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         return ndev;
 437 }
 438
 439 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 440 {
 441         struct inet6_dev *idev;
 442
 443         ASSERT_RTNL();
 444
 445         idev = __in6_dev_get(dev);
 446         if (!idev) {
 447                 idev = ipv6_add_dev(dev);
 448                 if (!idev)
 449                         return NULL;
 450         }
 451
 452         if (dev->flags&IFF_UP)
 453                 ipv6_mc_up(idev);
 454         return idev;
 455 }
 456
 457 #ifdef CONFIG_SYSCTL
 458 static void dev_forward_change(struct inet6_dev *idev)
 459 {
 460         struct net_device *dev;
 461         struct inet6_ifaddr *ifa;
 462
 463         if (!idev)
 464                 return;
 465         dev = idev->dev;
 466         if (idev->cnf.forwarding)
 467                 dev_disable_lro(dev);
 468         if (dev && (dev->flags & IFF_MULTICAST)) {
 469                 if (idev->cnf.forwarding)
 470                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 471                 else
 472                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 473         }
 474
 475         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 476                 if (ifa->flags&IFA_F_TENTATIVE)
 477                         continue;
 478                 if (idev->cnf.forwarding)
 479                         addrconf_join_anycast(ifa);
 480                 else
 481                         addrconf_leave_anycast(ifa);
 482         }
 483 }
 484
 485
 486 static void addrconf_forward_change(struct net *net, __s32 newf)
 487 {
 488         struct net_device *dev;
 489         struct inet6_dev *idev;
 490
 491         rcu_read_lock();
 492         for_each_netdev_rcu(net, dev) {
 493                 idev = __in6_dev_get(dev);
 494                 if (idev) {
 495                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 496                         idev->cnf.forwarding = newf;
 497                         if (changed)
 498                                 dev_forward_change(idev);
 499                 }
 500         }
 501         rcu_read_unlock();
 502 }
 503
 504 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 505 {
 506         struct net *net;
 507
 508         net = (struct net *)table->extra2;
 509         if (p == &net->ipv6.devconf_dflt->forwarding)
 510                 return 0;
 511
 512         if (!rtnl_trylock()) {
 513                 /* Restore the original values before restarting */
 514                 *p = old;
 515                 return restart_syscall();
 516         }
 517
 518         if (p == &net->ipv6.devconf_all->forwarding) {
 519                 __s32 newf = net->ipv6.devconf_all->forwarding;
 520                 net->ipv6.devconf_dflt->forwarding = newf;
 521                 addrconf_forward_change(net, newf);
 522         } else if ((!*p) ^ (!old))
 523                 dev_forward_change((struct inet6_dev *)table->extra1);
 524         rtnl_unlock();
 525
 526         if (*p)
 527                 rt6_purge_dflt_routers(net);
 528         return 1;
 529 }
 530 #endif
 531
 532 /* Nobody refers to this ifaddr, destroy it */
 533 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 534 {
 535         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 536
 537 #ifdef NET_REFCNT_DEBUG
 538         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 539 #endif
 540
 541         in6_dev_put(ifp->idev);
 542
 543         if (del_timer(&ifp->timer))
 544                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 545
 546         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 547                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 548                 return;
 549         }
 550         dst_release(&ifp->rt->dst);
 551
 552         kfree_rcu(ifp, rcu);
 553 }
 554
 555 static void
 556 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 557 {
 558         struct list_head *p;
 559         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 560
 561         /*
 562          * Each device address list is sorted in order of scope -
 563          * global before linklocal.
 564          */
 565         list_for_each(p, &idev->addr_list) {
 566                 struct inet6_ifaddr *ifa
 567                         = list_entry(p, struct inet6_ifaddr, if_list);
 568                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 569                         break;
 570         }
 571
 572         list_add_tail(&ifp->if_list, p);
 573 }
 574
 575 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 576 {
 577         /*
 578          * We perform the hash function over the last 64 bits of the address
 579          * This will include the IEEE address token on links that support it.
 580          */
 581         return jhash_2words((__force u32)addr->s6_addr32[2],
 582                             (__force u32)addr->s6_addr32[3], 0)
 583                 & (IN6_ADDR_HSIZE - 1);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         unsigned int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         spin_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         spin_lock_init(&ifa->state_lock);
 642         init_timer(&ifa->timer);
 643         INIT_HLIST_NODE(&ifa->addr_lst);
 644         ifa->timer.data = (unsigned long) ifa;
 645         ifa->scope = scope;
 646         ifa->prefix_len = pfxlen;
 647         ifa->flags = flags | IFA_F_TENTATIVE;
 648         ifa->cstamp = ifa->tstamp = jiffies;
 649
 650         ifa->rt = rt;
 651
 652         /*
 653          * part one of RFC 4429, section 3.3
 654          * We should not configure an address as
 655          * optimistic if we do not yet know the link
 656          * layer address of our nexhop router
 657          */
 658
 659         if (rt->rt6i_nexthop == NULL)
 660                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 661
 662         ifa->idev = idev;
 663         in6_dev_hold(idev);
 664         /* For caller */
 665         in6_ifa_hold(ifa);
 666
 667         /* Add to big hash table */
 668         hash = ipv6_addr_hash(addr);
 669
 670         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 671         spin_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 680                 in6_ifa_hold(ifa);
 681         }
 682 #endif
 683
 684         in6_ifa_hold(ifa);
 685         write_unlock(&idev->lock);
 686 out2:
 687         rcu_read_unlock_bh();
 688
 689         if (likely(err == 0))
 690                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 691         else {
 692                 kfree(ifa);
 693                 ifa = ERR_PTR(err);
 694         }
 695
 696         return ifa;
 697 out:
 698         spin_unlock(&addrconf_hash_lock);
 699         goto out2;
 700 }
 701
 702 /* This function wants to get referenced ifp and releases it before return */
 703
 704 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 705 {
 706         struct inet6_ifaddr *ifa, *ifn;
 707         struct inet6_dev *idev = ifp->idev;
 708         int state;
 709         int deleted = 0, onlink = 0;
 710         unsigned long expires = jiffies;
 711
 712         spin_lock_bh(&ifp->state_lock);
 713         state = ifp->state;
 714         ifp->state = INET6_IFADDR_STATE_DEAD;
 715         spin_unlock_bh(&ifp->state_lock);
 716
 717         if (state == INET6_IFADDR_STATE_DEAD)
 718                 goto out;
 719
 720         spin_lock_bh(&addrconf_hash_lock);
 721         hlist_del_init_rcu(&ifp->addr_lst);
 722         spin_unlock_bh(&addrconf_hash_lock);
 723
 724         write_lock_bh(&idev->lock);
 725 #ifdef CONFIG_IPV6_PRIVACY
 726         if (ifp->flags&IFA_F_TEMPORARY) {
 727                 list_del(&ifp->tmp_list);
 728                 if (ifp->ifpub) {
 729                         in6_ifa_put(ifp->ifpub);
 730                         ifp->ifpub = NULL;
 731                 }
 732                 __in6_ifa_put(ifp);
 733         }
 734 #endif
 735
 736         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 737                 if (ifa == ifp) {
 738                         list_del_init(&ifp->if_list);
 739                         __in6_ifa_put(ifp);
 740
 741                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 742                                 break;
 743                         deleted = 1;
 744                         continue;
 745                 } else if (ifp->flags & IFA_F_PERMANENT) {
 746                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 747                                               ifp->prefix_len)) {
 748                                 if (ifa->flags & IFA_F_PERMANENT) {
 749                                         onlink = 1;
 750                                         if (deleted)
 751                                                 break;
 752                                 } else {
 753                                         unsigned long lifetime;
 754
 755                                         if (!onlink)
 756                                                 onlink = -1;
 757
 758                                         spin_lock(&ifa->lock);
 759
 760                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 761                                         /*
 762                                          * Note: Because this address is
 763                                          * not permanent, lifetime <
 764                                          * LONG_MAX / HZ here.
 765                                          */
 766                                         if (time_before(expires,
 767                                                         ifa->tstamp + lifetime * HZ))
 768                                                 expires = ifa->tstamp + lifetime * HZ;
 769                                         spin_unlock(&ifa->lock);
 770                                 }
 771                         }
 772                 }
 773         }
 774         write_unlock_bh(&idev->lock);
 775
 776         addrconf_del_timer(ifp);
 777
 778         ipv6_ifa_notify(RTM_DELADDR, ifp);
 779
 780         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 781
 782         /*
 783          * Purge or update corresponding prefix
 784          *
 785          * 1) we don't purge prefix here if address was not permanent.
 786          *    prefix is managed by its own lifetime.
 787          * 2) if there're no addresses, delete prefix.
 788          * 3) if there're still other permanent address(es),
 789          *    corresponding prefix is still permanent.
 790          * 4) otherwise, update prefix lifetime to the
 791          *    longest valid lifetime among the corresponding
 792          *    addresses on the device.
 793          *    Note: subsequent RA will update lifetime.
 794          *
 795          * --yoshfuji
 796          */
 797         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 798                 struct in6_addr prefix;
 799                 struct rt6_info *rt;
 800                 struct net *net = dev_net(ifp->idev->dev);
 801                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 802                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 803
 804                 if (rt && addrconf_is_prefix_route(rt)) {
 805                         if (onlink == 0) {
 806                                 ip6_del_rt(rt);
 807                                 rt = NULL;
 808                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 809                                 rt->rt6i_expires = expires;
 810                                 rt->rt6i_flags |= RTF_EXPIRES;
 811                         }
 812                 }
 813                 dst_release(&rt->dst);
 814         }
 815
 816         /* clean up prefsrc entries */
 817         rt6_remove_prefsrc(ifp);
 818 out:
 819         in6_ifa_put(ifp);
 820 }
 821
 822 #ifdef CONFIG_IPV6_PRIVACY
 823 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 824 {
 825         struct inet6_dev *idev = ifp->idev;
 826         struct in6_addr addr, *tmpaddr;
 827         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp, age;
 828         unsigned long regen_advance;
 829         int tmp_plen;
 830         int ret = 0;
 831         int max_addresses;
 832         u32 addr_flags;
 833
 834         write_lock(&idev->lock);
 835         if (ift) {
 836                 spin_lock_bh(&ift->lock);
 837                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 838                 spin_unlock_bh(&ift->lock);
 839                 tmpaddr = &addr;
 840         } else {
 841                 tmpaddr = NULL;
 842         }
 843 retry:
 844         in6_dev_hold(idev);
 845         if (idev->cnf.use_tempaddr <= 0) {
 846                 write_unlock(&idev->lock);
 847                 printk(KERN_INFO
 848                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 849                 in6_dev_put(idev);
 850                 ret = -1;
 851                 goto out;
 852         }
 853         spin_lock_bh(&ifp->lock);
 854         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 855                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 856                 spin_unlock_bh(&ifp->lock);
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_WARNING
 859                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         in6_ifa_hold(ifp);
 865         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 866         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 871                 in6_ifa_put(ifp);
 872                 in6_dev_put(idev);
 873                 ret = -1;
 874                 goto out;
 875         }
 876         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 877         age = (jiffies - ifp->tstamp) / HZ;
 878         tmp_valid_lft = min_t(__u32,
 879                               ifp->valid_lft,
 880                               idev->cnf.temp_valid_lft + age);
 881         tmp_prefered_lft = min_t(__u32,
 882                                  ifp->prefered_lft,
 883                                  idev->cnf.temp_prefered_lft + age -
 884                                  idev->cnf.max_desync_factor);
 885         tmp_plen = ifp->prefix_len;
 886         max_addresses = idev->cnf.max_addresses;
 887         tmp_cstamp = ifp->cstamp;
 888         tmp_tstamp = ifp->tstamp;
 889         spin_unlock_bh(&ifp->lock);
 890
 891         regen_advance = idev->cnf.regen_max_retry *
 892                         idev->cnf.dad_transmits *
 893                         idev->nd_parms->retrans_time / HZ;
 894         write_unlock(&idev->lock);
 895
 896         /* A temporary address is created only if this calculated Preferred
 897          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 898          * an implementation must not create a temporary address with a zero
 899          * Preferred Lifetime.
 900          */
 901         if (tmp_prefered_lft <= regen_advance) {
 902                 in6_ifa_put(ifp);
 903                 in6_dev_put(idev);
 904                 ret = -1;
 905                 goto out;
 906         }
 907
 908         addr_flags = IFA_F_TEMPORARY;
 909         /* set in addrconf_prefix_rcv() */
 910         if (ifp->flags & IFA_F_OPTIMISTIC)
 911                 addr_flags |= IFA_F_OPTIMISTIC;
 912
 913         ift = !max_addresses ||
 914               ipv6_count_addresses(idev) < max_addresses ?
 915                 ipv6_add_addr(idev, &addr, tmp_plen,
 916                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 917                               addr_flags) : NULL;
 918         if (!ift || IS_ERR(ift)) {
 919                 in6_ifa_put(ifp);
 920                 in6_dev_put(idev);
 921                 printk(KERN_INFO
 922                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 923                 tmpaddr = &addr;
 924                 write_lock(&idev->lock);
 925                 goto retry;
 926         }
 927
 928         spin_lock_bh(&ift->lock);
 929         ift->ifpub = ifp;
 930         ift->valid_lft = tmp_valid_lft;
 931         ift->prefered_lft = tmp_prefered_lft;
 932         ift->cstamp = tmp_cstamp;
 933         ift->tstamp = tmp_tstamp;
 934         spin_unlock_bh(&ift->lock);
 935
 936         addrconf_dad_start(ift, 0);
 937         in6_ifa_put(ift);
 938         in6_dev_put(idev);
 939 out:
 940         return ret;
 941 }
 942 #endif
 943
 944 /*
 945  *      Choose an appropriate source address (RFC3484)
 946  */
 947 enum {
 948         IPV6_SADDR_RULE_INIT = 0,
 949         IPV6_SADDR_RULE_LOCAL,
 950         IPV6_SADDR_RULE_SCOPE,
 951         IPV6_SADDR_RULE_PREFERRED,
 952 #ifdef CONFIG_IPV6_MIP6
 953         IPV6_SADDR_RULE_HOA,
 954 #endif
 955         IPV6_SADDR_RULE_OIF,
 956         IPV6_SADDR_RULE_LABEL,
 957 #ifdef CONFIG_IPV6_PRIVACY
 958         IPV6_SADDR_RULE_PRIVACY,
 959 #endif
 960         IPV6_SADDR_RULE_ORCHID,
 961         IPV6_SADDR_RULE_PREFIX,
 962         IPV6_SADDR_RULE_MAX
 963 };
 964
 965 struct ipv6_saddr_score {
 966         int                     rule;
 967         int                     addr_type;
 968         struct inet6_ifaddr     *ifa;
 969         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 970         int                     scopedist;
 971         int                     matchlen;
 972 };
 973
 974 struct ipv6_saddr_dst {
 975         const struct in6_addr *addr;
 976         int ifindex;
 977         int scope;
 978         int label;
 979         unsigned int prefs;
 980 };
 981
 982 static inline int ipv6_saddr_preferred(int type)
 983 {
 984         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 985                 return 1;
 986         return 0;
 987 }
 988
 989 static int ipv6_get_saddr_eval(struct net *net,
 990                                struct ipv6_saddr_score *score,
 991                                struct ipv6_saddr_dst *dst,
 992                                int i)
 993 {
 994         int ret;
 995
 996         if (i <= score->rule) {
 997                 switch (i) {
 998                 case IPV6_SADDR_RULE_SCOPE:
 999                         ret = score->scopedist;
1000                         break;
1001                 case IPV6_SADDR_RULE_PREFIX:
1002                         ret = score->matchlen;
1003                         break;
1004                 default:
1005                         ret = !!test_bit(i, score->scorebits);
1006                 }
1007                 goto out;
1008         }
1009
1010         switch (i) {
1011         case IPV6_SADDR_RULE_INIT:
1012                 /* Rule 0: remember if hiscore is not ready yet */
1013                 ret = !!score->ifa;
1014                 break;
1015         case IPV6_SADDR_RULE_LOCAL:
1016                 /* Rule 1: Prefer same address */
1017                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1018                 break;
1019         case IPV6_SADDR_RULE_SCOPE:
1020                 /* Rule 2: Prefer appropriate scope
1021                  *
1022                  *      ret
1023                  *       ^
1024                  *    -1 |  d 15
1025                  *    ---+--+-+---> scope
1026                  *       |
1027                  *       |             d is scope of the destination.
