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 #define pr_fmt(fmt) "IPv6: " fmt
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/kernel.h>
  46 #include <linux/socket.h>
  47 #include <linux/sockios.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #include <linux/slab.h>
  59 #ifdef CONFIG_SYSCTL
  60 #include <linux/sysctl.h>
  61 #endif
  62 #include <linux/capability.h>
  63 #include <linux/delay.h>
  64 #include <linux/notifier.h>
  65 #include <linux/string.h>
  66
  67 #include <net/net_namespace.h>
  68 #include <net/sock.h>
  69 #include <net/snmp.h>
  70
  71 #include <net/af_ieee802154.h>
  72 #include <net/ipv6.h>
  73 #include <net/protocol.h>
  74 #include <net/ndisc.h>
  75 #include <net/ip6_route.h>
  76 #include <net/addrconf.h>
  77 #include <net/tcp.h>
  78 #include <net/ip.h>
  79 #include <net/netlink.h>
  80 #include <net/pkt_sched.h>
  81 #include <linux/if_tunnel.h>
  82 #include <linux/rtnetlink.h>
  83
  84 #ifdef CONFIG_IPV6_PRIVACY
  85 #include <linux/random.h>
  86 #endif
  87
  88 #include <linux/uaccess.h>
  89 #include <asm/unaligned.h>
  90
  91 #include <linux/proc_fs.h>
  92 #include <linux/seq_file.h>
  93 #include <linux/export.h>
  94
  95 /* Set to 3 to get tracing... */
  96 #define ACONF_DEBUG 2
  97
  98 #if ACONF_DEBUG >= 3
  99 #define ADBG(x) printk x
 100 #else
 101 #define ADBG(x)
 102 #endif
 103
 104 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 105
 106 static inline u32 cstamp_delta(unsigned long cstamp)
 107 {
 108         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 109 }
 110
 111 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 112 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 113 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 114
 115 #ifdef CONFIG_SYSCTL
 116 static void addrconf_sysctl_register(struct inet6_dev *idev);
 117 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 118 #else
 119 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 120 {
 121 }
 122
 123 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 124 {
 125 }
 126 #endif
 127
 128 #ifdef CONFIG_IPV6_PRIVACY
 129 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 130 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 131 static void ipv6_regen_rndid(unsigned long data);
 132 #endif
 133
 134 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 135 static int ipv6_count_addresses(struct inet6_dev *idev);
 136
 137 /*
 138  *      Configured unicast address hash table
 139  */
 140 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 141 static DEFINE_SPINLOCK(addrconf_hash_lock);
 142
 143 static void addrconf_verify(unsigned long);
 144
 145 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 146 static DEFINE_SPINLOCK(addrconf_verify_lock);
 147
 148 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 149 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 150
 151 static void addrconf_type_change(struct net_device *dev,
 152                                  unsigned long event);
 153 static int addrconf_ifdown(struct net_device *dev, int how);
 154
 155 static void addrconf_dad_start(struct inet6_ifaddr *ifp);
 156 static void addrconf_dad_timer(unsigned long data);
 157 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 158 static void addrconf_dad_run(struct inet6_dev *idev);
 159 static void addrconf_rs_timer(unsigned long data);
 160 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 161 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 162
 163 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 164                                 struct prefix_info *pinfo);
 165 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 166                                struct net_device *dev);
 167
 168 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 169
 170 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 171         .forwarding             = 0,
 172         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 173         .mtu6                   = IPV6_MIN_MTU,
 174         .accept_ra              = 1,
 175         .accept_redirects       = 1,
 176         .autoconf               = 1,
 177         .force_mld_version      = 0,
 178         .dad_transmits          = 1,
 179         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 180         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 181         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 182 #ifdef CONFIG_IPV6_PRIVACY
 183         .use_tempaddr           = 0,
 184         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 185         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 186         .regen_max_retry        = REGEN_MAX_RETRY,
 187         .max_desync_factor      = MAX_DESYNC_FACTOR,
 188 #endif
 189         .max_addresses          = IPV6_MAX_ADDRESSES,
 190         .accept_ra_defrtr       = 1,
 191         .accept_ra_pinfo        = 1,
 192 #ifdef CONFIG_IPV6_ROUTER_PREF
 193         .accept_ra_rtr_pref     = 1,
 194         .rtr_probe_interval     = 60 * HZ,
 195 #ifdef CONFIG_IPV6_ROUTE_INFO
 196         .accept_ra_rt_info_max_plen = 0,
 197 #endif
 198 #endif
 199         .proxy_ndp              = 0,
 200         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 201         .disable_ipv6           = 0,
 202         .accept_dad             = 1,
 203 };
 204
 205 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 206         .forwarding             = 0,
 207         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 208         .mtu6                   = IPV6_MIN_MTU,
 209         .accept_ra              = 1,
 210         .accept_redirects       = 1,
 211         .autoconf               = 1,
 212         .dad_transmits          = 1,
 213         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 214         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 215         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 216 #ifdef CONFIG_IPV6_PRIVACY
 217         .use_tempaddr           = 0,
 218         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 219         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 220         .regen_max_retry        = REGEN_MAX_RETRY,
 221         .max_desync_factor      = MAX_DESYNC_FACTOR,
 222 #endif
 223         .max_addresses          = IPV6_MAX_ADDRESSES,
 224         .accept_ra_defrtr       = 1,
 225         .accept_ra_pinfo        = 1,
 226 #ifdef CONFIG_IPV6_ROUTER_PREF
 227         .accept_ra_rtr_pref     = 1,
 228         .rtr_probe_interval     = 60 * HZ,
 229 #ifdef CONFIG_IPV6_ROUTE_INFO
 230         .accept_ra_rt_info_max_plen = 0,
 231 #endif
 232 #endif
 233         .proxy_ndp              = 0,
 234         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 235         .disable_ipv6           = 0,
 236         .accept_dad             = 1,
 237 };
 238
 239 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 240 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 241 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 242 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 243 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 244
 245 /* Check if a valid qdisc is available */
 246 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 247 {
 248         return !qdisc_tx_is_noop(dev);
 249 }
 250
 251 /* Check if a route is valid prefix route */
 252 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 253 {
 254         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 255 }
 256
 257 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 258 {
 259         if (del_timer(&ifp->timer))
 260                 __in6_ifa_put(ifp);
 261 }
 262
 263 enum addrconf_timer_t {
 264         AC_NONE,
 265         AC_DAD,
 266         AC_RS,
 267 };
 268
 269 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 270                                enum addrconf_timer_t what,
 271                                unsigned long when)
 272 {
 273         if (!del_timer(&ifp->timer))
 274                 in6_ifa_hold(ifp);
 275
 276         switch (what) {
 277         case AC_DAD:
 278                 ifp->timer.function = addrconf_dad_timer;
 279                 break;
 280         case AC_RS:
 281                 ifp->timer.function = addrconf_rs_timer;
 282                 break;
 283         default:
 284                 break;
 285         }
 286         ifp->timer.expires = jiffies + when;
 287         add_timer(&ifp->timer);
 288 }
 289
 290 static int snmp6_alloc_dev(struct inet6_dev *idev)
 291 {
 292         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 293                           sizeof(struct ipstats_mib),
 294                           __alignof__(struct ipstats_mib)) < 0)
 295                 goto err_ip;
 296         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 297                                         GFP_KERNEL);
 298         if (!idev->stats.icmpv6dev)
 299                 goto err_icmp;
 300         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 301                                            GFP_KERNEL);
 302         if (!idev->stats.icmpv6msgdev)
 303                 goto err_icmpmsg;
 304
 305         return 0;
 306
 307 err_icmpmsg:
 308         kfree(idev->stats.icmpv6dev);
 309 err_icmp:
 310         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 311 err_ip:
 312         return -ENOMEM;
 313 }
 314
 315 static void snmp6_free_dev(struct inet6_dev *idev)
 316 {
 317         kfree(idev->stats.icmpv6msgdev);
 318         kfree(idev->stats.icmpv6dev);
 319         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 320 }
 321
 322 /* Nobody refers to this device, we may destroy it. */
 323
 324 void in6_dev_finish_destroy(struct inet6_dev *idev)
 325 {
 326         struct net_device *dev = idev->dev;
 327
 328         WARN_ON(!list_empty(&idev->addr_list));
 329         WARN_ON(idev->mc_list != NULL);
 330
 331 #ifdef NET_REFCNT_DEBUG
 332         pr_debug("%s: %s\n", __func__, dev ? dev->name : "NIL");
 333 #endif
 334         dev_put(dev);
 335         if (!idev->dead) {
 336                 pr_warn("Freeing alive inet6 device %p\n", idev);
 337                 return;
 338         }
 339         snmp6_free_dev(idev);
 340         kfree_rcu(idev, rcu);
 341 }
 342 EXPORT_SYMBOL(in6_dev_finish_destroy);
 343
 344 static struct inet6_dev *ipv6_add_dev(struct net_device *dev)
 345 {
 346         struct inet6_dev *ndev;
 347
 348         ASSERT_RTNL();
 349
 350         if (dev->mtu < IPV6_MIN_MTU)
 351                 return NULL;
 352
 353         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 354
 355         if (ndev == NULL)
 356                 return NULL;
 357
 358         rwlock_init(&ndev->lock);
 359         ndev->dev = dev;
 360         INIT_LIST_HEAD(&ndev->addr_list);
 361
 362         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 363         ndev->cnf.mtu6 = dev->mtu;
 364         ndev->cnf.sysctl = NULL;
 365         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 366         if (ndev->nd_parms == NULL) {
 367                 kfree(ndev);
 368                 return NULL;
 369         }
 370         if (ndev->cnf.forwarding)
 371                 dev_disable_lro(dev);
 372         /* We refer to the device */
 373         dev_hold(dev);
 374
 375         if (snmp6_alloc_dev(ndev) < 0) {
 376                 ADBG((KERN_WARNING
 377                         "%s: cannot allocate memory for statistics; dev=%s.\n",
 378                         __func__, dev->name));
 379                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 380                 dev_put(dev);
 381                 kfree(ndev);
 382                 return NULL;
 383         }
 384
 385         if (snmp6_register_dev(ndev) < 0) {
 386                 ADBG((KERN_WARNING
 387                         "%s: cannot create /proc/net/dev_snmp6/%s\n",
 388                         __func__, dev->name));
 389                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 390                 ndev->dead = 1;
 391                 in6_dev_finish_destroy(ndev);
 392                 return NULL;
 393         }
 394
 395         /* One reference from device.  We must do this before
 396          * we invoke __ipv6_regen_rndid().
 397          */
 398         in6_dev_hold(ndev);
 399
 400         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 401                 ndev->cnf.accept_dad = -1;
 402
 403 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 404         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 405                 pr_info("%s: Disabled Multicast RS\n", 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_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev *ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         rcu_read_lock();
 497         for_each_netdev_rcu(net, dev) {
 498                 idev = __in6_dev_get(dev);
 499                 if (idev) {
 500                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 501                         idev->cnf.forwarding = newf;
 502                         if (changed)
 503                                 dev_forward_change(idev);
 504                 }
 505         }
 506         rcu_read_unlock();
 507 }
 508
 509 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf)
 510 {
 511         struct net *net;
 512         int old;
 513
 514         if (!rtnl_trylock())
 515                 return restart_syscall();
 516
 517         net = (struct net *)table->extra2;
 518         old = *p;
 519         *p = newf;
 520
 521         if (p == &net->ipv6.devconf_dflt->forwarding) {
 522                 rtnl_unlock();
 523                 return 0;
 524         }
 525
 526         if (p == &net->ipv6.devconf_all->forwarding) {
 527                 net->ipv6.devconf_dflt->forwarding = newf;
 528                 addrconf_forward_change(net, newf);
 529         } else if ((!newf) ^ (!old))
 530                 dev_forward_change((struct inet6_dev *)table->extra1);
 531         rtnl_unlock();
 532
 533         if (newf)
 534                 rt6_purge_dflt_routers(net);
 535         return 1;
 536 }
 537 #endif
 538
 539 /* Nobody refers to this ifaddr, destroy it */
 540 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 541 {
 542         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 543
 544 #ifdef NET_REFCNT_DEBUG
 545         pr_debug("%s\n", __func__);
 546 #endif
 547
 548         in6_dev_put(ifp->idev);
 549
 550         if (del_timer(&ifp->timer))
 551                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 552
 553         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 554                 pr_warn("Freeing alive inet6 address %p\n", ifp);
 555                 return;
 556         }
 557         dst_release(&ifp->rt->dst);
 558
 559         kfree_rcu(ifp, rcu);
 560 }
 561
 562 static void
 563 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 564 {
 565         struct list_head *p;
 566         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 567
 568         /*
 569          * Each device address list is sorted in order of scope -
 570          * global before linklocal.
 571          */
 572         list_for_each(p, &idev->addr_list) {
 573                 struct inet6_ifaddr *ifa
 574                         = list_entry(p, struct inet6_ifaddr, if_list);
 575                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 576                         break;
 577         }
 578
 579         list_add_tail(&ifp->if_list, p);
 580 }
 581
 582 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 583 {
 584         /*
 585          * We perform the hash function over the last 64 bits of the address
 586          * This will include the IEEE address token on links that support it.
 587          */
 588         return jhash_2words((__force u32)addr->s6_addr32[2],
 589                             (__force u32)addr->s6_addr32[3], 0)
 590                 & (IN6_ADDR_HSIZE - 1);
 591 }
 592
 593 /* On success it returns ifp with increased reference count */
 594
 595 static struct inet6_ifaddr *
 596 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 597               int scope, u32 flags)
 598 {
 599         struct inet6_ifaddr *ifa = NULL;
 600         struct rt6_info *rt;
 601         unsigned int hash;
 602         int err = 0;
 603         int addr_type = ipv6_addr_type(addr);
 604
 605         if (addr_type == IPV6_ADDR_ANY ||
 606             addr_type & IPV6_ADDR_MULTICAST ||
 607             (!(idev->dev->flags & IFF_LOOPBACK) &&
 608              addr_type & IPV6_ADDR_LOOPBACK))
 609                 return ERR_PTR(-EADDRNOTAVAIL);
 610
 611         rcu_read_lock_bh();
 612         if (idev->dead) {
 613                 err = -ENODEV;                  /*XXX*/
 614                 goto out2;
 615         }
 616
 617         if (idev->cnf.disable_ipv6) {
 618                 err = -EACCES;
 619                 goto out2;
 620         }
 621
 622         spin_lock(&addrconf_hash_lock);
 623
 624         /* Ignore adding duplicate addresses on an interface */
 625         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 626                 ADBG(("ipv6_add_addr: already assigned\n"));
 627                 err = -EEXIST;
 628                 goto out;
 629         }
 630
 631         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 632
 633         if (ifa == NULL) {
 634                 ADBG(("ipv6_add_addr: malloc failed\n"));
 635                 err = -ENOBUFS;
 636                 goto out;
 637         }
 638
 639         rt = addrconf_dst_alloc(idev, addr, false);
 640         if (IS_ERR(rt)) {
 641                 err = PTR_ERR(rt);
 642                 goto out;
 643         }
 644
 645         ifa->addr = *addr;
 646
 647         spin_lock_init(&ifa->lock);
 648         spin_lock_init(&ifa->state_lock);
 649         init_timer(&ifa->timer);
 650         INIT_HLIST_NODE(&ifa->addr_lst);
 651         ifa->timer.data = (unsigned long) ifa;
 652         ifa->scope = scope;
 653         ifa->prefix_len = pfxlen;
 654         ifa->flags = flags | IFA_F_TENTATIVE;
 655         ifa->cstamp = ifa->tstamp = jiffies;
 656
 657         ifa->rt = rt;
 658
 659         ifa->idev = idev;
 660         in6_dev_hold(idev);
 661         /* For caller */
 662         in6_ifa_hold(ifa);
 663
 664         /* Add to big hash table */
 665         hash = ipv6_addr_hash(addr);
 666
 667         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 668         spin_unlock(&addrconf_hash_lock);
 669
 670         write_lock(&idev->lock);
 671         /* Add to inet6_dev unicast addr list. */
 672         ipv6_link_dev_addr(idev, ifa);
 673
 674 #ifdef CONFIG_IPV6_PRIVACY
 675         if (ifa->flags&IFA_F_TEMPORARY) {
 676                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 677                 in6_ifa_hold(ifa);
 678         }
 679 #endif
 680
 681         in6_ifa_hold(ifa);
 682         write_unlock(&idev->lock);
 683 out2:
 684         rcu_read_unlock_bh();
 685
 686         if (likely(err == 0))
 687                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 688         else {
 689                 kfree(ifa);
 690                 ifa = ERR_PTR(err);
 691         }
 692
 693         return ifa;
 694 out:
 695         spin_unlock(&addrconf_hash_lock);
 696         goto out2;
 697 }
 698
 699 /* This function wants to get referenced ifp and releases it before return */
 700
 701 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 702 {
 703         struct inet6_ifaddr *ifa, *ifn;
 704         struct inet6_dev *idev = ifp->idev;
 705         int state;
 706         int deleted = 0, onlink = 0;
 707         unsigned long expires = jiffies;
 708
 709         spin_lock_bh(&ifp->state_lock);
 710         state = ifp->state;
 711         ifp->state = INET6_IFADDR_STATE_DEAD;
 712         spin_unlock_bh(&ifp->state_lock);
 713
 714         if (state == INET6_IFADDR_STATE_DEAD)
 715                 goto out;
 716
 717         spin_lock_bh(&addrconf_hash_lock);
 718         hlist_del_init_rcu(&ifp->addr_lst);
 719         spin_unlock_bh(&addrconf_hash_lock);
 720
 721         write_lock_bh(&idev->lock);
 722 #ifdef CONFIG_IPV6_PRIVACY
 723         if (ifp->flags&IFA_F_TEMPORARY) {
 724                 list_del(&ifp->tmp_list);
 725                 if (ifp->ifpub) {
 726                         in6_ifa_put(ifp->ifpub);
 727                         ifp->ifpub = NULL;
 728                 }
 729                 __in6_ifa_put(ifp);
 730         }
 731 #endif
 732
 733         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 734                 if (ifa == ifp) {
 735                         list_del_init(&ifp->if_list);
 736                         __in6_ifa_put(ifp);
 737
 738                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 739                                 break;
 740                         deleted = 1;
 741                         continue;
 742                 } else if (ifp->flags & IFA_F_PERMANENT) {
 743                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 744                                               ifp->prefix_len)) {
 745                                 if (ifa->flags & IFA_F_PERMANENT) {
 746                                         onlink = 1;
 747                                         if (deleted)
 748                                                 break;
 749                                 } else {
 750                                         unsigned long lifetime;
 751
 752                                         if (!onlink)
 753                                                 onlink = -1;
 754
 755                                         spin_lock(&ifa->lock);
 756
 757                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 758                                         /*
 759                                          * Note: Because this address is
 760                                          * not permanent, lifetime <
 761                                          * LONG_MAX / HZ here.
