net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <[email protected]>
   7  *      Alexey Kuznetsov        <[email protected]>
   8  *
   9  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
  10  *
  11  *      This program is free software; you can redistribute it and/or
  12  *      modify it under the terms of the GNU General Public License
  13  *      as published by the Free Software Foundation; either version
  14  *      2 of the License, or (at your option) any later version.
  15  */
  16
  17 /*
  18  *      Changes:
  19  *
  20  *      Janos Farkas                    :       delete timer on ifdown
  21  *      <[email protected]>
  22  *      Andi Kleen                      :       kill double kfree on module
  23  *                                              unload.
  24  *      Maciej W. Rozycki               :       FDDI support
  25  *      sekiya@USAGI                    :       Don't send too many RS
  26  *                                              packets.
  27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  28  *                                              packets.
  29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  30  *                                              address validation timer.
  31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  32  *                                              support.
  33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  34  *                                              address on a same interface.
  35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  37  *                                              seq_file.
  38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  39  *                                              selection; consider scope,
  40  *                                              status etc.
  41  */
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/socket.h>
  46 #include <linux/sockios.h>
  47 #include <linux/net.h>
  48 #include <linux/in6.h>
  49 #include <linux/netdevice.h>
  50 #include <linux/if_addr.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/if_arcnet.h>
  53 #include <linux/if_infiniband.h>
  54 #include <linux/route.h>
  55 #include <linux/inetdevice.h>
  56 #include <linux/init.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 <asm/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 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #ifdef CONFIG_SYSCTL
 104 static void addrconf_sysctl_register(struct inet6_dev *idev);
 105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 106 #else
 107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 108 {
 109 }
 110
 111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 112 {
 113 }
 114 #endif
 115
 116 #ifdef CONFIG_IPV6_PRIVACY
 117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 119 static void ipv6_regen_rndid(unsigned long data);
 120
 121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 122 #endif
 123
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188 };
 189
 190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201 #ifdef CONFIG_IPV6_PRIVACY
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207 #endif
 208         .max_addresses          = IPV6_MAX_ADDRESSES,
 209         .accept_ra_defrtr       = 1,
 210         .accept_ra_pinfo        = 1,
 211 #ifdef CONFIG_IPV6_ROUTER_PREF
 212         .accept_ra_rtr_pref     = 1,
 213         .rtr_probe_interval     = 60 * HZ,
 214 #ifdef CONFIG_IPV6_ROUTE_INFO
 215         .accept_ra_rt_info_max_plen = 0,
 216 #endif
 217 #endif
 218         .proxy_ndp              = 0,
 219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 220 };
 221
 222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 225
 226 /* Check if a valid qdisc is available */
 227 static inline int addrconf_qdisc_ok(struct net_device *dev)
 228 {
 229         return (dev->qdisc != &noop_qdisc);
 230 }
 231
 232 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 233 {
 234         if (del_timer(&ifp->timer))
 235                 __in6_ifa_put(ifp);
 236 }
 237
 238 enum addrconf_timer_t
 239 {
 240         AC_NONE,
 241         AC_DAD,
 242         AC_RS,
 243 };
 244
 245 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 246                                enum addrconf_timer_t what,