1028                  *  B-d  |  \
1029                  *       |   \      <- smaller scope is better if
1030                  *  B-15 |    \        if scope is enough for destinaion.
1031                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1032                  * d-C-1 | /
1033                  *       |/         <- greater is better
1034                  *   -C  /             if scope is not enough for destination.
1035                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1036                  *
1037                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1038                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1039                  * Assume B = 0 and we get C > 29.
1040                  */
1041                 ret = __ipv6_addr_src_scope(score->addr_type);
1042                 if (ret >= dst->scope)
1043                         ret = -ret;
1044                 else
1045                         ret -= 128;     /* 30 is enough */
1046                 score->scopedist = ret;
1047                 break;
1048         case IPV6_SADDR_RULE_PREFERRED:
1049                 /* Rule 3: Avoid deprecated and optimistic addresses */
1050                 ret = ipv6_saddr_preferred(score->addr_type) ||
1051                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1052                 break;
1053 #ifdef CONFIG_IPV6_MIP6
1054         case IPV6_SADDR_RULE_HOA:
1055             {
1056                 /* Rule 4: Prefer home address */
1057                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1058                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1059                 break;
1060             }
1061 #endif
1062         case IPV6_SADDR_RULE_OIF:
1063                 /* Rule 5: Prefer outgoing interface */
1064                 ret = (!dst->ifindex ||
1065                        dst->ifindex == score->ifa->idev->dev->ifindex);
1066                 break;
1067         case IPV6_SADDR_RULE_LABEL:
1068                 /* Rule 6: Prefer matching label */
1069                 ret = ipv6_addr_label(net,
1070                                       &score->ifa->addr, score->addr_type,
1071                                       score->ifa->idev->dev->ifindex) == dst->label;
1072                 break;
1073 #ifdef CONFIG_IPV6_PRIVACY
1074         case IPV6_SADDR_RULE_PRIVACY:
1075             {
1076                 /* Rule 7: Prefer public address
1077                  * Note: prefer temporary address if use_tempaddr >= 2
1078                  */
1079                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1080                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1081                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1082                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1083                 break;
1084             }
1085 #endif
1086         case IPV6_SADDR_RULE_ORCHID:
1087                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1088                  *          non-ORCHID vs non-ORCHID
1089                  */
1090                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1091                         ipv6_addr_orchid(dst->addr));
1092                 break;
1093         case IPV6_SADDR_RULE_PREFIX:
1094                 /* Rule 8: Use longest matching prefix */
1095                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1096                                                        dst->addr);
1097                 break;
1098         default:
1099                 ret = 0;
1100         }
1101
1102         if (ret)
1103                 __set_bit(i, score->scorebits);
1104         score->rule = i;
1105 out:
1106         return ret;
1107 }
1108
1109 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1110                        const struct in6_addr *daddr, unsigned int prefs,
1111                        struct in6_addr *saddr)
1112 {
1113         struct ipv6_saddr_score scores[2],
1114                                 *score = &scores[0], *hiscore = &scores[1];
1115         struct ipv6_saddr_dst dst;
1116         struct net_device *dev;
1117         int dst_type;
1118
1119         dst_type = __ipv6_addr_type(daddr);
1120         dst.addr = daddr;
1121         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1122         dst.scope = __ipv6_addr_src_scope(dst_type);
1123         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1124         dst.prefs = prefs;
1125
1126         hiscore->rule = -1;
1127         hiscore->ifa = NULL;
1128
1129         rcu_read_lock();
1130
1131         for_each_netdev_rcu(net, dev) {
1132                 struct inet6_dev *idev;
1133
1134                 /* Candidate Source Address (section 4)
1135                  *  - multicast and link-local destination address,
1136                  *    the set of candidate source address MUST only
1137                  *    include addresses assigned to interfaces
1138                  *    belonging to the same link as the outgoing
1139                  *    interface.
1140                  * (- For site-local destination addresses, the
1141                  *    set of candidate source addresses MUST only
1142                  *    include addresses assigned to interfaces
1143                  *    belonging to the same site as the outgoing
1144                  *    interface.)
1145                  */
1146                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1147                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1148                     dst.ifindex && dev->ifindex != dst.ifindex)
1149                         continue;
1150
1151                 idev = __in6_dev_get(dev);
1152                 if (!idev)
1153                         continue;
1154
1155                 read_lock_bh(&idev->lock);
1156                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1157                         int i;
1158
1159                         /*
1160                          * - Tentative Address (RFC2462 section 5.4)
1161                          *  - A tentative address is not considered
1162                          *    "assigned to an interface" in the traditional
1163                          *    sense, unless it is also flagged as optimistic.
1164                          * - Candidate Source Address (section 4)
1165                          *  - In any case, anycast addresses, multicast
1166                          *    addresses, and the unspecified address MUST
1167                          *    NOT be included in a candidate set.
1168                          */
1169                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1170                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1171                                 continue;
1172
1173                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1174
1175                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1176                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1177                                 LIMIT_NETDEBUG(KERN_DEBUG
1178                                                "ADDRCONF: unspecified / multicast address "
1179                                                "assigned as unicast address on %s",
1180                                                dev->name);
1181                                 continue;
1182                         }
1183
1184                         score->rule = -1;
1185                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1186
1187                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1188                                 int minihiscore, miniscore;
1189
1190                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1191                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1192
1193                                 if (minihiscore > miniscore) {
1194                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1195                                             score->scopedist > 0) {
1196                                                 /*
1197                                                  * special case:
1198                                                  * each remaining entry
1199                                                  * has too small (not enough)
1200                                                  * scope, because ifa entries
1201                                                  * are sorted by their scope
1202                                                  * values.
1203                                                  */
1204                                                 goto try_nextdev;
1205                                         }
1206                                         break;
1207                                 } else if (minihiscore < miniscore) {
1208                                         if (hiscore->ifa)
1209                                                 in6_ifa_put(hiscore->ifa);
1210
1211                                         in6_ifa_hold(score->ifa);
1212
1213                                         swap(hiscore, score);
1214
1215                                         /* restore our iterator */
1216                                         score->ifa = hiscore->ifa;
1217
1218                                         break;
1219                                 }
1220                         }
1221                 }
1222 try_nextdev:
1223                 read_unlock_bh(&idev->lock);
1224         }
1225         rcu_read_unlock();
1226
1227         if (!hiscore->ifa)
1228                 return -EADDRNOTAVAIL;
1229
1230         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1231         in6_ifa_put(hiscore->ifa);
1232         return 0;
1233 }
1234 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1235
1236 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1237                     unsigned char banned_flags)
1238 {
1239         struct inet6_dev *idev;
1240         int err = -EADDRNOTAVAIL;
1241
1242         rcu_read_lock();
1243         idev = __in6_dev_get(dev);
1244         if (idev) {
1245                 struct inet6_ifaddr *ifp;
1246
1247                 read_lock_bh(&idev->lock);
1248                 list_for_each_entry(ifp, &idev->addr_list, if_list) {
1249                         if (ifp->scope == IFA_LINK &&
1250                             !(ifp->flags & banned_flags)) {
1251                                 ipv6_addr_copy(addr, &ifp->addr);
1252                                 err = 0;
1253                                 break;
1254                         }
1255                 }
1256                 read_unlock_bh(&idev->lock);
1257         }
1258         rcu_read_unlock();
1259         return err;
1260 }
1261
1262 static int ipv6_count_addresses(struct inet6_dev *idev)
1263 {
1264         int cnt = 0;
1265         struct inet6_ifaddr *ifp;
1266
1267         read_lock_bh(&idev->lock);
1268         list_for_each_entry(ifp, &idev->addr_list, if_list)
1269                 cnt++;
1270         read_unlock_bh(&idev->lock);
1271         return cnt;
1272 }
1273
1274 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1275                   struct net_device *dev, int strict)
1276 {
1277         struct inet6_ifaddr *ifp;
1278         struct hlist_node *node;
1279         unsigned int hash = ipv6_addr_hash(addr);
1280
1281         rcu_read_lock_bh();
1282         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1283                 if (!net_eq(dev_net(ifp->idev->dev), net))
1284                         continue;
1285                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1286                     !(ifp->flags&IFA_F_TENTATIVE) &&
1287                     (dev == NULL || ifp->idev->dev == dev ||
1288                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1289                         rcu_read_unlock_bh();
1290                         return 1;
1291                 }
1292         }
1293
1294         rcu_read_unlock_bh();
1295         return 0;
1296 }
1297 EXPORT_SYMBOL(ipv6_chk_addr);
1298
1299 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1300                                struct net_device *dev)
1301 {
1302         unsigned int hash = ipv6_addr_hash(addr);
1303         struct inet6_ifaddr *ifp;
1304         struct hlist_node *node;
1305
1306         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1307                 if (!net_eq(dev_net(ifp->idev->dev), net))
1308                         continue;
1309                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1310                         if (dev == NULL || ifp->idev->dev == dev)
1311                                 return true;
1312                 }
1313         }
1314         return false;
1315 }
1316
1317 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1318 {
1319         struct inet6_dev *idev;
1320         struct inet6_ifaddr *ifa;
1321         int     onlink;
1322
1323         onlink = 0;
1324         rcu_read_lock();
1325         idev = __in6_dev_get(dev);
1326         if (idev) {
1327                 read_lock_bh(&idev->lock);
1328                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1329                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1330                                                    ifa->prefix_len);
1331                         if (onlink)
1332                                 break;
1333                 }
1334                 read_unlock_bh(&idev->lock);
1335         }
1336         rcu_read_unlock();
1337         return onlink;
1338 }
1339
1340 EXPORT_SYMBOL(ipv6_chk_prefix);
1341
1342 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1343                                      struct net_device *dev, int strict)
1344 {
1345         struct inet6_ifaddr *ifp, *result = NULL;
1346         unsigned int hash = ipv6_addr_hash(addr);
1347         struct hlist_node *node;
1348
1349         rcu_read_lock_bh();
1350         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1351                 if (!net_eq(dev_net(ifp->idev->dev), net))
1352                         continue;
1353                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1354                         if (dev == NULL || ifp->idev->dev == dev ||
1355                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1356                                 result = ifp;
1357                                 in6_ifa_hold(ifp);
1358                                 break;
1359                         }
1360                 }
1361         }
1362         rcu_read_unlock_bh();
1363
1364         return result;
1365 }
1366
1367 /* Gets referenced address, destroys ifaddr */
1368
1369 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1370 {
1371         if (ifp->flags&IFA_F_PERMANENT) {
1372                 spin_lock_bh(&ifp->lock);
1373                 addrconf_del_timer(ifp);
1374                 ifp->flags |= IFA_F_TENTATIVE;
1375                 if (dad_failed)
1376                         ifp->flags |= IFA_F_DADFAILED;
1377                 spin_unlock_bh(&ifp->lock);
1378                 if (dad_failed)
1379                         ipv6_ifa_notify(0, ifp);
1380                 in6_ifa_put(ifp);
1381 #ifdef CONFIG_IPV6_PRIVACY
1382         } else if (ifp->flags&IFA_F_TEMPORARY) {
1383                 struct inet6_ifaddr *ifpub;
1384                 spin_lock_bh(&ifp->lock);
1385                 ifpub = ifp->ifpub;
1386                 if (ifpub) {
1387                         in6_ifa_hold(ifpub);
1388                         spin_unlock_bh(&ifp->lock);
1389                         ipv6_create_tempaddr(ifpub, ifp);
1390                         in6_ifa_put(ifpub);
1391                 } else {
1392                         spin_unlock_bh(&ifp->lock);
1393                 }
1394                 ipv6_del_addr(ifp);
1395 #endif
1396         } else
1397                 ipv6_del_addr(ifp);
1398 }
1399
1400 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1401 {
1402         int err = -ENOENT;
1403
1404         spin_lock(&ifp->state_lock);
1405         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1406                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1407                 err = 0;
1408         }
1409         spin_unlock(&ifp->state_lock);
1410
1411         return err;
1412 }
1413
1414 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1415 {
1416         struct inet6_dev *idev = ifp->idev;
1417
1418         if (addrconf_dad_end(ifp)) {
1419                 in6_ifa_put(ifp);
1420                 return;
1421         }
1422
1423         if (net_ratelimit())
1424                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1425                         ifp->idev->dev->name, &ifp->addr);
1426
1427         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1428                 struct in6_addr addr;
1429
1430                 addr.s6_addr32[0] = htonl(0xfe800000);
1431                 addr.s6_addr32[1] = 0;
1432
1433                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1434                     ipv6_addr_equal(&ifp->addr, &addr)) {
1435                         /* DAD failed for link-local based on MAC address */
1436                         idev->cnf.disable_ipv6 = 1;
1437
1438                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1439                                 ifp->idev->dev->name);
1440                 }
1441         }
1442
1443         addrconf_dad_stop(ifp, 1);
1444 }
1445
1446 /* Join to solicited addr multicast group. */
1447
1448 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1449 {
1450         struct in6_addr maddr;
1451
1452         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1453                 return;
1454
1455         addrconf_addr_solict_mult(addr, &maddr);
1456         ipv6_dev_mc_inc(dev, &maddr);
1457 }
1458
1459 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         __ipv6_dev_mc_dec(idev, &maddr);
1468 }
1469
1470 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1471 {
1472         struct in6_addr addr;
1473         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1474         if (ipv6_addr_any(&addr))
1475                 return;
1476         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1477 }
1478
1479 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1480 {
1481         struct in6_addr addr;
1482         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1483         if (ipv6_addr_any(&addr))
1484                 return;
1485         __ipv6_dev_ac_dec(ifp->idev, &addr);
1486 }
1487
1488 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1489 {
1490         if (dev->addr_len != ETH_ALEN)
1491                 return -1;
1492         memcpy(eui, dev->dev_addr, 3);
1493         memcpy(eui + 5, dev->dev_addr + 3, 3);
1494
1495         /*
1496          * The zSeries OSA network cards can be shared among various
1497          * OS instances, but the OSA cards have only one MAC address.
1498          * This leads to duplicate address conflicts in conjunction
1499          * with IPv6 if more than one instance uses the same card.
1500          *
1501          * The driver for these cards can deliver a unique 16-bit
1502          * identifier for each instance sharing the same card.  It is
1503          * placed instead of 0xFFFE in the interface identifier.  The
1504          * "u" bit of the interface identifier is not inverted in this
1505          * case.  Hence the resulting interface identifier has local
1506          * scope according to RFC2373.