 762                                          */
 763                                         if (time_before(expires,
 764                                                         ifa->tstamp + lifetime * HZ))
 765                                                 expires = ifa->tstamp + lifetime * HZ;
 766                                         spin_unlock(&ifa->lock);
 767                                 }
 768                         }
 769                 }
 770         }
 771         write_unlock_bh(&idev->lock);
 772
 773         addrconf_del_timer(ifp);
 774
 775         ipv6_ifa_notify(RTM_DELADDR, ifp);
 776
 777         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 778
 779         /*
 780          * Purge or update corresponding prefix
 781          *
 782          * 1) we don't purge prefix here if address was not permanent.
 783          *    prefix is managed by its own lifetime.
 784          * 2) if there're no addresses, delete prefix.
 785          * 3) if there're still other permanent address(es),
 786          *    corresponding prefix is still permanent.
 787          * 4) otherwise, update prefix lifetime to the
 788          *    longest valid lifetime among the corresponding
 789          *    addresses on the device.
 790          *    Note: subsequent RA will update lifetime.
 791          *
 792          * --yoshfuji
 793          */
 794         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 795                 struct in6_addr prefix;
 796                 struct rt6_info *rt;
 797                 struct net *net = dev_net(ifp->idev->dev);
 798                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 799                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 800
 801                 if (rt && addrconf_is_prefix_route(rt)) {
 802                         if (onlink == 0) {
 803                                 ip6_del_rt(rt);
 804                                 rt = NULL;
 805                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 806                                 rt6_set_expires(rt, expires);
 807                         }
 808                 }
 809                 dst_release(&rt->dst);
 810         }
 811
 812         /* clean up prefsrc entries */
 813         rt6_remove_prefsrc(ifp);
 814 out:
 815         in6_ifa_put(ifp);
 816 }
 817
 818 #ifdef CONFIG_IPV6_PRIVACY
 819 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 820 {
 821         struct inet6_dev *idev = ifp->idev;
 822         struct in6_addr addr, *tmpaddr;
 823         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 824         unsigned long regen_advance;
 825         int tmp_plen;
 826         int ret = 0;
 827         int max_addresses;
 828         u32 addr_flags;
 829         unsigned long now = jiffies;
 830
 831         write_lock(&idev->lock);
 832         if (ift) {
 833                 spin_lock_bh(&ift->lock);
 834                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 835                 spin_unlock_bh(&ift->lock);
 836                 tmpaddr = &addr;
 837         } else {
 838                 tmpaddr = NULL;
 839         }
 840 retry:
 841         in6_dev_hold(idev);
 842         if (idev->cnf.use_tempaddr <= 0) {
 843                 write_unlock(&idev->lock);
 844                 pr_info("%s: use_tempaddr is disabled\n", __func__);
 845                 in6_dev_put(idev);
 846                 ret = -1;
 847                 goto out;
 848         }
 849         spin_lock_bh(&ifp->lock);
 850         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 851                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 852                 spin_unlock_bh(&ifp->lock);
 853                 write_unlock(&idev->lock);
 854                 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n",
 855                         __func__);
 856                 in6_dev_put(idev);
 857                 ret = -1;
 858                 goto out;
 859         }
 860         in6_ifa_hold(ifp);
 861         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 862         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 863                 spin_unlock_bh(&ifp->lock);
 864                 write_unlock(&idev->lock);
 865                 pr_warn("%s: regeneration of randomized interface id failed\n",
 866                         __func__);
 867                 in6_ifa_put(ifp);
 868                 in6_dev_put(idev);
 869                 ret = -1;
 870                 goto out;
 871         }
 872         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 873         age = (now - ifp->tstamp) / HZ;
 874         tmp_valid_lft = min_t(__u32,
 875                               ifp->valid_lft,
 876                               idev->cnf.temp_valid_lft + age);
 877         tmp_prefered_lft = min_t(__u32,
 878                                  ifp->prefered_lft,
 879                                  idev->cnf.temp_prefered_lft + age -
 880                                  idev->cnf.max_desync_factor);
 881         tmp_plen = ifp->prefix_len;
 882         max_addresses = idev->cnf.max_addresses;
 883         tmp_tstamp = ifp->tstamp;
 884         spin_unlock_bh(&ifp->lock);
 885
 886         regen_advance = idev->cnf.regen_max_retry *
 887                         idev->cnf.dad_transmits *
 888                         idev->nd_parms->retrans_time / HZ;
 889         write_unlock(&idev->lock);
 890
 891         /* A temporary address is created only if this calculated Preferred
 892          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 893          * an implementation must not create a temporary address with a zero
 894          * Preferred Lifetime.
 895          */
 896         if (tmp_prefered_lft <= regen_advance) {
 897                 in6_ifa_put(ifp);
 898                 in6_dev_put(idev);
 899                 ret = -1;
 900                 goto out;
 901         }
 902
 903         addr_flags = IFA_F_TEMPORARY;
 904         /* set in addrconf_prefix_rcv() */
 905         if (ifp->flags & IFA_F_OPTIMISTIC)
 906                 addr_flags |= IFA_F_OPTIMISTIC;
 907
 908         ift = !max_addresses ||
 909               ipv6_count_addresses(idev) < max_addresses ?
 910                 ipv6_add_addr(idev, &addr, tmp_plen,
 911                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 912                               addr_flags) : NULL;
 913         if (!ift || IS_ERR(ift)) {
 914                 in6_ifa_put(ifp);
 915                 in6_dev_put(idev);
 916                 pr_info("%s: retry temporary address regeneration\n", __func__);
 917                 tmpaddr = &addr;
 918                 write_lock(&idev->lock);
 919                 goto retry;
 920         }
 921
 922         spin_lock_bh(&ift->lock);
 923         ift->ifpub = ifp;
 924         ift->valid_lft = tmp_valid_lft;
 925         ift->prefered_lft = tmp_prefered_lft;
 926         ift->cstamp = now;
 927         ift->tstamp = tmp_tstamp;
 928         spin_unlock_bh(&ift->lock);
 929
 930         addrconf_dad_start(ift);
 931         in6_ifa_put(ift);
 932         in6_dev_put(idev);
 933 out:
 934         return ret;
 935 }
 936 #endif
 937
 938 /*
 939  *      Choose an appropriate source address (RFC3484)
 940  */
 941 enum {
 942         IPV6_SADDR_RULE_INIT = 0,
 943         IPV6_SADDR_RULE_LOCAL,
 944         IPV6_SADDR_RULE_SCOPE,
 945         IPV6_SADDR_RULE_PREFERRED,
 946 #ifdef CONFIG_IPV6_MIP6
 947         IPV6_SADDR_RULE_HOA,
 948 #endif
 949         IPV6_SADDR_RULE_OIF,
 950         IPV6_SADDR_RULE_LABEL,
 951 #ifdef CONFIG_IPV6_PRIVACY
 952         IPV6_SADDR_RULE_PRIVACY,
 953 #endif
 954         IPV6_SADDR_RULE_ORCHID,
 955         IPV6_SADDR_RULE_PREFIX,
 956         IPV6_SADDR_RULE_MAX
 957 };
 958
 959 struct ipv6_saddr_score {
 960         int                     rule;
 961         int                     addr_type;
 962         struct inet6_ifaddr     *ifa;
 963         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 964         int                     scopedist;
 965         int                     matchlen;
 966 };
 967
 968 struct ipv6_saddr_dst {
 969         const struct in6_addr *addr;
 970         int ifindex;
 971         int scope;
 972         int label;
 973         unsigned int prefs;
 974 };
 975
 976 static inline int ipv6_saddr_preferred(int type)
 977 {
 978         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 979                 return 1;
 980         return 0;
 981 }
 982
 983 static int ipv6_get_saddr_eval(struct net *net,
 984                                struct ipv6_saddr_score *score,
 985                                struct ipv6_saddr_dst *dst,
 986                                int i)
 987 {
 988         int ret;
 989
 990         if (i <= score->rule) {
 991                 switch (i) {
 992                 case IPV6_SADDR_RULE_SCOPE:
 993                         ret = score->scopedist;
 994                         break;
 995                 case IPV6_SADDR_RULE_PREFIX:
 996                         ret = score->matchlen;
 997                         break;
 998                 default:
 999                         ret = !!test_bit(i, score->scorebits);
1000                 }
1001                 goto out;
1002         }
1003
1004         switch (i) {
1005         case IPV6_SADDR_RULE_INIT:
1006                 /* Rule 0: remember if hiscore is not ready yet */
1007                 ret = !!score->ifa;
1008                 break;
1009         case IPV6_SADDR_RULE_LOCAL:
1010                 /* Rule 1: Prefer same address */
1011                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1012                 break;
1013         case IPV6_SADDR_RULE_SCOPE:
1014                 /* Rule 2: Prefer appropriate scope
1015                  *
1016                  *      ret
1017                  *       ^
1018                  *    -1 |  d 15
1019                  *    ---+--+-+---> scope
1020                  *       |
1021                  *       |             d is scope of the destination.
1022                  *  B-d  |  \
1023                  *       |   \      <- smaller scope is better if
1024                  *  B-15 |    \        if scope is enough for destinaion.
1025                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1026                  * d-C-1 | /
1027                  *       |/         <- greater is better
1028                  *   -C  /             if scope is not enough for destination.
1029                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1030                  *
1031                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1032                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1033                  * Assume B = 0 and we get C > 29.
1034                  */
1035                 ret = __ipv6_addr_src_scope(score->addr_type);
1036                 if (ret >= dst->scope)
1037                         ret = -ret;
1038                 else
1039                         ret -= 128;     /* 30 is enough */
1040                 score->scopedist = ret;
1041                 break;
1042         case IPV6_SADDR_RULE_PREFERRED:
1043                 /* Rule 3: Avoid deprecated and optimistic addresses */
1044                 ret = ipv6_saddr_preferred(score->addr_type) ||
1045                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1046                 break;
1047 #ifdef CONFIG_IPV6_MIP6
1048         case IPV6_SADDR_RULE_HOA:
1049             {
1050                 /* Rule 4: Prefer home address */
1051                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1052                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1053                 break;
1054             }
1055 #endif
1056         case IPV6_SADDR_RULE_OIF:
1057                 /* Rule 5: Prefer outgoing interface */
1058                 ret = (!dst->ifindex ||
1059                        dst->ifindex == score->ifa->idev->dev->ifindex);
1060                 break;
1061         case IPV6_SADDR_RULE_LABEL:
1062                 /* Rule 6: Prefer matching label */
1063                 ret = ipv6_addr_label(net,
1064                                       &score->ifa->addr, score->addr_type,
1065                                       score->ifa->idev->dev->ifindex) == dst->label;
1066                 break;
1067 #ifdef CONFIG_IPV6_PRIVACY
1068         case IPV6_SADDR_RULE_PRIVACY:
1069             {
1070                 /* Rule 7: Prefer public address
1071                  * Note: prefer temporary address if use_tempaddr >= 2
1072                  */
1073                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1074                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1075                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1076                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1077                 break;
1078             }
1079 #endif
1080         case IPV6_SADDR_RULE_ORCHID:
1081                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1082                  *          non-ORCHID vs non-ORCHID
1083                  */
1084                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1085                         ipv6_addr_orchid(dst->addr));
1086                 break;
1087         case IPV6_SADDR_RULE_PREFIX:
1088                 /* Rule 8: Use longest matching prefix */
1089                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1090                                                        dst->addr);
1091                 break;
1092         default:
1093                 ret = 0;
1094         }
1095
1096         if (ret)
1097                 __set_bit(i, score->scorebits);
1098         score->rule = i;
1099 out:
1100         return ret;
1101 }
1102
1103 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1104                        const struct in6_addr *daddr, unsigned int prefs,
1105                        struct in6_addr *saddr)
1106 {
1107         struct ipv6_saddr_score scores[2],
1108                                 *score = &scores[0], *hiscore = &scores[1];
1109         struct ipv6_saddr_dst dst;
1110         struct net_device *dev;
1111         int dst_type;
1112
1113         dst_type = __ipv6_addr_type(daddr);
1114         dst.addr = daddr;
1115         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1116         dst.scope = __ipv6_addr_src_scope(dst_type);
1117         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1118         dst.prefs = prefs;
1119
1120         hiscore->rule = -1;
1121         hiscore->ifa = NULL;
1122
1123         rcu_read_lock();
1124
1125         for_each_netdev_rcu(net, dev) {
1126                 struct inet6_dev *idev;
1127
1128                 /* Candidate Source Address (section 4)
1129                  *  - multicast and link-local destination address,
1130                  *    the set of candidate source address MUST only
1131                  *    include addresses assigned to interfaces
1132                  *    belonging to the same link as the outgoing
1133                  *    interface.
1134                  * (- For site-local destination addresses, the
1135                  *    set of candidate source addresses MUST only
1136                  *    include addresses assigned to interfaces
1137                  *    belonging to the same site as the outgoing
1138                  *    interface.)
1139                  */
1140                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1141                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1142                     dst.ifindex && dev->ifindex != dst.ifindex)
1143                         continue;
1144
1145                 idev = __in6_dev_get(dev);
1146                 if (!idev)
1147                         continue;
1148
1149                 read_lock_bh(&idev->lock);
1150                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1151                         int i;
1152
1153                         /*
1154                          * - Tentative Address (RFC2462 section 5.4)
1155                          *  - A tentative address is not considered
1156                          *    "assigned to an interface" in the traditional
1157                          *    sense, unless it is also flagged as optimistic.
1158                          * - Candidate Source Address (section 4)
1159                          *  - In any case, anycast addresses, multicast
1160                          *    addresses, and the unspecified address MUST
1161                          *    NOT be included in a candidate set.
1162                          */
1163                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1164                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1165                                 continue;
1166
1167                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1168
1169                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1170                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1171                                 LIMIT_NETDEBUG(KERN_DEBUG
1172                                                "ADDRCONF: unspecified / multicast address "
1173                                                "assigned as unicast address on %s",
1174                                                dev->name);
1175                                 continue;
1176                         }
1177
1178                         score->rule = -1;
1179                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1180
1181                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1182                                 int minihiscore, miniscore;
1183
1184                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1185                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1186
1187                                 if (minihiscore > miniscore) {
1188                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1189                                             score->scopedist > 0) {
1190                                                 /*
1191                                                  * special case:
1192                                                  * each remaining entry
1193                                                  * has too small (not enough)
1194                                                  * scope, because ifa entries
1195                                                  * are sorted by their scope
1196                                                  * values.
