 break;
3907         }
3908 }
3909
3910 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3911 {
3912         struct nlattr *nla;
3913         struct ifla_cacheinfo ci;
3914
3915         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3916
3917         ci.max_reasm_len = IPV6_MAXPLEN;
3918         ci.tstamp = cstamp_delta(idev->tstamp);
3919         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3920         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3921         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3922
3923         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3924         if (nla == NULL)
3925                 goto nla_put_failure;
3926         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3927
3928         /* XXX - MC not implemented */
3929
3930         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3931         if (nla == NULL)
3932                 goto nla_put_failure;
3933         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3934
3935         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3936         if (nla == NULL)
3937                 goto nla_put_failure;
3938         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3939
3940         return 0;
3941
3942 nla_put_failure:
3943         return -EMSGSIZE;
3944 }
3945
3946 static size_t inet6_get_link_af_size(const struct net_device *dev)
3947 {
3948         if (!__in6_dev_get(dev))
3949                 return 0;
3950
3951         return inet6_ifla6_size();
3952 }
3953
3954 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3955 {
3956         struct inet6_dev *idev = __in6_dev_get(dev);
3957
3958         if (!idev)
3959                 return -ENODATA;
3960
3961         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3962                 return -EMSGSIZE;
3963
3964         return 0;
3965 }
3966
3967 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3968                              u32 pid, u32 seq, int event, unsigned int flags)
3969 {
3970         struct net_device *dev = idev->dev;
3971         struct ifinfomsg *hdr;
3972         struct nlmsghdr *nlh;
3973         void *protoinfo;
3974
3975         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3976         if (nlh == NULL)
3977                 return -EMSGSIZE;
3978
3979         hdr = nlmsg_data(nlh);
3980         hdr->ifi_family = AF_INET6;
3981         hdr->__ifi_pad = 0;
3982         hdr->ifi_type = dev->type;
3983         hdr->ifi_index = dev->ifindex;
3984         hdr->ifi_flags = dev_get_flags(dev);
3985         hdr->ifi_change = 0;
3986
3987         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3988
3989         if (dev->addr_len)
3990                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3991
3992         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3993         if (dev->ifindex != dev->iflink)
3994                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3995
3996         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3997         if (protoinfo == NULL)
3998                 goto nla_put_failure;
3999
4000         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4001                 goto nla_put_failure;
4002
4003         nla_nest_end(skb, protoinfo);
4004         return nlmsg_end(skb, nlh);
4005
4006 nla_put_failure:
4007         nlmsg_cancel(skb, nlh);
4008         return -EMSGSIZE;
4009 }
4010
4011 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4012 {
4013         struct net *net = sock_net(skb->sk);
4014         int h, s_h;
4015         int idx = 0, s_idx;
4016         struct net_device *dev;
4017         struct inet6_dev *idev;
4018         struct hlist_head *head;
4019         struct hlist_node *node;
4020
4021         s_h = cb->args[0];
4022         s_idx = cb->args[1];
4023
4024         rcu_read_lock();
4025         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4026                 idx = 0;
4027                 head = &net->dev_index_head[h];
4028                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4029                         if (idx < s_idx)
4030                                 goto cont;
4031                         idev = __in6_dev_get(dev);
4032                         if (!idev)
4033                                 goto cont;
4034                         if (inet6_fill_ifinfo(skb, idev,
4035                                               NETLINK_CB(cb->skb).pid,
4036                                               cb->nlh->nlmsg_seq,
4037                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4038                                 goto out;
4039 cont:
4040                         idx++;
4041                 }
4042         }
4043 out:
4044         rcu_read_unlock();
4045         cb->args[1] = idx;
4046         cb->args[0] = h;
4047
4048         return skb->len;
4049 }
4050
4051 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4052 {
4053         struct sk_buff *skb;
4054         struct net *net = dev_net(idev->dev);
4055         int err = -ENOBUFS;
4056
4057         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4058         if (skb == NULL)
4059                 goto errout;
4060
4061         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4062         if (err < 0) {
4063                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4064                 WARN_ON(err == -EMSGSIZE);
4065                 kfree_skb(skb);
4066                 goto errout;
4067         }
4068         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4069         return;
4070 errout:
4071         if (err < 0)
4072                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4073 }
4074
4075 static inline size_t inet6_prefix_nlmsg_size(void)
4076 {
4077         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4078                + nla_total_size(sizeof(struct in6_addr))
4079                + nla_total_size(sizeof(struct prefix_cacheinfo));
4080 }
4081
4082 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4083                              struct prefix_info *pinfo, u32 pid, u32 seq,