1507          */
1508         if (dev->dev_id) {
1509                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1510                 eui[4] = dev->dev_id & 0xFF;
1511         } else {
1512                 eui[3] = 0xFF;
1513                 eui[4] = 0xFE;
1514                 eui[0] ^= 2;
1515         }
1516         return 0;
1517 }
1518
1519 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1520 {
1521         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1522         if (dev->addr_len != ARCNET_ALEN)
1523                 return -1;
1524         memset(eui, 0, 7);
1525         eui[7] = *(u8*)dev->dev_addr;
1526         return 0;
1527 }
1528
1529 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1530 {
1531         if (dev->addr_len != INFINIBAND_ALEN)
1532                 return -1;
1533         memcpy(eui, dev->dev_addr + 12, 8);
1534         eui[0] |= 2;
1535         return 0;
1536 }
1537
1538 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1539 {
1540         if (addr == 0)
1541                 return -1;
1542         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1543                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1544                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1545                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1546                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1547                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1548         eui[1] = 0;
1549         eui[2] = 0x5E;
1550         eui[3] = 0xFE;
1551         memcpy(eui + 4, &addr, 4);
1552         return 0;
1553 }
1554
1555 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1556 {
1557         if (dev->priv_flags & IFF_ISATAP)
1558                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1559         return -1;
1560 }
1561
1562 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1563 {
1564         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1565 }
1566
1567 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1568 {
1569         switch (dev->type) {
1570         case ARPHRD_ETHER:
1571         case ARPHRD_FDDI:
1572         case ARPHRD_IEEE802_TR:
1573                 return addrconf_ifid_eui48(eui, dev);
1574         case ARPHRD_ARCNET:
1575                 return addrconf_ifid_arcnet(eui, dev);
1576         case ARPHRD_INFINIBAND:
1577                 return addrconf_ifid_infiniband(eui, dev);
1578         case ARPHRD_SIT:
1579                 return addrconf_ifid_sit(eui, dev);
1580         case ARPHRD_IPGRE:
1581                 return addrconf_ifid_gre(eui, dev);
1582         }
1583         return -1;
1584 }
1585
1586 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1587 {
1588         int err = -1;
1589         struct inet6_ifaddr *ifp;
1590
1591         read_lock_bh(&idev->lock);
1592         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1593                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1594                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1595                         err = 0;
1596                         break;
1597                 }
1598         }
1599         read_unlock_bh(&idev->lock);
1600         return err;
1601 }
1602
1603 #ifdef CONFIG_IPV6_PRIVACY
1604 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1605 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1606 {
1607 regen:
1608         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1609         idev->rndid[0] &= ~0x02;
1610
1611         /*
1612          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1613          * check if generated address is not inappropriate
1614          *
1615          *  - Reserved subnet anycast (RFC 2526)
1616          *      11111101 11....11 1xxxxxxx
1617          *  - ISATAP (RFC4214) 6.1
1618          *      00-00-5E-FE-xx-xx-xx-xx
1619          *  - value 0
1620          *  - XXX: already assigned to an address on the device
1621          */
1622         if (idev->rndid[0] == 0xfd &&
1623             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1624             (idev->rndid[7]&0x80))
1625                 goto regen;
1626         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1627                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1628                         goto regen;
1629                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1630                         goto regen;
1631         }
1632
1633         return 0;
1634 }
1635
1636 static void ipv6_regen_rndid(unsigned long data)
1637 {
1638         struct inet6_dev *idev = (struct inet6_dev *) data;
1639         unsigned long expires;
1640
1641         rcu_read_lock_bh();
1642         write_lock_bh(&idev->lock);
1643
1644         if (idev->dead)
1645                 goto out;
1646
1647         if (__ipv6_regen_rndid(idev) < 0)
1648                 goto out;
1649
1650         expires = jiffies +
1651                 idev->cnf.temp_prefered_lft * HZ -
1652                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1653                 idev->cnf.max_desync_factor * HZ;
1654         if (time_before(expires, jiffies)) {
1655                 printk(KERN_WARNING
1656                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1657                         idev->dev->name);
1658                 goto out;
1659         }
1660
1661         if (!mod_timer(&idev->regen_timer, expires))
1662                 in6_dev_hold(idev);
1663
1664 out:
1665         write_unlock_bh(&idev->lock);
1666         rcu_read_unlock_bh();
1667         in6_dev_put(idev);
1668 }
1669
1670 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1671         int ret = 0;
1672
1673         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1674                 ret = __ipv6_regen_rndid(idev);
1675         return ret;
1676 }
1677 #endif
1678
1679 /*
1680  *      Add prefix route.
1681  */
1682
1683 static void
1684 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1685                       unsigned long expires, u32 flags)
1686 {
1687         struct fib6_config cfg = {
1688                 .fc_table = RT6_TABLE_PREFIX,
1689                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1690                 .fc_ifindex = dev->ifindex,
1691                 .fc_expires = expires,
1692                 .fc_dst_len = plen,
1693                 .fc_flags = RTF_UP | flags,
1694                 .fc_nlinfo.nl_net = dev_net(dev),
1695                 .fc_protocol = RTPROT_KERNEL,
1696         };
1697
1698         ipv6_addr_copy(&cfg.fc_dst, pfx);
1699
1700         /* Prevent useless cloning on PtP SIT.
1701            This thing is done here expecting that the whole
1702            class of non-broadcast devices need not cloning.
1703          */
1704 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1705         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1706                 cfg.fc_flags |= RTF_NONEXTHOP;
1707 #endif
1708
1709         ip6_route_add(&cfg);
1710 }
1711
1712 /* Create "default" multicast route to the interface */
1713
1714 static void addrconf_add_mroute(struct net_device *dev)
1715 {
1716         struct fib6_config cfg = {
1717                 .fc_table = RT6_TABLE_LOCAL,
1718                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1719                 .fc_ifindex = dev->ifindex,
1720                 .fc_dst_len = 8,
1721                 .fc_flags = RTF_UP,
1722                 .fc_nlinfo.nl_net = dev_net(dev),
1723         };
1724
1725         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1726
1727         ip6_route_add(&cfg);
1728 }
1729
1730 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1731 static void sit_route_add(struct net_device *dev)
1732 {
1733         struct fib6_config cfg = {
1734                 .fc_table = RT6_TABLE_MAIN,
1735                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1736                 .fc_ifindex = dev->ifindex,
1737                 .fc_dst_len = 96,
1738                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1739                 .fc_nlinfo.nl_net = dev_net(dev),
1740         };
1741
1742         /* prefix length - 96 bits "::d.d.d.d" */
1743         ip6_route_add(&cfg);
1744 }
1745 #endif
1746
1747 static void addrconf_add_lroute(struct net_device *dev)
1748 {
1749         struct in6_addr addr;
1750
1751         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1752         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1753 }
1754
1755 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1756 {
1757         struct inet6_dev *idev;
1758
1759         ASSERT_RTNL();
1760
1761         idev = ipv6_find_idev(dev);
1762         if (!idev)
1763                 return ERR_PTR(-ENOBUFS);
1764
1765         if (idev->cnf.disable_ipv6)
1766                 return ERR_PTR(-EACCES);
1767
1768         /* Add default multicast route */
1769         addrconf_add_mroute(dev);
1770
1771         /* Add link local route */
1772         addrconf_add_lroute(dev);
1773         return idev;
1774 }
1775
1776 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1777 {
1778         struct prefix_info *pinfo;
1779         __u32 valid_lft;
1780         __u32 prefered_lft;
1781         int addr_type;
1782         struct inet6_dev *in6_dev;
1783         struct net *net = dev_net(dev);
1784
1785         pinfo = (struct prefix_info *) opt;
1786
1787         if (len < sizeof(struct prefix_info)) {
1788                 ADBG(("addrconf: prefix option too short\n"));
1789                 return;
1790         }
1791
1792         /*
1793          *      Validation checks ([ADDRCONF], page 19)
1794          */
1795
1796         addr_type = ipv6_addr_type(&pinfo->prefix);
1797
1798         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1799                 return;
1800
1801         valid_lft = ntohl(pinfo->valid);
1802         prefered_lft = ntohl(pinfo->prefered);
1803
1804         if (prefered_lft > valid_lft) {
1805                 if (net_ratelimit())
1806                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1807                 return;
1808         }
1809
1810         in6_dev = in6_dev_get(dev);
1811
1812         if (in6_dev == NULL) {
1813                 if (net_ratelimit())
1814                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1815                 return;
1816         }
1817
1818         /*
1819          *      Two things going on here:
1820          *      1) Add routes for on-link prefixes
1821          *      2) Configure prefixes with the auto flag set
1822          */
1823
1824         if (pinfo->onlink) {
1825                 struct rt6_info *rt;
1826                 unsigned long rt_expires;
1827
1828                 /* Avoid arithmetic overflow. Really, we could
1829                  * save rt_expires in seconds, likely valid_lft,
1830                  * but it would require division in fib gc, that it
1831                  * not good.
1832                  */
1833                 if (HZ > USER_HZ)
1834                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1835                 else
1836                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1837
1838                 if (addrconf_finite_timeout(rt_expires))
1839                         rt_expires *= HZ;
1840
1841                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1842                                 dev->ifindex, 1);
1843
1844                 if (rt && addrconf_is_prefix_route(rt)) {
1845                         /* Autoconf prefix route */
1846                         if (valid_lft == 0) {
1847                                 ip6_del_rt(rt);
1848                                 rt = NULL;
1849                         } else if (addrconf_finite_timeout(rt_expires)) {
1850                                 /* not infinity */
1851                                 rt->rt6i_expires = jiffies + rt_expires;
1852                                 rt->rt6i_flags |= RTF_EXPIRES;
1853                         } else {
1854                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1855                                 rt->rt6i_expires = 0;
1856                         }
1857                 } else if (valid_lft) {
1858                         clock_t expires = 0;
1859                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1860                         if (addrconf_finite_timeout(rt_expires)) {
1861                                 /* not infinity */
1862                                 flags |= RTF_EXPIRES;
1863                                 expires = jiffies_to_clock_t(rt_expires);
1864                         }
1865                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1866                                               dev, expires, flags);
1867                 }
1868                 if (rt)
1869                         dst_release(&rt->dst);
1870         }
1871
1872         /* Try to figure out our local address for this prefix */
1873
1874         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1875                 struct inet6_ifaddr * ifp;
1876                 struct in6_addr addr;
1877                 int create = 0, update_lft = 0;
1878
1879                 if (pinfo->prefix_len == 64) {
1880                         memcpy(&addr, &pinfo->prefix, 8);
1881                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1882                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1883                                 in6_dev_put(in6_dev);
1884                                 return;
1885                         }
1886                         goto ok;
1887                 }
1888                 if (net_ratelimit())
1889                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1890                                pinfo->prefix_len);
1891                 in6_dev_put(in6_dev);
1892                 return;
1893
1894 ok:
1895
1896                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1897
1898                 if (ifp == NULL && valid_lft) {
1899                         int max_addresses = in6_dev->cnf.max_addresses;
1900                         u32 addr_flags = 0;
1901
1902 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1903                         if (in6_dev->cnf.optimistic_dad &&
1904                             !net->ipv6.devconf_all->forwarding)
1905                                 addr_flags = IFA_F_OPTIMISTIC;
1906 #endif
1907
1908                         /* Do not allow to create too much of autoconfigured
1909                          * addresses; this would be too easy way to crash kernel.
1910                          */
1911                         if (!max_addresses ||
1912                             ipv6_count_addresses(in6_dev) < max_addresses)
1913                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1914                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1915                                                     addr_flags);
1916
1917                         if (!ifp || IS_ERR(ifp)) {
1918                                 in6_dev_put(in6_dev);
1919                                 return;
1920                         }
1921
1922                         update_lft = create = 1;
1923                         ifp->cstamp = jiffies;
1924                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1925                 }
1926
1927                 if (ifp) {
1928                         int flags;
1929                         unsigned long now;
1930 #ifdef CONFIG_IPV6_PRIVACY
1931                         struct inet6_ifaddr *ift;
1932 #endif
1933                         u32 stored_lft;
1934
1935                         /* update lifetime (RFC2462 5.5.3 e) */
1936                         spin_lock(&ifp->lock);
1937                         now = jiffies;
1938                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1939                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1940                         else
1941                                 stored_lft = 0;
1942                         if (!update_lft && stored_lft) {
1943                                 if (valid_lft > MIN_VALID_LIFETIME ||
1944                                     valid_lft > stored_lft)
1945                                         update_lft = 1;
1946                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1947                                         /* valid_lft <= stored_lft is always true */
1948                                         /*
1949                                          * RFC 4862 Section 5.5.3e:
1950                                          * "Note that the preferred lifetime of
1951                                          *  the corresponding address is always
1952                                          *  reset to the Preferred Lifetime in
1953                                          *  the received Prefix Information
1954                                          *  option, regardless of whether the
1955                                          *  valid lifetime is also reset or
1956                                          *  ignored."
1957                                          *
1958                                          *  So if the preferred lifetime in
1959                                          *  this advertisement is different
1960                                          *  than what we have stored, but the
1961                                          *  valid lifetime is invalid, just
1962                                          *  reset prefered_lft.
1963                                          *
1964                                          *  We must set the valid lifetime
1965                                          *  to the stored lifetime since we'll
1966                                          *  be updating the timestamp below,
1967                                          *  else we'll set it back to the
1968                                          *  minimum.
1969                                          */
1970                                         if (prefered_lft != ifp->prefered_lft) {
1971                                                 valid_lft = stored_lft;
1972                                                 update_lft = 1;
1973                                         }
1974                                 } else {
1975                                         valid_lft = MIN_VALID_LIFETIME;
1976                                         if (valid_lft < prefered_lft)
1977                                                 prefered_lft = valid_lft;
1978                                         update_lft = 1;
1979                                 }
1980                         }
1981
1982                         if (update_lft) {
1983                                 ifp->valid_lft = valid_lft;
1984                                 ifp->prefered_lft = prefered_lft;
1985                                 ifp->tstamp = now;
1986                                 flags = ifp->flags;
1987                                 ifp->flags &= ~IFA_F_DEPRECATED;
1988                                 spin_unlock(&ifp->lock);
1989
1990                                 if (!(flags&IFA_F_TENTATIVE))
1991                                         ipv6_ifa_notify(0, ifp);
1992                         } else
1993                                 spin_unlock(&ifp->lock);
1994
1995 #ifdef CONFIG_IPV6_PRIVACY
1996                         read_lock_bh(&in6_dev->lock);
1997                         /* update all temporary addresses in the list */
1998                         list_for_each_entry(ift, &in6_dev->tempaddr_list, tmp_list) {
1999                                 /*
2000                                  * When adjusting the lifetimes of an existing
2001                                  * temporary address, only lower the lifetimes.
2002                                  * Implementations must not increase the
2003                                  * lifetimes of an existing temporary address
2004                                  * when processing a Prefix Information Option.
2005                                  */
2006                                 if (ifp != ift->ifpub)
2007                                         continue;
2008
2009                                 spin_lock(&ift->lock);
2010                                 flags = ift->flags;
2011                                 if (ift->valid_lft > valid_lft &&
2012                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
2013                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
2014                                 if (ift->prefered_lft > prefered_lft &&
2015                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
2016                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
2017                                 spin_unlock(&ift->lock);
2018                                 if (!(flags&IFA_F_TENTATIVE))
2019                                         ipv6_ifa_notify(0, ift);
2020                         }
2021
2022                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2023                                 /*
2024                                  * When a new public address is created as described in [ADDRCONF],
2025                                  * also create a new temporary address. Also create a temporary
2026                                  * address if it's enabled but no temporary address currently exists.
2027                                  */
2028                                 read_unlock_bh(&in6_dev->lock);
2029                                 ipv6_create_tempaddr(ifp, NULL);
2030                         } else {
2031                                 read_unlock_bh(&in6_dev->lock);
2032                         }
2033 #endif
2034                         in6_ifa_put(ifp);
2035                         addrconf_verify(0);
2036                 }
2037         }
2038         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2039         in6_dev_put(in6_dev);
2040 }
2041
2042 /*
2043  *      Set destination address.
2044  *      Special case for SIT interfaces where we create a new "virtual"
2045  *      device.