1197                                                  */
1198                                                 goto try_nextdev;
1199                                         }
1200                                         break;
1201                                 } else if (minihiscore < miniscore) {
1202                                         if (hiscore->ifa)
1203                                                 in6_ifa_put(hiscore->ifa);
1204
1205                                         in6_ifa_hold(score->ifa);
1206
1207                                         swap(hiscore, score);
1208
1209                                         /* restore our iterator */
1210                                         score->ifa = hiscore->ifa;
1211
1212                                         break;
1213                                 }
1214                         }
1215                 }
1216 try_nextdev:
1217                 read_unlock_bh(&idev->lock);
1218         }
1219         rcu_read_unlock();
1220
1221         if (!hiscore->ifa)
1222                 return -EADDRNOTAVAIL;
1223
1224         *saddr = hiscore->ifa->addr;
1225         in6_ifa_put(hiscore->ifa);
1226         return 0;
1227 }
1228 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1229
1230 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1231                     unsigned char banned_flags)
1232 {
1233         struct inet6_dev *idev;
1234         int err = -EADDRNOTAVAIL;
1235
1236         rcu_read_lock();
1237         idev = __in6_dev_get(dev);
1238         if (idev) {
1239                 struct inet6_ifaddr *ifp;
1240
1241                 read_lock_bh(&idev->lock);
1242                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1243                         if (ifp->scope == IFA_LINK &&
1244                             !(ifp->flags & banned_flags)) {
1245                                 *addr = ifp->addr;
1246                                 err = 0;
1247                                 break;
1248                         }
1249                 }
1250                 read_unlock_bh(&idev->lock);
1251         }
1252         rcu_read_unlock();
1253         return err;
1254 }
1255
1256 static int ipv6_count_addresses(struct inet6_dev *idev)
1257 {
1258         int cnt = 0;
1259         struct inet6_ifaddr *ifp;
1260
1261         read_lock_bh(&idev->lock);
1262         list_for_each_entry(ifp, &idev->addr_list, if_list)
1263                 cnt++;
1264         read_unlock_bh(&idev->lock);
1265         return cnt;
1266 }
1267
1268 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1269                   struct net_device *dev, int strict)
1270 {
1271         struct inet6_ifaddr *ifp;
1272         struct hlist_node *node;
1273         unsigned int hash = ipv6_addr_hash(addr);
1274
1275         rcu_read_lock_bh();
1276         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1277                 if (!net_eq(dev_net(ifp->idev->dev), net))
1278                         continue;
1279                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1280                     !(ifp->flags&IFA_F_TENTATIVE) &&
1281                     (dev == NULL || ifp->idev->dev == dev ||
1282                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1283                         rcu_read_unlock_bh();
1284                         return 1;
1285                 }
1286         }
1287
1288         rcu_read_unlock_bh();
1289         return 0;
1290 }
1291 EXPORT_SYMBOL(ipv6_chk_addr);
1292
1293 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1294                                struct net_device *dev)
1295 {
1296         unsigned int hash = ipv6_addr_hash(addr);
1297         struct inet6_ifaddr *ifp;
1298         struct hlist_node *node;
1299
1300         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1301                 if (!net_eq(dev_net(ifp->idev->dev), net))
1302                         continue;
1303                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1304                         if (dev == NULL || ifp->idev->dev == dev)
1305                                 return true;
1306                 }
1307         }
1308         return false;
1309 }
1310
1311 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1312 {
1313         struct inet6_dev *idev;
1314         struct inet6_ifaddr *ifa;
1315         int     onlink;
1316
1317         onlink = 0;
1318         rcu_read_lock();
1319         idev = __in6_dev_get(dev);
1320         if (idev) {
1321                 read_lock_bh(&idev->lock);
1322                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1323                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1324                                                    ifa->prefix_len);
1325                         if (onlink)
1326                                 break;
1327                 }
1328                 read_unlock_bh(&idev->lock);
1329         }
1330         rcu_read_unlock();
1331         return onlink;
1332 }
1333 EXPORT_SYMBOL(ipv6_chk_prefix);
1334
1335 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1336                                      struct net_device *dev, int strict)
1337 {
1338         struct inet6_ifaddr *ifp, *result = NULL;
1339         unsigned int hash = ipv6_addr_hash(addr);
1340         struct hlist_node *node;
1341
1342         rcu_read_lock_bh();
1343         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1344                 if (!net_eq(dev_net(ifp->idev->dev), net))
1345                         continue;
1346                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1347                         if (dev == NULL || ifp->idev->dev == dev ||
1348                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1349                                 result = ifp;
1350                                 in6_ifa_hold(ifp);
1351                                 break;
1352                         }
1353                 }
1354         }
1355         rcu_read_unlock_bh();
1356
1357         return result;
1358 }
1359
1360 /* Gets referenced address, destroys ifaddr */
1361
1362 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1363 {
1364         if (ifp->flags&IFA_F_PERMANENT) {
1365                 spin_lock_bh(&ifp->lock);
1366                 addrconf_del_timer(ifp);
1367                 ifp->flags |= IFA_F_TENTATIVE;
1368                 if (dad_failed)
1369                         ifp->flags |= IFA_F_DADFAILED;
1370                 spin_unlock_bh(&ifp->lock);
1371                 if (dad_failed)
1372                         ipv6_ifa_notify(0, ifp);
1373                 in6_ifa_put(ifp);
1374 #ifdef CONFIG_IPV6_PRIVACY
1375         } else if (ifp->flags&IFA_F_TEMPORARY) {
1376                 struct inet6_ifaddr *ifpub;
1377                 spin_lock_bh(&ifp->lock);
1378                 ifpub = ifp->ifpub;
1379                 if (ifpub) {
1380                         in6_ifa_hold(ifpub);
1381                         spin_unlock_bh(&ifp->lock);
1382                         ipv6_create_tempaddr(ifpub, ifp);
1383                         in6_ifa_put(ifpub);
1384                 } else {
1385                         spin_unlock_bh(&ifp->lock);
1386                 }
1387                 ipv6_del_addr(ifp);
1388 #endif
1389         } else
1390                 ipv6_del_addr(ifp);
1391 }
1392
1393 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1394 {
1395         int err = -ENOENT;
1396
1397         spin_lock(&ifp->state_lock);
1398         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1399                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1400                 err = 0;
1401         }
1402         spin_unlock(&ifp->state_lock);
1403
1404         return err;
1405 }
1406
1407 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1408 {
1409         struct inet6_dev *idev = ifp->idev;
1410
1411         if (addrconf_dad_end(ifp)) {
1412                 in6_ifa_put(ifp);
1413                 return;
1414         }
1415
1416         net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n",
1417                              ifp->idev->dev->name, &ifp->addr);
1418
1419         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1420                 struct in6_addr addr;
1421
1422                 addr.s6_addr32[0] = htonl(0xfe800000);
1423                 addr.s6_addr32[1] = 0;
1424
1425                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1426                     ipv6_addr_equal(&ifp->addr, &addr)) {
1427                         /* DAD failed for link-local based on MAC address */
1428                         idev->cnf.disable_ipv6 = 1;
1429
1430                         pr_info("%s: IPv6 being disabled!\n",
1431                                 ifp->idev->dev->name);
1432                 }
1433         }
1434
1435         addrconf_dad_stop(ifp, 1);
1436 }
1437
1438 /* Join to solicited addr multicast group. */
1439
1440 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1441 {
1442         struct in6_addr maddr;
1443
1444         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1445                 return;
1446
1447         addrconf_addr_solict_mult(addr, &maddr);
1448         ipv6_dev_mc_inc(dev, &maddr);
1449 }
1450
1451 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1452 {
1453         struct in6_addr maddr;
1454
1455         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1456                 return;
1457
1458         addrconf_addr_solict_mult(addr, &maddr);
1459         __ipv6_dev_mc_dec(idev, &maddr);
1460 }
1461
1462 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1463 {
1464         struct in6_addr addr;
1465         if (ifp->prefix_len == 127) /* RFC 6164 */
1466                 return;
1467         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1468         if (ipv6_addr_any(&addr))
1469                 return;
1470         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1471 }
1472
1473 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1474 {
1475         struct in6_addr addr;
1476         if (ifp->prefix_len == 127) /* RFC 6164 */
1477                 return;
1478         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479         if (ipv6_addr_any(&addr))
1480                 return;
1481         __ipv6_dev_ac_dec(ifp->idev, &addr);
1482 }
1483
1484 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1485 {
1486         if (dev->addr_len != ETH_ALEN)
1487                 return -1;
1488         memcpy(eui, dev->dev_addr, 3);
1489         memcpy(eui + 5, dev->dev_addr + 3, 3);
1490
1491         /*
1492          * The zSeries OSA network cards can be shared among various
1493          * OS instances, but the OSA cards have only one MAC address.
1494          * This leads to duplicate address conflicts in conjunction
1495          * with IPv6 if more than one instance uses the same card.
1496          *
1497          * The driver for these cards can deliver a unique 16-bit
1498          * identifier for each instance sharing the same card.  It is
1499          * placed instead of 0xFFFE in the interface identifier.  The
1500          * "u" bit of the interface identifier is not inverted in this
1501          * case.  Hence the resulting interface identifier has local
1502          * scope according to RFC2373.
1503          */
1504         if (dev->dev_id) {
1505                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1506                 eui[4] = dev->dev_id & 0xFF;
1507         } else {
1508                 eui[3] = 0xFF;
1509                 eui[4] = 0xFE;
1510                 eui[0] ^= 2;
1511         }
1512         return 0;
1513 }
1514
1515 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev)
1516 {
1517         if (dev->addr_len != IEEE802154_ADDR_LEN)
1518                 return -1;
1519         memcpy(eui, dev->dev_addr, 8);
1520         return 0;
1521 }
1522
1523 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1524 {
1525         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1526         if (dev->addr_len != ARCNET_ALEN)
1527                 return -1;
1528         memset(eui, 0, 7);
1529         eui[7] = *(u8 *)dev->dev_addr;
1530         return 0;
1531 }
1532
1533 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1534 {
1535         if (dev->addr_len != INFINIBAND_ALEN)
1536                 return -1;
1537         memcpy(eui, dev->dev_addr + 12, 8);
1538         eui[0] |= 2;
1539         return 0;
1540 }
1541
1542 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1543 {
1544         if (addr == 0)
1545                 return -1;
1546         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1547                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1548                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1549                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1550                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1551                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1552         eui[1] = 0;
1553         eui[2] = 0x5E;
1554         eui[3] = 0xFE;
1555         memcpy(eui + 4, &addr, 4);
1556         return 0;
1557 }
1558
1559 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1560 {
1561         if (dev->priv_flags & IFF_ISATAP)
1562                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1563         return -1;
1564 }
1565
1566 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1567 {
1568         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1569 }
1570
1571 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1572 {
1573         switch (dev->type) {
1574         case ARPHRD_ETHER:
1575         case ARPHRD_FDDI:
1576                 return addrconf_ifid_eui48(eui, dev);
1577         case ARPHRD_ARCNET:
1578                 return addrconf_ifid_arcnet(eui, dev);
1579         case ARPHRD_INFINIBAND:
1580                 return addrconf_ifid_infiniband(eui, dev);
1581         case ARPHRD_SIT:
1582                 return addrconf_ifid_sit(eui, dev);
1583         case ARPHRD_IPGRE:
1584                 return addrconf_ifid_gre(eui, dev);
1585         case ARPHRD_IEEE802154:
1586                 return addrconf_ifid_eui64(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                 pr_warn("%s: too short regeneration interval; timer disabled for %s\n",
1661                         __func__, idev->dev->name);
1662                 goto out;
1663         }
1664
1665         if (!mod_timer(&idev->regen_timer, expires))
1666                 in6_dev_hold(idev);
1667
1668 out:
1669         write_unlock_bh(&idev->lock);
1670         rcu_read_unlock_bh();
1671         in6_dev_put(idev);
1672 }
1673
1674 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr)
1675 {
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         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->dst.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, bool sllao)
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                 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n");
1846                 return;
1847         }
1848
1849         in6_dev = in6_dev_get(dev);
1850
1851         if (in6_dev == NULL) {
1852                 net_dbg_ratelimited("addrconf: device %s not configured\n",
1853                                     dev->name);
1854                 return;
1855         }
1856
1857         /*
1858          *      Two things going on here:
1859          *      1) Add routes for on-link prefixes
1860          *      2) Configure prefixes with the auto flag set
1861          */
1862
1863         if (pinfo->onlink) {
1864                 struct rt6_info *rt;
1865                 unsigned long rt_expires;
1866
1867                 /* Avoid arithmetic overflow. Really, we could
1868                  * save rt_expires in seconds, likely valid_lft,
1869                  * but it would require division in fib gc, that it
1870                  * not good.
1871                  */
1872                 if (HZ > USER_HZ)
1873                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1874                 else
1875                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1876
1877                 if (addrconf_finite_timeout(rt_expires))
1878                         rt_expires *= HZ;
1879
1880                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1881                                                pinfo->prefix_len,
1882                                                dev,
1883                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1884                                                RTF_GATEWAY | RTF_DEFAULT);
1885
1886                 if (rt) {
1887                         /* Autoconf prefix route */
1888                         if (valid_lft == 0) {
1889                                 ip6_del_rt(rt);
1890                                 rt = NULL;
1891                         } else if (addrconf_finite_timeout(rt_expires)) {
1892                                 /* not infinity */
1893                                 rt6_set_expires(rt, jiffies + rt_expires);
1894                         } else {
1895                                 rt6_clean_expires(rt);
1896                         }
1897                 } else if (valid_lft) {
1898                         clock_t expires = 0;
1899                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1900                         if (addrconf_finite_timeout(rt_expires)) {
1901                                 /* not infinity */
1902                                 flags |= RTF_EXPIRES;
1903                                 expires = jiffies_to_clock_t(rt_expires);
1904                         }
1905                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1906                                               dev, expires, flags);
1907                 }
1908                 if (rt)
1909                         dst_release(&rt->dst);
1910         }
1911
1912         /* Try to figure out our local address for this prefix */
1913
1914         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1915                 struct inet6_ifaddr *ifp;
1916                 struct in6_addr addr;
1917                 int create = 0, update_lft = 0;
1918
1919                 if (pinfo->prefix_len == 64) {
1920                         memcpy(&addr, &pinfo->prefix, 8);
1921                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1922                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1923                                 in6_dev_put(in6_dev);
1924                                 return;
1925                         }
1926                         goto ok;
1927                 }
1928                 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n",
1929                                     pinfo->prefix_len);
1930                 in6_dev_put(in6_dev);
1931                 return;
1932
1933 ok:
1934
1935                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1936
1937                 if (ifp == NULL && valid_lft) {
1938                         int max_addresses = in6_dev->cnf.max_addresses;
1939                         u32 addr_flags = 0;
1940
1941 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1942                         if (in6_dev->cnf.optimistic_dad &&
1943                             !net->ipv6.devconf_all->forwarding && sllao)
1944                                 addr_flags = IFA_F_OPTIMISTIC;
1945 #endif
1946
1947                         /* Do not allow to create too much of autoconfigured
1948                          * addresses; this would be too easy way to crash kernel.
1949                          */
1950                         if (!max_addresses ||
1951                             ipv6_count_addresses(in6_dev) < max_addresses)
1952                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1953                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1954                                                     addr_flags);
1955
1956                         if (!ifp || IS_ERR(ifp)) {
1957                                 in6_dev_put(in6_dev);
1958                                 return;
1959                         }
1960
1961                         update_lft = create = 1;
1962                         ifp->cstamp = jiffies;
1963                         addrconf_dad_start(ifp);
1964                 }
1965
1966                 if (ifp) {
1967                         int flags;
1968                         unsigned long now;
1969 #ifdef CONFIG_IPV6_PRIVACY
1970                         struct inet6_ifaddr *ift;
1971 #endif
1972                         u32 stored_lft;
1973
1974                         /* update lifetime (RFC2462 5.5.3 e) */
1975                         spin_lock(&ifp->lock);
1976                         now = jiffies;
1977                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1978                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1979                         else
1980                                 stored_lft = 0;
1981                         if (!update_lft && stored_lft) {
1982                                 if (valid_lft > MIN_VALID_LIFETIME ||
1983                                     valid_lft > stored_lft)
1984                                         update_lft = 1;
1985                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1986                                         /* valid_lft <= stored_lft is always true */
1987                                         /*
1988                                          * RFC 4862 Section 5.5.3e:
1989                                          * "Note that the preferred lifetime of
1990                                          *  the corresponding address is always
1991                                          *  reset to the Preferred Lifetime in
1992                                          *  the received Prefix Information
1993                                          *  option, regardless of whether the
1994                                          *  valid lifetime is also reset or
1995                                          *  ignored."
1996                                          *
1997                                          *  So if the preferred lifetime in
1998                                          *  this advertisement is different
1999                                          *  than what we have stored, but the
2000                                          *  valid lifetime is invalid, just
2001                                          *  reset prefered_lft.
2002                                          *
2003                                          *  We must set the valid lifetime
2004                                          *  to the stored lifetime since we'll
2005                                          *  be updating the timestamp below,
2006                                          *  else we'll set it back to the
2007                                          *  minimum.
