SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_SPINLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static void addrconf_type_change(struct net_device *dev,
 141                                  unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void __percpu **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void __percpu **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void __percpu **)idev->stats.icmpv6);
 305         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         INIT_LIST_HEAD(&ndev->tempaddr_list);
 405         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 406         if ((dev->flags&IFF_LOOPBACK) ||
 407             dev->type == ARPHRD_TUNNEL ||
 408             dev->type == ARPHRD_TUNNEL6 ||
 409             dev->type == ARPHRD_SIT ||
 410             dev->type == ARPHRD_NONE) {
 411                 printk(KERN_INFO
 412                        "%s: Disabled Privacy Extensions\n",
 413                        dev->name);
 414                 ndev->cnf.use_tempaddr = -1;
 415         } else {
 416                 in6_dev_hold(ndev);
 417                 ipv6_regen_rndid((unsigned long) ndev);
 418         }
 419 #endif
 420
 421         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 422                 ndev->if_flags |= IF_READY;
 423
 424         ipv6_mc_init_dev(ndev);
 425         ndev->tstamp = jiffies;
 426         addrconf_sysctl_register(ndev);
 427         /* protected by rtnl_lock */
 428         rcu_assign_pointer(dev->ip6_ptr, ndev);
 429
 430         /* Join all-node multicast group */
 431         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 432
 433         return ndev;
 434 }
 435
 436 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 437 {
 438         struct inet6_dev *idev;
 439
 440         ASSERT_RTNL();
 441
 442         if ((idev = __in6_dev_get(dev)) == NULL) {
 443                 if ((idev = ipv6_add_dev(dev)) == NULL)
 444                         return NULL;
 445         }
 446
 447         if (dev->flags&IFF_UP)
 448                 ipv6_mc_up(idev);
 449         return idev;
 450 }
 451
 452 #ifdef CONFIG_SYSCTL
 453 static void dev_forward_change(struct inet6_dev *idev)
 454 {
 455         struct net_device *dev;
 456         struct inet6_ifaddr *ifa;
 457
 458         if (!idev)
 459                 return;
 460         dev = idev->dev;
 461         if (idev->cnf.forwarding)
 462                 dev_disable_lro(dev);
 463         if (dev && (dev->flags & IFF_MULTICAST)) {
 464                 if (idev->cnf.forwarding)
 465                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 466                 else
 467                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 468         }
 469         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 470                 if (ifa->flags&IFA_F_TENTATIVE)
 471                         continue;
 472                 if (idev->cnf.forwarding)
 473                         addrconf_join_anycast(ifa);
 474                 else
 475                         addrconf_leave_anycast(ifa);
 476         }
 477 }
 478
 479
 480 static void addrconf_forward_change(struct net *net, __s32 newf)
 481 {
 482         struct net_device *dev;
 483         struct inet6_dev *idev;
 484
 485         rcu_read_lock();
 486         for_each_netdev_rcu(net, dev) {
 487                 idev = __in6_dev_get(dev);
 488                 if (idev) {
 489                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 490                         idev->cnf.forwarding = newf;
 491                         if (changed)
 492                                 dev_forward_change(idev);
 493                 }
 494         }
 495         rcu_read_unlock();
 496 }
 497
 498 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 499 {
 500         struct net *net;
 501
 502         net = (struct net *)table->extra2;
 503         if (p == &net->ipv6.devconf_dflt->forwarding)
 504                 return 0;
 505
 506         if (!rtnl_trylock()) {
 507                 /* Restore the original values before restarting */
 508                 *p = old;
 509                 return restart_syscall();
 510         }
 511
 512         if (p == &net->ipv6.devconf_all->forwarding) {
 513                 __s32 newf = net->ipv6.devconf_all->forwarding;
 514                 net->ipv6.devconf_dflt->forwarding = newf;
 515                 addrconf_forward_change(net, newf);
 516         } else if ((!*p) ^ (!old))
 517                 dev_forward_change((struct inet6_dev *)table->extra1);
 518         rtnl_unlock();
 519
 520         if (*p)
 521                 rt6_purge_dflt_routers(net);
 522         return 1;
 523 }
 524 #endif
 525
 526 static void inet6_ifa_finish_destroy_rcu(struct rcu_head *head)
 527 {
 528         struct inet6_ifaddr *ifp = container_of(head, struct inet6_ifaddr, rcu);
 529         kfree(ifp);
 530 }
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(ifp->if_next != NULL);
 536         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 537
 538 #ifdef NET_REFCNT_DEBUG
 539         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 540 #endif
 541
 542         in6_dev_put(ifp->idev);
 543
 544         if (del_timer(&ifp->timer))
 545                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 546
 547         if (!ifp->dead) {
 548                 printk("Freeing alive inet6 address %p\n", ifp);
 549                 return;
 550         }
 551         dst_release(&ifp->rt->u.dst);
 552
 553         call_rcu(&ifp->rcu, inet6_ifa_finish_destroy_rcu);
 554 }
 555
 556 static void
 557 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 558 {
 559         struct inet6_ifaddr *ifa, **ifap;
 560         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 561
 562         /*
 563          * Each device address list is sorted in order of scope -
 564          * global before linklocal.
 565          */
 566         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 567              ifap = &ifa->if_next) {
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         ifp->if_next = *ifap;
 573         *ifap = ifp;
 574 }
 575
 576 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 577 {
 578         /*
 579          * We perform the hash function over the last 64 bits of the address
 580          * This will include the IEEE address token on links that support it.
 581          */
 582         return jhash_2words(addr->s6_addr32[2],  addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         init_timer(&ifa->timer);
 642         INIT_HLIST_NODE(&ifa->addr_lst);
 643         ifa->timer.data = (unsigned long) ifa;
 644         ifa->scope = scope;
 645         ifa->prefix_len = pfxlen;
 646         ifa->flags = flags | IFA_F_TENTATIVE;
 647         ifa->cstamp = ifa->tstamp = jiffies;
 648
 649         ifa->rt = rt;
 650
 651         /*
 652          * part one of RFC 4429, section 3.3
 653          * We should not configure an address as
 654          * optimistic if we do not yet know the link
 655          * layer address of our nexhop router
 656          */
 657
 658         if (rt->rt6i_nexthop == NULL)
 659                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 660
 661         ifa->idev = idev;
 662         in6_dev_hold(idev);
 663         /* For caller */
 664         in6_ifa_hold(ifa);
 665
 666         /* Add to big hash table */
 667         hash = ipv6_addr_hash(addr);
 668
 669         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 670         in6_ifa_hold(ifa);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, **ifap;
 707         struct inet6_dev *idev = ifp->idev;
 708         int hash;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         hash = ipv6_addr_hash(&ifp->addr);
 713
 714         ifp->dead = 1;
 715
 716         spin_lock_bh(&addrconf_hash_lock);
 717         hlist_del_init_rcu(&ifp->addr_lst);
 718         __in6_ifa_put(ifp);
 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         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 734                 if (ifa == ifp) {
 735                         *ifap = ifa->if_next;
 736                         __in6_ifa_put(ifp);
 737                         ifa->if_next = NULL;
 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                 ifap = &ifa->if_next;
 771         }
 772         write_unlock_bh(&idev->lock);
 773
 774         addrconf_del_timer(ifp);
 775
 776         ipv6_ifa_notify(RTM_DELADDR, ifp);
 777
 778         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 779
 780         /*
 781          * Purge or update corresponding prefix
 782          *
 783          * 1) we don't purge prefix here if address was not permanent.
 784          *    prefix is managed by its own lifetime.
 785          * 2) if there're no addresses, delete prefix.
 786          * 3) if there're still other permanent address(es),
 787          *    corresponding prefix is still permanent.
 788          * 4) otherwise, update prefix lifetime to the
 789          *    longest valid lifetime among the corresponding
 790          *    addresses on the device.
 791          *    Note: subsequent RA will update lifetime.
 792          *
 793          * --yoshfuji
 794          */
 795         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 796                 struct in6_addr prefix;
 797                 struct rt6_info *rt;
 798                 struct net *net = dev_net(ifp->idev->dev);
 799                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 800                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 801
 802                 if (rt && addrconf_is_prefix_route(rt)) {
 803                         if (onlink == 0) {
 804                                 ip6_del_rt(rt);
 805                                 rt = NULL;
 806                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 807                                 rt->rt6i_expires = expires;
 808                                 rt->rt6i_flags |= RTF_EXPIRES;
 809                         }
 810                 }
 811                 dst_release(&rt->u.dst);
 812         }
 813
 814         in6_ifa_put(ifp);
 815 }
 816
 817 #ifdef CONFIG_IPV6_PRIVACY
 818 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 819 {
 820         struct inet6_dev *idev = ifp->idev;
 821         struct in6_addr addr, *tmpaddr;
 822         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 823         unsigned long regen_advance;
 824         int tmp_plen;
 825         int ret = 0;
 826         int max_addresses;
 827         u32 addr_flags;
 828
 829         write_lock(&idev->lock);
 830         if (ift) {
 831                 spin_lock_bh(&ift->lock);
 832                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 833                 spin_unlock_bh(&ift->lock);
 834                 tmpaddr = &addr;
 835         } else {
 836                 tmpaddr = NULL;
 837         }
 838 retry:
 839         in6_dev_hold(idev);
 840         if (idev->cnf.use_tempaddr <= 0) {
 841                 write_unlock(&idev->lock);
 842                 printk(KERN_INFO
 843                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 844                 in6_dev_put(idev);
 845                 ret = -1;
 846                 goto out;
 847         }
 848         spin_lock_bh(&ifp->lock);
 849         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 850                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 851                 spin_unlock_bh(&ifp->lock);
 852                 write_unlock(&idev->lock);
 853                 printk(KERN_WARNING
 854                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 855                 in6_dev_put(idev);
 856                 ret = -1;
 857                 goto out;
 858         }
 859         in6_ifa_hold(ifp);
 860         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 861         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 862                 spin_unlock_bh(&ifp->lock);
 863                 write_unlock(&idev->lock);
 864                 printk(KERN_WARNING
 865                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 866                 in6_ifa_put(ifp);
 867                 in6_dev_put(idev);
 868                 ret = -1;
 869                 goto out;
 870         }
 871         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 872         tmp_valid_lft = min_t(__u32,
 873                               ifp->valid_lft,
 874                               idev->cnf.temp_valid_lft);
 875         tmp_prefered_lft = min_t(__u32,
 876                                  ifp->prefered_lft,
 877                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 878         tmp_plen = ifp->prefix_len;
 879         max_addresses = idev->cnf.max_addresses;
 880         tmp_cstamp = ifp->cstamp;
 881         tmp_tstamp = ifp->tstamp;
 882         spin_unlock_bh(&ifp->lock);
 883
 884         regen_advance = idev->cnf.regen_max_retry *
 885                         idev->cnf.dad_transmits *
 886                         idev->nd_parms->retrans_time / HZ;
 887         write_unlock(&idev->lock);
 888
 889         /* A temporary address is created only if this calculated Preferred
 890          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 891          * an implementation must not create a temporary address with a zero
 892          * Preferred Lifetime.
 893          */
 894         if (tmp_prefered_lft <= regen_advance) {
 895                 in6_ifa_put(ifp);
 896                 in6_dev_put(idev);
 897                 ret = -1;
 898                 goto out;
 899         }
 900
 901         addr_flags = IFA_F_TEMPORARY;
 902         /* set in addrconf_prefix_rcv() */
 903         if (ifp->flags & IFA_F_OPTIMISTIC)
 904                 addr_flags |= IFA_F_OPTIMISTIC;
 905
 906         ift = !max_addresses ||
 907               ipv6_count_addresses(idev) < max_addresses ?
 908                 ipv6_add_addr(idev, &addr, tmp_plen,
 909                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 910                               addr_flags) : NULL;
 911         if (!ift || IS_ERR(ift)) {
 912                 in6_ifa_put(ifp);
 913                 in6_dev_put(idev);
 914                 printk(KERN_INFO
 915                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 916                 tmpaddr = &addr;
 917                 write_lock(&idev->lock);
 918                 goto retry;
 919         }
 920
 921         spin_lock_bh(&ift->lock);
 922         ift->ifpub = ifp;
 923         ift->valid_lft = tmp_valid_lft;
 924         ift->prefered_lft = tmp_prefered_lft;
 925         ift->cstamp = tmp_cstamp;
 926         ift->tstamp = tmp_tstamp;
 927         spin_unlock_bh(&ift->lock);
 928
 929         addrconf_dad_start(ift, 0);
 930         in6_ifa_put(ift);
 931         in6_dev_put(idev);
 932 out:
 933         return ret;
 934 }
 935 #endif
 936
 937 /*
 938  *      Choose an appropriate source address (RFC3484)
 939  */
 940 enum {
 941         IPV6_SADDR_RULE_INIT = 0,
 942         IPV6_SADDR_RULE_LOCAL,
 943         IPV6_SADDR_RULE_SCOPE,
 944         IPV6_SADDR_RULE_PREFERRED,
 945 #ifdef CONFIG_IPV6_MIP6
 946         IPV6_SADDR_RULE_HOA,
 947 #endif
 948         IPV6_SADDR_RULE_OIF,
 949         IPV6_SADDR_RULE_LABEL,
 950 #ifdef CONFIG_IPV6_PRIVACY
 951         IPV6_SADDR_RULE_PRIVACY,
 952 #endif
 953         IPV6_SADDR_RULE_ORCHID,
 954         IPV6_SADDR_RULE_PREFIX,
 955         IPV6_SADDR_RULE_MAX
 956 };
 957
 958 struct ipv6_saddr_score {
 959         int                     rule;
 960         int                     addr_type;
 961         struct inet6_ifaddr     *ifa;
 962         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 963         int                     scopedist;
 964         int                     matchlen;
 965 };
 966
 967 struct ipv6_saddr_dst {
 968         const struct in6_addr *addr;
 969         int ifindex;
 970         int scope;
 971         int label;
 972         unsigned int prefs;
 973 };
 974
 975 static inline int ipv6_saddr_preferred(int type)
 976 {
 977         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 978                 return 1;
 979         return 0;
 980 }
 981
 982 static int ipv6_get_saddr_eval(struct net *net,
 983                                struct ipv6_saddr_score *score,
 984                                struct ipv6_saddr_dst *dst,
 985                                int i)
 986 {
 987         int ret;
 988
 989         if (i <= score->rule) {
 990                 switch (i) {
 991                 case IPV6_SADDR_RULE_SCOPE:
 992                         ret = score->scopedist;
 993                         break;
 994                 case IPV6_SADDR_RULE_PREFIX:
 995                         ret = score->matchlen;
 996                         break;
 997                 default:
 998                         ret = !!test_bit(i, score->scorebits);
 999                 }
1000                 goto out;
1001         }
1002
1003         switch (i) {
1004         case IPV6_SADDR_RULE_INIT:
1005                 /* Rule 0: remember if hiscore is not ready yet */
1006                 ret = !!score->ifa;
1007                 break;
1008         case IPV6_SADDR_RULE_LOCAL:
1009                 /* Rule 1: Prefer same address */
1010                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1011                 break;
1012         case IPV6_SADDR_RULE_SCOPE:
1013                 /* Rule 2: Prefer appropriate scope
1014                  *
1015                  *      ret
1016                  *       ^
1017                  *    -1 |  d 15
1018                  *    ---+--+-+---> scope
1019                  *       |
1020                  *       |             d is scope of the destination.