2046  */
2047 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2048 {
2049         struct in6_ifreq ireq;
2050         struct net_device *dev;
2051         int err = -EINVAL;
2052
2053         rtnl_lock();
2054
2055         err = -EFAULT;
2056         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2057                 goto err_exit;
2058
2059         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2060
2061         err = -ENODEV;
2062         if (dev == NULL)
2063                 goto err_exit;
2064
2065 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2066         if (dev->type == ARPHRD_SIT) {
2067                 const struct net_device_ops *ops = dev->netdev_ops;
2068                 struct ifreq ifr;
2069                 struct ip_tunnel_parm p;
2070
2071                 err = -EADDRNOTAVAIL;
2072                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2073                         goto err_exit;
2074
2075                 memset(&p, 0, sizeof(p));
2076                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2077                 p.iph.saddr = 0;
2078                 p.iph.version = 4;
2079                 p.iph.ihl = 5;
2080                 p.iph.protocol = IPPROTO_IPV6;
2081                 p.iph.ttl = 64;
2082                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2083
2084                 if (ops->ndo_do_ioctl) {
2085                         mm_segment_t oldfs = get_fs();
2086
2087                         set_fs(KERNEL_DS);
2088                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2089                         set_fs(oldfs);
2090                 } else
2091                         err = -EOPNOTSUPP;
2092
2093                 if (err == 0) {
2094                         err = -ENOBUFS;
2095                         dev = __dev_get_by_name(net, p.name);
2096                         if (!dev)
2097                                 goto err_exit;
2098                         err = dev_open(dev);
2099                 }
2100         }
2101 #endif
2102
2103 err_exit:
2104         rtnl_unlock();
2105         return err;
2106 }
2107
2108 /*
2109  *      Manual configuration of address on an interface
2110  */
2111 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2112                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2113                           __u32 valid_lft)
2114 {
2115         struct inet6_ifaddr *ifp;
2116         struct inet6_dev *idev;
2117         struct net_device *dev;
2118         int scope;
2119         u32 flags;
2120         clock_t expires;
2121         unsigned long timeout;
2122
2123         ASSERT_RTNL();
2124
2125         if (plen > 128)
2126                 return -EINVAL;
2127
2128         /* check the lifetime */
2129         if (!valid_lft || prefered_lft > valid_lft)
2130                 return -EINVAL;
2131
2132         dev = __dev_get_by_index(net, ifindex);
2133         if (!dev)
2134                 return -ENODEV;
2135
2136         idev = addrconf_add_dev(dev);
2137         if (IS_ERR(idev))
2138                 return PTR_ERR(idev);
2139
2140         scope = ipv6_addr_scope(pfx);
2141
2142         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2143         if (addrconf_finite_timeout(timeout)) {
2144                 expires = jiffies_to_clock_t(timeout * HZ);
2145                 valid_lft = timeout;
2146                 flags = RTF_EXPIRES;
2147         } else {
2148                 expires = 0;
2149                 flags = 0;
2150                 ifa_flags |= IFA_F_PERMANENT;
2151         }
2152
2153         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2154         if (addrconf_finite_timeout(timeout)) {
2155                 if (timeout == 0)
2156                         ifa_flags |= IFA_F_DEPRECATED;
2157                 prefered_lft = timeout;
2158         }
2159
2160         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2161
2162         if (!IS_ERR(ifp)) {
2163                 spin_lock_bh(&ifp->lock);
2164                 ifp->valid_lft = valid_lft;
2165                 ifp->prefered_lft = prefered_lft;
2166                 ifp->tstamp = jiffies;
2167                 spin_unlock_bh(&ifp->lock);
2168
2169                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2170                                       expires, flags);
2171                 /*
2172                  * Note that section 3.1 of RFC 4429 indicates
2173                  * that the Optimistic flag should not be set for
2174                  * manually configured addresses
2175                  */
2176                 addrconf_dad_start(ifp, 0);
2177                 in6_ifa_put(ifp);
2178                 addrconf_verify(0);
2179                 return 0;
2180         }
2181
2182         return PTR_ERR(ifp);
2183 }
2184
2185 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2186                           unsigned int plen)
2187 {
2188         struct inet6_ifaddr *ifp;
2189         struct inet6_dev *idev;
2190         struct net_device *dev;
2191
2192         if (plen > 128)
2193                 return -EINVAL;
2194
2195         dev = __dev_get_by_index(net, ifindex);
2196         if (!dev)
2197                 return -ENODEV;
2198
2199         if ((idev = __in6_dev_get(dev)) == NULL)
2200                 return -ENXIO;
2201
2202         read_lock_bh(&idev->lock);
2203         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2204                 if (ifp->prefix_len == plen &&
2205                     ipv6_addr_equal(pfx, &ifp->addr)) {
2206                         in6_ifa_hold(ifp);
2207                         read_unlock_bh(&idev->lock);
2208
2209                         ipv6_del_addr(ifp);
2210
2211                         /* If the last address is deleted administratively,
2212                            disable IPv6 on this interface.
2213                          */
2214                         if (list_empty(&idev->addr_list))
2215                                 addrconf_ifdown(idev->dev, 1);
2216                         return 0;
2217                 }
2218         }
2219         read_unlock_bh(&idev->lock);
2220         return -EADDRNOTAVAIL;
2221 }
2222
2223
2224 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2225 {
2226         struct in6_ifreq ireq;
2227         int err;
2228
2229         if (!capable(CAP_NET_ADMIN))
2230                 return -EPERM;
2231
2232         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2233                 return -EFAULT;
2234
2235         rtnl_lock();
2236         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2237                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2238                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2239         rtnl_unlock();
2240         return err;
2241 }
2242
2243 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2244 {
2245         struct in6_ifreq ireq;
2246         int err;
2247
2248         if (!capable(CAP_NET_ADMIN))
2249                 return -EPERM;
2250
2251         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2252                 return -EFAULT;
2253
2254         rtnl_lock();
2255         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2256                              ireq.ifr6_prefixlen);
2257         rtnl_unlock();
2258         return err;
2259 }
2260
2261 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2262                      int plen, int scope)
2263 {
2264         struct inet6_ifaddr *ifp;
2265
2266         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2267         if (!IS_ERR(ifp)) {
2268                 spin_lock_bh(&ifp->lock);
2269                 ifp->flags &= ~IFA_F_TENTATIVE;
2270                 spin_unlock_bh(&ifp->lock);
2271                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2272                 in6_ifa_put(ifp);
2273         }
2274 }
2275
2276 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2277 static void sit_add_v4_addrs(struct inet6_dev *idev)
2278 {
2279         struct in6_addr addr;
2280         struct net_device *dev;
2281         struct net *net = dev_net(idev->dev);
2282         int scope;
2283
2284         ASSERT_RTNL();
2285
2286         memset(&addr, 0, sizeof(struct in6_addr));
2287         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2288
2289         if (idev->dev->flags&IFF_POINTOPOINT) {
2290                 addr.s6_addr32[0] = htonl(0xfe800000);
2291                 scope = IFA_LINK;
2292         } else {
2293                 scope = IPV6_ADDR_COMPATv4;
2294         }
2295
2296         if (addr.s6_addr32[3]) {
2297                 add_addr(idev, &addr, 128, scope);
2298                 return;
2299         }
2300
2301         for_each_netdev(net, dev) {
2302                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2303                 if (in_dev && (dev->flags & IFF_UP)) {
2304                         struct in_ifaddr * ifa;
2305
2306                         int flag = scope;
2307
2308                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2309                                 int plen;
2310
2311                                 addr.s6_addr32[3] = ifa->ifa_local;
2312
2313                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2314                                         continue;
2315                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2316                                         if (idev->dev->flags&IFF_POINTOPOINT)
2317                                                 continue;
2318                                         flag |= IFA_HOST;
2319                                 }
2320                                 if (idev->dev->flags&IFF_POINTOPOINT)
2321                                         plen = 64;
2322                                 else
2323                                         plen = 96;
2324
2325                                 add_addr(idev, &addr, plen, flag);
2326                         }
2327                 }
2328         }
2329 }
2330 #endif
2331
2332 static void init_loopback(struct net_device *dev)
2333 {
2334         struct inet6_dev  *idev;
2335
2336         /* ::1 */
2337
2338         ASSERT_RTNL();
2339
2340         if ((idev = ipv6_find_idev(dev)) == NULL) {
2341                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2342                 return;
2343         }
2344
2345         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2346 }
2347
2348 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2349 {
2350         struct inet6_ifaddr * ifp;
2351         u32 addr_flags = IFA_F_PERMANENT;
2352
2353 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2354         if (idev->cnf.optimistic_dad &&
2355             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2356                 addr_flags |= IFA_F_OPTIMISTIC;
2357 #endif
2358
2359
2360         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2361         if (!IS_ERR(ifp)) {
2362                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2363                 addrconf_dad_start(ifp, 0);
2364                 in6_ifa_put(ifp);
2365         }
2366 }
2367
2368 static void addrconf_dev_config(struct net_device *dev)
2369 {
2370         struct in6_addr addr;
2371         struct inet6_dev    * idev;
2372
2373         ASSERT_RTNL();
2374
2375         if ((dev->type != ARPHRD_ETHER) &&
2376             (dev->type != ARPHRD_FDDI) &&
2377             (dev->type != ARPHRD_IEEE802_TR) &&
2378             (dev->type != ARPHRD_ARCNET) &&
2379             (dev->type != ARPHRD_INFINIBAND)) {
2380                 /* Alas, we support only Ethernet autoconfiguration. */
2381                 return;
2382         }
2383
2384         idev = addrconf_add_dev(dev);
2385         if (IS_ERR(idev))
2386                 return;
2387
2388         memset(&addr, 0, sizeof(struct in6_addr));
2389         addr.s6_addr32[0] = htonl(0xFE800000);
2390
2391         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2392                 addrconf_add_linklocal(idev, &addr);
2393 }
2394
2395 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2396 static void addrconf_sit_config(struct net_device *dev)
2397 {
2398         struct inet6_dev *idev;
2399
2400         ASSERT_RTNL();
2401
2402         /*
2403          * Configure the tunnel with one of our IPv4
2404          * addresses... we should configure all of
2405          * our v4 addrs in the tunnel
2406          */
2407
2408         if ((idev = ipv6_find_idev(dev)) == NULL) {
2409                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2410                 return;
2411         }
2412
2413         if (dev->priv_flags & IFF_ISATAP) {
2414                 struct in6_addr addr;
2415
2416                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2417                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2418                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2419                         addrconf_add_linklocal(idev, &addr);
2420                 return;
2421         }
2422
2423         sit_add_v4_addrs(idev);
2424
2425         if (dev->flags&IFF_POINTOPOINT) {
2426                 addrconf_add_mroute(dev);
2427                 addrconf_add_lroute(dev);
2428         } else
2429                 sit_route_add(dev);
2430 }
2431 #endif
2432
2433 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2434 static void addrconf_gre_config(struct net_device *dev)
2435 {
2436         struct inet6_dev *idev;
2437         struct in6_addr addr;
2438
2439         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2440
2441         ASSERT_RTNL();
2442
2443         if ((idev = ipv6_find_idev(dev)) == NULL) {
2444                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2445                 return;
2446         }
2447
2448         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2449         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2450
2451         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2452                 addrconf_add_linklocal(idev, &addr);
2453 }
2454 #endif
2455
2456 static inline int
2457 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2458 {
2459         struct in6_addr lladdr;
2460
2461         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2462                 addrconf_add_linklocal(idev, &lladdr);
2463                 return 0;
2464         }
2465         return -1;
2466 }
2467
2468 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2469 {
2470         struct net_device *link_dev;
2471         struct net *net = dev_net(idev->dev);
2472
2473         /* first try to inherit the link-local address from the link device */
2474         if (idev->dev->iflink &&
2475             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2476                 if (!ipv6_inherit_linklocal(idev, link_dev))
2477                         return;
2478         }
2479         /* then try to inherit it from any device */
2480         for_each_netdev(net, link_dev) {
2481                 if (!ipv6_inherit_linklocal(idev, link_dev))
2482                         return;
2483         }
2484         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2485 }
2486
2487 /*
2488  * Autoconfigure tunnel with a link-local address so routing protocols,
2489  * DHCPv6, MLD etc. can be run over the virtual link
2490  */
2491
2492 static void addrconf_ip6_tnl_config(struct net_device *dev)
2493 {
2494         struct inet6_dev *idev;
2495
2496         ASSERT_RTNL();
2497
2498         idev = addrconf_add_dev(dev);
2499         if (IS_ERR(idev)) {
2500                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2501                 return;
2502         }
2503         ip6_tnl_add_linklocal(idev);
2504 }
2505
2506 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2507                            void * data)
2508 {
2509         struct net_device *dev = (struct net_device *) data;
2510         struct inet6_dev *idev = __in6_dev_get(dev);
2511         int run_pending = 0;
2512         int err;
2513
2514         switch (event) {
2515         case NETDEV_REGISTER:
2516                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2517                         idev = ipv6_add_dev(dev);
2518                         if (!idev)
2519                                 return notifier_from_errno(-ENOMEM);
2520                 }
2521                 break;
2522
2523         case NETDEV_UP:
2524         case NETDEV_CHANGE:
2525                 if (dev->flags & IFF_SLAVE)
2526                         break;
2527
2528                 if (event == NETDEV_UP) {
2529                         if (!addrconf_qdisc_ok(dev)) {
2530                                 /* device is not ready yet. */
2531                                 printk(KERN_INFO
2532                                         "ADDRCONF(NETDEV_UP): %s: "
2533                                         "link is not ready\n",
2534                                         dev->name);
2535                                 break;
2536                         }
2537
2538                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2539                                 idev = ipv6_add_dev(dev);
2540
2541                         if (idev) {
2542                                 idev->if_flags |= IF_READY;
2543                                 run_pending = 1;
2544                         }
2545                 } else {
2546                         if (!addrconf_qdisc_ok(dev)) {
2547                                 /* device is still not ready. */
2548                                 break;
2549                         }
2550
2551                         if (idev) {
2552                                 if (idev->if_flags & IF_READY)
2553                                         /* device is already configured. */
2554                                         break;
2555                                 idev->if_flags |= IF_READY;
2556                         }
2557
2558                         printk(KERN_INFO
2559                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2560                                         "link becomes ready\n",
2561                                         dev->name);
2562
2563                         run_pending = 1;
2564                 }
2565
2566                 switch (dev->type) {
2567 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2568                 case ARPHRD_SIT:
2569                         addrconf_sit_config(dev);
2570                         break;
2571 #endif
2572 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2573                 case ARPHRD_IPGRE:
2574                         addrconf_gre_config(dev);
2575                         break;
2576 #endif
2577                 case ARPHRD_TUNNEL6:
2578                         addrconf_ip6_tnl_config(dev);
2579                         break;
2580                 case ARPHRD_LOOPBACK:
2581                         init_loopback(dev);
2582                         break;
2583
2584                 default:
2585                         addrconf_dev_config(dev);
2586                         break;
2587                 }
2588
2589                 if (idev) {
2590                         if (run_pending)
2591                                 addrconf_dad_run(idev);
2592
2593                         /*
2594                          * If the MTU changed during the interface down,
2595                          * when the interface up, the changed MTU must be
2596                          * reflected in the idev as well as routers.
2597                          */
2598                         if (idev->cnf.mtu6 != dev->mtu &&
2599                             dev->mtu >= IPV6_MIN_MTU) {
2600                                 rt6_mtu_change(dev, dev->mtu);
2601                                 idev->cnf.mtu6 = dev->mtu;
2602                         }
2603                         idev->tstamp = jiffies;
2604                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2605
2606                         /*
2607                          * If the changed mtu during down is lower than
2608                          * IPV6_MIN_MTU stop IPv6 on this interface.
2609                          */
2610                         if (dev->mtu < IPV6_MIN_MTU)
2611                                 addrconf_ifdown(dev, 1);
2612                 }
2613                 break;
2614
2615         case NETDEV_CHANGEMTU:
2616                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2617                         rt6_mtu_change(dev, dev->mtu);
2618                         idev->cnf.mtu6 = dev->mtu;
2619                         break;
2620                 }
2621
2622                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2623                         idev = ipv6_add_dev(dev);
2624                         if (idev)
2625                                 break;
2626                 }
2627
2628                 /*
2629                  * MTU falled under IPV6_MIN_MTU.
2630                  * Stop IPv6 on this interface.
2631                  */
2632
2633         case NETDEV_DOWN:
2634         case NETDEV_UNREGISTER:
2635                 /*
2636                  *      Remove all addresses from this interface.
