net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         RCU_INIT_POINTER(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         return ndev;
 438 }
 439
 440 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 441 {
 442         struct inet6_dev *idev;
 443
 444         ASSERT_RTNL();
 445
 446         idev = __in6_dev_get(dev);
 447         if (!idev) {
 448                 idev = ipv6_add_dev(dev);
 449                 if (!idev)
 450                         return NULL;
 451         }
 452
 453         if (dev->flags&IFF_UP)
 454                 ipv6_mc_up(idev);
 455         return idev;
 456 }
 457
 458 #ifdef CONFIG_SYSCTL
 459 static void dev_forward_change(struct inet6_dev *idev)
 460 {
 461         struct net_device *dev;
 462         struct inet6_ifaddr *ifa;
 463
 464         if (!idev)
 465                 return;
 466         dev = idev->dev;
 467         if (idev->cnf.forwarding)
 468                 dev_disable_lro(dev);
 469         if (dev && (dev->flags & IFF_MULTICAST)) {
 470                 if (idev->cnf.forwarding)
 471                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 472                 else
 473                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 474         }
 475
 476         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 477                 if (ifa->flags&IFA_F_TENTATIVE)
 478                         continue;
 479                 if (idev->cnf.forwarding)
 480                         addrconf_join_anycast(ifa);
 481                 else
 482                         addrconf_leave_anycast(ifa);
 483         }
 484 }
 485
 486
 487 static void addrconf_forward_change(struct net *net, __s32 newf)
 488 {
 489         struct net_device *dev;
 490         struct inet6_dev *idev;
 491
 492         rcu_read_lock();
 493         for_each_netdev_rcu(net, dev) {
 494                 idev = __in6_dev_get(dev);
 495                 if (idev) {
 496                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 497                         idev->cnf.forwarding = newf;
 498                         if (changed)
 499                                 dev_forward_change(idev);
 500                 }
 501         }
 502         rcu_read_unlock();
 503 }
 504
 505 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 506 {
 507         struct net *net;
 508
 509         net = (struct net *)table->extra2;
 510         if (p == &net->ipv6.devconf_dflt->forwarding)
 511                 return 0;
 512
 513         if (!rtnl_trylock()) {
 514                 /* Restore the original values before restarting */
 515                 *p = old;
 516                 return restart_syscall();
 517         }
 518
 519         if (p == &net->ipv6.devconf_all->forwarding) {
 520                 __s32 newf = net->ipv6.devconf_all->forwarding;
 521                 net->ipv6.devconf_dflt->forwarding = newf;
 522                 addrconf_forward_change(net, newf);
 523         } else if ((!*p) ^ (!old))
 524                 dev_forward_change((struct inet6_dev *)table->extra1);
 525         rtnl_unlock();
 526
 527         if (*p)
 528                 rt6_purge_dflt_routers(net);
 529         return 1;
 530 }
 531 #endif
 532
 533 /* Nobody refers to this ifaddr, destroy it */
 534 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 535 {
 536         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 537
 538 #ifdef NET_REFCNT_DEBUG
 539         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 540 #endif
 541
 542         in6_dev_put(ifp->idev);
 543
 544         if (del_timer(&ifp->timer))
 545                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 546
 547         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 548                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 549                 return;
 550         }
 551         dst_release(&ifp->rt->dst);
 552
 553         kfree_rcu(ifp, rcu);
 554 }
 555
 556 static void
 557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 558 {
 559         struct list_head *p;
 560         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 561
 562         /*
 563          * Each device address list is sorted in order of scope -
 564          * global before linklocal.
 565          */
 566         list_for_each(p, &idev->addr_list) {
 567                 struct inet6_ifaddr *ifa
 568                         = list_entry(p, struct inet6_ifaddr, if_list);
 569                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 570                         break;
 571         }
 572
 573         list_add_tail(&ifp->if_list, p);
 574 }
 575
 576 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 577 {
 578         /*
 579          * We perform the hash function over the last 64 bits of the address
 580          * This will include the IEEE address token on links that support it.
 581          */
 582         return jhash_2words((__force u32)addr->s6_addr32[2],
 583                             (__force u32)addr->s6_addr32[3], 0)
 584                 & (IN6_ADDR_HSIZE - 1);
 585 }
 586
 587 /* On success it returns ifp with increased reference count */
 588
 589 static struct inet6_ifaddr *
 590 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 591               int scope, u32 flags)
 592 {
 593         struct inet6_ifaddr *ifa = NULL;
 594         struct rt6_info *rt;
 595         unsigned int hash;
 596         int err = 0;
 597         int addr_type = ipv6_addr_type(addr);
 598
 599         if (addr_type == IPV6_ADDR_ANY ||
 600             addr_type & IPV6_ADDR_MULTICAST ||
 601             (!(idev->dev->flags & IFF_LOOPBACK) &&
 602              addr_type & IPV6_ADDR_LOOPBACK))
 603                 return ERR_PTR(-EADDRNOTAVAIL);
 604
 605         rcu_read_lock_bh();
 606         if (idev->dead) {
 607                 err = -ENODEV;                  /*XXX*/
 608                 goto out2;
 609         }
 610
 611         if (idev->cnf.disable_ipv6) {
 612                 err = -EACCES;
 613                 goto out2;
 614         }
 615
 616         spin_lock(&addrconf_hash_lock);
 617
 618         /* Ignore adding duplicate addresses on an interface */
 619         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 620                 ADBG(("ipv6_add_addr: already assigned\n"));
 621                 err = -EEXIST;
 622                 goto out;
 623         }
 624
 625         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 626
 627         if (ifa == NULL) {
 628                 ADBG(("ipv6_add_addr: malloc failed\n"));
 629                 err = -ENOBUFS;
 630                 goto out;
 631         }
 632
 633         rt = addrconf_dst_alloc(idev, addr, 0);
 634         if (IS_ERR(rt)) {
 635                 err = PTR_ERR(rt);
 636                 goto out;
 637         }
 638
 639         ipv6_addr_copy(&ifa->addr, addr);
 640
 641         spin_lock_init(&ifa->lock);
 642         spin_lock_init(&ifa->state_lock);
 643         init_timer(&ifa->timer);
 644         INIT_HLIST_NODE(&ifa->addr_lst);
 645         ifa->timer.data = (unsigned long) ifa;
 646         ifa->scope = scope;
 647         ifa->prefix_len = pfxlen;
 648         ifa->flags = flags | IFA_F_TENTATIVE;
 649         ifa->cstamp = ifa->tstamp = jiffies;
 650
 651         ifa->rt = rt;
 652
 653         /*
 654          * part one of RFC 4429, section 3.3
 655          * We should not configure an address as
 656          * optimistic if we do not yet know the link
 657          * layer address of our nexhop router
 658          */
 659
 660         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 661                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 662
 663         ifa->idev = idev;
 664         in6_dev_hold(idev);
 665         /* For caller */
 666         in6_ifa_hold(ifa);
 667
 668         /* Add to big hash table */
 669         hash = ipv6_addr_hash(addr);
 670
 671         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 672         spin_unlock(&addrconf_hash_lock);
 673
 674         write_lock(&idev->lock);
 675         /* Add to inet6_dev unicast addr list. */
 676         ipv6_link_dev_addr(idev, ifa);
 677
 678 #ifdef CONFIG_IPV6_PRIVACY
 679         if (ifa->flags&IFA_F_TEMPORARY) {
 680                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 681                 in6_ifa_hold(ifa);
 682         }
 683 #endif
 684
 685         in6_ifa_hold(ifa);
 686         write_unlock(&idev->lock);
 687 out2:
 688         rcu_read_unlock_bh();
 689
 690         if (likely(err == 0))
 691                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 692         else {
 693                 kfree(ifa);
 694                 ifa = ERR_PTR(err);
 695         }
 696
 697         return ifa;
 698 out:
 699         spin_unlock(&addrconf_hash_lock);
 700         goto out2;
 701 }
 702
 703 /* This function wants to get referenced ifp and releases it before return */
 704
 705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 706 {
 707         struct inet6_ifaddr *ifa, *ifn;
 708         struct inet6_dev *idev = ifp->idev;
 709         int state;
 710         int deleted = 0, onlink = 0;
 711         unsigned long expires = jiffies;
 712
 713         spin_lock_bh(&ifp->state_lock);
 714         state = ifp->state;
 715         ifp->state = INET6_IFADDR_STATE_DEAD;
 716         spin_unlock_bh(&ifp->state_lock);
 717
 718         if (state == INET6_IFADDR_STATE_DEAD)
 719                 goto out;
 720
 721         spin_lock_bh(&addrconf_hash_lock);
 722         hlist_del_init_rcu(&ifp->addr_lst);
 723         spin_unlock_bh(&addrconf_hash_lock);
 724
 725         write_lock_bh(&idev->lock);
 726 #ifdef CONFIG_IPV6_PRIVACY
 727         if (ifp->flags&IFA_F_TEMPORARY) {
 728                 list_del(&ifp->tmp_list);
 729                 if (ifp->ifpub) {
 730                         in6_ifa_put(ifp->ifpub);
 731                         ifp->ifpub = NULL;
 732                 }
 733                 __in6_ifa_put(ifp);
 734         }
 735 #endif
 736
 737         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 738                 if (ifa == ifp) {
 739                         list_del_init(&ifp->if_list);
 740                         __in6_ifa_put(ifp);
 741
 742                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 743                                 break;
 744                         deleted = 1;
 745                         continue;
 746                 } else if (ifp->flags & IFA_F_PERMANENT) {
 747                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 748                                               ifp->prefix_len)) {
 749                                 if (ifa->flags & IFA_F_PERMANENT) {
 750                                         onlink = 1;
 751                                         if (deleted)
 752                                                 break;
 753                                 } else {
 754                                         unsigned long lifetime;
 755
 756                                         if (!onlink)
 757                                                 onlink = -1;
 758
 759                                         spin_lock(&ifa->lock);
 760
 761                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 762                                         /*
 763                                          * Note: Because this address is
 764                                          * not permanent, lifetime <
 765                                          * LONG_MAX / HZ here.
 766                                          */
 767                                         if (time_before(expires,
 768                                                         ifa->tstamp + lifetime * HZ))
 769                                                 expires = ifa->tstamp + lifetime * HZ;
 770                                         spin_unlock(&ifa->lock);
 771                                 }
 772                         }
 773                 }
 774         }
 775         write_unlock_bh(&idev->lock);
 776
 777         addrconf_del_timer(ifp);
 778
 779         ipv6_ifa_notify(RTM_DELADDR, ifp);
 780
 781         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 782
 783         /*
 784          * Purge or update corresponding prefix
 785          *
 786          * 1) we don't purge prefix here if address was not permanent.
 787          *    prefix is managed by its own lifetime.
 788          * 2) if there're no addresses, delete prefix.
 789          * 3) if there're still other permanent address(es),
 790          *    corresponding prefix is still permanent.
 791          * 4) otherwise, update prefix lifetime to the
 792          *    longest valid lifetime among the corresponding
 793          *    addresses on the device.
 794          *    Note: subsequent RA will update lifetime.
 795          *
 796          * --yoshfuji
 797          */
 798         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 799                 struct in6_addr prefix;
 800                 struct rt6_info *rt;
 801                 struct net *net = dev_net(ifp->idev->dev);
 802                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 803                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 804
 805                 if (rt && addrconf_is_prefix_route(rt)) {
 806                         if (onlink == 0) {
 807                                 ip6_del_rt(rt);
 808                                 rt = NULL;
 809                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 810                                 rt->rt6i_expires = expires;
 811                                 rt->rt6i_flags |= RTF_EXPIRES;
 812                         }
 813                 }
 814                 dst_release(&rt->dst);
 815         }
 816
 817         /* clean up prefsrc entries */
 818         rt6_remove_prefsrc(ifp);
 819 out:
 820         in6_ifa_put(ifp);
 821 }
 822
 823 #ifdef CONFIG_IPV6_PRIVACY
 824 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 825 {
 826         struct inet6_dev *idev = ifp->idev;
 827         struct in6_addr addr, *tmpaddr;
 828         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 829         unsigned long regen_advance;
 830         int tmp_plen;
 831         int ret = 0;
 832         int max_addresses;
 833         u32 addr_flags;
 834         unsigned long now = jiffies;
 835
 836         write_lock(&idev->lock);
 837         if (ift) {
 838                 spin_lock_bh(&ift->lock);
 839                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 840                 spin_unlock_bh(&ift->lock);
 841                 tmpaddr = &addr;
 842         } else {
 843                 tmpaddr = NULL;
 844         }
 845 retry:
 846         in6_dev_hold(idev);
 847         if (idev->cnf.use_tempaddr <= 0) {
 848                 write_unlock(&idev->lock);
 849                 printk(KERN_INFO
 850                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 851                 in6_dev_put(idev);
 852                 ret = -1;
 853                 goto out;
 854         }
 855         spin_lock_bh(&ifp->lock);
 856         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 857                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 858                 spin_unlock_bh(&ifp->lock);
 859                 write_unlock(&idev->lock);
 860                 printk(KERN_WARNING
 861                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         in6_ifa_hold(ifp);
 867         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 868         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 869                 spin_unlock_bh(&ifp->lock);
 870                 write_unlock(&idev->lock);
 871                 printk(KERN_WARNING
 872                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 873                 in6_ifa_put(ifp);
 874                 in6_dev_put(idev);
 875                 ret = -1;
 876                 goto out;
 877         }
 878         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 879         age = (now - ifp->tstamp) / HZ;
 880         tmp_valid_lft = min_t(__u32,
 881                               ifp->valid_lft,
 882                               idev->cnf.temp_valid_lft + age);
 883         tmp_prefered_lft = min_t(__u32,
 884                                  ifp->prefered_lft,
 885                                  idev->cnf.temp_prefered_lft + age -
 886                                  idev->cnf.max_desync_factor);
 887         tmp_plen = ifp->prefix_len;
 888         max_addresses = idev->cnf.max_addresses;
 889         tmp_tstamp = ifp->tstamp;
 890         spin_unlock_bh(&ifp->lock);
 891
 892         regen_advance = idev->cnf.regen_max_retry *
 893                         idev->cnf.dad_transmits *
 894                         idev->nd_parms->retrans_time / HZ;
 895         write_unlock(&idev->lock);
 896
 897         /* A temporary address is created only if this calculated Preferred
 898          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 899          * an implementation must not create a temporary address with a zero
 900          * Preferred Lifetime.
 901          */
 902         if (tmp_prefered_lft <= regen_advance) {
 903                 in6_ifa_put(ifp);
 904                 in6_dev_put(idev);
 905                 ret = -1;
 906                 goto out;
 907         }
 908
 909         addr_flags = IFA_F_TEMPORARY;
 910         /* set in addrconf_prefix_rcv() */
 911         if (ifp->flags & IFA_F_OPTIMISTIC)
 912                 addr_flags |= IFA_F_OPTIMISTIC;
 913
 914         ift = !max_addresses ||
 915               ipv6_count_addresses(idev) < max_addresses ?