1021                  *  B-d  |  \
1022                  *       |   \      <- smaller scope is better if
1023                  *  B-15 |    \        if scope is enough for destinaion.
1024                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1025                  * d-C-1 | /
1026                  *       |/         <- greater is better
1027                  *   -C  /             if scope is not enough for destination.
1028                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1029                  *
1030                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1031                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1032                  * Assume B = 0 and we get C > 29.
1033                  */
1034                 ret = __ipv6_addr_src_scope(score->addr_type);
1035                 if (ret >= dst->scope)
1036                         ret = -ret;
1037                 else
1038                         ret -= 128;     /* 30 is enough */
1039                 score->scopedist = ret;
1040                 break;
1041         case IPV6_SADDR_RULE_PREFERRED:
1042                 /* Rule 3: Avoid deprecated and optimistic addresses */
1043                 ret = ipv6_saddr_preferred(score->addr_type) ||
1044                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1045                 break;
1046 #ifdef CONFIG_IPV6_MIP6
1047         case IPV6_SADDR_RULE_HOA:
1048             {
1049                 /* Rule 4: Prefer home address */
1050                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1051                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1052                 break;
1053             }
1054 #endif
1055         case IPV6_SADDR_RULE_OIF:
1056                 /* Rule 5: Prefer outgoing interface */
1057                 ret = (!dst->ifindex ||
1058                        dst->ifindex == score->ifa->idev->dev->ifindex);
1059                 break;
1060         case IPV6_SADDR_RULE_LABEL:
1061                 /* Rule 6: Prefer matching label */
1062                 ret = ipv6_addr_label(net,
1063                                       &score->ifa->addr, score->addr_type,
1064                                       score->ifa->idev->dev->ifindex) == dst->label;
1065                 break;
1066 #ifdef CONFIG_IPV6_PRIVACY
1067         case IPV6_SADDR_RULE_PRIVACY:
1068             {
1069                 /* Rule 7: Prefer public address
1070                  * Note: prefer temprary address if use_tempaddr >= 2
1071                  */
1072                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1073                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1074                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1075                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1076                 break;
1077             }
1078 #endif
1079         case IPV6_SADDR_RULE_ORCHID:
1080                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1081                  *          non-ORCHID vs non-ORCHID
1082                  */
1083                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1084                         ipv6_addr_orchid(dst->addr));
1085                 break;
1086         case IPV6_SADDR_RULE_PREFIX:
1087                 /* Rule 8: Use longest matching prefix */
1088                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1089                                                        dst->addr);
1090                 break;
1091         default:
1092                 ret = 0;
1093         }
1094
1095         if (ret)
1096                 __set_bit(i, score->scorebits);
1097         score->rule = i;
1098 out:
1099         return ret;
1100 }
1101
1102 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1103                        const struct in6_addr *daddr, unsigned int prefs,
1104                        struct in6_addr *saddr)
1105 {
1106         struct ipv6_saddr_score scores[2],
1107                                 *score = &scores[0], *hiscore = &scores[1];
1108         struct ipv6_saddr_dst dst;
1109         struct net_device *dev;
1110         int dst_type;
1111
1112         dst_type = __ipv6_addr_type(daddr);
1113         dst.addr = daddr;
1114         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1115         dst.scope = __ipv6_addr_src_scope(dst_type);
1116         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1117         dst.prefs = prefs;
1118
1119         hiscore->rule = -1;
1120         hiscore->ifa = NULL;
1121
1122         rcu_read_lock();
1123
1124         for_each_netdev_rcu(net, dev) {
1125                 struct inet6_dev *idev;
1126
1127                 /* Candidate Source Address (section 4)
1128                  *  - multicast and link-local destination address,
1129                  *    the set of candidate source address MUST only
1130                  *    include addresses assigned to interfaces
1131                  *    belonging to the same link as the outgoing
1132                  *    interface.
1133                  * (- For site-local destination addresses, the
1134                  *    set of candidate source addresses MUST only
1135                  *    include addresses assigned to interfaces
1136                  *    belonging to the same site as the outgoing
1137                  *    interface.)
1138                  */
1139                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1140                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1141                     dst.ifindex && dev->ifindex != dst.ifindex)
1142                         continue;
1143
1144                 idev = __in6_dev_get(dev);
1145                 if (!idev)
1146                         continue;
1147
1148                 read_lock_bh(&idev->lock);
1149                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1150                         int i;
1151
1152                         /*
1153                          * - Tentative Address (RFC2462 section 5.4)
1154                          *  - A tentative address is not considered
1155                          *    "assigned to an interface" in the traditional
1156                          *    sense, unless it is also flagged as optimistic.
1157                          * - Candidate Source Address (section 4)
1158                          *  - In any case, anycast addresses, multicast
1159                          *    addresses, and the unspecified address MUST
1160                          *    NOT be included in a candidate set.
1161                          */
1162                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1163                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1164                                 continue;
1165
1166                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1167
1168                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1169                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1170                                 LIMIT_NETDEBUG(KERN_DEBUG
1171                                                "ADDRCONF: unspecified / multicast address "
1172                                                "assigned as unicast address on %s",
1173                                                dev->name);
1174                                 continue;
1175                         }
1176
1177                         score->rule = -1;
1178                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1179
1180                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1181                                 int minihiscore, miniscore;
1182
1183                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1184                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1185
1186                                 if (minihiscore > miniscore) {
1187                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1188                                             score->scopedist > 0) {
1189                                                 /*
1190                                                  * special case:
1191                                                  * each remaining entry
1192                                                  * has too small (not enough)
1193                                                  * scope, because ifa entries
1194                                                  * are sorted by their scope
1195                                                  * values.
1196                                                  */
1197                                                 goto try_nextdev;
1198                                         }
1199                                         break;
1200                                 } else if (minihiscore < miniscore) {
1201                                         if (hiscore->ifa)
1202                                                 in6_ifa_put(hiscore->ifa);
1203
1204                                         in6_ifa_hold(score->ifa);
1205
1206                                         swap(hiscore, score);
1207
1208                                         /* restore our iterator */
1209                                         score->ifa = hiscore->ifa;
1210
1211                                         break;
1212                                 }
1213                         }
1214                 }
1215 try_nextdev:
1216                 read_unlock_bh(&idev->lock);
1217         }
1218         rcu_read_unlock();
1219
1220         if (!hiscore->ifa)
1221                 return -EADDRNOTAVAIL;
1222
1223         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1224         in6_ifa_put(hiscore->ifa);
1225         return 0;
1226 }
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         if ((idev = __in6_dev_get(dev)) != NULL) {
1238                 struct inet6_ifaddr *ifp;
1239
1240                 read_lock_bh(&idev->lock);
1241                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1242                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1243                                 ipv6_addr_copy(addr, &ifp->addr);
1244                                 err = 0;
1245                                 break;
1246                         }
1247                 }
1248                 read_unlock_bh(&idev->lock);
1249         }
1250         rcu_read_unlock();
1251         return err;
1252 }
1253
1254 static int ipv6_count_addresses(struct inet6_dev *idev)
1255 {
1256         int cnt = 0;
1257         struct inet6_ifaddr *ifp;
1258
1259         read_lock_bh(&idev->lock);
1260         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1261                 cnt++;
1262         read_unlock_bh(&idev->lock);
1263         return cnt;
1264 }
1265
1266 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1267                   struct net_device *dev, int strict)
1268 {
1269         struct inet6_ifaddr *ifp = NULL;
1270         struct hlist_node *node;
1271         unsigned int hash = ipv6_addr_hash(addr);
1272
1273         rcu_read_lock_bh();
1274         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1275                 if (!net_eq(dev_net(ifp->idev->dev), net))
1276                         continue;
1277                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1278                     !(ifp->flags&IFA_F_TENTATIVE)) {
1279                         if (dev == NULL || ifp->idev->dev == dev ||
1280                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1281                                 break;
1282                 }
1283         }
1284         rcu_read_unlock_bh();
1285
1286         return ifp != NULL;
1287 }
1288 EXPORT_SYMBOL(ipv6_chk_addr);
1289
1290 static
1291 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1292                        struct net_device *dev)
1293 {
1294         struct inet6_ifaddr *ifp;
1295         struct hlist_node *node;
1296         unsigned int hash = ipv6_addr_hash(addr);
1297
1298         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1299                 if (!net_eq(dev_net(ifp->idev->dev), net))
1300                         continue;
1301                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1302                         if (dev == NULL || ifp->idev->dev == dev)
1303                                 break;
1304                 }
1305         }
1306         return ifp != NULL;
1307 }
1308
1309 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1310 {
1311         struct inet6_dev *idev;
1312         struct inet6_ifaddr *ifa;
1313         int     onlink;
1314
1315         onlink = 0;
1316         rcu_read_lock();
1317         idev = __in6_dev_get(dev);
1318         if (idev) {
1319                 read_lock_bh(&idev->lock);
1320                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1321                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1322                                                    ifa->prefix_len);
1323                         if (onlink)
1324                                 break;
1325                 }
1326                 read_unlock_bh(&idev->lock);
1327         }
1328         rcu_read_unlock();
1329         return onlink;
1330 }
1331
1332 EXPORT_SYMBOL(ipv6_chk_prefix);
1333
1334 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1335                                      struct net_device *dev, int strict)
1336 {
1337         struct inet6_ifaddr *ifp = NULL;
1338         struct hlist_node *node;
1339         unsigned int hash = ipv6_addr_hash(addr);
1340
1341         rcu_read_lock_bh();
1342         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1343                 if (!net_eq(dev_net(ifp->idev->dev), net))
1344                         continue;
1345                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1346                         if (dev == NULL || ifp->idev->dev == dev ||
1347                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1348                                 in6_ifa_hold(ifp);
1349                                 break;
1350                         }
1351                 }
1352         }
1353         rcu_read_unlock_bh();
1354
1355         return ifp;
1356 }
1357
1358 /* Gets referenced address, destroys ifaddr */
1359
1360 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1361 {
1362         if (ifp->flags&IFA_F_PERMANENT) {
1363                 spin_lock_bh(&ifp->lock);
1364                 addrconf_del_timer(ifp);
1365                 ifp->flags |= IFA_F_TENTATIVE;
1366                 if (dad_failed)
1367                         ifp->flags |= IFA_F_DADFAILED;
1368                 spin_unlock_bh(&ifp->lock);
1369                 if (dad_failed)
1370                         ipv6_ifa_notify(0, ifp);
1371                 in6_ifa_put(ifp);
1372 #ifdef CONFIG_IPV6_PRIVACY
1373         } else if (ifp->flags&IFA_F_TEMPORARY) {
1374                 struct inet6_ifaddr *ifpub;
1375                 spin_lock_bh(&ifp->lock);
1376                 ifpub = ifp->ifpub;
1377                 if (ifpub) {
1378                         in6_ifa_hold(ifpub);
1379                         spin_unlock_bh(&ifp->lock);
1380                         ipv6_create_tempaddr(ifpub, ifp);
1381                         in6_ifa_put(ifpub);
1382                 } else {
1383                         spin_unlock_bh(&ifp->lock);
1384                 }
1385                 ipv6_del_addr(ifp);
1386 #endif
1387         } else
1388                 ipv6_del_addr(ifp);
1389 }
1390
1391 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1392 {
1393         struct inet6_dev *idev = ifp->idev;
1394
1395         if (net_ratelimit())
1396                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1397                         ifp->idev->dev->name, &ifp->addr);
1398
1399         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1400                 struct in6_addr addr;
1401
1402                 addr.s6_addr32[0] = htonl(0xfe800000);
1403                 addr.s6_addr32[1] = 0;
1404
1405                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1406                     ipv6_addr_equal(&ifp->addr, &addr)) {
1407                         /* DAD failed for link-local based on MAC address */
1408                         idev->cnf.disable_ipv6 = 1;
1409
1410                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1411                                 ifp->idev->dev->name);
1412                 }
1413         }
1414
1415         addrconf_dad_stop(ifp, 1);
1416 }
1417
1418 /* Join to solicited addr multicast group. */
1419
1420 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1421 {
1422         struct in6_addr maddr;
1423
1424         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1425                 return;
1426
1427         addrconf_addr_solict_mult(addr, &maddr);
1428         ipv6_dev_mc_inc(dev, &maddr);
1429 }
1430
1431 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1432 {
1433         struct in6_addr maddr;
1434
1435         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1436                 return;
1437
1438         addrconf_addr_solict_mult(addr, &maddr);
1439         __ipv6_dev_mc_dec(idev, &maddr);
1440 }
1441
1442 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1443 {
1444         struct in6_addr addr;
1445         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1446         if (ipv6_addr_any(&addr))
1447                 return;
1448         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1449 }
1450
1451 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1452 {
1453         struct in6_addr addr;
1454         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1455         if (ipv6_addr_any(&addr))
1456                 return;
1457         __ipv6_dev_ac_dec(ifp->idev, &addr);
1458 }
1459
1460 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1461 {
1462         if (dev->addr_len != ETH_ALEN)
1463                 return -1;
1464         memcpy(eui, dev->dev_addr, 3);
1465         memcpy(eui + 5, dev->dev_addr + 3, 3);
1466
1467         /*
1468          * The zSeries OSA network cards can be shared among various
1469          * OS instances, but the OSA cards have only one MAC address.
1470          * This leads to duplicate address conflicts in conjunction
1471          * with IPv6 if more than one instance uses the same card.
1472          *
1473          * The driver for these cards can deliver a unique 16-bit
1474          * identifier for each instance sharing the same card.  It is
1475          * placed instead of 0xFFFE in the interface identifier.  The
1476          * "u" bit of the interface identifier is not inverted in this
1477          * case.  Hence the resulting interface identifier has local
1478          * scope according to RFC2373.