2637                  */
2638                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2639                 break;
2640
2641         case NETDEV_CHANGENAME:
2642                 if (idev) {
2643                         snmp6_unregister_dev(idev);
2644                         addrconf_sysctl_unregister(idev);
2645                         addrconf_sysctl_register(idev);
2646                         err = snmp6_register_dev(idev);
2647                         if (err)
2648                                 return notifier_from_errno(err);
2649                 }
2650                 break;
2651
2652         case NETDEV_PRE_TYPE_CHANGE:
2653         case NETDEV_POST_TYPE_CHANGE:
2654                 addrconf_type_change(dev, event);
2655                 break;
2656         }
2657
2658         return NOTIFY_OK;
2659 }
2660
2661 /*
2662  *      addrconf module should be notified of a device going up
2663  */
2664 static struct notifier_block ipv6_dev_notf = {
2665         .notifier_call = addrconf_notify,
2666 };
2667
2668 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2669 {
2670         struct inet6_dev *idev;
2671         ASSERT_RTNL();
2672
2673         idev = __in6_dev_get(dev);
2674
2675         if (event == NETDEV_POST_TYPE_CHANGE)
2676                 ipv6_mc_remap(idev);
2677         else if (event == NETDEV_PRE_TYPE_CHANGE)
2678                 ipv6_mc_unmap(idev);
2679 }
2680
2681 static int addrconf_ifdown(struct net_device *dev, int how)
2682 {
2683         struct net *net = dev_net(dev);
2684         struct inet6_dev *idev;
2685         struct inet6_ifaddr *ifa;
2686         int state, i;
2687
2688         ASSERT_RTNL();
2689
2690         rt6_ifdown(net, dev);
2691         neigh_ifdown(&nd_tbl, dev);
2692
2693         idev = __in6_dev_get(dev);
2694         if (idev == NULL)
2695                 return -ENODEV;
2696
2697         /*
2698          * Step 1: remove reference to ipv6 device from parent device.
2699          *         Do not dev_put!
2700          */
2701         if (how) {
2702                 idev->dead = 1;
2703
2704                 /* protected by rtnl_lock */
2705                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2706
2707                 /* Step 1.5: remove snmp6 entry */
2708                 snmp6_unregister_dev(idev);
2709
2710         }
2711
2712         /* Step 2: clear hash table */
2713         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2714                 struct hlist_head *h = &inet6_addr_lst[i];
2715                 struct hlist_node *n;
2716
2717                 spin_lock_bh(&addrconf_hash_lock);
2718         restart:
2719                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2720                         if (ifa->idev == idev) {
2721                                 hlist_del_init_rcu(&ifa->addr_lst);
2722                                 addrconf_del_timer(ifa);
2723                                 goto restart;
2724                         }
2725                 }
2726                 spin_unlock_bh(&addrconf_hash_lock);
2727         }
2728
2729         write_lock_bh(&idev->lock);
2730
2731         /* Step 2: clear flags for stateless addrconf */
2732         if (!how)
2733                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2734
2735 #ifdef CONFIG_IPV6_PRIVACY
2736         if (how && del_timer(&idev->regen_timer))
2737                 in6_dev_put(idev);
2738
2739         /* Step 3: clear tempaddr list */
2740         while (!list_empty(&idev->tempaddr_list)) {
2741                 ifa = list_first_entry(&idev->tempaddr_list,
2742                                        struct inet6_ifaddr, tmp_list);
2743                 list_del(&ifa->tmp_list);
2744                 write_unlock_bh(&idev->lock);
2745                 spin_lock_bh(&ifa->lock);
2746
2747                 if (ifa->ifpub) {
2748                         in6_ifa_put(ifa->ifpub);
2749                         ifa->ifpub = NULL;
2750                 }
2751                 spin_unlock_bh(&ifa->lock);
2752                 in6_ifa_put(ifa);
2753                 write_lock_bh(&idev->lock);
2754         }
2755 #endif
2756
2757         while (!list_empty(&idev->addr_list)) {
2758                 ifa = list_first_entry(&idev->addr_list,
2759                                        struct inet6_ifaddr, if_list);
2760                 addrconf_del_timer(ifa);
2761
2762                 list_del(&ifa->if_list);
2763
2764                 write_unlock_bh(&idev->lock);
2765
2766                 spin_lock_bh(&ifa->state_lock);
2767                 state = ifa->state;
2768                 ifa->state = INET6_IFADDR_STATE_DEAD;
2769                 spin_unlock_bh(&ifa->state_lock);
2770
2771                 if (state != INET6_IFADDR_STATE_DEAD) {
2772                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2773                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2774                 }
2775                 in6_ifa_put(ifa);
2776
2777                 write_lock_bh(&idev->lock);
2778         }
2779
2780         write_unlock_bh(&idev->lock);
2781
2782         /* Step 5: Discard multicast list */
2783         if (how)
2784                 ipv6_mc_destroy_dev(idev);
2785         else
2786                 ipv6_mc_down(idev);
2787
2788         idev->tstamp = jiffies;
2789
2790         /* Last: Shot the device (if unregistered) */
2791         if (how) {
2792                 addrconf_sysctl_unregister(idev);
2793                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2794                 neigh_ifdown(&nd_tbl, dev);
2795                 in6_dev_put(idev);
2796         }
2797         return 0;
2798 }
2799
2800 static void addrconf_rs_timer(unsigned long data)
2801 {
2802         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2803         struct inet6_dev *idev = ifp->idev;
2804
2805         read_lock(&idev->lock);
2806         if (idev->dead || !(idev->if_flags & IF_READY))
2807                 goto out;
2808
2809         if (idev->cnf.forwarding)
2810                 goto out;
2811
2812         /* Announcement received after solicitation was sent */
2813         if (idev->if_flags & IF_RA_RCVD)
2814                 goto out;
2815
2816         spin_lock(&ifp->lock);
2817         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2818                 /* The wait after the last probe can be shorter */
2819                 addrconf_mod_timer(ifp, AC_RS,
2820                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2821                                    idev->cnf.rtr_solicit_delay :
2822                                    idev->cnf.rtr_solicit_interval);
2823                 spin_unlock(&ifp->lock);
2824
2825                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2826         } else {
2827                 spin_unlock(&ifp->lock);
2828                 /*
2829                  * Note: we do not support deprecated "all on-link"
2830                  * assumption any longer.
2831                  */
2832                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2833                        idev->dev->name);
2834         }
2835
2836 out:
2837         read_unlock(&idev->lock);
2838         in6_ifa_put(ifp);
2839 }
2840
2841 /*
2842  *      Duplicate Address Detection
2843  */
2844 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2845 {
2846         unsigned long rand_num;
2847         struct inet6_dev *idev = ifp->idev;
2848
2849         if (ifp->flags & IFA_F_OPTIMISTIC)
2850                 rand_num = 0;
2851         else
2852                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2853
2854         ifp->probes = idev->cnf.dad_transmits;
2855         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2856 }
2857
2858 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2859 {
2860         struct inet6_dev *idev = ifp->idev;
2861         struct net_device *dev = idev->dev;
2862
2863         addrconf_join_solict(dev, &ifp->addr);
2864
2865         net_srandom(ifp->addr.s6_addr32[3]);
2866
2867         read_lock_bh(&idev->lock);
2868         spin_lock(&ifp->lock);
2869         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2870                 goto out;
2871
2872         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2873             idev->cnf.accept_dad < 1 ||
2874             !(ifp->flags&IFA_F_TENTATIVE) ||
2875             ifp->flags & IFA_F_NODAD) {
2876                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2877                 spin_unlock(&ifp->lock);
2878                 read_unlock_bh(&idev->lock);
2879
2880                 addrconf_dad_completed(ifp);
2881                 return;
2882         }
2883
2884         if (!(idev->if_flags & IF_READY)) {
2885                 spin_unlock(&ifp->lock);
2886                 read_unlock_bh(&idev->lock);
2887                 /*
2888                  * If the device is not ready:
2889                  * - keep it tentative if it is a permanent address.
2890                  * - otherwise, kill it.
2891                  */
2892                 in6_ifa_hold(ifp);
2893                 addrconf_dad_stop(ifp, 0);
2894                 return;
2895         }
2896
2897         /*
2898          * Optimistic nodes can start receiving
2899          * Frames right away
2900          */
2901         if (ifp->flags & IFA_F_OPTIMISTIC)
2902                 ip6_ins_rt(ifp->rt);
2903
2904         addrconf_dad_kick(ifp);
2905 out:
2906         spin_unlock(&ifp->lock);
2907         read_unlock_bh(&idev->lock);
2908 }
2909
2910 static void addrconf_dad_timer(unsigned long data)
2911 {
2912         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2913         struct inet6_dev *idev = ifp->idev;
2914         struct in6_addr mcaddr;
2915
2916         if (!ifp->probes && addrconf_dad_end(ifp))
2917                 goto out;
2918
2919         read_lock(&idev->lock);
2920         if (idev->dead || !(idev->if_flags & IF_READY)) {
2921                 read_unlock(&idev->lock);
2922                 goto out;
2923         }
2924
2925         spin_lock(&ifp->lock);
2926         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
2927                 spin_unlock(&ifp->lock);
2928                 read_unlock(&idev->lock);
2929                 goto out;
2930         }
2931
2932         if (ifp->probes == 0) {
2933                 /*
2934                  * DAD was successful
2935                  */
2936
2937                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2938                 spin_unlock(&ifp->lock);
2939                 read_unlock(&idev->lock);
2940
2941                 addrconf_dad_completed(ifp);
2942
2943                 goto out;
2944         }
2945
2946         ifp->probes--;
2947         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2948         spin_unlock(&ifp->lock);
2949         read_unlock(&idev->lock);
2950
2951         /* send a neighbour solicitation for our addr */
2952         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2953         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2954 out:
2955         in6_ifa_put(ifp);
2956 }
2957
2958 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2959 {
2960         struct net_device *dev = ifp->idev->dev;
2961
2962         /*
2963          *      Configure the address for reception. Now it is valid.
2964          */
2965
2966         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2967
2968         /* If added prefix is link local and forwarding is off,
2969            start sending router solicitations.
2970          */
2971
2972         if ((ifp->idev->cnf.forwarding == 0 ||
2973              ifp->idev->cnf.forwarding == 2) &&
2974             ifp->idev->cnf.rtr_solicits > 0 &&
2975             (dev->flags&IFF_LOOPBACK) == 0 &&
2976             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2977                 /*
2978                  *      If a host as already performed a random delay
2979                  *      [...] as part of DAD [...] there is no need
2980                  *      to delay again before sending the first RS
2981                  */
2982                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2983
2984                 spin_lock_bh(&ifp->lock);
2985                 ifp->probes = 1;
2986                 ifp->idev->if_flags |= IF_RS_SENT;
2987                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2988                 spin_unlock_bh(&ifp->lock);
2989         }
2990 }
2991
2992 static void addrconf_dad_run(struct inet6_dev *idev)
2993 {
2994         struct inet6_ifaddr *ifp;
2995
2996         read_lock_bh(&idev->lock);
2997         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2998                 spin_lock(&ifp->lock);
2999                 if (ifp->flags & IFA_F_TENTATIVE &&
3000                     ifp->state == INET6_IFADDR_STATE_DAD)
3001                         addrconf_dad_kick(ifp);
3002                 spin_unlock(&ifp->lock);
3003         }
3004         read_unlock_bh(&idev->lock);
3005 }
3006
3007 #ifdef CONFIG_PROC_FS
3008 struct if6_iter_state {
3009         struct seq_net_private p;
3010         int bucket;
3011 };
3012
3013 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3014 {
3015         struct inet6_ifaddr *ifa = NULL;
3016         struct if6_iter_state *state = seq->private;
3017         struct net *net = seq_file_net(seq);
3018
3019         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3020                 struct hlist_node *n;
3021                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3022                                          addr_lst)
3023                         if (net_eq(dev_net(ifa->idev->dev), net))
3024                                 return ifa;
3025         }
3026         return NULL;
3027 }
3028
3029 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq,
3030                                          struct inet6_ifaddr *ifa)
3031 {
3032         struct if6_iter_state *state = seq->private;
3033         struct net *net = seq_file_net(seq);
3034         struct hlist_node *n = &ifa->addr_lst;
3035
3036         hlist_for_each_entry_continue_rcu_bh(ifa, n, addr_lst)
3037                 if (net_eq(dev_net(ifa->idev->dev), net))
3038                         return ifa;
3039
3040         while (++state->bucket < IN6_ADDR_HSIZE) {
3041                 hlist_for_each_entry_rcu_bh(ifa, n,
3042                                      &inet6_addr_lst[state->bucket], addr_lst) {
3043                         if (net_eq(dev_net(ifa->idev->dev), net))
3044                                 return ifa;
3045                 }
3046         }
3047
3048         return NULL;
3049 }
3050
3051 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
3052 {
3053         struct inet6_ifaddr *ifa = if6_get_first(seq);
3054
3055         if (ifa)
3056                 while (pos && (ifa = if6_get_next(seq, ifa)) != NULL)
3057                         --pos;
3058         return pos ? NULL : ifa;
3059 }
3060
3061 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
3062         __acquires(rcu_bh)
3063 {
3064         rcu_read_lock_bh();
3065         return if6_get_idx(seq, *pos);
3066 }
3067
3068 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3069 {
3070         struct inet6_ifaddr *ifa;
3071
3072         ifa = if6_get_next(seq, v);
3073         ++*pos;
3074         return ifa;
3075 }
3076
3077 static void if6_seq_stop(struct seq_file *seq, void *v)
3078         __releases(rcu_bh)
3079 {
3080         rcu_read_unlock_bh();
3081 }
3082
3083 static int if6_seq_show(struct seq_file *seq, void *v)
3084 {
3085         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3086         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3087                    &ifp->addr,
3088                    ifp->idev->dev->ifindex,
3089                    ifp->prefix_len,
3090                    ifp->scope,
3091                    ifp->flags,
3092                    ifp->idev->dev->name);
3093         return 0;
3094 }
3095
3096 static const struct seq_operations if6_seq_ops = {
3097         .start  = if6_seq_start,
3098         .next   = if6_seq_next,
3099         .show   = if6_seq_show,
3100         .stop   = if6_seq_stop,
3101 };
3102
3103 static int if6_seq_open(struct inode *inode, struct file *file)
3104 {
3105         return seq_open_net(inode, file, &if6_seq_ops,
3106                             sizeof(struct if6_iter_state));
3107 }
3108
3109 static const struct file_operations if6_fops = {
3110         .owner          = THIS_MODULE,
3111         .open           = if6_seq_open,
3112         .read           = seq_read,
3113         .llseek         = seq_lseek,
3114         .release        = seq_release_net,
3115 };
3116
3117 static int __net_init if6_proc_net_init(struct net *net)
3118 {
3119         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3120                 return -ENOMEM;
3121         return 0;
3122 }
3123
3124 static void __net_exit if6_proc_net_exit(struct net *net)
3125 {
3126        proc_net_remove(net, "if_inet6");
3127 }
3128
3129 static struct pernet_operations if6_proc_net_ops = {
3130        .init = if6_proc_net_init,
3131        .exit = if6_proc_net_exit,
3132 };
3133
3134 int __init if6_proc_init(void)
3135 {
3136         return register_pernet_subsys(&if6_proc_net_ops);
3137 }
3138
3139 void if6_proc_exit(void)
3140 {
3141         unregister_pernet_subsys(&if6_proc_net_ops);
3142 }
3143 #endif  /* CONFIG_PROC_FS */
3144
3145 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3146 /* Check if address is a home address configured on any interface. */
3147 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr)
3148 {
3149         int ret = 0;
3150         struct inet6_ifaddr *ifp = NULL;
3151         struct hlist_node *n;
3152         unsigned int hash = ipv6_addr_hash(addr);
3153
3154         rcu_read_lock_bh();
3155         hlist_for_each_entry_rcu_bh(ifp, n, &inet6_addr_lst[hash], addr_lst) {
3156                 if (!net_eq(dev_net(ifp->idev->dev), net))
3157                         continue;
3158                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3159                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3160                         ret = 1;
3161                         break;
3162                 }
3163         }
3164         rcu_read_unlock_bh();
3165         return ret;
3166 }
3167 #endif
3168
3169 /*
3170  *      Periodic address status verification
3171  */
3172
3173 static void addrconf_verify(unsigned long foo)
3174 {
3175         unsigned long now, next, next_sec, next_sched;
3176         struct inet6_ifaddr *ifp;
3177         struct hlist_node *node;
3178         int i;
3179
3180         rcu_read_lock_bh();
3181         spin_lock(&addrconf_verify_lock);
3182         now = jiffies;
3183         next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
3184
3185         del_timer(&addr_chk_timer);
3186
3187         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
3188 restart:
3189                 hlist_for_each_entry_rcu_bh(ifp, node,
3190                                          &inet6_addr_lst[i], addr_lst) {
3191                         unsigned long age;
3192
3193                         if (ifp->flags & IFA_F_PERMANENT)
3194                                 continue;
3195
3196                         spin_lock(&ifp->lock);
3197                         /* We try to batch several events at once. */
3198                         age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ;
3199
3200                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3201                             age >= ifp->valid_lft) {
3202                                 spin_unlock(&ifp->lock);
3203                                 in6_ifa_hold(ifp);
3204                                 ipv6_del_addr(ifp);
3205                                 goto restart;
3206                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3207                                 spin_unlock(&ifp->lock);
3208                                 continue;
3209                         } else if (age >= ifp->prefered_lft) {
3210                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3211                                 int deprecate = 0;
3212
3213                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3214                                         deprecate = 1;