 916                 ipv6_add_addr(idev, &addr, tmp_plen,
 917                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 918                               addr_flags) : NULL;
 919         if (!ift || IS_ERR(ift)) {
 920                 in6_ifa_put(ifp);
 921                 in6_dev_put(idev);
 922                 printk(KERN_INFO
 923                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 924                 tmpaddr = &addr;
 925                 write_lock(&idev->lock);
 926                 goto retry;
 927         }
 928
 929         spin_lock_bh(&ift->lock);
 930         ift->ifpub = ifp;
 931         ift->valid_lft = tmp_valid_lft;
 932         ift->prefered_lft = tmp_prefered_lft;
 933         ift->cstamp = now;
 934         ift->tstamp = tmp_tstamp;
 935         spin_unlock_bh(&ift->lock);
 936
 937         addrconf_dad_start(ift, 0);
 938         in6_ifa_put(ift);
 939         in6_dev_put(idev);
 940 out:
 941         return ret;
 942 }
 943 #endif
 944
 945 /*
 946  *      Choose an appropriate source address (RFC3484)
 947  */
 948 enum {
 949         IPV6_SADDR_RULE_INIT = 0,
 950         IPV6_SADDR_RULE_LOCAL,
 951         IPV6_SADDR_RULE_SCOPE,
 952         IPV6_SADDR_RULE_PREFERRED,
 953 #ifdef CONFIG_IPV6_MIP6
 954         IPV6_SADDR_RULE_HOA,
 955 #endif
 956         IPV6_SADDR_RULE_OIF,
 957         IPV6_SADDR_RULE_LABEL,
 958 #ifdef CONFIG_IPV6_PRIVACY
 959         IPV6_SADDR_RULE_PRIVACY,
 960 #endif
 961         IPV6_SADDR_RULE_ORCHID,
 962         IPV6_SADDR_RULE_PREFIX,
 963         IPV6_SADDR_RULE_MAX
 964 };
 965
 966 struct ipv6_saddr_score {
 967         int                     rule;
 968         int                     addr_type;
 969         struct inet6_ifaddr     *ifa;
 970         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 971         int                     scopedist;
 972         int                     matchlen;
 973 };
 974
 975 struct ipv6_saddr_dst {
 976         const struct in6_addr *addr;
 977         int ifindex;
 978         int scope;
 979         int label;
 980         unsigned int prefs;
 981 };
 982
 983 static inline int ipv6_saddr_preferred(int type)
 984 {
 985         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 986                 return 1;
 987         return 0;
 988 }
 989
 990 static int ipv6_get_saddr_eval(struct net *net,
 991                                struct ipv6_saddr_score *score,
 992                                struct ipv6_saddr_dst *dst,
 993                                int i)
 994 {
 995         int ret;
 996
 997         if (i <= score->rule) {
 998                 switch (i) {
 999                 case IPV6_SADDR_RULE_SCOPE:
1000                         ret = score->scopedist;
1001                         break;
1002                 case IPV6_SADDR_RULE_PREFIX:
1003                         ret = score->matchlen;
1004                         break;
1005                 default:
1006                         ret = !!test_bit(i, score->scorebits);
1007                 }
1008                 goto out;
1009         }
1010
1011         switch (i) {
1012         case IPV6_SADDR_RULE_INIT:
1013                 /* Rule 0: remember if hiscore is not ready yet */
1014                 ret = !!score->ifa;
1015                 break;
1016         case IPV6_SADDR_RULE_LOCAL:
1017                 /* Rule 1: Prefer same address */
1018                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1019                 break;
1020         case IPV6_SADDR_RULE_SCOPE:
1021                 /* Rule 2: Prefer appropriate scope
1022                  *
1023                  *      ret
1024                  *       ^
1025                  *    -1 |  d 15
1026                  *    ---+--+-+---> scope
1027                  *       |
1028                  *       |             d is scope of the destination.
1029                  *  B-d  |  \
1030                  *       |   \      <- smaller scope is better if
1031                  *  B-15 |    \        if scope is enough for destinaion.
1032                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1033                  * d-C-1 | /
1034                  *       |/         <- greater is better
1035                  *   -C  /             if scope is not enough for destination.
1036                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1037                  *
1038                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1039                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1040                  * Assume B = 0 and we get C > 29.
1041                  */
1042                 ret = __ipv6_addr_src_scope(score->addr_type);
1043                 if (ret >= dst->scope)
1044                         ret = -ret;
1045                 else
1046                         ret -= 128;     /* 30 is enough */
1047                 score->scopedist = ret;
1048                 break;
1049         case IPV6_SADDR_RULE_PREFERRED:
1050                 /* Rule 3: Avoid deprecated and optimistic addresses */
1051                 ret = ipv6_saddr_preferred(score->addr_type) ||
1052                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1053                 break;
1054 #ifdef CONFIG_IPV6_MIP6
1055         case IPV6_SADDR_RULE_HOA:
1056             {
1057                 /* Rule 4: Prefer home address */
1058                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1059                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1060                 break;
1061             }
1062 #endif
1063         case IPV6_SADDR_RULE_OIF:
1064                 /* Rule 5: Prefer outgoing interface */
1065                 ret = (!dst->ifindex ||
1066                        dst->ifindex == score->ifa->idev->dev->ifindex);
1067                 break;
1068         case IPV6_SADDR_RULE_LABEL:
1069                 /* Rule 6: Prefer matching label */
1070                 ret = ipv6_addr_label(net,
1071                                       &score->ifa->addr, score->addr_type,
1072                                       score->ifa->idev->dev->ifindex) == dst->label;
1073                 break;
1074 #ifdef CONFIG_IPV6_PRIVACY
1075         case IPV6_SADDR_RULE_PRIVACY:
1076             {
1077                 /* Rule 7: Prefer public address
1078                  * Note: prefer temporary address if use_tempaddr >= 2
1079                  */
1080                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1081                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1082                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1083                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1084                 break;
1085             }
1086 #endif
1087         case IPV6_SADDR_RULE_ORCHID:
1088                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1089                  *          non-ORCHID vs non-ORCHID
1090                  */
1091                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1092                         ipv6_addr_orchid(dst->addr));
1093                 break;
1094         case IPV6_SADDR_RULE_PREFIX:
1095                 /* Rule 8: Use longest matching prefix */
1096                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1097                                                        dst->addr);
1098                 break;
1099         default:
1100                 ret = 0;
1101         }
1102
1103         if (ret)
1104                 __set_bit(i, score->scorebits);
1105         score->rule = i;
1106 out:
1107         return ret;
1108 }
1109
1110 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1111                        const struct in6_addr *daddr, unsigned int prefs,
1112                        struct in6_addr *saddr)
1113 {
1114         struct ipv6_saddr_score scores[2],
1115                                 *score = &scores[0], *hiscore = &scores[1];
1116         struct ipv6_saddr_dst dst;
1117         struct net_device *dev;
1118         int dst_type;
1119
1120         dst_type = __ipv6_addr_type(daddr);
1121         dst.addr = daddr;
1122         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1123         dst.scope = __ipv6_addr_src_scope(dst_type);
1124         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1125         dst.prefs = prefs;
1126
1127         hiscore->rule = -1;
1128         hiscore->ifa = NULL;
1129
1130         rcu_read_lock();
1131
1132         for_each_netdev_rcu(net, dev) {
1133                 struct inet6_dev *idev;
1134
1135                 /* Candidate Source Address (section 4)
1136                  *  - multicast and link-local destination address,
1137                  *    the set of candidate source address MUST only
1138                  *    include addresses assigned to interfaces
1139                  *    belonging to the same link as the outgoing
1140                  *    interface.
1141                  * (- For site-local destination addresses, the
1142                  *    set of candidate source addresses MUST only
1143                  *    include addresses assigned to interfaces
1144                  *    belonging to the same site as the outgoing
1145                  *    interface.)
1146                  */
1147                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1148                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1149                     dst.ifindex && dev->ifindex != dst.ifindex)
1150                         continue;
1151
1152                 idev = __in6_dev_get(dev);
1153                 if (!idev)
1154                         continue;
1155
1156                 read_lock_bh(&idev->lock);
1157                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1158                         int i;
1159
1160                         /*
1161                          * - Tentative Address (RFC2462 section 5.4)
1162                          *  - A tentative address is not considered
1163                          *    "assigned to an interface" in the traditional
1164                          *    sense, unless it is also flagged as optimistic.
1165                          * - Candidate Source Address (section 4)
1166                          *  - In any case, anycast addresses, multicast
1167                          *    addresses, and the unspecified address MUST
1168                          *    NOT be included in a candidate set.
1169                          */
1170                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1171                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1172                                 continue;
1173
1174                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1175
1176                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1177                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1178                                 LIMIT_NETDEBUG(KERN_DEBUG
1179                                                "ADDRCONF: unspecified / multicast address "
1180                                                "assigned as unicast address on %s",
1181                                                dev->name);
1182                                 continue;
1183                         }
1184
1185                         score->rule = -1;
1186                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1187
1188                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1189                                 int minihiscore, miniscore;
1190
1191                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1192                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1193
1194                                 if (minihiscore > miniscore) {
1195                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1196                                             score->scopedist > 0) {
1197                                                 /*
1198                                                  * special case:
1199                                                  * each remaining entry
1200                                                  * has too small (not enough)
1201                                                  * scope, because ifa entries
1202                                                  * are sorted by their scope
1203                                                  * values.
1204                                                  */
1205                                                 goto try_nextdev;
1206                                         }
1207                                         break;
1208                                 } else if (minihiscore < miniscore) {
1209                                         if (hiscore->ifa)
1210                                                 in6_ifa_put(hiscore->ifa);
1211
1212                                         in6_ifa_hold(score->ifa);
1213
1214                                         swap(hiscore, score);
1215
1216                                         /* restore our iterator */
1217                                         score->ifa = hiscore->ifa;
1218
1219                                         break;
1220                                 }
1221                         }
1222                 }
1223 try_nextdev:
1224                 read_unlock_bh(&idev->lock);
1225         }
1226         rcu_read_unlock();
1227
1228         if (!hiscore->ifa)
1229                 return -EADDRNOTAVAIL;
1230
1231         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1232         in6_ifa_put(hiscore->ifa);
1233         return 0;
1234 }
1235 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1236
1237 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1238                     unsigned char banned_flags)
1239 {
1240         struct inet6_dev *idev;
1241         int err = -EADDRNOTAVAIL;
1242
1243         rcu_read_lock();
1244         idev = __in6_dev_get(dev);
1245         if (idev) {
1246                 struct inet6_ifaddr *ifp;
1247
1248                 read_lock_bh(&idev->lock);
1249                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1250                         if (ifp->scope == IFA_LINK &&
1251                             !(ifp->flags & banned_flags)) {
1252                                 ipv6_addr_copy(addr, &ifp->addr);
1253                                 err = 0;
1254                                 break;
1255                         }
1256                 }
1257                 read_unlock_bh(&idev->lock);
1258         }
1259         rcu_read_unlock();
1260         return err;
1261 }
1262
1263 static int ipv6_count_addresses(struct inet6_dev *idev)
1264 {
1265         int cnt = 0;
1266         struct inet6_ifaddr *ifp;
1267
1268         read_lock_bh(&idev->lock);
1269         list_for_each_entry(ifp, &idev->addr_list, if_list)
1270                 cnt++;
1271         read_unlock_bh(&idev->lock);
1272         return cnt;
1273 }
1274
1275 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1276                   struct net_device *dev, int strict)
1277 {
1278         struct inet6_ifaddr *ifp;
1279         struct hlist_node *node;
1280         unsigned int hash = ipv6_addr_hash(addr);
1281
1282         rcu_read_lock_bh();
1283         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1284                 if (!net_eq(dev_net(ifp->idev->dev), net))
1285                         continue;
1286                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1287                     !(ifp->flags&IFA_F_TENTATIVE) &&
1288                     (dev == NULL || ifp->idev->dev == dev ||
1289                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1290                         rcu_read_unlock_bh();
1291                         return 1;
1292                 }
1293         }
1294
1295         rcu_read_unlock_bh();
1296         return 0;
1297 }
1298 EXPORT_SYMBOL(ipv6_chk_addr);
1299
1300 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1301                                struct net_device *dev)
1302 {
1303         unsigned int hash = ipv6_addr_hash(addr);
1304         struct inet6_ifaddr *ifp;
1305         struct hlist_node *node;
1306
1307         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1308                 if (!net_eq(dev_net(ifp->idev->dev), net))
1309                         continue;
1310                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1311                         if (dev == NULL || ifp->idev->dev == dev)
1312                                 return true;
1313                 }
1314         }
1315         return false;
1316 }
1317
1318 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1319 {
1320         struct inet6_dev *idev;
1321         struct inet6_ifaddr *ifa;
1322         int     onlink;
1323
1324         onlink = 0;
1325         rcu_read_lock();
1326         idev = __in6_dev_get(dev);
1327         if (idev) {
1328                 read_lock_bh(&idev->lock);
1329                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1330                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1331                                                    ifa->prefix_len);
1332                         if (onlink)
1333                                 break;
1334                 }
1335                 read_unlock_bh(&idev->lock);
1336         }
1337         rcu_read_unlock();
1338         return onlink;
1339 }
1340
1341 EXPORT_SYMBOL(ipv6_chk_prefix);
1342
1343 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1344                                      struct net_device *dev, int strict)
1345 {
1346         struct inet6_ifaddr *ifp, *result = NULL;
1347         unsigned int hash = ipv6_addr_hash(addr);
1348         struct hlist_node *node;
1349
1350         rcu_read_lock_bh();
1351         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1352                 if (!net_eq(dev_net(ifp->idev->dev), net))
1353                         continue;
1354                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1355                         if (dev == NULL || ifp->idev->dev == dev ||
1356                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1357                                 result = ifp;
1358                                 in6_ifa_hold(ifp);
1359                                 break;
1360                         }
1361                 }
1362         }
1363         rcu_read_unlock_bh();
1364
1365         return result;
1366 }
1367
1368 /* Gets referenced address, destroys ifaddr */
1369
1370 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1371 {
1372         if (ifp->flags&IFA_F_PERMANENT) {
1373                 spin_lock_bh(&ifp->lock);
1374                 addrconf_del_timer(ifp);
1375                 ifp->flags |= IFA_F_TENTATIVE;
1376                 if (dad_failed)
1377                         ifp->flags |= IFA_F_DADFAILED;
1378                 spin_unlock_bh(&ifp->lock);
1379                 if (dad_failed)
1380                         ipv6_ifa_notify(0, ifp);
1381                 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383         } else if (ifp->flags&IFA_F_TEMPORARY) {
1384                 struct inet6_ifaddr *ifpub;
1385                 spin_lock_bh(&ifp->lock);
1386                 ifpub = ifp->ifpub;
1387                 if (ifpub) {
1388                         in6_ifa_hold(ifpub);
1389                         spin_unlock_bh(&ifp->lock);
1390                         ipv6_create_tempaddr(ifpub, ifp);
1391                         in6_ifa_put(ifpub);
1392                 } else {
1393                         spin_unlock_bh(&ifp->lock);
1394                 }
1395                 ipv6_del_addr(ifp);
1396 #endif
1397         } else
1398                 ipv6_del_addr(ifp);
1399 }
1400
1401 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1402 {
1403         int err = -ENOENT;
1404
1405         spin_lock(&ifp->state_lock);
1406         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1407                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1408                 err = 0;
1409         }
1410         spin_unlock(&ifp->state_lock);
1411
1412         return err;
1413 }
1414
1415 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1416 {
1417         struct inet6_dev *idev = ifp->idev;
1418
1419         if (addrconf_dad_end(ifp)) {
1420                 in6_ifa_put(ifp);
1421                 return;
1422         }
1423
1424         if (net_ratelimit())
1425                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1426                         ifp->idev->dev->name, &ifp->addr);
1427
1428         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1429                 struct in6_addr addr;
1430
1431                 addr.s6_addr32[0] = htonl(0xfe800000);
1432                 addr.s6_addr32[1] = 0;
1433
1434                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1435                     ipv6_addr_equal(&ifp->addr, &addr)) {
1436                         /* DAD failed for link-local based on MAC address */
1437                         idev->cnf.disable_ipv6 = 1;
1438
1439                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1440                                 ifp->idev->dev->name);
1441                 }
1442         }
1443
1444         addrconf_dad_stop(ifp, 1);
1445 }
1446
1447 /* Join to solicited addr multicast group. */
1448
1449 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1450 {
1451         struct in6_addr maddr;
1452
1453         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1454                 return;
1455
1456         addrconf_addr_solict_mult(addr, &maddr);
1457         ipv6_dev_mc_inc(dev, &maddr);
1458 }
1459
1460 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1461 {
1462         struct in6_addr maddr;
1463
1464         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1465                 return;
1466
1467         addrconf_addr_solict_mult(addr, &maddr);
1468         __ipv6_dev_mc_dec(idev, &maddr);
1469 }
1470
1471 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1472 {
1473         struct in6_addr addr;
1474         if (ifp->prefix_len == 127) /* RFC 6164 */
1475                 return;
1476         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1477         if (ipv6_addr_any(&addr))
1478                 return;
1479         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1480 }
1481
1482 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1483 {
1484         struct in6_addr addr;
1485         if (ifp->prefix_len == 127) /* RFC 6164 */
1486                 return;
1487         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1488         if (ipv6_addr_any(&addr))
1489                 return;
1490         __ipv6_dev_ac_dec(ifp->idev, &addr);
1491 }
1492
1493 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1494 {
1495         if (dev->addr_len != ETH_ALEN)
1496                 return -1;
1497         memcpy(eui, dev->dev_addr, 3);
1498         memcpy(eui + 5, dev->dev_addr + 3, 3);
1499
1500         /*
1501          * The zSeries OSA network cards can be shared among various
1502          * OS instances, but the OSA cards have only one MAC address.
