net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         for_each_netdev(net, dev) {
 497                 idev = __in6_dev_get(dev);
 498                 if (idev) {
 499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 500                         idev->cnf.forwarding = newf;
 501                         if (changed)
 502                                 dev_forward_change(idev);
 503                 }
 504         }
 505 }
 506
 507 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 508 {
 509         struct net *net;
 510
 511         net = (struct net *)table->extra2;
 512         if (p == &net->ipv6.devconf_dflt->forwarding)
 513                 return 0;
 514
 515         if (!rtnl_trylock()) {
 516                 /* Restore the original values before restarting */
 517                 *p = old;
 518                 return restart_syscall();
 519         }
 520
 521         if (p == &net->ipv6.devconf_all->forwarding) {
 522                 __s32 newf = net->ipv6.devconf_all->forwarding;
 523                 net->ipv6.devconf_dflt->forwarding = newf;
 524                 addrconf_forward_change(net, newf);
 525         } else if ((!*p) ^ (!old))
 526                 dev_forward_change((struct inet6_dev *)table->extra1);
 527         rtnl_unlock();
 528
 529         if (*p)
 530                 rt6_purge_dflt_routers(net);
 531         return 1;
 532 }
 533 #endif
 534
 535 /* Nobody refers to this ifaddr, destroy it */
 536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 537 {
 538         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 539
 540 #ifdef NET_REFCNT_DEBUG
 541         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 542 #endif
 543
 544         in6_dev_put(ifp->idev);
 545
 546         if (del_timer(&ifp->timer))
 547                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 548
 549         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 550                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 551                 return;
 552         }
 553         dst_release(&ifp->rt->dst);
 554
 555         kfree_rcu(ifp, rcu);
 556 }
 557
 558 static void
 559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 560 {
 561         struct list_head *p;
 562         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 563
 564         /*
 565          * Each device address list is sorted in order of scope -
 566          * global before linklocal.
 567          */
 568         list_for_each(p, &idev->addr_list) {
 569                 struct inet6_ifaddr *ifa
 570                         = list_entry(p, struct inet6_ifaddr, if_list);
 571                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 572                         break;
 573         }
 574
 575         list_add_tail(&ifp->if_list, p);
 576 }
 577
 578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 579 {
 580         /*
 581          * We perform the hash function over the last 64 bits of the address
 582          * This will include the IEEE address token on links that support it.
 583          */
 584         return jhash_2words((__force u32)addr->s6_addr32[2],
 585                             (__force u32)addr->s6_addr32[3], 0)
 586                 & (IN6_ADDR_HSIZE - 1);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         unsigned int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         spin_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         spin_lock_init(&ifa->state_lock);
 645         init_timer(&ifa->timer);
 646         INIT_HLIST_NODE(&ifa->addr_lst);
 647         ifa->timer.data = (unsigned long) ifa;
 648         ifa->scope = scope;
 649         ifa->prefix_len = pfxlen;
 650         ifa->flags = flags | IFA_F_TENTATIVE;
 651         ifa->cstamp = ifa->tstamp = jiffies;
 652
 653         ifa->rt = rt;
 654
 655         /*
 656          * part one of RFC 4429, section 3.3
 657          * We should not configure an address as
 658          * optimistic if we do not yet know the link
 659          * layer address of our nexhop router
 660          */
 661
 662         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 663                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 664
 665         ifa->idev = idev;
 666         in6_dev_hold(idev);
 667         /* For caller */
 668         in6_ifa_hold(ifa);
 669
 670         /* Add to big hash table */
 671         hash = ipv6_addr_hash(addr);
 672
 673         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 674         spin_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 683                 in6_ifa_hold(ifa);
 684         }
 685 #endif
 686
 687         in6_ifa_hold(ifa);
 688         write_unlock(&idev->lock);
 689 out2:
 690         rcu_read_unlock_bh();
 691
 692         if (likely(err == 0))
 693                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 694         else {
 695                 kfree(ifa);
 696                 ifa = ERR_PTR(err);
 697         }
 698
 699         return ifa;
 700 out:
 701         spin_unlock(&addrconf_hash_lock);
 702         goto out2;
 703 }
 704
 705 /* This function wants to get referenced ifp and releases it before return */
 706
 707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 708 {
 709         struct inet6_ifaddr *ifa, *ifn;
 710         struct inet6_dev *idev = ifp->idev;
 711         int state;
 712         int deleted = 0, onlink = 0;
 713         unsigned long expires = jiffies;
 714
 715         spin_lock_bh(&ifp->state_lock);
 716         state = ifp->state;
 717         ifp->state = INET6_IFADDR_STATE_DEAD;
 718         spin_unlock_bh(&ifp->state_lock);
 719
 720         if (state == INET6_IFADDR_STATE_DEAD)
 721                 goto out;
 722
 723         spin_lock_bh(&addrconf_hash_lock);
 724         hlist_del_init_rcu(&ifp->addr_lst);
 725         spin_unlock_bh(&addrconf_hash_lock);
 726
 727         write_lock_bh(&idev->lock);
 728 #ifdef CONFIG_IPV6_PRIVACY
 729         if (ifp->flags&IFA_F_TEMPORARY) {
 730                 list_del(&ifp->tmp_list);
 731                 if (ifp->ifpub) {
 732                         in6_ifa_put(ifp->ifpub);
 733                         ifp->ifpub = NULL;
 734                 }
 735                 __in6_ifa_put(ifp);
 736         }
 737 #endif
 738
 739         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 740                 if (ifa == ifp) {
 741                         list_del_init(&ifp->if_list);
 742                         __in6_ifa_put(ifp);
 743
 744                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 745                                 break;
 746                         deleted = 1;
 747                         continue;
 748                 } else if (ifp->flags & IFA_F_PERMANENT) {
 749                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 750                                               ifp->prefix_len)) {
 751                                 if (ifa->flags & IFA_F_PERMANENT) {
 752                                         onlink = 1;
 753                                         if (deleted)
 754                                                 break;
 755                                 } else {
 756                                         unsigned long lifetime;
 757
 758                                         if (!onlink)
 759                                                 onlink = -1;
 760
 761                                         spin_lock(&ifa->lock);
 762
 763                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 764                                         /*
 765                                          * Note: Because this address is
 766                                          * not permanent, lifetime <
 767                                          * LONG_MAX / HZ here.
 768                                          */
 769                                         if (time_before(expires,
 770                                                         ifa->tstamp + lifetime * HZ))
 771                                                 expires = ifa->tstamp + lifetime * HZ;
 772                                         spin_unlock(&ifa->lock);
 773                                 }
 774                         }
 775                 }
 776         }
 777         write_unlock_bh(&idev->lock);
 778
 779         addrconf_del_timer(ifp);
 780
 781         ipv6_ifa_notify(RTM_DELADDR, ifp);
 782
 783         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 784
 785         /*
 786          * Purge or update corresponding prefix
 787          *
 788          * 1) we don't purge prefix here if address was not permanent.
 789          *    prefix is managed by its own lifetime.
 790          * 2) if there're no addresses, delete prefix.
 791          * 3) if there're still other permanent address(es),
 792          *    corresponding prefix is still permanent.
 793          * 4) otherwise, update prefix lifetime to the
 794          *    longest valid lifetime among the corresponding
 795          *    addresses on the device.
 796          *    Note: subsequent RA will update lifetime.
 797          *
 798          * --yoshfuji
 799          */
 800         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 801                 struct in6_addr prefix;
 802                 struct rt6_info *rt;
 803                 struct net *net = dev_net(ifp->idev->dev);
 804                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 805                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 806
 807                 if (rt && addrconf_is_prefix_route(rt)) {
 808                         if (onlink == 0) {
 809                                 ip6_del_rt(rt);
 810                                 rt = NULL;
 811                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 812                                 rt->rt6i_expires = expires;
 813                                 rt->rt6i_flags |= RTF_EXPIRES;
 814                         }
 815                 }
 816                 dst_release(&rt->dst);
 817         }
 818
 819         /* clean up prefsrc entries */
 820         rt6_remove_prefsrc(ifp);
 821 out:
 822         in6_ifa_put(ifp);
 823 }
 824
 825 #ifdef CONFIG_IPV6_PRIVACY
 826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 827 {
 828         struct inet6_dev *idev = ifp->idev;
 829         struct in6_addr addr, *tmpaddr;
 830         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 831         unsigned long regen_advance;
 832         int tmp_plen;
 833         int ret = 0;
 834         int max_addresses;
 835         u32 addr_flags;
 836         unsigned long now = jiffies;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         age = (now - ifp->tstamp) / HZ;
 882         tmp_valid_lft = min_t(__u32,
 883                               ifp->valid_lft,
 884                               idev->cnf.temp_valid_lft + age);
 885         tmp_prefered_lft = min_t(__u32,
 886                                  ifp->prefered_lft,
 887                                  idev->cnf.temp_prefered_lft + age -
 888                                  idev->cnf.max_desync_factor);
 889         tmp_plen = ifp->prefix_len;
 890         max_addresses = idev->cnf.max_addresses;
 891         tmp_tstamp = ifp->tstamp;
 892         spin_unlock_bh(&ifp->lock);
 893
 894         regen_advance = idev->cnf.regen_max_retry *
 895                         idev->cnf.dad_transmits *
 896                         idev->nd_parms->retrans_time / HZ;
 897         write_unlock(&idev->lock);
 898
 899         /* A temporary address is created only if this calculated Preferred
 900          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 901          * an implementation must not create a temporary address with a zero
 902          * Preferred Lifetime.
 903          */
 904         if (tmp_prefered_lft <= regen_advance) {
 905                 in6_ifa_put(ifp);
 906                 in6_dev_put(idev);
 907                 ret = -1;
 908                 goto out;
 909         }
 910
 911         addr_flags = IFA_F_TEMPORARY;
 912         /* set in addrconf_prefix_rcv() */
 913         if (ifp->flags & IFA_F_OPTIMISTIC)
 914                 addr_flags |= IFA_F_OPTIMISTIC;
 915
 916         ift = !max_addresses ||
 917               ipv6_count_addresses(idev) < max_addresses ?
 918                 ipv6_add_addr(idev, &addr, tmp_plen,
 919                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 920                               addr_flags) : NULL;
 921         if (!ift || IS_ERR(ift)) {
 922                 in6_ifa_put(ifp);
 923                 in6_dev_put(idev);
 924                 printk(KERN_INFO
 925                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 926                 tmpaddr = &addr;
 927                 write_lock(&idev->lock);
 928                 goto retry;
 929         }
 930
 931         spin_lock_bh(&ift->lock);
 932         ift->ifpub = ifp;
 933         ift->valid_lft = tmp_valid_lft;
 934         ift->prefered_lft = tmp_prefered_lft;
 935         ift->cstamp = now;
 936         ift->tstamp = tmp_tstamp;
 937         spin_unlock_bh(&ift->lock);
 938
 939         addrconf_dad_start(ift, 0);
 940         in6_ifa_put(ift);
 941         in6_dev_put(idev);
 942 out:
 943         return ret;
 944 }
 945 #endif
 946
 947 /*
 948  *      Choose an appropriate source address (RFC3484)
 949  */
 950 enum {
 951         IPV6_SADDR_RULE_INIT = 0,
 952         IPV6_SADDR_RULE_LOCAL,
 953         IPV6_SADDR_RULE_SCOPE,
 954         IPV6_SADDR_RULE_PREFERRED,
 955 #ifdef CONFIG_IPV6_MIP6
 956         IPV6_SADDR_RULE_HOA,
 957 #endif
 958         IPV6_SADDR_RULE_OIF,
 959         IPV6_SADDR_RULE_LABEL,
 960 #ifdef CONFIG_IPV6_PRIVACY
 961         IPV6_SADDR_RULE_PRIVACY,
 962 #endif
 963         IPV6_SADDR_RULE_ORCHID,
 964         IPV6_SADDR_RULE_PREFIX,
 965         IPV6_SADDR_RULE_MAX
 966 };
 967
 968 struct ipv6_saddr_score {
 969         int                     rule;
 970         int                     addr_type;
 971         struct inet6_ifaddr     *ifa;
 972         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 973         int                     scopedist;
 974         int                     matchlen;
 975 };
 976
 977 struct ipv6_saddr_dst {
 978         const struct in6_addr *addr;
 979         int ifindex;
 980         int scope;
 981         int label;
 982         unsigned int prefs;
 983 };
 984
 985 static inline int ipv6_saddr_preferred(int type)
 986 {
 987         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 988                 return 1;
 989         return 0;
 990 }
 991
 992 static int ipv6_get_saddr_eval(struct net *net,
 993                                struct ipv6_saddr_score *score,
 994                                struct ipv6_saddr_dst *dst,
 995                                int i)
 996 {
 997         int ret;
 998
 999         if (i <= score->rule) {
1000                 switch (i) {
1001                 case IPV6_SADDR_RULE_SCOPE:
1002                         ret = score->scopedist;
1003                         break;
1004                 case IPV6_SADDR_RULE_PREFIX:
1005                         ret = score->matchlen;
1006                         break;
1007                 default:
1008                         ret = !!test_bit(i, score->scorebits);
1009                 }
1010                 goto out;
1011         }
1012
1013         switch (i) {
1014         case IPV6_SADDR_RULE_INIT:
1015                 /* Rule 0: remember if hiscore is not ready yet */
1016                 ret = !!score->ifa;
1017                 break;
1018         case IPV6_SADDR_RULE_LOCAL:
1019                 /* Rule 1: Prefer same address */
1020                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1021                 break;
1022         case IPV6_SADDR_RULE_SCOPE:
1023                 /* Rule 2: Prefer appropriate scope
1024                  *
1025                  *      ret
1026                  *       ^
1027                  *    -1 |  d 15
1028                  *    ---+--+-+---> scope
1029                  *       |
1030                  *       |             d is scope of the destination.
1031                  *  B-d  |  \
1032                  *       |   \      <- smaller scope is better if
1033                  *  B-15 |    \        if scope is enough for destinaion.
1034                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1035                  * d-C-1 | /
1036                  *       |/         <- greater is better
1037                  *   -C  /             if scope is not enough for destination.
1038                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1039                  *
1040                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1041                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1042                  * Assume B = 0 and we get C > 29.