2008                                          */
2009                                         if (prefered_lft != ifp->prefered_lft) {
2010                                                 valid_lft = stored_lft;
2011                                                 update_lft = 1;
2012                                         }
2013                                 } else {
2014                                         valid_lft = MIN_VALID_LIFETIME;
2015                                         if (valid_lft < prefered_lft)
2016                                                 prefered_lft = valid_lft;
2017                                         update_lft = 1;
2018                                 }
2019                         }
2020
2021                         if (update_lft) {
2022                                 ifp->valid_lft = valid_lft;
2023                                 ifp->prefered_lft = prefered_lft;
2024                                 ifp->tstamp = now;
2025                                 flags = ifp->flags;
2026                                 ifp->flags &= ~IFA_F_DEPRECATED;
2027                                 spin_unlock(&ifp->lock);
2028
2029                                 if (!(flags&IFA_F_TENTATIVE))
2030                                         ipv6_ifa_notify(0, ifp);
2031                         } else
2032                                 spin_unlock(&ifp->lock);
2033
2034 #ifdef CONFIG_IPV6_PRIVACY
2035                         read_lock_bh(&in6_dev->lock);
2036                         /* update all temporary addresses in the list */
2037                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2038                                             tmp_list) {
2039                                 int age, max_valid, max_prefered;
2040
2041                                 if (ifp != ift->ifpub)
2042                                         continue;
2043
2044                                 /*
2045                                  * RFC 4941 section 3.3:
2046                                  * If a received option will extend the lifetime
2047                                  * of a public address, the lifetimes of
2048                                  * temporary addresses should be extended,
2049                                  * subject to the overall constraint that no
2050                                  * temporary addresses should ever remain
2051                                  * "valid" or "preferred" for a time longer than
2052                                  * (TEMP_VALID_LIFETIME) or
2053                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2054                                  * respectively.
2055                                  */
2056                                 age = (now - ift->cstamp) / HZ;
2057                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2058                                 if (max_valid < 0)
2059                                         max_valid = 0;
2060
2061                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2062                                                in6_dev->cnf.max_desync_factor -
2063                                                age;
2064                                 if (max_prefered < 0)
2065                                         max_prefered = 0;
2066
2067                                 if (valid_lft > max_valid)
2068                                         valid_lft = max_valid;
2069
2070                                 if (prefered_lft > max_prefered)
2071                                         prefered_lft = max_prefered;
2072
2073                                 spin_lock(&ift->lock);
2074                                 flags = ift->flags;
2075                                 ift->valid_lft = valid_lft;
2076                                 ift->prefered_lft = prefered_lft;
2077                                 ift->tstamp = now;
2078                                 if (prefered_lft > 0)
2079                                         ift->flags &= ~IFA_F_DEPRECATED;
2080
2081                                 spin_unlock(&ift->lock);
2082                                 if (!(flags&IFA_F_TENTATIVE))
2083                                         ipv6_ifa_notify(0, ift);
2084                         }
2085
2086                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2087                                 /*
2088                                  * When a new public address is created as
2089                                  * described in [ADDRCONF], also create a new
2090                                  * temporary address. Also create a temporary
2091                                  * address if it's enabled but no temporary
2092                                  * address currently exists.
2093                                  */
2094                                 read_unlock_bh(&in6_dev->lock);
2095                                 ipv6_create_tempaddr(ifp, NULL);
2096                         } else {
2097                                 read_unlock_bh(&in6_dev->lock);
2098                         }
2099 #endif
2100                         in6_ifa_put(ifp);
2101                         addrconf_verify(0);
2102                 }
2103         }
2104         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2105         in6_dev_put(in6_dev);
2106 }
2107
2108 /*
2109  *      Set destination address.
2110  *      Special case for SIT interfaces where we create a new "virtual"
2111  *      device.
2112  */
2113 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2114 {
2115         struct in6_ifreq ireq;
2116         struct net_device *dev;
2117         int err = -EINVAL;
2118
2119         rtnl_lock();
2120
2121         err = -EFAULT;
2122         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2123                 goto err_exit;
2124
2125         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2126
2127         err = -ENODEV;
2128         if (dev == NULL)
2129                 goto err_exit;
2130
2131 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2132         if (dev->type == ARPHRD_SIT) {
2133                 const struct net_device_ops *ops = dev->netdev_ops;
2134                 struct ifreq ifr;
2135                 struct ip_tunnel_parm p;
2136
2137                 err = -EADDRNOTAVAIL;
2138                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2139                         goto err_exit;
2140
2141                 memset(&p, 0, sizeof(p));
2142                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2143                 p.iph.saddr = 0;
2144                 p.iph.version = 4;
2145                 p.iph.ihl = 5;
2146                 p.iph.protocol = IPPROTO_IPV6;
2147                 p.iph.ttl = 64;
2148                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2149
2150                 if (ops->ndo_do_ioctl) {
2151                         mm_segment_t oldfs = get_fs();
2152
2153                         set_fs(KERNEL_DS);
2154                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2155                         set_fs(oldfs);
2156                 } else
2157                         err = -EOPNOTSUPP;
2158
2159                 if (err == 0) {
2160                         err = -ENOBUFS;
2161                         dev = __dev_get_by_name(net, p.name);
2162                         if (!dev)
2163                                 goto err_exit;
2164                         err = dev_open(dev);
2165                 }
2166         }
2167 #endif
2168
2169 err_exit:
2170         rtnl_unlock();
2171         return err;
2172 }
2173
2174 /*
2175  *      Manual configuration of address on an interface
2176  */
2177 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2178                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2179                           __u32 valid_lft)
2180 {
2181         struct inet6_ifaddr *ifp;
2182         struct inet6_dev *idev;
2183         struct net_device *dev;
2184         int scope;
2185         u32 flags;
2186         clock_t expires;
2187         unsigned long timeout;
2188
2189         ASSERT_RTNL();
2190
2191         if (plen > 128)
2192                 return -EINVAL;
2193
2194         /* check the lifetime */
2195         if (!valid_lft || prefered_lft > valid_lft)
2196                 return -EINVAL;
2197
2198         dev = __dev_get_by_index(net, ifindex);
2199         if (!dev)
2200                 return -ENODEV;
2201
2202         idev = addrconf_add_dev(dev);
2203         if (IS_ERR(idev))
2204                 return PTR_ERR(idev);
2205
2206         scope = ipv6_addr_scope(pfx);
2207
2208         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2209         if (addrconf_finite_timeout(timeout)) {
2210                 expires = jiffies_to_clock_t(timeout * HZ);
2211                 valid_lft = timeout;
2212                 flags = RTF_EXPIRES;
2213         } else {
2214                 expires = 0;
2215                 flags = 0;
2216                 ifa_flags |= IFA_F_PERMANENT;
2217         }
2218
2219         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2220         if (addrconf_finite_timeout(timeout)) {
2221                 if (timeout == 0)
2222                         ifa_flags |= IFA_F_DEPRECATED;
2223                 prefered_lft = timeout;
2224         }
2225
2226         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2227
2228         if (!IS_ERR(ifp)) {
2229                 spin_lock_bh(&ifp->lock);
2230                 ifp->valid_lft = valid_lft;
2231                 ifp->prefered_lft = prefered_lft;
2232                 ifp->tstamp = jiffies;
2233                 spin_unlock_bh(&ifp->lock);
2234
2235                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2236                                       expires, flags);
2237                 /*
2238                  * Note that section 3.1 of RFC 4429 indicates
2239                  * that the Optimistic flag should not be set for
2240                  * manually configured addresses
2241                  */
2242                 addrconf_dad_start(ifp);
2243                 in6_ifa_put(ifp);
2244                 addrconf_verify(0);
2245                 return 0;
2246         }
2247
2248         return PTR_ERR(ifp);
2249 }
2250
2251 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2252                           unsigned int plen)
2253 {
2254         struct inet6_ifaddr *ifp;
2255         struct inet6_dev *idev;
2256         struct net_device *dev;
2257
2258         if (plen > 128)
2259                 return -EINVAL;
2260
2261         dev = __dev_get_by_index(net, ifindex);
2262         if (!dev)
2263                 return -ENODEV;
2264
2265         if ((idev = __in6_dev_get(dev)) == NULL)
2266                 return -ENXIO;
2267
2268         read_lock_bh(&idev->lock);
2269         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2270                 if (ifp->prefix_len == plen &&
2271                     ipv6_addr_equal(pfx, &ifp->addr)) {
2272                         in6_ifa_hold(ifp);
2273                         read_unlock_bh(&idev->lock);
2274
2275                         ipv6_del_addr(ifp);
2276
2277                         /* If the last address is deleted administratively,
2278                            disable IPv6 on this interface.
2279                          */
2280                         if (list_empty(&idev->addr_list))
2281                                 addrconf_ifdown(idev->dev, 1);
2282                         return 0;
2283                 }
2284         }
2285         read_unlock_bh(&idev->lock);
2286         return -EADDRNOTAVAIL;
2287 }
2288
2289
2290 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2291 {
2292         struct in6_ifreq ireq;
2293         int err;
2294
2295         if (!capable(CAP_NET_ADMIN))
2296                 return -EPERM;
2297
2298         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2299                 return -EFAULT;
2300
2301         rtnl_lock();
2302         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2303                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2304                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2305         rtnl_unlock();
2306         return err;
2307 }
2308
2309 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2310 {
2311         struct in6_ifreq ireq;
2312         int err;
2313
2314         if (!capable(CAP_NET_ADMIN))
2315                 return -EPERM;
2316
2317         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2318                 return -EFAULT;
2319
2320         rtnl_lock();
2321         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2322                              ireq.ifr6_prefixlen);
2323         rtnl_unlock();
2324         return err;
2325 }
2326
2327 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2328                      int plen, int scope)
2329 {
2330         struct inet6_ifaddr *ifp;
2331
2332         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2333         if (!IS_ERR(ifp)) {
2334                 spin_lock_bh(&ifp->lock);
2335                 ifp->flags &= ~IFA_F_TENTATIVE;
2336                 spin_unlock_bh(&ifp->lock);
2337                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2338                 in6_ifa_put(ifp);
2339         }
2340 }
2341
2342 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2343 static void sit_add_v4_addrs(struct inet6_dev *idev)
2344 {
2345         struct in6_addr addr;
2346         struct net_device *dev;
2347         struct net *net = dev_net(idev->dev);
2348         int scope;
2349
2350         ASSERT_RTNL();
2351
2352         memset(&addr, 0, sizeof(struct in6_addr));
2353         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2354
2355         if (idev->dev->flags&IFF_POINTOPOINT) {
2356                 addr.s6_addr32[0] = htonl(0xfe800000);
2357                 scope = IFA_LINK;
2358         } else {
2359                 scope = IPV6_ADDR_COMPATv4;
2360         }
2361
2362         if (addr.s6_addr32[3]) {
2363                 add_addr(idev, &addr, 128, scope);
2364                 return;
2365         }
2366
2367         for_each_netdev(net, dev) {
2368                 struct in_device *in_dev = __in_dev_get_rtnl(dev);
2369                 if (in_dev && (dev->flags & IFF_UP)) {
2370                         struct in_ifaddr *ifa;
2371
2372                         int flag = scope;
2373
2374                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2375                                 int plen;
2376
2377                                 addr.s6_addr32[3] = ifa->ifa_local;
2378
2379                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2380                                         continue;
2381                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2382                                         if (idev->dev->flags&IFF_POINTOPOINT)
2383                                                 continue;
2384                                         flag |= IFA_HOST;
2385                                 }
2386                                 if (idev->dev->flags&IFF_POINTOPOINT)
2387                                         plen = 64;
2388                                 else
2389                                         plen = 96;
2390
2391                                 add_addr(idev, &addr, plen, flag);
2392                         }
2393                 }
2394         }
2395 }
2396 #endif
2397
2398 static void init_loopback(struct net_device *dev)
2399 {
2400         struct inet6_dev  *idev;
2401
2402         /* ::1 */
2403
2404         ASSERT_RTNL();
2405
2406         if ((idev = ipv6_find_idev(dev)) == NULL) {
2407                 pr_debug("%s: add_dev failed\n", __func__);
2408                 return;
2409         }
2410
2411         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2412 }
2413
2414 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2415 {
2416         struct inet6_ifaddr *ifp;
2417         u32 addr_flags = IFA_F_PERMANENT;
2418
2419 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2420         if (idev->cnf.optimistic_dad &&
2421             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2422                 addr_flags |= IFA_F_OPTIMISTIC;
2423 #endif
2424
2425
2426         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2427         if (!IS_ERR(ifp)) {
2428                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2429                 addrconf_dad_start(ifp);
2430                 in6_ifa_put(ifp);
2431         }
2432 }
2433
2434 static void addrconf_dev_config(struct net_device *dev)
2435 {
2436         struct in6_addr addr;
2437         struct inet6_dev *idev;
2438
2439         ASSERT_RTNL();
2440
2441         if ((dev->type != ARPHRD_ETHER) &&
2442             (dev->type != ARPHRD_FDDI) &&
2443             (dev->type != ARPHRD_ARCNET) &&
2444             (dev->type != ARPHRD_INFINIBAND) &&
2445             (dev->type != ARPHRD_IEEE802154)) {
2446                 /* Alas, we support only Ethernet autoconfiguration. */
2447                 return;
2448         }
2449
2450         idev = addrconf_add_dev(dev);
2451         if (IS_ERR(idev))
2452                 return;
2453
2454         memset(&addr, 0, sizeof(struct in6_addr));
2455         addr.s6_addr32[0] = htonl(0xFE800000);
2456
2457         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2458                 addrconf_add_linklocal(idev, &addr);
2459 }
2460
2461 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2462 static void addrconf_sit_config(struct net_device *dev)
2463 {
2464         struct inet6_dev *idev;
2465
2466         ASSERT_RTNL();
2467
2468         /*
2469          * Configure the tunnel with one of our IPv4
2470          * addresses... we should configure all of
2471          * our v4 addrs in the tunnel
2472          */
2473
2474         if ((idev = ipv6_find_idev(dev)) == NULL) {
2475                 pr_debug("%s: add_dev failed\n", __func__);
2476                 return;
2477         }
2478
2479         if (dev->priv_flags & IFF_ISATAP) {
2480                 struct in6_addr addr;
2481
2482                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2483                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2484                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2485                         addrconf_add_linklocal(idev, &addr);
2486                 return;
2487         }
2488
2489         sit_add_v4_addrs(idev);
2490
2491         if (dev->flags&IFF_POINTOPOINT) {
2492                 addrconf_add_mroute(dev);
2493                 addrconf_add_lroute(dev);
2494         } else
2495                 sit_route_add(dev);
2496 }
2497 #endif
2498
2499 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2500 static void addrconf_gre_config(struct net_device *dev)
2501 {
2502         struct inet6_dev *idev;
2503         struct in6_addr addr;
2504
2505         pr_info("%s(%s)\n", __func__, dev->name);
2506
2507         ASSERT_RTNL();
2508
2509         if ((idev = ipv6_find_idev(dev)) == NULL) {
2510                 pr_debug("%s: add_dev failed\n", __func__);
2511                 return;
2512         }
2513
2514         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2515         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2516
2517         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2518                 addrconf_add_linklocal(idev, &addr);
2519 }
2520 #endif
2521
2522 static inline int
2523 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2524 {
2525         struct in6_addr lladdr;
2526
2527         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2528                 addrconf_add_linklocal(idev, &lladdr);
2529                 return 0;
2530         }
2531         return -1;
2532 }
2533
2534 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2535 {
2536         struct net_device *link_dev;
2537         struct net *net = dev_net(idev->dev);
2538
2539         /* first try to inherit the link-local address from the link device */
2540         if (idev->dev->iflink &&
2541             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2542                 if (!ipv6_inherit_linklocal(idev, link_dev))
2543                         return;
2544         }
2545         /* then try to inherit it from any device */
2546         for_each_netdev(net, link_dev) {
2547                 if (!ipv6_inherit_linklocal(idev, link_dev))
2548                         return;
2549         }
2550         pr_debug("init ip6-ip6: add_linklocal failed\n");
2551 }
2552
2553 /*
2554  * Autoconfigure tunnel with a link-local address so routing protocols,
2555  * DHCPv6, MLD etc. can be run over the virtual link
2556  */
2557
2558 static void addrconf_ip6_tnl_config(struct net_device *dev)
2559 {
2560         struct inet6_dev *idev;
2561
2562         ASSERT_RTNL();
2563
2564         idev = addrconf_add_dev(dev);
2565         if (IS_ERR(idev)) {
2566                 pr_debug("init ip6-ip6: add_dev failed\n");
2567                 return;
2568         }
2569         ip6_tnl_add_linklocal(idev);
2570 }
2571
2572 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2573                            void *data)
2574 {
2575         struct net_device *dev = (struct net_device *) data;
2576         struct inet6_dev *idev = __in6_dev_get(dev);
2577         int run_pending = 0;
2578         int err;
2579
2580         switch (event) {
2581         case NETDEV_REGISTER:
2582                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2583                         idev = ipv6_add_dev(dev);
2584                         if (!idev)
2585                                 return notifier_from_errno(-ENOMEM);
2586                 }
2587                 break;
2588
2589         case NETDEV_UP:
2590         case NETDEV_CHANGE:
2591                 if (dev->flags & IFF_SLAVE)
2592                         break;
2593
2594                 if (event == NETDEV_UP) {
2595                         if (!addrconf_qdisc_ok(dev)) {
2596                                 /* device is not ready yet. */
2597                                 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n",
2598                                         dev->name);
2599                                 break;
2600                         }
2601
2602                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2603                                 idev = ipv6_add_dev(dev);
2604
2605                         if (idev) {
2606                                 idev->if_flags |= IF_READY;
2607                                 run_pending = 1;
2608                         }
2609                 } else {
2610                         if (!addrconf_qdisc_ok(dev)) {
2611                                 /* device is still not ready. */
2612                                 break;
2613                         }
2614
2615                         if (idev) {
2616                                 if (idev->if_flags & IF_READY)
2617                                         /* device is already configured. */
2618                                         break;
2619                                 idev->if_flags |= IF_READY;
2620                         }
2621
2622                         pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n",
2623                                 dev->name);
2624
2625                         run_pending = 1;
2626                 }
2627
2628                 switch (dev->type) {
2629 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2630                 case ARPHRD_SIT:
2631                         addrconf_sit_config(dev);
2632                         break;
2633 #endif
2634 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2635                 case ARPHRD_IPGRE:
2636                         addrconf_gre_config(dev);
2637                         break;
2638 #endif
2639                 case ARPHRD_TUNNEL6:
2640                         addrconf_ip6_tnl_config(dev);
2641                         break;
2642                 case ARPHRD_LOOPBACK:
2643                         init_loopback(dev);
2644                         break;
2645
2646                 default:
2647                         addrconf_dev_config(dev);
2648                         break;
2649                 }
2650
2651                 if (idev) {
2652                         if (run_pending)
2653                                 addrconf_dad_run(idev);
2654
2655                         /*
2656                          * If the MTU changed during the interface down,
2657                          * when the interface up, the changed MTU must be
2658                          * reflected in the idev as well as routers.