1503          * This leads to duplicate address conflicts in conjunction
1504          * with IPv6 if more than one instance uses the same card.
1505          *
1506          * The driver for these cards can deliver a unique 16-bit
1507          * identifier for each instance sharing the same card.  It is
1508          * placed instead of 0xFFFE in the interface identifier.  The
1509          * "u" bit of the interface identifier is not inverted in this
1510          * case.  Hence the resulting interface identifier has local
1511          * scope according to RFC2373.
1512          */
1513         if (dev->dev_id) {
1514                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1515                 eui[4] = dev->dev_id & 0xFF;
1516         } else {
1517                 eui[3] = 0xFF;
1518                 eui[4] = 0xFE;
1519                 eui[0] ^= 2;
1520         }
1521         return 0;
1522 }
1523
1524 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1525 {
1526         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1527         if (dev->addr_len != ARCNET_ALEN)
1528                 return -1;
1529         memset(eui, 0, 7);
1530         eui[7] = *(u8*)dev->dev_addr;
1531         return 0;
1532 }
1533
1534 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1535 {
1536         if (dev->addr_len != INFINIBAND_ALEN)
1537                 return -1;
1538         memcpy(eui, dev->dev_addr + 12, 8);
1539         eui[0] |= 2;
1540         return 0;
1541 }
1542
1543 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1544 {
1545         if (addr == 0)
1546                 return -1;
1547         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1548                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1549                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1550                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1551                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1552                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1553         eui[1] = 0;
1554         eui[2] = 0x5E;
1555         eui[3] = 0xFE;
1556         memcpy(eui + 4, &addr, 4);
1557         return 0;
1558 }
1559
1560 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1561 {
1562         if (dev->priv_flags & IFF_ISATAP)
1563                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1564         return -1;
1565 }
1566
1567 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1568 {
1569         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1570 }
1571
1572 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1573 {
1574         switch (dev->type) {
1575         case ARPHRD_ETHER:
1576         case ARPHRD_FDDI:
1577         case ARPHRD_IEEE802_TR:
1578                 return addrconf_ifid_eui48(eui, dev);
1579         case ARPHRD_ARCNET:
1580                 return addrconf_ifid_arcnet(eui, dev);
1581         case ARPHRD_INFINIBAND:
1582                 return addrconf_ifid_infiniband(eui, dev);
1583         case ARPHRD_SIT:
1584                 return addrconf_ifid_sit(eui, dev);
1585         case ARPHRD_IPGRE:
1586                 return addrconf_ifid_gre(eui, dev);
1587         }
1588         return -1;
1589 }
1590
1591 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1592 {
1593         int err = -1;
1594         struct inet6_ifaddr *ifp;
1595
1596         read_lock_bh(&idev->lock);
1597         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1598                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1599                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1600                         err = 0;
1601                         break;
1602                 }
1603         }
1604         read_unlock_bh(&idev->lock);
1605         return err;
1606 }
1607
1608 #ifdef CONFIG_IPV6_PRIVACY
1609 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1610 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1611 {
1612 regen:
1613         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1614         idev->rndid[0] &= ~0x02;
1615
1616         /*
1617          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1618          * check if generated address is not inappropriate
1619          *
1620          *  - Reserved subnet anycast (RFC 2526)
1621          *      11111101 11....11 1xxxxxxx
1622          *  - ISATAP (RFC4214) 6.1
1623          *      00-00-5E-FE-xx-xx-xx-xx
1624          *  - value 0
1625          *  - XXX: already assigned to an address on the device
1626          */
1627         if (idev->rndid[0] == 0xfd &&
1628             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1629             (idev->rndid[7]&0x80))
1630                 goto regen;
1631         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1632                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1633                         goto regen;
1634                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1635                         goto regen;
1636         }
1637
1638         return 0;
1639 }
1640
1641 static void ipv6_regen_rndid(unsigned long data)
1642 {
1643         struct inet6_dev *idev = (struct inet6_dev *) data;
1644         unsigned long expires;
1645
1646         rcu_read_lock_bh();
1647         write_lock_bh(&idev->lock);
1648
1649         if (idev->dead)
1650                 goto out;
1651
1652         if (__ipv6_regen_rndid(idev) < 0)
1653                 goto out;
1654
1655         expires = jiffies +
1656                 idev->cnf.temp_prefered_lft * HZ -
1657                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1658                 idev->cnf.max_desync_factor * HZ;
1659         if (time_before(expires, jiffies)) {
1660                 printk(KERN_WARNING
1661                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1662                         idev->dev->name);
1663                 goto out;
1664         }
1665
1666         if (!mod_timer(&idev->regen_timer, expires))
1667                 in6_dev_hold(idev);
1668
1669 out:
1670         write_unlock_bh(&idev->lock);
1671         rcu_read_unlock_bh();
1672         in6_dev_put(idev);
1673 }
1674
1675 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1676         int ret = 0;
1677
1678         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1679                 ret = __ipv6_regen_rndid(idev);
1680         return ret;
1681 }
1682 #endif
1683
1684 /*
1685  *      Add prefix route.
1686  */
1687
1688 static void
1689 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1690                       unsigned long expires, u32 flags)
1691 {
1692         struct fib6_config cfg = {
1693                 .fc_table = RT6_TABLE_PREFIX,
1694                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1695                 .fc_ifindex = dev->ifindex,
1696                 .fc_expires = expires,
1697                 .fc_dst_len = plen,
1698                 .fc_flags = RTF_UP | flags,
1699                 .fc_nlinfo.nl_net = dev_net(dev),
1700                 .fc_protocol = RTPROT_KERNEL,
1701         };
1702
1703         ipv6_addr_copy(&cfg.fc_dst, pfx);
1704
1705         /* Prevent useless cloning on PtP SIT.
1706            This thing is done here expecting that the whole
1707            class of non-broadcast devices need not cloning.
1708          */
1709 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1710         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1711                 cfg.fc_flags |= RTF_NONEXTHOP;
1712 #endif
1713
1714         ip6_route_add(&cfg);
1715 }
1716
1717
1718 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1719                                                   int plen,
1720                                                   const struct net_device *dev,
1721                                                   u32 flags, u32 noflags)
1722 {
1723         struct fib6_node *fn;
1724         struct rt6_info *rt = NULL;
1725         struct fib6_table *table;
1726
1727         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1728         if (table == NULL)
1729                 return NULL;
1730
1731         write_lock_bh(&table->tb6_lock);
1732         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1733         if (!fn)
1734                 goto out;
1735         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1736                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1737                         continue;
1738                 if ((rt->rt6i_flags & flags) != flags)
1739                         continue;
1740                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1741                         continue;
1742                 dst_hold(&rt->dst);
1743                 break;
1744         }
1745 out:
1746         write_unlock_bh(&table->tb6_lock);
1747         return rt;
1748 }
1749
1750
1751 /* Create "default" multicast route to the interface */
1752
1753 static void addrconf_add_mroute(struct net_device *dev)
1754 {
1755         struct fib6_config cfg = {
1756                 .fc_table = RT6_TABLE_LOCAL,
1757                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1758                 .fc_ifindex = dev->ifindex,
1759                 .fc_dst_len = 8,
1760                 .fc_flags = RTF_UP,
1761                 .fc_nlinfo.nl_net = dev_net(dev),
1762         };
1763
1764         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1765
1766         ip6_route_add(&cfg);
1767 }
1768
1769 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1770 static void sit_route_add(struct net_device *dev)
1771 {
1772         struct fib6_config cfg = {
1773                 .fc_table = RT6_TABLE_MAIN,
1774                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1775                 .fc_ifindex = dev->ifindex,
1776                 .fc_dst_len = 96,
1777                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1778                 .fc_nlinfo.nl_net = dev_net(dev),
1779         };
1780
1781         /* prefix length - 96 bits "::d.d.d.d" */
1782         ip6_route_add(&cfg);
1783 }
1784 #endif
1785
1786 static void addrconf_add_lroute(struct net_device *dev)
1787 {
1788         struct in6_addr addr;
1789
1790         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1791         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1792 }
1793
1794 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1795 {
1796         struct inet6_dev *idev;
1797
1798         ASSERT_RTNL();
1799
1800         idev = ipv6_find_idev(dev);
1801         if (!idev)
1802                 return ERR_PTR(-ENOBUFS);
1803
1804         if (idev->cnf.disable_ipv6)
1805                 return ERR_PTR(-EACCES);
1806
1807         /* Add default multicast route */
1808         if (!(dev->flags & IFF_LOOPBACK))
1809                 addrconf_add_mroute(dev);
1810
1811         /* Add link local route */
1812         addrconf_add_lroute(dev);
1813         return idev;
1814 }
1815
1816 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1817 {
1818         struct prefix_info *pinfo;
1819         __u32 valid_lft;
1820         __u32 prefered_lft;
1821         int addr_type;
1822         struct inet6_dev *in6_dev;
1823         struct net *net = dev_net(dev);
1824
1825         pinfo = (struct prefix_info *) opt;
1826
1827         if (len < sizeof(struct prefix_info)) {
1828                 ADBG(("addrconf: prefix option too short\n"));
1829                 return;
1830         }
1831
1832         /*
1833          *      Validation checks ([ADDRCONF], page 19)
1834          */
1835
1836         addr_type = ipv6_addr_type(&pinfo->prefix);
1837
1838         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1839                 return;
1840
1841         valid_lft = ntohl(pinfo->valid);
1842         prefered_lft = ntohl(pinfo->prefered);
1843
1844         if (prefered_lft > valid_lft) {
1845                 if (net_ratelimit())
1846                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1847                 return;
1848         }
1849
1850         in6_dev = in6_dev_get(dev);
1851
1852         if (in6_dev == NULL) {
1853                 if (net_ratelimit())
1854                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1855                 return;
1856         }
1857
1858         /*
1859          *      Two things going on here:
1860          *      1) Add routes for on-link prefixes
1861          *      2) Configure prefixes with the auto flag set
1862          */
1863
1864         if (pinfo->onlink) {
1865                 struct rt6_info *rt;
1866                 unsigned long rt_expires;
1867
1868                 /* Avoid arithmetic overflow. Really, we could
1869                  * save rt_expires in seconds, likely valid_lft,
1870                  * but it would require division in fib gc, that it
1871                  * not good.
1872                  */
1873                 if (HZ > USER_HZ)
1874                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1875                 else
1876                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1877
1878                 if (addrconf_finite_timeout(rt_expires))
1879                         rt_expires *= HZ;
1880
1881                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1882                                                pinfo->prefix_len,
1883                                                dev,
1884                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1885                                                RTF_GATEWAY | RTF_DEFAULT);
1886
1887                 if (rt) {
1888                         /* Autoconf prefix route */
1889                         if (valid_lft == 0) {
1890                                 ip6_del_rt(rt);
1891                                 rt = NULL;
1892                         } else if (addrconf_finite_timeout(rt_expires)) {
1893                                 /* not infinity */
1894                                 rt->rt6i_expires = jiffies + rt_expires;
1895                                 rt->rt6i_flags |= RTF_EXPIRES;
1896                         } else {
1897                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1898                                 rt->rt6i_expires = 0;
1899                         }
1900                 } else if (valid_lft) {
1901                         clock_t expires = 0;
1902                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1903                         if (addrconf_finite_timeout(rt_expires)) {
1904                                 /* not infinity */
1905                                 flags |= RTF_EXPIRES;
1906                                 expires = jiffies_to_clock_t(rt_expires);
1907                         }
1908                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1909                                               dev, expires, flags);
1910                 }
1911                 if (rt)
1912                         dst_release(&rt->dst);
1913         }
1914
1915         /* Try to figure out our local address for this prefix */
1916
1917         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1918                 struct inet6_ifaddr * ifp;
1919                 struct in6_addr addr;
1920                 int create = 0, update_lft = 0;
1921
1922                 if (pinfo->prefix_len == 64) {
1923                         memcpy(&addr, &pinfo->prefix, 8);
1924                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1925                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1926                                 in6_dev_put(in6_dev);
1927                                 return;
1928                         }
1929                         goto ok;
1930                 }
1931                 if (net_ratelimit())
1932                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1933                                pinfo->prefix_len);
1934                 in6_dev_put(in6_dev);
1935                 return;
1936
1937 ok:
1938
1939                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1940
1941                 if (ifp == NULL && valid_lft) {
1942                         int max_addresses = in6_dev->cnf.max_addresses;
1943                         u32 addr_flags = 0;
1944
1945 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1946                         if (in6_dev->cnf.optimistic_dad &&
1947                             !net->ipv6.devconf_all->forwarding)
1948                                 addr_flags = IFA_F_OPTIMISTIC;
1949 #endif
1950
1951                         /* Do not allow to create too much of autoconfigured
1952                          * addresses; this would be too easy way to crash kernel.
1953                          */
1954                         if (!max_addresses ||
1955                             ipv6_count_addresses(in6_dev) < max_addresses)
1956                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1957                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1958                                                     addr_flags);
1959
1960                         if (!ifp || IS_ERR(ifp)) {
1961                                 in6_dev_put(in6_dev);
1962                                 return;
1963                         }
1964
1965                         update_lft = create = 1;
1966                         ifp->cstamp = jiffies;
1967                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1968                 }
1969
1970                 if (ifp) {
1971                         int flags;
1972                         unsigned long now;
1973 #ifdef CONFIG_IPV6_PRIVACY
1974                         struct inet6_ifaddr *ift;
1975 #endif
1976                         u32 stored_lft;
1977
1978                         /* update lifetime (RFC2462 5.5.3 e) */
1979                         spin_lock(&ifp->lock);
1980                         now = jiffies;
1981                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1982                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1983                         else
1984                                 stored_lft = 0;
1985                         if (!update_lft && stored_lft) {
1986                                 if (valid_lft > MIN_VALID_LIFETIME ||
1987                                     valid_lft > stored_lft)
1988                                         update_lft = 1;
1989                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1990                                         /* valid_lft <= stored_lft is always true */
1991                                         /*
1992                                          * RFC 4862 Section 5.5.3e:
1993                                          * "Note that the preferred lifetime of
1994                                          *  the corresponding address is always
1995                                          *  reset to the Preferred Lifetime in
1996                                          *  the received Prefix Information
1997                                          *  option, regardless of whether the
1998                                          *  valid lifetime is also reset or
1999                                          *  ignored."
2000                                          *
2001                                          *  So if the preferred lifetime in
2002                                          *  this advertisement is different
2003                                          *  than what we have stored, but the
2004                                          *  valid lifetime is invalid, just
2005                                          *  reset prefered_lft.
2006                                          *
2007                                          *  We must set the valid lifetime
2008                                          *  to the stored lifetime since we'll
2009                                          *  be updating the timestamp below,
2010                                          *  else we'll set it back to the
2011                                          *  minimum.
2012                                          */
2013                                         if (prefered_lft != ifp->prefered_lft) {
2014                                                 valid_lft = stored_lft;
2015                                                 update_lft = 1;
2016                                         }
2017                                 } else {
2018                                         valid_lft = MIN_VALID_LIFETIME;
2019                                         if (valid_lft < prefered_lft)
2020                                                 prefered_lft = valid_lft;
2021                                         update_lft = 1;
2022                                 }
2023                         }
2024
2025                         if (update_lft) {
2026                                 ifp->valid_lft = valid_lft;
2027                                 ifp->prefered_lft = prefered_lft;
2028                                 ifp->tstamp = now;
2029                                 flags = ifp->flags;
2030                                 ifp->flags &= ~IFA_F_DEPRECATED;
2031                                 spin_unlock(&ifp->lock);
2032
2033                                 if (!(flags&IFA_F_TENTATIVE))
2034                                         ipv6_ifa_notify(0, ifp);
2035                         } else
2036                                 spin_unlock(&ifp->lock);
2037
2038 #ifdef CONFIG_IPV6_PRIVACY
2039                         read_lock_bh(&in6_dev->lock);
2040                         /* update all temporary addresses in the list */
2041                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2042                                             tmp_list) {
2043                                 int age, max_valid, max_prefered;
2044
2045                                 if (ifp != ift->ifpub)
2046                                         continue;
2047
2048                                 /*
2049                                  * RFC 4941 section 3.3:
2050                                  * If a received option will extend the lifetime
2051                                  * of a public address, the lifetimes of
2052                                  * temporary addresses should be extended,
2053                                  * subject to the overall constraint that no
2054                                  * temporary addresses should ever remain
2055                                  * "valid" or "preferred" for a time longer than
2056                                  * (TEMP_VALID_LIFETIME) or
2057                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2058                                  * respectively.