1043                  */
1044                 ret = __ipv6_addr_src_scope(score->addr_type);
1045                 if (ret >= dst->scope)
1046                         ret = -ret;
1047                 else
1048                         ret -= 128;     /* 30 is enough */
1049                 score->scopedist = ret;
1050                 break;
1051         case IPV6_SADDR_RULE_PREFERRED:
1052                 /* Rule 3: Avoid deprecated and optimistic addresses */
1053                 ret = ipv6_saddr_preferred(score->addr_type) ||
1054                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1055                 break;
1056 #ifdef CONFIG_IPV6_MIP6
1057         case IPV6_SADDR_RULE_HOA:
1058             {
1059                 /* Rule 4: Prefer home address */
1060                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1061                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1062                 break;
1063             }
1064 #endif
1065         case IPV6_SADDR_RULE_OIF:
1066                 /* Rule 5: Prefer outgoing interface */
1067                 ret = (!dst->ifindex ||
1068                        dst->ifindex == score->ifa->idev->dev->ifindex);
1069                 break;
1070         case IPV6_SADDR_RULE_LABEL:
1071                 /* Rule 6: Prefer matching label */
1072                 ret = ipv6_addr_label(net,
1073                                       &score->ifa->addr, score->addr_type,
1074                                       score->ifa->idev->dev->ifindex) == dst->label;
1075                 break;
1076 #ifdef CONFIG_IPV6_PRIVACY
1077         case IPV6_SADDR_RULE_PRIVACY:
1078             {
1079                 /* Rule 7: Prefer public address
1080                  * Note: prefer temporary address if use_tempaddr >= 2
1081                  */
1082                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1083                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1084                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1085                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1086                 break;
1087             }
1088 #endif
1089         case IPV6_SADDR_RULE_ORCHID:
1090                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1091                  *          non-ORCHID vs non-ORCHID
1092                  */
1093                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1094                         ipv6_addr_orchid(dst->addr));
1095                 break;
1096         case IPV6_SADDR_RULE_PREFIX:
1097                 /* Rule 8: Use longest matching prefix */
1098                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1099                                                        dst->addr);
1100                 break;
1101         default:
1102                 ret = 0;
1103         }
1104
1105         if (ret)
1106                 __set_bit(i, score->scorebits);
1107         score->rule = i;
1108 out:
1109         return ret;
1110 }
1111
1112 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1113                        const struct in6_addr *daddr, unsigned int prefs,
1114                        struct in6_addr *saddr)
1115 {
1116         struct ipv6_saddr_score scores[2],
1117                                 *score = &scores[0], *hiscore = &scores[1];
1118         struct ipv6_saddr_dst dst;
1119         struct net_device *dev;
1120         int dst_type;
1121
1122         dst_type = __ipv6_addr_type(daddr);
1123         dst.addr = daddr;
1124         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1125         dst.scope = __ipv6_addr_src_scope(dst_type);
1126         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1127         dst.prefs = prefs;
1128
1129         hiscore->rule = -1;
1130         hiscore->ifa = NULL;
1131
1132         rcu_read_lock();
1133
1134         for_each_netdev_rcu(net, dev) {
1135                 struct inet6_dev *idev;
1136
1137                 /* Candidate Source Address (section 4)
1138                  *  - multicast and link-local destination address,
1139                  *    the set of candidate source address MUST only
1140                  *    include addresses assigned to interfaces
1141                  *    belonging to the same link as the outgoing
1142                  *    interface.
1143                  * (- For site-local destination addresses, the
1144                  *    set of candidate source addresses MUST only
1145                  *    include addresses assigned to interfaces
1146                  *    belonging to the same site as the outgoing
1147                  *    interface.)
1148                  */
1149                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1150                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1151                     dst.ifindex && dev->ifindex != dst.ifindex)
1152                         continue;
1153
1154                 idev = __in6_dev_get(dev);
1155                 if (!idev)
1156                         continue;
1157
1158                 read_lock_bh(&idev->lock);
1159                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1160                         int i;
1161
1162                         /*
1163                          * - Tentative Address (RFC2462 section 5.4)
1164                          *  - A tentative address is not considered
1165                          *    "assigned to an interface" in the traditional
1166                          *    sense, unless it is also flagged as optimistic.
1167                          * - Candidate Source Address (section 4)
1168                          *  - In any case, anycast addresses, multicast
1169                          *    addresses, and the unspecified address MUST
1170                          *    NOT be included in a candidate set.
1171                          */
1172                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1173                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1174                                 continue;
1175
1176                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1177
1178                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1179                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1180                                 LIMIT_NETDEBUG(KERN_DEBUG
1181                                                "ADDRCONF: unspecified / multicast address "
1182                                                "assigned as unicast address on %s",
1183                                                dev->name);
1184                                 continue;
1185                         }
1186
1187                         score->rule = -1;
1188                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1189
1190                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1191                                 int minihiscore, miniscore;
1192
1193                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1194                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1195
1196                                 if (minihiscore > miniscore) {
1197                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1198                                             score->scopedist > 0) {
1199                                                 /*
1200                                                  * special case:
1201                                                  * each remaining entry
1202                                                  * has too small (not enough)
1203                                                  * scope, because ifa entries
1204                                                  * are sorted by their scope
1205                                                  * values.
1206                                                  */
1207                                                 goto try_nextdev;
1208                                         }
1209                                         break;
1210                                 } else if (minihiscore < miniscore) {
1211                                         if (hiscore->ifa)
1212                                                 in6_ifa_put(hiscore->ifa);
1213
1214                                         in6_ifa_hold(score->ifa);
1215
1216                                         swap(hiscore, score);
1217
1218                                         /* restore our iterator */
1219                                         score->ifa = hiscore->ifa;
1220
1221                                         break;
1222                                 }
1223                         }
1224                 }
1225 try_nextdev:
1226                 read_unlock_bh(&idev->lock);
1227         }
1228         rcu_read_unlock();
1229
1230         if (!hiscore->ifa)
1231                 return -EADDRNOTAVAIL;
1232
1233         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1234         in6_ifa_put(hiscore->ifa);
1235         return 0;
1236 }
1237 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1238
1239 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1240                     unsigned char banned_flags)
1241 {
1242         struct inet6_dev *idev;
1243         int err = -EADDRNOTAVAIL;
1244
1245         rcu_read_lock();
1246         idev = __in6_dev_get(dev);
1247         if (idev) {
1248                 struct inet6_ifaddr *ifp;
1249
1250                 read_lock_bh(&idev->lock);
1251                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1252                         if (ifp->scope == IFA_LINK &&
1253                             !(ifp->flags & banned_flags)) {
1254                                 ipv6_addr_copy(addr, &ifp->addr);
1255                                 err = 0;
1256                                 break;
1257                         }
1258                 }
1259                 read_unlock_bh(&idev->lock);
1260         }
1261         rcu_read_unlock();
1262         return err;
1263 }
1264
1265 static int ipv6_count_addresses(struct inet6_dev *idev)
1266 {
1267         int cnt = 0;
1268         struct inet6_ifaddr *ifp;
1269
1270         read_lock_bh(&idev->lock);
1271         list_for_each_entry(ifp, &idev->addr_list, if_list)
1272                 cnt++;
1273         read_unlock_bh(&idev->lock);
1274         return cnt;
1275 }
1276
1277 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1278                   struct net_device *dev, int strict)
1279 {
1280         struct inet6_ifaddr *ifp;
1281         struct hlist_node *node;
1282         unsigned int hash = ipv6_addr_hash(addr);
1283
1284         rcu_read_lock_bh();
1285         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1286                 if (!net_eq(dev_net(ifp->idev->dev), net))
1287                         continue;
1288                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1289                     !(ifp->flags&IFA_F_TENTATIVE) &&
1290                     (dev == NULL || ifp->idev->dev == dev ||
1291                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1292                         rcu_read_unlock_bh();
1293                         return 1;
1294                 }
1295         }
1296
1297         rcu_read_unlock_bh();
1298         return 0;
1299 }
1300 EXPORT_SYMBOL(ipv6_chk_addr);
1301
1302 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1303                                struct net_device *dev)
1304 {
1305         unsigned int hash = ipv6_addr_hash(addr);
1306         struct inet6_ifaddr *ifp;
1307         struct hlist_node *node;
1308
1309         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1310                 if (!net_eq(dev_net(ifp->idev->dev), net))
1311                         continue;
1312                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1313                         if (dev == NULL || ifp->idev->dev == dev)
1314                                 return true;
1315                 }
1316         }
1317         return false;
1318 }
1319
1320 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1321 {
1322         struct inet6_dev *idev;
1323         struct inet6_ifaddr *ifa;
1324         int     onlink;
1325
1326         onlink = 0;
1327         rcu_read_lock();
1328         idev = __in6_dev_get(dev);
1329         if (idev) {
1330                 read_lock_bh(&idev->lock);
1331                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1332                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1333                                                    ifa->prefix_len);
1334                         if (onlink)
1335                                 break;
1336                 }
1337                 read_unlock_bh(&idev->lock);
1338         }
1339         rcu_read_unlock();
1340         return onlink;
1341 }
1342
1343 EXPORT_SYMBOL(ipv6_chk_prefix);
1344
1345 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1346                                      struct net_device *dev, int strict)
1347 {
1348         struct inet6_ifaddr *ifp, *result = NULL;
1349         unsigned int hash = ipv6_addr_hash(addr);
1350         struct hlist_node *node;
1351
1352         rcu_read_lock_bh();
1353         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1354                 if (!net_eq(dev_net(ifp->idev->dev), net))
1355                         continue;
1356                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1357                         if (dev == NULL || ifp->idev->dev == dev ||
1358                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1359                                 result = ifp;
1360                                 in6_ifa_hold(ifp);
1361                                 break;
1362                         }
1363                 }
1364         }
1365         rcu_read_unlock_bh();
1366
1367         return result;
1368 }
1369
1370 /* Gets referenced address, destroys ifaddr */
1371
1372 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1373 {
1374         if (ifp->flags&IFA_F_PERMANENT) {
1375                 spin_lock_bh(&ifp->lock);
1376                 addrconf_del_timer(ifp);
1377                 ifp->flags |= IFA_F_TENTATIVE;
1378                 if (dad_failed)
1379                         ifp->flags |= IFA_F_DADFAILED;
1380                 spin_unlock_bh(&ifp->lock);
1381                 if (dad_failed)
1382                         ipv6_ifa_notify(0, ifp);
1383                 in6_ifa_put(ifp);
1384 #ifdef CONFIG_IPV6_PRIVACY
1385         } else if (ifp->flags&IFA_F_TEMPORARY) {
1386                 struct inet6_ifaddr *ifpub;
1387                 spin_lock_bh(&ifp->lock);
1388                 ifpub = ifp->ifpub;
1389                 if (ifpub) {
1390                         in6_ifa_hold(ifpub);
1391                         spin_unlock_bh(&ifp->lock);
1392                         ipv6_create_tempaddr(ifpub, ifp);
1393                         in6_ifa_put(ifpub);
1394                 } else {
1395                         spin_unlock_bh(&ifp->lock);
1396                 }
1397                 ipv6_del_addr(ifp);
1398 #endif
1399         } else
1400                 ipv6_del_addr(ifp);
1401 }
1402
1403 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1404 {
1405         int err = -ENOENT;
1406
1407         spin_lock(&ifp->state_lock);
1408         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1409                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1410                 err = 0;
1411         }
1412         spin_unlock(&ifp->state_lock);
1413
1414         return err;
1415 }
1416
1417 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1418 {
1419         struct inet6_dev *idev = ifp->idev;
1420
1421         if (addrconf_dad_end(ifp)) {
1422                 in6_ifa_put(ifp);
1423                 return;
1424         }
1425
1426         if (net_ratelimit())
1427                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1428                         ifp->idev->dev->name, &ifp->addr);
1429
1430         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1431                 struct in6_addr addr;
1432
1433                 addr.s6_addr32[0] = htonl(0xfe800000);
1434                 addr.s6_addr32[1] = 0;
1435
1436                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1437                     ipv6_addr_equal(&ifp->addr, &addr)) {
1438                         /* DAD failed for link-local based on MAC address */
1439                         idev->cnf.disable_ipv6 = 1;
1440
1441                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1442                                 ifp->idev->dev->name);
1443                 }
1444         }
1445
1446         addrconf_dad_stop(ifp, 1);
1447 }
1448
1449 /* Join to solicited addr multicast group. */
1450
1451 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1452 {
1453         struct in6_addr maddr;
1454
1455         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1456                 return;
1457
1458         addrconf_addr_solict_mult(addr, &maddr);
1459         ipv6_dev_mc_inc(dev, &maddr);
1460 }
1461
1462 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1463 {
1464         struct in6_addr maddr;
1465
1466         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1467                 return;
1468
1469         addrconf_addr_solict_mult(addr, &maddr);
1470         __ipv6_dev_mc_dec(idev, &maddr);
1471 }
1472
1473 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1474 {
1475         struct in6_addr addr;
1476         if (ifp->prefix_len == 127) /* RFC 6164 */
1477                 return;
1478         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1479         if (ipv6_addr_any(&addr))
1480                 return;
1481         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1482 }
1483
1484 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1485 {
1486         struct in6_addr addr;
1487         if (ifp->prefix_len == 127) /* RFC 6164 */
1488                 return;
1489         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1490         if (ipv6_addr_any(&addr))
1491                 return;
1492         __ipv6_dev_ac_dec(ifp->idev, &addr);
1493 }
1494
1495 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1496 {
1497         if (dev->addr_len != ETH_ALEN)
1498                 return -1;
1499         memcpy(eui, dev->dev_addr, 3);
1500         memcpy(eui + 5, dev->dev_addr + 3, 3);
1501
1502         /*
1503          * The zSeries OSA network cards can be shared among various
1504          * OS instances, but the OSA cards have only one MAC address.
1505          * This leads to duplicate address conflicts in conjunction
1506          * with IPv6 if more than one instance uses the same card.
1507          *
1508          * The driver for these cards can deliver a unique 16-bit
1509          * identifier for each instance sharing the same card.  It is
1510          * placed instead of 0xFFFE in the interface identifier.  The
1511          * "u" bit of the interface identifier is not inverted in this
1512          * case.  Hence the resulting interface identifier has local
1513          * scope according to RFC2373.
1514          */
1515         if (dev->dev_id) {
1516                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1517                 eui[4] = dev->dev_id & 0xFF;
1518         } else {
1519                 eui[3] = 0xFF;
1520                 eui[4] = 0xFE;
1521                 eui[0] ^= 2;
1522         }
1523         return 0;
1524 }
1525
1526 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1527 {
1528         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1529         if (dev->addr_len != ARCNET_ALEN)
1530                 return -1;
1531         memset(eui, 0, 7);
1532         eui[7] = *(u8*)dev->dev_addr;
1533         return 0;
1534 }
1535
1536 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1537 {
1538         if (dev->addr_len != INFINIBAND_ALEN)
1539                 return -1;
1540         memcpy(eui, dev->dev_addr + 12, 8);
1541         eui[0] |= 2;
1542         return 0;
1543 }
1544
1545 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1546 {
1547         if (addr == 0)
1548                 return -1;
1549         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1550                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1551                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1552                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1553                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1554                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1555         eui[1] = 0;
1556         eui[2] = 0x5E;
1557         eui[3] = 0xFE;
1558         memcpy(eui + 4, &addr, 4);
1559         return 0;
1560 }
1561
1562 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1563 {
1564         if (dev->priv_flags & IFF_ISATAP)
1565                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1566         return -1;
1567 }
1568
1569 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1570 {
1571         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1572 }
1573
1574 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1575 {
1576         switch (dev->type) {
1577         case ARPHRD_ETHER:
1578         case ARPHRD_FDDI:
1579         case ARPHRD_IEEE802_TR:
1580                 return addrconf_ifid_eui48(eui, dev);
1581         case ARPHRD_ARCNET:
1582                 return addrconf_ifid_arcnet(eui, dev);
1583         case ARPHRD_INFINIBAND:
1584                 return addrconf_ifid_infiniband(eui, dev);
1585         case ARPHRD_SIT:
1586                 return addrconf_ifid_sit(eui, dev);
1587         case ARPHRD_IPGRE:
1588                 return addrconf_ifid_gre(eui, dev);
1589         }
1590         return -1;
1591 }
1592
1593 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1594 {
1595         int err = -1;
1596         struct inet6_ifaddr *ifp;
1597
1598         read_lock_bh(&idev->lock);
1599         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1600                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1601                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1602                         err = 0;
1603                         break;
1604                 }
1605         }
1606         read_unlock_bh(&idev->lock);
1607         return err;
1608 }
1609
1610 #ifdef CONFIG_IPV6_PRIVACY
1611 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1612 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1613 {
1614 regen:
1615         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1616         idev->rndid[0] &= ~0x02;
1617
1618         /*
1619          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1620          * check if generated address is not inappropriate
1621          *
1622          *  - Reserved subnet anycast (RFC 2526)
1623          *      11111101 11....11 1xxxxxxx
1624          *  - ISATAP (RFC4214) 6.1
1625          *      00-00-5E-FE-xx-xx-xx-xx
1626          *  - value 0
1627          *  - XXX: already assigned to an address on the device
1628          */
1629         if (idev->rndid[0] == 0xfd &&
1630             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1631             (idev->rndid[7]&0x80))
1632                 goto regen;
1633         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1634                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1635                         goto regen;
1636                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1637                         goto regen;
1638         }
1639
1640         return 0;
1641 }
1642
1643 static void ipv6_regen_rndid(unsigned long data)
1644 {
1645         struct inet6_dev *idev = (struct inet6_dev *) data;
1646         unsigned long expires;
1647
1648         rcu_read_lock_bh();
1649         write_lock_bh(&idev->lock);
1650
1651         if (idev->dead)
1652                 goto out;
1653
1654         if (__ipv6_regen_rndid(idev) < 0)
1655                 goto out;
1656
1657         expires = jiffies +
1658                 idev->cnf.temp_prefered_lft * HZ -
1659                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1660                 idev->cnf.max_desync_factor * HZ;
1661         if (time_before(expires, jiffies)) {
1662                 printk(KERN_WARNING
1663                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1664                         idev->dev->name);
1665                 goto out;
1666         }
1667
1668         if (!mod_timer(&idev->regen_timer, expires))
1669                 in6_dev_hold(idev);
1670
1671 out:
1672         write_unlock_bh(&idev->lock);
1673         rcu_read_unlock_bh();
1674         in6_dev_put(idev);
1675 }
1676
1677 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1678         int ret = 0;
1679
1680         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1681                 ret = __ipv6_regen_rndid(idev);
1682         return ret;
1683 }
1684 #endif
1685
1686 /*
1687  *      Add prefix route.