3215                                         ifp->flags |= IFA_F_DEPRECATED;
3216                                 }
3217
3218                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3219                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3220
3221                                 spin_unlock(&ifp->lock);
3222
3223                                 if (deprecate) {
3224                                         in6_ifa_hold(ifp);
3225
3226                                         ipv6_ifa_notify(0, ifp);
3227                                         in6_ifa_put(ifp);
3228                                         goto restart;
3229                                 }
3230 #ifdef CONFIG_IPV6_PRIVACY
3231                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3232                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3233                                 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry *
3234                                         ifp->idev->cnf.dad_transmits *
3235                                         ifp->idev->nd_parms->retrans_time / HZ;
3236
3237                                 if (age >= ifp->prefered_lft - regen_advance) {
3238                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3239                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3240                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3241                                         if (!ifp->regen_count && ifpub) {
3242                                                 ifp->regen_count++;
3243                                                 in6_ifa_hold(ifp);
3244                                                 in6_ifa_hold(ifpub);
3245                                                 spin_unlock(&ifp->lock);
3246
3247                                                 spin_lock(&ifpub->lock);
3248                                                 ifpub->regen_count = 0;
3249                                                 spin_unlock(&ifpub->lock);
3250                                                 ipv6_create_tempaddr(ifpub, ifp);
3251                                                 in6_ifa_put(ifpub);
3252                                                 in6_ifa_put(ifp);
3253                                                 goto restart;
3254                                         }
3255                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3256                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3257                                 spin_unlock(&ifp->lock);
3258 #endif
3259                         } else {
3260                                 /* ifp->prefered_lft <= ifp->valid_lft */
3261                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3262                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3263                                 spin_unlock(&ifp->lock);
3264                         }
3265                 }
3266         }
3267
3268         next_sec = round_jiffies_up(next);
3269         next_sched = next;
3270
3271         /* If rounded timeout is accurate enough, accept it. */
3272         if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ))
3273                 next_sched = next_sec;
3274
3275         /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */
3276         if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX))
3277                 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX;
3278
3279         ADBG((KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n",
3280               now, next, next_sec, next_sched));
3281
3282         addr_chk_timer.expires = next_sched;
3283         add_timer(&addr_chk_timer);
3284         spin_unlock(&addrconf_verify_lock);
3285         rcu_read_unlock_bh();
3286 }
3287
3288 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3289 {
3290         struct in6_addr *pfx = NULL;
3291
3292         if (addr)
3293                 pfx = nla_data(addr);
3294
3295         if (local) {
3296                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3297                         pfx = NULL;
3298                 else
3299                         pfx = nla_data(local);
3300         }
3301
3302         return pfx;
3303 }
3304
3305 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3306         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3307         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3308         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3309 };
3310
3311 static int
3312 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3313 {
3314         struct net *net = sock_net(skb->sk);
3315         struct ifaddrmsg *ifm;
3316         struct nlattr *tb[IFA_MAX+1];
3317         struct in6_addr *pfx;
3318         int err;
3319
3320         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3321         if (err < 0)
3322                 return err;
3323
3324         ifm = nlmsg_data(nlh);
3325         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3326         if (pfx == NULL)
3327                 return -EINVAL;
3328
3329         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3330 }
3331
3332 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3333                              u32 prefered_lft, u32 valid_lft)
3334 {
3335         u32 flags;
3336         clock_t expires;
3337         unsigned long timeout;
3338
3339         if (!valid_lft || (prefered_lft > valid_lft))
3340                 return -EINVAL;
3341
3342         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3343         if (addrconf_finite_timeout(timeout)) {
3344                 expires = jiffies_to_clock_t(timeout * HZ);
3345                 valid_lft = timeout;
3346                 flags = RTF_EXPIRES;
3347         } else {
3348                 expires = 0;
3349                 flags = 0;
3350                 ifa_flags |= IFA_F_PERMANENT;
3351         }
3352
3353         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3354         if (addrconf_finite_timeout(timeout)) {
3355                 if (timeout == 0)
3356                         ifa_flags |= IFA_F_DEPRECATED;
3357                 prefered_lft = timeout;
3358         }
3359
3360         spin_lock_bh(&ifp->lock);
3361         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3362         ifp->tstamp = jiffies;
3363         ifp->valid_lft = valid_lft;
3364         ifp->prefered_lft = prefered_lft;
3365
3366         spin_unlock_bh(&ifp->lock);
3367         if (!(ifp->flags&IFA_F_TENTATIVE))
3368                 ipv6_ifa_notify(0, ifp);
3369
3370         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3371                               expires, flags);
3372         addrconf_verify(0);
3373
3374         return 0;
3375 }
3376
3377 static int
3378 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3379 {
3380         struct net *net = sock_net(skb->sk);
3381         struct ifaddrmsg *ifm;
3382         struct nlattr *tb[IFA_MAX+1];
3383         struct in6_addr *pfx;
3384         struct inet6_ifaddr *ifa;
3385         struct net_device *dev;
3386         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3387         u8 ifa_flags;
3388         int err;
3389
3390         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3391         if (err < 0)
3392                 return err;
3393
3394         ifm = nlmsg_data(nlh);
3395         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3396         if (pfx == NULL)
3397                 return -EINVAL;
3398
3399         if (tb[IFA_CACHEINFO]) {
3400                 struct ifa_cacheinfo *ci;
3401
3402                 ci = nla_data(tb[IFA_CACHEINFO]);
3403                 valid_lft = ci->ifa_valid;
3404                 preferred_lft = ci->ifa_prefered;
3405         } else {
3406                 preferred_lft = INFINITY_LIFE_TIME;
3407                 valid_lft = INFINITY_LIFE_TIME;
3408         }
3409
3410         dev =  __dev_get_by_index(net, ifm->ifa_index);
3411         if (dev == NULL)
3412                 return -ENODEV;
3413
3414         /* We ignore other flags so far. */
3415         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3416
3417         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3418         if (ifa == NULL) {
3419                 /*
3420                  * It would be best to check for !NLM_F_CREATE here but
3421                  * userspace alreay relies on not having to provide this.
3422                  */
3423                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3424                                       ifm->ifa_prefixlen, ifa_flags,
3425                                       preferred_lft, valid_lft);
3426         }
3427
3428         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3429             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3430                 err = -EEXIST;
3431         else
3432                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3433
3434         in6_ifa_put(ifa);
3435
3436         return err;
3437 }
3438
3439 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3440                           u8 scope, int ifindex)
3441 {
3442         struct ifaddrmsg *ifm;
3443
3444         ifm = nlmsg_data(nlh);
3445         ifm->ifa_family = AF_INET6;
3446         ifm->ifa_prefixlen = prefixlen;
3447         ifm->ifa_flags = flags;
3448         ifm->ifa_scope = scope;
3449         ifm->ifa_index = ifindex;
3450 }
3451
3452 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3453                          unsigned long tstamp, u32 preferred, u32 valid)
3454 {
3455         struct ifa_cacheinfo ci;
3456
3457         ci.cstamp = cstamp_delta(cstamp);
3458         ci.tstamp = cstamp_delta(tstamp);
3459         ci.ifa_prefered = preferred;
3460         ci.ifa_valid = valid;
3461
3462         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3463 }
3464
3465 static inline int rt_scope(int ifa_scope)
3466 {
3467         if (ifa_scope & IFA_HOST)
3468                 return RT_SCOPE_HOST;
3469         else if (ifa_scope & IFA_LINK)
3470                 return RT_SCOPE_LINK;
3471         else if (ifa_scope & IFA_SITE)
3472                 return RT_SCOPE_SITE;
3473         else
3474                 return RT_SCOPE_UNIVERSE;
3475 }
3476
3477 static inline int inet6_ifaddr_msgsize(void)
3478 {
3479         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3480                + nla_total_size(16) /* IFA_ADDRESS */
3481                + nla_total_size(sizeof(struct ifa_cacheinfo));
3482 }
3483
3484 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3485                              u32 pid, u32 seq, int event, unsigned int flags)
3486 {
3487         struct nlmsghdr  *nlh;
3488         u32 preferred, valid;
3489
3490         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3491         if (nlh == NULL)
3492                 return -EMSGSIZE;
3493
3494         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3495                       ifa->idev->dev->ifindex);
3496
3497         if (!(ifa->flags&IFA_F_PERMANENT)) {
3498                 preferred = ifa->prefered_lft;
3499                 valid = ifa->valid_lft;
3500                 if (preferred != INFINITY_LIFE_TIME) {
3501                         long tval = (jiffies - ifa->tstamp)/HZ;
3502                         if (preferred > tval)
3503                                 preferred -= tval;
3504                         else
3505                                 preferred = 0;
3506                         if (valid != INFINITY_LIFE_TIME) {
3507                                 if (valid > tval)
3508                                         valid -= tval;
3509                                 else
3510                                         valid = 0;
3511                         }
3512                 }
3513         } else {
3514                 preferred = INFINITY_LIFE_TIME;
3515                 valid = INFINITY_LIFE_TIME;
3516         }
3517
3518         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3519             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3520                 nlmsg_cancel(skb, nlh);
3521                 return -EMSGSIZE;
3522         }
3523
3524         return nlmsg_end(skb, nlh);
3525 }
3526
3527 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3528                                 u32 pid, u32 seq, int event, u16 flags)
3529 {
3530         struct nlmsghdr  *nlh;
3531         u8 scope = RT_SCOPE_UNIVERSE;
3532         int ifindex = ifmca->idev->dev->ifindex;
3533
3534         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3535                 scope = RT_SCOPE_SITE;
3536
3537         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3538         if (nlh == NULL)
3539                 return -EMSGSIZE;
3540
3541         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3542         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3543             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3544                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3545                 nlmsg_cancel(skb, nlh);
3546                 return -EMSGSIZE;
3547         }
3548
3549         return nlmsg_end(skb, nlh);
3550 }
3551
3552 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3553                                 u32 pid, u32 seq, int event, unsigned int flags)
3554 {
3555         struct nlmsghdr  *nlh;
3556         u8 scope = RT_SCOPE_UNIVERSE;
3557         int ifindex = ifaca->aca_idev->dev->ifindex;
3558
3559         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3560                 scope = RT_SCOPE_SITE;
3561
3562         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3563         if (nlh == NULL)
3564                 return -EMSGSIZE;
3565
3566         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3567         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3568             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3569                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3570                 nlmsg_cancel(skb, nlh);
3571                 return -EMSGSIZE;
3572         }
3573
3574         return nlmsg_end(skb, nlh);
3575 }
3576
3577 enum addr_type_t {
3578         UNICAST_ADDR,
3579         MULTICAST_ADDR,
3580         ANYCAST_ADDR,
3581 };
3582
3583 /* called with rcu_read_lock() */
3584 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3585                           struct netlink_callback *cb, enum addr_type_t type,
3586                           int s_ip_idx, int *p_ip_idx)
3587 {
3588         struct ifmcaddr6 *ifmca;
3589         struct ifacaddr6 *ifaca;
3590         int err = 1;
3591         int ip_idx = *p_ip_idx;
3592
3593         read_lock_bh(&idev->lock);
3594         switch (type) {
3595         case UNICAST_ADDR: {
3596                 struct inet6_ifaddr *ifa;
3597
3598                 /* unicast address incl. temp addr */
3599                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
3600                         if (++ip_idx < s_ip_idx)
3601                                 continue;
3602                         err = inet6_fill_ifaddr(skb, ifa,
3603                                                 NETLINK_CB(cb->skb).pid,
3604                                                 cb->nlh->nlmsg_seq,
3605                                                 RTM_NEWADDR,
3606                                                 NLM_F_MULTI);
3607                         if (err <= 0)
3608                                 break;
3609                 }
3610                 break;
3611         }
3612         case MULTICAST_ADDR:
3613                 /* multicast address */
3614                 for (ifmca = idev->mc_list; ifmca;
3615                      ifmca = ifmca->next, ip_idx++) {
3616                         if (ip_idx < s_ip_idx)
3617                                 continue;
3618                         err = inet6_fill_ifmcaddr(skb, ifmca,
3619                                                   NETLINK_CB(cb->skb).pid,
3620                                                   cb->nlh->nlmsg_seq,
3621                                                   RTM_GETMULTICAST,
3622                                                   NLM_F_MULTI);
3623                         if (err <= 0)
3624                                 break;
3625                 }
3626                 break;
3627         case ANYCAST_ADDR:
3628                 /* anycast address */
3629                 for (ifaca = idev->ac_list; ifaca;
3630                      ifaca = ifaca->aca_next, ip_idx++) {
3631                         if (ip_idx < s_ip_idx)
3632                                 continue;
3633                         err = inet6_fill_ifacaddr(skb, ifaca,
3634                                                   NETLINK_CB(cb->skb).pid,
3635                                                   cb->nlh->nlmsg_seq,
3636                                                   RTM_GETANYCAST,
3637                                                   NLM_F_MULTI);
3638                         if (err <= 0)
3639                                 break;
3640                 }
3641                 break;
3642         default:
3643                 break;
3644         }
3645         read_unlock_bh(&idev->lock);
3646         *p_ip_idx = ip_idx;
3647         return err;
3648 }
3649
3650 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3651                            enum addr_type_t type)
3652 {
3653         struct net *net = sock_net(skb->sk);
3654         int h, s_h;
3655         int idx, ip_idx;
3656         int s_idx, s_ip_idx;
3657         struct net_device *dev;
3658         struct inet6_dev *idev;
3659         struct hlist_head *head;
3660         struct hlist_node *node;
3661
3662         s_h = cb->args[0];
3663         s_idx = idx = cb->args[1];
3664         s_ip_idx = ip_idx = cb->args[2];
3665
3666         rcu_read_lock();
3667         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3668                 idx = 0;
3669                 head = &net->dev_index_head[h];
3670                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3671                         if (idx < s_idx)
3672                                 goto cont;
3673                         if (h > s_h || idx > s_idx)
3674                                 s_ip_idx = 0;
3675                         ip_idx = 0;
3676                         idev = __in6_dev_get(dev);
3677                         if (!idev)
3678                                 goto cont;
3679
3680                         if (in6_dump_addrs(idev, skb, cb, type,
3681                                            s_ip_idx, &ip_idx) <= 0)
3682                                 goto done;
3683 cont:
3684                         idx++;
3685                 }
3686         }
3687 done:
3688         rcu_read_unlock();
3689         cb->args[0] = h;
3690         cb->args[1] = idx;
3691         cb->args[2] = ip_idx;
3692
3693         return skb->len;
3694 }
3695
3696 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3697 {
3698         enum addr_type_t type = UNICAST_ADDR;
3699
3700         return inet6_dump_addr(skb, cb, type);
3701 }
3702
3703 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3704 {
3705         enum addr_type_t type = MULTICAST_ADDR;
3706
3707         return inet6_dump_addr(skb, cb, type);
3708 }
3709
3710
3711 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3712 {
3713         enum addr_type_t type = ANYCAST_ADDR;
3714
3715         return inet6_dump_addr(skb, cb, type);
3716 }
3717
3718 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3719                              void *arg)
3720 {
3721         struct net *net = sock_net(in_skb->sk);
3722         struct ifaddrmsg *ifm;
3723         struct nlattr *tb[IFA_MAX+1];
3724         struct in6_addr *addr = NULL;
3725         struct net_device *dev = NULL;
3726         struct inet6_ifaddr *ifa;
3727         struct sk_buff *skb;
3728         int err;
3729
3730         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3731         if (err < 0)
3732                 goto errout;
3733
3734         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3735         if (addr == NULL) {
3736                 err = -EINVAL;