2059                                  */
2060                                 age = (now - ift->cstamp) / HZ;
2061                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2062                                 if (max_valid < 0)
2063                                         max_valid = 0;
2064
2065                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2066                                                in6_dev->cnf.max_desync_factor -
2067                                                age;
2068                                 if (max_prefered < 0)
2069                                         max_prefered = 0;
2070
2071                                 if (valid_lft > max_valid)
2072                                         valid_lft = max_valid;
2073
2074                                 if (prefered_lft > max_prefered)
2075                                         prefered_lft = max_prefered;
2076
2077                                 spin_lock(&ift->lock);
2078                                 flags = ift->flags;
2079                                 ift->valid_lft = valid_lft;
2080                                 ift->prefered_lft = prefered_lft;
2081                                 ift->tstamp = now;
2082                                 if (prefered_lft > 0)
2083                                         ift->flags &= ~IFA_F_DEPRECATED;
2084
2085                                 spin_unlock(&ift->lock);
2086                                 if (!(flags&IFA_F_TENTATIVE))
2087                                         ipv6_ifa_notify(0, ift);
2088                         }
2089
2090                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2091                                 /*
2092                                  * When a new public address is created as
2093                                  * described in [ADDRCONF], also create a new
2094                                  * temporary address. Also create a temporary
2095                                  * address if it's enabled but no temporary
2096                                  * address currently exists.
2097                                  */
2098                                 read_unlock_bh(&in6_dev->lock);
2099                                 ipv6_create_tempaddr(ifp, NULL);
2100                         } else {
2101                                 read_unlock_bh(&in6_dev->lock);
2102                         }
2103 #endif
2104                         in6_ifa_put(ifp);
2105                         addrconf_verify(0);
2106                 }
2107         }
2108         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2109         in6_dev_put(in6_dev);
2110 }
2111
2112 /*
2113  *      Set destination address.
2114  *      Special case for SIT interfaces where we create a new "virtual"
2115  *      device.
2116  */
2117 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2118 {
2119         struct in6_ifreq ireq;
2120         struct net_device *dev;
2121         int err = -EINVAL;
2122
2123         rtnl_lock();
2124
2125         err = -EFAULT;
2126         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2127                 goto err_exit;
2128
2129         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2130
2131         err = -ENODEV;
2132         if (dev == NULL)
2133                 goto err_exit;
2134
2135 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2136         if (dev->type == ARPHRD_SIT) {
2137                 const struct net_device_ops *ops = dev->netdev_ops;
2138                 struct ifreq ifr;
2139                 struct ip_tunnel_parm p;
2140
2141                 err = -EADDRNOTAVAIL;
2142                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2143                         goto err_exit;
2144
2145                 memset(&p, 0, sizeof(p));
2146                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2147                 p.iph.saddr = 0;
2148                 p.iph.version = 4;
2149                 p.iph.ihl = 5;
2150                 p.iph.protocol = IPPROTO_IPV6;
2151                 p.iph.ttl = 64;
2152                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2153
2154                 if (ops->ndo_do_ioctl) {
2155                         mm_segment_t oldfs = get_fs();
2156
2157                         set_fs(KERNEL_DS);
2158                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2159                         set_fs(oldfs);
2160                 } else
2161                         err = -EOPNOTSUPP;
2162
2163                 if (err == 0) {
2164                         err = -ENOBUFS;
2165                         dev = __dev_get_by_name(net, p.name);
2166                         if (!dev)
2167                                 goto err_exit;
2168                         err = dev_open(dev);
2169                 }
2170         }
2171 #endif
2172
2173 err_exit:
2174         rtnl_unlock();
2175         return err;
2176 }
2177
2178 /*
2179  *      Manual configuration of address on an interface
2180  */
2181 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2182                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2183                           __u32 valid_lft)
2184 {
2185         struct inet6_ifaddr *ifp;
2186         struct inet6_dev *idev;
2187         struct net_device *dev;
2188         int scope;
2189         u32 flags;
2190         clock_t expires;
2191         unsigned long timeout;
2192
2193         ASSERT_RTNL();
2194
2195         if (plen > 128)
2196                 return -EINVAL;
2197
2198         /* check the lifetime */
2199         if (!valid_lft || prefered_lft > valid_lft)
2200                 return -EINVAL;
2201
2202         dev = __dev_get_by_index(net, ifindex);
2203         if (!dev)
2204                 return -ENODEV;
2205
2206         idev = addrconf_add_dev(dev);
2207         if (IS_ERR(idev))
2208                 return PTR_ERR(idev);
2209
2210         scope = ipv6_addr_scope(pfx);
2211
2212         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2213         if (addrconf_finite_timeout(timeout)) {
2214                 expires = jiffies_to_clock_t(timeout * HZ);
2215                 valid_lft = timeout;
2216                 flags = RTF_EXPIRES;
2217         } else {
2218                 expires = 0;
2219                 flags = 0;
2220                 ifa_flags |= IFA_F_PERMANENT;
2221         }
2222
2223         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2224         if (addrconf_finite_timeout(timeout)) {
2225                 if (timeout == 0)
2226                         ifa_flags |= IFA_F_DEPRECATED;
2227                 prefered_lft = timeout;
2228         }
2229
2230         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2231
2232         if (!IS_ERR(ifp)) {
2233                 spin_lock_bh(&ifp->lock);
2234                 ifp->valid_lft = valid_lft;
2235                 ifp->prefered_lft = prefered_lft;
2236                 ifp->tstamp = jiffies;
2237                 spin_unlock_bh(&ifp->lock);
2238
2239                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2240                                       expires, flags);
2241                 /*
2242                  * Note that section 3.1 of RFC 4429 indicates
2243                  * that the Optimistic flag should not be set for
2244                  * manually configured addresses
2245                  */
2246                 addrconf_dad_start(ifp, 0);
2247                 in6_ifa_put(ifp);
2248                 addrconf_verify(0);
2249                 return 0;
2250         }
2251
2252         return PTR_ERR(ifp);
2253 }
2254
2255 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2256                           unsigned int plen)
2257 {
2258         struct inet6_ifaddr *ifp;
2259         struct inet6_dev *idev;
2260         struct net_device *dev;
2261
2262         if (plen > 128)
2263                 return -EINVAL;
2264
2265         dev = __dev_get_by_index(net, ifindex);
2266         if (!dev)
2267                 return -ENODEV;
2268
2269         if ((idev = __in6_dev_get(dev)) == NULL)
2270                 return -ENXIO;
2271
2272         read_lock_bh(&idev->lock);
2273         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2274                 if (ifp->prefix_len == plen &&
2275                     ipv6_addr_equal(pfx, &ifp->addr)) {
2276                         in6_ifa_hold(ifp);
2277                         read_unlock_bh(&idev->lock);
2278
2279                         ipv6_del_addr(ifp);
2280
2281                         /* If the last address is deleted administratively,
2282                            disable IPv6 on this interface.
2283                          */
2284                         if (list_empty(&idev->addr_list))
2285                                 addrconf_ifdown(idev->dev, 1);
2286                         return 0;
2287                 }
2288         }
2289         read_unlock_bh(&idev->lock);
2290         return -EADDRNOTAVAIL;
2291 }
2292
2293
2294 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2295 {
2296         struct in6_ifreq ireq;
2297         int err;
2298
2299         if (!capable(CAP_NET_ADMIN))
2300                 return -EPERM;
2301
2302         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2303                 return -EFAULT;
2304
2305         rtnl_lock();
2306         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2307                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2308                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2309         rtnl_unlock();
2310         return err;
2311 }
2312
2313 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2314 {
2315         struct in6_ifreq ireq;
2316         int err;
2317
2318         if (!capable(CAP_NET_ADMIN))
2319                 return -EPERM;
2320
2321         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2322                 return -EFAULT;
2323
2324         rtnl_lock();
2325         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2326                              ireq.ifr6_prefixlen);
2327         rtnl_unlock();
2328         return err;
2329 }
2330
2331 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2332                      int plen, int scope)
2333 {
2334         struct inet6_ifaddr *ifp;
2335
2336         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2337         if (!IS_ERR(ifp)) {
2338                 spin_lock_bh(&ifp->lock);
2339                 ifp->flags &= ~IFA_F_TENTATIVE;
2340                 spin_unlock_bh(&ifp->lock);
2341                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2342                 in6_ifa_put(ifp);
2343         }
2344 }
2345
2346 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2347 static void sit_add_v4_addrs(struct inet6_dev *idev)
2348 {
2349         struct in6_addr addr;
2350         struct net_device *dev;
2351         struct net *net = dev_net(idev->dev);
2352         int scope;
2353
2354         ASSERT_RTNL();
2355
2356         memset(&addr, 0, sizeof(struct in6_addr));
2357         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2358
2359         if (idev->dev->flags&IFF_POINTOPOINT) {
2360                 addr.s6_addr32[0] = htonl(0xfe800000);
2361                 scope = IFA_LINK;
2362         } else {
2363                 scope = IPV6_ADDR_COMPATv4;
2364         }
2365
2366         if (addr.s6_addr32[3]) {
2367                 add_addr(idev, &addr, 128, scope);
2368                 return;
2369         }
2370
2371         for_each_netdev(net, dev) {
2372                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2373                 if (in_dev && (dev->flags & IFF_UP)) {
2374                         struct in_ifaddr * ifa;
2375
2376                         int flag = scope;
2377
2378                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2379                                 int plen;
2380
2381                                 addr.s6_addr32[3] = ifa->ifa_local;
2382
2383                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2384                                         continue;
2385                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2386                                         if (idev->dev->flags&IFF_POINTOPOINT)
2387                                                 continue;
2388                                         flag |= IFA_HOST;
2389                                 }
2390                                 if (idev->dev->flags&IFF_POINTOPOINT)
2391                                         plen = 64;
2392                                 else
2393                                         plen = 96;
2394
2395                                 add_addr(idev, &addr, plen, flag);
2396                         }
2397                 }
2398         }
2399 }
2400 #endif
2401
2402 static void init_loopback(struct net_device *dev)
2403 {
2404         struct inet6_dev  *idev;
2405
2406         /* ::1 */
2407
2408         ASSERT_RTNL();
2409
2410         if ((idev = ipv6_find_idev(dev)) == NULL) {
2411                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2412                 return;
2413         }
2414
2415         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2416 }
2417
2418 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2419 {
2420         struct inet6_ifaddr * ifp;
2421         u32 addr_flags = IFA_F_PERMANENT;
2422
2423 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2424         if (idev->cnf.optimistic_dad &&
2425             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2426                 addr_flags |= IFA_F_OPTIMISTIC;
2427 #endif
2428
2429
2430         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2431         if (!IS_ERR(ifp)) {
2432                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2433                 addrconf_dad_start(ifp, 0);
2434                 in6_ifa_put(ifp);
2435         }
2436 }
2437
2438 static void addrconf_dev_config(struct net_device *dev)
2439 {
2440         struct in6_addr addr;
2441         struct inet6_dev    * idev;
2442
2443         ASSERT_RTNL();
2444
2445         if ((dev->type != ARPHRD_ETHER) &&
2446             (dev->type != ARPHRD_FDDI) &&
2447             (dev->type != ARPHRD_IEEE802_TR) &&
2448             (dev->type != ARPHRD_ARCNET) &&
2449             (dev->type != ARPHRD_INFINIBAND)) {
2450                 /* Alas, we support only Ethernet autoconfiguration. */
2451                 return;
2452         }
2453
2454         idev = addrconf_add_dev(dev);
2455         if (IS_ERR(idev))
2456                 return;
2457
2458         memset(&addr, 0, sizeof(struct in6_addr));
2459         addr.s6_addr32[0] = htonl(0xFE800000);
2460
2461         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2462                 addrconf_add_linklocal(idev, &addr);
2463 }
2464
2465 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2466 static void addrconf_sit_config(struct net_device *dev)
2467 {
2468         struct inet6_dev *idev;
2469
2470         ASSERT_RTNL();
2471
2472         /*
2473          * Configure the tunnel with one of our IPv4
2474          * addresses... we should configure all of
2475          * our v4 addrs in the tunnel
2476          */
2477
2478         if ((idev = ipv6_find_idev(dev)) == NULL) {
2479                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2480                 return;
2481         }
2482
2483         if (dev->priv_flags & IFF_ISATAP) {
2484                 struct in6_addr addr;
2485
2486                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2487                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2488                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2489                         addrconf_add_linklocal(idev, &addr);
2490                 return;
2491         }
2492
2493         sit_add_v4_addrs(idev);
2494
2495         if (dev->flags&IFF_POINTOPOINT) {
2496                 addrconf_add_mroute(dev);
2497                 addrconf_add_lroute(dev);
2498         } else
2499                 sit_route_add(dev);
2500 }
2501 #endif
2502
2503 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2504 static void addrconf_gre_config(struct net_device *dev)
2505 {
2506         struct inet6_dev *idev;
2507         struct in6_addr addr;
2508
2509         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2510
2511         ASSERT_RTNL();
2512
2513         if ((idev = ipv6_find_idev(dev)) == NULL) {
2514                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2515                 return;
2516         }
2517
2518         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2519         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2520
2521         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2522                 addrconf_add_linklocal(idev, &addr);
2523 }
2524 #endif
2525
2526 static inline int
2527 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2528 {
2529         struct in6_addr lladdr;
2530
2531         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2532                 addrconf_add_linklocal(idev, &lladdr);
2533                 return 0;
2534         }
2535         return -1;
2536 }
2537
2538 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2539 {
2540         struct net_device *link_dev;
2541         struct net *net = dev_net(idev->dev);
2542
2543         /* first try to inherit the link-local address from the link device */
2544         if (idev->dev->iflink &&
2545             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2546                 if (!ipv6_inherit_linklocal(idev, link_dev))
2547                         return;
2548         }
2549         /* then try to inherit it from any device */
2550         for_each_netdev(net, link_dev) {
2551                 if (!ipv6_inherit_linklocal(idev, link_dev))
2552                         return;
2553         }
2554         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2555 }
2556
2557 /*
2558  * Autoconfigure tunnel with a link-local address so routing protocols,
2559  * DHCPv6, MLD etc. can be run over the virtual link
2560  */
2561
2562 static void addrconf_ip6_tnl_config(struct net_device *dev)
2563 {
2564         struct inet6_dev *idev;
2565
2566         ASSERT_RTNL();
2567
2568         idev = addrconf_add_dev(dev);
2569         if (IS_ERR(idev)) {
2570                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2571                 return;
2572         }
2573         ip6_tnl_add_linklocal(idev);
2574 }
2575
2576 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2577                            void * data)
2578 {
2579         struct net_device *dev = (struct net_device *) data;
2580         struct inet6_dev *idev = __in6_dev_get(dev);
2581         int run_pending = 0;
2582         int err;
2583
2584         switch (event) {
2585         case NETDEV_REGISTER:
2586                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2587                         idev = ipv6_add_dev(dev);
2588                         if (!idev)
2589                                 return notifier_from_errno(-ENOMEM);
2590                 }
2591                 break;
2592
2593         case NETDEV_UP:
2594         case NETDEV_CHANGE:
2595                 if (dev->flags & IFF_SLAVE)
2596                         break;
2597
2598                 if (event == NETDEV_UP) {
2599                         if (!addrconf_qdisc_ok(dev)) {
2600                                 /* device is not ready yet. */
2601                                 printk(KERN_INFO
2602                                         "ADDRCONF(NETDEV_UP): %s: "
2603                                         "link is not ready\n",
2604                                         dev->name);
2605                                 break;
2606                         }
2607
2608                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2609                                 idev = ipv6_add_dev(dev);
2610
2611                         if (idev) {
2612                                 idev->if_flags |= IF_READY;
2613                                 run_pending = 1;
2614                         }
2615                 } else {
2616                         if (!addrconf_qdisc_ok(dev)) {
2617                                 /* device is still not ready. */
2618                                 break;
2619                         }
2620
2621                         if (idev) {
2622                                 if (idev->if_flags & IF_READY)
2623                                         /* device is already configured. */
2624                                         break;
2625                                 idev->if_flags |= IF_READY;
2626                         }
2627
2628                         printk(KERN_INFO
2629                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2630                                         "link becomes ready\n",
2631                                         dev->name);
2632
2633                         run_pending = 1;
2634                 }
2635
2636                 switch (dev->type) {
2637 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2638                 case ARPHRD_SIT:
2639                         addrconf_sit_config(dev);
2640                         break;
2641 #endif
2642 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2643                 case ARPHRD_IPGRE:
2644                         addrconf_gre_config(dev);
2645                         break;
2646 #endif
2647                 case ARPHRD_TUNNEL6:
2648                         addrconf_ip6_tnl_config(dev);
2649                         break;
2650                 case ARPHRD_LOOPBACK:
2651                         init_loopback(dev);
2652                         break;
2653
2654                 default:
2655                         addrconf_dev_config(dev);
2656                         break;
2657                 }
2658
2659                 if (idev) {
2660                         if (run_pending)
2661                                 addrconf_dad_run(idev);
2662
2663                         /*
2664                          * If the MTU changed during the interface down,
2665                          * when the interface up, the changed MTU must be
2666                          * reflected in the idev as well as routers.