1688  */
1689
1690 static void
1691 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1692                       unsigned long expires, u32 flags)
1693 {
1694         struct fib6_config cfg = {
1695                 .fc_table = RT6_TABLE_PREFIX,
1696                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1697                 .fc_ifindex = dev->ifindex,
1698                 .fc_expires = expires,
1699                 .fc_dst_len = plen,
1700                 .fc_flags = RTF_UP | flags,
1701                 .fc_nlinfo.nl_net = dev_net(dev),
1702                 .fc_protocol = RTPROT_KERNEL,
1703         };
1704
1705         ipv6_addr_copy(&cfg.fc_dst, pfx);
1706
1707         /* Prevent useless cloning on PtP SIT.
1708            This thing is done here expecting that the whole
1709            class of non-broadcast devices need not cloning.
1710          */
1711 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1712         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1713                 cfg.fc_flags |= RTF_NONEXTHOP;
1714 #endif
1715
1716         ip6_route_add(&cfg);
1717 }
1718
1719
1720 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1721                                                   int plen,
1722                                                   const struct net_device *dev,
1723                                                   u32 flags, u32 noflags)
1724 {
1725         struct fib6_node *fn;
1726         struct rt6_info *rt = NULL;
1727         struct fib6_table *table;
1728
1729         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1730         if (table == NULL)
1731                 return NULL;
1732
1733         write_lock_bh(&table->tb6_lock);
1734         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1735         if (!fn)
1736                 goto out;
1737         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1738                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1739                         continue;
1740                 if ((rt->rt6i_flags & flags) != flags)
1741                         continue;
1742                 if ((noflags != 0) && ((rt->rt6i_flags & flags) != 0))
1743                         continue;
1744                 dst_hold(&rt->dst);
1745                 break;
1746         }
1747 out:
1748         write_unlock_bh(&table->tb6_lock);
1749         return rt;
1750 }
1751
1752
1753 /* Create "default" multicast route to the interface */
1754
1755 static void addrconf_add_mroute(struct net_device *dev)
1756 {
1757         struct fib6_config cfg = {
1758                 .fc_table = RT6_TABLE_LOCAL,
1759                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1760                 .fc_ifindex = dev->ifindex,
1761                 .fc_dst_len = 8,
1762                 .fc_flags = RTF_UP,
1763                 .fc_nlinfo.nl_net = dev_net(dev),
1764         };
1765
1766         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1767
1768         ip6_route_add(&cfg);
1769 }
1770
1771 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1772 static void sit_route_add(struct net_device *dev)
1773 {
1774         struct fib6_config cfg = {
1775                 .fc_table = RT6_TABLE_MAIN,
1776                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1777                 .fc_ifindex = dev->ifindex,
1778                 .fc_dst_len = 96,
1779                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1780                 .fc_nlinfo.nl_net = dev_net(dev),
1781         };
1782
1783         /* prefix length - 96 bits "::d.d.d.d" */
1784         ip6_route_add(&cfg);
1785 }
1786 #endif
1787
1788 static void addrconf_add_lroute(struct net_device *dev)
1789 {
1790         struct in6_addr addr;
1791
1792         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1793         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1794 }
1795
1796 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1797 {
1798         struct inet6_dev *idev;
1799
1800         ASSERT_RTNL();
1801
1802         idev = ipv6_find_idev(dev);
1803         if (!idev)
1804                 return ERR_PTR(-ENOBUFS);
1805
1806         if (idev->cnf.disable_ipv6)
1807                 return ERR_PTR(-EACCES);
1808
1809         /* Add default multicast route */
1810         if (!(dev->flags & IFF_LOOPBACK))
1811                 addrconf_add_mroute(dev);
1812
1813         /* Add link local route */
1814         addrconf_add_lroute(dev);
1815         return idev;
1816 }
1817
1818 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1819 {
1820         struct prefix_info *pinfo;
1821         __u32 valid_lft;
1822         __u32 prefered_lft;
1823         int addr_type;
1824         struct inet6_dev *in6_dev;
1825         struct net *net = dev_net(dev);
1826
1827         pinfo = (struct prefix_info *) opt;
1828
1829         if (len < sizeof(struct prefix_info)) {
1830                 ADBG(("addrconf: prefix option too short\n"));
1831                 return;
1832         }
1833
1834         /*
1835          *      Validation checks ([ADDRCONF], page 19)
1836          */
1837
1838         addr_type = ipv6_addr_type(&pinfo->prefix);
1839
1840         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1841                 return;
1842
1843         valid_lft = ntohl(pinfo->valid);
1844         prefered_lft = ntohl(pinfo->prefered);
1845
1846         if (prefered_lft > valid_lft) {
1847                 if (net_ratelimit())
1848                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1849                 return;
1850         }
1851
1852         in6_dev = in6_dev_get(dev);
1853
1854         if (in6_dev == NULL) {
1855                 if (net_ratelimit())
1856                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1857                 return;
1858         }
1859
1860         /*
1861          *      Two things going on here:
1862          *      1) Add routes for on-link prefixes
1863          *      2) Configure prefixes with the auto flag set
1864          */
1865
1866         if (pinfo->onlink) {
1867                 struct rt6_info *rt;
1868                 unsigned long rt_expires;
1869
1870                 /* Avoid arithmetic overflow. Really, we could
1871                  * save rt_expires in seconds, likely valid_lft,
1872                  * but it would require division in fib gc, that it
1873                  * not good.
1874                  */
1875                 if (HZ > USER_HZ)
1876                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1877                 else
1878                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1879
1880                 if (addrconf_finite_timeout(rt_expires))
1881                         rt_expires *= HZ;
1882
1883                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1884                                                pinfo->prefix_len,
1885                                                dev,
1886                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1887                                                RTF_GATEWAY | RTF_DEFAULT);
1888
1889                 if (rt) {
1890                         /* Autoconf prefix route */
1891                         if (valid_lft == 0) {
1892                                 ip6_del_rt(rt);
1893                                 rt = NULL;
1894                         } else if (addrconf_finite_timeout(rt_expires)) {
1895                                 /* not infinity */
1896                                 rt->rt6i_expires = jiffies + rt_expires;
1897                                 rt->rt6i_flags |= RTF_EXPIRES;
1898                         } else {
1899                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1900                                 rt->rt6i_expires = 0;
1901                         }
1902                 } else if (valid_lft) {
1903                         clock_t expires = 0;
1904                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1905                         if (addrconf_finite_timeout(rt_expires)) {
1906                                 /* not infinity */
1907                                 flags |= RTF_EXPIRES;
1908                                 expires = jiffies_to_clock_t(rt_expires);
1909                         }
1910                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1911                                               dev, expires, flags);
1912                 }
1913                 if (rt)
1914                         dst_release(&rt->dst);
1915         }
1916
1917         /* Try to figure out our local address for this prefix */
1918
1919         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1920                 struct inet6_ifaddr * ifp;
1921                 struct in6_addr addr;
1922                 int create = 0, update_lft = 0;
1923
1924                 if (pinfo->prefix_len == 64) {
1925                         memcpy(&addr, &pinfo->prefix, 8);
1926                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1927                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1928                                 in6_dev_put(in6_dev);
1929                                 return;
1930                         }
1931                         goto ok;
1932                 }
1933                 if (net_ratelimit())
1934                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1935                                pinfo->prefix_len);
1936                 in6_dev_put(in6_dev);
1937                 return;
1938
1939 ok:
1940
1941                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1942
1943                 if (ifp == NULL && valid_lft) {
1944                         int max_addresses = in6_dev->cnf.max_addresses;
1945                         u32 addr_flags = 0;
1946
1947 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1948                         if (in6_dev->cnf.optimistic_dad &&
1949                             !net->ipv6.devconf_all->forwarding)
1950                                 addr_flags = IFA_F_OPTIMISTIC;
1951 #endif
1952
1953                         /* Do not allow to create too much of autoconfigured
1954                          * addresses; this would be too easy way to crash kernel.
1955                          */
1956                         if (!max_addresses ||
1957                             ipv6_count_addresses(in6_dev) < max_addresses)
1958                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1959                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1960                                                     addr_flags);
1961
1962                         if (!ifp || IS_ERR(ifp)) {
1963                                 in6_dev_put(in6_dev);
1964                                 return;
1965                         }
1966
1967                         update_lft = create = 1;
1968                         ifp->cstamp = jiffies;
1969                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1970                 }
1971
1972                 if (ifp) {
1973                         int flags;
1974                         unsigned long now;
1975 #ifdef CONFIG_IPV6_PRIVACY
1976                         struct inet6_ifaddr *ift;
1977 #endif
1978                         u32 stored_lft;
1979
1980                         /* update lifetime (RFC2462 5.5.3 e) */
1981                         spin_lock(&ifp->lock);
1982                         now = jiffies;
1983                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1984                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1985                         else
1986                                 stored_lft = 0;
1987                         if (!update_lft && stored_lft) {
1988                                 if (valid_lft > MIN_VALID_LIFETIME ||
1989                                     valid_lft > stored_lft)
1990                                         update_lft = 1;
1991                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1992                                         /* valid_lft <= stored_lft is always true */
1993                                         /*
1994                                          * RFC 4862 Section 5.5.3e:
1995                                          * "Note that the preferred lifetime of
1996                                          *  the corresponding address is always
1997                                          *  reset to the Preferred Lifetime in
1998                                          *  the received Prefix Information
1999                                          *  option, regardless of whether the
2000                                          *  valid lifetime is also reset or
2001                                          *  ignored."
2002                                          *
2003                                          *  So if the preferred lifetime in
2004                                          *  this advertisement is different
2005                                          *  than what we have stored, but the
2006                                          *  valid lifetime is invalid, just
2007                                          *  reset prefered_lft.
2008                                          *
2009                                          *  We must set the valid lifetime
2010                                          *  to the stored lifetime since we'll
2011                                          *  be updating the timestamp below,
2012                                          *  else we'll set it back to the
2013                                          *  minimum.
2014                                          */
2015                                         if (prefered_lft != ifp->prefered_lft) {
2016                                                 valid_lft = stored_lft;
2017                                                 update_lft = 1;
2018                                         }
2019                                 } else {
2020                                         valid_lft = MIN_VALID_LIFETIME;
2021                                         if (valid_lft < prefered_lft)
2022                                                 prefered_lft = valid_lft;
2023                                         update_lft = 1;
2024                                 }
2025                         }
2026
2027                         if (update_lft) {
2028                                 ifp->valid_lft = valid_lft;
2029                                 ifp->prefered_lft = prefered_lft;
2030                                 ifp->tstamp = now;
2031                                 flags = ifp->flags;
2032                                 ifp->flags &= ~IFA_F_DEPRECATED;
2033                                 spin_unlock(&ifp->lock);
2034
2035                                 if (!(flags&IFA_F_TENTATIVE))
2036                                         ipv6_ifa_notify(0, ifp);
2037                         } else
2038                                 spin_unlock(&ifp->lock);
2039
2040 #ifdef CONFIG_IPV6_PRIVACY
2041                         read_lock_bh(&in6_dev->lock);
2042                         /* update all temporary addresses in the list */
2043                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2044                                             tmp_list) {
2045                                 int age, max_valid, max_prefered;
2046
2047                                 if (ifp != ift->ifpub)
2048                                         continue;
2049
2050                                 /*
2051                                  * RFC 4941 section 3.3:
2052                                  * If a received option will extend the lifetime
2053                                  * of a public address, the lifetimes of
2054                                  * temporary addresses should be extended,
2055                                  * subject to the overall constraint that no
2056                                  * temporary addresses should ever remain
2057                                  * "valid" or "preferred" for a time longer than
2058                                  * (TEMP_VALID_LIFETIME) or
2059                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2060                                  * respectively.
2061                                  */
2062                                 age = (now - ift->cstamp) / HZ;
2063                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2064                                 if (max_valid < 0)
2065                                         max_valid = 0;
2066
2067                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2068                                                in6_dev->cnf.max_desync_factor -
2069                                                age;
2070                                 if (max_prefered < 0)
2071                                         max_prefered = 0;
2072
2073                                 if (valid_lft > max_valid)
2074                                         valid_lft = max_valid;
2075
2076                                 if (prefered_lft > max_prefered)
2077                                         prefered_lft = max_prefered;
2078
2079                                 spin_lock(&ift->lock);
2080                                 flags = ift->flags;
2081                                 ift->valid_lft = valid_lft;
2082                                 ift->prefered_lft = prefered_lft;
2083                                 ift->tstamp = now;
2084                                 if (prefered_lft > 0)
2085                                         ift->flags &= ~IFA_F_DEPRECATED;
2086
2087                                 spin_unlock(&ift->lock);
2088                                 if (!(flags&IFA_F_TENTATIVE))
2089                                         ipv6_ifa_notify(0, ift);
2090                         }
2091
2092                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2093                                 /*
2094                                  * When a new public address is created as
2095                                  * described in [ADDRCONF], also create a new
2096                                  * temporary address. Also create a temporary
2097                                  * address if it's enabled but no temporary
2098                                  * address currently exists.
2099                                  */
2100                                 read_unlock_bh(&in6_dev->lock);
2101                                 ipv6_create_tempaddr(ifp, NULL);
2102                         } else {
2103                                 read_unlock_bh(&in6_dev->lock);
2104                         }
2105 #endif
2106                         in6_ifa_put(ifp);
2107                         addrconf_verify(0);
2108                 }
2109         }
2110         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2111         in6_dev_put(in6_dev);
2112 }
2113
2114 /*
2115  *      Set destination address.
2116  *      Special case for SIT interfaces where we create a new "virtual"
2117  *      device.