 247                                unsigned long when)
 248 {
 249         if (!del_timer(&ifp->timer))
 250                 in6_ifa_hold(ifp);
 251
 252         switch (what) {
 253         case AC_DAD:
 254                 ifp->timer.function = addrconf_dad_timer;
 255                 break;
 256         case AC_RS:
 257                 ifp->timer.function = addrconf_rs_timer;
 258                 break;
 259         default:;
 260         }
 261         ifp->timer.expires = jiffies + when;
 262         add_timer(&ifp->timer);
 263 }
 264
 265 static int snmp6_alloc_dev(struct inet6_dev *idev)
 266 {
 267         if (snmp_mib_init((void **)idev->stats.ipv6,
 268                           sizeof(struct ipstats_mib)) < 0)
 269                 goto err_ip;
 270         if (snmp_mib_init((void **)idev->stats.icmpv6,
 271                           sizeof(struct icmpv6_mib)) < 0)
 272                 goto err_icmp;
 273         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 274                           sizeof(struct icmpv6msg_mib)) < 0)
 275                 goto err_icmpmsg;
 276
 277         return 0;
 278
 279 err_icmpmsg:
 280         snmp_mib_free((void **)idev->stats.icmpv6);
 281 err_icmp:
 282         snmp_mib_free((void **)idev->stats.ipv6);
 283 err_ip:
 284         return -ENOMEM;
 285 }
 286
 287 static void snmp6_free_dev(struct inet6_dev *idev)
 288 {
 289         snmp_mib_free((void **)idev->stats.icmpv6msg);
 290         snmp_mib_free((void **)idev->stats.icmpv6);
 291         snmp_mib_free((void **)idev->stats.ipv6);
 292 }
 293
 294 /* Nobody refers to this device, we may destroy it. */
 295
 296 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 297 {
 298         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 299         kfree(idev);
 300 }
 301
 302 void in6_dev_finish_destroy(struct inet6_dev *idev)
 303 {
 304         struct net_device *dev = idev->dev;
 305         BUG_TRAP(idev->addr_list==NULL);
 306         BUG_TRAP(idev->mc_list==NULL);
 307 #ifdef NET_REFCNT_DEBUG
 308         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 309 #endif
 310         dev_put(dev);
 311         if (!idev->dead) {
 312                 printk("Freeing alive inet6 device %p\n", idev);
 313                 return;
 314         }
 315         snmp6_free_dev(idev);
 316         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 317 }
 318
 319 EXPORT_SYMBOL(in6_dev_finish_destroy);
 320
 321 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 322 {
 323         struct inet6_dev *ndev;
 324         struct in6_addr maddr;
 325
 326         ASSERT_RTNL();
 327
 328         if (dev->mtu < IPV6_MIN_MTU)
 329                 return NULL;
 330
 331         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 332
 333         if (ndev == NULL)
 334                 return NULL;
 335
 336         rwlock_init(&ndev->lock);
 337         ndev->dev = dev;
 338         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 339         ndev->cnf.mtu6 = dev->mtu;
 340         ndev->cnf.sysctl = NULL;
 341         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 342         if (ndev->nd_parms == NULL) {
 343                 kfree(ndev);
 344                 return NULL;
 345         }
 346         /* We refer to the device */
 347         dev_hold(dev);
 348
 349         if (snmp6_alloc_dev(ndev) < 0) {
 350                 ADBG((KERN_WARNING
 351                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 352                         __func__, dev->name));
 353                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 354                 ndev->dead = 1;
 355                 in6_dev_finish_destroy(ndev);
 356                 return NULL;
 357         }
 358
 359         if (snmp6_register_dev(ndev) < 0) {
 360                 ADBG((KERN_WARNING
 361                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 362                         __func__, dev->name));
 363                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 364                 ndev->dead = 1;
 365                 in6_dev_finish_destroy(ndev);
 366                 return NULL;
 367         }
 368
 369         /* One reference from device.  We must do this before
 370          * we invoke __ipv6_regen_rndid().