2667                          */
2668                         if (idev->cnf.mtu6 != dev->mtu &&
2669                             dev->mtu >= IPV6_MIN_MTU) {
2670                                 rt6_mtu_change(dev, dev->mtu);
2671                                 idev->cnf.mtu6 = dev->mtu;
2672                         }
2673                         idev->tstamp = jiffies;
2674                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2675
2676                         /*
2677                          * If the changed mtu during down is lower than
2678                          * IPV6_MIN_MTU stop IPv6 on this interface.
2679                          */
2680                         if (dev->mtu < IPV6_MIN_MTU)
2681                                 addrconf_ifdown(dev, 1);
2682                 }
2683                 break;
2684
2685         case NETDEV_CHANGEMTU:
2686                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2687                         rt6_mtu_change(dev, dev->mtu);
2688                         idev->cnf.mtu6 = dev->mtu;
2689                         break;
2690                 }
2691
2692                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2693                         idev = ipv6_add_dev(dev);
2694                         if (idev)
2695                                 break;
2696                 }
2697
2698                 /*
2699                  * MTU falled under IPV6_MIN_MTU.
2700                  * Stop IPv6 on this interface.
2701                  */
2702
2703         case NETDEV_DOWN:
2704         case NETDEV_UNREGISTER:
2705                 /*
2706                  *      Remove all addresses from this interface.
2707                  */
2708                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2709                 break;
2710
2711         case NETDEV_CHANGENAME:
2712                 if (idev) {
2713                         snmp6_unregister_dev(idev);
2714                         addrconf_sysctl_unregister(idev);
2715                         addrconf_sysctl_register(idev);
2716                         err = snmp6_register_dev(idev);
2717                         if (err)
2718                                 return notifier_from_errno(err);
2719                 }
2720                 break;
2721
2722         case NETDEV_PRE_TYPE_CHANGE:
2723         case NETDEV_POST_TYPE_CHANGE:
2724                 addrconf_type_change(dev, event);
2725                 break;
2726         }
2727
2728         return NOTIFY_OK;
2729 }
2730
2731 /*
2732  *      addrconf module should be notified of a device going up
2733  */
2734 static struct notifier_block ipv6_dev_notf = {
2735         .notifier_call = addrconf_notify,
2736 };
2737
2738 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2739 {
2740         struct inet6_dev *idev;
2741         ASSERT_RTNL();
2742
2743         idev = __in6_dev_get(dev);
2744
2745         if (event == NETDEV_POST_TYPE_CHANGE)
2746                 ipv6_mc_remap(idev);
2747         else if (event == NETDEV_PRE_TYPE_CHANGE)
2748                 ipv6_mc_unmap(idev);
2749 }
2750
2751 static int addrconf_ifdown(struct net_device *dev, int how)
2752 {
2753         struct net *net = dev_net(dev);
2754         struct inet6_dev *idev;
2755         struct inet6_ifaddr *ifa;
2756         int state, i;
2757
2758         ASSERT_RTNL();
2759
2760         rt6_ifdown(net, dev);
2761         neigh_ifdown(&nd_tbl, dev);
2762
2763         idev = __in6_dev_get(dev);
2764         if (idev == NULL)
2765                 return -ENODEV;
2766
2767         /*
2768          * Step 1: remove reference to ipv6 device from parent device.
2769          *         Do not dev_put!
2770          */
2771         if (how) {
2772                 idev->dead = 1;
2773
2774                 /* protected by rtnl_lock */
2775                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2776
2777                 /* Step 1.5: remove snmp6 entry */
2778                 snmp6_unregister_dev(idev);
2779
2780         }
2781
2782         /* Step 2: clear hash table */
2783         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2784                 struct hlist_head *h = &inet6_addr_lst[i];
2785                 struct hlist_node *n;
2786
2787                 spin_lock_bh(&addrconf_hash_lock);
2788         restart:
2789                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2790                         if (ifa->idev == idev) {
2791                                 hlist_del_init_rcu(&ifa->addr_lst);
2792                                 addrconf_del_timer(ifa);
2793                                 goto restart;
2794                         }
2795                 }
2796                 spin_unlock_bh(&addrconf_hash_lock);
2797         }
2798
2799         write_lock_bh(&idev->lock);
2800
2801         /* Step 2: clear flags for stateless addrconf */
2802         if (!how)
2803                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2804
2805 #ifdef CONFIG_IPV6_PRIVACY
2806         if (how && del_timer(&idev->regen_timer))
2807                 in6_dev_put(idev);
2808
2809         /* Step 3: clear tempaddr list */
2810         while (!list_empty(&idev->tempaddr_list)) {
2811                 ifa = list_first_entry(&idev->tempaddr_list,
2812                                        struct inet6_ifaddr, tmp_list);
2813                 list_del(&ifa->tmp_list);
2814                 write_unlock_bh(&idev->lock);
2815                 spin_lock_bh(&ifa->lock);
2816
2817                 if (ifa->ifpub) {
2818                         in6_ifa_put(ifa->ifpub);
2819                         ifa->ifpub = NULL;
2820                 }
2821                 spin_unlock_bh(&ifa->lock);
2822                 in6_ifa_put(ifa);
2823                 write_lock_bh(&idev->lock);
2824         }
2825 #endif
2826
2827         while (!list_empty(&idev->addr_list)) {
2828                 ifa = list_first_entry(&idev->addr_list,
2829                                        struct inet6_ifaddr, if_list);
2830                 addrconf_del_timer(ifa);
2831
2832                 list_del(&ifa->if_list);
2833
2834                 write_unlock_bh(&idev->lock);
2835
2836                 spin_lock_bh(&ifa->state_lock);
2837                 state = ifa->state;
2838                 ifa->state = INET6_IFADDR_STATE_DEAD;
2839                 spin_unlock_bh(&ifa->state_lock);
2840
2841                 if (state != INET6_IFADDR_STATE_DEAD) {
2842                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2843                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2844                 }
2845                 in6_ifa_put(ifa);
2846
2847                 write_lock_bh(&idev->lock);
2848         }
2849
2850         write_unlock_bh(&idev->lock);
2851
2852         /* Step 5: Discard multicast list */
2853         if (how)
2854                 ipv6_mc_destroy_dev(idev);
2855         else
2856                 ipv6_mc_down(idev);
2857
2858         idev->tstamp = jiffies;
2859
2860         /* Last: Shot the device (if unregistered) */
2861         if (how) {
2862                 addrconf_sysctl_unregister(idev);
2863                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2864                 neigh_ifdown(&nd_tbl, dev);
2865                 in6_dev_put(idev);
2866         }
2867         return 0;
2868 }
2869
2870 static void addrconf_rs_timer(unsigned long data)
2871 {
2872         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2873         struct inet6_dev *idev = ifp->idev;
2874
2875         read_lock(&idev->lock);
2876         if (idev->dead || !(idev->if_flags & IF_READY))
2877                 goto out;
2878
2879         if (idev->cnf.forwarding)
2880                 goto out;
2881
2882         /* Announcement received after solicitation was sent */
2883         if (idev->if_flags & IF_RA_RCVD)
2884                 goto out;
2885
2886         spin_lock(&ifp->lock);
2887         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2888                 /* The wait after the last probe can be shorter */
2889                 addrconf_mod_timer(ifp, AC_RS,
2890                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2891                                    idev->cnf.rtr_solicit_delay :
2892                                    idev->cnf.rtr_solicit_interval);
2893                 spin_unlock(&ifp->lock);
2894
2895                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2896         } else {
2897                 spin_unlock(&ifp->lock);
2898                 /*
2899                  * Note: we do not support deprecated "all on-link"
2900                  * assumption any longer.
2901                  */
2902                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2903                        idev->dev->name);
2904         }
2905
2906 out:
2907         read_unlock(&idev->lock);
2908         in6_ifa_put(ifp);
2909 }
2910
2911 /*
2912  *      Duplicate Address Detection
2913  */
2914 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2915 {
2916         unsigned long rand_num;
2917         struct inet6_dev *idev = ifp->idev;
2918
2919         if (ifp->flags & IFA_F_OPTIMISTIC)
2920                 rand_num = 0;
2921         else
2922                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2923
2924         ifp->probes = idev->cnf.dad_transmits;
2925         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2926 }
2927
2928 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2929 {
2930         struct inet6_dev *idev = ifp->idev;
2931         struct net_device *dev = idev->dev;
2932
2933         addrconf_join_solict(dev, &ifp->addr);
2934
2935         net_srandom(ifp->addr.s6_addr32[3]);
2936
2937         read_lock_bh(&idev->lock);
2938         spin_lock(&ifp->lock);
2939         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2940                 goto out;
2941
2942         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2943             idev->cnf.accept_dad < 1 ||
2944             !(ifp->flags&IFA_F_TENTATIVE) ||
2945             ifp->flags & IFA_F_NODAD) {
2946                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2947                 spin_unlock(&ifp->lock);
2948                 read_unlock_bh(&idev->lock);
2949
2950                 addrconf_dad_completed(ifp);
2951                 return;
2952         }
2953
2954         if (!(idev->if_flags & IF_READY)) {
2955                 spin_unlock(&ifp->lock);
2956                 read_unlock_bh(&idev->lock);
2957                 /*
2958                  * If the device is not ready:
2959                  * - keep it tentative if it is a permanent address.
2960                  * - otherwise, kill it.
2961                  */
2962                 in6_ifa_hold(ifp);
2963                 addrconf_dad_stop(ifp, 0);
2964                 return;
2965         }
2966
2967         /*
2968          * Optimistic nodes can start receiving
2969          * Frames right away
2970          */
2971         if (ifp->flags & IFA_F_OPTIMISTIC)
2972                 ip6_ins_rt(ifp->rt);
2973
2974         addrconf_dad_kick(ifp);
2975 out:
2976         spin_unlock(&ifp->lock);
2977         read_unlock_bh(&idev->lock);
2978 }
2979
2980 static void addrconf_dad_timer(unsigned long data)
2981 {
2982         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2983         struct inet6_dev *idev = ifp->idev;
2984         struct in6_addr mcaddr;
2985
2986         if (!ifp->probes && addrconf_dad_end(ifp))
2987                 goto out;
2988
2989         read_lock(&idev->lock);
2990         if (idev->dead || !(idev->if_flags & IF_READY)) {
2991                 read_unlock(&idev->lock);
2992                 goto out;
2993         }
2994
2995         spin_lock(&ifp->lock);
2996         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2997                 spin_unlock(&ifp->lock);
2998                 read_unlock(&idev->lock);
2999                 goto out;
3000         }
3001
3002         if (ifp->probes == 0) {
3003                 /*
3004                  * DAD was successful
3005                  */
3006
3007                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3008                 spin_unlock(&ifp->lock);
3009                 read_unlock(&idev->lock);
3010
3011                 addrconf_dad_completed(ifp);
3012
3013                 goto out;
3014         }
3015
3016         ifp->probes--;
3017         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3018         spin_unlock(&ifp->lock);
3019         read_unlock(&idev->lock);
3020
3021         /* send a neighbour solicitation for our addr */
3022         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3023         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3024 out:
3025         in6_ifa_put(ifp);
3026 }
3027
3028 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3029 {
3030         struct net_device *dev = ifp->idev->dev;
3031
3032         /*
3033          *      Configure the address for reception. Now it is valid.
3034          */
3035
3036         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3037
3038         /* If added prefix is link local and we are prepared to process
3039            router advertisements, start sending router solicitations.