2118  */
2119 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2120 {
2121         struct in6_ifreq ireq;
2122         struct net_device *dev;
2123         int err = -EINVAL;
2124
2125         rtnl_lock();
2126
2127         err = -EFAULT;
2128         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2129                 goto err_exit;
2130
2131         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2132
2133         err = -ENODEV;
2134         if (dev == NULL)
2135                 goto err_exit;
2136
2137 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2138         if (dev->type == ARPHRD_SIT) {
2139                 const struct net_device_ops *ops = dev->netdev_ops;
2140                 struct ifreq ifr;
2141                 struct ip_tunnel_parm p;
2142
2143                 err = -EADDRNOTAVAIL;
2144                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2145                         goto err_exit;
2146
2147                 memset(&p, 0, sizeof(p));
2148                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2149                 p.iph.saddr = 0;
2150                 p.iph.version = 4;
2151                 p.iph.ihl = 5;
2152                 p.iph.protocol = IPPROTO_IPV6;
2153                 p.iph.ttl = 64;
2154                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2155
2156                 if (ops->ndo_do_ioctl) {
2157                         mm_segment_t oldfs = get_fs();
2158
2159                         set_fs(KERNEL_DS);
2160                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2161                         set_fs(oldfs);
2162                 } else
2163                         err = -EOPNOTSUPP;
2164
2165                 if (err == 0) {
2166                         err = -ENOBUFS;
2167                         dev = __dev_get_by_name(net, p.name);
2168                         if (!dev)
2169                                 goto err_exit;
2170                         err = dev_open(dev);
2171                 }
2172         }
2173 #endif
2174
2175 err_exit:
2176         rtnl_unlock();
2177         return err;
2178 }
2179
2180 /*
2181  *      Manual configuration of address on an interface
2182  */
2183 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2184                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2185                           __u32 valid_lft)
2186 {
2187         struct inet6_ifaddr *ifp;
2188         struct inet6_dev *idev;
2189         struct net_device *dev;
2190         int scope;
2191         u32 flags;
2192         clock_t expires;
2193         unsigned long timeout;
2194
2195         ASSERT_RTNL();
2196
2197         if (plen > 128)
2198                 return -EINVAL;
2199
2200         /* check the lifetime */
2201         if (!valid_lft || prefered_lft > valid_lft)
2202                 return -EINVAL;
2203
2204         dev = __dev_get_by_index(net, ifindex);
2205         if (!dev)
2206                 return -ENODEV;
2207
2208         idev = addrconf_add_dev(dev);
2209         if (IS_ERR(idev))
2210                 return PTR_ERR(idev);
2211
2212         scope = ipv6_addr_scope(pfx);
2213
2214         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2215         if (addrconf_finite_timeout(timeout)) {
2216                 expires = jiffies_to_clock_t(timeout * HZ);
2217                 valid_lft = timeout;
2218                 flags = RTF_EXPIRES;
2219         } else {
2220                 expires = 0;
2221                 flags = 0;
2222                 ifa_flags |= IFA_F_PERMANENT;
2223         }
2224
2225         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2226         if (addrconf_finite_timeout(timeout)) {
2227                 if (timeout == 0)
2228                         ifa_flags |= IFA_F_DEPRECATED;
2229                 prefered_lft = timeout;
2230         }
2231
2232         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2233
2234         if (!IS_ERR(ifp)) {
2235                 spin_lock_bh(&ifp->lock);
2236                 ifp->valid_lft = valid_lft;
2237                 ifp->prefered_lft = prefered_lft;
2238                 ifp->tstamp = jiffies;
2239                 spin_unlock_bh(&ifp->lock);
2240
2241                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2242                                       expires, flags);
2243                 /*
2244                  * Note that section 3.1 of RFC 4429 indicates
2245                  * that the Optimistic flag should not be set for
2246                  * manually configured addresses
2247                  */
2248                 addrconf_dad_start(ifp, 0);
2249                 in6_ifa_put(ifp);
2250                 addrconf_verify(0);
2251                 return 0;
2252         }
2253
2254         return PTR_ERR(ifp);
2255 }
2256
2257 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2258                           unsigned int plen)
2259 {
2260         struct inet6_ifaddr *ifp;
2261         struct inet6_dev *idev;
2262         struct net_device *dev;
2263
2264         if (plen > 128)
2265                 return -EINVAL;
2266
2267         dev = __dev_get_by_index(net, ifindex);
2268         if (!dev)
2269                 return -ENODEV;
2270
2271         if ((idev = __in6_dev_get(dev)) == NULL)
2272                 return -ENXIO;
2273
2274         read_lock_bh(&idev->lock);
2275         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2276                 if (ifp->prefix_len == plen &&
2277                     ipv6_addr_equal(pfx, &ifp->addr)) {
2278                         in6_ifa_hold(ifp);
2279                         read_unlock_bh(&idev->lock);
2280
2281                         ipv6_del_addr(ifp);
2282
2283                         /* If the last address is deleted administratively,
2284                            disable IPv6 on this interface.
2285                          */
2286                         if (list_empty(&idev->addr_list))
2287                                 addrconf_ifdown(idev->dev, 1);
2288                         return 0;
2289                 }
2290         }
2291         read_unlock_bh(&idev->lock);
2292         return -EADDRNOTAVAIL;
2293 }
2294
2295
2296 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2297 {
2298         struct in6_ifreq ireq;
2299         int err;
2300
2301         if (!capable(CAP_NET_ADMIN))
2302                 return -EPERM;
2303
2304         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2305                 return -EFAULT;
2306
2307         rtnl_lock();
2308         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2309                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2310                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2311         rtnl_unlock();
2312         return err;
2313 }
2314
2315 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2316 {
2317         struct in6_ifreq ireq;
2318         int err;
2319
2320         if (!capable(CAP_NET_ADMIN))
2321                 return -EPERM;
2322
2323         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2324                 return -EFAULT;
2325
2326         rtnl_lock();
2327         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2328                              ireq.ifr6_prefixlen);
2329         rtnl_unlock();
2330         return err;
2331 }
2332
2333 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2334                      int plen, int scope)
2335 {
2336         struct inet6_ifaddr *ifp;
2337
2338         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2339         if (!IS_ERR(ifp)) {
2340                 spin_lock_bh(&ifp->lock);
2341                 ifp->flags &= ~IFA_F_TENTATIVE;
2342                 spin_unlock_bh(&ifp->lock);
2343                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2344                 in6_ifa_put(ifp);
2345         }
2346 }
2347
2348 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2349 static void sit_add_v4_addrs(struct inet6_dev *idev)
2350 {
2351         struct in6_addr addr;
2352         struct net_device *dev;
2353         struct net *net = dev_net(idev->dev);
2354         int scope;
2355
2356         ASSERT_RTNL();
2357
2358         memset(&addr, 0, sizeof(struct in6_addr));
2359         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2360
2361         if (idev->dev->flags&IFF_POINTOPOINT) {
2362                 addr.s6_addr32[0] = htonl(0xfe800000);
2363                 scope = IFA_LINK;
2364         } else {
2365                 scope = IPV6_ADDR_COMPATv4;
2366         }
2367
2368         if (addr.s6_addr32[3]) {
2369                 add_addr(idev, &addr, 128, scope);
2370                 return;
2371         }
2372
2373         for_each_netdev(net, dev) {
2374                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2375                 if (in_dev && (dev->flags & IFF_UP)) {
2376                         struct in_ifaddr * ifa;
2377
2378                         int flag = scope;
2379
2380                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2381                                 int plen;
2382
2383                                 addr.s6_addr32[3] = ifa->ifa_local;
2384
2385                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2386                                         continue;
2387                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2388                                         if (idev->dev->flags&IFF_POINTOPOINT)
2389                                                 continue;
2390                                         flag |= IFA_HOST;
2391                                 }
2392                                 if (idev->dev->flags&IFF_POINTOPOINT)
2393                                         plen = 64;
2394                                 else
2395                                         plen = 96;
2396
2397                                 add_addr(idev, &addr, plen, flag);
2398                         }
2399                 }
2400         }
2401 }
2402 #endif
2403
2404 static void init_loopback(struct net_device *dev)
2405 {
2406         struct inet6_dev  *idev;
2407
2408         /* ::1 */
2409
2410         ASSERT_RTNL();
2411
2412         if ((idev = ipv6_find_idev(dev)) == NULL) {
2413                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2414                 return;
2415         }
2416
2417         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2418 }
2419
2420 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2421 {
2422         struct inet6_ifaddr * ifp;
2423         u32 addr_flags = IFA_F_PERMANENT;
2424
2425 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2426         if (idev->cnf.optimistic_dad &&
2427             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2428                 addr_flags |= IFA_F_OPTIMISTIC;
2429 #endif
2430
2431
2432         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2433         if (!IS_ERR(ifp)) {
2434                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2435                 addrconf_dad_start(ifp, 0);
2436                 in6_ifa_put(ifp);
2437         }
2438 }
2439
2440 static void addrconf_dev_config(struct net_device *dev)
2441 {
2442         struct in6_addr addr;
2443         struct inet6_dev    * idev;
2444
2445         ASSERT_RTNL();
2446
2447         if ((dev->type != ARPHRD_ETHER) &&
2448             (dev->type != ARPHRD_FDDI) &&
2449             (dev->type != ARPHRD_IEEE802_TR) &&
2450             (dev->type != ARPHRD_ARCNET) &&
2451             (dev->type != ARPHRD_INFINIBAND)) {
2452                 /* Alas, we support only Ethernet autoconfiguration. */
2453                 return;
2454         }
2455
2456         idev = addrconf_add_dev(dev);
2457         if (IS_ERR(idev))
2458                 return;
2459
2460         memset(&addr, 0, sizeof(struct in6_addr));
2461         addr.s6_addr32[0] = htonl(0xFE800000);
2462
2463         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2464                 addrconf_add_linklocal(idev, &addr);
2465 }
2466
2467 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2468 static void addrconf_sit_config(struct net_device *dev)
2469 {
2470         struct inet6_dev *idev;
2471
2472         ASSERT_RTNL();
2473
2474         /*
2475          * Configure the tunnel with one of our IPv4
2476          * addresses... we should configure all of
2477          * our v4 addrs in the tunnel
2478          */
2479
2480         if ((idev = ipv6_find_idev(dev)) == NULL) {
2481                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2482                 return;
2483         }
2484
2485         if (dev->priv_flags & IFF_ISATAP) {
2486                 struct in6_addr addr;
2487
2488                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2489                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2490                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2491                         addrconf_add_linklocal(idev, &addr);
2492                 return;
2493         }
2494
2495         sit_add_v4_addrs(idev);
2496
2497         if (dev->flags&IFF_POINTOPOINT) {
2498                 addrconf_add_mroute(dev);
2499                 addrconf_add_lroute(dev);
2500         } else
2501                 sit_route_add(dev);
2502 }
2503 #endif
2504
2505 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2506 static void addrconf_gre_config(struct net_device *dev)
2507 {
2508         struct inet6_dev *idev;
2509         struct in6_addr addr;
2510
2511         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2512
2513         ASSERT_RTNL();
2514
2515         if ((idev = ipv6_find_idev(dev)) == NULL) {
2516                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2517                 return;
2518         }
2519
2520         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2521         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2522
2523         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2524                 addrconf_add_linklocal(idev, &addr);
2525 }
2526 #endif
2527
2528 static inline int
2529 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2530 {
2531         struct in6_addr lladdr;
2532
2533         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2534                 addrconf_add_linklocal(idev, &lladdr);
2535                 return 0;
2536         }
2537         return -1;
2538 }
2539
2540 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2541 {
2542         struct net_device *link_dev;
2543         struct net *net = dev_net(idev->dev);
2544
2545         /* first try to inherit the link-local address from the link device */
2546         if (idev->dev->iflink &&
2547             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2548                 if (!ipv6_inherit_linklocal(idev, link_dev))
2549                         return;
2550         }
2551         /* then try to inherit it from any device */
2552         for_each_netdev(net, link_dev) {
2553                 if (!ipv6_inherit_linklocal(idev, link_dev))
2554                         return;
2555         }
2556         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2557 }
2558
2559 /*
2560  * Autoconfigure tunnel with a link-local address so routing protocols,
2561  * DHCPv6, MLD etc. can be run over the virtual link
2562  */
2563
2564 static void addrconf_ip6_tnl_config(struct net_device *dev)
2565 {
2566         struct inet6_dev *idev;
2567
2568         ASSERT_RTNL();
2569
2570         idev = addrconf_add_dev(dev);
2571         if (IS_ERR(idev)) {
2572                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2573                 return;
2574         }
2575         ip6_tnl_add_linklocal(idev);
2576 }
2577
2578 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2579                            void * data)
2580 {
2581         struct net_device *dev = (struct net_device *) data;
2582         struct inet6_dev *idev = __in6_dev_get(dev);
2583         int run_pending = 0;
2584         int err;
2585
2586         switch (event) {
2587         case NETDEV_REGISTER:
2588                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2589                         idev = ipv6_add_dev(dev);
2590                         if (!idev)
2591                                 return notifier_from_errno(-ENOMEM);
2592                 }
2593                 break;
2594
2595         case NETDEV_UP:
2596         case NETDEV_CHANGE:
2597                 if (dev->flags & IFF_SLAVE)
2598                         break;
2599
2600                 if (event == NETDEV_UP) {
2601                         if (!addrconf_qdisc_ok(dev)) {
2602                                 /* device is not ready yet. */
2603                                 printk(KERN_INFO
2604                                         "ADDRCONF(NETDEV_UP): %s: "
2605                                         "link is not ready\n",
2606                                         dev->name);
2607                                 break;
2608                         }
2609
2610                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2611                                 idev = ipv6_add_dev(dev);
2612
2613                         if (idev) {
2614                                 idev->if_flags |= IF_READY;
2615                                 run_pending = 1;
2616                         }
2617                 } else {
2618                         if (!addrconf_qdisc_ok(dev)) {
2619                                 /* device is still not ready. */
2620                                 break;
2621                         }
2622
2623                         if (idev) {
2624                                 if (idev->if_flags & IF_READY)
2625                                         /* device is already configured. */
2626                                         break;
2627                                 idev->if_flags |= IF_READY;
2628                         }
2629
2630                         printk(KERN_INFO
2631                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2632                                         "link becomes ready\n",
2633                                         dev->name);
2634
2635                         run_pending = 1;
2636                 }
2637
2638                 switch (dev->type) {
2639 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2640                 case ARPHRD_SIT:
2641                         addrconf_sit_config(dev);
2642                         break;
2643 #endif
2644 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2645                 case ARPHRD_IPGRE:
2646                         addrconf_gre_config(dev);
2647                         break;
2648 #endif
2649                 case ARPHRD_TUNNEL6:
2650                         addrconf_ip6_tnl_config(dev);
2651                         break;
2652                 case ARPHRD_LOOPBACK:
2653                         init_loopback(dev);
2654                         break;
2655
2656                 default:
2657                         addrconf_dev_config(dev);
2658                         break;
2659                 }
2660
2661                 if (idev) {
2662                         if (run_pending)
2663                                 addrconf_dad_run(idev);
2664
2665                         /*
2666                          * If the MTU changed during the interface down,
2667                          * when the interface up, the changed MTU must be
2668                          * reflected in the idev as well as routers.
2669                          */
2670                         if (idev->cnf.mtu6 != dev->mtu &&
2671                             dev->mtu >= IPV6_MIN_MTU) {
2672                                 rt6_mtu_change(dev, dev->mtu);
2673                                 idev->cnf.mtu6 = dev->mtu;
2674                         }
2675                         idev->tstamp = jiffies;
2676                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2677
2678                         /*
2679                          * If the changed mtu during down is lower than
2680                          * IPV6_MIN_MTU stop IPv6 on this interface.
2681                          */
2682                         if (dev->mtu < IPV6_MIN_MTU)
2683                                 addrconf_ifdown(dev, 1);
2684                 }
2685                 break;
2686
2687         case NETDEV_CHANGEMTU:
2688                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2689                         rt6_mtu_change(dev, dev->mtu);
2690                         idev->cnf.mtu6 = dev->mtu;
2691                         break;
2692                 }
2693
2694                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2695                         idev = ipv6_add_dev(dev);
2696                         if (idev)
2697                                 break;
2698                 }
2699
2700                 /*
2701                  * MTU falled under IPV6_MIN_MTU.
2702                  * Stop IPv6 on this interface.
2703                  */
2704
2705         case NETDEV_DOWN:
2706         case NETDEV_UNREGISTER:
2707                 /*
2708                  *      Remove all addresses from this interface.