 371          */
 372         in6_dev_hold(ndev);
 373
 374 #ifdef CONFIG_IPV6_PRIVACY
 375         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 376         if ((dev->flags&IFF_LOOPBACK) ||
 377             dev->type == ARPHRD_TUNNEL ||
 378 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 379             dev->type == ARPHRD_SIT ||
 380 #endif
 381             dev->type == ARPHRD_NONE) {
 382                 printk(KERN_INFO
 383                        "%s: Disabled Privacy Extensions\n",
 384                        dev->name);
 385                 ndev->cnf.use_tempaddr = -1;
 386
 387                 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 388                         printk(KERN_INFO
 389                                "%s: Disabled Multicast RS\n",
 390                                dev->name);
 391                         ndev->cnf.rtr_solicits = 0;
 392                 }
 393         } else {
 394                 in6_dev_hold(ndev);
 395                 ipv6_regen_rndid((unsigned long) ndev);
 396         }
 397 #endif
 398
 399         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 400                 ndev->if_flags |= IF_READY;
 401
 402         ipv6_mc_init_dev(ndev);
 403         ndev->tstamp = jiffies;
 404         addrconf_sysctl_register(ndev);
 405         /* protected by rtnl_lock */
 406         rcu_assign_pointer(dev->ip6_ptr, ndev);
 407
 408         /* Join all-node multicast group */
 409         ipv6_addr_all_nodes(&maddr);
 410         ipv6_dev_mc_inc(dev, &maddr);
 411
 412         return ndev;
 413 }
 414
 415 struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 416 {
 417         struct inet6_dev *idev;
 418
 419         ASSERT_RTNL();
 420
 421         if ((idev = __in6_dev_get(dev)) == NULL) {
 422                 if ((idev = ipv6_add_dev(dev)) == NULL)
 423                         return NULL;
 424         }
 425
 426         if (dev->flags&IFF_UP)
 427                 ipv6_mc_up(idev);
 428         return idev;
 429 }
 430
 431 #ifdef CONFIG_SYSCTL
 432 static void dev_forward_change(struct inet6_dev *idev)
 433 {
 434         struct net_device *dev;
 435         struct inet6_ifaddr *ifa;
 436         struct in6_addr addr;
 437
 438         if (!idev)
 439                 return;
 440         dev = idev->dev;
 441         if (dev && (dev->flags & IFF_MULTICAST)) {
 442                 ipv6_addr_all_routers(&addr);
 443
 444                 if (idev->cnf.forwarding)
 445                         ipv6_dev_mc_inc(dev, &addr);
 446                 else
 447                         ipv6_dev_mc_dec(dev, &addr);
 448         }
 449         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 450                 if (ifa->flags&IFA_F_TENTATIVE)
 451                         continue;
 452                 if (idev->cnf.forwarding)
 453                         addrconf_join_anycast(ifa);
 454                 else
 455                         addrconf_leave_anycast(ifa);
 456         }
 457 }
 458
 459
 460 static void addrconf_forward_change(struct net *net, __s32 newf)
 461 {
 462         struct net_device *dev;
 463         struct inet6_dev *idev;
 464
 465         read_lock(&dev_base_lock);
 466         for_each_netdev(net, dev) {
 467                 rcu_read_lock();
 468                 idev = __in6_dev_get(dev);
 469                 if (idev) {
 470                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 471                         idev->cnf.forwarding = newf;
 472                         if (changed)
 473                                 dev_forward_change(idev);
 474                 }
 475                 rcu_read_unlock();
 476         }
 477         read_unlock(&dev_base_lock);
 478 }
 479
 480 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 481 {
 482         struct net *net;
 483
 484         net = (struct net *)table->extra2;
 485         if (p == &net->ipv6.devconf_dflt->forwarding)
 486                 return;
 487
 488         if (p == &net->ipv6.devconf_all->forwarding) {
 489                 __s32 newf = net->ipv6.devconf_all->forwarding;
 490                 net->ipv6.devconf_dflt->forwarding = newf;
 491                 addrconf_forward_change(net, newf);
 492         } else if ((!*p) ^ (!old))
 493                 dev_forward_change((struct inet6_dev *)table->extra1);
 494
 495         if (*p)
 496                 rt6_purge_dflt_routers(net);
 497 }
 498 #endif
 499
 500 /* Nobody refers to this ifaddr, destroy it */
 501
 502 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 503 {
 504         BUG_TRAP(ifp->if_next==NULL);
 505         BUG_TRAP(ifp->lst_next==NULL);
 506 #ifdef NET_REFCNT_DEBUG
 507         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 508 #endif
 509
 510         in6_dev_put(ifp->idev);
 511
 512         if (del_timer(&ifp->timer))
 513                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 514
 515         if (!ifp->dead) {
 516                 printk("Freeing alive inet6 address %p\n", ifp);
 517                 return;
 518         }
 519         dst_release(&ifp->rt->u.dst);
 520
 521         kfree(ifp);
 522 }
 523
 524 static void
 525 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 526 {
 527         struct inet6_ifaddr *ifa, **ifap;
 528         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 529
 530         /*
 531          * Each device address list is sorted in order of scope -
 532          * global before linklocal.