3737                 goto errout;
3738         }
3739
3740         ifm = nlmsg_data(nlh);
3741         if (ifm->ifa_index)
3742                 dev = __dev_get_by_index(net, ifm->ifa_index);
3743
3744         ifa = ipv6_get_ifaddr(net, addr, dev, 1);
3745         if (!ifa) {
3746                 err = -EADDRNOTAVAIL;
3747                 goto errout;
3748         }
3749
3750         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL);
3751         if (!skb) {
3752                 err = -ENOBUFS;
3753                 goto errout_ifa;
3754         }
3755
3756         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3757                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3758         if (err < 0) {
3759                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3760                 WARN_ON(err == -EMSGSIZE);
3761                 kfree_skb(skb);
3762                 goto errout_ifa;
3763         }
3764         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3765 errout_ifa:
3766         in6_ifa_put(ifa);
3767 errout:
3768         return err;
3769 }
3770
3771 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3772 {
3773         struct sk_buff *skb;
3774         struct net *net = dev_net(ifa->idev->dev);
3775         int err = -ENOBUFS;
3776
3777         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3778         if (skb == NULL)
3779                 goto errout;
3780
3781         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3782         if (err < 0) {
3783                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3784                 WARN_ON(err == -EMSGSIZE);
3785                 kfree_skb(skb);
3786                 goto errout;
3787         }
3788         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3789         return;
3790 errout:
3791         if (err < 0)
3792                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3793 }
3794
3795 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3796                                 __s32 *array, int bytes)
3797 {
3798         BUG_ON(bytes < (DEVCONF_MAX * 4));
3799
3800         memset(array, 0, bytes);
3801         array[DEVCONF_FORWARDING] = cnf->forwarding;
3802         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3803         array[DEVCONF_MTU6] = cnf->mtu6;
3804         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3805         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3806         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3807         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3808         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3809         array[DEVCONF_RTR_SOLICIT_INTERVAL] =
3810                 jiffies_to_msecs(cnf->rtr_solicit_interval);
3811         array[DEVCONF_RTR_SOLICIT_DELAY] =
3812                 jiffies_to_msecs(cnf->rtr_solicit_delay);
3813         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3814 #ifdef CONFIG_IPV6_PRIVACY
3815         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3816         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3817         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3818         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3819         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3820 #endif
3821         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3822         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3823         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3824 #ifdef CONFIG_IPV6_ROUTER_PREF
3825         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3826         array[DEVCONF_RTR_PROBE_INTERVAL] =
3827                 jiffies_to_msecs(cnf->rtr_probe_interval);
3828 #ifdef CONFIG_IPV6_ROUTE_INFO
3829         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3830 #endif
3831 #endif
3832         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3833         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3834 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3835         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3836 #endif
3837 #ifdef CONFIG_IPV6_MROUTE
3838         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3839 #endif
3840         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3841         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3842         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3843 }
3844
3845 static inline size_t inet6_ifla6_size(void)
3846 {
3847         return nla_total_size(4) /* IFLA_INET6_FLAGS */
3848              + nla_total_size(sizeof(struct ifla_cacheinfo))
3849              + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3850              + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3851              + nla_total_size(ICMP6_MIB_MAX * 8); /* IFLA_INET6_ICMP6STATS */
3852 }
3853
3854 static inline size_t inet6_if_nlmsg_size(void)
3855 {
3856         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3857                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3858                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3859                + nla_total_size(4) /* IFLA_MTU */
3860                + nla_total_size(4) /* IFLA_LINK */
3861                + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */
3862 }
3863
3864 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib,
3865                                       int items, int bytes)
3866 {
3867         int i;
3868         int pad = bytes - sizeof(u64) * items;
3869         BUG_ON(pad < 0);
3870
3871         /* Use put_unaligned() because stats may not be aligned for u64. */
3872         put_unaligned(items, &stats[0]);
3873         for (i = 1; i < items; i++)
3874                 put_unaligned(atomic_long_read(&mib[i]), &stats[i]);
3875
3876         memset(&stats[items], 0, pad);
3877 }
3878
3879 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib,
3880                                       int items, int bytes, size_t syncpoff)
3881 {
3882         int i;
3883         int pad = bytes - sizeof(u64) * items;
3884         BUG_ON(pad < 0);
3885
3886         /* Use put_unaligned() because stats may not be aligned for u64. */
3887         put_unaligned(items, &stats[0]);
3888         for (i = 1; i < items; i++)
3889                 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]);
3890
3891         memset(&stats[items], 0, pad);
3892 }
3893
3894 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3895                              int bytes)
3896 {
3897         switch (attrtype) {
3898         case IFLA_INET6_STATS:
3899                 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6,
3900                                      IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp));
3901                 break;
3902         case IFLA_INET6_ICMP6STATS:
3903                 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes);
3904                 break;
3905         }
3906 }
3907
3908 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev)
3909 {
3910         struct nlattr *nla;
3911         struct ifla_cacheinfo ci;
3912
3913         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3914
3915         ci.max_reasm_len = IPV6_MAXPLEN;
3916         ci.tstamp = cstamp_delta(idev->tstamp);
3917         ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time);
3918         ci.retrans_time = jiffies_to_msecs(idev->nd_parms->retrans_time);
3919         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3920
3921         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3922         if (nla == NULL)
3923                 goto nla_put_failure;
3924         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3925
3926         /* XXX - MC not implemented */
3927
3928         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3929         if (nla == NULL)
3930                 goto nla_put_failure;
3931         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3932
3933         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3934         if (nla == NULL)
3935                 goto nla_put_failure;
3936         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3937
3938         return 0;
3939
3940 nla_put_failure:
3941         return -EMSGSIZE;
3942 }
3943
3944 static size_t inet6_get_link_af_size(const struct net_device *dev)
3945 {
3946         if (!__in6_dev_get(dev))
3947                 return 0;
3948
3949         return inet6_ifla6_size();
3950 }
3951
3952 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev)
3953 {
3954         struct inet6_dev *idev = __in6_dev_get(dev);
3955
3956         if (!idev)
3957                 return -ENODATA;
3958
3959         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3960                 return -EMSGSIZE;
3961
3962         return 0;
3963 }
3964
3965 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3966                              u32 pid, u32 seq, int event, unsigned int flags)
3967 {
3968         struct net_device *dev = idev->dev;
3969         struct ifinfomsg *hdr;
3970         struct nlmsghdr *nlh;
3971         void *protoinfo;
3972
3973         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3974         if (nlh == NULL)
3975                 return -EMSGSIZE;
3976
3977         hdr = nlmsg_data(nlh);
3978         hdr->ifi_family = AF_INET6;
3979         hdr->__ifi_pad = 0;
3980         hdr->ifi_type = dev->type;
3981         hdr->ifi_index = dev->ifindex;
3982         hdr->ifi_flags = dev_get_flags(dev);
3983         hdr->ifi_change = 0;
3984
3985         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3986
3987         if (dev->addr_len)
3988                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3989
3990         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3991         if (dev->ifindex != dev->iflink)
3992                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3993
3994         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3995         if (protoinfo == NULL)
3996                 goto nla_put_failure;
3997
3998         if (inet6_fill_ifla6_attrs(skb, idev) < 0)
3999                 goto nla_put_failure;
4000
4001         nla_nest_end(skb, protoinfo);
4002         return nlmsg_end(skb, nlh);
4003
4004 nla_put_failure:
4005         nlmsg_cancel(skb, nlh);
4006         return -EMSGSIZE;
4007 }
4008
4009 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
4010 {
4011         struct net *net = sock_net(skb->sk);
4012         int h, s_h;
4013         int idx = 0, s_idx;
4014         struct net_device *dev;
4015         struct inet6_dev *idev;
4016         struct hlist_head *head;
4017         struct hlist_node *node;
4018
4019         s_h = cb->args[0];
4020         s_idx = cb->args[1];
4021
4022         rcu_read_lock();
4023         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
4024                 idx = 0;
4025                 head = &net->dev_index_head[h];
4026                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
4027                         if (idx < s_idx)
4028                                 goto cont;
4029                         idev = __in6_dev_get(dev);
4030                         if (!idev)
4031                                 goto cont;
4032                         if (inet6_fill_ifinfo(skb, idev,
4033                                               NETLINK_CB(cb->skb).pid,
4034                                               cb->nlh->nlmsg_seq,
4035                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
4036                                 goto out;
4037 cont:
4038                         idx++;
4039                 }
4040         }
4041 out:
4042         rcu_read_unlock();
4043         cb->args[1] = idx;
4044         cb->args[0] = h;
4045
4046         return skb->len;
4047 }
4048
4049 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
4050 {
4051         struct sk_buff *skb;
4052         struct net *net = dev_net(idev->dev);
4053         int err = -ENOBUFS;
4054
4055         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
4056         if (skb == NULL)
4057                 goto errout;
4058
4059         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
4060         if (err < 0) {
4061                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
4062                 WARN_ON(err == -EMSGSIZE);
4063                 kfree_skb(skb);
4064                 goto errout;
4065         }
4066         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC);
4067         return;
4068 errout:
4069         if (err < 0)
4070                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err);
4071 }
4072
4073 static inline size_t inet6_prefix_nlmsg_size(void)
4074 {
4075         return NLMSG_ALIGN(sizeof(struct prefixmsg))
4076                + nla_total_size(sizeof(struct in6_addr))
4077                + nla_total_size(sizeof(struct prefix_cacheinfo));
4078 }
4079
4080 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
4081                              struct prefix_info *pinfo, u32 pid, u32 seq,
4082                              int event, unsigned int flags)
4083 {
4084         struct prefixmsg *pmsg;
4085         struct nlmsghdr *nlh;
4086         struct prefix_cacheinfo ci;
4087
4088         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
4089         if (nlh == NULL)
4090                 return -EMSGSIZE;
4091
4092         pmsg = nlmsg_data(nlh);
4093         pmsg->prefix_family = AF_INET6;
4094         pmsg->prefix_pad1 = 0;
4095         pmsg->prefix_pad2 = 0;
4096         pmsg->prefix_ifindex = idev->dev->ifindex;
4097         pmsg->prefix_len = pinfo->prefix_len;
4098         pmsg->prefix_type = pinfo->type;
4099         pmsg->prefix_pad3 = 0;
4100         pmsg->prefix_flags = 0;
4101         if (pinfo->onlink)
4102                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
4103         if (pinfo->autoconf)
4104                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
4105
4106         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
4107
4108         ci.preferred_time = ntohl(pinfo->prefered);
4109         ci.valid_time = ntohl(pinfo->valid);
4110         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
4111
4112         return nlmsg_end(skb, nlh);
4113
4114 nla_put_failure:
4115         nlmsg_cancel(skb, nlh);
4116         return -EMSGSIZE;
4117 }
4118
4119 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
4120                          struct prefix_info *pinfo)
4121 {
4122         struct sk_buff *skb;
4123         struct net *net = dev_net(idev->dev);
4124         int err = -ENOBUFS;
4125
4126         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
4127         if (skb == NULL)
4128                 goto errout;
4129
4130         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
4131         if (err < 0) {
4132                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
4133                 WARN_ON(err == -EMSGSIZE);
4134                 kfree_skb(skb);
4135                 goto errout;
4136         }
4137         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
4138         return;
4139 errout:
4140         if (err < 0)
4141                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
4142 }
4143
4144 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4145 {
4146         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4147
4148         switch (event) {
4149         case RTM_NEWADDR:
4150                 /*
4151                  * If the address was optimistic
4152                  * we inserted the route at the start of
4153                  * our DAD process, so we don't need
4154                  * to do it again
4155                  */
4156                 if (!(ifp->rt->rt6i_node))
4157                         ip6_ins_rt(ifp->rt);
4158                 if (ifp->idev->cnf.forwarding)
4159                         addrconf_join_anycast(ifp);
4160                 break;
4161         case RTM_DELADDR:
4162                 if (ifp->idev->cnf.forwarding)
4163                         addrconf_leave_anycast(ifp);
4164                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4165                 dst_hold(&ifp->rt->dst);
4166
4167                 if (ip6_del_rt(ifp->rt))
4168                         dst_free(&ifp->rt->dst);
4169                 break;
4170         }
4171 }
4172
4173 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4174 {
4175         rcu_read_lock_bh();
4176         if (likely(ifp->idev->dead == 0))
4177                 __ipv6_ifa_notify(event, ifp);
4178         rcu_read_unlock_bh();
4179 }
4180
4181 #ifdef CONFIG_SYSCTL
4182
4183 static
4184 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4185                            void __user *buffer, size_t *lenp, loff_t *ppos)
4186 {
4187         int *valp = ctl->data;
4188         int val = *valp;
4189         loff_t pos = *ppos;
4190         int ret;
4191
4192         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4193
4194         if (write)
4195                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4196         if (ret)
4197                 *ppos = pos;
4198         return ret;
4199 }
4200
4201 static void dev_disable_change(struct inet6_dev *idev)
4202 {
4203         if (!idev || !idev->dev)
4204                 return;
4205
4206         if (idev->cnf.disable_ipv6)
4207                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4208         else
4209                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4210 }
4211
4212 static void addrconf_disable_change(struct net *net, __s32 newf)
4213 {
4214         struct net_device *dev;
4215         struct inet6_dev *idev;
4216
4217         rcu_read_lock();
4218         for_each_netdev_rcu(net, dev) {
4219                 idev = __in6_dev_get(dev);
4220                 if (idev) {
4221                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4222                         idev->cnf.disable_ipv6 = newf;
4223                         if (changed)
4224                                 dev_disable_change(idev);
4225                 }
4226         }
4227         rcu_read_unlock();
4228 }
4229
4230 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4231 {
4232         struct net *net;
4233
4234         net = (struct net *)table->extra2;
4235
4236         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4237                 return 0;
4238
4239         if (!rtnl_trylock()) {
4240                 /* Restore the original values before restarting */
4241                 *p = old;
4242                 return restart_syscall();
4243         }
4244
4245         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4246                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4247                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4248                 addrconf_disable_change(net, newf);
4249         } else if ((!*p) ^ (!old))
4250                 dev_disable_change((struct inet6_dev *)table->extra1);
4251
4252         rtnl_unlock();
4253         return 0;
4254 }
4255
4256 static
4257 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4258                             void __user *buffer, size_t *lenp, loff_t *ppos)
4259 {
4260         int *valp = ctl->data;
4261         int val = *valp;
4262         loff_t pos = *ppos;
4263         int ret;
4264
4265         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4266
4267         if (write)
4268                 ret = addrconf_disable_ipv6(ctl, valp, val);
4269         if (ret)
4270                 *ppos = pos;
4271         return ret;
4272 }
4273
4274 static struct addrconf_sysctl_table
4275 {
4276         struct ctl_table_header *sysctl_header;
4277         ctl_table addrconf_vars[DEVCONF_MAX+1];
4278         char *dev_name;
4279 } addrconf_sysctl __read_mostly = {
4280         .sysctl_header = NULL,
4281         .addrconf_vars = {
4282                 {
4283                         .procname       = "forwarding",
4284                         .data           = &ipv6_devconf.forwarding,
4285                         .maxlen         = sizeof(int),
4286                         .mode           = 0644,
4287                         .proc_handler   = addrconf_sysctl_forward,
4288                 },
4289                 {