2709                  */
2710                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2711                 break;
2712
2713         case NETDEV_CHANGENAME:
2714                 if (idev) {
2715                         snmp6_unregister_dev(idev);
2716                         addrconf_sysctl_unregister(idev);
2717                         addrconf_sysctl_register(idev);
2718                         err = snmp6_register_dev(idev);
2719                         if (err)
2720                                 return notifier_from_errno(err);
2721                 }
2722                 break;
2723
2724         case NETDEV_PRE_TYPE_CHANGE:
2725         case NETDEV_POST_TYPE_CHANGE:
2726                 addrconf_type_change(dev, event);
2727                 break;
2728         }
2729
2730         return NOTIFY_OK;
2731 }
2732
2733 /*
2734  *      addrconf module should be notified of a device going up
2735  */
2736 static struct notifier_block ipv6_dev_notf = {
2737         .notifier_call = addrconf_notify,
2738 };
2739
2740 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2741 {
2742         struct inet6_dev *idev;
2743         ASSERT_RTNL();
2744
2745         idev = __in6_dev_get(dev);
2746
2747         if (event == NETDEV_POST_TYPE_CHANGE)
2748                 ipv6_mc_remap(idev);
2749         else if (event == NETDEV_PRE_TYPE_CHANGE)
2750                 ipv6_mc_unmap(idev);
2751 }
2752
2753 static int addrconf_ifdown(struct net_device *dev, int how)
2754 {
2755         struct net *net = dev_net(dev);
2756         struct inet6_dev *idev;
2757         struct inet6_ifaddr *ifa;
2758         int state, i;
2759
2760         ASSERT_RTNL();
2761
2762         rt6_ifdown(net, dev);
2763         neigh_ifdown(&nd_tbl, dev);
2764
2765         idev = __in6_dev_get(dev);
2766         if (idev == NULL)
2767                 return -ENODEV;
2768
2769         /*
2770          * Step 1: remove reference to ipv6 device from parent device.
2771          *         Do not dev_put!
2772          */
2773         if (how) {
2774                 idev->dead = 1;
2775
2776                 /* protected by rtnl_lock */
2777                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2778
2779                 /* Step 1.5: remove snmp6 entry */
2780                 snmp6_unregister_dev(idev);
2781
2782         }
2783
2784         /* Step 2: clear hash table */
2785         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2786                 struct hlist_head *h = &inet6_addr_lst[i];
2787                 struct hlist_node *n;
2788
2789                 spin_lock_bh(&addrconf_hash_lock);
2790         restart:
2791                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2792                         if (ifa->idev == idev) {
2793                                 hlist_del_init_rcu(&ifa->addr_lst);
2794                                 addrconf_del_timer(ifa);
2795                                 goto restart;
2796                         }
2797                 }
2798                 spin_unlock_bh(&addrconf_hash_lock);
2799         }
2800
2801         write_lock_bh(&idev->lock);
2802
2803         /* Step 2: clear flags for stateless addrconf */
2804         if (!how)
2805                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2806
2807 #ifdef CONFIG_IPV6_PRIVACY
2808         if (how && del_timer(&idev->regen_timer))
2809                 in6_dev_put(idev);
2810
2811         /* Step 3: clear tempaddr list */
2812         while (!list_empty(&idev->tempaddr_list)) {
2813                 ifa = list_first_entry(&idev->tempaddr_list,
2814                                        struct inet6_ifaddr, tmp_list);
2815                 list_del(&ifa->tmp_list);
2816                 write_unlock_bh(&idev->lock);
2817                 spin_lock_bh(&ifa->lock);
2818
2819                 if (ifa->ifpub) {
2820                         in6_ifa_put(ifa->ifpub);
2821                         ifa->ifpub = NULL;
2822                 }
2823                 spin_unlock_bh(&ifa->lock);
2824                 in6_ifa_put(ifa);
2825                 write_lock_bh(&idev->lock);
2826         }
2827 #endif
2828
2829         while (!list_empty(&idev->addr_list)) {
2830                 ifa = list_first_entry(&idev->addr_list,
2831                                        struct inet6_ifaddr, if_list);
2832                 addrconf_del_timer(ifa);
2833
2834                 list_del(&ifa->if_list);
2835
2836                 write_unlock_bh(&idev->lock);
2837
2838                 spin_lock_bh(&ifa->state_lock);
2839                 state = ifa->state;
2840                 ifa->state = INET6_IFADDR_STATE_DEAD;
2841                 spin_unlock_bh(&ifa->state_lock);
2842
2843                 if (state != INET6_IFADDR_STATE_DEAD) {
2844                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2845                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2846                 }
2847                 in6_ifa_put(ifa);
2848
2849                 write_lock_bh(&idev->lock);
2850         }
2851
2852         write_unlock_bh(&idev->lock);
2853
2854         /* Step 5: Discard multicast list */
2855         if (how)
2856                 ipv6_mc_destroy_dev(idev);
2857         else
2858                 ipv6_mc_down(idev);
2859
2860         idev->tstamp = jiffies;
2861
2862         /* Last: Shot the device (if unregistered) */
2863         if (how) {
2864                 addrconf_sysctl_unregister(idev);
2865                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2866                 neigh_ifdown(&nd_tbl, dev);
2867                 in6_dev_put(idev);
2868         }
2869         return 0;
2870 }
2871
2872 static void addrconf_rs_timer(unsigned long data)
2873 {
2874         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2875         struct inet6_dev *idev = ifp->idev;
2876
2877         read_lock(&idev->lock);
2878         if (idev->dead || !(idev->if_flags & IF_READY))
2879                 goto out;
2880
2881         if (idev->cnf.forwarding)
2882                 goto out;
2883
2884         /* Announcement received after solicitation was sent */
2885         if (idev->if_flags & IF_RA_RCVD)
2886                 goto out;
2887
2888         spin_lock(&ifp->lock);
2889         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2890                 /* The wait after the last probe can be shorter */
2891                 addrconf_mod_timer(ifp, AC_RS,
2892                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2893                                    idev->cnf.rtr_solicit_delay :
2894                                    idev->cnf.rtr_solicit_interval);
2895                 spin_unlock(&ifp->lock);
2896
2897                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2898         } else {
2899                 spin_unlock(&ifp->lock);
2900                 /*
2901                  * Note: we do not support deprecated "all on-link"
2902                  * assumption any longer.
2903                  */
2904                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2905                        idev->dev->name);
2906         }
2907
2908 out:
2909         read_unlock(&idev->lock);
2910         in6_ifa_put(ifp);
2911 }
2912
2913 /*
2914  *      Duplicate Address Detection
2915  */
2916 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2917 {
2918         unsigned long rand_num;
2919         struct inet6_dev *idev = ifp->idev;
2920
2921         if (ifp->flags & IFA_F_OPTIMISTIC)
2922                 rand_num = 0;
2923         else
2924                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2925
2926         ifp->probes = idev->cnf.dad_transmits;
2927         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2928 }
2929
2930 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2931 {
2932         struct inet6_dev *idev = ifp->idev;
2933         struct net_device *dev = idev->dev;
2934
2935         addrconf_join_solict(dev, &ifp->addr);
2936
2937         net_srandom(ifp->addr.s6_addr32[3]);
2938
2939         read_lock_bh(&idev->lock);
2940         spin_lock(&ifp->lock);
2941         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2942                 goto out;
2943
2944         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2945             idev->cnf.accept_dad < 1 ||
2946             !(ifp->flags&IFA_F_TENTATIVE) ||
2947             ifp->flags & IFA_F_NODAD) {
2948                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2949                 spin_unlock(&ifp->lock);
2950                 read_unlock_bh(&idev->lock);
2951
2952                 addrconf_dad_completed(ifp);
2953                 return;
2954         }
2955
2956         if (!(idev->if_flags & IF_READY)) {
2957                 spin_unlock(&ifp->lock);
2958                 read_unlock_bh(&idev->lock);
2959                 /*
2960                  * If the device is not ready:
2961                  * - keep it tentative if it is a permanent address.
2962                  * - otherwise, kill it.
2963                  */
2964                 in6_ifa_hold(ifp);
2965                 addrconf_dad_stop(ifp, 0);
2966                 return;
2967         }
2968
2969         /*
2970          * Optimistic nodes can start receiving
2971          * Frames right away
2972          */
2973         if (ifp->flags & IFA_F_OPTIMISTIC)
2974                 ip6_ins_rt(ifp->rt);
2975
2976         addrconf_dad_kick(ifp);
2977 out:
2978         spin_unlock(&ifp->lock);
2979         read_unlock_bh(&idev->lock);
2980 }
2981
2982 static void addrconf_dad_timer(unsigned long data)
2983 {
2984         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2985         struct inet6_dev *idev = ifp->idev;
2986         struct in6_addr mcaddr;
2987
2988         if (!ifp->probes && addrconf_dad_end(ifp))
2989                 goto out;
2990
2991         read_lock(&idev->lock);
2992         if (idev->dead || !(idev->if_flags & IF_READY)) {
2993                 read_unlock(&idev->lock);
2994                 goto out;
2995         }
2996
2997         spin_lock(&ifp->lock);
2998         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2999                 spin_unlock(&ifp->lock);
3000                 read_unlock(&idev->lock);
3001                 goto out;
3002         }
3003
3004         if (ifp->probes == 0) {
3005                 /*
3006                  * DAD was successful
3007                  */
3008
3009                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3010                 spin_unlock(&ifp->lock);
3011                 read_unlock(&idev->lock);
3012
3013                 addrconf_dad_completed(ifp);
3014
3015                 goto out;
3016         }
3017
3018         ifp->probes--;
3019         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3020         spin_unlock(&ifp->lock);
3021         read_unlock(&idev->lock);
3022
3023         /* send a neighbour solicitation for our addr */
3024         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3025         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3026 out:
3027         in6_ifa_put(ifp);
3028 }
3029
3030 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3031 {
3032         struct net_device *dev = ifp->idev->dev;
3033
3034         /*
3035          *      Configure the address for reception. Now it is valid.
3036          */
3037
3038         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3039
3040         /* If added prefix is link local and we are prepared to process
3041            router advertisements, start sending router solicitations.
3042          */
3043
3044         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3045              ifp->idev->cnf.accept_ra == 2) &&
3046             ifp->idev->cnf.rtr_solicits > 0 &&
3047             (dev->flags&IFF_LOOPBACK) == 0 &&
3048             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3049                 /*
3050                  *      If a host as already performed a random delay
3051                  *      [...] as part of DAD [...] there is no need
3052                  *      to delay again before sending the first RS
3053                  */
3054                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3055
3056                 spin_lock_bh(&ifp->lock);
3057                 ifp->probes = 1;
3058                 ifp->idev->if_flags |= IF_RS_SENT;
3059                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3060                 spin_unlock_bh(&ifp->lock);
3061         }
3062 }
3063
3064 static void addrconf_dad_run(struct inet6_dev *idev)
3065 {
3066         struct inet6_ifaddr *ifp;
3067
3068         read_lock_bh(&idev->lock);
3069         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3070                 spin_lock(&ifp->lock);
3071                 if (ifp->flags & IFA_F_TENTATIVE &&
3072                     ifp->state == INET6_IFADDR_STATE_DAD)
3073                         addrconf_dad_kick(ifp);
3074                 spin_unlock(&ifp->lock);
3075         }
3076         read_unlock_bh(&idev->lock);
3077 }
3078
3079 #ifdef CONFIG_PROC_FS
3080 struct if6_iter_state {
3081         struct seq_net_private p;
3082         int bucket;
3083 };
3084
3085 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3086 {
3087         struct inet6_ifaddr *ifa = NULL;
3088         struct if6_iter_state *state = seq->private;
3089         struct net *net = seq_file_net(seq);
3090
3091         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3092                 struct hlist_node *n;
3093                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3094                                          addr_lst)
3095                         if (net_eq(dev_net(ifa->idev->dev), net))
3096                                 return ifa;
3097         }
3098         return NULL;
3099 }
3100
3101 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3102                                          struct inet6_ifaddr *ifa)
3103 {
3104         struct if6_iter_state *state = seq->private;
3105         struct net *net = seq_file_net(seq);
3106         struct hlist_node *n = &ifa->addr_lst;
3107
3108         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3109                 if (net_eq(dev_net(ifa->idev->dev), net))
3110                         return ifa;
3111
3112         while (++state->bucket < IN6_ADDR_HSIZE) {
3113                 hlist_for_each_entry_rcu_bh(ifa, n,
3114                                      &inet6_addr_lst[state->bucket], addr_lst) {
3115                         if (net_eq(dev_net(ifa->idev->dev), net))
3116                                 return ifa;
3117                 }
3118         }
3119
3120         return NULL;
3121 }
3122
3123 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3124 {
3125         struct inet6_ifaddr *ifa = if6_get_first(seq);
3126
3127         if (ifa)
3128                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3129                         --pos;
3130         return pos ? NULL : ifa;
3131 }
3132
3133 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3134         __acquires(rcu_bh)
3135 {
3136         rcu_read_lock_bh();
3137         return if6_get_idx(seq, *pos);
3138 }
3139
3140 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3141 {
3142         struct inet6_ifaddr *ifa;
3143
3144         ifa = if6_get_next(seq, v);
3145         ++*pos;
3146         return ifa;
3147 }
3148
3149 static void if6_seq_stop(struct seq_file *seq, void *v)
3150         __releases(rcu_bh)
3151 {
3152         rcu_read_unlock_bh();
3153 }
3154
3155 static int if6_seq_show(struct seq_file *seq, void *v)
3156 {
3157         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3158         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3159                    &ifp->addr,
3160                    ifp->idev->dev->ifindex,
3161                    ifp->prefix_len,
3162                    ifp->scope,
3163                    ifp->flags,
3164                    ifp->idev->dev->name);
3165         return 0;
3166 }
3167
3168 static const struct seq_operations if6_seq_ops = {
3169         .start  = if6_seq_start,
3170         .next   = if6_seq_next,
3171         .show   = if6_seq_show,
3172         .stop   = if6_seq_stop,
3173 };
3174
3175 static int if6_seq_open(struct inode *inode, struct file *file)
3176 {
3177         return seq_open_net(inode, file, &if6_seq_ops,
3178                             sizeof(struct if6_iter_state));
3179 }
3180
3181 static const struct file_operations if6_fops = {
3182         .owner          = THIS_MODULE,
3183         .open           = if6_seq_open,
3184         .read           = seq_read,
3185         .llseek         = seq_lseek,
3186         .release        = seq_release_net,
3187 };
3188
3189 static int __net_init if6_proc_net_init(struct net *net)
3190 {
3191         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3192                 return -ENOMEM;
3193         return 0;
3194 }
3195
3196 static void __net_exit if6_proc_net_exit(struct net *net)
3197 {
3198        proc_net_remove(net, "if_inet6");
3199 }
3200
3201 static struct pernet_operations if6_proc_net_ops = {
3202        .init = if6_proc_net_init,
3203        .exit = if6_proc_net_exit,
3204 };
3205
3206 int __init if6_proc_init(void)
3207 {
3208         return register_pernet_subsys(&if6_proc_net_ops);
3209 }
3210
3211 void if6_proc_exit(void)
3212 {
3213         unregister_pernet_subsys(&if6_proc_net_ops);
3214 }
3215 #endif  /* CONFIG_PROC_FS */
3216
3217 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3218 /* Check if address is a home address configured on any interface. */
3219 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3220 {
3221         int ret = 0;
3222         struct inet6_ifaddr *ifp = NULL;
3223         struct hlist_node *n;
3224         unsigned int hash = ipv6_addr_hash(addr);
3225
3226         rcu_read_lock_bh();
3227         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3228                 if (!net_eq(dev_net(ifp->idev->dev), net))
3229                         continue;
3230                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3231                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3232                         ret = 1;
3233                         break;
3234                 }
3235         }
3236         rcu_read_unlock_bh();
3237         return ret;
3238 }
3239 #endif
3240
3241 /*
3242  *      Periodic address status verification
3243  */
3244
3245 static void addrconf_verify(unsigned long foo)
3246 {
3247         unsigned long now, next, next_sec, next_sched;
3248         struct inet6_ifaddr *ifp;
3249         struct hlist_node *node;
3250         int i;
3251
3252         rcu_read_lock_bh();
3253         spin_lock(&addrconf_verify_lock);
3254         now = jiffies;
3255         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3256
3257         del_timer(&addr_chk_timer);
3258
3259         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3260 restart:
3261                 hlist_for_each_entry_rcu_bh(ifp, node,
3262                                          &inet6_addr_lst[i], addr_lst) {
3263                         unsigned long age;
3264
3265                         if (ifp->flags & IFA_F_PERMANENT)
3266                                 continue;
3267
3268                         spin_lock(&ifp->lock);
3269                         /* We try to batch several events at once. */
3270                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3271
3272                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3273                             age >= ifp->valid_lft) {
3274                                 spin_unlock(&ifp->lock);
3275                                 in6_ifa_hold(ifp);
3276                                 ipv6_del_addr(ifp);
3277                                 goto restart;
3278                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3279                                 spin_unlock(&ifp->lock);
3280                                 continue;
3281                         } else if (age >= ifp->prefered_lft) {
3282                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3283                                 int deprecate = 0;
3284