4084                              int event, unsigned int flags)
4085 {
4086         struct prefixmsg *pmsg;
4087         struct nlmsghdr *nlh;
4088         struct prefix_cacheinfo ci;
4089
4090         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4091         if (nlh == NULL)
4092                 return -EMSGSIZE;
4093
4094         pmsg = nlmsg_data(nlh);
4095         pmsg->prefix_family = AF_INET6;
4096         pmsg->prefix_pad1 = 0;
4097         pmsg->prefix_pad2 = 0;
4098         pmsg->prefix_ifindex = idev->dev->ifindex;
4099         pmsg->prefix_len = pinfo->prefix_len;
4100         pmsg->prefix_type = pinfo->type;
4101         pmsg->prefix_pad3 = 0;
4102         pmsg->prefix_flags = 0;
4103         if (pinfo->onlink)
4104                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4105         if (pinfo->autoconf)
4106                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4107
4108         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4109
4110         ci.preferred_time = ntohl(pinfo->prefered);
4111         ci.valid_time = ntohl(pinfo->valid);
4112         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4113
4114         return nlmsg_end(skb, nlh);
4115
4116 nla_put_failure:
4117         nlmsg_cancel(skb, nlh);
4118         return -EMSGSIZE;
4119 }
4120
4121 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4122                          struct prefix_info *pinfo)
4123 {
4124         struct sk_buff *skb;
4125         struct net *net = dev_net(idev->dev);
4126         int err = -ENOBUFS;
4127
4128         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4129         if (skb == NULL)
4130                 goto errout;
4131
4132         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4133         if (err < 0) {
4134                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4135                 WARN_ON(err == -EMSGSIZE);
4136                 kfree_skb(skb);
4137                 goto errout;
4138         }
4139         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4140         return;
4141 errout:
4142         if (err < 0)
4143                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4144 }
4145
4146 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4147 {
4148         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4149
4150         switch (event) {
4151         case RTM_NEWADDR:
4152                 /*
4153                  * If the address was optimistic
4154                  * we inserted the route at the start of
4155                  * our DAD process, so we don't need
4156                  * to do it again
4157                  */
4158                 if (!(ifp->rt->rt6i_node))
4159                         ip6_ins_rt(ifp->rt);
4160                 if (ifp->idev->cnf.forwarding)
4161                         addrconf_join_anycast(ifp);
4162                 break;
4163         case RTM_DELADDR:
4164                 if (ifp->idev->cnf.forwarding)
4165                         addrconf_leave_anycast(ifp);
4166                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4167                 dst_hold(&ifp->rt->dst);
4168
4169                 if (ip6_del_rt(ifp->rt))
4170                         dst_free(&ifp->rt->dst);
4171                 break;
4172         }
4173 }
4174
4175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4176 {
4177         rcu_read_lock_bh();
4178         if (likely(ifp->idev->dead == 0))
4179                 __ipv6_ifa_notify(event, ifp);
4180         rcu_read_unlock_bh();
4181 }
4182
4183 #ifdef CONFIG_SYSCTL
4184
4185 static
4186 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4187                            void __user *buffer, size_t *lenp, loff_t *ppos)
4188 {
4189         int *valp = ctl->data;
4190         int val = *valp;
4191         loff_t pos = *ppos;
4192         int ret;
4193
4194         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4195
4196         if (write)
4197                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4198         if (ret)
4199                 *ppos = pos;
4200         return ret;
4201 }
4202
4203 static void dev_disable_change(struct inet6_dev *idev)
4204 {
4205         if (!idev || !idev->dev)
4206                 return;
4207
4208         if (idev->cnf.disable_ipv6)
4209                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4210         else
4211                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4212 }
4213
4214 static void addrconf_disable_change(struct net *net, __s32 newf)
4215 {
4216         struct net_device *dev;
4217         struct inet6_dev *idev;
4218
4219         rcu_read_lock();
4220         for_each_netdev_rcu(net, dev) {
4221                 idev = __in6_dev_get(dev);
4222                 if (idev) {
4223                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4224                         idev->cnf.disable_ipv6 = newf;
4225                         if (changed)
4226                                 dev_disable_change(idev);
4227                 }
4228         }
4229         rcu_read_unlock();
4230 }
4231
4232 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4233 {
4234         struct net *net;
4235
4236         net = (struct net *)table->extra2;
4237
4238         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4239                 return 0;
4240
4241         if (!rtnl_trylock()) {
4242                 /* Restore the original values before restarting */
4243                 *p = old;
4244                 return restart_syscall();
4245         }
4246
4247         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4248                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4249                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4250                 addrconf_disable_change(net, newf);
4251         } else if ((!*p) ^ (!old))
4252                 dev_disable_change((struct inet6_dev *)table->extra1);
4253
4254         rtnl_unlock();
4255         return 0;
4256 }
4257
4258 static
4259 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4260                             void __user *buffer, size_t *lenp, loff_t *ppos)