3040          */
3041
3042         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3043              ifp->idev->cnf.accept_ra == 2) &&
3044             ifp->idev->cnf.rtr_solicits > 0 &&
3045             (dev->flags&IFF_LOOPBACK) == 0 &&
3046             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3047                 /*
3048                  *      If a host as already performed a random delay
3049                  *      [...] as part of DAD [...] there is no need
3050                  *      to delay again before sending the first RS
3051                  */
3052                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3053
3054                 spin_lock_bh(&ifp->lock);
3055                 ifp->probes = 1;
3056                 ifp->idev->if_flags |= IF_RS_SENT;
3057                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3058                 spin_unlock_bh(&ifp->lock);
3059         }
3060 }
3061
3062 static void addrconf_dad_run(struct inet6_dev *idev)
3063 {
3064         struct inet6_ifaddr *ifp;
3065
3066         read_lock_bh(&idev->lock);
3067         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3068                 spin_lock(&ifp->lock);
3069                 if (ifp->flags & IFA_F_TENTATIVE &&
3070                     ifp->state == INET6_IFADDR_STATE_DAD)
3071                         addrconf_dad_kick(ifp);
3072                 spin_unlock(&ifp->lock);
3073         }
3074         read_unlock_bh(&idev->lock);
3075 }
3076
3077 #ifdef CONFIG_PROC_FS
3078 struct if6_iter_state {
3079         struct seq_net_private p;
3080         int bucket;
3081 };
3082
3083 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3084 {
3085         struct inet6_ifaddr *ifa = NULL;
3086         struct if6_iter_state *state = seq->private;
3087         struct net *net = seq_file_net(seq);
3088
3089         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3090                 struct hlist_node *n;
3091                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3092                                          addr_lst)
3093                         if (net_eq(dev_net(ifa->idev->dev), net))
3094                                 return ifa;
3095         }
3096         return NULL;
3097 }
3098
3099 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3100                                          struct inet6_ifaddr *ifa)
3101 {
3102         struct if6_iter_state *state = seq->private;
3103         struct net *net = seq_file_net(seq);
3104         struct hlist_node *n = &ifa->addr_lst;
3105
3106         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3107                 if (net_eq(dev_net(ifa->idev->dev), net))
3108                         return ifa;
3109
3110         while (++state->bucket < IN6_ADDR_HSIZE) {
3111                 hlist_for_each_entry_rcu_bh(ifa, n,
3112                                      &inet6_addr_lst[state->bucket], addr_lst) {
3113                         if (net_eq(dev_net(ifa->idev->dev), net))
3114                                 return ifa;
3115                 }
3116         }
3117
3118         return NULL;
3119 }
3120
3121 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3122 {
3123         struct inet6_ifaddr *ifa = if6_get_first(seq);
3124
3125         if (ifa)
3126                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3127                         --pos;
3128         return pos ? NULL : ifa;
3129 }
3130
3131 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3132         __acquires(rcu_bh)
3133 {
3134         rcu_read_lock_bh();
3135         return if6_get_idx(seq, *pos);
3136 }
3137
3138 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3139 {
3140         struct inet6_ifaddr *ifa;
3141
3142         ifa = if6_get_next(seq, v);
3143         ++*pos;
3144         return ifa;
3145 }
3146
3147 static void if6_seq_stop(struct seq_file *seq, void *v)
3148         __releases(rcu_bh)
3149 {
3150         rcu_read_unlock_bh();
3151 }
3152
3153 static int if6_seq_show(struct seq_file *seq, void *v)
3154 {
3155         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3156         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3157                    &ifp->addr,
3158                    ifp->idev->dev->ifindex,
3159                    ifp->prefix_len,
3160                    ifp->scope,
3161                    ifp->flags,
3162                    ifp->idev->dev->name);
3163         return 0;
3164 }
3165
3166 static const struct seq_operations if6_seq_ops = {
3167         .start  = if6_seq_start,
3168         .next   = if6_seq_next,
3169         .show   = if6_seq_show,
3170         .stop   = if6_seq_stop,
3171 };
3172
3173 static int if6_seq_open(struct inode *inode, struct file *file)
3174 {
3175         return seq_open_net(inode, file, &if6_seq_ops,
3176                             sizeof(struct if6_iter_state));
3177 }
3178
3179 static const struct file_operations if6_fops = {
3180         .owner          = THIS_MODULE,
3181         .open           = if6_seq_open,
3182         .read           = seq_read,
3183         .llseek         = seq_lseek,
3184         .release        = seq_release_net,
3185 };
3186
3187 static int __net_init if6_proc_net_init(struct net *net)
3188 {
3189         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3190                 return -ENOMEM;
3191         return 0;
3192 }
3193
3194 static void __net_exit if6_proc_net_exit(struct net *net)
3195 {
3196        proc_net_remove(net, "if_inet6");
3197 }
3198
3199 static struct pernet_operations if6_proc_net_ops = {
3200        .init = if6_proc_net_init,
3201        .exit = if6_proc_net_exit,
3202 };
3203
3204 int __init if6_proc_init(void)
3205 {
3206         return register_pernet_subsys(&if6_proc_net_ops);
3207 }
3208
3209 void if6_proc_exit(void)
3210 {
3211         unregister_pernet_subsys(&if6_proc_net_ops);
3212 }
3213 #endif  /* CONFIG_PROC_FS */
3214
3215 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3216 /* Check if address is a home address configured on any interface. */
3217 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3218 {
3219         int ret = 0;
3220         struct inet6_ifaddr *ifp = NULL;
3221         struct hlist_node *n;
3222         unsigned int hash = ipv6_addr_hash(addr);
3223
3224         rcu_read_lock_bh();
3225         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3226                 if (!net_eq(dev_net(ifp->idev->dev), net))
3227                         continue;
3228                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3229                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3230                         ret = 1;
3231                         break;
3232                 }
3233         }
3234         rcu_read_unlock_bh();
3235         return ret;
3236 }
3237 #endif
3238
3239 /*
3240  *      Periodic address status verification
3241  */
3242
3243 static void addrconf_verify(unsigned long foo)
3244 {
3245         unsigned long now, next, next_sec, next_sched;
3246         struct inet6_ifaddr *ifp;
3247         struct hlist_node *node;
3248         int i;
3249
3250         rcu_read_lock_bh();
3251         spin_lock(&addrconf_verify_lock);
3252         now = jiffies;
3253         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3254
3255         del_timer(&addr_chk_timer);
3256
3257         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3258 restart:
3259                 hlist_for_each_entry_rcu_bh(ifp, node,
3260                                          &inet6_addr_lst[i], addr_lst) {
3261                         unsigned long age;
3262
3263                         if (ifp->flags & IFA_F_PERMANENT)
3264                                 continue;
3265
3266                         spin_lock(&ifp->lock);
3267                         /* We try to batch several events at once. */
3268                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3269
3270                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3271                             age >= ifp->valid_lft) {
3272                                 spin_unlock(&ifp->lock);
3273                                 in6_ifa_hold(ifp);
3274                                 ipv6_del_addr(ifp);
3275                                 goto restart;
3276                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3277                                 spin_unlock(&ifp->lock);
3278                                 continue;
3279                         } else if (age >= ifp->prefered_lft) {
3280                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3281                                 int deprecate = 0;
3282
3283                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3284                                         deprecate = 1;
3285                                         ifp->flags |= IFA_F_DEPRECATED;
3286                                 }
3287
3288                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3289                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3290
3291                                 spin_unlock(&ifp->lock);
3292
3293                                 if (deprecate) {
3294                                         in6_ifa_hold(ifp);
3295
3296                                         ipv6_ifa_notify(0, ifp);
3297                                         in6_ifa_put(ifp);
3298                                         goto restart;
3299                                 }
3300 #ifdef CONFIG_IPV6_PRIVACY
3301                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3302                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3303                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3304                                         ifp->idev->cnf.dad_transmits *
3305                                         ifp->idev->nd_parms->retrans_time / HZ;
3306
3307                                 if (age >= ifp->prefered_lft - regen_advance) {
3308                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3309                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3310                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3311                                         if (!ifp->regen_count && ifpub) {
3312                                                 ifp->regen_count++;
3313                                                 in6_ifa_hold(ifp);
3314                                                 in6_ifa_hold(ifpub);
3315                                                 spin_unlock(&ifp->lock);
3316
3317                                                 spin_lock(&ifpub->lock);
3318                                                 ifpub->regen_count = 0;
3319                                                 spin_unlock(&ifpub->lock);
3320                                                 ipv6_create_tempaddr(ifpub, ifp);
3321                                                 in6_ifa_put(ifpub);
3322                                                 in6_ifa_put(ifp);
3323                                                 goto restart;
3324                                         }
3325                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3326                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3327                                 spin_unlock(&ifp->lock);
3328 #endif
3329                         } else {
3330                                 /* ifp->prefered_lft <= ifp->valid_lft */
3331                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3332                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3333                                 spin_unlock(&ifp->lock);
3334                         }
3335                 }
3336         }
3337
3338         next_sec = round_jiffies_up(next);
3339         next_sched = next;
3340
3341         /* If rounded timeout is accurate enough, accept it. */
3342         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3343                 next_sched = next_sec;
3344
3345         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3346         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3347                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3348
3349         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3350               now, next, next_sec, next_sched));
3351
3352         addr_chk_timer.expires = next_sched;
3353         add_timer(&addr_chk_timer);
3354         spin_unlock(&addrconf_verify_lock);
3355         rcu_read_unlock_bh();
3356 }
3357
3358 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3359 {
3360         struct in6_addr *pfx = NULL;
3361
3362         if (addr)
3363                 pfx = nla_data(addr);
3364
3365         if (local) {
3366                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3367                         pfx = NULL;
3368                 else
3369                         pfx = nla_data(local);
3370         }
3371
3372         return pfx;
3373 }
3374
3375 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3376         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3377         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3378         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3379 };
3380
3381 static int
3382 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3383 {
3384         struct net *net = sock_net(skb->sk);
3385         struct ifaddrmsg *ifm;
3386         struct nlattr *tb[IFA_MAX+1];
3387         struct in6_addr *pfx;
3388         int err;
3389
3390         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3391         if (err < 0)
3392                 return err;
3393
3394         ifm = nlmsg_data(nlh);
3395         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3396         if (pfx == NULL)
3397                 return -EINVAL;
3398
3399         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3400 }
3401
3402 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3403                              u32 prefered_lft, u32 valid_lft)
3404 {
3405         u32 flags;
3406         clock_t expires;
3407         unsigned long timeout;
3408
3409         if (!valid_lft || (prefered_lft > valid_lft))
3410                 return -EINVAL;
3411
3412         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3413         if (addrconf_finite_timeout(timeout)) {
3414                 expires = jiffies_to_clock_t(timeout * HZ);
3415                 valid_lft = timeout;
3416                 flags = RTF_EXPIRES;
3417         } else {
3418                 expires = 0;
3419                 flags = 0;
3420                 ifa_flags |= IFA_F_PERMANENT;
3421         }
3422
3423         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3424         if (addrconf_finite_timeout(timeout)) {
3425                 if (timeout == 0)
3426                         ifa_flags |= IFA_F_DEPRECATED;
3427                 prefered_lft = timeout;
3428         }
3429
3430         spin_lock_bh(&ifp->lock);
3431         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3432         ifp->tstamp = jiffies;
3433         ifp->valid_lft = valid_lft;
3434         ifp->prefered_lft = prefered_lft;
3435
3436         spin_unlock_bh(&ifp->lock);
3437         if (!(ifp->flags&IFA_F_TENTATIVE))
3438                 ipv6_ifa_notify(0, ifp);
3439
3440         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3441                               expires, flags);
3442         addrconf_verify(0);
3443
3444         return 0;
3445 }
3446
3447 static int
3448 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3449 {
3450         struct net *net = sock_net(skb->sk);
3451         struct ifaddrmsg *ifm;
3452         struct nlattr *tb[IFA_MAX+1];
3453         struct in6_addr *pfx;
3454         struct inet6_ifaddr *ifa;
3455         struct net_device *dev;
3456         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3457         u8 ifa_flags;
3458         int err;
3459
3460         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3461         if (err < 0)
3462                 return err;
3463
3464         ifm = nlmsg_data(nlh);
3465         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3466         if (pfx == NULL)
3467                 return -EINVAL;
3468
3469         if (tb[IFA_CACHEINFO]) {
3470                 struct ifa_cacheinfo *ci;
3471
3472                 ci = nla_data(tb[IFA_CACHEINFO]);
3473                 valid_lft = ci->ifa_valid;
3474                 preferred_lft = ci->ifa_prefered;
3475         } else {
3476                 preferred_lft = INFINITY_LIFE_TIME;
3477                 valid_lft = INFINITY_LIFE_TIME;
3478         }
3479
3480         dev =  __dev_get_by_index(net, ifm->ifa_index);
3481         if (dev == NULL)
3482                 return -ENODEV;
3483
3484         /* We ignore other flags so far. */
3485         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3486
3487         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3488         if (ifa == NULL) {
3489                 /*
3490                  * It would be best to check for !NLM_F_CREATE here but
3491                  * userspace alreay relies on not having to provide this.
3492                  */
3493                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3494                                       ifm->ifa_prefixlen, ifa_flags,
3495                                       preferred_lft, valid_lft);
3496         }
3497
3498         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3499             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3500                 err = -EEXIST;
3501         else
3502                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3503
3504         in6_ifa_put(ifa);
3505
3506         return err;
3507 }
3508
3509 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3510                           u8 scope, int ifindex)
3511 {
3512         struct ifaddrmsg *ifm;
3513
3514         ifm = nlmsg_data(nlh);
3515         ifm->ifa_family = AF_INET6;
3516         ifm->ifa_prefixlen = prefixlen;
3517         ifm->ifa_flags = flags;
3518         ifm->ifa_scope = scope;
3519         ifm->ifa_index = ifindex;
3520 }
3521
3522 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3523                          unsigned long tstamp, u32 preferred, u32 valid)
3524 {
3525         struct ifa_cacheinfo ci;
3526
3527         ci.cstamp = cstamp_delta(cstamp);
3528         ci.tstamp = cstamp_delta(tstamp);
3529         ci.ifa_prefered = preferred;
3530         ci.ifa_valid = valid;
3531
3532         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3533 }
3534
3535 static inline int rt_scope(int ifa_scope)
3536 {
3537         if (ifa_scope & IFA_HOST)
3538                 return RT_SCOPE_HOST;
3539         else if (ifa_scope & IFA_LINK)
3540                 return RT_SCOPE_LINK;
3541         else if (ifa_scope & IFA_SITE)
3542                 return RT_SCOPE_SITE;
3543         else
3544                 return RT_SCOPE_UNIVERSE;
3545 }
3546
3547 static inline int inet6_ifaddr_msgsize(void)
3548 {
3549         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3550                + nla_total_size(16) /* IFA_ADDRESS */
3551                + nla_total_size(sizeof(struct ifa_cacheinfo));
3552 }
3553
3554 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3555                              u32 pid, u32 seq, int event, unsigned int flags)
3556 {
3557         struct nlmsghdr  *nlh;
3558         u32 preferred, valid;
3559
3560         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3561         if (nlh == NULL)
3562                 return -EMSGSIZE;
3563
3564         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3565                       ifa->idev->dev->ifindex);
3566
3567         if (!(ifa->flags&IFA_F_PERMANENT)) {
3568                 preferred = ifa->prefered_lft;
3569                 valid = ifa->valid_lft;
3570                 if (preferred != INFINITY_LIFE_TIME) {
3571                         long tval = (jiffies - ifa->tstamp)/HZ;
3572                         if (preferred > tval)
3573                                 preferred -= tval;
3574                         else
3575                                 preferred = 0;
3576                         if (valid != INFINITY_LIFE_TIME) {
3577                                 if (valid > tval)
3578                                         valid -= tval;
3579                                 else
3580                                         valid = 0;
3581                         }
3582                 }
3583         } else {
3584                 preferred = INFINITY_LIFE_TIME;
3585                 valid = INFINITY_LIFE_TIME;
3586         }
3587
3588         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3589             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3590                 nlmsg_cancel(skb, nlh);
3591                 return -EMSGSIZE;
3592         }
3593
3594         return nlmsg_end(skb, nlh);
3595 }
3596
3597 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3598                                 u32 pid, u32 seq, int event, u16 flags)
3599 {
3600         struct nlmsghdr  *nlh;
3601         u8 scope = RT_SCOPE_UNIVERSE;
3602         int ifindex = ifmca->idev->dev->ifindex;
3603
3604         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3605                 scope = RT_SCOPE_SITE;
3606
3607         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3608         if (nlh == NULL)
3609                 return -EMSGSIZE;
3610
3611         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3612         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3613             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3614                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3615                 nlmsg_cancel(skb, nlh);
3616                 return -EMSGSIZE;
3617         }
3618
3619         return nlmsg_end(skb, nlh);
3620 }
3621
3622 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3623                                 u32 pid, u32 seq, int event, unsigned int flags)
3624 {
3625         struct nlmsghdr  *nlh;
3626         u8 scope = RT_SCOPE_UNIVERSE;
3627         int ifindex = ifaca->aca_idev->dev->ifindex;
3628
3629         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3630                 scope = RT_SCOPE_SITE;
3631
3632         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3633         if (nlh == NULL)
3634                 return -EMSGSIZE;
3635
3636         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3637         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3638             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3639                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3640                 nlmsg_cancel(skb, nlh);
3641                 return -EMSGSIZE;
3642         }
3643
3644         return nlmsg_end(skb, nlh);
3645 }
3646
3647 enum addr_type_t {
3648         UNICAST_ADDR,
3649         MULTICAST_ADDR,
3650         ANYCAST_ADDR,
3651 };
3652
3653 /* called with rcu_read_lock() */