3285                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3286                                         deprecate = 1;
3287                                         ifp->flags |= IFA_F_DEPRECATED;
3288                                 }
3289
3290                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3291                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3292
3293                                 spin_unlock(&ifp->lock);
3294
3295                                 if (deprecate) {
3296                                         in6_ifa_hold(ifp);
3297
3298                                         ipv6_ifa_notify(0, ifp);
3299                                         in6_ifa_put(ifp);
3300                                         goto restart;
3301                                 }
3302 #ifdef CONFIG_IPV6_PRIVACY
3303                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3304                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3305                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3306                                         ifp->idev->cnf.dad_transmits *
3307                                         ifp->idev->nd_parms->retrans_time / HZ;
3308
3309                                 if (age >= ifp->prefered_lft - regen_advance) {
3310                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3311                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3312                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3313                                         if (!ifp->regen_count && ifpub) {
3314                                                 ifp->regen_count++;
3315                                                 in6_ifa_hold(ifp);
3316                                                 in6_ifa_hold(ifpub);
3317                                                 spin_unlock(&ifp->lock);
3318
3319                                                 spin_lock(&ifpub->lock);
3320                                                 ifpub->regen_count = 0;
3321                                                 spin_unlock(&ifpub->lock);
3322                                                 ipv6_create_tempaddr(ifpub, ifp);
3323                                                 in6_ifa_put(ifpub);
3324                                                 in6_ifa_put(ifp);
3325                                                 goto restart;
3326                                         }
3327                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3328                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3329                                 spin_unlock(&ifp->lock);
3330 #endif
3331                         } else {
3332                                 /* ifp->prefered_lft <= ifp->valid_lft */
3333                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3334                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3335                                 spin_unlock(&ifp->lock);
3336                         }
3337                 }
3338         }
3339
3340         next_sec = round_jiffies_up(next);
3341         next_sched = next;
3342
3343         /* If rounded timeout is accurate enough, accept it. */
3344         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3345                 next_sched = next_sec;
3346
3347         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3348         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3349                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3350
3351         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3352               now, next, next_sec, next_sched));
3353
3354         addr_chk_timer.expires = next_sched;
3355         add_timer(&addr_chk_timer);
3356         spin_unlock(&addrconf_verify_lock);
3357         rcu_read_unlock_bh();
3358 }
3359
3360 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3361 {
3362         struct in6_addr *pfx = NULL;
3363
3364         if (addr)
3365                 pfx = nla_data(addr);
3366
3367         if (local) {
3368                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3369                         pfx = NULL;
3370                 else
3371                         pfx = nla_data(local);
3372         }
3373
3374         return pfx;
3375 }
3376
3377 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3378         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3379         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3380         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3381 };
3382
3383 static int
3384 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3385 {
3386         struct net *net = sock_net(skb->sk);
3387         struct ifaddrmsg *ifm;
3388         struct nlattr *tb[IFA_MAX+1];
3389         struct in6_addr *pfx;
3390         int err;
3391
3392         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3393         if (err < 0)
3394                 return err;
3395
3396         ifm = nlmsg_data(nlh);
3397         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3398         if (pfx == NULL)
3399                 return -EINVAL;
3400
3401         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3402 }
3403
3404 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3405                              u32 prefered_lft, u32 valid_lft)
3406 {
3407         u32 flags;
3408         clock_t expires;
3409         unsigned long timeout;
3410
3411         if (!valid_lft || (prefered_lft > valid_lft))
3412                 return -EINVAL;
3413
3414         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3415         if (addrconf_finite_timeout(timeout)) {
3416                 expires = jiffies_to_clock_t(timeout * HZ);
3417                 valid_lft = timeout;
3418                 flags = RTF_EXPIRES;
3419         } else {
3420                 expires = 0;
3421                 flags = 0;
3422                 ifa_flags |= IFA_F_PERMANENT;
3423         }
3424
3425         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3426         if (addrconf_finite_timeout(timeout)) {
3427                 if (timeout == 0)
3428                         ifa_flags |= IFA_F_DEPRECATED;
3429                 prefered_lft = timeout;
3430         }
3431
3432         spin_lock_bh(&ifp->lock);
3433         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3434         ifp->tstamp = jiffies;
3435         ifp->valid_lft = valid_lft;
3436         ifp->prefered_lft = prefered_lft;
3437
3438         spin_unlock_bh(&ifp->lock);
3439         if (!(ifp->flags&IFA_F_TENTATIVE))
3440                 ipv6_ifa_notify(0, ifp);
3441
3442         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3443                               expires, flags);
3444         addrconf_verify(0);
3445
3446         return 0;
3447 }
3448
3449 static int
3450 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3451 {
3452         struct net *net = sock_net(skb->sk);
3453         struct ifaddrmsg *ifm;
3454         struct nlattr *tb[IFA_MAX+1];
3455         struct in6_addr *pfx;
3456         struct inet6_ifaddr *ifa;
3457         struct net_device *dev;
3458         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3459         u8 ifa_flags;
3460         int err;
3461
3462         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3463         if (err < 0)
3464                 return err;
3465
3466         ifm = nlmsg_data(nlh);
3467         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3468         if (pfx == NULL)
3469                 return -EINVAL;
3470
3471         if (tb[IFA_CACHEINFO]) {
3472                 struct ifa_cacheinfo *ci;
3473
3474                 ci = nla_data(tb[IFA_CACHEINFO]);
3475                 valid_lft = ci->ifa_valid;
3476                 preferred_lft = ci->ifa_prefered;
3477         } else {
3478                 preferred_lft = INFINITY_LIFE_TIME;
3479                 valid_lft = INFINITY_LIFE_TIME;
3480         }
3481
3482         dev =  __dev_get_by_index(net, ifm->ifa_index);
3483         if (dev == NULL)
3484                 return -ENODEV;
3485
3486         /* We ignore other flags so far. */
3487         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3488
3489         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3490         if (ifa == NULL) {
3491                 /*
3492                  * It would be best to check for !NLM_F_CREATE here but
3493                  * userspace alreay relies on not having to provide this.
3494                  */
3495                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3496                                       ifm->ifa_prefixlen, ifa_flags,
3497                                       preferred_lft, valid_lft);
3498         }
3499
3500         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3501             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3502                 err = -EEXIST;
3503         else
3504                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3505
3506         in6_ifa_put(ifa);
3507
3508         return err;
3509 }
3510
3511 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3512                           u8 scope, int ifindex)
3513 {
3514         struct ifaddrmsg *ifm;
3515
3516         ifm = nlmsg_data(nlh);
3517         ifm->ifa_family = AF_INET6;
3518         ifm->ifa_prefixlen = prefixlen;
3519         ifm->ifa_flags = flags;
3520         ifm->ifa_scope = scope;
3521         ifm->ifa_index = ifindex;
3522 }
3523
3524 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3525                          unsigned long tstamp, u32 preferred, u32 valid)
3526 {
3527         struct ifa_cacheinfo ci;
3528
3529         ci.cstamp = cstamp_delta(cstamp);
3530         ci.tstamp = cstamp_delta(tstamp);
3531         ci.ifa_prefered = preferred;
3532         ci.ifa_valid = valid;
3533
3534         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3535 }
3536
3537 static inline int rt_scope(int ifa_scope)
3538 {
3539         if (ifa_scope & IFA_HOST)
3540                 return RT_SCOPE_HOST;
3541         else if (ifa_scope & IFA_LINK)
3542                 return RT_SCOPE_LINK;
3543         else if (ifa_scope & IFA_SITE)
3544                 return RT_SCOPE_SITE;
3545         else
3546                 return RT_SCOPE_UNIVERSE;
3547 }
3548
3549 static inline int inet6_ifaddr_msgsize(void)
3550 {
3551         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3552                + nla_total_size(16) /* IFA_ADDRESS */
3553                + nla_total_size(sizeof(struct ifa_cacheinfo));
3554 }
3555
3556 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3557                              u32 pid, u32 seq, int event, unsigned int flags)
3558 {
3559         struct nlmsghdr  *nlh;
3560         u32 preferred, valid;
3561
3562         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3563         if (nlh == NULL)
3564                 return -EMSGSIZE;
3565
3566         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3567                       ifa->idev->dev->ifindex);
3568
3569         if (!(ifa->flags&IFA_F_PERMANENT)) {
3570                 preferred = ifa->prefered_lft;
3571                 valid = ifa->valid_lft;
3572                 if (preferred != INFINITY_LIFE_TIME) {
3573                         long tval = (jiffies - ifa->tstamp)/HZ;
3574                         if (preferred > tval)
3575                                 preferred -= tval;
3576                         else
3577                                 preferred = 0;
3578                         if (valid != INFINITY_LIFE_TIME) {
3579                                 if (valid > tval)
3580                                         valid -= tval;
3581                                 else
3582                                         valid = 0;
3583                         }
3584                 }
3585         } else {
3586                 preferred = INFINITY_LIFE_TIME;
3587                 valid = INFINITY_LIFE_TIME;
3588         }
3589
3590         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3591             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3592                 nlmsg_cancel(skb, nlh);
3593                 return -EMSGSIZE;
3594         }
3595
3596         return nlmsg_end(skb, nlh);
3597 }
3598
3599 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3600                                 u32 pid, u32 seq, int event, u16 flags)
3601 {
3602         struct nlmsghdr  *nlh;
3603         u8 scope = RT_SCOPE_UNIVERSE;
3604         int ifindex = ifmca->idev->dev->ifindex;
3605
3606         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3607                 scope = RT_SCOPE_SITE;
3608
3609         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3610         if (nlh == NULL)
3611                 return -EMSGSIZE;
3612
3613         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3614         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3615             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3616                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3617                 nlmsg_cancel(skb, nlh);
3618                 return -EMSGSIZE;
3619         }
3620
3621         return nlmsg_end(skb, nlh);
3622 }
3623
3624 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3625                                 u32 pid, u32 seq, int event, unsigned int flags)
3626 {
3627         struct nlmsghdr  *nlh;
3628         u8 scope = RT_SCOPE_UNIVERSE;
3629         int ifindex = ifaca->aca_idev->dev->ifindex;
3630
3631         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3632                 scope = RT_SCOPE_SITE;
3633
3634         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3635         if (nlh == NULL)
3636                 return -EMSGSIZE;
3637
3638         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3639         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3640             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3641                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3642                 nlmsg_cancel(skb, nlh);
3643                 return -EMSGSIZE;
3644         }
3645
3646         return nlmsg_end(skb, nlh);
3647 }
3648
3649 enum addr_type_t {
3650         UNICAST_ADDR,
3651         MULTICAST_ADDR,
3652         ANYCAST_ADDR,
3653 };
3654
3655 /* called with rcu_read_lock() */
3656 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3657                           struct netlink_callback *cb, enum addr_type_t type,
3658                           int s_ip_idx, int *p_ip_idx)
3659 {
3660         struct ifmcaddr6 *ifmca;
3661         struct ifacaddr6 *ifaca;
3662         int err = 1;
3663         int ip_idx = *p_ip_idx;
3664
3665         read_lock_bh(&idev->lock);
3666         switch (type) {
3667         case UNICAST_ADDR: {
3668                 struct inet6_ifaddr *ifa;
3669
3670                 /* unicast address incl. temp addr */
3671                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3672                         if (++ip_idx < s_ip_idx)
3673                                 continue;
3674                         err = inet6_fill_ifaddr(skb, ifa,
3675                                                 NETLINK_CB(cb->skb).pid,
3676                                                 cb->nlh->nlmsg_seq,
3677                                                 RTM_NEWADDR,
3678                                                 NLM_F_MULTI);
3679                         if (err <= 0)
3680                                 break;
3681                 }
3682                 break;
3683         }
3684         case MULTICAST_ADDR:
3685                 /* multicast address */
3686                 for (ifmca = idev->mc_list; ifmca;
3687                      ifmca = ifmca->next, ip_idx++) {
3688                         if (ip_idx < s_ip_idx)
3689                                 continue;
3690                         err = inet6_fill_ifmcaddr(skb, ifmca,
3691                                                   NETLINK_CB(cb->skb).pid,
3692                                                   cb->nlh->nlmsg_seq,
3693                                                   RTM_GETMULTICAST,
3694                                                   NLM_F_MULTI);
3695                         if (err <= 0)
3696                                 break;
3697                 }
3698                 break;
3699         case ANYCAST_ADDR:
3700                 /* anycast address */
3701                 for (ifaca = idev->ac_list; ifaca;
3702                      ifaca = ifaca->aca_next, ip_idx++) {
3703                         if (ip_idx < s_ip_idx)
3704                                 continue;
3705                         err = inet6_fill_ifacaddr(skb, ifaca,
3706                                                   NETLINK_CB(cb->skb).pid,
3707                                                   cb->nlh->nlmsg_seq,
3708                                                   RTM_GETANYCAST,
3709                                                   NLM_F_MULTI);
3710                         if (err <= 0)
3711                                 break;
3712                 }
3713                 break;
3714         default:
3715                 break;
3716         }
3717         read_unlock_bh(&idev->lock);
3718         *p_ip_idx = ip_idx;
3719         return err;
3720 }
3721
3722 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3723                            enum addr_type_t type)
3724 {
3725         struct net *net = sock_net(skb->sk);
3726         int h, s_h;
3727         int idx, ip_idx;
3728         int s_idx, s_ip_idx;
3729         struct net_device *dev;
3730         struct inet6_dev *idev;
3731         struct hlist_head *head;
3732         struct hlist_node *node;
3733
3734         s_h = cb->args[0];
3735         s_idx = idx = cb->args[1];
3736         s_ip_idx = ip_idx = cb->args[2];
3737
3738         rcu_read_lock();
3739         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3740                 idx = 0;
3741                 head = &net->dev_index_head[h];
3742                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3743                         if (idx < s_idx)
3744                                 goto cont;
3745                         if (h > s_h || idx > s_idx)
3746                                 s_ip_idx = 0;
3747                         ip_idx = 0;
3748                         idev = __in6_dev_get(dev);
3749                         if (!idev)
3750                                 goto cont;
3751
3752                         if (in6_dump_addrs(idev, skb, cb, type,
3753                                            s_ip_idx, &ip_idx) <= 0)
3754                                 goto done;
3755 cont:
3756                         idx++;
3757                 }
3758         }
3759 done:
3760         rcu_read_unlock();
3761         cb->args[0] = h;
3762         cb->args[1] = idx;
3763         cb->args[2] = ip_idx;
3764
3765         return skb->len;
3766 }
3767
3768 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3769 {
3770         enum addr_type_t type = UNICAST_ADDR;
3771
3772         return inet6_dump_addr(skb, cb, type);
3773 }
3774
3775 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3776 {
3777         enum addr_type_t type = MULTICAST_ADDR;
3778
3779         return inet6_dump_addr(skb, cb, type);
3780 }
3781
3782
3783 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3784 {
3785         enum addr_type_t type = ANYCAST_ADDR;
3786
3787         return inet6_dump_addr(skb, cb, type);
3788 }
3789
3790 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3791                              void *arg)
3792 {
3793         struct net *net = sock_net(in_skb->sk);
3794         struct ifaddrmsg *ifm;
3795         struct nlattr *tb[IFA_MAX+1];
3796         struct in6_addr *addr = NULL;
3797         struct net_device *dev = NULL;
3798         struct inet6_ifaddr *ifa;
3799         struct sk_buff *skb;
3800         int err;
3801
3802         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3803         if (err < 0)
3804                 goto errout;
3805
3806         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3807         if (addr == NULL) {