4261 {
4262         int *valp = ctl->data;
4263         int val = *valp;
4264         loff_t pos = *ppos;
4265         int ret;
4266
4267         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4268
4269         if (write)
4270                 ret = addrconf_disable_ipv6(ctl, valp, val);
4271         if (ret)
4272                 *ppos = pos;
4273         return ret;
4274 }
4275
4276 static struct addrconf_sysctl_table
4277 {
4278         struct ctl_table_header *sysctl_header;
4279         ctl_table addrconf_vars[DEVCONF_MAX+1];
4280         char *dev_name;
4281 } addrconf_sysctl __read_mostly = {
4282         .sysctl_header = NULL,
4283         .addrconf_vars = {
4284                 {
4285                         .procname       = "forwarding",
4286                         .data           = &ipv6_devconf.forwarding,
4287                         .maxlen         = sizeof(int),
4288                         .mode           = 0644,
4289                         .proc_handler   = addrconf_sysctl_forward,
4290                 },
4291                 {
4292                         .procname       = "hop_limit",
4293                         .data           = &ipv6_devconf.hop_limit,
4294                         .maxlen         = sizeof(int),
4295                         .mode           = 0644,
4296                         .proc_handler   = proc_dointvec,
4297                 },
4298                 {
4299                         .procname       = "mtu",
4300                         .data           = &ipv6_devconf.mtu6,
4301                         .maxlen         = sizeof(int),
4302                         .mode           = 0644,
4303                         .proc_handler   = proc_dointvec,
4304                 },
4305                 {
4306                         .procname       = "accept_ra",
4307                         .data           = &ipv6_devconf.accept_ra,
4308                         .maxlen         = sizeof(int),
4309                         .mode           = 0644,
4310                         .proc_handler   = proc_dointvec,
4311                 },
4312                 {
4313                         .procname       = "accept_redirects",
4314                         .data           = &ipv6_devconf.accept_redirects,
4315                         .maxlen         = sizeof(int),
4316                         .mode           = 0644,
4317                         .proc_handler   = proc_dointvec,
4318                 },
4319                 {
4320                         .procname       = "autoconf",
4321                         .data           = &ipv6_devconf.autoconf,
4322                         .maxlen         = sizeof(int),
4323                         .mode           = 0644,
4324                         .proc_handler   = proc_dointvec,
4325                 },
4326                 {
4327                         .procname       = "dad_transmits",
4328                         .data           = &ipv6_devconf.dad_transmits,
4329                         .maxlen         = sizeof(int),
4330                         .mode           = 0644,
4331                         .proc_handler   = proc_dointvec,
4332                 },
4333                 {
4334                         .procname       = "router_solicitations",
4335                         .data           = &ipv6_devconf.rtr_solicits,
4336                         .maxlen         = sizeof(int),
4337                         .mode           = 0644,
4338                         .proc_handler   = proc_dointvec,
4339                 },
4340                 {
4341                         .procname       = "router_solicitation_interval",
4342                         .data           = &ipv6_devconf.rtr_solicit_interval,
4343                         .maxlen         = sizeof(int),
4344                         .mode           = 0644,
4345                         .proc_handler   = proc_dointvec_jiffies,
4346                 },
4347                 {
4348                         .procname       = "router_solicitation_delay",
4349                         .data           = &ipv6_devconf.rtr_solicit_delay,
4350                         .maxlen         = sizeof(int),
4351                         .mode           = 0644,
4352                         .proc_handler   = proc_dointvec_jiffies,
4353                 },
4354                 {
4355                         .procname       = "force_mld_version",
4356                         .data           = &ipv6_devconf.force_mld_version,
4357                         .maxlen         = sizeof(int),
4358                         .mode           = 0644,
4359                         .proc_handler   = proc_dointvec,
4360                 },
4361 #ifdef CONFIG_IPV6_PRIVACY
4362                 {
4363                         .procname       = "use_tempaddr",
4364                         .data           = &ipv6_devconf.use_tempaddr,
4365                         .maxlen         = sizeof(int),
4366                         .mode           = 0644,
4367                         .proc_handler   = proc_dointvec,
4368                 },
4369                 {
4370                         .procname       = "temp_valid_lft",
4371                         .data           = &ipv6_devconf.temp_valid_lft,
4372                         .maxlen         = sizeof(int),
4373                         .mode           = 0644,
4374                         .proc_handler   = proc_dointvec,
4375                 },
4376                 {
4377                         .procname       = "temp_prefered_lft",
4378                         .data           = &ipv6_devconf.temp_prefered_lft,
4379                         .maxlen         = sizeof(int),
4380                         .mode           = 0644,
4381                         .proc_handler   = proc_dointvec,
4382                 },
4383                 {
4384                         .procname       = "regen_max_retry",
4385                         .data           = &ipv6_devconf.regen_max_retry,
4386                         .maxlen         = sizeof(int),
4387                         .mode           = 0644,