1479          */
1480         if (dev->dev_id) {
1481                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1482                 eui[4] = dev->dev_id & 0xFF;
1483         } else {
1484                 eui[3] = 0xFF;
1485                 eui[4] = 0xFE;
1486                 eui[0] ^= 2;
1487         }
1488         return 0;
1489 }
1490
1491 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1492 {
1493         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1494         if (dev->addr_len != ARCNET_ALEN)
1495                 return -1;
1496         memset(eui, 0, 7);
1497         eui[7] = *(u8*)dev->dev_addr;
1498         return 0;
1499 }
1500
1501 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1502 {
1503         if (dev->addr_len != INFINIBAND_ALEN)
1504                 return -1;
1505         memcpy(eui, dev->dev_addr + 12, 8);
1506         eui[0] |= 2;
1507         return 0;
1508 }
1509
1510 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1511 {
1512         if (addr == 0)
1513                 return -1;
1514         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1515                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1516                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1517                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1518                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1519                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1520         eui[1] = 0;
1521         eui[2] = 0x5E;
1522         eui[3] = 0xFE;
1523         memcpy(eui + 4, &addr, 4);
1524         return 0;
1525 }
1526 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1527
1528 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1529 {
1530         if (dev->priv_flags & IFF_ISATAP)
1531                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1532         return -1;
1533 }
1534
1535 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1536 {
1537         switch (dev->type) {
1538         case ARPHRD_ETHER:
1539         case ARPHRD_FDDI:
1540         case ARPHRD_IEEE802_TR:
1541                 return addrconf_ifid_eui48(eui, dev);
1542         case ARPHRD_ARCNET:
1543                 return addrconf_ifid_arcnet(eui, dev);
1544         case ARPHRD_INFINIBAND:
1545                 return addrconf_ifid_infiniband(eui, dev);
1546         case ARPHRD_SIT:
1547                 return addrconf_ifid_sit(eui, dev);
1548         }
1549         return -1;
1550 }
1551
1552 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1553 {
1554         int err = -1;
1555         struct inet6_ifaddr *ifp;
1556
1557         read_lock_bh(&idev->lock);
1558         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1559                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1560                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1561                         err = 0;
1562                         break;
1563                 }
1564         }
1565         read_unlock_bh(&idev->lock);
1566         return err;
1567 }
1568
1569 #ifdef CONFIG_IPV6_PRIVACY
1570 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1571 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1572 {
1573 regen:
1574         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1575         idev->rndid[0] &= ~0x02;
1576
1577         /*
1578          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1579          * check if generated address is not inappropriate
1580          *
1581          *  - Reserved subnet anycast (RFC 2526)
1582          *      11111101 11....11 1xxxxxxx
1583          *  - ISATAP (RFC4214) 6.1
1584          *      00-00-5E-FE-xx-xx-xx-xx
1585          *  - value 0
1586          *  - XXX: already assigned to an address on the device
1587          */
1588         if (idev->rndid[0] == 0xfd &&
1589             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1590             (idev->rndid[7]&0x80))
1591                 goto regen;
1592         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1593                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1594                         goto regen;
1595                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1596                         goto regen;
1597         }
1598
1599         return 0;
1600 }
1601
1602 static void ipv6_regen_rndid(unsigned long data)
1603 {
1604         struct inet6_dev *idev = (struct inet6_dev *) data;
1605         unsigned long expires;
1606
1607         rcu_read_lock_bh();
1608         write_lock_bh(&idev->lock);
1609
1610         if (idev->dead)
1611                 goto out;
1612
1613         if (__ipv6_regen_rndid(idev) < 0)
1614                 goto out;
1615
1616         expires = jiffies +
1617                 idev->cnf.temp_prefered_lft * HZ -
1618                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1619         if (time_before(expires, jiffies)) {
1620                 printk(KERN_WARNING
1621                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1622                         idev->dev->name);
1623                 goto out;
1624         }
1625
1626         if (!mod_timer(&idev->regen_timer, expires))
1627                 in6_dev_hold(idev);
1628
1629 out:
1630         write_unlock_bh(&idev->lock);
1631         rcu_read_unlock_bh();
1632         in6_dev_put(idev);
1633 }
1634
1635 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1636         int ret = 0;
1637
1638         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1639                 ret = __ipv6_regen_rndid(idev);
1640         return ret;
1641 }
1642 #endif
1643
1644 /*
1645  *      Add prefix route.
1646  */
1647
1648 static void
1649 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1650                       unsigned long expires, u32 flags)
1651 {
1652         struct fib6_config cfg = {
1653                 .fc_table = RT6_TABLE_PREFIX,
1654                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1655                 .fc_ifindex = dev->ifindex,
1656                 .fc_expires = expires,
1657                 .fc_dst_len = plen,
1658                 .fc_flags = RTF_UP | flags,
1659                 .fc_nlinfo.nl_net = dev_net(dev),
1660                 .fc_protocol = RTPROT_KERNEL,
1661         };
1662
1663         ipv6_addr_copy(&cfg.fc_dst, pfx);
1664
1665         /* Prevent useless cloning on PtP SIT.
1666            This thing is done here expecting that the whole
1667            class of non-broadcast devices need not cloning.
1668          */
1669 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1670         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1671                 cfg.fc_flags |= RTF_NONEXTHOP;
1672 #endif
1673
1674         ip6_route_add(&cfg);
1675 }
1676
1677 /* Create "default" multicast route to the interface */
1678
1679 static void addrconf_add_mroute(struct net_device *dev)
1680 {
1681         struct fib6_config cfg = {
1682                 .fc_table = RT6_TABLE_LOCAL,
1683                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1684                 .fc_ifindex = dev->ifindex,
1685                 .fc_dst_len = 8,
1686                 .fc_flags = RTF_UP,
1687                 .fc_nlinfo.nl_net = dev_net(dev),
1688         };
1689
1690         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1691
1692         ip6_route_add(&cfg);
1693 }
1694
1695 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1696 static void sit_route_add(struct net_device *dev)
1697 {
1698         struct fib6_config cfg = {
1699                 .fc_table = RT6_TABLE_MAIN,
1700                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1701                 .fc_ifindex = dev->ifindex,
1702                 .fc_dst_len = 96,
1703                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1704                 .fc_nlinfo.nl_net = dev_net(dev),
1705         };
1706
1707         /* prefix length - 96 bits "::d.d.d.d" */
1708         ip6_route_add(&cfg);
1709 }
1710 #endif
1711
1712 static void addrconf_add_lroute(struct net_device *dev)
1713 {
1714         struct in6_addr addr;
1715
1716         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1717         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1718 }
1719
1720 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1721 {
1722         struct inet6_dev *idev;
1723
1724         ASSERT_RTNL();
1725
1726         if ((idev = ipv6_find_idev(dev)) == NULL)
1727                 return NULL;
1728
1729         /* Add default multicast route */
1730         addrconf_add_mroute(dev);
1731
1732         /* Add link local route */
1733         addrconf_add_lroute(dev);
1734         return idev;
1735 }
1736
1737 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1738 {
1739         struct prefix_info *pinfo;
1740         __u32 valid_lft;
1741         __u32 prefered_lft;
1742         int addr_type;
1743         struct inet6_dev *in6_dev;
1744         struct net *net = dev_net(dev);
1745
1746         pinfo = (struct prefix_info *) opt;
1747
1748         if (len < sizeof(struct prefix_info)) {
1749                 ADBG(("addrconf: prefix option too short\n"));
1750                 return;
1751         }
1752
1753         /*
1754          *      Validation checks ([ADDRCONF], page 19)
1755          */
1756
1757         addr_type = ipv6_addr_type(&pinfo->prefix);
1758
1759         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1760                 return;
1761
1762         valid_lft = ntohl(pinfo->valid);
1763         prefered_lft = ntohl(pinfo->prefered);
1764
1765         if (prefered_lft > valid_lft) {
1766                 if (net_ratelimit())
1767                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1768                 return;
1769         }
1770
1771         in6_dev = in6_dev_get(dev);
1772
1773         if (in6_dev == NULL) {
1774                 if (net_ratelimit())
1775                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1776                 return;
1777         }
1778
1779         /*
1780          *      Two things going on here:
1781          *      1) Add routes for on-link prefixes
1782          *      2) Configure prefixes with the auto flag set
1783          */
1784
1785         if (pinfo->onlink) {
1786                 struct rt6_info *rt;
1787                 unsigned long rt_expires;
1788
1789                 /* Avoid arithmetic overflow. Really, we could
1790                  * save rt_expires in seconds, likely valid_lft,
1791                  * but it would require division in fib gc, that it
1792                  * not good.
1793                  */
1794                 if (HZ > USER_HZ)
1795                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1796                 else
1797                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1798
1799                 if (addrconf_finite_timeout(rt_expires))
1800                         rt_expires *= HZ;
1801
1802                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1803                                 dev->ifindex, 1);
1804
1805                 if (rt && addrconf_is_prefix_route(rt)) {
1806                         /* Autoconf prefix route */
1807                         if (valid_lft == 0) {
1808                                 ip6_del_rt(rt);
1809                                 rt = NULL;
1810                         } else if (addrconf_finite_timeout(rt_expires)) {
1811                                 /* not infinity */
1812                                 rt->rt6i_expires = jiffies + rt_expires;
1813                                 rt->rt6i_flags |= RTF_EXPIRES;
1814                         } else {
1815                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1816                                 rt->rt6i_expires = 0;
1817                         }
1818                 } else if (valid_lft) {
1819                         clock_t expires = 0;
1820                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1821                         if (addrconf_finite_timeout(rt_expires)) {
1822                                 /* not infinity */
1823                                 flags |= RTF_EXPIRES;
1824                                 expires = jiffies_to_clock_t(rt_expires);
1825                         }
1826                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1827                                               dev, expires, flags);
1828                 }
1829                 if (rt)
1830                         dst_release(&rt->u.dst);
1831         }
1832
1833         /* Try to figure out our local address for this prefix */
1834
1835         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1836                 struct inet6_ifaddr * ifp;
1837                 struct in6_addr addr;
1838                 int create = 0, update_lft = 0;
1839
1840                 if (pinfo->prefix_len == 64) {
1841                         memcpy(&addr, &pinfo->prefix, 8);
1842                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1843                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1844                                 in6_dev_put(in6_dev);
1845                                 return;
1846                         }
1847                         goto ok;
1848                 }
1849                 if (net_ratelimit())
1850                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1851                                pinfo->prefix_len);
1852                 in6_dev_put(in6_dev);
1853                 return;
1854
1855 ok:
1856
1857                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1858
1859                 if (ifp == NULL && valid_lft) {
1860                         int max_addresses = in6_dev->cnf.max_addresses;
1861                         u32 addr_flags = 0;
1862
1863 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1864                         if (in6_dev->cnf.optimistic_dad &&
1865                             !net->ipv6.devconf_all->forwarding)
1866                                 addr_flags = IFA_F_OPTIMISTIC;
1867 #endif
1868
1869                         /* Do not allow to create too much of autoconfigured
1870                          * addresses; this would be too easy way to crash kernel.
1871                          */
1872                         if (!max_addresses ||
1873                             ipv6_count_addresses(in6_dev) < max_addresses)
1874                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1875                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1876                                                     addr_flags);
1877
1878                         if (!ifp || IS_ERR(ifp)) {
1879                                 in6_dev_put(in6_dev);
1880                                 return;
1881                         }
1882
1883                         update_lft = create = 1;
1884                         ifp->cstamp = jiffies;
1885                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1886                 }
1887
1888                 if (ifp) {
1889                         int flags;
1890                         unsigned long now;
1891 #ifdef CONFIG_IPV6_PRIVACY
1892                         struct inet6_ifaddr *ift;
1893 #endif
1894                         u32 stored_lft;
1895
1896                         /* update lifetime (RFC2462 5.5.3 e) */
1897                         spin_lock(&ifp->lock);
1898                         now = jiffies;
1899                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1900                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1901                         else
1902                                 stored_lft = 0;
1903                         if (!update_lft && stored_lft) {
1904                                 if (valid_lft > MIN_VALID_LIFETIME ||
1905                                     valid_lft > stored_lft)
1906                                         update_lft = 1;
1907                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1908                                         /* valid_lft <= stored_lft is always true */
1909                                         /*
1910                                          * RFC 4862 Section 5.5.3e:
1911                                          * "Note that the preferred lifetime of
1912                                          *  the corresponding address is always
1913                                          *  reset to the Preferred Lifetime in
1914                                          *  the received Prefix Information
1915                                          *  option, regardless of whether the
1916                                          *  valid lifetime is also reset or
1917                                          *  ignored."
1918                                          *
1919                                          *  So if the preferred lifetime in
1920                                          *  this advertisement is different
1921                                          *  than what we have stored, but the
1922                                          *  valid lifetime is invalid, just
1923                                          *  reset prefered_lft.
1924                                          *
1925                                          *  We must set the valid lifetime
1926                                          *  to the stored lifetime since we'll
1927                                          *  be updating the timestamp below,
1928                                          *  else we'll set it back to the
1929                                          *  minumum.
1930                                          */
1931                                         if (prefered_lft != ifp->prefered_lft) {
1932                                                 valid_lft = stored_lft;
1933                                                 update_lft = 1;
1934                                         }
1935                                 } else {
1936                                         valid_lft = MIN_VALID_LIFETIME;
1937                                         if (valid_lft < prefered_lft)
1938                                                 prefered_lft = valid_lft;
1939                                         update_lft = 1;
1940                                 }
1941                         }
1942
1943                         if (update_lft) {
1944                                 ifp->valid_lft = valid_lft;
1945                                 ifp->prefered_lft = prefered_lft;
1946                                 ifp->tstamp = now;
1947                                 flags = ifp->flags;
1948                                 ifp->flags &= ~IFA_F_DEPRECATED;
1949                                 spin_unlock(&ifp->lock);
1950
1951                                 if (!(flags&IFA_F_TENTATIVE))
1952                                         ipv6_ifa_notify(0, ifp);
1953                         } else
1954                                 spin_unlock(&ifp->lock);
1955
1956 #ifdef CONFIG_IPV6_PRIVACY
1957                         read_lock_bh(&in6_dev->lock);
1958                         /* update all temporary addresses in the list */
1959                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1960                                 /*
1961                                  * When adjusting the lifetimes of an existing
1962                                  * temporary address, only lower the lifetimes.
1963                                  * Implementations must not increase the
1964                                  * lifetimes of an existing temporary address
1965                                  * when processing a Prefix Information Option.
1966                                  */
1967                                 if (ifp != ift->ifpub)
1968                                         continue;
1969
1970                                 spin_lock(&ift->lock);
1971                                 flags = ift->flags;
1972                                 if (ift->valid_lft > valid_lft &&
1973                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1974                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1975                                 if (ift->prefered_lft > prefered_lft &&
1976                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1977                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1978                                 spin_unlock(&ift->lock);
1979                                 if (!(flags&IFA_F_TENTATIVE))
1980                                         ipv6_ifa_notify(0, ift);
1981                         }
1982
1983                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1984                                 /*
1985                                  * When a new public address is created as described in [ADDRCONF],
1986                                  * also create a new temporary address.
1987                                  */
1988                                 read_unlock_bh(&in6_dev->lock);
1989                                 ipv6_create_tempaddr(ifp, NULL);
1990                         } else {
1991                                 read_unlock_bh(&in6_dev->lock);
1992                         }
1993 #endif
1994                         in6_ifa_put(ifp);
1995                         addrconf_verify(0);
1996                 }
1997         }
1998         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1999         in6_dev_put(in6_dev);
2000 }
2001
2002 /*
2003  *      Set destination address.
2004  *      Special case for SIT interfaces where we create a new "virtual"
2005  *      device.