3808                 err = -EINVAL;
3809                 goto errout;
3810         }
3811
3812         ifm = nlmsg_data(nlh);
3813         if (ifm->ifa_index)
3814                 dev = __dev_get_by_index(net, ifm->ifa_index);
3815
3816         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3817         if (!ifa) {
3818                 err = -EADDRNOTAVAIL;
3819                 goto errout;
3820         }
3821
3822         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3823         if (!skb) {
3824                 err = -ENOBUFS;
3825                 goto errout_ifa;
3826         }
3827
3828         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3829                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3830         if (err < 0) {
3831                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3832                 WARN_ON(err == -EMSGSIZE);
3833                 kfree_skb(skb);
3834                 goto errout_ifa;
3835         }
3836         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3837 errout_ifa:
3838         in6_ifa_put(ifa);
3839 errout:
3840         return err;
3841 }
3842
3843 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3844 {
3845         struct sk_buff *skb;
3846         struct net *net = dev_net(ifa->idev->dev);
3847         int err = -ENOBUFS;
3848
3849         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3850         if (skb == NULL)
3851                 goto errout;
3852
3853         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3854         if (err < 0) {
3855                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3856                 WARN_ON(err == -EMSGSIZE);
3857                 kfree_skb(skb);
3858                 goto errout;
3859         }
3860         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3861         return;
3862 errout:
3863         if (err < 0)
3864                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3865 }
3866
3867 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3868                                 __s32 *array, int bytes)
3869 {
3870         BUG_ON(bytes < (DEVCONF_MAX * 4));
3871
3872         memset(array, 0, bytes);
3873         array[DEVCONF_FORWARDING] = cnf->forwarding;
3874         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3875         array[DEVCONF_MTU6] = cnf->mtu6;
3876         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3877         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3878         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3879         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3880         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3881         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3882                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3883         array[DEVCONF_RTR_SOLICIT_DELAY] =
3884                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3885         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3886 #ifdef CONFIG_IPV6_PRIVACY
3887         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3888         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3889         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3890         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3891         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3892 #endif
3893         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3894         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3895         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3896 #ifdef CONFIG_IPV6_ROUTER_PREF
3897         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3898         array[DEVCONF_RTR_PROBE_INTERVAL] =
3899                 jiffies_to_msecs(cnf->rtr_probe_interval);
3900 #ifdef CONFIG_IPV6_ROUTE_INFO
3901         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3902 #endif
3903 #endif
3904         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3905         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3906 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3907         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3908 #endif
3909 #ifdef CONFIG_IPV6_MROUTE
3910         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3911 #endif
3912         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3913         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3914         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3915 }
3916
3917 static inline size_t inet6_ifla6_size(void)
3918 {
3919         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3920              + nla_total_size(sizeof(struct ifla_cacheinfo))
3921              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3922              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3923              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3924 }
3925
3926 static inline size_t inet6_if_nlmsg_size(void)
3927 {
3928         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3929                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3930                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3931                + nla_total_size(4) /* IFLA_MTU */
3932                + nla_total_size(4) /* IFLA_LINK */
3933                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3934 }
3935
3936 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3937                                       int items, int bytes)
3938 {
3939         int i;
3940         int pad = bytes - sizeof(u64) * items;
3941         BUG_ON(pad < 0);
3942
3943         /* Use put_unaligned() because stats may not be aligned for u64. */
3944         put_unaligned(items, &stats[0]);
3945         for (i = 1; i < items; i++)
3946                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3947
3948         memset(&stats[items], 0, pad);
3949 }
3950
3951 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3952                                       int items, int bytes, size_t syncpoff)
3953 {
3954         int i;
3955         int pad = bytes - sizeof(u64) * items;
3956         BUG_ON(pad < 0);
3957
3958         /* Use put_unaligned() because stats may not be aligned for u64. */
3959         put_unaligned(items, &stats[0]);
3960         for (i = 1; i < items; i++)
3961                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3962
3963         memset(&stats[items], 0, pad);
3964 }
3965
3966 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3967                              int bytes)
3968 {
3969         switch (attrtype) {
3970         case IFLA_INET6_STATS:
3971                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3972                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3973                 break;
3974         case IFLA_INET6_ICMP6STATS:
3975                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3976                 break;
3977         }
3978 }
3979
3980 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3981 {
3982         struct nlattr *nla;
3983         struct ifla_cacheinfo ci;
3984
3985         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3986
3987         ci.max_reasm_len = IPV6_MAXPLEN;
3988         ci.tstamp = cstamp_delta(idev->tstamp);
3989         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3990         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3991         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3992
3993         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3994         if (nla == NULL)
3995                 goto nla_put_failure;
3996         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3997
3998         /* XXX - MC not implemented */
3999
4000         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
4001         if (nla == NULL)
4002                 goto nla_put_failure;
4003         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
4004
4005         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
4006         if (nla == NULL)
4007                 goto nla_put_failure;
4008         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
4009
4010         return 0;
4011
4012 nla_put_failure:
4013         return -EMSGSIZE;
4014 }
4015
4016 static size_t inet6_get_link_af_size(const struct net_device *dev)
4017 {
4018         if (!__in6_dev_get(dev))
4019                 return 0;
4020
4021         return inet6_ifla6_size();
4022 }
4023
4024 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
4025 {
4026         struct inet6_dev *idev = __in6_dev_get(dev);
4027
4028         if (!idev)
4029                 return -ENODATA;
4030
4031         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4032                 return -EMSGSIZE;
4033
4034         return 0;
4035 }
4036
4037 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
4038                              u32 pid, u32 seq, int event, unsigned int flags)
4039 {
4040         struct net_device *dev = idev->dev;
4041         struct ifinfomsg *hdr;
4042         struct nlmsghdr *nlh;
4043         void *protoinfo;
4044
4045         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
4046         if (nlh == NULL)
4047                 return -EMSGSIZE;
4048
4049         hdr = nlmsg_data(nlh);
4050         hdr->ifi_family = AF_INET6;
4051         hdr->__ifi_pad = 0;
4052         hdr->ifi_type = dev->type;
4053         hdr->ifi_index = dev->ifindex;
4054         hdr->ifi_flags = dev_get_flags(dev);
4055         hdr->ifi_change = 0;
4056
4057         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
4058
4059         if (dev->addr_len)
4060                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
4061
4062         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
4063         if (dev->ifindex != dev->iflink)
4064                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
4065
4066         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
4067         if (protoinfo == NULL)
4068                 goto nla_put_failure;
4069
4070         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
4071                 goto nla_put_failure;
4072
4073         nla_nest_end(skb, protoinfo);
4074         return nlmsg_end(skb, nlh);
4075
4076 nla_put_failure:
4077         nlmsg_cancel(skb, nlh);
4078         return -EMSGSIZE;
4079 }
4080
4081 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4082 {
4083         struct net *net = sock_net(skb->sk);
4084         int h, s_h;
4085         int idx = 0, s_idx;
4086         struct net_device *dev;
4087         struct inet6_dev *idev;
4088         struct hlist_head *head;
4089         struct hlist_node *node;
4090
4091         s_h = cb->args[0];
4092         s_idx = cb->args[1];
4093
4094         rcu_read_lock();
4095         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4096                 idx = 0;
4097                 head = &net->dev_index_head[h];
4098                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4099                         if (idx < s_idx)
4100                                 goto cont;
4101                         idev = __in6_dev_get(dev);
4102                         if (!idev)
4103                                 goto cont;
4104                         if (inet6_fill_ifinfo(skb, idev,
4105                                               NETLINK_CB(cb->skb).pid,
4106                                               cb->nlh->nlmsg_seq,
4107                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4108                                 goto out;
4109 cont:
4110                         idx++;
4111                 }
4112         }
4113 out:
4114         rcu_read_unlock();
4115         cb->args[1] = idx;
4116         cb->args[0] = h;
4117
4118         return skb->len;
4119 }
4120
4121 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4122 {
4123         struct sk_buff *skb;
4124         struct net *net = dev_net(idev->dev);
4125         int err = -ENOBUFS;
4126
4127         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4128         if (skb == NULL)
4129                 goto errout;
4130
4131         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4132         if (err < 0) {
4133                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4134                 WARN_ON(err == -EMSGSIZE);
4135                 kfree_skb(skb);
4136                 goto errout;
4137         }
4138         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4139         return;
4140 errout:
4141         if (err < 0)
4142                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4143 }
4144
4145 static inline size_t inet6_prefix_nlmsg_size(void)
4146 {
4147         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4148                + nla_total_size(sizeof(struct in6_addr))
4149                + nla_total_size(sizeof(struct prefix_cacheinfo));
4150 }
4151
4152 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4153                              struct prefix_info *pinfo, u32 pid, u32 seq,
4154                              int event, unsigned int flags)
4155 {
4156         struct prefixmsg *pmsg;
4157         struct nlmsghdr *nlh;
4158         struct prefix_cacheinfo ci;
4159
4160         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4161         if (nlh == NULL)
4162                 return -EMSGSIZE;
4163
4164         pmsg = nlmsg_data(nlh);
4165         pmsg->prefix_family = AF_INET6;
4166         pmsg->prefix_pad1 = 0;
4167         pmsg->prefix_pad2 = 0;
4168         pmsg->prefix_ifindex = idev->dev->ifindex;
4169         pmsg->prefix_len = pinfo->prefix_len;
4170         pmsg->prefix_type = pinfo->type;
4171         pmsg->prefix_pad3 = 0;
4172         pmsg->prefix_flags = 0;
4173         if (pinfo->onlink)
4174                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4175         if (pinfo->autoconf)
4176                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4177
4178         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4179
4180         ci.preferred_time = ntohl(pinfo->prefered);
4181         ci.valid_time = ntohl(pinfo->valid);
4182         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4183
4184         return nlmsg_end(skb, nlh);
4185
4186 nla_put_failure:
4187         nlmsg_cancel(skb, nlh);
4188         return -EMSGSIZE;
4189 }
4190
4191 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4192                          struct prefix_info *pinfo)
4193 {
4194         struct sk_buff *skb;
4195         struct net *net = dev_net(idev->dev);
4196         int err = -ENOBUFS;
4197
4198         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4199         if (skb == NULL)
4200                 goto errout;
4201
4202         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4203         if (err < 0) {
4204                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4205                 WARN_ON(err == -EMSGSIZE);
4206                 kfree_skb(skb);
4207                 goto errout;
4208         }
4209         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4210         return;
4211 errout:
4212         if (err < 0)
4213                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4214 }
4215
4216 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4217 {
4218         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4219
4220         switch (event) {
4221         case RTM_NEWADDR:
4222                 /*
4223                  * If the address was optimistic
4224                  * we inserted the route at the start of
4225                  * our DAD process, so we don't need
4226                  * to do it again
4227                  */
4228                 if (!(ifp->rt->rt6i_node))
4229                         ip6_ins_rt(ifp->rt);
4230                 if (ifp->idev->cnf.forwarding)
4231                         addrconf_join_anycast(ifp);
4232                 break;
4233         case RTM_DELADDR:
4234                 if (ifp->idev->cnf.forwarding)
4235                         addrconf_leave_anycast(ifp);
4236                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4237                 dst_hold(&ifp->rt->dst);
4238
4239                 if (ip6_del_rt(ifp->rt))
4240                         dst_free(&ifp->rt->dst);
4241                 break;
4242         }
4243 }
4244
4245 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4246 {
4247         rcu_read_lock_bh();
4248         if (likely(ifp->idev->dead == 0))
4249                 __ipv6_ifa_notify(event, ifp);
4250         rcu_read_unlock_bh();
4251 }
4252
4253 #ifdef CONFIG_SYSCTL
4254
4255 static
4256 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4257                            void __user *buffer, size_t *lenp, loff_t *ppos)
4258 {
4259         int *valp = ctl->data;
4260         int val = *valp;
4261         loff_t pos = *ppos;
4262         int ret;
4263
4264         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4265
4266         if (write)
4267                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4268         if (ret)
4269                 *ppos = pos;
4270         return ret;
4271 }
4272
4273 static void dev_disable_change(struct inet6_dev *idev)
4274 {
4275         if (!idev || !idev->dev)
4276                 return;
4277
4278         if (idev->cnf.disable_ipv6)
4279                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4280         else
4281                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4282 }
4283
4284 static void addrconf_disable_change(struct net *net, __s32 newf)
4285 {
4286         struct net_device *dev;
4287         struct inet6_dev *idev;
4288
4289         rcu_read_lock();
4290         for_each_netdev_rcu(net, dev) {
4291                 idev = __in6_dev_get(dev);
4292                 if (idev) {
4293                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4294                         idev->cnf.disable_ipv6 = newf;
4295                         if (changed)
4296                                 dev_disable_change(idev);
4297                 }
4298         }
4299         rcu_read_unlock();
4300 }
4301
4302 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4303 {
4304         struct net *net;
4305
4306         net = (struct net *)table->extra2;
4307
4308         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4309                 return 0;
4310
4311         if (!rtnl_trylock()) {
4312                 /* Restore the original values before restarting */
4313                 *p = old;
4314                 return restart_syscall();
4315         }
4316
4317         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4318                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4319                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4320                 addrconf_disable_change(net, newf);
4321         } else if ((!*p) ^ (!old))
4322                 dev_disable_change((struct inet6_dev *)table->extra1);
4323
4324         rtnl_unlock();
4325         return 0;
4326 }
4327
4328 static
4329 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4330                             void __user *buffer, size_t *lenp, loff_t *ppos)
4331 {
4332         int *valp = ctl->data;
4333         int val = *valp;
4334         loff_t pos = *ppos;
4335         int ret;
4336
4337         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4338
4339         if (write)
4340                 ret = addrconf_disable_ipv6(ctl, valp, val);
4341         if (ret)
4342                 *ppos = pos;
4343         return ret;
4344 }
4345
4346 static struct addrconf_sysctl_table
4347 {
4348         struct ctl_table_header *sysctl_header;
4349         ctl_table addrconf_vars[DEVCONF_MAX+1];
4350         char *dev_name;
4351 } addrconf_sysctl __read_mostly = {
4352         .sysctl_header = NULL,
4353         .addrconf_vars = {
4354                 {
4355                         .procname       = "forwarding",
4356                         .data           = &ipv6_devconf.forwarding,
4357                         .maxlen         = sizeof(int),
4358                         .mode           = 0644,