4388                         .proc_handler   = proc_dointvec,
4389                 },
4390                 {
4391                         .procname       = "max_desync_factor",
4392                         .data           = &ipv6_devconf.max_desync_factor,
4393                         .maxlen         = sizeof(int),
4394                         .mode           = 0644,
4395                         .proc_handler   = proc_dointvec,
4396                 },
4397 #endif
4398                 {
4399                         .procname       = "max_addresses",
4400                         .data           = &ipv6_devconf.max_addresses,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "accept_ra_defrtr",
4407                         .data           = &ipv6_devconf.accept_ra_defrtr,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "accept_ra_pinfo",
4414                         .data           = &ipv6_devconf.accept_ra_pinfo,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419 #ifdef CONFIG_IPV6_ROUTER_PREF
4420                 {
4421                         .procname       = "accept_ra_rtr_pref",
4422                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4423                         .maxlen         = sizeof(int),
4424                         .mode           = 0644,
4425                         .proc_handler   = proc_dointvec,
4426                 },
4427                 {
4428                         .procname       = "router_probe_interval",
4429                         .data           = &ipv6_devconf.rtr_probe_interval,
4430                         .maxlen         = sizeof(int),
4431                         .mode           = 0644,
4432                         .proc_handler   = proc_dointvec_jiffies,
4433                 },
4434 #ifdef CONFIG_IPV6_ROUTE_INFO
4435                 {
4436                         .procname       = "accept_ra_rt_info_max_plen",
4437                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4438                         .maxlen         = sizeof(int),
4439                         .mode           = 0644,
4440                         .proc_handler   = proc_dointvec,
4441                 },
4442 #endif
4443 #endif
4444                 {
4445                         .procname       = "proxy_ndp",
4446                         .data           = &ipv6_devconf.proxy_ndp,
4447                         .maxlen         = sizeof(int),
4448                         .mode           = 0644,
4449                         .proc_handler   = proc_dointvec,
4450                 },
4451                 {
4452                         .procname       = "accept_source_route",
4453                         .data           = &ipv6_devconf.accept_source_route,
4454                         .maxlen         = sizeof(int),
4455                         .mode           = 0644,
4456                         .proc_handler   = proc_dointvec,
4457                 },
4458 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4459                 {
4460                         .procname       = "optimistic_dad",
4461                         .data           = &ipv6_devconf.optimistic_dad,
4462                         .maxlen         = sizeof(int),
4463                         .mode           = 0644,
4464                         .proc_handler   = proc_dointvec,
4465
4466                 },
4467 #endif
4468 #ifdef CONFIG_IPV6_MROUTE
4469                 {
4470                         .procname       = "mc_forwarding",
4471                         .data           = &ipv6_devconf.mc_forwarding,
4472                         .maxlen         = sizeof(int),
4473                         .mode           = 0444,
4474                         .proc_handler   = proc_dointvec,
4475                 },
4476 #endif
4477                 {
4478                         .procname       = "disable_ipv6",
4479                         .data           = &ipv6_devconf.disable_ipv6,
4480                         .maxlen         = sizeof(int),
4481                         .mode           = 0644,
4482                         .proc_handler   = addrconf_sysctl_disable,
4483                 },
4484                 {
4485                         .procname       = "accept_dad",
4486                         .data           = &ipv6_devconf.accept_dad,
4487                         .maxlen         = sizeof(int),
4488                         .mode           = 0644,
4489                         .proc_handler   = proc_dointvec,
4490                 },
4491                 {
4492                         .procname       = "force_tllao",
4493                         .data           = &ipv6_devconf.force_tllao,
4494                         .maxlen         = sizeof(int),
4495                         .mode           = 0644,
4496                         .proc_handler   = proc_dointvec
4497                 },
4498                 {
4499                         /* sentinel */
4500                 }
4501         },
4502 };
4503
4504 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4505                 struct inet6_dev *idev, struct ipv6_devconf *p)
4506 {
4507         int i;
4508         struct addrconf_sysctl_table *t;
4509
4510 #define ADDRCONF_CTL_PATH_DEV   3
4511
4512         struct ctl_path addrconf_ctl_path[] = {
4513                 { .procname = "net", },
4514                 { .procname = "ipv6", },
4515                 { .procname = "conf", },
4516                 { /* to be set */ },
4517                 { },
4518         };
4519
4520
4521         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4522         if (t == NULL)
4523                 goto out;
4524
4525         for (i = 0; t->addrconf_vars[i].data; i++) {
4526                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4527                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4528                 t->addrconf_vars[i].extra2 = net;
4529         }
4530
4531         /*
4532          * Make a copy of dev_name, because '.procname' is regarded as const
4533          * by sysctl and we wouldn't want anyone to change it under our feet
4534          * (see SIOCSIFNAME).