4290                         .procname       = "hop_limit",
4291                         .data           = &ipv6_devconf.hop_limit,
4292                         .maxlen         = sizeof(int),
4293                         .mode           = 0644,
4294                         .proc_handler   = proc_dointvec,
4295                 },
4296                 {
4297                         .procname       = "mtu",
4298                         .data           = &ipv6_devconf.mtu6,
4299                         .maxlen         = sizeof(int),
4300                         .mode           = 0644,
4301                         .proc_handler   = proc_dointvec,
4302                 },
4303                 {
4304                         .procname       = "accept_ra",
4305                         .data           = &ipv6_devconf.accept_ra,
4306                         .maxlen         = sizeof(int),
4307                         .mode           = 0644,
4308                         .proc_handler   = proc_dointvec,
4309                 },
4310                 {
4311                         .procname       = "accept_redirects",
4312                         .data           = &ipv6_devconf.accept_redirects,
4313                         .maxlen         = sizeof(int),
4314                         .mode           = 0644,
4315                         .proc_handler   = proc_dointvec,
4316                 },
4317                 {
4318                         .procname       = "autoconf",
4319                         .data           = &ipv6_devconf.autoconf,
4320                         .maxlen         = sizeof(int),
4321                         .mode           = 0644,
4322                         .proc_handler   = proc_dointvec,
4323                 },
4324                 {
4325                         .procname       = "dad_transmits",
4326                         .data           = &ipv6_devconf.dad_transmits,
4327                         .maxlen         = sizeof(int),
4328                         .mode           = 0644,
4329                         .proc_handler   = proc_dointvec,
4330                 },
4331                 {
4332                         .procname       = "router_solicitations",
4333                         .data           = &ipv6_devconf.rtr_solicits,
4334                         .maxlen         = sizeof(int),
4335                         .mode           = 0644,
4336                         .proc_handler   = proc_dointvec,
4337                 },
4338                 {
4339                         .procname       = "router_solicitation_interval",
4340                         .data           = &ipv6_devconf.rtr_solicit_interval,
4341                         .maxlen         = sizeof(int),
4342                         .mode           = 0644,
4343                         .proc_handler   = proc_dointvec_jiffies,
4344                 },
4345                 {
4346                         .procname       = "router_solicitation_delay",
4347                         .data           = &ipv6_devconf.rtr_solicit_delay,
4348                         .maxlen         = sizeof(int),
4349                         .mode           = 0644,
4350                         .proc_handler   = proc_dointvec_jiffies,
4351                 },
4352                 {
4353                         .procname       = "force_mld_version",
4354                         .data           = &ipv6_devconf.force_mld_version,
4355                         .maxlen         = sizeof(int),
4356                         .mode           = 0644,
4357                         .proc_handler   = proc_dointvec,
4358                 },
4359 #ifdef CONFIG_IPV6_PRIVACY
4360                 {
4361                         .procname       = "use_tempaddr",
4362                         .data           = &ipv6_devconf.use_tempaddr,
4363                         .maxlen         = sizeof(int),
4364                         .mode           = 0644,
4365                         .proc_handler   = proc_dointvec,
4366                 },
4367                 {
4368                         .procname       = "temp_valid_lft",
4369                         .data           = &ipv6_devconf.temp_valid_lft,
4370                         .maxlen         = sizeof(int),
4371                         .mode           = 0644,
4372                         .proc_handler   = proc_dointvec,
4373                 },
4374                 {
4375                         .procname       = "temp_prefered_lft",
4376                         .data           = &ipv6_devconf.temp_prefered_lft,
4377                         .maxlen         = sizeof(int),
4378                         .mode           = 0644,
4379                         .proc_handler   = proc_dointvec,
4380                 },
4381                 {
4382                         .procname       = "regen_max_retry",
4383                         .data           = &ipv6_devconf.regen_max_retry,
4384                         .maxlen         = sizeof(int),
4385                         .mode           = 0644,
4386                         .proc_handler   = proc_dointvec,
4387                 },
4388                 {
4389                         .procname       = "max_desync_factor",
4390                         .data           = &ipv6_devconf.max_desync_factor,
4391                         .maxlen         = sizeof(int),
4392                         .mode           = 0644,
4393                         .proc_handler   = proc_dointvec,
4394                 },
4395 #endif
4396                 {
4397                         .procname       = "max_addresses",
4398                         .data           = &ipv6_devconf.max_addresses,
4399                         .maxlen         = sizeof(int),
4400                         .mode           = 0644,
4401                         .proc_handler   = proc_dointvec,
4402                 },
4403                 {
4404                         .procname       = "accept_ra_defrtr",
4405                         .data           = &ipv6_devconf.accept_ra_defrtr,
4406                         .maxlen         = sizeof(int),
4407                         .mode           = 0644,
4408                         .proc_handler   = proc_dointvec,
4409                 },
4410                 {
4411                         .procname       = "accept_ra_pinfo",
4412                         .data           = &ipv6_devconf.accept_ra_pinfo,
4413                         .maxlen         = sizeof(int),
4414                         .mode           = 0644,
4415                         .proc_handler   = proc_dointvec,
4416                 },
4417 #ifdef CONFIG_IPV6_ROUTER_PREF
4418                 {
4419                         .procname       = "accept_ra_rtr_pref",
4420                         .data           = &ipv6_devconf.accept_ra_rtr_pref,
4421                         .maxlen         = sizeof(int),
4422                         .mode           = 0644,
4423                         .proc_handler   = proc_dointvec,
4424                 },
4425                 {
4426                         .procname       = "router_probe_interval",
4427                         .data           = &ipv6_devconf.rtr_probe_interval,
4428                         .maxlen         = sizeof(int),
4429                         .mode           = 0644,
4430                         .proc_handler   = proc_dointvec_jiffies,
4431                 },
4432 #ifdef CONFIG_IPV6_ROUTE_INFO
4433                 {
4434                         .procname       = "accept_ra_rt_info_max_plen",
4435                         .data           = &ipv6_devconf.accept_ra_rt_info_max_plen,
4436                         .maxlen         = sizeof(int),
4437                         .mode           = 0644,
4438                         .proc_handler   = proc_dointvec,
4439                 },
4440 #endif
4441 #endif
4442                 {
4443                         .procname       = "proxy_ndp",
4444                         .data           = &ipv6_devconf.proxy_ndp,
4445                         .maxlen         = sizeof(int),
4446                         .mode           = 0644,
4447                         .proc_handler   = proc_dointvec,
4448                 },
4449                 {
4450                         .procname       = "accept_source_route",
4451                         .data           = &ipv6_devconf.accept_source_route,
4452                         .maxlen         = sizeof(int),
4453                         .mode           = 0644,
4454                         .proc_handler   = proc_dointvec,
4455                 },
4456 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4457                 {
4458                         .procname       = "optimistic_dad",
4459                         .data           = &ipv6_devconf.optimistic_dad,
4460                         .maxlen         = sizeof(int),
4461                         .mode           = 0644,
4462                         .proc_handler   = proc_dointvec,
4463
4464                 },
4465 #endif
4466 #ifdef CONFIG_IPV6_MROUTE
4467                 {
4468                         .procname       = "mc_forwarding",
4469                         .data           = &ipv6_devconf.mc_forwarding,
4470                         .maxlen         = sizeof(int),
4471                         .mode           = 0444,
4472                         .proc_handler   = proc_dointvec,
4473                 },
4474 #endif
4475                 {
4476                         .procname       = "disable_ipv6",
4477                         .data           = &ipv6_devconf.disable_ipv6,
4478                         .maxlen         = sizeof(int),
4479                         .mode           = 0644,
4480                         .proc_handler   = addrconf_sysctl_disable,
4481                 },
4482                 {
4483                         .procname       = "accept_dad",
4484                         .data           = &ipv6_devconf.accept_dad,
4485                         .maxlen         = sizeof(int),
4486                         .mode           = 0644,
4487                         .proc_handler   = proc_dointvec,
4488                 },
4489                 {
4490                         .procname       = "force_tllao",
4491                         .data           = &ipv6_devconf.force_tllao,
4492                         .maxlen         = sizeof(int),
4493                         .mode           = 0644,
4494                         .proc_handler   = proc_dointvec
4495                 },
4496                 {
4497                         /* sentinel */
4498                 }
4499         },
4500 };
4501
4502 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4503                 struct inet6_dev *idev, struct ipv6_devconf *p)
4504 {
4505         int i;
4506         struct addrconf_sysctl_table *t;
4507
4508 #define ADDRCONF_CTL_PATH_DEV   3
4509
4510         struct ctl_path addrconf_ctl_path[] = {
4511                 { .procname = "net", },
4512                 { .procname = "ipv6", },
4513                 { .procname = "conf", },
4514                 { /* to be set */ },
4515                 { },
4516         };
4517
4518
4519         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4520         if (t == NULL)
4521                 goto out;
4522
4523         for (i = 0; t->addrconf_vars[i].data; i++) {
4524                 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf;
4525                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4526                 t->addrconf_vars[i].extra2 = net;
4527         }
4528
4529         /*
4530          * Make a copy of dev_name, because '.procname' is regarded as const
4531          * by sysctl and we wouldn't want anyone to change it under our feet
4532          * (see SIOCSIFNAME).
4533          */
4534         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4535         if (!t->dev_name)
4536                 goto free;
4537
4538         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4539
4540         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4541                         t->addrconf_vars);
4542         if (t->sysctl_header == NULL)
4543                 goto free_procname;
4544
4545         p->sysctl = t;
4546         return 0;
4547
4548 free_procname:
4549         kfree(t->dev_name);
4550 free:
4551         kfree(t);
4552 out:
4553         return -ENOBUFS;
4554 }
4555
4556 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4557 {
4558         struct addrconf_sysctl_table *t;
4559
4560         if (p->sysctl == NULL)
4561                 return;
4562
4563         t = p->sysctl;
4564         p->sysctl = NULL;
4565         unregister_net_sysctl_table(t->sysctl_header);
4566         kfree(t->dev_name);
4567         kfree(t);
4568 }
4569
4570 static void addrconf_sysctl_register(struct inet6_dev *idev)
4571 {
4572         neigh_sysctl_register(idev->dev, idev->nd_parms, "ipv6",
4573                               &ndisc_ifinfo_sysctl_change);
4574         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4575                                         idev, &idev->cnf);
4576 }
4577
4578 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4579 {
4580         __addrconf_sysctl_unregister(&idev->cnf);
4581         neigh_sysctl_unregister(idev->nd_parms);
4582 }
4583
4584
4585 #endif
4586
4587 static int __net_init addrconf_init_net(struct net *net)
4588 {
4589         int err;
4590         struct ipv6_devconf *all, *dflt;
4591
4592         err = -ENOMEM;
4593         all = &ipv6_devconf;
4594         dflt = &ipv6_devconf_dflt;
4595
4596         if (!net_eq(net, &init_net)) {
4597                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4598                 if (all == NULL)
4599                         goto err_alloc_all;
4600
4601                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4602                 if (dflt == NULL)
4603                         goto err_alloc_dflt;
4604         } else {
4605                 /* these will be inherited by all namespaces */
4606                 dflt->autoconf = ipv6_defaults.autoconf;
4607                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4608         }
4609
4610         net->ipv6.devconf_all = all;
4611         net->ipv6.devconf_dflt = dflt;
4612
4613 #ifdef CONFIG_SYSCTL
4614         err = __addrconf_sysctl_register(net, "all", NULL, all);
4615         if (err < 0)
4616                 goto err_reg_all;
4617
4618         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4619         if (err < 0)
4620                 goto err_reg_dflt;
4621 #endif
4622         return 0;
4623
4624 #ifdef CONFIG_SYSCTL
4625 err_reg_dflt:
4626         __addrconf_sysctl_unregister(all);
4627 err_reg_all:
4628         kfree(dflt);
4629 #endif
4630 err_alloc_dflt:
4631         kfree(all);
4632 err_alloc_all:
4633         return err;
4634 }
4635
4636 static void __net_exit addrconf_exit_net(struct net *net)
4637 {
4638 #ifdef CONFIG_SYSCTL
4639         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4640         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4641 #endif
4642         if (!net_eq(net, &init_net)) {
4643                 kfree(net->ipv6.devconf_dflt);
4644                 kfree(net->ipv6.devconf_all);
4645         }
4646 }
4647
4648 static struct pernet_operations addrconf_ops = {
4649         .init = addrconf_init_net,
4650         .exit = addrconf_exit_net,
4651 };
4652
4653 /*
4654  *      Device notifier
4655  */
4656
4657 int register_inet6addr_notifier(struct notifier_block *nb)
4658 {
4659         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4660 }
4661 EXPORT_SYMBOL(register_inet6addr_notifier);
4662
4663 int unregister_inet6addr_notifier(struct notifier_block *nb)
4664 {
4665         return atomic_notifier_chain_unregister(&inet6addr_chain, nb);
4666 }
4667 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4668
4669 static struct rtnl_af_ops inet6_ops = {
4670         .family           = AF_INET6,
4671         .fill_link_af     = inet6_fill_link_af,
4672         .get_link_af_size = inet6_get_link_af_size,
4673 };
4674
4675 /*
4676  *      Init / cleanup code
4677  */
4678
4679 int __init addrconf_init(void)
4680 {
4681         int i, err;
4682
4683         err = ipv6_addr_label_init();
4684         if (err < 0) {
4685                 printk(KERN_CRIT "IPv6 Addrconf:"
4686                        " cannot initialize default policy table: %d.\n", err);
4687                 goto out;
4688         }
4689
4690         err = register_pernet_subsys(&addrconf_ops);
4691         if (err < 0)
4692                 goto out_addrlabel;
4693
4694         /* The addrconf netdev notifier requires that loopback_dev
4695          * has it's ipv6 private information allocated and setup
4696          * before it can bring up and give link-local addresses
4697          * to other devices which are up.
4698          *
4699          * Unfortunately, loopback_dev is not necessarily the first
4700          * entry in the global dev_base list of net devices.  In fact,
4701          * it is likely to be the very last entry on that list.
4702          * So this causes the notifier registry below to try and
4703          * give link-local addresses to all devices besides loopback_dev
4704          * first, then loopback_dev, which cases all the non-loopback_dev
4705          * devices to fail to get a link-local address.
4706          *
4707          * So, as a temporary fix, allocate the ipv6 structure for
4708          * loopback_dev first by hand.
4709          * Longer term, all of the dependencies ipv6 has upon the loopback
4710          * device and it being up should be removed.
4711          */
4712         rtnl_lock();
4713         if (!ipv6_add_dev(init_net.loopback_dev))
4714                 err = -ENOMEM;
4715         rtnl_unlock();
4716         if (err)
4717                 goto errlo;
4718
4719         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4720                 INIT_HLIST_HEAD(&inet6_addr_lst[i]);
4721
4722         register_netdevice_notifier(&ipv6_dev_notf);
4723
4724         addrconf_verify(0);
4725
4726         err = rtnl_af_register(&inet6_ops);
4727         if (err < 0)
4728                 goto errout_af;
4729
4730         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo,
4731                               NULL);
4732         if (err < 0)
4733                 goto errout;
4734
4735         /* Only the first call to __rtnl_register can fail */
4736         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL);
4737         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL);
4738         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr,
4739                         inet6_dump_ifaddr, NULL);
4740         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL,
4741                         inet6_dump_ifmcaddr, NULL);
4742         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL,
4743                         inet6_dump_ifacaddr, NULL);
4744
4745         ipv6_addr_label_rtnl_register();
4746
4747         return 0;
4748 errout:
4749         rtnl_af_unregister(&inet6_ops);
4750 errout_af:
4751         unregister_netdevice_notifier(&ipv6_dev_notf);
4752 errlo:
4753         unregister_pernet_subsys(&addrconf_ops);
4754 out_addrlabel:
4755         ipv6_addr_label_cleanup();
4756 out:
4757         return err;
4758 }
4759
4760 void addrconf_cleanup(void)
4761 {
4762         struct net_device *dev;
4763         int i;
4764
4765         unregister_netdevice_notifier(&ipv6_dev_notf);
4766         unregister_pernet_subsys(&addrconf_ops);
4767         ipv6_addr_label_cleanup();
4768
4769         rtnl_lock();
4770
4771         __rtnl_af_unregister(&inet6_ops);
4772
4773         /* clean dev list */
4774         for_each_netdev(&init_net, dev) {
4775                 if (__in6_dev_get(dev) == NULL)
4776                         continue;
4777                 addrconf_ifdown(dev, 1);
4778         }
4779         addrconf_ifdown(init_net.loopback_dev, 2);
4780
4781         /*
4782          *      Check hash table.
4783          */
4784         spin_lock_bh(&addrconf_hash_lock);
4785         for (i = 0; i < IN6_ADDR_HSIZE; i++)
4786                 WARN_ON(!hlist_empty(&inet6_addr_lst[i]));
4787         spin_unlock_bh(&addrconf_hash_lock);
4788
4789         del_timer(&addr_chk_timer);
4790         rtnl_unlock();
4791 }