4359                         .proc_handler   = addrconf_sysctl_forward,
4360                 },
4361                 {
4362                         .procname       = "hop_limit",
4363                         .data           = &ipv6_devconf.hop_limit,
4364                         .maxlen         = sizeof(int),
4365                         .mode           = 0644,
4366                         .proc_handler   = proc_dointvec,
4367                 },
4368                 {
4369                         .procname       = "mtu",
4370                         .data           = &ipv6_devconf.mtu6,
4371                         .maxlen         = sizeof(int),
4372                         .mode           = 0644,
4373                         .proc_handler   = proc_dointvec,
4374                 },
4375                 {
4376                         .procname       = "accept_ra",
4377                         .data           = &ipv6_devconf.accept_ra,
4378                         .maxlen         = sizeof(int),
4379                         .mode           = 0644,
4380                         .proc_handler   = proc_dointvec,
4381                 },
4382                 {
4383                         .procname       = "accept_redirects",
4384                         .data           = &ipv6_devconf.accept_redirects,
4385                         .maxlen         = sizeof(int),
4386                         .mode           = 0644,
4387                         .proc_handler   = proc_dointvec,
4388                 },
4389                 {
4390                         .procname       = "autoconf",
4391                         .data           = &ipv6_devconf.autoconf,
4392                         .maxlen         = sizeof(int),
4393                         .mode           = 0644,
4394                         .proc_handler   = proc_dointvec,
4395                 },
4396                 {
4397                         .procname       = "dad_transmits",
4398                         .data           = &ipv6_devconf.dad_transmits,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec,
4402                 },
4403                 {
4404                         .procname       = "router_solicitations",
4405                         .data           = &ipv6_devconf.rtr_solicits,
4406                         .maxlen         = sizeof(int),
4407                         .mode           = 0644,
4408                         .proc_handler   = proc_dointvec,
4409                 },
4410                 {
4411                         .procname       = "router_solicitation_interval",
4412                         .data           = &ipv6_devconf.rtr_solicit_interval,
4413                         .maxlen         = sizeof(int),
4414                         .mode           = 0644,
4415                         .proc_handler   = proc_dointvec_jiffies,
4416                 },
4417                 {
4418                         .procname       = "router_solicitation_delay",
4419                         .data           = &ipv6_devconf.rtr_solicit_delay,
4420                         .maxlen         = sizeof(int),
4421                         .mode           = 0644,
4422                         .proc_handler   = proc_dointvec_jiffies,
4423                 },
4424                 {
4425                         .procname       = "force_mld_version",
4426                         .data           = &ipv6_devconf.force_mld_version,
4427                         .maxlen         = sizeof(int),
4428                         .mode           = 0644,
4429                         .proc_handler   = proc_dointvec,
4430                 },
4431 #ifdef CONFIG_IPV6_PRIVACY
4432                 {
4433                         .procname       = "use_tempaddr",
4434                         .data           = &ipv6_devconf.use_tempaddr,
4435                         .maxlen         = sizeof(int),
4436                         .mode           = 0644,
4437                         .proc_handler   = proc_dointvec,
4438                 },
4439                 {
4440                         .procname       = "temp_valid_lft",
4441                         .data           = &ipv6_devconf.temp_valid_lft,
4442                         .maxlen         = sizeof(int),
4443                         .mode           = 0644,
4444                         .proc_handler   = proc_dointvec,
4445                 },
4446                 {
4447                         .procname       = "temp_prefered_lft",
4448                         .data           = &ipv6_devconf.temp_prefered_lft,
4449                         .maxlen         = sizeof(int),
4450                         .mode           = 0644,
4451                         .proc_handler   = proc_dointvec,
4452                 },
4453                 {
4454                         .procname       = "regen_max_retry",
4455                         .data           = &ipv6_devconf.regen_max_retry,
4456                         .maxlen         = sizeof(int),
4457                         .mode           = 0644,
4458                         .proc_handler   = proc_dointvec,
4459                 },
4460                 {
4461                         .procname       = "max_desync_factor",
4462                         .data           = &ipv6_devconf.max_desync_factor,
4463                         .maxlen         = sizeof(int),
4464                         .mode           = 0644,
4465                         .proc_handler   = proc_dointvec,
4466                 },
4467 #endif
4468                 {
4469                         .procname       = "max_addresses",
4470                         .data           = &ipv6_devconf.max_addresses,
4471                         .maxlen         = sizeof(int),
4472                         .mode           = 0644,
4473                         .proc_handler   = proc_dointvec,
4474                 },
4475                 {
4476                         .procname       = "accept_ra_defrtr",
4477                         .data           = &ipv6_devconf.accept_ra_defrtr,
4478                         .maxlen         = sizeof(int),
4479                         .mode           = 0644,
4480                         .proc_handler   = proc_dointvec,
4481                 },
4482                 {
4483                         .procname       = "accept_ra_pinfo",
4484                         .data           = &ipv6_devconf.accept_ra_pinfo,
4485                         .maxlen         = sizeof(int),
4486                         .mode           = 0644,
4487                         .proc_handler   = proc_dointvec,
4488                 },
4489 #ifdef CONFIG_IPV6_ROUTER_PREF
4490                 {
4491                         .procname       = "accept_ra_rtr_pref",
4492                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4493                         .maxlen         = sizeof(int),
4494                         .mode           = 0644,
4495                         .proc_handler   = proc_dointvec,
4496                 },
4497                 {
4498                         .procname       = "router_probe_interval",
4499                         .data           = &ipv6_devconf.rtr_probe_interval,
4500                         .maxlen         = sizeof(int),
4501                         .mode           = 0644,
4502                         .proc_handler   = proc_dointvec_jiffies,
4503                 },
4504 #ifdef CONFIG_IPV6_ROUTE_INFO
4505                 {
4506                         .procname       = "accept_ra_rt_info_max_plen",
4507                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4508                         .maxlen         = sizeof(int),
4509                         .mode           = 0644,
4510                         .proc_handler   = proc_dointvec,
4511                 },
4512 #endif
4513 #endif
4514                 {
4515                         .procname       = "proxy_ndp",
4516                         .data           = &ipv6_devconf.proxy_ndp,
4517                         .maxlen         = sizeof(int),
4518                         .mode           = 0644,
4519                         .proc_handler   = proc_dointvec,
4520                 },
4521                 {
4522                         .procname       = "accept_source_route",
4523                         .data           = &ipv6_devconf.accept_source_route,
4524                         .maxlen         = sizeof(int),
4525                         .mode           = 0644,
4526                         .proc_handler   = proc_dointvec,
4527                 },
4528 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4529                 {
4530                         .procname       = "optimistic_dad",
4531                         .data           = &ipv6_devconf.optimistic_dad,
4532                         .maxlen         = sizeof(int),
4533                         .mode           = 0644,
4534                         .proc_handler   = proc_dointvec,
4535
4536                 },
4537 #endif
4538 #ifdef CONFIG_IPV6_MROUTE
4539                 {
4540                         .procname       = "mc_forwarding",
4541                         .data           = &ipv6_devconf.mc_forwarding,
4542                         .maxlen         = sizeof(int),
4543                         .mode           = 0444,
4544                         .proc_handler   = proc_dointvec,
4545                 },
4546 #endif
4547                 {
4548                         .procname       = "disable_ipv6",
4549                         .data           = &ipv6_devconf.disable_ipv6,
4550                         .maxlen         = sizeof(int),
4551                         .mode           = 0644,
4552                         .proc_handler   = addrconf_sysctl_disable,
4553                 },
4554                 {
4555                         .procname       = "accept_dad",
4556                         .data           = &ipv6_devconf.accept_dad,
4557                         .maxlen         = sizeof(int),
4558                         .mode           = 0644,
4559                         .proc_handler   = proc_dointvec,
4560                 },
4561                 {
4562                         .procname       = "force_tllao",
4563                         .data           = &ipv6_devconf.force_tllao,
4564                         .maxlen         = sizeof(int),
4565                         .mode           = 0644,
4566                         .proc_handler   = proc_dointvec
4567                 },
4568                 {
4569                         /* sentinel */
4570                 }
4571         },
4572 };
4573
4574 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4575                 struct inet6_dev *idev, struct ipv6_devconf *p)
4576 {
4577         int i;
4578         struct addrconf_sysctl_table *t;
4579
4580 #define ADDRCONF_CTL_PATH_DEV   3
4581
4582         struct ctl_path addrconf_ctl_path[] = {
4583                 { .procname = "net", },
4584                 { .procname = "ipv6", },
4585                 { .procname = "conf", },
4586                 { /* to be set */ },
4587                 { },
4588         };
4589
4590
4591         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4592         if (t == NULL)
4593                 goto out;
4594
4595         for (i = 0; t->addrconf_vars[i].data; i++) {
4596                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4597                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4598                 t->addrconf_vars[i].extra2 = net;
4599         }
4600
4601         /*
4602          * Make a copy of dev_name, because '.procname' is regarded as const
4603          * by sysctl and we wouldn't want anyone to change it under our feet
4604          * (see SIOCSIFNAME).
4605          */
4606         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4607         if (!t->dev_name)
4608                 goto free;
4609
4610         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4611
4612         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4613                         t->addrconf_vars);
4614         if (t->sysctl_header == NULL)
4615                 goto free_procname;
4616
4617         p->sysctl = t;
4618         return 0;
4619
4620 free_procname:
4621         kfree(t->dev_name);
4622 free:
4623         kfree(t);
4624 out:
4625         return -ENOBUFS;
4626 }
4627
4628 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4629 {
4630         struct addrconf_sysctl_table *t;
4631
4632         if (p->sysctl == NULL)
4633                 return;
4634
4635         t = p->sysctl;
4636         p->sysctl = NULL;
4637         unregister_net_sysctl_table(t->sysctl_header);
4638         kfree(t->dev_name);
4639         kfree(t);
4640 }
4641
4642 static void addrconf_sysctl_register(struct inet6_dev *idev)
4643 {
4644         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4645                               &ndisc_ifinfo_sysctl_change);
4646         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4647                                         idev, &idev->cnf);
4648 }
4649
4650 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4651 {
4652         __addrconf_sysctl_unregister(&idev->cnf);
4653         neigh_sysctl_unregister(idev->nd_parms);
4654 }
4655
4656
4657 #endif
4658
4659 static int __net_init addrconf_init_net(struct net *net)
4660 {
4661         int err;
4662         struct ipv6_devconf *all, *dflt;
4663
4664         err = -ENOMEM;
4665         all = &ipv6_devconf;
4666         dflt = &ipv6_devconf_dflt;
4667
4668         if (!net_eq(net, &init_net)) {
4669                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4670                 if (all == NULL)
4671                         goto err_alloc_all;
4672
4673                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4674                 if (dflt == NULL)
4675                         goto err_alloc_dflt;
4676         } else {
4677                 /* these will be inherited by all namespaces */
4678                 dflt->autoconf = ipv6_defaults.autoconf;
4679                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4680         }
4681
4682         net->ipv6.devconf_all = all;
4683         net->ipv6.devconf_dflt = dflt;
4684
4685 #ifdef CONFIG_SYSCTL
4686         err = __addrconf_sysctl_register(net, "all", NULL, all);
4687         if (err < 0)
4688                 goto err_reg_all;
4689
4690         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4691         if (err < 0)
4692                 goto err_reg_dflt;
4693 #endif
4694         return 0;
4695
4696 #ifdef CONFIG_SYSCTL
4697 err_reg_dflt:
4698         __addrconf_sysctl_unregister(all);
4699 err_reg_all:
4700         kfree(dflt);
4701 #endif
4702 err_alloc_dflt:
4703         kfree(all);
4704 err_alloc_all:
4705         return err;
4706 }
4707
4708 static void __net_exit addrconf_exit_net(struct net *net)
4709 {
4710 #ifdef CONFIG_SYSCTL
4711         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4712         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4713 #endif
4714         if (!net_eq(net, &init_net)) {
4715                 kfree(net->ipv6.devconf_dflt);
4716                 kfree(net->ipv6.devconf_all);
4717         }
4718 }
4719
4720 static struct pernet_operations addrconf_ops = {
4721         .init = addrconf_init_net,
4722         .exit = addrconf_exit_net,
4723 };
4724
4725 /*
4726  *      Device notifier
4727  */
4728
4729 int register_inet6addr_notifier(struct notifier_block *nb)
4730 {
4731         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4732 }
4733 EXPORT_SYMBOL(register_inet6addr_notifier);
4734
4735 int unregister_inet6addr_notifier(struct notifier_block *nb)
4736 {
4737         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4738 }
4739 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4740
4741 static struct rtnl_af_ops inet6_ops = {
4742         .family           = AF_INET6,
4743         .fill_link_af     = inet6_fill_link_af,
4744         .get_link_af_size = inet6_get_link_af_size,
4745 };
4746
4747 /*
4748  *      Init / cleanup code
4749  */
4750
4751 int __init addrconf_init(void)
4752 {
4753         int i, err;
4754
4755         err = ipv6_addr_label_init();
4756         if (err < 0) {
4757                 printk(KERN_CRIT "IPv6 Addrconf:"
4758                        " cannot initialize default policy table: %d.\n", err);
4759                 goto out;
4760         }
4761
4762         err = register_pernet_subsys(&addrconf_ops);
4763         if (err < 0)
4764                 goto out_addrlabel;
4765
4766         /* The addrconf netdev notifier requires that loopback_dev
4767          * has it's ipv6 private information allocated and setup
4768          * before it can bring up and give link-local addresses
4769          * to other devices which are up.
4770          *
4771          * Unfortunately, loopback_dev is not necessarily the first
4772          * entry in the global dev_base list of net devices.  In fact,
4773          * it is likely to be the very last entry on that list.
4774          * So this causes the notifier registry below to try and
4775          * give link-local addresses to all devices besides loopback_dev
4776          * first, then loopback_dev, which cases all the non-loopback_dev
4777          * devices to fail to get a link-local address.
4778          *
4779          * So, as a temporary fix, allocate the ipv6 structure for
4780          * loopback_dev first by hand.
4781          * Longer term, all of the dependencies ipv6 has upon the loopback
4782          * device and it being up should be removed.
4783          */
4784         rtnl_lock();
4785         if (!ipv6_add_dev(init_net.loopback_dev))
4786                 err = -ENOMEM;
4787         rtnl_unlock();
4788         if (err)
4789                 goto errlo;
4790
4791         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4792                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4793
4794         register_netdevice_notifier(&ipv6_dev_notf);
4795
4796         addrconf_verify(0);
4797
4798         err = rtnl_af_register(&inet6_ops);
4799         if (err < 0)
4800                 goto errout_af;
4801
4802         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4803                               NULL);
4804         if (err < 0)
4805                 goto errout;
4806
4807         /* Only the first call to __rtnl_register can fail */
4808         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4809         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4810         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4811                         inet6_dump_ifaddr, NULL);
4812         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4813                         inet6_dump_ifmcaddr, NULL);
4814         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4815                         inet6_dump_ifacaddr, NULL);
4816
4817         ipv6_addr_label_rtnl_register();
4818
4819         return 0;
4820 errout:
4821         rtnl_af_unregister(&inet6_ops);
4822 errout_af:
4823         unregister_netdevice_notifier(&ipv6_dev_notf);
4824 errlo:
4825         unregister_pernet_subsys(&addrconf_ops);
4826 out_addrlabel:
4827         ipv6_addr_label_cleanup();
4828 out:
4829         return err;
4830 }
4831
4832 void addrconf_cleanup(void)
4833 {
4834         struct net_device *dev;
4835         int i;
4836
4837         unregister_netdevice_notifier(&ipv6_dev_notf);
4838         unregister_pernet_subsys(&addrconf_ops);
4839         ipv6_addr_label_cleanup();
4840
4841         rtnl_lock();
4842
4843         __rtnl_af_unregister(&inet6_ops);
4844
4845         /* clean dev list */
4846         for_each_netdev(&init_net, dev) {
4847                 if (__in6_dev_get(dev) == NULL)
4848                         continue;
4849                 addrconf_ifdown(dev, 1);
4850         }
4851         addrconf_ifdown(init_net.loopback_dev, 2);
4852
4853         /*
4854          *      Check hash table.
4855          */
4856         spin_lock_bh(&addrconf_hash_lock);
4857         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4858                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4859         spin_unlock_bh(&addrconf_hash_lock);
4860
4861         del_timer(&addr_chk_timer);
4862         rtnl_unlock();
4863 }