3654 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3655                           struct netlink_callback *cb, enum addr_type_t type,
3656                           int s_ip_idx, int *p_ip_idx)
3657 {
3658         struct ifmcaddr6 *ifmca;
3659         struct ifacaddr6 *ifaca;
3660         int err = 1;
3661         int ip_idx = *p_ip_idx;
3662
3663         read_lock_bh(&idev->lock);
3664         switch (type) {
3665         case UNICAST_ADDR: {
3666                 struct inet6_ifaddr *ifa;
3667
3668                 /* unicast address incl. temp addr */
3669                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3670                         if (++ip_idx < s_ip_idx)
3671                                 continue;
3672                         err = inet6_fill_ifaddr(skb, ifa,
3673                                                 NETLINK_CB(cb->skb).pid,
3674                                                 cb->nlh->nlmsg_seq,
3675                                                 RTM_NEWADDR,
3676                                                 NLM_F_MULTI);
3677                         if (err <= 0)
3678                                 break;
3679                 }
3680                 break;
3681         }
3682         case MULTICAST_ADDR:
3683                 /* multicast address */
3684                 for (ifmca = idev->mc_list; ifmca;
3685                      ifmca = ifmca->next, ip_idx++) {
3686                         if (ip_idx < s_ip_idx)
3687                                 continue;
3688                         err = inet6_fill_ifmcaddr(skb, ifmca,
3689                                                   NETLINK_CB(cb->skb).pid,
3690                                                   cb->nlh->nlmsg_seq,
3691                                                   RTM_GETMULTICAST,
3692                                                   NLM_F_MULTI);
3693                         if (err <= 0)
3694                                 break;
3695                 }
3696                 break;
3697         case ANYCAST_ADDR:
3698                 /* anycast address */
3699                 for (ifaca = idev->ac_list; ifaca;
3700                      ifaca = ifaca->aca_next, ip_idx++) {
3701                         if (ip_idx < s_ip_idx)
3702                                 continue;
3703                         err = inet6_fill_ifacaddr(skb, ifaca,
3704                                                   NETLINK_CB(cb->skb).pid,
3705                                                   cb->nlh->nlmsg_seq,
3706                                                   RTM_GETANYCAST,
3707                                                   NLM_F_MULTI);
3708                         if (err <= 0)
3709                                 break;
3710                 }
3711                 break;
3712         default:
3713                 break;
3714         }
3715         read_unlock_bh(&idev->lock);
3716         *p_ip_idx = ip_idx;
3717         return err;
3718 }
3719
3720 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3721                            enum addr_type_t type)
3722 {
3723         struct net *net = sock_net(skb->sk);
3724         int h, s_h;
3725         int idx, ip_idx;
3726         int s_idx, s_ip_idx;
3727         struct net_device *dev;
3728         struct inet6_dev *idev;
3729         struct hlist_head *head;
3730         struct hlist_node *node;
3731
3732         s_h = cb->args[0];
3733         s_idx = idx = cb->args[1];
3734         s_ip_idx = ip_idx = cb->args[2];
3735
3736         rcu_read_lock();
3737         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3738                 idx = 0;
3739                 head = &net->dev_index_head[h];
3740                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3741                         if (idx < s_idx)
3742                                 goto cont;
3743                         if (h > s_h || idx > s_idx)
3744                                 s_ip_idx = 0;
3745                         ip_idx = 0;
3746                         idev = __in6_dev_get(dev);
3747                         if (!idev)
3748                                 goto cont;
3749
3750                         if (in6_dump_addrs(idev, skb, cb, type,
3751                                            s_ip_idx, &ip_idx) <= 0)
3752                                 goto done;
3753 cont:
3754                         idx++;
3755                 }
3756         }
3757 done:
3758         rcu_read_unlock();
3759         cb->args[0] = h;
3760         cb->args[1] = idx;
3761         cb->args[2] = ip_idx;
3762
3763         return skb->len;
3764 }
3765
3766 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3767 {
3768         enum addr_type_t type = UNICAST_ADDR;
3769
3770         return inet6_dump_addr(skb, cb, type);
3771 }
3772
3773 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3774 {
3775         enum addr_type_t type = MULTICAST_ADDR;
3776
3777         return inet6_dump_addr(skb, cb, type);
3778 }
3779
3780
3781 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3782 {
3783         enum addr_type_t type = ANYCAST_ADDR;
3784
3785         return inet6_dump_addr(skb, cb, type);
3786 }
3787
3788 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3789                              void *arg)
3790 {
3791         struct net *net = sock_net(in_skb->sk);
3792         struct ifaddrmsg *ifm;
3793         struct nlattr *tb[IFA_MAX+1];
3794         struct in6_addr *addr = NULL;
3795         struct net_device *dev = NULL;
3796         struct inet6_ifaddr *ifa;
3797         struct sk_buff *skb;
3798         int err;
3799
3800         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3801         if (err < 0)
3802                 goto errout;
3803
3804         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3805         if (addr == NULL) {
3806                 err = -EINVAL;
3807                 goto errout;
3808         }
3809
3810         ifm = nlmsg_data(nlh);
3811         if (ifm->ifa_index)
3812                 dev = __dev_get_by_index(net, ifm->ifa_index);
3813
3814         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3815         if (!ifa) {
3816                 err = -EADDRNOTAVAIL;
3817                 goto errout;
3818         }
3819
3820         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3821         if (!skb) {
3822                 err = -ENOBUFS;
3823                 goto errout_ifa;
3824         }
3825
3826         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3827                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3828         if (err < 0) {
3829                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3830                 WARN_ON(err == -EMSGSIZE);
3831                 kfree_skb(skb);
3832                 goto errout_ifa;
3833         }
3834         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3835 errout_ifa:
3836         in6_ifa_put(ifa);
3837 errout:
3838         return err;
3839 }
3840
3841 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3842 {
3843         struct sk_buff *skb;
3844         struct net *net = dev_net(ifa->idev->dev);
3845         int err = -ENOBUFS;
3846
3847         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3848         if (skb == NULL)
3849                 goto errout;
3850
3851         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3852         if (err < 0) {
3853                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3854                 WARN_ON(err == -EMSGSIZE);
3855                 kfree_skb(skb);
3856                 goto errout;
3857         }
3858         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3859         return;
3860 errout:
3861         if (err < 0)
3862                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3863 }
3864
3865 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3866                                 __s32 *array, int bytes)
3867 {
3868         BUG_ON(bytes < (DEVCONF_MAX * 4));
3869
3870         memset(array, 0, bytes);
3871         array[DEVCONF_FORWARDING] = cnf->forwarding;
3872         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3873         array[DEVCONF_MTU6] = cnf->mtu6;
3874         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3875         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3876         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3877         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3878         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3879         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3880                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3881         array[DEVCONF_RTR_SOLICIT_DELAY] =
3882                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3883         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3884 #ifdef CONFIG_IPV6_PRIVACY
3885         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3886         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3887         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3888         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3889         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3890 #endif
3891         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3892         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3893         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3894 #ifdef CONFIG_IPV6_ROUTER_PREF
3895         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3896         array[DEVCONF_RTR_PROBE_INTERVAL] =
3897                 jiffies_to_msecs(cnf->rtr_probe_interval);
3898 #ifdef CONFIG_IPV6_ROUTE_INFO
3899         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3900 #endif
3901 #endif
3902         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3903         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3904 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3905         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3906 #endif
3907 #ifdef CONFIG_IPV6_MROUTE
3908         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3909 #endif
3910         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3911         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3912         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3913 }
3914
3915 static inline size_t inet6_ifla6_size(void)
3916 {
3917         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3918              + nla_total_size(sizeof(struct ifla_cacheinfo))
3919              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3920              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3921              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3922 }
3923
3924 static inline size_t inet6_if_nlmsg_size(void)
3925 {
3926         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3927                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3928                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3929                + nla_total_size(4) /* IFLA_MTU */
3930                + nla_total_size(4) /* IFLA_LINK */
3931                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3932 }
3933
3934 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3935                                       int items, int bytes)
3936 {
3937         int i;
3938         int pad = bytes - sizeof(u64) * items;
3939         BUG_ON(pad < 0);
3940
3941         /* Use put_unaligned() because stats may not be aligned for u64. */
3942         put_unaligned(items, &stats[0]);
3943         for (i = 1; i < items; i++)
3944                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3945
3946         memset(&stats[items], 0, pad);
3947 }
3948
3949 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3950                                       int items, int bytes, size_t syncpoff)
3951 {
3952         int i;
3953         int pad = bytes - sizeof(u64) * items;
3954         BUG_ON(pad < 0);
3955
3956         /* Use put_unaligned() because stats may not be aligned for u64. */
3957         put_unaligned(items, &stats[0]);
3958         for (i = 1; i < items; i++)
3959                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3960
3961         memset(&stats[items], 0, pad);
3962 }
3963
3964 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3965                              int bytes)
3966 {
3967         switch (attrtype) {
3968         case IFLA_INET6_STATS:
3969                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3970                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3971                 break;
3972         case IFLA_INET6_ICMP6STATS:
3973                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3974                 break;
3975         }
3976 }
3977
3978 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3979 {
3980         struct nlattr *nla;
3981         struct ifla_cacheinfo ci;
3982
3983         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3984
3985         ci.max_reasm_len = IPV6_MAXPLEN;
3986         ci.tstamp = cstamp_delta(idev->tstamp);
3987         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3988         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3989         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3990
3991         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3992         if (nla == NULL)
3993                 goto nla_put_failure;
3994         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3995
3996         /* XXX - MC not implemented */
3997
3998         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3999         if (nla == NULL)
4000                 goto nla_put_failure;
4001         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4002
4003         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4004         if (nla == NULL)
4005                 goto nla_put_failure;
4006         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4007
4008         return 0;
4009
4010 nla_put_failure:
4011         return -EMSGSIZE;
4012 }
4013
4014 static size_t inet6_get_link_af_size(const struct net_device *dev)
4015 {
4016         if (!__in6_dev_get(dev))
4017                 return 0;
4018
4019         return inet6_ifla6_size();
4020 }
4021
4022 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4023 {
4024         struct inet6_dev *idev = __in6_dev_get(dev);
4025
4026         if (!idev)
4027                 return -ENODATA;
4028
4029         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4030                 return -EMSGSIZE;
4031
4032         return 0;
4033 }
4034
4035 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4036                              u32 pid, u32 seq, int event, unsigned int flags)
4037 {
4038         struct net_device *dev = idev->dev;
4039         struct ifinfomsg *hdr;
4040         struct nlmsghdr *nlh;
4041         void *protoinfo;
4042
4043         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4044         if (nlh == NULL)
4045                 return -EMSGSIZE;
4046
4047         hdr = nlmsg_data(nlh);
4048         hdr->ifi_family = AF_INET6;
4049         hdr->__ifi_pad = 0;
4050         hdr->ifi_type = dev->type;
4051         hdr->ifi_index = dev->ifindex;
4052         hdr->ifi_flags = dev_get_flags(dev);
4053         hdr->ifi_change = 0;
4054
4055         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4056
4057         if (dev->addr_len)
4058                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4059
4060         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4061         if (dev->ifindex != dev->iflink)
4062                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4063
4064         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4065         if (protoinfo == NULL)
4066                 goto nla_put_failure;
4067
4068         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4069                 goto nla_put_failure;
4070
4071         nla_nest_end(skb, protoinfo);
4072         return nlmsg_end(skb, nlh);
4073
4074 nla_put_failure:
4075         nlmsg_cancel(skb, nlh);
4076         return -EMSGSIZE;
4077 }
4078
4079 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4080 {
4081         struct net *net = sock_net(skb->sk);
4082         int h, s_h;
4083         int idx = 0, s_idx;
4084         struct net_device *dev;
4085         struct inet6_dev *idev;
4086         struct hlist_head *head;
4087         struct hlist_node *node;
4088
4089         s_h = cb->args[0];
4090         s_idx = cb->args[1];
4091
4092         rcu_read_lock();
4093         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4094                 idx = 0;
4095                 head = &net->dev_index_head[h];
4096                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4097                         if (idx < s_idx)
4098                                 goto cont;
4099                         idev = __in6_dev_get(dev);
4100                         if (!idev)
4101                                 goto cont;
4102                         if (inet6_fill_ifinfo(skb, idev,
4103                                               NETLINK_CB(cb->skb).pid,
4104                                               cb->nlh->nlmsg_seq,
4105                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4106                                 goto out;
4107 cont:
4108                         idx++;
4109                 }
4110         }
4111 out:
4112         rcu_read_unlock();
4113         cb->args[1] = idx;
4114         cb->args[0] = h;
4115
4116         return skb->len;
4117 }
4118
4119 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4120 {
4121         struct sk_buff *skb;
4122         struct net *net = dev_net(idev->dev);
4123         int err = -ENOBUFS;
4124
4125         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4126         if (skb == NULL)
4127                 goto errout;
4128
4129         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4130         if (err < 0) {
4131                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4132                 WARN_ON(err == -EMSGSIZE);
4133                 kfree_skb(skb);
4134                 goto errout;
4135         }
4136         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4137         return;
4138 errout:
4139         if (err < 0)
4140                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4141 }
4142
4143 static inline size_t inet6_prefix_nlmsg_size(void)
4144 {
4145         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4146                + nla_total_size(sizeof(struct in6_addr))
4147                + nla_total_size(sizeof(struct prefix_cacheinfo));
4148 }
4149
4150 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4151                              struct prefix_info *pinfo, u32 pid, u32 seq,
4152                              int event, unsigned int flags)
4153 {
4154         struct prefixmsg *pmsg;
4155         struct nlmsghdr *nlh;
4156         struct prefix_cacheinfo ci;
4157
4158         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4159         if (nlh == NULL)
4160                 return -EMSGSIZE;
4161
4162         pmsg = nlmsg_data(nlh);
4163         pmsg->prefix_family = AF_INET6;
4164         pmsg->prefix_pad1 = 0;
4165         pmsg->prefix_pad2 = 0;
4166         pmsg->prefix_ifindex = idev->dev->ifindex;
4167         pmsg->prefix_len = pinfo->prefix_len;
4168         pmsg->prefix_type = pinfo->type;
4169         pmsg->prefix_pad3 = 0;
4170         pmsg->prefix_flags = 0;
4171         if (pinfo->onlink)
4172                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4173         if (pinfo->autoconf)
4174                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4175
4176         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4177
4178         ci.preferred_time = ntohl(pinfo->prefered);
4179         ci.valid_time = ntohl(pinfo->valid);
4180         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4181
4182         return nlmsg_end(skb, nlh);
4183
4184 nla_put_failure:
4185         nlmsg_cancel(skb, nlh);
4186         return -EMSGSIZE;
4187 }
4188
4189 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4190                          struct prefix_info *pinfo)
4191 {
4192         struct sk_buff *skb;
4193         struct net *net = dev_net(idev->dev);
4194         int err = -ENOBUFS;
4195
4196         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4197         if (skb == NULL)
4198                 goto errout;
4199
4200         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4201         if (err < 0) {
4202                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4203                 WARN_ON(err == -EMSGSIZE);
4204                 kfree_skb(skb);
4205                 goto errout;
4206         }
4207         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4208         return;
4209 errout:
4210         if (err < 0)
4211                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4212 }
4213
4214 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4215 {
4216         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4217
4218         switch (event) {
4219         case RTM_NEWADDR:
4220                 /*
4221                  * If the address was optimistic
4222                  * we inserted the route at the start of
4223                  * our DAD process, so we don't need
4224                  * to do it again
4225                  */
4226                 if (!(ifp->rt->rt6i_node))
4227                         ip6_ins_rt(ifp->rt);
4228                 if (ifp->idev->cnf.forwarding)
4229                         addrconf_join_anycast(ifp);
4230                 break;
4231         case RTM_DELADDR:
4232                 if (ifp->idev->cnf.forwarding)
4233                         addrconf_leave_anycast(ifp);
4234                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4235                 dst_hold(&ifp->rt->dst);
4236
4237                 if (ip6_del_rt(ifp->rt))
4238                         dst_free(&ifp->rt->dst);
4239                 break;
4240         }
4241 }
4242
4243 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4244 {
4245         rcu_read_lock_bh();
4246         if (likely(ifp->idev->dead == 0))
4247                 __ipv6_ifa_notify(event, ifp);
4248         rcu_read_unlock_bh();
4249 }
4250
4251 #ifdef CONFIG_SYSCTL
4252
4253 static
4254 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4255                            void __user *buffer, size_t *lenp, loff_t *ppos)
4256 {
4257         int *valp = ctl->data;
4258         int val = *valp;
4259         loff_t pos = *ppos;
4260         int ret;
4261
4262         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4263
4264         if (write)
4265                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4266         if (ret)
4267                 *ppos = pos;
4268         return ret;
4269 }
4270
4271 static void dev_disable_change(struct inet6_dev *idev)
4272 {
4273         if (!idev || !idev->dev)
4274                 return;
4275
