net/ipv6/addrconf.c

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