2659                          */
2660                         if (idev->cnf.mtu6 != dev->mtu &&
2661                             dev->mtu >= IPV6_MIN_MTU) {
2662                                 rt6_mtu_change(dev, dev->mtu);
2663                                 idev->cnf.mtu6 = dev->mtu;
2664                         }
2665                         idev->tstamp = jiffies;
2666                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2667
2668                         /*
2669                          * If the changed mtu during down is lower than
2670                          * IPV6_MIN_MTU stop IPv6 on this interface.
2671                          */
2672                         if (dev->mtu < IPV6_MIN_MTU)
2673                                 addrconf_ifdown(dev, 1);
2674                 }
2675                 break;
2676
2677         case NETDEV_CHANGEMTU:
2678                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2679                         rt6_mtu_change(dev, dev->mtu);
2680                         idev->cnf.mtu6 = dev->mtu;
2681                         break;
2682                 }
2683
2684                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2685                         idev = ipv6_add_dev(dev);
2686                         if (idev)
2687                                 break;
2688                 }
2689
2690                 /*
2691                  * MTU falled under IPV6_MIN_MTU.
2692                  * Stop IPv6 on this interface.
2693                  */
2694
2695         case NETDEV_DOWN:
2696         case NETDEV_UNREGISTER:
2697                 /*
2698                  *      Remove all addresses from this interface.
2699                  */
2700                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2701                 break;
2702
2703         case NETDEV_CHANGENAME:
2704                 if (idev) {
2705                         snmp6_unregister_dev(idev);
2706                         addrconf_sysctl_unregister(idev);
2707                         addrconf_sysctl_register(idev);
2708                         err = snmp6_register_dev(idev);
2709                         if (err)
2710                                 return notifier_from_errno(err);
2711                 }
2712                 break;
2713
2714         case NETDEV_PRE_TYPE_CHANGE:
2715         case NETDEV_POST_TYPE_CHANGE:
2716                 addrconf_type_change(dev, event);
2717                 break;
2718         }
2719
2720         return NOTIFY_OK;
2721 }
2722
2723 /*
2724  *      addrconf module should be notified of a device going up
2725  */
2726 static struct notifier_block ipv6_dev_notf = {
2727         .notifier_call = addrconf_notify,
2728 };
2729
2730 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2731 {
2732         struct inet6_dev *idev;
2733         ASSERT_RTNL();
2734
2735         idev = __in6_dev_get(dev);
2736
2737         if (event == NETDEV_POST_TYPE_CHANGE)
2738                 ipv6_mc_remap(idev);
2739         else if (event == NETDEV_PRE_TYPE_CHANGE)
2740                 ipv6_mc_unmap(idev);
2741 }
2742
2743 static int addrconf_ifdown(struct net_device *dev, int how)
2744 {
2745         struct net *net = dev_net(dev);
2746         struct inet6_dev *idev;
2747         struct inet6_ifaddr *ifa;
2748         int state, i;
2749
2750         ASSERT_RTNL();
2751
2752         rt6_ifdown(net, dev);
2753         neigh_ifdown(&nd_tbl, dev);
2754
2755         idev = __in6_dev_get(dev);
2756         if (idev == NULL)
2757                 return -ENODEV;
2758
2759         /*
2760          * Step 1: remove reference to ipv6 device from parent device.
2761          *         Do not dev_put!
2762          */
2763         if (how) {
2764                 idev->dead = 1;
2765
2766                 /* protected by rtnl_lock */
2767                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2768
2769                 /* Step 1.5: remove snmp6 entry */
2770                 snmp6_unregister_dev(idev);
2771
2772         }
2773
2774         /* Step 2: clear hash table */
2775         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2776                 struct hlist_head *h = &inet6_addr_lst[i];
2777                 struct hlist_node *n;
2778
2779                 spin_lock_bh(&addrconf_hash_lock);
2780         restart:
2781                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2782                         if (ifa->idev == idev) {
2783                                 hlist_del_init_rcu(&ifa->addr_lst);
2784                                 addrconf_del_timer(ifa);
2785                                 goto restart;
2786                         }
2787                 }
2788                 spin_unlock_bh(&addrconf_hash_lock);
2789         }
2790
2791         write_lock_bh(&idev->lock);
2792
2793         /* Step 2: clear flags for stateless addrconf */
2794         if (!how)
2795                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2796
2797 #ifdef CONFIG_IPV6_PRIVACY
2798         if (how && del_timer(&idev->regen_timer))
2799                 in6_dev_put(idev);
2800
2801         /* Step 3: clear tempaddr list */
2802         while (!list_empty(&idev->tempaddr_list)) {
2803                 ifa = list_first_entry(&idev->tempaddr_list,
2804                                        struct inet6_ifaddr, tmp_list);
2805                 list_del(&ifa->tmp_list);
2806                 write_unlock_bh(&idev->lock);
2807                 spin_lock_bh(&ifa->lock);
2808
2809                 if (ifa->ifpub) {
2810                         in6_ifa_put(ifa->ifpub);
2811                         ifa->ifpub = NULL;
2812                 }
2813                 spin_unlock_bh(&ifa->lock);
2814                 in6_ifa_put(ifa);
2815                 write_lock_bh(&idev->lock);
2816         }
2817 #endif
2818
2819         while (!list_empty(&idev->addr_list)) {
2820                 ifa = list_first_entry(&idev->addr_list,
2821                                        struct inet6_ifaddr, if_list);
2822                 addrconf_del_timer(ifa);
2823
2824                 list_del(&ifa->if_list);
2825
2826                 write_unlock_bh(&idev->lock);
2827
2828                 spin_lock_bh(&ifa->state_lock);
2829                 state = ifa->state;
2830                 ifa->state = INET6_IFADDR_STATE_DEAD;
2831                 spin_unlock_bh(&ifa->state_lock);
2832
2833                 if (state != INET6_IFADDR_STATE_DEAD) {
2834                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2835                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2836                 }
2837                 in6_ifa_put(ifa);
2838
2839                 write_lock_bh(&idev->lock);
2840         }
2841
2842         write_unlock_bh(&idev->lock);
2843
2844         /* Step 5: Discard multicast list */
2845         if (how)
2846                 ipv6_mc_destroy_dev(idev);
2847         else
2848                 ipv6_mc_down(idev);
2849
2850         idev->tstamp = jiffies;
2851
2852         /* Last: Shot the device (if unregistered) */
2853         if (how) {
2854                 addrconf_sysctl_unregister(idev);
2855                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2856                 neigh_ifdown(&nd_tbl, dev);
2857                 in6_dev_put(idev);
2858         }
2859         return 0;
2860 }
2861
2862 static void addrconf_rs_timer(unsigned long data)
2863 {
2864         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2865         struct inet6_dev *idev = ifp->idev;
2866
2867         read_lock(&idev->lock);
2868         if (idev->dead || !(idev->if_flags & IF_READY))
2869                 goto out;
2870
2871         if (idev->cnf.forwarding)
2872                 goto out;
2873
2874         /* Announcement received after solicitation was sent */
2875         if (idev->if_flags & IF_RA_RCVD)
2876                 goto out;
2877
2878         spin_lock(&ifp->lock);
2879         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2880                 /* The wait after the last probe can be shorter */
2881                 addrconf_mod_timer(ifp, AC_RS,
2882                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2883                                    idev->cnf.rtr_solicit_delay :
2884                                    idev->cnf.rtr_solicit_interval);
2885                 spin_unlock(&ifp->lock);
2886
2887                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2888         } else {
2889                 spin_unlock(&ifp->lock);
2890                 /*
2891                  * Note: we do not support deprecated "all on-link"
2892                  * assumption any longer.
2893                  */
2894                 pr_debug("%s: no IPv6 routers present\n", idev->dev->name);
2895         }
2896
2897 out:
2898         read_unlock(&idev->lock);
2899         in6_ifa_put(ifp);
2900 }
2901
2902 /*
2903  *      Duplicate Address Detection
2904  */
2905 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2906 {
2907         unsigned long rand_num;
2908         struct inet6_dev *idev = ifp->idev;
2909
2910         if (ifp->flags & IFA_F_OPTIMISTIC)
2911                 rand_num = 0;
2912         else
2913                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2914
2915         ifp->probes = idev->cnf.dad_transmits;
2916         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2917 }
2918
2919 static void addrconf_dad_start(struct inet6_ifaddr *ifp)
2920 {
2921         struct inet6_dev *idev = ifp->idev;
2922         struct net_device *dev = idev->dev;
2923
2924         addrconf_join_solict(dev, &ifp->addr);
2925
2926         net_srandom(ifp->addr.s6_addr32[3]);
2927
2928         read_lock_bh(&idev->lock);
2929         spin_lock(&ifp->lock);
2930         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2931                 goto out;
2932
2933         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2934             idev->cnf.accept_dad < 1 ||
2935             !(ifp->flags&IFA_F_TENTATIVE) ||
2936             ifp->flags & IFA_F_NODAD) {
2937                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2938                 spin_unlock(&ifp->lock);
2939                 read_unlock_bh(&idev->lock);
2940
2941                 addrconf_dad_completed(ifp);
2942                 return;
2943         }
2944
2945         if (!(idev->if_flags & IF_READY)) {
2946                 spin_unlock(&ifp->lock);
2947                 read_unlock_bh(&idev->lock);
2948                 /*
2949                  * If the device is not ready:
2950                  * - keep it tentative if it is a permanent address.
2951                  * - otherwise, kill it.
2952                  */
2953                 in6_ifa_hold(ifp);
2954                 addrconf_dad_stop(ifp, 0);
2955                 return;
2956         }
2957
2958         /*
2959          * Optimistic nodes can start receiving
2960          * Frames right away
2961          */
2962         if (ifp->flags & IFA_F_OPTIMISTIC)
2963                 ip6_ins_rt(ifp->rt);
2964
2965         addrconf_dad_kick(ifp);
2966 out:
2967         spin_unlock(&ifp->lock);
2968         read_unlock_bh(&idev->lock);
2969 }
2970
2971 static void addrconf_dad_timer(unsigned long data)
2972 {
2973         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2974         struct inet6_dev *idev = ifp->idev;
2975         struct in6_addr mcaddr;
2976
2977         if (!ifp->probes && addrconf_dad_end(ifp))
2978                 goto out;
2979
2980         read_lock(&idev->lock);
2981         if (idev->dead || !(idev->if_flags & IF_READY)) {
2982                 read_unlock(&idev->lock);
2983                 goto out;
2984         }
2985
2986         spin_lock(&ifp->lock);
2987         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2988                 spin_unlock(&ifp->lock);
2989                 read_unlock(&idev->lock);
2990                 goto out;
2991         }
2992
2993         if (ifp->probes == 0) {
2994                 /*
2995                  * DAD was successful
2996                  */
2997
2998                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2999                 spin_unlock(&ifp->lock);
3000                 read_unlock(&idev->lock);
3001
3002                 addrconf_dad_completed(ifp);
3003
3004                 goto out;
3005         }
3006
3007         ifp->probes--;
3008         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3009         spin_unlock(&ifp->lock);
3010         read_unlock(&idev->lock);
3011
3012         /* send a neighbour solicitation for our addr */
3013         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3014         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3015 out:
3016         in6_ifa_put(ifp);
3017 }
3018
3019 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3020 {
3021         struct net_device *dev = ifp->idev->dev;
3022
3023         /*
3024          *      Configure the address for reception. Now it is valid.
3025          */
3026
3027         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3028
3029         /* If added prefix is link local and we are prepared to process
3030            router advertisements, start sending router solicitations.