2006  */
2007 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2008 {
2009         struct in6_ifreq ireq;
2010         struct net_device *dev;
2011         int err = -EINVAL;
2012
2013         rtnl_lock();
2014
2015         err = -EFAULT;
2016         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2017                 goto err_exit;
2018
2019         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2020
2021         err = -ENODEV;
2022         if (dev == NULL)
2023                 goto err_exit;
2024
2025 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2026         if (dev->type == ARPHRD_SIT) {
2027                 const struct net_device_ops *ops = dev->netdev_ops;
2028                 struct ifreq ifr;
2029                 struct ip_tunnel_parm p;
2030
2031                 err = -EADDRNOTAVAIL;
2032                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2033                         goto err_exit;
2034
2035                 memset(&p, 0, sizeof(p));
2036                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2037                 p.iph.saddr = 0;
2038                 p.iph.version = 4;
2039                 p.iph.ihl = 5;
2040                 p.iph.protocol = IPPROTO_IPV6;
2041                 p.iph.ttl = 64;
2042                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2043
2044                 if (ops->ndo_do_ioctl) {
2045                         mm_segment_t oldfs = get_fs();
2046
2047                         set_fs(KERNEL_DS);
2048                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2049                         set_fs(oldfs);
2050                 } else
2051                         err = -EOPNOTSUPP;
2052
2053                 if (err == 0) {
2054                         err = -ENOBUFS;
2055                         dev = __dev_get_by_name(net, p.name);
2056                         if (!dev)
2057                                 goto err_exit;
2058                         err = dev_open(dev);
2059                 }
2060         }
2061 #endif
2062
2063 err_exit:
2064         rtnl_unlock();
2065         return err;
2066 }
2067
2068 /*
2069  *      Manual configuration of address on an interface
2070  */
2071 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2072                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2073                           __u32 valid_lft)
2074 {
2075         struct inet6_ifaddr *ifp;
2076         struct inet6_dev *idev;
2077         struct net_device *dev;
2078         int scope;
2079         u32 flags;
2080         clock_t expires;
2081         unsigned long timeout;
2082
2083         ASSERT_RTNL();
2084
2085         if (plen > 128)
2086                 return -EINVAL;
2087
2088         /* check the lifetime */
2089         if (!valid_lft || prefered_lft > valid_lft)
2090                 return -EINVAL;
2091
2092         dev = __dev_get_by_index(net, ifindex);
2093         if (!dev)
2094                 return -ENODEV;
2095
2096         if ((idev = addrconf_add_dev(dev)) == NULL)
2097                 return -ENOBUFS;
2098
2099         scope = ipv6_addr_scope(pfx);
2100
2101         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2102         if (addrconf_finite_timeout(timeout)) {
2103                 expires = jiffies_to_clock_t(timeout * HZ);
2104                 valid_lft = timeout;
2105                 flags = RTF_EXPIRES;
2106         } else {
2107                 expires = 0;
2108                 flags = 0;
2109                 ifa_flags |= IFA_F_PERMANENT;
2110         }
2111
2112         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2113         if (addrconf_finite_timeout(timeout)) {
2114                 if (timeout == 0)
2115                         ifa_flags |= IFA_F_DEPRECATED;
2116                 prefered_lft = timeout;
2117         }
2118
2119         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2120
2121         if (!IS_ERR(ifp)) {
2122                 spin_lock_bh(&ifp->lock);
2123                 ifp->valid_lft = valid_lft;
2124                 ifp->prefered_lft = prefered_lft;
2125                 ifp->tstamp = jiffies;
2126                 spin_unlock_bh(&ifp->lock);
2127
2128                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2129                                       expires, flags);
2130                 /*
2131                  * Note that section 3.1 of RFC 4429 indicates
2132                  * that the Optimistic flag should not be set for
2133                  * manually configured addresses
2134                  */
2135                 addrconf_dad_start(ifp, 0);
2136                 in6_ifa_put(ifp);
2137                 addrconf_verify(0);
2138                 return 0;
2139         }
2140
2141         return PTR_ERR(ifp);
2142 }
2143
2144 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2145                           unsigned int plen)
2146 {
2147         struct inet6_ifaddr *ifp;
2148         struct inet6_dev *idev;
2149         struct net_device *dev;
2150
2151         if (plen > 128)
2152                 return -EINVAL;
2153
2154         dev = __dev_get_by_index(net, ifindex);
2155         if (!dev)
2156                 return -ENODEV;
2157
2158         if ((idev = __in6_dev_get(dev)) == NULL)
2159                 return -ENXIO;
2160
2161         read_lock_bh(&idev->lock);
2162         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2163                 if (ifp->prefix_len == plen &&
2164                     ipv6_addr_equal(pfx, &ifp->addr)) {
2165                         in6_ifa_hold(ifp);
2166                         read_unlock_bh(&idev->lock);
2167
2168                         ipv6_del_addr(ifp);
2169
2170                         /* If the last address is deleted administratively,
2171                            disable IPv6 on this interface.
2172                          */
2173                         if (idev->addr_list == NULL)
2174                                 addrconf_ifdown(idev->dev, 1);
2175                         return 0;
2176                 }
2177         }
2178         read_unlock_bh(&idev->lock);
2179         return -EADDRNOTAVAIL;
2180 }
2181
2182
2183 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2184 {
2185         struct in6_ifreq ireq;
2186         int err;
2187
2188         if (!capable(CAP_NET_ADMIN))
2189                 return -EPERM;
2190
2191         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2192                 return -EFAULT;
2193
2194         rtnl_lock();
2195         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2196                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2197                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2198         rtnl_unlock();
2199         return err;
2200 }
2201
2202 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2203 {
2204         struct in6_ifreq ireq;
2205         int err;
2206
2207         if (!capable(CAP_NET_ADMIN))
2208                 return -EPERM;
2209
2210         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2211                 return -EFAULT;
2212
2213         rtnl_lock();
2214         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2215                              ireq.ifr6_prefixlen);
2216         rtnl_unlock();
2217         return err;
2218 }
2219
2220 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2221                      int plen, int scope)
2222 {
2223         struct inet6_ifaddr *ifp;
2224
2225         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2226         if (!IS_ERR(ifp)) {
2227                 spin_lock_bh(&ifp->lock);
2228                 ifp->flags &= ~IFA_F_TENTATIVE;
2229                 spin_unlock_bh(&ifp->lock);
2230                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2231                 in6_ifa_put(ifp);
2232         }
2233 }
2234
2235 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2236 static void sit_add_v4_addrs(struct inet6_dev *idev)
2237 {
2238         struct in6_addr addr;
2239         struct net_device *dev;
2240         struct net *net = dev_net(idev->dev);
2241         int scope;
2242
2243         ASSERT_RTNL();
2244
2245         memset(&addr, 0, sizeof(struct in6_addr));
2246         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2247
2248         if (idev->dev->flags&IFF_POINTOPOINT) {
2249                 addr.s6_addr32[0] = htonl(0xfe800000);
2250                 scope = IFA_LINK;
2251         } else {
2252                 scope = IPV6_ADDR_COMPATv4;
2253         }
2254
2255         if (addr.s6_addr32[3]) {
2256                 add_addr(idev, &addr, 128, scope);
2257                 return;
2258         }
2259
2260         for_each_netdev(net, dev) {
2261                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2262                 if (in_dev && (dev->flags & IFF_UP)) {
2263                         struct in_ifaddr * ifa;
2264
2265                         int flag = scope;
2266
2267                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2268                                 int plen;
2269
2270                                 addr.s6_addr32[3] = ifa->ifa_local;
2271
2272                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2273                                         continue;
2274                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2275                                         if (idev->dev->flags&IFF_POINTOPOINT)
2276                                                 continue;
2277                                         flag |= IFA_HOST;
2278                                 }
2279                                 if (idev->dev->flags&IFF_POINTOPOINT)
2280                                         plen = 64;
2281                                 else
2282                                         plen = 96;
2283
2284                                 add_addr(idev, &addr, plen, flag);
2285                         }
2286                 }
2287         }
2288 }
2289 #endif
2290
2291 static void init_loopback(struct net_device *dev)
2292 {
2293         struct inet6_dev  *idev;
2294
2295         /* ::1 */
2296
2297         ASSERT_RTNL();
2298
2299         if ((idev = ipv6_find_idev(dev)) == NULL) {
2300                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2301                 return;
2302         }
2303
2304         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2305 }
2306
2307 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2308 {
2309         struct inet6_ifaddr * ifp;
2310         u32 addr_flags = IFA_F_PERMANENT;
2311
2312 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2313         if (idev->cnf.optimistic_dad &&
2314             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2315                 addr_flags |= IFA_F_OPTIMISTIC;
2316 #endif
2317
2318
2319         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2320         if (!IS_ERR(ifp)) {
2321                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2322                 addrconf_dad_start(ifp, 0);
2323                 in6_ifa_put(ifp);
2324         }
2325 }
2326
2327 static void addrconf_dev_config(struct net_device *dev)
2328 {
2329         struct in6_addr addr;
2330         struct inet6_dev    * idev;
2331
2332         ASSERT_RTNL();
2333
2334         if ((dev->type != ARPHRD_ETHER) &&
2335             (dev->type != ARPHRD_FDDI) &&
2336             (dev->type != ARPHRD_IEEE802_TR) &&
2337             (dev->type != ARPHRD_ARCNET) &&
2338             (dev->type != ARPHRD_INFINIBAND)) {
2339                 /* Alas, we support only Ethernet autoconfiguration. */
2340                 return;
2341         }
2342
2343         idev = addrconf_add_dev(dev);
2344         if (idev == NULL)
2345                 return;
2346
2347         memset(&addr, 0, sizeof(struct in6_addr));
2348         addr.s6_addr32[0] = htonl(0xFE800000);
2349
2350         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2351                 addrconf_add_linklocal(idev, &addr);
2352 }
2353
2354 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2355 static void addrconf_sit_config(struct net_device *dev)
2356 {
2357         struct inet6_dev *idev;
2358
2359         ASSERT_RTNL();
2360
2361         /*
2362          * Configure the tunnel with one of our IPv4
2363          * addresses... we should configure all of
2364          * our v4 addrs in the tunnel
2365          */
2366
2367         if ((idev = ipv6_find_idev(dev)) == NULL) {
2368                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2369                 return;
2370         }
2371
2372         if (dev->priv_flags & IFF_ISATAP) {
2373                 struct in6_addr addr;
2374
2375                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2376                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2377                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2378                         addrconf_add_linklocal(idev, &addr);
2379                 return;
2380         }
2381
2382         sit_add_v4_addrs(idev);
2383
2384         if (dev->flags&IFF_POINTOPOINT) {
2385                 addrconf_add_mroute(dev);
2386                 addrconf_add_lroute(dev);
2387         } else
2388                 sit_route_add(dev);
2389 }
2390 #endif
2391
2392 static inline int
2393 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2394 {
2395         struct in6_addr lladdr;
2396
2397         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2398                 addrconf_add_linklocal(idev, &lladdr);
2399                 return 0;
2400         }
2401         return -1;
2402 }
2403
2404 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2405 {
2406         struct net_device *link_dev;
2407         struct net *net = dev_net(idev->dev);
2408
2409         /* first try to inherit the link-local address from the link device */
2410         if (idev->dev->iflink &&
2411             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2412                 if (!ipv6_inherit_linklocal(idev, link_dev))
2413                         return;
2414         }
2415         /* then try to inherit it from any device */
2416         for_each_netdev(net, link_dev) {
2417                 if (!ipv6_inherit_linklocal(idev, link_dev))
2418                         return;
2419         }
2420         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2421 }
2422
2423 /*
2424  * Autoconfigure tunnel with a link-local address so routing protocols,
2425  * DHCPv6, MLD etc. can be run over the virtual link
2426  */
2427
2428 static void addrconf_ip6_tnl_config(struct net_device *dev)
2429 {
2430         struct inet6_dev *idev;
2431
2432         ASSERT_RTNL();
2433
2434         if ((idev = addrconf_add_dev(dev)) == NULL) {
2435                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2436                 return;
2437         }
2438         ip6_tnl_add_linklocal(idev);
2439 }
2440
2441 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2442                            void * data)
2443 {
2444         struct net_device *dev = (struct net_device *) data;
2445         struct inet6_dev *idev = __in6_dev_get(dev);
2446         int run_pending = 0;
2447         int err;
2448
2449         switch(event) {
2450         case NETDEV_REGISTER:
2451                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2452                         idev = ipv6_add_dev(dev);
2453                         if (!idev)
2454                                 return notifier_from_errno(-ENOMEM);
2455                 }
2456                 break;
2457         case NETDEV_UP:
2458         case NETDEV_CHANGE:
2459                 if (dev->flags & IFF_SLAVE)
2460                         break;
2461
2462                 if (event == NETDEV_UP) {
2463                         if (!addrconf_qdisc_ok(dev)) {
2464                                 /* device is not ready yet. */
2465                                 printk(KERN_INFO
2466                                         "ADDRCONF(NETDEV_UP): %s: "
2467                                         "link is not ready\n",
2468                                         dev->name);
2469                                 break;
2470                         }
2471
2472                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2473                                 idev = ipv6_add_dev(dev);
2474
2475                         if (idev) {
2476                                 idev->if_flags |= IF_READY;
2477                                 run_pending = 1;
2478                         }
2479                 } else {
2480                         if (!addrconf_qdisc_ok(dev)) {
2481                                 /* device is still not ready. */
2482                                 break;
2483                         }
2484
2485                         if (idev) {
2486                                 if (idev->if_flags & IF_READY) {
2487                                         /* device is already configured. */
2488                                         break;
2489                                 }
2490                                 idev->if_flags |= IF_READY;
2491                         }
2492
2493                         printk(KERN_INFO
2494                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2495                                         "link becomes ready\n",
2496                                         dev->name);
2497
2498                         run_pending = 1;
2499                 }
2500
2501                 switch(dev->type) {
2502 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2503                 case ARPHRD_SIT:
2504                         addrconf_sit_config(dev);
2505                         break;
2506 #endif
2507                 case ARPHRD_TUNNEL6:
2508                         addrconf_ip6_tnl_config(dev);
2509                         break;
2510                 case ARPHRD_LOOPBACK:
2511                         init_loopback(dev);
2512                         break;
2513
2514                 default:
2515                         addrconf_dev_config(dev);
2516                         break;
2517                 }
2518                 if (idev) {
2519                         if (run_pending)
2520                                 addrconf_dad_run(idev);
2521
2522                         /* If the MTU changed during the interface down, when the
2523                            interface up, the changed MTU must be reflected in the
2524                            idev as well as routers.
2525                          */
2526                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2527                                 rt6_mtu_change(dev, dev->mtu);
2528                                 idev->cnf.mtu6 = dev->mtu;
2529                         }
2530                         idev->tstamp = jiffies;
2531                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2532                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2533                            stop IPv6 on this interface.
2534                          */
2535                         if (dev->mtu < IPV6_MIN_MTU)
2536                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2537                 }
2538                 break;
2539
2540         case NETDEV_CHANGEMTU:
2541                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2542                         rt6_mtu_change(dev, dev->mtu);
2543                         idev->cnf.mtu6 = dev->mtu;
2544                         break;
2545                 }
2546
2547                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2548                         idev = ipv6_add_dev(dev);
2549                         if (idev)
2550                                 break;
2551                 }
2552
2553                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2554
2555         case NETDEV_DOWN:
2556         case NETDEV_UNREGISTER:
2557                 /*
2558                  *      Remove all addresses from this interface.