4276         if (idev->cnf.disable_ipv6)
4277                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4278         else
4279                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4280 }
4281
4282 static void addrconf_disable_change(struct net *net, __s32 newf)
4283 {
4284         struct net_device *dev;
4285         struct inet6_dev *idev;
4286
4287         rcu_read_lock();
4288         for_each_netdev_rcu(net, dev) {
4289                 idev = __in6_dev_get(dev);
4290                 if (idev) {
4291                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4292                         idev->cnf.disable_ipv6 = newf;
4293                         if (changed)
4294                                 dev_disable_change(idev);
4295                 }
4296         }
4297         rcu_read_unlock();
4298 }
4299
4300 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4301 {
4302         struct net *net;
4303
4304         net = (struct net *)table->extra2;
4305
4306         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4307                 return 0;
4308
4309         if (!rtnl_trylock()) {
4310                 /* Restore the original values before restarting */
4311                 *p = old;
4312                 return restart_syscall();
4313         }
4314
4315         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4316                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4317                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4318                 addrconf_disable_change(net, newf);
4319         } else if ((!*p) ^ (!old))
4320                 dev_disable_change((struct inet6_dev *)table->extra1);
4321
4322         rtnl_unlock();
4323         return 0;
4324 }
4325
4326 static
4327 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4328                             void __user *buffer, size_t *lenp, loff_t *ppos)
4329 {
4330         int *valp = ctl->data;
4331         int val = *valp;
4332         loff_t pos = *ppos;
4333         int ret;
4334
4335         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4336
4337         if (write)
4338                 ret = addrconf_disable_ipv6(ctl, valp, val);
4339         if (ret)
4340                 *ppos = pos;
4341         return ret;
4342 }
4343
4344 static struct addrconf_sysctl_table
4345 {
4346         struct ctl_table_header *sysctl_header;
4347         ctl_table addrconf_vars[DEVCONF_MAX+1];
4348         char *dev_name;
4349 } addrconf_sysctl __read_mostly = {
4350         .sysctl_header = NULL,
4351         .addrconf_vars = {
4352                 {
4353                         .procname       = "forwarding",
4354                         .data           = &ipv6_devconf.forwarding,
4355                         .maxlen         = sizeof(int),
4356                         .mode           = 0644,
4357                         .proc_handler   = addrconf_sysctl_forward,
4358                 },
4359                 {
4360                         .procname       = "hop_limit",
4361                         .data           = &ipv6_devconf.hop_limit,
4362                         .maxlen         = sizeof(int),
4363                         .mode           = 0644,
4364                         .proc_handler   = proc_dointvec,
4365                 },
4366                 {
4367                         .procname       = "mtu",
4368                         .data           = &ipv6_devconf.mtu6,
4369                         .maxlen         = sizeof(int),
4370                         .mode           = 0644,
4371                         .proc_handler   = proc_dointvec,
4372                 },
4373                 {
4374                         .procname       = "accept_ra",
4375                         .data           = &ipv6_devconf.accept_ra,
4376                         .maxlen         = sizeof(int),
4377                         .mode           = 0644,
4378                         .proc_handler   = proc_dointvec,
4379                 },
4380                 {
4381                         .procname       = "accept_redirects",
4382                         .data           = &ipv6_devconf.accept_redirects,
4383                         .maxlen         = sizeof(int),
4384                         .mode           = 0644,
4385                         .proc_handler   = proc_dointvec,
4386                 },
4387                 {
4388                         .procname       = "autoconf",
4389                         .data           = &ipv6_devconf.autoconf,
4390                         .maxlen         = sizeof(int),
4391                         .mode           = 0644,
4392                         .proc_handler   = proc_dointvec,
4393                 },
4394                 {
4395                         .procname       = "dad_transmits",
4396                         .data           = &ipv6_devconf.dad_transmits,
4397                         .maxlen         = sizeof(int),
4398                         .mode           = 0644,
4399                         .proc_handler   = proc_dointvec,
4400                 },
4401                 {
4402                         .procname       = "router_solicitations",
4403                         .data           = &ipv6_devconf.rtr_solicits,
4404                         .maxlen         = sizeof(int),
4405                         .mode           = 0644,
4406                         .proc_handler   = proc_dointvec,
4407                 },
4408                 {
4409                         .procname       = "router_solicitation_interval",
4410                         .data           = &ipv6_devconf.rtr_solicit_interval,
4411                         .maxlen         = sizeof(int),
4412                         .mode           = 0644,
4413                         .proc_handler   = proc_dointvec_jiffies,
4414                 },
4415                 {
4416                         .procname       = "router_solicitation_delay",
4417                         .data           = &ipv6_devconf.rtr_solicit_delay,
4418                         .maxlen         = sizeof(int),
4419                         .mode           = 0644,
4420                         .proc_handler   = proc_dointvec_jiffies,
4421                 },
4422                 {
4423                         .procname       = "force_mld_version",
4424                         .data           = &ipv6_devconf.force_mld_version,
4425                         .maxlen         = sizeof(int),
4426                         .mode           = 0644,
4427                         .proc_handler   = proc_dointvec,
4428                 },
4429 #ifdef CONFIG_IPV6_PRIVACY
4430                 {
4431                         .procname       = "use_tempaddr",
4432                         .data           = &ipv6_devconf.use_tempaddr,
4433                         .maxlen         = sizeof(int),
4434                         .mode           = 0644,
4435                         .proc_handler   = proc_dointvec,
4436                 },
4437                 {
4438                         .procname       = "temp_valid_lft",
4439                         .data           = &ipv6_devconf.temp_valid_lft,
4440                         .maxlen         = sizeof(int),
4441                         .mode           = 0644,
4442                         .proc_handler   = proc_dointvec,
4443                 },
4444                 {
4445                         .procname       = "temp_prefered_lft",
4446                         .data           = &ipv6_devconf.temp_prefered_lft,
4447                         .maxlen         = sizeof(int),
4448                         .mode           = 0644,
4449                         .proc_handler   = proc_dointvec,
4450                 },
4451                 {
4452                         .procname       = "regen_max_retry",
4453                         .data           = &ipv6_devconf.regen_max_retry,
4454                         .maxlen         = sizeof(int),
4455                         .mode           = 0644,
4456                         .proc_handler   = proc_dointvec,
4457                 },
4458                 {
4459                         .procname       = "max_desync_factor",
4460                         .data           = &ipv6_devconf.max_desync_factor,
4461                         .maxlen         = sizeof(int),
4462                         .mode           = 0644,
4463                         .proc_handler   = proc_dointvec,
4464                 },
4465 #endif
4466                 {
4467                         .procname       = "max_addresses",
4468                         .data           = &ipv6_devconf.max_addresses,
4469                         .maxlen         = sizeof(int),
4470                         .mode           = 0644,
4471                         .proc_handler   = proc_dointvec,
4472                 },
4473                 {
4474                         .procname       = "accept_ra_defrtr",
4475                         .data           = &ipv6_devconf.accept_ra_defrtr,
4476                         .maxlen         = sizeof(int),
4477                         .mode           = 0644,
4478                         .proc_handler   = proc_dointvec,
4479                 },
4480                 {
4481                         .procname       = "accept_ra_pinfo",
4482                         .data           = &ipv6_devconf.accept_ra_pinfo,
4483                         .maxlen         = sizeof(int),
4484                         .mode           = 0644,
4485                         .proc_handler   = proc_dointvec,
4486                 },
4487 #ifdef CONFIG_IPV6_ROUTER_PREF
4488                 {
4489                         .procname       = "accept_ra_rtr_pref",
4490                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4491                         .maxlen         = sizeof(int),
4492                         .mode           = 0644,
4493                         .proc_handler   = proc_dointvec,
4494                 },
4495                 {
4496                         .procname       = "router_probe_interval",
4497                         .data           = &ipv6_devconf.rtr_probe_interval,
4498                         .maxlen         = sizeof(int),
4499                         .mode           = 0644,
4500                         .proc_handler   = proc_dointvec_jiffies,
4501                 },
4502 #ifdef CONFIG_IPV6_ROUTE_INFO
4503                 {
4504                         .procname       = "accept_ra_rt_info_max_plen",
4505                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4506                         .maxlen         = sizeof(int),
4507                         .mode           = 0644,
4508                         .proc_handler   = proc_dointvec,
4509                 },
4510 #endif
4511 #endif
4512                 {
4513                         .procname       = "proxy_ndp",
4514                         .data           = &ipv6_devconf.proxy_ndp,
4515                         .maxlen         = sizeof(int),
4516                         .mode           = 0644,
4517                         .proc_handler   = proc_dointvec,
4518                 },
4519                 {
4520                         .procname       = "accept_source_route",
4521                         .data           = &ipv6_devconf.accept_source_route,
4522                         .maxlen         = sizeof(int),
4523                         .mode           = 0644,
4524                         .proc_handler   = proc_dointvec,
4525                 },
4526 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4527                 {
4528                         .procname       = "optimistic_dad",
4529                         .data           = &ipv6_devconf.optimistic_dad,
4530                         .maxlen         = sizeof(int),
4531                         .mode           = 0644,
4532                         .proc_handler   = proc_dointvec,
4533
4534                 },
4535 #endif
4536 #ifdef CONFIG_IPV6_MROUTE
4537                 {
4538                         .procname       = "mc_forwarding",
4539                         .data           = &ipv6_devconf.mc_forwarding,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0444,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544 #endif
4545                 {
4546                         .procname       = "disable_ipv6",
4547                         .data           = &ipv6_devconf.disable_ipv6,
4548                         .maxlen         = sizeof(int),
4549                         .mode           = 0644,
4550                         .proc_handler   = addrconf_sysctl_disable,
4551                 },
4552                 {
4553                         .procname       = "accept_dad",
4554                         .data           = &ipv6_devconf.accept_dad,
4555                         .maxlen         = sizeof(int),
4556                         .mode           = 0644,
4557                         .proc_handler   = proc_dointvec,
4558                 },
4559                 {
4560                         .procname       = "force_tllao",
4561                         .data           = &ipv6_devconf.force_tllao,
4562                         .maxlen         = sizeof(int),
4563                         .mode           = 0644,
4564                         .proc_handler   = proc_dointvec
4565                 },
4566                 {
4567                         /* sentinel */
4568                 }
4569         },
4570 };
4571
4572 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4573                 struct inet6_dev *idev, struct ipv6_devconf *p)
4574 {
4575         int i;
4576         struct addrconf_sysctl_table *t;
4577
4578 #define ADDRCONF_CTL_PATH_DEV   3
4579
4580         struct ctl_path addrconf_ctl_path[] = {
4581                 { .procname = "net", },
4582                 { .procname = "ipv6", },
4583                 { .procname = "conf", },
4584                 { /* to be set */ },
4585                 { },
4586         };
4587
4588
4589         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4590         if (t == NULL)
4591                 goto out;
4592
4593         for (i = 0; t->addrconf_vars[i].data; i++) {
4594                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4595                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4596                 t->addrconf_vars[i].extra2 = net;
4597         }
4598
4599         /*
4600          * Make a copy of dev_name, because '.procname' is regarded as const
4601          * by sysctl and we wouldn't want anyone to change it under our feet
4602          * (see SIOCSIFNAME).
4603          */
4604         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4605         if (!t->dev_name)
4606                 goto free;
4607
4608         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4609
4610         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4611                         t->addrconf_vars);
4612         if (t->sysctl_header == NULL)
4613                 goto free_procname;
4614
4615         p->sysctl = t;
4616         return 0;
4617
4618 free_procname:
4619         kfree(t->dev_name);
4620 free:
4621         kfree(t);
4622 out:
4623         return -ENOBUFS;
4624 }
4625
4626 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4627 {
4628         struct addrconf_sysctl_table *t;
4629
4630         if (p->sysctl == NULL)
4631                 return;
4632
4633         t = p->sysctl;
4634         p->sysctl = NULL;
4635         unregister_net_sysctl_table(t->sysctl_header);
4636         kfree(t->dev_name);
4637         kfree(t);
4638 }
4639
4640 static void addrconf_sysctl_register(struct inet6_dev *idev)
4641 {
4642         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4643                               &ndisc_ifinfo_sysctl_change);
4644         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4645                                         idev, &idev->cnf);
4646 }
4647
4648 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4649 {
4650         __addrconf_sysctl_unregister(&idev->cnf);
4651         neigh_sysctl_unregister(idev->nd_parms);
4652 }
4653
4654
4655 #endif
4656
4657 static int __net_init addrconf_init_net(struct net *net)
4658 {
4659         int err;
4660         struct ipv6_devconf *all, *dflt;
4661
4662         err = -ENOMEM;
4663         all = &ipv6_devconf;
4664         dflt = &ipv6_devconf_dflt;
4665
4666         if (!net_eq(net, &init_net)) {
4667                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4668                 if (all == NULL)
4669                         goto err_alloc_all;
4670
4671                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4672                 if (dflt == NULL)
4673                         goto err_alloc_dflt;
4674         } else {
4675                 /* these will be inherited by all namespaces */
4676                 dflt->autoconf = ipv6_defaults.autoconf;
4677                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4678         }
4679
4680         net->ipv6.devconf_all = all;
4681         net->ipv6.devconf_dflt = dflt;
4682
4683 #ifdef CONFIG_SYSCTL
4684         err = __addrconf_sysctl_register(net, "all", NULL, all);
4685         if (err < 0)
4686                 goto err_reg_all;
4687
4688         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4689         if (err < 0)
4690                 goto err_reg_dflt;
4691 #endif
4692         return 0;
4693
4694 #ifdef CONFIG_SYSCTL
4695 err_reg_dflt:
4696         __addrconf_sysctl_unregister(all);
4697 err_reg_all:
4698         kfree(dflt);
4699 #endif
4700 err_alloc_dflt:
4701         kfree(all);
4702 err_alloc_all:
4703         return err;
4704 }
4705
4706 static void __net_exit addrconf_exit_net(struct net *net)
4707 {
4708 #ifdef CONFIG_SYSCTL
4709         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4710         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4711 #endif
4712         if (!net_eq(net, &init_net)) {
4713                 kfree(net->ipv6.devconf_dflt);
4714                 kfree(net->ipv6.devconf_all);
4715         }
4716 }
4717
4718 static struct pernet_operations addrconf_ops = {
4719         .init = addrconf_init_net,
4720         .exit = addrconf_exit_net,
4721 };
4722
4723 /*
4724  *      Device notifier
4725  */
4726
4727 int register_inet6addr_notifier(struct notifier_block *nb)
4728 {
4729         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4730 }
4731 EXPORT_SYMBOL(register_inet6addr_notifier);
4732
4733 int unregister_inet6addr_notifier(struct notifier_block *nb)
4734 {
4735         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4736 }
4737 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4738
4739 static struct rtnl_af_ops inet6_ops = {
4740         .family           = AF_INET6,
4741         .fill_link_af     = inet6_fill_link_af,
4742         .get_link_af_size = inet6_get_link_af_size,
4743 };
4744
4745 /*
4746  *      Init / cleanup code
4747  */
4748
4749 int __init addrconf_init(void)
4750 {
4751         int i, err;
4752
4753         err = ipv6_addr_label_init();
4754         if (err < 0) {
4755                 printk(KERN_CRIT "IPv6 Addrconf:"
4756                        " cannot initialize default policy table: %d.\n", err);
4757                 goto out;
4758         }
4759
4760         err = register_pernet_subsys(&addrconf_ops);
4761         if (err < 0)
4762                 goto out_addrlabel;
4763
4764         /* The addrconf netdev notifier requires that loopback_dev
4765          * has it's ipv6 private information allocated and setup
4766          * before it can bring up and give link-local addresses
4767          * to other devices which are up.
4768          *
4769          * Unfortunately, loopback_dev is not necessarily the first
4770          * entry in the global dev_base list of net devices.  In fact,
4771          * it is likely to be the very last entry on that list.
4772          * So this causes the notifier registry below to try and
4773          * give link-local addresses to all devices besides loopback_dev
4774          * first, then loopback_dev, which cases all the non-loopback_dev
4775          * devices to fail to get a link-local address.
4776          *
4777          * So, as a temporary fix, allocate the ipv6 structure for
4778          * loopback_dev first by hand.
4779          * Longer term, all of the dependencies ipv6 has upon the loopback
4780          * device and it being up should be removed.
4781          */
4782         rtnl_lock();
4783         if (!ipv6_add_dev(init_net.loopback_dev))
4784                 err = -ENOMEM;
4785         rtnl_unlock();
4786         if (err)
4787                 goto errlo;
4788
4789         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4790                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4791
4792         register_netdevice_notifier(&ipv6_dev_notf);
4793
4794         addrconf_verify(0);
4795
4796         err = rtnl_af_register(&inet6_ops);
4797         if (err < 0)
4798                 goto errout_af;
4799
4800         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4801                               NULL);
4802         if (err < 0)
4803                 goto errout;
4804
4805         /* Only the first call to __rtnl_register can fail */
4806         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4807         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4808         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4809                         inet6_dump_ifaddr, NULL);
4810         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4811                         inet6_dump_ifmcaddr, NULL);
4812         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4813                         inet6_dump_ifacaddr, NULL);
4814
4815         ipv6_addr_label_rtnl_register();
4816
4817         return 0;
4818 errout:
4819         rtnl_af_unregister(&inet6_ops);
4820 errout_af:
4821         unregister_netdevice_notifier(&ipv6_dev_notf);
4822 errlo:
4823         unregister_pernet_subsys(&addrconf_ops);
4824 out_addrlabel:
4825         ipv6_addr_label_cleanup();
4826 out:
4827         return err;
4828 }
4829
4830 void addrconf_cleanup(void)
4831 {
4832         struct net_device *dev;
4833         int i;
4834
4835         unregister_netdevice_notifier(&ipv6_dev_notf);
4836         unregister_pernet_subsys(&addrconf_ops);
4837         ipv6_addr_label_cleanup();
4838
4839         rtnl_lock();
4840
4841         __rtnl_af_unregister(&inet6_ops);
4842
4843         /* clean dev list */
4844         for_each_netdev(&init_net, dev) {
4845                 if (__in6_dev_get(dev) == NULL)
4846                         continue;
4847                 addrconf_ifdown(dev, 1);
4848         }
4849         addrconf_ifdown(init_net.loopback_dev, 2);
4850
4851         /*
4852          *      Check hash table.
4853          */
4854         spin_lock_bh(&addrconf_hash_lock);
4855         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4856                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4857         spin_unlock_bh(&addrconf_hash_lock);
4858
4859         del_timer(&addr_chk_timer);
4860         rtnl_unlock();
4861 }