3031          */
3032
3033         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3034              ifp->idev->cnf.accept_ra == 2) &&
3035             ifp->idev->cnf.rtr_solicits > 0 &&
3036             (dev->flags&IFF_LOOPBACK) == 0 &&
3037             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3038                 /*
3039                  *      If a host as already performed a random delay
3040                  *      [...] as part of DAD [...] there is no need
3041                  *      to delay again before sending the first RS
3042                  */
3043                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3044
3045                 spin_lock_bh(&ifp->lock);
3046                 ifp->probes = 1;
3047                 ifp->idev->if_flags |= IF_RS_SENT;
3048                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3049                 spin_unlock_bh(&ifp->lock);
3050         }
3051 }
3052
3053 static void addrconf_dad_run(struct inet6_dev *idev)
3054 {
3055         struct inet6_ifaddr *ifp;
3056
3057         read_lock_bh(&idev->lock);
3058         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3059                 spin_lock(&ifp->lock);
3060                 if (ifp->flags & IFA_F_TENTATIVE &&
3061                     ifp->state == INET6_IFADDR_STATE_DAD)
3062                         addrconf_dad_kick(ifp);
3063                 spin_unlock(&ifp->lock);
3064         }
3065         read_unlock_bh(&idev->lock);
3066 }
3067
3068 #ifdef CONFIG_PROC_FS
3069 struct if6_iter_state {
3070         struct seq_net_private p;
3071         int bucket;
3072         int offset;
3073 };
3074
3075 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos)
3076 {
3077         struct inet6_ifaddr *ifa = NULL;
3078         struct if6_iter_state *state = seq->private;
3079         struct net *net = seq_file_net(seq);
3080         int p = 0;
3081
3082         /* initial bucket if pos is 0 */
3083         if (pos == 0) {
3084                 state->bucket = 0;
3085                 state->offset = 0;
3086         }
3087
3088         for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3089                 struct hlist_node *n;
3090                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3091                                          addr_lst) {
3092                         /* sync with offset */
3093                         if (p < state->offset) {
3094                                 p++;
3095                                 continue;
3096                         }
3097                         state->offset++;
3098                         if (net_eq(dev_net(ifa->idev->dev), net))
3099                                 return ifa;
3100                 }
3101
3102                 /* prepare for next bucket */
3103                 state->offset = 0;
3104                 p = 0;
3105         }
3106         return NULL;
3107 }
3108
3109 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3110                                          struct inet6_ifaddr *ifa)
3111 {
3112         struct if6_iter_state *state = seq->private;
3113         struct net *net = seq_file_net(seq);
3114         struct hlist_node *n = &ifa->addr_lst;
3115
3116         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst) {
3117                 state->offset++;
3118                 if (net_eq(dev_net(ifa->idev->dev), net))
3119                         return ifa;
3120         }
3121
3122         while (++state->bucket < IN6_ADDR_HSIZE) {
3123                 state->offset = 0;
3124                 hlist_for_each_entry_rcu_bh(ifa, n,
3125                                      &inet6_addr_lst[state->bucket], addr_lst) {
3126                         state->offset++;
3127                         if (net_eq(dev_net(ifa->idev->dev), net))
3128                                 return ifa;
3129                 }
3130         }
3131
3132         return NULL;
3133 }
3134
3135 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3136         __acquires(rcu_bh)
3137 {
3138         rcu_read_lock_bh();
3139         return if6_get_first(seq, *pos);
3140 }
3141
3142 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3143 {
3144         struct inet6_ifaddr *ifa;
3145
3146         ifa = if6_get_next(seq, v);
3147         ++*pos;
3148         return ifa;
3149 }
3150
3151 static void if6_seq_stop(struct seq_file *seq, void *v)
3152         __releases(rcu_bh)
3153 {
3154         rcu_read_unlock_bh();
3155 }
3156
3157 static int if6_seq_show(struct seq_file *seq, void *v)
3158 {
3159         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3160         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3161                    &ifp->addr,
3162                    ifp->idev->dev->ifindex,
3163                    ifp->prefix_len,
3164                    ifp->scope,
3165                    ifp->flags,
3166                    ifp->idev->dev->name);
3167         return 0;
3168 }
3169
3170 static const struct seq_operations if6_seq_ops = {
3171         .start  = if6_seq_start,
3172         .next   = if6_seq_next,
3173         .show   = if6_seq_show,
3174         .stop   = if6_seq_stop,
3175 };
3176
3177 static int if6_seq_open(struct inode *inode, struct file *file)
3178 {
3179         return seq_open_net(inode, file, &if6_seq_ops,
3180                             sizeof(struct if6_iter_state));
3181 }
3182
3183 static const struct file_operations if6_fops = {
3184         .owner          = THIS_MODULE,
3185         .open           = if6_seq_open,
3186         .read           = seq_read,
3187         .llseek         = seq_lseek,
3188         .release        = seq_release_net,
3189 };
3190
3191 static int __net_init if6_proc_net_init(struct net *net)
3192 {
3193         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3194                 return -ENOMEM;
3195         return 0;
3196 }
3197
3198 static void __net_exit if6_proc_net_exit(struct net *net)
3199 {
3200        proc_net_remove(net, "if_inet6");
3201 }
3202
3203 static struct pernet_operations if6_proc_net_ops = {
3204        .init = if6_proc_net_init,
3205        .exit = if6_proc_net_exit,
3206 };
3207
3208 int __init if6_proc_init(void)
3209 {
3210         return register_pernet_subsys(&if6_proc_net_ops);
3211 }
3212
3213 void if6_proc_exit(void)
3214 {
3215         unregister_pernet_subsys(&if6_proc_net_ops);
3216 }
3217 #endif  /* CONFIG_PROC_FS */
3218
3219 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3220 /* Check if address is a home address configured on any interface. */
3221 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3222 {
3223         int ret = 0;
3224         struct inet6_ifaddr *ifp = NULL;
3225         struct hlist_node *n;
3226         unsigned int hash = ipv6_addr_hash(addr);
3227
3228         rcu_read_lock_bh();
3229         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3230                 if (!net_eq(dev_net(ifp->idev->dev), net))
3231                         continue;
3232                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3233                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3234                         ret = 1;
3235                         break;
3236                 }
3237         }
3238         rcu_read_unlock_bh();
3239         return ret;
3240 }
3241 #endif
3242
3243 /*
3244  *      Periodic address status verification
3245  */
3246
3247 static void addrconf_verify(unsigned long foo)
3248 {
3249         unsigned long now, next, next_sec, next_sched;
3250         struct inet6_ifaddr *ifp;
3251         struct hlist_node *node;
3252         int i;
3253
3254         rcu_read_lock_bh();
3255         spin_lock(&addrconf_verify_lock);
3256         now = jiffies;
3257         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3258
3259         del_timer(&addr_chk_timer);
3260
3261         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3262 restart:
3263                 hlist_for_each_entry_rcu_bh(ifp, node,
3264                                          &inet6_addr_lst[i], addr_lst) {
3265                         unsigned long age;
3266
3267                         if (ifp->flags & IFA_F_PERMANENT)
3268                                 continue;
3269
3270                         spin_lock(&ifp->lock);
3271                         /* We try to batch several events at once. */
3272                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3273
3274                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3275                             age >= ifp->valid_lft) {
3276                                 spin_unlock(&ifp->lock);
3277                                 in6_ifa_hold(ifp);
3278                                 ipv6_del_addr(ifp);
3279                                 goto restart;
3280                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3281                                 spin_unlock(&ifp->lock);
3282                                 continue;
3283                         } else if (age >= ifp->prefered_lft) {
3284                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3285                                 int deprecate = 0;
3286
3287                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3288                                         deprecate = 1;
3289                                         ifp->flags |= IFA_F_DEPRECATED;
3290                                 }
3291
3292                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3293                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3294
3295                                 spin_unlock(&ifp->lock);
3296
3297                                 if (deprecate) {
3298                                         in6_ifa_hold(ifp);
3299
3300                                         ipv6_ifa_notify(0, ifp);
3301                                         in6_ifa_put(ifp);
3302                                         goto restart;
3303                                 }
3304 #ifdef CONFIG_IPV6_PRIVACY
3305                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3306                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3307                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3308                                         ifp->idev->cnf.dad_transmits *
3309                                         ifp->idev->nd_parms->retrans_time / HZ;
3310
3311                                 if (age >= ifp->prefered_lft - regen_advance) {
3312                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3313                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3314                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3315                                         if (!ifp->regen_count && ifpub) {
3316                                                 ifp->regen_count++;
3317                                                 in6_ifa_hold(ifp);
3318                                                 in6_ifa_hold(ifpub);
3319                                                 spin_unlock(&ifp->lock);
3320
3321                                                 spin_lock(&ifpub->lock);
3322                                                 ifpub->regen_count = 0;
3323                                                 spin_unlock(&ifpub->lock);
3324                                                 ipv6_create_tempaddr(ifpub, ifp);
3325                                                 in6_ifa_put(ifpub);
3326                                                 in6_ifa_put(ifp);
3327                                                 goto restart;
3328                                         }
3329                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3330                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3331                                 spin_unlock(&ifp->lock);
3332 #endif
3333                         } else {
3334                                 /* ifp->prefered_lft <= ifp->valid_lft */
3335                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3336                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3337                                 spin_unlock(&ifp->lock);
3338                         }
3339                 }
3340         }
3341
3342         next_sec = round_jiffies_up(next);
3343         next_sched = next;
3344
3345         /* If rounded timeout is accurate enough, accept it. */
3346         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3347                 next_sched = next_sec;
3348
3349         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3350         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3351                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3352
3353         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3354               now, next, next_sec, next_sched));
3355
3356         addr_chk_timer.expires = next_sched;
3357         add_timer(&addr_chk_timer);
3358         spin_unlock(&addrconf_verify_lock);
3359         rcu_read_unlock_bh();
3360 }
3361
3362 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3363 {
3364         struct in6_addr *pfx = NULL;
3365
3366         if (addr)
3367                 pfx = nla_data(addr);
3368
3369         if (local) {
3370                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3371                         pfx = NULL;
3372                 else
3373                         pfx = nla_data(local);
3374         }
3375
3376         return pfx;
3377 }
3378
3379 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3380         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3381         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3382         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3383 };
3384
3385 static int
3386 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3387 {
3388         struct net *net = sock_net(skb->sk);
3389         struct ifaddrmsg *ifm;
3390         struct nlattr *tb[IFA_MAX+1];
3391         struct in6_addr *pfx;
3392         int err;
3393
3394         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3395         if (err < 0)
3396                 return err;
3397
3398         ifm = nlmsg_data(nlh);
3399         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3400         if (pfx == NULL)
3401                 return -EINVAL;
3402
3403         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3404 }
3405
3406 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3407                              u32 prefered_lft, u32 valid_lft)
3408 {
3409         u32 flags;
3410         clock_t expires;
3411         unsigned long timeout;
3412
3413         if (!valid_lft || (prefered_lft > valid_lft))
3414                 return -EINVAL;
3415
3416         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3417         if (addrconf_finite_timeout(timeout)) {
3418                 expires = jiffies_to_clock_t(timeout * HZ);
3419                 valid_lft = timeout;
3420                 flags = RTF_EXPIRES;
3421         } else {
3422                 expires = 0;
3423                 flags = 0;
3424                 ifa_flags |= IFA_F_PERMANENT;
3425         }
3426
3427         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3428         if (addrconf_finite_timeout(timeout)) {
3429                 if (timeout == 0)
3430                         ifa_flags |= IFA_F_DEPRECATED;
3431                 prefered_lft = timeout;
3432         }
3433
3434         spin_lock_bh(&ifp->lock);
3435         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3436         ifp->tstamp = jiffies;
3437         ifp->valid_lft = valid_lft;
3438         ifp->prefered_lft = prefered_lft;
3439
3440         spin_unlock_bh(&ifp->lock);
3441         if (!(ifp->flags&IFA_F_TENTATIVE))
3442                 ipv6_ifa_notify(0, ifp);
3443
3444         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3445                               expires, flags);
3446         addrconf_verify(0);
3447
3448         return 0;
3449 }
3450
3451 static int
3452 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3453 {
3454         struct net *net = sock_net(skb->sk);
3455         struct ifaddrmsg *ifm;
3456         struct nlattr *tb[IFA_MAX+1];
3457         struct in6_addr *pfx;
3458         struct inet6_ifaddr *ifa;
3459         struct net_device *dev;
3460         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3461         u8 ifa_flags;
3462         int err;
3463
3464         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3465         if (err < 0)
3466                 return err;
3467
3468         ifm = nlmsg_data(nlh);
3469         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3470         if (pfx == NULL)
3471                 return -EINVAL;
3472
3473         if (tb[IFA_CACHEINFO]) {
3474                 struct ifa_cacheinfo *ci;
3475
3476                 ci = nla_data(tb[IFA_CACHEINFO]);
3477                 valid_lft = ci->ifa_valid;
3478                 preferred_lft = ci->ifa_prefered;
3479         } else {
3480                 preferred_lft = INFINITY_LIFE_TIME;
3481                 valid_lft = INFINITY_LIFE_TIME;
3482         }
3483
3484         dev =  __dev_get_by_index(net, ifm->ifa_index);
3485         if (dev == NULL)
3486                 return -ENODEV;
3487
3488         /* We ignore other flags so far. */
3489         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3490
3491         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3492         if (ifa == NULL) {
3493                 /*
3494                  * It would be best to check for !NLM_F_CREATE here but
3495                  * userspace alreay relies on not having to provide this.
3496                  */
3497                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3498                                       ifm->ifa_prefixlen, ifa_flags,
3499                                       preferred_lft, valid_lft);
3500         }
3501
3502         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3503             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3504                 err = -EEXIST;
3505         else
3506                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3507
3508         in6_ifa_put(ifa);
3509
3510         return err;
3511 }
3512
3513 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3514                           u8 scope, int ifindex)
3515 {
3516         struct ifaddrmsg *ifm;
3517
3518         ifm = nlmsg_data(nlh);
3519         ifm->ifa_family = AF_INET6;
3520         ifm->ifa_prefixlen = prefixlen;
3521         ifm->ifa_flags = flags;
3522         ifm->ifa_scope = scope;
3523         ifm->ifa_index = ifindex;
3524 }
3525
3526 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3527                          unsigned long tstamp, u32 preferred, u32 valid)
3528 {
3529         struct ifa_cacheinfo ci;
3530
3531         ci.cstamp = cstamp_delta(cstamp);
3532         ci.tstamp = cstamp_delta(tstamp);
3533         ci.ifa_prefered = preferred;
3534         ci.ifa_valid = valid;
3535
3536         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3537 }
3538
3539 static inline int rt_scope(int ifa_scope)
3540 {
3541         if (ifa_scope & IFA_HOST)
3542                 return RT_SCOPE_HOST;
3543         else if (ifa_scope & IFA_LINK)
3544                 return RT_SCOPE_LINK;
3545         else if (ifa_scope & IFA_SITE)
3546                 return RT_SCOPE_SITE;
3547         else
3548                 return RT_SCOPE_UNIVERSE;
3549 }
3550
3551 static inline int inet6_ifaddr_msgsize(void)
3552 {
3553         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3554                + nla_total_size(16) /* IFA_ADDRESS */
3555                + nla_total_size(sizeof(struct ifa_cacheinfo));
3556 }
3557
3558 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3559                              u32 pid, u32 seq, int event, unsigned int flags)
3560 {
3561         struct nlmsghdr  *nlh;
3562         u32 preferred, valid;
3563
3564         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3565         if (nlh == NULL)
3566                 return -EMSGSIZE;
3567
3568         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3569                       ifa->idev->dev->ifindex);
3570
3571         if (!(ifa->flags&IFA_F_PERMANENT)) {
3572                 preferred = ifa->prefered_lft;
3573                 valid = ifa->valid_lft;
3574                 if (preferred != INFINITY_LIFE_TIME) {
3575                         long tval = (jiffies - ifa->tstamp)/HZ;
3576                         if (preferred > tval)
3577                                 preferred -= tval;
3578                         else
3579                                 preferred = 0;
3580                         if (valid != INFINITY_LIFE_TIME) {
3581                                 if (valid > tval)
3582                                         valid -= tval;
3583                                 else
3584                                         valid = 0;
3585                         }
3586                 }
3587         } else {
3588                 preferred = INFINITY_LIFE_TIME;
3589                 valid = INFINITY_LIFE_TIME;
3590         }
3591
3592         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3593             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3594                 nlmsg_cancel(skb, nlh);
3595                 return -EMSGSIZE;
3596         }
3597
3598         return nlmsg_end(skb, nlh);
3599 }
3600
3601 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3602                                 u32 pid, u32 seq, int event, u16 flags)
3603 {
3604         struct nlmsghdr  *nlh;
3605         u8 scope = RT_SCOPE_UNIVERSE;
3606         int ifindex = ifmca->idev->dev->ifindex;
3607
3608         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3609                 scope = RT_SCOPE_SITE;
3610
3611         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3612         if (nlh == NULL)
3613                 return -EMSGSIZE;
3614
3615         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3616         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3617             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3618                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3619                 nlmsg_cancel(skb, nlh);
3620                 return -EMSGSIZE;
3621         }
3622
3623         return nlmsg_end(skb, nlh);
3624 }
3625
3626 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3627                                 u32 pid, u32 seq, int event, unsigned int flags)
3628 {
3629         struct nlmsghdr  *nlh;
3630         u8 scope = RT_SCOPE_UNIVERSE;
3631         int ifindex = ifaca->aca_idev->dev->ifindex;
3632
3633         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3634                 scope = RT_SCOPE_SITE;
3635
3636         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3637         if (nlh == NULL)
3638                 return -EMSGSIZE;
3639
3640         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3641         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3642             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3643                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3644                 nlmsg_cancel(skb, nlh);
3645                 return -EMSGSIZE;
3646         }
3647
3648         return nlmsg_end(skb, nlh);
3649 }
3650
3651 enum addr_type_t {
3652         UNICAST_ADDR,
3653         MULTICAST_ADDR,
3654         ANYCAST_ADDR,
3655 };
3656
3657 /* called with rcu_read_lock() */
3658 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3659                           struct netlink_callback *cb, enum addr_type_t type,
3660                           int s_ip_idx, int *p_ip_idx)
3661 {
3662         struct ifmcaddr6 *ifmca;
3663         struct ifacaddr6 *ifaca;
3664         int err = 1;
3665         int ip_idx = *p_ip_idx;
3666
3667         read_lock_bh(&idev->lock);
3668         switch (type) {
3669         case UNICAST_ADDR: {
3670                 struct inet6_ifaddr *ifa;
3671
3672                 /* unicast address incl. temp addr */
3673                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3674                         if (++ip_idx < s_ip_idx)
3675                                 continue;
3676                         err = inet6_fill_ifaddr(skb, ifa,
3677                                                 NETLINK_CB(cb->skb).pid,
3678                                                 cb->nlh->nlmsg_seq,
3679                                                 RTM_NEWADDR,
3680                                                 NLM_F_MULTI);