2559                  */
2560                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2561                 break;
2562
2563         case NETDEV_CHANGENAME:
2564                 if (idev) {
2565                         snmp6_unregister_dev(idev);
2566                         addrconf_sysctl_unregister(idev);
2567                         addrconf_sysctl_register(idev);
2568                         err = snmp6_register_dev(idev);
2569                         if (err)
2570                                 return notifier_from_errno(err);
2571                 }
2572                 break;
2573         case NETDEV_PRE_TYPE_CHANGE:
2574         case NETDEV_POST_TYPE_CHANGE:
2575                 addrconf_type_change(dev, event);
2576                 break;
2577         }
2578
2579         return NOTIFY_OK;
2580 }
2581
2582 /*
2583  *      addrconf module should be notified of a device going up
2584  */
2585 static struct notifier_block ipv6_dev_notf = {
2586         .notifier_call = addrconf_notify,
2587         .priority = 0
2588 };
2589
2590 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2591 {
2592         struct inet6_dev *idev;
2593         ASSERT_RTNL();
2594
2595         idev = __in6_dev_get(dev);
2596
2597         if (event == NETDEV_POST_TYPE_CHANGE)
2598                 ipv6_mc_remap(idev);
2599         else if (event == NETDEV_PRE_TYPE_CHANGE)
2600                 ipv6_mc_unmap(idev);
2601 }
2602
2603 static int addrconf_ifdown(struct net_device *dev, int how)
2604 {
2605         struct inet6_dev *idev;
2606         struct inet6_ifaddr *ifa, *keep_list, **bifa;
2607         struct net *net = dev_net(dev);
2608
2609         ASSERT_RTNL();
2610
2611         rt6_ifdown(net, dev);
2612         neigh_ifdown(&nd_tbl, dev);
2613
2614         idev = __in6_dev_get(dev);
2615         if (idev == NULL)
2616                 return -ENODEV;
2617
2618         /* Step 1: remove reference to ipv6 device from parent device.
2619                    Do not dev_put!
2620          */
2621         if (how) {
2622                 idev->dead = 1;
2623
2624                 /* protected by rtnl_lock */
2625                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2626
2627                 /* Step 1.5: remove snmp6 entry */
2628                 snmp6_unregister_dev(idev);
2629
2630         }
2631
2632         write_lock_bh(&idev->lock);
2633
2634         /* Step 3: clear flags for stateless addrconf */
2635         if (!how)
2636                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2637
2638         /* Step 4: clear address list */
2639 #ifdef CONFIG_IPV6_PRIVACY
2640         if (how && del_timer(&idev->regen_timer))
2641                 in6_dev_put(idev);
2642
2643         /* clear tempaddr list */
2644         while (!list_empty(&idev->tempaddr_list)) {
2645                 ifa = list_first_entry(&idev->tempaddr_list,
2646                                        struct inet6_ifaddr, tmp_list);
2647                 list_del(&ifa->tmp_list);
2648                 ifa->dead = 1;
2649                 write_unlock_bh(&idev->lock);
2650                 spin_lock_bh(&ifa->lock);
2651
2652                 if (ifa->ifpub) {
2653                         in6_ifa_put(ifa->ifpub);
2654                         ifa->ifpub = NULL;
2655                 }
2656                 spin_unlock_bh(&ifa->lock);
2657                 in6_ifa_put(ifa);
2658                 write_lock_bh(&idev->lock);
2659         }
2660 #endif
2661         keep_list = NULL;
2662         bifa = &keep_list;
2663         while ((ifa = idev->addr_list) != NULL) {
2664                 idev->addr_list = ifa->if_next;
2665                 ifa->if_next = NULL;
2666
2667                 addrconf_del_timer(ifa);
2668
2669                 /* If just doing link down, and address is permanent
2670                    and not link-local, then retain it. */
2671                 if (how == 0 &&
2672                     (ifa->flags&IFA_F_PERMANENT) &&
2673                     !(ipv6_addr_type(&ifa->addr) & IPV6_ADDR_LINKLOCAL)) {
2674
2675                         /* Move to holding list */
2676                         *bifa = ifa;
2677                         bifa = &ifa->if_next;
2678
2679                         /* If not doing DAD on this address, just keep it. */
2680                         if ((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2681                             idev->cnf.accept_dad <= 0 ||
2682                             (ifa->flags & IFA_F_NODAD))
2683                                 continue;
2684
2685                         /* If it was tentative already, no need to notify */
2686                         if (ifa->flags & IFA_F_TENTATIVE)
2687                                 continue;
2688
2689                         /* Flag it for later restoration when link comes up */
2690                         ifa->flags |= IFA_F_TENTATIVE;
2691                         in6_ifa_hold(ifa);
2692                 } else {
2693                         ifa->dead = 1;
2694                 }
2695                 write_unlock_bh(&idev->lock);
2696
2697                 /* clear hash table */
2698                 spin_lock_bh(&addrconf_hash_lock);
2699                 hlist_del_init_rcu(&ifa->addr_lst);
2700                 __in6_ifa_put(ifa);
2701                 spin_unlock_bh(&addrconf_hash_lock);
2702
2703                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2704                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2705                 in6_ifa_put(ifa);
2706
2707                 write_lock_bh(&idev->lock);
2708         }
2709
2710         idev->addr_list = keep_list;
2711
2712         write_unlock_bh(&idev->lock);
2713
2714         /* Step 5: Discard multicast list */
2715
2716         if (how)
2717                 ipv6_mc_destroy_dev(idev);
2718         else
2719                 ipv6_mc_down(idev);
2720
2721         idev->tstamp = jiffies;
2722
2723         /* Shot the device (if unregistered) */
2724
2725         if (how) {
2726                 addrconf_sysctl_unregister(idev);
2727                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2728                 neigh_ifdown(&nd_tbl, dev);
2729                 in6_dev_put(idev);
2730         }
2731         return 0;
2732 }
2733
2734 static void addrconf_rs_timer(unsigned long data)
2735 {
2736         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2737         struct inet6_dev *idev = ifp->idev;
2738
2739         read_lock(&idev->lock);
2740         if (idev->dead || !(idev->if_flags & IF_READY))
2741                 goto out;
2742
2743         if (idev->cnf.forwarding)
2744                 goto out;
2745
2746         /* Announcement received after solicitation was sent */
2747         if (idev->if_flags & IF_RA_RCVD)
2748                 goto out;
2749
2750         spin_lock(&ifp->lock);
2751         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2752                 /* The wait after the last probe can be shorter */
2753                 addrconf_mod_timer(ifp, AC_RS,
2754                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2755                                    idev->cnf.rtr_solicit_delay :
2756                                    idev->cnf.rtr_solicit_interval);
2757                 spin_unlock(&ifp->lock);
2758
2759                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2760         } else {
2761                 spin_unlock(&ifp->lock);
2762                 /*
2763                  * Note: we do not support deprecated "all on-link"
2764                  * assumption any longer.
2765                  */
2766                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2767                        idev->dev->name);
2768         }
2769
2770 out:
2771         read_unlock(&idev->lock);
2772         in6_ifa_put(ifp);
2773 }
2774
2775 /*
2776  *      Duplicate Address Detection
2777  */
2778 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2779 {
2780         unsigned long rand_num;
2781         struct inet6_dev *idev = ifp->idev;
2782
2783         if (ifp->flags & IFA_F_OPTIMISTIC)
2784                 rand_num = 0;
2785         else
2786                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2787
2788         ifp->probes = idev->cnf.dad_transmits;
2789         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2790 }
2791
2792 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2793 {
2794         struct inet6_dev *idev = ifp->idev;
2795         struct net_device *dev = idev->dev;
2796
2797         addrconf_join_solict(dev, &ifp->addr);
2798
2799         net_srandom(ifp->addr.s6_addr32[3]);
2800
2801         read_lock_bh(&idev->lock);
2802         if (ifp->dead)
2803                 goto out;
2804
2805         spin_lock(&ifp->lock);
2806         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2807             idev->cnf.accept_dad < 1 ||
2808             !(ifp->flags&IFA_F_TENTATIVE) ||
2809             ifp->flags & IFA_F_NODAD) {
2810                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2811                 spin_unlock(&ifp->lock);
2812                 read_unlock_bh(&idev->lock);
2813
2814                 addrconf_dad_completed(ifp);
2815                 return;
2816         }
2817
2818         if (!(idev->if_flags & IF_READY)) {
2819                 spin_unlock(&ifp->lock);
2820                 read_unlock_bh(&idev->lock);
2821                 /*
2822                  * If the device is not ready:
2823                  * - keep it tentative if it is a permanent address.
2824                  * - otherwise, kill it.
2825                  */
2826                 in6_ifa_hold(ifp);
2827                 addrconf_dad_stop(ifp, 0);
2828                 return;
2829         }
2830
2831         /*
2832          * Optimistic nodes can start receiving
2833          * Frames right away
2834          */
2835         if(ifp->flags & IFA_F_OPTIMISTIC)
2836                 ip6_ins_rt(ifp->rt);
2837
2838         addrconf_dad_kick(ifp);
2839         spin_unlock(&ifp->lock);
2840 out:
2841         read_unlock_bh(&idev->lock);
2842 }
2843
2844 static void addrconf_dad_timer(unsigned long data)
2845 {
2846         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2847         struct inet6_dev *idev = ifp->idev;
2848         struct in6_addr mcaddr;
2849
2850         read_lock(&idev->lock);
2851         if (idev->dead || !(idev->if_flags & IF_READY)) {
2852                 read_unlock(&idev->lock);
2853                 goto out;
2854         }
2855
2856         spin_lock(&ifp->lock);
2857         if (ifp->probes == 0) {
2858                 /*
2859                  * DAD was successful
2860                  */
2861
2862                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2863                 spin_unlock(&ifp->lock);
2864                 read_unlock(&idev->lock);
2865
2866                 addrconf_dad_completed(ifp);
2867
2868                 goto out;
2869         }
2870
2871         ifp->probes--;
2872         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2873         spin_unlock(&ifp->lock);
2874         read_unlock(&idev->lock);
2875
2876         /* send a neighbour solicitation for our addr */
2877         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2878         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2879 out:
2880         in6_ifa_put(ifp);
2881 }
2882
2883 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2884 {
2885         struct net_device *     dev = ifp->idev->dev;
2886
2887         /*
2888          *      Configure the address for reception. Now it is valid.
2889          */
2890
2891         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2892
2893         /* If added prefix is link local and forwarding is off,
2894            start sending router solicitations.
2895          */
2896
2897         if (ifp->idev->cnf.forwarding == 0 &&
2898             ifp->idev->cnf.rtr_solicits > 0 &&
2899             (dev->flags&IFF_LOOPBACK) == 0 &&
2900             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2901                 /*
2902                  *      If a host as already performed a random delay
2903                  *      [...] as part of DAD [...] there is no need
2904                  *      to delay again before sending the first RS
2905                  */
2906                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2907
2908                 spin_lock_bh(&ifp->lock);
2909                 ifp->probes = 1;
2910                 ifp->idev->if_flags |= IF_RS_SENT;
2911                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2912                 spin_unlock_bh(&ifp->lock);
2913         }
2914 }
2915
2916 static void addrconf_dad_run(struct inet6_dev *idev) {
2917         struct inet6_ifaddr *ifp;
2918
2919         read_lock_bh(&idev->lock);
2920         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2921                 spin_lock(&ifp->lock);
2922                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2923                         spin_unlock(&ifp->lock);
2924                         continue;
2925                 }
2926                 spin_unlock(&ifp->lock);
2927                 addrconf_dad_kick(ifp);
2928         }
2929         read_unlock_bh(&idev->lock);
2930 }
2931
2932 #ifdef CONFIG_PROC_FS
2933 struct if6_iter_state {
2934         struct seq_net_private p;
2935         int bucket;
2936 };
2937
2938 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2939 {
2940         struct inet6_ifaddr *ifa = NULL;
2941         struct if6_iter_state *state = seq->private;
2942         struct net *net = seq_file_net(seq);
2943
2944         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2945                 struct hlist_node *n;
2946                 hlist_for_each_entry_rcu(ifa, n, &inet6_addr_lst[state->bucket],
2947                                          addr_lst)
2948                         if (net_eq(dev_net(ifa->idev->dev), net))
2949                                 return ifa;
2950         }
2951         return NULL;
2952 }
2953
2954 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
2955                                          struct inet6_ifaddr *ifa)
2956 {
2957         struct if6_iter_state *state = seq->private;
2958         struct net *net = seq_file_net(seq);
2959         struct hlist_node *n = &ifa->addr_lst;
2960
2961         hlist_for_each_entry_continue_rcu(ifa, n, addr_lst)
2962                 if (net_eq(dev_net(ifa->idev->dev), net))
2963                         return ifa;
2964
2965         while (++state->bucket < IN6_ADDR_HSIZE) {
2966                 hlist_for_each_entry(ifa, n,
2967                                      &inet6_addr_lst[state->bucket], addr_lst) {
2968                         if (net_eq(dev_net(ifa->idev->dev), net))
2969                                 return ifa;
2970                 }
2971         }
2972
2973         return NULL;
2974 }
2975
2976 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2977 {
2978         struct inet6_ifaddr *ifa = if6_get_first(seq);
2979
2980         if (ifa)
2981                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2982                         --pos;
2983         return pos ? NULL : ifa;
2984 }
2985
2986 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2987         __acquires(rcu)
2988 {
2989         rcu_read_lock_bh();
2990         return if6_get_idx(seq, *pos);
2991 }
2992
2993 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2994 {
2995         struct inet6_ifaddr *ifa;
2996
2997         ifa = if6_get_next(seq, v);
2998         ++*pos;
2999         return ifa;
3000 }
3001
3002 static void if6_seq_stop(struct seq_file *seq, void *v)
3003         __releases(rcu)
3004 {
3005         rcu_read_unlock_bh();
3006 }
3007
3008 static int if6_seq_show(struct seq_file *seq, void *v)
3009 {
3010         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3011         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3012                    &ifp->addr,
3013                    ifp->idev->dev->ifindex,
3014                    ifp->prefix_len,
3015                    ifp->scope,
3016                    ifp->flags,
3017                    ifp->idev->dev->name);
3018         return 0;
3019 }
3020
3021 static const struct seq_operations if6_seq_ops = {
3022         .start  = if6_seq_start,
3023         .next   = if6_seq_next,
3024         .show   = if6_seq_show,
3025         .stop   = if6_seq_stop,
3026 };
3027
3028 static int if6_seq_open(struct inode *inode, struct file *file)
3029 {
3030         return seq_open_net(inode, file, &if6_seq_ops,
3031                             sizeof(struct if6_iter_state));
3032 }
3033
3034 static const struct file_operations if6_fops = {
3035         .owner          = THIS_MODULE,
3036         .open           = if6_seq_open,
3037         .read           = seq_read,
3038         .llseek         = seq_lseek,
3039         .release        = seq_release_net,
3040 };
3041
3042 static int __net_init if6_proc_net_init(struct net *net)
3043 {
3044         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3045                 return -ENOMEM;
3046         return 0;
3047 }
3048
3049 static void __net_exit if6_proc_net_exit(struct net *net)
3050 {
3051        proc_net_remove(net, "if_inet6");
3052 }
3053
3054 static struct pernet_operations if6_proc_net_ops = {
3055        .init = if6_proc_net_init,
3056        .exit = if6_proc_net_exit,
3057 };
3058
3059 int __init if6_proc_init(void)
3060 {
3061         return register_pernet_subsys(&if6_proc_net_ops);
3062 }
3063
3064 void if6_proc_exit(void)
3065 {
3066         unregister_pernet_subsys(&if6_proc_net_ops);
3067 }
3068 #endif  /* CONFIG_PROC_FS */
3069
3070 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3071 /* Check if address is a home address configured on any interface. */
3072 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3073 {
3074         int ret = 0;
3075         struct inet6_ifaddr *ifp = NULL;
3076         struct hlist_node *n;
3077         unsigned int hash = ipv6_addr_hash(addr);
3078
3079         rcu_read_lock_bh();
3080         hlist_for_each_entry_rcu(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3081                 if (!net_eq(dev_net(ifp->idev->dev), net))
3082                         continue;
3083                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3084                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3085                         ret = 1;
3086                         break;
3087                 }
3088         }
3089         rcu_read_unlock_bh();
3090         return ret;
3091 }
3092 #endif
3093
3094 /*
3095  *      Periodic address status verification
3096  */
3097
3098 static void addrconf_verify(unsigned long foo)
3099 {
3100         struct inet6_ifaddr *ifp;
3101         struct hlist_node *node;
3102         unsigned long now, next;
3103         int i;
3104
3105         rcu_read_lock_bh();
3106         spin_lock(&addrconf_verify_lock);
3107         now = jiffies;
3108         next = now + ADDR_CHECK_FREQUENCY;
3109
3110         del_timer(&addr_chk_timer);
3111
3112         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3113
3114 restart:
3115                 hlist_for_each_entry_rcu(ifp, node,
3116                                          &inet6_addr_lst[i], addr_lst) {
3117                         unsigned long age;
3118 #ifdef CONFIG_IPV6_PRIVACY
3119                         unsigned long regen_advance;
3120 #endif
3121
3122                         if (ifp->flags & IFA_F_PERMANENT)
3123                                 continue;
3124
3125                         spin_lock(&ifp->lock);
3126                         age = (now - ifp->tstamp) / HZ;
3127
3128 #ifdef CONFIG_IPV6_PRIVACY
