net/ipv6/addrconf.c

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