3681                         if (err <= 0)
3682                                 break;
3683                 }
3684                 break;
3685         }
3686         case MULTICAST_ADDR:
3687                 /* multicast address */
3688                 for (ifmca = idev->mc_list; ifmca;
3689                      ifmca = ifmca->next, ip_idx++) {
3690                         if (ip_idx < s_ip_idx)
3691                                 continue;
3692                         err = inet6_fill_ifmcaddr(skb, ifmca,
3693                                                   NETLINK_CB(cb->skb).pid,
3694                                                   cb->nlh->nlmsg_seq,
3695                                                   RTM_GETMULTICAST,
3696                                                   NLM_F_MULTI);
3697                         if (err <= 0)
3698                                 break;
3699                 }
3700                 break;
3701         case ANYCAST_ADDR:
3702                 /* anycast address */
3703                 for (ifaca = idev->ac_list; ifaca;
3704                      ifaca = ifaca->aca_next, ip_idx++) {
3705                         if (ip_idx < s_ip_idx)
3706                                 continue;
3707                         err = inet6_fill_ifacaddr(skb, ifaca,
3708                                                   NETLINK_CB(cb->skb).pid,
3709                                                   cb->nlh->nlmsg_seq,
3710                                                   RTM_GETANYCAST,
3711                                                   NLM_F_MULTI);
3712                         if (err <= 0)
3713                                 break;
3714                 }
3715                 break;
3716         default:
3717                 break;
3718         }
3719         read_unlock_bh(&idev->lock);
3720         *p_ip_idx = ip_idx;
3721         return err;
3722 }
3723
3724 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3725                            enum addr_type_t type)
3726 {
3727         struct net *net = sock_net(skb->sk);
3728         int h, s_h;
3729         int idx, ip_idx;
3730         int s_idx, s_ip_idx;
3731         struct net_device *dev;
3732         struct inet6_dev *idev;
3733         struct hlist_head *head;
3734         struct hlist_node *node;
3735
3736         s_h = cb->args[0];
3737         s_idx = idx = cb->args[1];
3738         s_ip_idx = ip_idx = cb->args[2];
3739
3740         rcu_read_lock();
3741         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3742                 idx = 0;
3743                 head = &net->dev_index_head[h];
3744                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3745                         if (idx < s_idx)
3746                                 goto cont;
3747                         if (h > s_h || idx > s_idx)
3748                                 s_ip_idx = 0;
3749                         ip_idx = 0;
3750                         idev = __in6_dev_get(dev);
3751                         if (!idev)
3752                                 goto cont;
3753
3754                         if (in6_dump_addrs(idev, skb, cb, type,
3755                                            s_ip_idx, &ip_idx) <= 0)
3756                                 goto done;
3757 cont:
3758                         idx++;
3759                 }
3760         }
3761 done:
3762         rcu_read_unlock();
3763         cb->args[0] = h;
3764         cb->args[1] = idx;
3765         cb->args[2] = ip_idx;
3766
3767         return skb->len;
3768 }
3769
3770 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3771 {
3772         enum addr_type_t type = UNICAST_ADDR;
3773
3774         return inet6_dump_addr(skb, cb, type);
3775 }
3776
3777 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3778 {
3779         enum addr_type_t type = MULTICAST_ADDR;
3780
3781         return inet6_dump_addr(skb, cb, type);
3782 }
3783
3784
3785 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3786 {
3787         enum addr_type_t type = ANYCAST_ADDR;
3788
3789         return inet6_dump_addr(skb, cb, type);
3790 }
3791
3792 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3793                              void *arg)
3794 {
3795         struct net *net = sock_net(in_skb->sk);
3796         struct ifaddrmsg *ifm;
3797         struct nlattr *tb[IFA_MAX+1];
3798         struct in6_addr *addr = NULL;
3799         struct net_device *dev = NULL;
3800         struct inet6_ifaddr *ifa;
3801         struct sk_buff *skb;
3802         int err;
3803
3804         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3805         if (err < 0)
3806                 goto errout;
3807
3808         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3809         if (addr == NULL) {
3810                 err = -EINVAL;
3811                 goto errout;
3812         }
3813
3814         ifm = nlmsg_data(nlh);
3815         if (ifm->ifa_index)
3816                 dev = __dev_get_by_index(net, ifm->ifa_index);
3817
3818         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3819         if (!ifa) {
3820                 err = -EADDRNOTAVAIL;
3821                 goto errout;
3822         }
3823
3824         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3825         if (!skb) {
3826                 err = -ENOBUFS;
3827                 goto errout_ifa;
3828         }
3829
3830         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3831                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3832         if (err < 0) {
3833                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3834                 WARN_ON(err == -EMSGSIZE);
3835                 kfree_skb(skb);
3836                 goto errout_ifa;
3837         }
3838         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3839 errout_ifa:
3840         in6_ifa_put(ifa);
3841 errout:
3842         return err;
3843 }
3844
3845 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3846 {
3847         struct sk_buff *skb;
3848         struct net *net = dev_net(ifa->idev->dev);
3849         int err = -ENOBUFS;
3850
3851         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3852         if (skb == NULL)
3853                 goto errout;
3854
3855         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3856         if (err < 0) {
3857                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3858                 WARN_ON(err == -EMSGSIZE);
3859                 kfree_skb(skb);
3860                 goto errout;
3861         }
3862         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3863         return;
3864 errout:
3865         if (err < 0)
3866                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3867 }
3868
3869 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3870                                 __s32 *array, int bytes)
3871 {
3872         BUG_ON(bytes < (DEVCONF_MAX * 4));
3873
3874         memset(array, 0, bytes);
3875         array[DEVCONF_FORWARDING] = cnf->forwarding;
3876         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3877         array[DEVCONF_MTU6] = cnf->mtu6;
3878         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3879         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3880         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3881         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3882         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3883         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3884                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3885         array[DEVCONF_RTR_SOLICIT_DELAY] =
3886                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3887         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3888 #ifdef CONFIG_IPV6_PRIVACY
3889         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3890         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3891         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3892         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3893         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3894 #endif
3895         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3896         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3897         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3898 #ifdef CONFIG_IPV6_ROUTER_PREF
3899         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3900         array[DEVCONF_RTR_PROBE_INTERVAL] =
3901                 jiffies_to_msecs(cnf->rtr_probe_interval);
3902 #ifdef CONFIG_IPV6_ROUTE_INFO
3903         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3904 #endif
3905 #endif
3906         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3907         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3908 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3909         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3910 #endif
3911 #ifdef CONFIG_IPV6_MROUTE
3912         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3913 #endif
3914         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3915         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3916         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3917 }
3918
3919 static inline size_t inet6_ifla6_size(void)
3920 {
3921         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3922              + nla_total_size(sizeof(struct ifla_cacheinfo))
3923              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3924              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3925              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3926 }
3927
3928 static inline size_t inet6_if_nlmsg_size(void)
3929 {
3930         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3931                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3932                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3933                + nla_total_size(4) /* IFLA_MTU */
3934                + nla_total_size(4) /* IFLA_LINK */
3935                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3936 }
3937
3938 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3939                                       int items, int bytes)
3940 {
3941         int i;
3942         int pad = bytes - sizeof(u64) * items;
3943         BUG_ON(pad < 0);
3944
3945         /* Use put_unaligned() because stats may not be aligned for u64. */
3946         put_unaligned(items, &stats[0]);
3947         for (i = 1; i < items; i++)
3948                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3949
3950         memset(&stats[items], 0, pad);
3951 }
3952
3953 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3954                                       int items, int bytes, size_t syncpoff)
3955 {
3956         int i;
3957         int pad = bytes - sizeof(u64) * items;
3958         BUG_ON(pad < 0);
3959
3960         /* Use put_unaligned() because stats may not be aligned for u64. */
3961         put_unaligned(items, &stats[0]);
3962         for (i = 1; i < items; i++)
3963                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3964
3965         memset(&stats[items], 0, pad);
3966 }
3967
3968 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3969                              int bytes)
3970 {
3971         switch (attrtype) {
3972         case IFLA_INET6_STATS:
3973                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3974                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3975                 break;
3976         case IFLA_INET6_ICMP6STATS:
3977                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3978                 break;
3979         }
3980 }
3981
3982 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3983 {
3984         struct nlattr *nla;
3985         struct ifla_cacheinfo ci;
3986
3987         if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags))
3988                 goto nla_put_failure;
3989         ci.max_reasm_len = IPV6_MAXPLEN;
3990         ci.tstamp = cstamp_delta(idev->tstamp);
3991         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3992         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3993         if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci))
3994                 goto nla_put_failure;
3995         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3996         if (nla == NULL)
3997                 goto nla_put_failure;
3998         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3999
4000         /* XXX - MC not implemented */
4001
4002         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4003         if (nla == NULL)
4004                 goto nla_put_failure;
4005         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4006
4007         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4008         if (nla == NULL)
4009                 goto nla_put_failure;
4010         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4011
4012         return 0;
4013
4014 nla_put_failure:
4015         return -EMSGSIZE;
4016 }
4017
4018 static size_t inet6_get_link_af_size(const struct net_device *dev)
4019 {
4020         if (!__in6_dev_get(dev))
4021                 return 0;
4022
4023         return inet6_ifla6_size();
4024 }
4025
4026 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4027 {
4028         struct inet6_dev *idev = __in6_dev_get(dev);
4029
4030         if (!idev)
4031                 return -ENODATA;
4032
4033         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4034                 return -EMSGSIZE;
4035
4036         return 0;
4037 }
4038
4039 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4040                              u32 pid, u32 seq, int event, unsigned int flags)
4041 {
4042         struct net_device *dev = idev->dev;
4043         struct ifinfomsg *hdr;
4044         struct nlmsghdr *nlh;
4045         void *protoinfo;
4046
4047         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4048         if (nlh == NULL)
4049                 return -EMSGSIZE;
4050
4051         hdr = nlmsg_data(nlh);
4052         hdr->ifi_family = AF_INET6;
4053         hdr->__ifi_pad = 0;
4054         hdr->ifi_type = dev->type;
4055         hdr->ifi_index = dev->ifindex;
4056         hdr->ifi_flags = dev_get_flags(dev);
4057         hdr->ifi_change = 0;
4058
4059         if (nla_put_string(skb, IFLA_IFNAME, dev->name) ||
4060             (dev->addr_len &&
4061              nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) ||
4062             nla_put_u32(skb, IFLA_MTU, dev->mtu) ||
4063             (dev->ifindex != dev->iflink &&
4064              nla_put_u32(skb, IFLA_LINK, dev->iflink)))
4065                 goto nla_put_failure;
4066         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4067         if (protoinfo == NULL)
4068                 goto nla_put_failure;
4069
4070         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4071                 goto nla_put_failure;
4072
4073         nla_nest_end(skb, protoinfo);
4074         return nlmsg_end(skb, nlh);
4075
4076 nla_put_failure:
4077         nlmsg_cancel(skb, nlh);
4078         return -EMSGSIZE;
4079 }
4080
4081 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4082 {
4083         struct net *net = sock_net(skb->sk);
4084         int h, s_h;
4085         int idx = 0, s_idx;
4086         struct net_device *dev;
4087         struct inet6_dev *idev;
4088         struct hlist_head *head;
4089         struct hlist_node *node;
4090
4091         s_h = cb->args[0];
4092         s_idx = cb->args[1];
4093
4094         rcu_read_lock();
4095         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4096                 idx = 0;
4097                 head = &net->dev_index_head[h];
4098                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4099                         if (idx < s_idx)
4100                                 goto cont;
4101                         idev = __in6_dev_get(dev);
4102                         if (!idev)
4103                                 goto cont;
4104                         if (inet6_fill_ifinfo(skb, idev,
4105                                               NETLINK_CB(cb->skb).pid,
4106                                               cb->nlh->nlmsg_seq,
4107                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4108                                 goto out;
4109 cont:
4110                         idx++;
4111                 }
4112         }
4113 out:
4114         rcu_read_unlock();
4115         cb->args[1] = idx;
4116         cb->args[0] = h;
4117
4118         return skb->len;
4119 }
4120
4121 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4122 {
4123         struct sk_buff *skb;
4124         struct net *net = dev_net(idev->dev);
4125         int err = -ENOBUFS;
4126
4127         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4128         if (skb == NULL)
4129                 goto errout;
4130
4131         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4132         if (err < 0) {
4133                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4134                 WARN_ON(err == -EMSGSIZE);
4135                 kfree_skb(skb);
4136                 goto errout;
4137         }
4138         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4139         return;
4140 errout:
4141         if (err < 0)
4142                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4143 }
4144
4145 static inline size_t inet6_prefix_nlmsg_size(void)
4146 {
4147         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4148                + nla_total_size(sizeof(struct in6_addr))
4149                + nla_total_size(sizeof(struct prefix_cacheinfo));
4150 }
4151
4152 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4153                              struct prefix_info *pinfo, u32 pid, u32 seq,
4154                              int event, unsigned int flags)
4155 {
4156         struct prefixmsg *pmsg;
4157         struct nlmsghdr *nlh;
4158         struct prefix_cacheinfo ci;
4159
4160         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4161         if (nlh == NULL)
4162                 return -EMSGSIZE;
4163
4164         pmsg = nlmsg_data(nlh);
4165         pmsg->prefix_family = AF_INET6;
4166         pmsg->prefix_pad1 = 0;
4167         pmsg->prefix_pad2 = 0;
4168         pmsg->prefix_ifindex = idev->dev->ifindex;
4169         pmsg->prefix_len = pinfo->prefix_len;
4170         pmsg->prefix_type = pinfo->type;
4171         pmsg->prefix_pad3 = 0;
4172         pmsg->prefix_flags = 0;
4173         if (pinfo->onlink)
4174                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4175         if (pinfo->autoconf)
4176                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4177
4178         if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix))
4179                 goto nla_put_failure;
4180         ci.preferred_time = ntohl(pinfo->prefered);
4181         ci.valid_time = ntohl(pinfo->valid);
4182         if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci))
4183                 goto nla_put_failure;
4184         return nlmsg_end(skb, nlh);
4185
4186 nla_put_failure:
4187         nlmsg_cancel(skb, nlh);
4188         return -EMSGSIZE;
4189 }
4190
4191 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4192                          struct prefix_info *pinfo)
4193 {
4194         struct sk_buff *skb;
4195         struct net *net = dev_net(idev->dev);
4196         int err = -ENOBUFS;
4197
4198         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4199         if (skb == NULL)
4200                 goto errout;
4201
4202         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4203         if (err < 0) {
4204                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4205                 WARN_ON(err == -EMSGSIZE);
4206                 kfree_skb(skb);
4207                 goto errout;
4208         }
4209         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4210         return;
4211 errout:
4212         if (err < 0)
4213                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4214 }
4215
4216 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4217 {
4218         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4219
4220         switch (event) {
4221         case RTM_NEWADDR:
4222                 /*
4223                  * If the address was optimistic
4224                  * we inserted the route at the start of
4225                  * our DAD process, so we don't need
4226                  * to do it again
4227                  */
4228                 if (!(ifp->rt->rt6i_node))
4229                         ip6_ins_rt(ifp->rt);
4230                 if (ifp->idev->cnf.forwarding)
4231                         addrconf_join_anycast(ifp);
4232                 break;
4233         case RTM_DELADDR:
4234                 if (ifp->idev->cnf.forwarding)
4235                         addrconf_leave_anycast(ifp);
4236                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4237                 dst_hold(&ifp->rt->dst);
4238
4239                 if (ip6_del_rt(ifp->rt))
4240                         dst_free(&ifp->rt->dst);
4241                 break;
4242         }
4243 }
4244
4245 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4246 {
4247         rcu_read_lock_bh();
4248         if (likely(ifp->idev->dead == 0))
4249                 __ipv6_ifa_notify(event, ifp);
4250         rcu_read_unlock_bh();
4251 }
4252
4253 #ifdef CONFIG_SYSCTL
4254
4255 static
4256 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4257                            void __user *buffer, size_t *lenp, loff_t *ppos)
4258 {
4259         int *valp = ctl->data;
4260         int val = *valp;
4261         loff_t pos = *ppos;
4262         ctl_table lctl;
4263         int ret;
4264
4265         /*
4266          * ctl->data points to idev->cnf.forwarding, we should
4267          * not modify it until we get the rtnl lock.
4268          */
4269         lctl = *ctl;
4270         lctl.data = &val;
4271
4272         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4273
4274         if (write)
4275                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4276         if (ret)
4277                 *ppos = pos;
4278         return ret;
4279 }
4280
4281 static void dev_disable_change(struct inet6_dev *idev)
4282 {
4283         if (!idev || !idev->dev)
4284                 return;
4285
4286         if (idev->cnf.disable_ipv6)
4287                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4288         else
4289                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4290 }
4291
4292 static void addrconf_disable_change(struct net *net, __s32 newf)
4293 {
4294         struct net_device *dev;
4295         struct inet6_dev *idev;
4296
4297         rcu_read_lock();
4298         for_each_netdev_rcu(net, dev) {
4299                 idev = __in6_dev_get(dev);
4300                 if (idev) {
4301                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4302                         idev->cnf.disable_ipv6 = newf;
4303                         if (changed)
4304                                 dev_disable_change(idev);
4305                 }
4306         }
4307         rcu_read_unlock();
4308 }
4309
4310 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf)
4311 {
4312         struct net *net;
4313         int old;
4314
4315         if (!rtnl_trylock())
4316                 return restart_syscall();
4317
4318         net = (struct net *)table->extra2;
4319         old = *p;
4320         *p = newf;
4321
4322         if (p == &net->ipv6.devconf_dflt->disable_ipv6) {
4323                 rtnl_unlock();
4324                 return 0;
4325         }
4326
4327         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4328                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4329                 addrconf_disable_change(net, newf);
4330         } else if ((!newf) ^ (!old))
4331                 dev_disable_change((struct inet6_dev *)table->extra1);
4332
4333         rtnl_unlock();
4334         return 0;
4335 }
4336
4337 static
4338 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4339                             void __user *buffer, size_t *lenp, loff_t *ppos)
4340 {
4341         int *valp = ctl->data;
4342         int val = *valp;
4343         loff_t pos = *ppos;
4344         ctl_table lctl;
4345         int ret;
4346
4347         /*
4348          * ctl->data points to idev->cnf.disable_ipv6, we should
4349          * not modify it until we get the rtnl lock.