3129                         regen_advance = ifp->idev->cnf.regen_max_retry *
3130                                         ifp->idev->cnf.dad_transmits *
3131                                         ifp->idev->nd_parms->retrans_time / HZ;
3132 #endif
3133
3134                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3135                             age >= ifp->valid_lft) {
3136                                 spin_unlock(&ifp->lock);
3137                                 in6_ifa_hold(ifp);
3138                                 ipv6_del_addr(ifp);
3139                                 goto restart;
3140                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3141                                 spin_unlock(&ifp->lock);
3142                                 continue;
3143                         } else if (age >= ifp->prefered_lft) {
3144                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3145                                 int deprecate = 0;
3146
3147                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3148                                         deprecate = 1;
3149                                         ifp->flags |= IFA_F_DEPRECATED;
3150                                 }
3151
3152                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3153                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3154
3155                                 spin_unlock(&ifp->lock);
3156
3157                                 if (deprecate) {
3158                                         in6_ifa_hold(ifp);
3159
3160                                         ipv6_ifa_notify(0, ifp);
3161                                         in6_ifa_put(ifp);
3162                                         goto restart;
3163                                 }
3164 #ifdef CONFIG_IPV6_PRIVACY
3165                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3166                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3167                                 if (age >= ifp->prefered_lft - regen_advance) {
3168                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3169                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3170                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3171                                         if (!ifp->regen_count && ifpub) {
3172                                                 ifp->regen_count++;
3173                                                 in6_ifa_hold(ifp);
3174                                                 in6_ifa_hold(ifpub);
3175                                                 spin_unlock(&ifp->lock);
3176
3177                                                 spin_lock(&ifpub->lock);
3178                                                 ifpub->regen_count = 0;
3179                                                 spin_unlock(&ifpub->lock);
3180                                                 ipv6_create_tempaddr(ifpub, ifp);
3181                                                 in6_ifa_put(ifpub);
3182                                                 in6_ifa_put(ifp);
3183                                                 goto restart;
3184                                         }
3185                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3186                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3187                                 spin_unlock(&ifp->lock);
3188 #endif
3189                         } else {
3190                                 /* ifp->prefered_lft <= ifp->valid_lft */
3191                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3192                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3193                                 spin_unlock(&ifp->lock);
3194                         }
3195                 }
3196         }
3197
3198         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3199         add_timer(&addr_chk_timer);
3200         spin_unlock(&addrconf_verify_lock);
3201         rcu_read_unlock_bh();
3202 }
3203
3204 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3205 {
3206         struct in6_addr *pfx = NULL;
3207
3208         if (addr)
3209                 pfx = nla_data(addr);
3210
3211         if (local) {
3212                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3213                         pfx = NULL;
3214                 else
3215                         pfx = nla_data(local);
3216         }
3217
3218         return pfx;
3219 }
3220
3221 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3222         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3223         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3224         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3225 };
3226
3227 static int
3228 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3229 {
3230         struct net *net = sock_net(skb->sk);
3231         struct ifaddrmsg *ifm;
3232         struct nlattr *tb[IFA_MAX+1];
3233         struct in6_addr *pfx;
3234         int err;
3235
3236         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3237         if (err < 0)
3238                 return err;
3239
3240         ifm = nlmsg_data(nlh);
3241         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3242         if (pfx == NULL)
3243                 return -EINVAL;
3244
3245         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3246 }
3247
3248 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3249                              u32 prefered_lft, u32 valid_lft)
3250 {
3251         u32 flags;
3252         clock_t expires;
3253         unsigned long timeout;
3254
3255         if (!valid_lft || (prefered_lft > valid_lft))
3256                 return -EINVAL;
3257
3258         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3259         if (addrconf_finite_timeout(timeout)) {
3260                 expires = jiffies_to_clock_t(timeout * HZ);
3261                 valid_lft = timeout;
3262                 flags = RTF_EXPIRES;
3263         } else {
3264                 expires = 0;
3265                 flags = 0;
3266                 ifa_flags |= IFA_F_PERMANENT;
3267         }
3268
3269         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3270         if (addrconf_finite_timeout(timeout)) {
3271                 if (timeout == 0)
3272                         ifa_flags |= IFA_F_DEPRECATED;
3273                 prefered_lft = timeout;
3274         }
3275
3276         spin_lock_bh(&ifp->lock);
3277         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3278         ifp->tstamp = jiffies;
3279         ifp->valid_lft = valid_lft;
3280         ifp->prefered_lft = prefered_lft;
3281
3282         spin_unlock_bh(&ifp->lock);
3283         if (!(ifp->flags&IFA_F_TENTATIVE))
3284                 ipv6_ifa_notify(0, ifp);
3285
3286         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3287                               expires, flags);
3288         addrconf_verify(0);
3289
3290         return 0;
3291 }
3292
3293 static int
3294 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3295 {
3296         struct net *net = sock_net(skb->sk);
3297         struct ifaddrmsg *ifm;
3298         struct nlattr *tb[IFA_MAX+1];
3299         struct in6_addr *pfx;
3300         struct inet6_ifaddr *ifa;
3301         struct net_device *dev;
3302         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3303         u8 ifa_flags;
3304         int err;
3305
3306         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3307         if (err < 0)
3308                 return err;
3309
3310         ifm = nlmsg_data(nlh);
3311         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3312         if (pfx == NULL)
3313                 return -EINVAL;
3314
3315         if (tb[IFA_CACHEINFO]) {
3316                 struct ifa_cacheinfo *ci;
3317
3318                 ci = nla_data(tb[IFA_CACHEINFO]);
3319                 valid_lft = ci->ifa_valid;
3320                 preferred_lft = ci->ifa_prefered;
3321         } else {
3322                 preferred_lft = INFINITY_LIFE_TIME;
3323                 valid_lft = INFINITY_LIFE_TIME;
3324         }
3325
3326         dev =  __dev_get_by_index(net, ifm->ifa_index);
3327         if (dev == NULL)
3328                 return -ENODEV;
3329
3330         /* We ignore other flags so far. */
3331         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3332
3333         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3334         if (ifa == NULL) {
3335                 /*
3336                  * It would be best to check for !NLM_F_CREATE here but
3337                  * userspace alreay relies on not having to provide this.
3338                  */
3339                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3340                                       ifm->ifa_prefixlen, ifa_flags,
3341                                       preferred_lft, valid_lft);
3342         }
3343
3344         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3345             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3346                 err = -EEXIST;
3347         else
3348                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3349
3350         in6_ifa_put(ifa);
3351
3352         return err;
3353 }
3354
3355 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3356                           u8 scope, int ifindex)
3357 {
3358         struct ifaddrmsg *ifm;
3359
3360         ifm = nlmsg_data(nlh);
3361         ifm->ifa_family = AF_INET6;
3362         ifm->ifa_prefixlen = prefixlen;
3363         ifm->ifa_flags = flags;
3364         ifm->ifa_scope = scope;
3365         ifm->ifa_index = ifindex;
3366 }
3367
3368 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3369                          unsigned long tstamp, u32 preferred, u32 valid)
3370 {
3371         struct ifa_cacheinfo ci;
3372
3373         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3374                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3375         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3376                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3377         ci.ifa_prefered = preferred;
3378         ci.ifa_valid = valid;
3379
3380         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3381 }
3382
3383 static inline int rt_scope(int ifa_scope)
3384 {
3385         if (ifa_scope & IFA_HOST)
3386                 return RT_SCOPE_HOST;
3387         else if (ifa_scope & IFA_LINK)
3388                 return RT_SCOPE_LINK;
3389         else if (ifa_scope & IFA_SITE)
3390                 return RT_SCOPE_SITE;
3391         else
3392                 return RT_SCOPE_UNIVERSE;
3393 }
3394
3395 static inline int inet6_ifaddr_msgsize(void)
3396 {
3397         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3398                + nla_total_size(16) /* IFA_ADDRESS */
3399                + nla_total_size(sizeof(struct ifa_cacheinfo));
3400 }
3401
3402 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3403                              u32 pid, u32 seq, int event, unsigned int flags)
3404 {
3405         struct nlmsghdr  *nlh;
3406         u32 preferred, valid;
3407
3408         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3409         if (nlh == NULL)
3410                 return -EMSGSIZE;
3411
3412         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3413                       ifa->idev->dev->ifindex);
3414
3415         if (!(ifa->flags&IFA_F_PERMANENT)) {
3416                 preferred = ifa->prefered_lft;
3417                 valid = ifa->valid_lft;
3418                 if (preferred != INFINITY_LIFE_TIME) {
3419                         long tval = (jiffies - ifa->tstamp)/HZ;
3420                         if (preferred > tval)
3421                                 preferred -= tval;
3422                         else
3423                                 preferred = 0;
3424                         if (valid != INFINITY_LIFE_TIME)
3425                                 valid -= tval;
3426                 }
3427         } else {
3428                 preferred = INFINITY_LIFE_TIME;
3429                 valid = INFINITY_LIFE_TIME;
3430         }
3431
3432         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3433             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3434                 nlmsg_cancel(skb, nlh);
3435                 return -EMSGSIZE;
3436         }
3437
3438         return nlmsg_end(skb, nlh);
3439 }
3440
3441 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3442                                 u32 pid, u32 seq, int event, u16 flags)
3443 {
3444         struct nlmsghdr  *nlh;
3445         u8 scope = RT_SCOPE_UNIVERSE;
3446         int ifindex = ifmca->idev->dev->ifindex;
3447
3448         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3449                 scope = RT_SCOPE_SITE;
3450
3451         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3452         if (nlh == NULL)
3453                 return -EMSGSIZE;
3454
3455         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3456         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3457             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3458                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3459                 nlmsg_cancel(skb, nlh);
3460                 return -EMSGSIZE;
3461         }
3462
3463         return nlmsg_end(skb, nlh);
3464 }
3465
3466 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3467                                 u32 pid, u32 seq, int event, unsigned int flags)
3468 {
3469         struct nlmsghdr  *nlh;
3470         u8 scope = RT_SCOPE_UNIVERSE;
3471         int ifindex = ifaca->aca_idev->dev->ifindex;
3472
3473         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3474                 scope = RT_SCOPE_SITE;
3475
3476         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3477         if (nlh == NULL)
3478                 return -EMSGSIZE;
3479
3480         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3481         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3482             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3483                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3484                 nlmsg_cancel(skb, nlh);
3485                 return -EMSGSIZE;
3486         }
3487
3488         return nlmsg_end(skb, nlh);
3489 }
3490
3491 enum addr_type_t
3492 {
3493         UNICAST_ADDR,
3494         MULTICAST_ADDR,
3495         ANYCAST_ADDR,
3496 };
3497
3498 /* called with rcu_read_lock() */
3499 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3500                           struct netlink_callback *cb, enum addr_type_t type,
3501                           int s_ip_idx, int *p_ip_idx)
3502 {
3503         struct inet6_ifaddr *ifa;
3504         struct ifmcaddr6 *ifmca;
3505         struct ifacaddr6 *ifaca;
3506         int err = 1;
3507         int ip_idx = *p_ip_idx;
3508
3509         read_lock_bh(&idev->lock);
3510         switch (type) {
3511         case UNICAST_ADDR:
3512                 /* unicast address incl. temp addr */
3513                 for (ifa = idev->addr_list; ifa;
3514                      ifa = ifa->if_next, ip_idx++) {
3515                         if (ip_idx < s_ip_idx)
3516                                 continue;
3517                         err = inet6_fill_ifaddr(skb, ifa,
3518                                                 NETLINK_CB(cb->skb).pid,
3519                                                 cb->nlh->nlmsg_seq,
3520                                                 RTM_NEWADDR,
3521                                                 NLM_F_MULTI);
3522                         if (err <= 0)
3523                                 break;
3524                 }
3525                 break;
3526         case MULTICAST_ADDR:
3527                 /* multicast address */
3528                 for (ifmca = idev->mc_list; ifmca;
3529                      ifmca = ifmca->next, ip_idx++) {
3530                         if (ip_idx < s_ip_idx)
3531                                 continue;
3532                         err = inet6_fill_ifmcaddr(skb, ifmca,
3533                                                   NETLINK_CB(cb->skb).pid,
3534                                                   cb->nlh->nlmsg_seq,
3535                                                   RTM_GETMULTICAST,
3536                                                   NLM_F_MULTI);
3537                         if (err <= 0)
3538                                 break;
3539                 }
3540                 break;
3541         case ANYCAST_ADDR:
3542                 /* anycast address */
3543                 for (ifaca = idev->ac_list; ifaca;
3544                      ifaca = ifaca->aca_next, ip_idx++) {
3545                         if (ip_idx < s_ip_idx)
3546                                 continue;
3547                         err = inet6_fill_ifacaddr(skb, ifaca,
3548                                                   NETLINK_CB(cb->skb).pid,
3549                                                   cb->nlh->nlmsg_seq,
3550                                                   RTM_GETANYCAST,
3551                                                   NLM_F_MULTI);
3552                         if (err <= 0)
3553                                 break;
3554                 }
3555                 break;
3556         default:
3557                 break;
3558         }
3559         read_unlock_bh(&idev->lock);
3560         *p_ip_idx = ip_idx;
3561         return err;
3562 }
3563
3564 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3565                            enum addr_type_t type)
3566 {
3567         struct net *net = sock_net(skb->sk);
3568         int h, s_h;
3569         int idx, ip_idx;
3570         int s_idx, s_ip_idx;
3571         struct net_device *dev;
3572         struct inet6_dev *idev;
3573         struct hlist_head *head;
3574         struct hlist_node *node;
3575
3576         s_h = cb->args[0];
3577         s_idx = idx = cb->args[1];
3578         s_ip_idx = ip_idx = cb->args[2];
3579
3580         rcu_read_lock();
3581         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3582                 idx = 0;
3583                 head = &net->dev_index_head[h];
3584                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3585                         if (idx < s_idx)
3586                                 goto cont;
3587                         if (idx > s_idx)
3588                                 s_ip_idx = 0;
3589                         ip_idx = 0;
3590                         if ((idev = __in6_dev_get(dev)) == NULL)
3591                                 goto cont;
3592
3593                         if (in6_dump_addrs(idev, skb, cb, type,
3594                                            s_ip_idx, &ip_idx) <= 0)
3595                                 goto done;
3596 cont:
3597                         idx++;
3598                 }
3599         }
3600 done:
3601         rcu_read_unlock();
3602         cb->args[0] = h;
3603         cb->args[1] = idx;
3604         cb->args[2] = ip_idx;
3605
3606         return skb->len;
3607 }
3608
3609 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3610 {
3611         enum addr_type_t type = UNICAST_ADDR;
3612
3613         return inet6_dump_addr(skb, cb, type);
3614 }
3615
3616 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3617 {
3618         enum addr_type_t type = MULTICAST_ADDR;
3619
3620         return inet6_dump_addr(skb, cb, type);
3621 }
3622
3623
3624 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3625 {
3626         enum addr_type_t type = ANYCAST_ADDR;
3627
3628         return inet6_dump_addr(skb, cb, type);
3629 }
3630