4535          */
4536         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4537         if (!t->dev_name)
4538                 goto free;
4539
4540         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4541
4542         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4543                         t->addrconf_vars);
4544         if (t->sysctl_header == NULL)
4545                 goto free_procname;
4546
4547         p->sysctl = t;
4548         return 0;
4549
4550 free_procname:
4551         kfree(t->dev_name);
4552 free:
4553         kfree(t);
4554 out:
4555         return -ENOBUFS;
4556 }
4557
4558 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4559 {
4560         struct addrconf_sysctl_table *t;
4561
4562         if (p->sysctl == NULL)
4563                 return;
4564
4565         t = p->sysctl;
4566         p->sysctl = NULL;
4567         unregister_net_sysctl_table(t->sysctl_header);
4568         kfree(t->dev_name);
4569         kfree(t);
4570 }
4571
4572 static void addrconf_sysctl_register(struct inet6_dev *idev)
4573 {
4574         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4575                               &ndisc_ifinfo_sysctl_change);
4576         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4577                                         idev, &idev->cnf);
4578 }
4579
4580 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4581 {
4582         __addrconf_sysctl_unregister(&idev->cnf);
4583         neigh_sysctl_unregister(idev->nd_parms);
4584 }
4585
4586
4587 #endif
4588
4589 static int __net_init addrconf_init_net(struct net *net)
4590 {
4591         int err;
4592         struct ipv6_devconf *all, *dflt;
4593
4594         err = -ENOMEM;
4595         all = &ipv6_devconf;
4596         dflt = &ipv6_devconf_dflt;
4597
4598         if (!net_eq(net, &init_net)) {
4599                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4600                 if (all == NULL)
4601                         goto err_alloc_all;
4602
4603                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4604                 if (dflt == NULL)
4605                         goto err_alloc_dflt;
4606         } else {
4607                 /* these will be inherited by all namespaces */
4608                 dflt->autoconf = ipv6_defaults.autoconf;
4609                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4610         }
4611
4612         net->ipv6.devconf_all = all;
4613         net->ipv6.devconf_dflt = dflt;
4614
4615 #ifdef CONFIG_SYSCTL
4616         err = __addrconf_sysctl_register(net, "all", NULL, all);
4617         if (err < 0)
4618                 goto err_reg_all;
4619
4620         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4621         if (err < 0)
4622                 goto err_reg_dflt;
4623 #endif
4624         return 0;
4625
4626 #ifdef CONFIG_SYSCTL
4627 err_reg_dflt:
4628         __addrconf_sysctl_unregister(all);
4629 err_reg_all:
4630         kfree(dflt);
4631 #endif
4632 err_alloc_dflt:
4633         kfree(all);
4634 err_alloc_all:
4635         return err;
4636 }
4637
4638 static void __net_exit addrconf_exit_net(struct net *net)
4639 {
4640 #ifdef CONFIG_SYSCTL
4641         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4642         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4643 #endif
4644         if (!net_eq(net, &init_net)) {
4645                 kfree(net->ipv6.devconf_dflt);
4646                 kfree(net->ipv6.devconf_all);
4647         }
4648 }
4649
4650 static struct pernet_operations addrconf_ops = {
4651         .init = addrconf_init_net,
4652         .exit = addrconf_exit_net,
4653 };
4654
4655 /*
4656  *      Device notifier
4657  */
4658
4659 int register_inet6addr_notifier(struct notifier_block *nb)
4660 {
4661         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4662 }
4663 EXPORT_SYMBOL(register_inet6addr_notifier);
4664
4665 int unregister_inet6addr_notifier(struct notifier_block *nb)
4666 {
4667         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4668 }
4669 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4670
4671 static struct rtnl_af_ops inet6_ops = {
4672         .family           = AF_INET6,
4673         .fill_link_af     = inet6_fill_link_af,
4674         .get_link_af_size = inet6_get_link_af_size,