4350          */
4351         lctl = *ctl;
4352         lctl.data = &val;
4353
4354         ret = proc_dointvec(&lctl, write, buffer, lenp, ppos);
4355
4356         if (write)
4357                 ret = addrconf_disable_ipv6(ctl, valp, val);
4358         if (ret)
4359                 *ppos = pos;
4360         return ret;
4361 }
4362
4363 static struct addrconf_sysctl_table
4364 {
4365         struct ctl_table_header *sysctl_header;
4366         ctl_table addrconf_vars[DEVCONF_MAX+1];
4367 } addrconf_sysctl __read_mostly = {
4368         .sysctl_header = NULL,
4369         .addrconf_vars = {
4370                 {
4371                         .procname       = "forwarding",
4372                         .data           = &ipv6_devconf.forwarding,
4373                         .maxlen         = sizeof(int),
4374                         .mode           = 0644,
4375                         .proc_handler   = addrconf_sysctl_forward,
4376                 },
4377                 {
4378                         .procname       = "hop_limit",
4379                         .data           = &ipv6_devconf.hop_limit,
4380                         .maxlen         = sizeof(int),
4381                         .mode           = 0644,
4382                         .proc_handler   = proc_dointvec,
4383                 },
4384                 {
4385                         .procname       = "mtu",
4386                         .data           = &ipv6_devconf.mtu6,
4387                         .maxlen         = sizeof(int),
4388                         .mode           = 0644,
4389                         .proc_handler   = proc_dointvec,
4390                 },
4391                 {
4392                         .procname       = "accept_ra",
4393                         .data           = &ipv6_devconf.accept_ra,
4394                         .maxlen         = sizeof(int),
4395                         .mode           = 0644,
4396                         .proc_handler   = proc_dointvec,
4397                 },
4398                 {
4399                         .procname       = "accept_redirects",
4400                         .data           = &ipv6_devconf.accept_redirects,
4401                         .maxlen         = sizeof(int),
4402                         .mode           = 0644,
4403                         .proc_handler   = proc_dointvec,
4404                 },
4405                 {
4406                         .procname       = "autoconf",
4407                         .data           = &ipv6_devconf.autoconf,
4408                         .maxlen         = sizeof(int),
4409                         .mode           = 0644,
4410                         .proc_handler   = proc_dointvec,
4411                 },
4412                 {
4413                         .procname       = "dad_transmits",
4414                         .data           = &ipv6_devconf.dad_transmits,
4415                         .maxlen         = sizeof(int),
4416                         .mode           = 0644,
4417                         .proc_handler   = proc_dointvec,
4418                 },
4419                 {
4420                         .procname       = "router_solicitations",
4421                         .data           = &ipv6_devconf.rtr_solicits,
4422                         .maxlen         = sizeof(int),
4423                         .mode           = 0644,
4424                         .proc_handler   = proc_dointvec,
4425                 },
4426                 {
4427                         .procname       = "router_solicitation_interval",
4428                         .data           = &ipv6_devconf.rtr_solicit_interval,
4429                         .maxlen         = sizeof(int),
4430                         .mode           = 0644,
4431                         .proc_handler   = proc_dointvec_jiffies,
4432                 },
4433                 {
4434                         .procname       = "router_solicitation_delay",
4435                         .data           = &ipv6_devconf.rtr_solicit_delay,
4436                         .maxlen         = sizeof(int),
4437                         .mode           = 0644,
4438                         .proc_handler   = proc_dointvec_jiffies,
4439                 },
4440                 {
4441                         .procname       = "force_mld_version",
4442                         .data           = &ipv6_devconf.force_mld_version,
4443                         .maxlen         = sizeof(int),
4444                         .mode           = 0644,
4445                         .proc_handler   = proc_dointvec,
4446                 },
4447 #ifdef CONFIG_IPV6_PRIVACY
4448                 {
4449                         .procname       = "use_tempaddr",
4450                         .data           = &ipv6_devconf.use_tempaddr,
4451                         .maxlen         = sizeof(int),
4452                         .mode           = 0644,
4453                         .proc_handler   = proc_dointvec,
4454                 },
4455                 {
4456                         .procname       = "temp_valid_lft",
4457                         .data           = &ipv6_devconf.temp_valid_lft,
4458                         .maxlen         = sizeof(int),
4459                         .mode           = 0644,
4460                         .proc_handler   = proc_dointvec,
4461                 },
4462                 {
4463                         .procname       = "temp_prefered_lft",
4464                         .data           = &ipv6_devconf.temp_prefered_lft,
4465                         .maxlen         = sizeof(int),
4466                         .mode           = 0644,
4467                         .proc_handler   = proc_dointvec,
4468                 },
4469                 {
4470                         .procname       = "regen_max_retry",
4471                         .data           = &ipv6_devconf.regen_max_retry,
4472                         .maxlen         = sizeof(int),
4473                         .mode           = 0644,
4474                         .proc_handler   = proc_dointvec,
4475                 },
4476                 {
4477                         .procname       = "max_desync_factor",
4478                         .data           = &ipv6_devconf.max_desync_factor,
4479                         .maxlen         = sizeof(int),
4480                         .mode           = 0644,
4481                         .proc_handler   = proc_dointvec,
4482                 },
4483 #endif
4484                 {
4485                         .procname       = "max_addresses",
4486                         .data           = &ipv6_devconf.max_addresses,
4487                         .maxlen         = sizeof(int),
4488                         .mode           = 0644,
4489                         .proc_handler   = proc_dointvec,
4490                 },
4491                 {
4492                         .procname       = "accept_ra_defrtr",
4493                         .data           = &ipv6_devconf.accept_ra_defrtr,
4494                         .maxlen         = sizeof(int),
4495                         .mode           = 0644,
4496                         .proc_handler   = proc_dointvec,
4497                 },
4498                 {
4499                         .procname       = "accept_ra_pinfo",
4500                         .data           = &ipv6_devconf.accept_ra_pinfo,
4501                         .maxlen         = sizeof(int),
4502                         .mode           = 0644,
4503                         .proc_handler   = proc_dointvec,
4504                 },
4505 #ifdef CONFIG_IPV6_ROUTER_PREF
4506                 {
4507                         .procname       = "accept_ra_rtr_pref",
4508                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4509                         .maxlen         = sizeof(int),
4510                         .mode           = 0644,
4511                         .proc_handler   = proc_dointvec,
4512                 },
4513                 {
4514                         .procname       = "router_probe_interval",
4515                         .data           = &ipv6_devconf.rtr_probe_interval,
4516                         .maxlen         = sizeof(int),
4517                         .mode           = 0644,
4518                         .proc_handler   = proc_dointvec_jiffies,
4519                 },
4520 #ifdef CONFIG_IPV6_ROUTE_INFO
4521                 {
4522                         .procname       = "accept_ra_rt_info_max_plen",
4523                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4524                         .maxlen         = sizeof(int),
4525                         .mode           = 0644,
4526                         .proc_handler   = proc_dointvec,
4527                 },
4528 #endif
4529 #endif
4530                 {
4531                         .procname       = "proxy_ndp",
4532                         .data           = &ipv6_devconf.proxy_ndp,
4533                         .maxlen         = sizeof(int),
4534                         .mode           = 0644,
4535                         .proc_handler   = proc_dointvec,
4536                 },
4537                 {
4538                         .procname       = "accept_source_route",
4539                         .data           = &ipv6_devconf.accept_source_route,
4540                         .maxlen         = sizeof(int),
4541                         .mode           = 0644,
4542                         .proc_handler   = proc_dointvec,
4543                 },
4544 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4545                 {
4546                         .procname       = "optimistic_dad",
4547                         .data           = &ipv6_devconf.optimistic_dad,
4548                         .maxlen         = sizeof(int),
4549                         .mode           = 0644,
4550                         .proc_handler   = proc_dointvec,
4551
4552                 },
4553 #endif
4554 #ifdef CONFIG_IPV6_MROUTE
4555                 {
4556                         .procname       = "mc_forwarding",
4557                         .data           = &ipv6_devconf.mc_forwarding,
4558                         .maxlen         = sizeof(int),
4559                         .mode           = 0444,
4560                         .proc_handler   = proc_dointvec,
4561                 },
4562 #endif
4563                 {
4564                         .procname       = "disable_ipv6",
4565                         .data           = &ipv6_devconf.disable_ipv6,
4566                         .maxlen         = sizeof(int),
4567                         .mode           = 0644,
4568                         .proc_handler   = addrconf_sysctl_disable,
4569                 },
4570                 {
4571                         .procname       = "accept_dad",
4572                         .data           = &ipv6_devconf.accept_dad,
4573                         .maxlen         = sizeof(int),
4574                         .mode           = 0644,
4575                         .proc_handler   = proc_dointvec,
4576                 },
4577                 {
4578                         .procname       = "force_tllao",
4579                         .data           = &ipv6_devconf.force_tllao,
4580                         .maxlen         = sizeof(int),
4581                         .mode           = 0644,
4582                         .proc_handler   = proc_dointvec
4583                 },
4584                 {
4585                         /* sentinel */
4586                 }
4587         },
4588 };
4589
4590 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4591                 struct inet6_dev *idev, struct ipv6_devconf *p)
4592 {
4593         int i;
4594         struct addrconf_sysctl_table *t;
4595         char path[sizeof("net/ipv6/conf/") + IFNAMSIZ];
4596
4597         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4598         if (t == NULL)
4599                 goto out;
4600
4601         for (i = 0; t->addrconf_vars[i].data; i++) {
4602                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4603                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4604                 t->addrconf_vars[i].extra2 = net;
4605         }
4606
4607         snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name);
4608
4609         t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars);
4610         if (t->sysctl_header == NULL)
4611                 goto free;
4612
4613         p->sysctl = t;
4614         return 0;
4615
4616 free:
4617         kfree(t);
4618 out:
4619         return -ENOBUFS;
4620 }
4621
4622 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4623 {
4624         struct addrconf_sysctl_table *t;
4625
4626         if (p->sysctl == NULL)
4627                 return;
4628
4629         t = p->sysctl;
4630         p->sysctl = NULL;
4631         unregister_net_sysctl_table(t->sysctl_header);
4632         kfree(t);
4633 }
4634
4635 static void addrconf_sysctl_register(struct inet6_dev *idev)
4636 {
4637         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4638                               &ndisc_ifinfo_sysctl_change);
4639         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4640                                         idev, &idev->cnf);
4641 }
4642
4643 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4644 {
4645         __addrconf_sysctl_unregister(&idev->cnf);
4646         neigh_sysctl_unregister(idev->nd_parms);
4647 }
4648
4649
4650 #endif
4651
4652 static int __net_init addrconf_init_net(struct net *net)
4653 {
4654         int err;
4655         struct ipv6_devconf *all, *dflt;
4656
4657         err = -ENOMEM;
4658         all = &ipv6_devconf;
4659         dflt = &ipv6_devconf_dflt;
4660
4661         if (!net_eq(net, &init_net)) {
4662                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4663                 if (all == NULL)
4664                         goto err_alloc_all;
4665
4666                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4667                 if (dflt == NULL)
4668                         goto err_alloc_dflt;
4669         } else {
4670                 /* these will be inherited by all namespaces */
4671                 dflt->autoconf = ipv6_defaults.autoconf;
4672                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4673         }
4674
4675         net->ipv6.devconf_all = all;
4676         net->ipv6.devconf_dflt = dflt;
4677
4678 #ifdef CONFIG_SYSCTL
4679         err = __addrconf_sysctl_register(net, "all", NULL, all);
4680         if (err < 0)
4681                 goto err_reg_all;
4682
4683         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4684         if (err < 0)
4685                 goto err_reg_dflt;
4686 #endif
4687         return 0;
4688
4689 #ifdef CONFIG_SYSCTL
4690 err_reg_dflt:
4691         __addrconf_sysctl_unregister(all);
4692 err_reg_all:
4693         kfree(dflt);
4694 #endif
4695 err_alloc_dflt:
4696         kfree(all);
4697 err_alloc_all:
4698         return err;
4699 }
4700
4701 static void __net_exit addrconf_exit_net(struct net *net)
4702 {
4703 #ifdef CONFIG_SYSCTL
4704         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4705         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4706 #endif
4707         if (!net_eq(net, &init_net)) {
4708                 kfree(net->ipv6.devconf_dflt);
4709                 kfree(net->ipv6.devconf_all);
4710         }
4711 }
4712
4713 static struct pernet_operations addrconf_ops = {
4714         .init = addrconf_init_net,
4715         .exit = addrconf_exit_net,
4716 };
4717
4718 /*
4719  *      Device notifier
4720  */
4721
4722 int register_inet6addr_notifier(struct notifier_block *nb)
4723 {
4724         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4725 }
4726 EXPORT_SYMBOL(register_inet6addr_notifier);
4727
4728 int unregister_inet6addr_notifier(struct notifier_block *nb)
4729 {
4730         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4731 }
4732 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4733
4734 static struct rtnl_af_ops inet6_ops = {
4735         .family           = AF_INET6,
4736         .fill_link_af     = inet6_fill_link_af,
4737         .get_link_af_size = inet6_get_link_af_size,
4738 };
4739
4740 /*
4741  *      Init / cleanup code
4742  */
4743
4744 int __init addrconf_init(void)
4745 {
4746         int i, err;
4747
4748         err = ipv6_addr_label_init();
4749         if (err < 0) {
4750                 pr_crit("%s: cannot initialize default policy table: %d\n",
4751                         __func__, err);
4752                 goto out;
4753         }
4754
4755         err = register_pernet_subsys(&addrconf_ops);
4756         if (err < 0)
4757                 goto out_addrlabel;
4758
4759         /* The addrconf netdev notifier requires that loopback_dev
4760          * has it's ipv6 private information allocated and setup
4761          * before it can bring up and give link-local addresses
4762          * to other devices which are up.
4763          *
4764          * Unfortunately, loopback_dev is not necessarily the first
4765          * entry in the global dev_base list of net devices.  In fact,
4766          * it is likely to be the very last entry on that list.
4767          * So this causes the notifier registry below to try and
4768          * give link-local addresses to all devices besides loopback_dev
4769          * first, then loopback_dev, which cases all the non-loopback_dev
4770          * devices to fail to get a link-local address.
4771          *
4772          * So, as a temporary fix, allocate the ipv6 structure for
4773          * loopback_dev first by hand.
4774          * Longer term, all of the dependencies ipv6 has upon the loopback
4775          * device and it being up should be removed.
4776          */
4777         rtnl_lock();
4778         if (!ipv6_add_dev(init_net.loopback_dev))
4779                 err = -ENOMEM;
4780         rtnl_unlock();
4781         if (err)
4782                 goto errlo;
4783
4784         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4785                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4786
4787         register_netdevice_notifier(&ipv6_dev_notf);
4788
4789         addrconf_verify(0);
4790
4791         err = rtnl_af_register(&inet6_ops);
4792         if (err < 0)
4793                 goto errout_af;
4794
4795         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4796                               NULL);
4797         if (err < 0)
4798                 goto errout;
4799
4800         /* Only the first call to __rtnl_register can fail */
4801         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4802         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4803         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4804                         inet6_dump_ifaddr, NULL);
4805         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4806                         inet6_dump_ifmcaddr, NULL);
4807         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4808                         inet6_dump_ifacaddr, NULL);
4809
4810         ipv6_addr_label_rtnl_register();
4811
4812         return 0;
4813 errout:
4814         rtnl_af_unregister(&inet6_ops);
4815 errout_af:
4816         unregister_netdevice_notifier(&ipv6_dev_notf);
4817 errlo:
4818         unregister_pernet_subsys(&addrconf_ops);
4819 out_addrlabel:
4820         ipv6_addr_label_cleanup();
4821 out:
4822         return err;
4823 }
4824
4825 void addrconf_cleanup(void)
4826 {
4827         struct net_device *dev;
4828         int i;
4829
4830         unregister_netdevice_notifier(&ipv6_dev_notf);
4831         unregister_pernet_subsys(&addrconf_ops);
4832         ipv6_addr_label_cleanup();
4833
4834         rtnl_lock();
4835
4836         __rtnl_af_unregister(&inet6_ops);
4837
4838         /* clean dev list */
4839         for_each_netdev(&init_net, dev) {
4840                 if (__in6_dev_get(dev) == NULL)
4841                         continue;
4842                 addrconf_ifdown(dev, 1);
4843         }
4844         addrconf_ifdown(init_net.loopback_dev, 2);
4845
4846         /*
4847          *      Check hash table.
4848          */
4849         spin_lock_bh(&addrconf_hash_lock);
4850         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4851                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4852         spin_unlock_bh(&addrconf_hash_lock);
4853
4854         del_timer(&addr_chk_timer);
4855         rtnl_unlock();
4856 }