3631 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3632                              void *arg)
3633 {
3634         struct net *net = sock_net(in_skb->sk);
3635         struct ifaddrmsg *ifm;
3636         struct nlattr *tb[IFA_MAX+1];
3637         struct in6_addr *addr = NULL;
3638         struct net_device *dev = NULL;
3639         struct inet6_ifaddr *ifa;
3640         struct sk_buff *skb;
3641         int err;
3642
3643         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3644         if (err < 0)
3645                 goto errout;
3646
3647         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3648         if (addr == NULL) {
3649                 err = -EINVAL;
3650                 goto errout;
3651         }
3652
3653         ifm = nlmsg_data(nlh);
3654         if (ifm->ifa_index)
3655                 dev = __dev_get_by_index(net, ifm->ifa_index);
3656
3657         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3658                 err = -EADDRNOTAVAIL;
3659                 goto errout;
3660         }
3661
3662         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3663                 err = -ENOBUFS;
3664                 goto errout_ifa;
3665         }
3666
3667         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3668                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3669         if (err < 0) {
3670                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3671                 WARN_ON(err == -EMSGSIZE);
3672                 kfree_skb(skb);
3673                 goto errout_ifa;
3674         }
3675         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3676 errout_ifa:
3677         in6_ifa_put(ifa);
3678 errout:
3679         return err;
3680 }
3681
3682 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3683 {
3684         struct sk_buff *skb;
3685         struct net *net = dev_net(ifa->idev->dev);
3686         int err = -ENOBUFS;
3687
3688         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3689         if (skb == NULL)
3690                 goto errout;
3691
3692         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3693         if (err < 0) {
3694                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3695                 WARN_ON(err == -EMSGSIZE);
3696                 kfree_skb(skb);
3697                 goto errout;
3698         }
3699         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3700         return;
3701 errout:
3702         if (err < 0)
3703                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3704 }
3705
3706 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3707                                 __s32 *array, int bytes)
3708 {
3709         BUG_ON(bytes < (DEVCONF_MAX * 4));
3710
3711         memset(array, 0, bytes);
3712         array[DEVCONF_FORWARDING] = cnf->forwarding;
3713         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3714         array[DEVCONF_MTU6] = cnf->mtu6;
3715         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3716         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3717         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3718         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3719         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3720         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3721         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3722         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3723 #ifdef CONFIG_IPV6_PRIVACY
3724         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3725         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3726         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3727         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3728         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3729 #endif
3730         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3731         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3732         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3733 #ifdef CONFIG_IPV6_ROUTER_PREF
3734         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3735         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3736 #ifdef CONFIG_IPV6_ROUTE_INFO
3737         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3738 #endif
3739 #endif
3740         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3741         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3742 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3743         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3744 #endif
3745 #ifdef CONFIG_IPV6_MROUTE
3746         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3747 #endif
3748         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3749         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3750         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3751 }
3752
3753 static inline size_t inet6_if_nlmsg_size(void)
3754 {
3755         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3756                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3757                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3758                + nla_total_size(4) /* IFLA_MTU */
3759                + nla_total_size(4) /* IFLA_LINK */
3760                + nla_total_size( /* IFLA_PROTINFO */
3761                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3762                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3763                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3764                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3765                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3766                  );
3767 }
3768
3769 static inline void __snmp6_fill_stats(u64 *stats, void __percpu **mib,
3770                                       int items, int bytes)
3771 {
3772         int i;
3773         int pad = bytes - sizeof(u64) * items;
3774         BUG_ON(pad < 0);
3775
3776         /* Use put_unaligned() because stats may not be aligned for u64. */
3777         put_unaligned(items, &stats[0]);
3778         for (i = 1; i < items; i++)
3779                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3780
3781         memset(&stats[items], 0, pad);
3782 }
3783
3784 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3785                              int bytes)
3786 {
3787         switch(attrtype) {
3788         case IFLA_INET6_STATS:
3789                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3790                 break;
3791         case IFLA_INET6_ICMP6STATS:
3792                 __snmp6_fill_stats(stats, (void __percpu **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3793                 break;
3794         }
3795 }
3796
3797 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3798                              u32 pid, u32 seq, int event, unsigned int flags)
3799 {
3800         struct net_device *dev = idev->dev;
3801         struct nlattr *nla;
3802         struct ifinfomsg *hdr;
3803         struct nlmsghdr *nlh;
3804         void *protoinfo;
3805         struct ifla_cacheinfo ci;
3806
3807         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3808         if (nlh == NULL)
3809                 return -EMSGSIZE;
3810
3811         hdr = nlmsg_data(nlh);
3812         hdr->ifi_family = AF_INET6;
3813         hdr->__ifi_pad = 0;
3814         hdr->ifi_type = dev->type;
3815         hdr->ifi_index = dev->ifindex;
3816         hdr->ifi_flags = dev_get_flags(dev);
3817         hdr->ifi_change = 0;
3818
3819         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3820
3821         if (dev->addr_len)
3822                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3823
3824         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3825         if (dev->ifindex != dev->iflink)
3826                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3827
3828         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3829         if (protoinfo == NULL)
3830                 goto nla_put_failure;
3831
3832         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3833
3834         ci.max_reasm_len = IPV6_MAXPLEN;
3835         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3836                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3837         ci.reachable_time = idev->nd_parms->reachable_time;
3838         ci.retrans_time = idev->nd_parms->retrans_time;
3839         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3840
3841         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3842         if (nla == NULL)
3843                 goto nla_put_failure;
3844         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3845
3846         /* XXX - MC not implemented */
3847
3848         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3849         if (nla == NULL)
3850                 goto nla_put_failure;
3851         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3852
3853         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3854         if (nla == NULL)
3855                 goto nla_put_failure;
3856         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3857
3858         nla_nest_end(skb, protoinfo);
3859         return nlmsg_end(skb, nlh);
3860
3861 nla_put_failure:
3862         nlmsg_cancel(skb, nlh);
3863         return -EMSGSIZE;
3864 }
3865
3866 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3867 {
3868         struct net *net = sock_net(skb->sk);
3869         int h, s_h;
3870         int idx = 0, s_idx;
3871         struct net_device *dev;
3872         struct inet6_dev *idev;
3873         struct hlist_head *head;
3874         struct hlist_node *node;
3875
3876         s_h = cb->args[0];
3877         s_idx = cb->args[1];
3878
3879         rcu_read_lock();
3880         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3881                 idx = 0;
3882                 head = &net->dev_index_head[h];
3883                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3884                         if (idx < s_idx)
3885                                 goto cont;
3886                         idev = __in6_dev_get(dev);
3887                         if (!idev)
3888                                 goto cont;
3889                         if (inet6_fill_ifinfo(skb, idev,
3890                                               NETLINK_CB(cb->skb).pid,
3891                                               cb->nlh->nlmsg_seq,
3892                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3893                                 goto out;
3894 cont:
3895                         idx++;
3896                 }
3897         }
3898 out:
3899         rcu_read_unlock();
3900         cb->args[1] = idx;
3901         cb->args[0] = h;
3902
3903         return skb->len;
3904 }
3905
3906 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3907 {
3908         struct sk_buff *skb;
3909         struct net *net = dev_net(idev->dev);
3910         int err = -ENOBUFS;
3911
3912         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3913         if (skb == NULL)
3914                 goto errout;
3915
3916         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3917         if (err < 0) {
3918                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3919                 WARN_ON(err == -EMSGSIZE);
3920                 kfree_skb(skb);
3921                 goto errout;
3922         }
3923         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3924         return;
3925 errout:
3926         if (err < 0)
3927                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3928 }
3929
3930 static inline size_t inet6_prefix_nlmsg_size(void)
3931 {
3932         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3933                + nla_total_size(sizeof(struct in6_addr))
3934                + nla_total_size(sizeof(struct prefix_cacheinfo));
3935 }
3936
3937 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3938                              struct prefix_info *pinfo, u32 pid, u32 seq,
3939                              int event, unsigned int flags)
3940 {
3941         struct prefixmsg *pmsg;
3942         struct nlmsghdr *nlh;
3943         struct prefix_cacheinfo ci;
3944
3945         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3946         if (nlh == NULL)
3947                 return -EMSGSIZE;
3948
3949         pmsg = nlmsg_data(nlh);
3950         pmsg->prefix_family = AF_INET6;
3951         pmsg->prefix_pad1 = 0;
3952         pmsg->prefix_pad2 = 0;
3953         pmsg->prefix_ifindex = idev->dev->ifindex;
3954         pmsg->prefix_len = pinfo->prefix_len;
3955         pmsg->prefix_type = pinfo->type;
3956         pmsg->prefix_pad3 = 0;
3957         pmsg->prefix_flags = 0;
3958         if (pinfo->onlink)
3959                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3960         if (pinfo->autoconf)
3961                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3962
3963         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3964
3965         ci.preferred_time = ntohl(pinfo->prefered);
3966         ci.valid_time = ntohl(pinfo->valid);
3967         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3968
3969         return nlmsg_end(skb, nlh);
3970
3971 nla_put_failure:
3972         nlmsg_cancel(skb, nlh);
3973         return -EMSGSIZE;
3974 }
3975
3976 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3977                          struct prefix_info *pinfo)
3978 {
3979         struct sk_buff *skb;
3980         struct net *net = dev_net(idev->dev);
3981         int err = -ENOBUFS;
3982
3983         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3984         if (skb == NULL)
3985                 goto errout;
3986
3987         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3988         if (err < 0) {
3989                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3990                 WARN_ON(err == -EMSGSIZE);
3991                 kfree_skb(skb);
3992                 goto errout;
3993         }
3994         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3995         return;
3996 errout:
3997         if (err < 0)
3998                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3999 }
4000
4001 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4002 {
4003         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4004
4005         switch (event) {
4006         case RTM_NEWADDR:
4007                 /*
4008                  * If the address was optimistic
4009                  * we inserted the route at the start of
4010                  * our DAD process, so we don't need
4011                  * to do it again
4012                  */
4013                 if (!(ifp->rt->rt6i_node))
4014                         ip6_ins_rt(ifp->rt);
4015                 if (ifp->idev->cnf.forwarding)
4016                         addrconf_join_anycast(ifp);
4017                 break;
4018         case RTM_DELADDR:
4019                 if (ifp->idev->cnf.forwarding)
4020                         addrconf_leave_anycast(ifp);
4021                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4022                 dst_hold(&ifp->rt->u.dst);
4023                 if (ip6_del_rt(ifp->rt))
4024                         dst_free(&ifp->rt->u.dst);
4025                 break;
4026         }
4027 }
4028
4029 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4030 {
4031         rcu_read_lock_bh();
4032         if (likely(ifp->idev->dead == 0))
4033                 __ipv6_ifa_notify(event, ifp);
4034         rcu_read_unlock_bh();
4035 }
4036
4037 #ifdef CONFIG_SYSCTL
4038
4039 static
4040 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4041                            void __user *buffer, size_t *lenp, loff_t *ppos)
4042 {
4043         int *valp = ctl->data;
4044         int val = *valp;
4045         loff_t pos = *ppos;
4046         int ret;
4047
4048         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4049
4050         if (write)
4051                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4052         if (ret)
4053                 *ppos = pos;
4054         return ret;
4055 }
4056
4057 static void dev_disable_change(struct inet6_dev *idev)
4058 {
4059         if (!idev || !idev->dev)
4060                 return;
4061
4062         if (idev->cnf.disable_ipv6)
4063                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4064         else
4065                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4066 }
4067
4068 static void addrconf_disable_change(struct net *net, __s32 newf)
4069 {
4070         struct net_device *dev;
4071         struct inet6_dev *idev;
4072
4073         rcu_read_lock();
4074         for_each_netdev_rcu(net, dev) {
4075                 idev = __in6_dev_get(dev);
4076                 if (idev) {
4077                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4078                         idev->cnf.disable_ipv6 = newf;
4079                         if (changed)
4080                                 dev_disable_change(idev);
4081                 }
4082         }
4083         rcu_read_unlock();
4084 }
4085
4086 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4087 {
4088         struct net *net;
4089
4090         net = (struct net *)table->extra2;
4091
4092         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4093                 return 0;
4094
4095         if (!rtnl_trylock()) {
4096                 /* Restore the original values before restarting */
4097                 *p = old;
4098                 return restart_syscall();
4099         }
4100
4101         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4102                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4103                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4104                 addrconf_disable_change(net, newf);
4105         } else if ((!*p) ^ (!old))
4106                 dev_disable_change((struct inet6_dev *)table->extra1);
4107
4108         rtnl_unlock();
4109         return 0;
4110 }
4111
4112 static
4113 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4114                             void __user *buffer, size_t *lenp, loff_t *ppos)
4115 {
4116         int *valp = ctl->data;
4117         int val = *valp;
4118         loff_t pos = *ppos;
4119         int ret;
4120
4121         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4122
4123         if (write)
4124                 ret = addrconf_disable_ipv6(ctl, valp, val);
4125         if (ret)
4126                 *ppos = pos;
4127         return ret;
4128 }
4129
4130 static struct addrconf_sysctl_table
4131 {
4132         struct ctl_table_header *sysctl_header;
4133         ctl_table addrconf_vars[DEVCONF_MAX+1];
4134         char *dev_name;
4135 } addrconf_sysctl __read_mostly = {
4136         .sysctl_header = NULL,
4137         .addrconf_vars = {
4138                 {
4139                         .procname       =       "forwarding",
4140                         .data           =       &ipv6_devconf.forwarding,
4141                         .maxlen         =       sizeof(int),
4142                         .mode           =       0644,
4143                         .proc_handler   =       addrconf_sysctl_forward,
4144                 },
4145                 {
4146                         .procname       =       "hop_limit",
4147                         .data           =       &ipv6_devconf.hop_limit,
4148                         .maxlen         =       sizeof(int),
4149                         .mode           =       0644,
4150                         .proc_handler   =       proc_dointvec,
4151                 },
4152                 {