4675 };
4676
4677 /*
4678  *      Init / cleanup code
4679  */
4680
4681 int __init addrconf_init(void)
4682 {
4683         int i, err;
4684
4685         err = ipv6_addr_label_init();
4686         if (err < 0) {
4687                 printk(KERN_CRIT "IPv6 Addrconf:"
4688                        " cannot initialize default policy table: %d.\n", err);
4689                 goto out;
4690         }
4691
4692         err = register_pernet_subsys(&addrconf_ops);
4693         if (err < 0)
4694                 goto out_addrlabel;
4695
4696         /* The addrconf netdev notifier requires that loopback_dev
4697          * has it's ipv6 private information allocated and setup
4698          * before it can bring up and give link-local addresses
4699          * to other devices which are up.
4700          *
4701          * Unfortunately, loopback_dev is not necessarily the first
4702          * entry in the global dev_base list of net devices.  In fact,
4703          * it is likely to be the very last entry on that list.
4704          * So this causes the notifier registry below to try and
4705          * give link-local addresses to all devices besides loopback_dev
4706          * first, then loopback_dev, which cases all the non-loopback_dev
4707          * devices to fail to get a link-local address.
4708          *
4709          * So, as a temporary fix, allocate the ipv6 structure for
4710          * loopback_dev first by hand.
4711          * Longer term, all of the dependencies ipv6 has upon the loopback
4712          * device and it being up should be removed.
4713          */
4714         rtnl_lock();
4715         if (!ipv6_add_dev(init_net.loopback_dev))
4716                 err = -ENOMEM;
4717         rtnl_unlock();
4718         if (err)
4719                 goto errlo;
4720
4721         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4722                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4723
4724         register_netdevice_notifier(&ipv6_dev_notf);
4725
4726         addrconf_verify(0);
4727
4728         err = rtnl_af_register(&inet6_ops);
4729         if (err < 0)
4730                 goto errout_af;
4731
4732         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4733                               NULL);
4734         if (err < 0)
4735                 goto errout;
4736
4737         /* Only the first call to __rtnl_register can fail */
4738         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4739         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4740         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4741                         inet6_dump_ifaddr, NULL);
4742         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4743                         inet6_dump_ifmcaddr, NULL);
4744         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4745                         inet6_dump_ifacaddr, NULL);
4746
4747         ipv6_addr_label_rtnl_register();
4748
4749         return 0;
4750 errout:
4751         rtnl_af_unregister(&inet6_ops);
4752 errout_af:
4753         unregister_netdevice_notifier(&ipv6_dev_notf);
4754 errlo:
4755         unregister_pernet_subsys(&addrconf_ops);
4756 out_addrlabel:
4757         ipv6_addr_label_cleanup();
4758 out:
4759         return err;
4760 }
4761
4762 void addrconf_cleanup(void)
4763 {
4764         struct net_device *dev;
4765         int i;
4766
4767         unregister_netdevice_notifier(&ipv6_dev_notf);
4768         unregister_pernet_subsys(&addrconf_ops);
4769         ipv6_addr_label_cleanup();
4770
4771         rtnl_lock();
4772
4773         __rtnl_af_unregister(&inet6_ops);
4774
4775         /* clean dev list */
4776         for_each_netdev(&init_net, dev) {
4777                 if (__in6_dev_get(dev) == NULL)
4778                         continue;
4779                 addrconf_ifdown(dev, 1);
4780         }
4781         addrconf_ifdown(init_net.loopback_dev, 2);
4782
4783         /*
4784          *      Check hash table.
4785          */
4786         spin_lock_bh(&addrconf_hash_lock);
4787         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4788                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4789         spin_unlock_bh(&addrconf_hash_lock);
4790
4791         del_timer(&addr_chk_timer);
4792         rtnl_unlock();
4793 }