4153                         .procname       =       "mtu",
4154                         .data           =       &ipv6_devconf.mtu6,
4155                         .maxlen         =       sizeof(int),
4156                         .mode           =       0644,
4157                         .proc_handler   =       proc_dointvec,
4158                 },
4159                 {
4160                         .procname       =       "accept_ra",
4161                         .data           =       &ipv6_devconf.accept_ra,
4162                         .maxlen         =       sizeof(int),
4163                         .mode           =       0644,
4164                         .proc_handler   =       proc_dointvec,
4165                 },
4166                 {
4167                         .procname       =       "accept_redirects",
4168                         .data           =       &ipv6_devconf.accept_redirects,
4169                         .maxlen         =       sizeof(int),
4170                         .mode           =       0644,
4171                         .proc_handler   =       proc_dointvec,
4172                 },
4173                 {
4174                         .procname       =       "autoconf",
4175                         .data           =       &ipv6_devconf.autoconf,
4176                         .maxlen         =       sizeof(int),
4177                         .mode           =       0644,
4178                         .proc_handler   =       proc_dointvec,
4179                 },
4180                 {
4181                         .procname       =       "dad_transmits",
4182                         .data           =       &ipv6_devconf.dad_transmits,
4183                         .maxlen         =       sizeof(int),
4184                         .mode           =       0644,
4185                         .proc_handler   =       proc_dointvec,
4186                 },
4187                 {
4188                         .procname       =       "router_solicitations",
4189                         .data           =       &ipv6_devconf.rtr_solicits,
4190                         .maxlen         =       sizeof(int),
4191                         .mode           =       0644,
4192                         .proc_handler   =       proc_dointvec,
4193                 },
4194                 {
4195                         .procname       =       "router_solicitation_interval",
4196                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4197                         .maxlen         =       sizeof(int),
4198                         .mode           =       0644,
4199                         .proc_handler   =       proc_dointvec_jiffies,
4200                 },
4201                 {
4202                         .procname       =       "router_solicitation_delay",
4203                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4204                         .maxlen         =       sizeof(int),
4205                         .mode           =       0644,
4206                         .proc_handler   =       proc_dointvec_jiffies,
4207                 },
4208                 {
4209                         .procname       =       "force_mld_version",
4210                         .data           =       &ipv6_devconf.force_mld_version,
4211                         .maxlen         =       sizeof(int),
4212                         .mode           =       0644,
4213                         .proc_handler   =       proc_dointvec,
4214                 },
4215 #ifdef CONFIG_IPV6_PRIVACY
4216                 {
4217                         .procname       =       "use_tempaddr",
4218                         .data           =       &ipv6_devconf.use_tempaddr,
4219                         .maxlen         =       sizeof(int),
4220                         .mode           =       0644,
4221                         .proc_handler   =       proc_dointvec,
4222                 },
4223                 {
4224                         .procname       =       "temp_valid_lft",
4225                         .data           =       &ipv6_devconf.temp_valid_lft,
4226                         .maxlen         =       sizeof(int),
4227                         .mode           =       0644,
4228                         .proc_handler   =       proc_dointvec,
4229                 },
4230                 {
4231                         .procname       =       "temp_prefered_lft",
4232                         .data           =       &ipv6_devconf.temp_prefered_lft,
4233                         .maxlen         =       sizeof(int),
4234                         .mode           =       0644,
4235                         .proc_handler   =       proc_dointvec,
4236                 },
4237                 {
4238                         .procname       =       "regen_max_retry",
4239                         .data           =       &ipv6_devconf.regen_max_retry,
4240                         .maxlen         =       sizeof(int),
4241                         .mode           =       0644,
4242                         .proc_handler   =       proc_dointvec,
4243                 },
4244                 {
4245                         .procname       =       "max_desync_factor",
4246                         .data           =       &ipv6_devconf.max_desync_factor,
4247                         .maxlen         =       sizeof(int),
4248                         .mode           =       0644,
4249                         .proc_handler   =       proc_dointvec,
4250                 },
4251 #endif
4252                 {
4253                         .procname       =       "max_addresses",
4254                         .data           =       &ipv6_devconf.max_addresses,
4255                         .maxlen         =       sizeof(int),
4256                         .mode           =       0644,
4257                         .proc_handler   =       proc_dointvec,
4258                 },
4259                 {
4260                         .procname       =       "accept_ra_defrtr",
4261                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4262                         .maxlen         =       sizeof(int),
4263                         .mode           =       0644,
4264                         .proc_handler   =       proc_dointvec,
4265                 },
4266                 {
4267                         .procname       =       "accept_ra_pinfo",
4268                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4269                         .maxlen         =       sizeof(int),
4270                         .mode           =       0644,
4271                         .proc_handler   =       proc_dointvec,
4272                 },
4273 #ifdef CONFIG_IPV6_ROUTER_PREF
4274                 {
4275                         .procname       =       "accept_ra_rtr_pref",
4276                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4277                         .maxlen         =       sizeof(int),
4278                         .mode           =       0644,
4279                         .proc_handler   =       proc_dointvec,
4280                 },
4281                 {
4282                         .procname       =       "router_probe_interval",
4283                         .data           =       &ipv6_devconf.rtr_probe_interval,
4284                         .maxlen         =       sizeof(int),
4285                         .mode           =       0644,
4286                         .proc_handler   =       proc_dointvec_jiffies,
4287                 },
4288 #ifdef CONFIG_IPV6_ROUTE_INFO
4289                 {
4290                         .procname       =       "accept_ra_rt_info_max_plen",
4291                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4292                         .maxlen         =       sizeof(int),
4293                         .mode           =       0644,
4294                         .proc_handler   =       proc_dointvec,
4295                 },
4296 #endif
4297 #endif
4298                 {
4299                         .procname       =       "proxy_ndp",
4300                         .data           =       &ipv6_devconf.proxy_ndp,
4301                         .maxlen         =       sizeof(int),
4302                         .mode           =       0644,
4303                         .proc_handler   =       proc_dointvec,
4304                 },
4305                 {
4306                         .procname       =       "accept_source_route",
4307                         .data           =       &ipv6_devconf.accept_source_route,
4308                         .maxlen         =       sizeof(int),
4309                         .mode           =       0644,
4310                         .proc_handler   =       proc_dointvec,
4311                 },
4312 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4313                 {
4314                         .procname       =       "optimistic_dad",
4315                         .data           =       &ipv6_devconf.optimistic_dad,
4316                         .maxlen         =       sizeof(int),
4317                         .mode           =       0644,
4318                         .proc_handler   =       proc_dointvec,
4319
4320                 },
4321 #endif
4322 #ifdef CONFIG_IPV6_MROUTE
4323                 {
4324                         .procname       =       "mc_forwarding",
4325                         .data           =       &ipv6_devconf.mc_forwarding,
4326                         .maxlen         =       sizeof(int),
4327                         .mode           =       0444,
4328                         .proc_handler   =       proc_dointvec,
4329                 },
4330 #endif
4331                 {
4332                         .procname       =       "disable_ipv6",
4333                         .data           =       &ipv6_devconf.disable_ipv6,
4334                         .maxlen         =       sizeof(int),
4335                         .mode           =       0644,
4336                         .proc_handler   =       addrconf_sysctl_disable,
4337                 },
4338                 {
4339                         .procname       =       "accept_dad",
4340                         .data           =       &ipv6_devconf.accept_dad,
4341                         .maxlen         =       sizeof(int),
4342                         .mode           =       0644,
4343                         .proc_handler   =       proc_dointvec,
4344                 },
4345                 {
4346                         .procname       = "force_tllao",
4347                         .data           = &ipv6_devconf.force_tllao,
4348                         .maxlen         = sizeof(int),
4349                         .mode           = 0644,
4350                         .proc_handler   = proc_dointvec
4351                 },
4352                 {
4353                         /* sentinel */
4354                 }
4355         },
4356 };
4357
4358 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4359                 struct inet6_dev *idev, struct ipv6_devconf *p)
4360 {
4361         int i;
4362         struct addrconf_sysctl_table *t;
4363
4364 #define ADDRCONF_CTL_PATH_DEV   3
4365
4366         struct ctl_path addrconf_ctl_path[] = {
4367                 { .procname = "net", },
4368                 { .procname = "ipv6", },
4369                 { .procname = "conf", },
4370                 { /* to be set */ },
4371                 { },
4372         };
4373
4374
4375         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4376         if (t == NULL)
4377                 goto out;
4378
4379         for (i=0; t->addrconf_vars[i].data; i++) {
4380                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4381                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4382                 t->addrconf_vars[i].extra2 = net;
4383         }
4384
4385         /*
4386          * Make a copy of dev_name, because '.procname' is regarded as const
4387          * by sysctl and we wouldn't want anyone to change it under our feet
4388          * (see SIOCSIFNAME).
4389          */
4390         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4391         if (!t->dev_name)
4392                 goto free;
4393
4394         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4395
4396         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4397                         t->addrconf_vars);
4398         if (t->sysctl_header == NULL)
4399                 goto free_procname;
4400
4401         p->sysctl = t;
4402         return 0;
4403
4404 free_procname:
4405         kfree(t->dev_name);
4406 free:
4407         kfree(t);
4408 out:
4409         return -ENOBUFS;
4410 }
4411
4412 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4413 {
4414         struct addrconf_sysctl_table *t;
4415
4416         if (p->sysctl == NULL)
4417                 return;
4418
4419         t = p->sysctl;
4420         p->sysctl = NULL;
4421         unregister_sysctl_table(t->sysctl_header);
4422         kfree(t->dev_name);
4423         kfree(t);
4424 }
4425
4426 static void addrconf_sysctl_register(struct inet6_dev *idev)
4427 {
4428         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4429                               &ndisc_ifinfo_sysctl_change);
4430         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4431                                         idev, &idev->cnf);
4432 }
4433
4434 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4435 {
4436         __addrconf_sysctl_unregister(&idev->cnf);
4437         neigh_sysctl_unregister(idev->nd_parms);
4438 }
4439
4440
4441 #endif
4442
4443 static int __net_init addrconf_init_net(struct net *net)
4444 {
4445         int err;
4446         struct ipv6_devconf *all, *dflt;
4447
4448         err = -ENOMEM;
4449         all = &ipv6_devconf;
4450         dflt = &ipv6_devconf_dflt;
4451
4452         if (!net_eq(net, &init_net)) {
4453                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4454                 if (all == NULL)
4455                         goto err_alloc_all;
4456
4457                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4458                 if (dflt == NULL)
4459                         goto err_alloc_dflt;
4460         } else {
4461                 /* these will be inherited by all namespaces */
4462                 dflt->autoconf = ipv6_defaults.autoconf;
4463                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4464         }
4465
4466         net->ipv6.devconf_all = all;
4467         net->ipv6.devconf_dflt = dflt;
4468
4469 #ifdef CONFIG_SYSCTL
4470         err = __addrconf_sysctl_register(net, "all", NULL, all);
4471         if (err < 0)
4472                 goto err_reg_all;
4473
4474         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4475         if (err < 0)
4476                 goto err_reg_dflt;
4477 #endif
4478         return 0;
4479
4480 #ifdef CONFIG_SYSCTL
4481 err_reg_dflt:
4482         __addrconf_sysctl_unregister(all);
4483 err_reg_all:
4484         kfree(dflt);
4485 #endif
4486 err_alloc_dflt:
4487         kfree(all);
4488 err_alloc_all:
4489         return err;
4490 }
4491
4492 static void __net_exit addrconf_exit_net(struct net *net)
4493 {
4494 #ifdef CONFIG_SYSCTL
4495         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4496         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4497 #endif
4498         if (!net_eq(net, &init_net)) {
4499                 kfree(net->ipv6.devconf_dflt);
4500                 kfree(net->ipv6.devconf_all);
4501         }
4502 }
4503
4504 static struct pernet_operations addrconf_ops = {
4505         .init = addrconf_init_net,
4506         .exit = addrconf_exit_net,
4507 };
4508
4509 /*
4510  *      Device notifier
4511  */
4512
4513 int register_inet6addr_notifier(struct notifier_block *nb)
4514 {
4515         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4516 }
4517
4518 EXPORT_SYMBOL(register_inet6addr_notifier);
4519
4520 int unregister_inet6addr_notifier(struct notifier_block *nb)
4521 {
4522         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4523 }
4524
4525 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4526
4527 /*
4528  *      Init / cleanup code
4529  */
4530
4531 int __init addrconf_init(void)
4532 {
4533         int i, err;
4534
4535         if ((err = ipv6_addr_label_init()) < 0) {
4536                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4537                         err);
4538                 return err;
4539         }
4540
4541         register_pernet_subsys(&addrconf_ops);
4542
4543         /* The addrconf netdev notifier requires that loopback_dev
4544          * has it's ipv6 private information allocated and setup
4545          * before it can bring up and give link-local addresses
4546          * to other devices which are up.
4547          *
4548          * Unfortunately, loopback_dev is not necessarily the first
4549          * entry in the global dev_base list of net devices.  In fact,
4550          * it is likely to be the very last entry on that list.
4551          * So this causes the notifier registry below to try and
4552          * give link-local addresses to all devices besides loopback_dev
4553          * first, then loopback_dev, which cases all the non-loopback_dev
4554          * devices to fail to get a link-local address.
4555          *
4556          * So, as a temporary fix, allocate the ipv6 structure for
4557          * loopback_dev first by hand.
4558          * Longer term, all of the dependencies ipv6 has upon the loopback
4559          * device and it being up should be removed.
4560          */
4561         rtnl_lock();
4562         if (!ipv6_add_dev(init_net.loopback_dev))
4563                 err = -ENOMEM;
4564         rtnl_unlock();
4565         if (err)
4566                 goto errlo;
4567
4568         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4569                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4570
4571         register_netdevice_notifier(&ipv6_dev_notf);
4572
4573         addrconf_verify(0);
4574
4575         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4576         if (err < 0)
4577                 goto errout;
4578
4579         /* Only the first call to __rtnl_register can fail */
4580         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4581         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4582         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4583         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4584         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4585
4586         ipv6_addr_label_rtnl_register();
4587
4588         return 0;
4589 errout:
4590         unregister_netdevice_notifier(&ipv6_dev_notf);
4591 errlo:
4592         unregister_pernet_subsys(&addrconf_ops);
4593
4594         return err;
4595 }
4596
4597 void addrconf_cleanup(void)
4598 {
4599         struct net_device *dev;
4600         int i;
4601
4602         unregister_netdevice_notifier(&ipv6_dev_notf);
4603         unregister_pernet_subsys(&addrconf_ops);
4604
4605         rtnl_lock();
4606
4607         /* clean dev list */
4608         for_each_netdev(&init_net, dev) {
4609                 if (__in6_dev_get(dev) == NULL)
4610                         continue;
4611                 addrconf_ifdown(dev, 1);
4612         }
4613         addrconf_ifdown(init_net.loopback_dev, 2);
4614
4615         /*
4616          *      Check hash table.
4617          */
4618         spin_lock_bh(&addrconf_hash_lock);
4619         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4620                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4621         spin_unlock_bh(&addrconf_hash_lock);
4622
4623         del_timer(&addr_chk_timer);
4624         rtnl_unlock();
4625 }