 533          */
 534         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 535              ifap = &ifa->if_next) {
 536                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 537                         break;
 538         }
 539
 540         ifp->if_next = *ifap;
 541         *ifap = ifp;
 542 }
 543
 544 /*
 545  *      Hash function taken from net_alias.c
 546  */
 547 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 548 {
 549         __u32 word;
 550
 551         /*
 552          * We perform the hash function over the last 64 bits of the address
 553          * This will include the IEEE address token on links that support it.
 554          */
 555
 556         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 557         word ^= (word >> 16);
 558         word ^= (word >> 8);
 559
 560         return ((word ^ (word >> 4)) & 0x0f);
 561 }
 562
 563 /* On success it returns ifp with increased reference count */
 564
 565 static struct inet6_ifaddr *
 566 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 567               int scope, u32 flags)
 568 {
 569         struct inet6_ifaddr *ifa = NULL;
 570         struct rt6_info *rt;
 571         int hash;
 572         int err = 0;
 573
 574         rcu_read_lock_bh();
 575         if (idev->dead) {
 576                 err = -ENODEV;                  /*XXX*/
 577                 goto out2;
 578         }
 579
 580         write_lock(&addrconf_hash_lock);
 581
 582         /* Ignore adding duplicate addresses on an interface */
 583         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 584                 ADBG(("ipv6_add_addr: already assigned\n"));
 585                 err = -EEXIST;
 586                 goto out;
 587         }
 588
 589         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 590
 591         if (ifa == NULL) {
 592                 ADBG(("ipv6_add_addr: malloc failed\n"));
 593                 err = -ENOBUFS;
 594                 goto out;
 595         }
 596
 597         rt = addrconf_dst_alloc(idev, addr, 0);
 598         if (IS_ERR(rt)) {
 599                 err = PTR_ERR(rt);
 600                 goto out;
 601         }
 602
 603         ipv6_addr_copy(&ifa->addr, addr);
 604
 605         spin_lock_init(&ifa->lock);
 606         init_timer(&ifa->timer);
 607         ifa->timer.data = (unsigned long) ifa;
 608         ifa->scope = scope;
 609         ifa->prefix_len = pfxlen;
 610         ifa->flags = flags | IFA_F_TENTATIVE;
 611         ifa->cstamp = ifa->tstamp = jiffies;
 612
 613         ifa->rt = rt;
 614
 615         /*
 616          * part one of RFC 4429, section 3.3
 617          * We should not configure an address as
 618          * optimistic if we do not yet know the link
 619          * layer address of our nexhop router
 620          */
 621
 622         if (rt->rt6i_nexthop == NULL)
 623                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 624
 625         ifa->idev = idev;
 626         in6_dev_hold(idev);
 627         /* For caller */
 628         in6_ifa_hold(ifa);
 629
 630         /* Add to big hash table */
 631         hash = ipv6_addr_hash(addr);
 632
 633         ifa->lst_next = inet6_addr_lst[hash];
 634         inet6_addr_lst[hash] = ifa;
 635         in6_ifa_hold(ifa);
 636         write_unlock(&addrconf_hash_lock);
 637
 638         write_lock(&idev->lock);
 639         /* Add to inet6_dev unicast addr list. */
 640         ipv6_link_dev_addr(idev, ifa);
 641
 642 #ifdef CONFIG_IPV6_PRIVACY
 643         if (ifa->flags&IFA_F_TEMPORARY) {
 644                 ifa->tmp_next = idev->tempaddr_list;
 645                 idev->tempaddr_list = ifa;
 646                 in6_ifa_hold(ifa);
 647         }
 648 #endif
 649
 650         in6_ifa_hold(ifa);
 651         write_unlock(&idev->lock);
 652 out2:
 653         rcu_read_unlock_bh();
 654
 655         if (likely(err == 0))
 656                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 657         else {
 658                 kfree(ifa);
 659                 ifa = ERR_PTR(err);
 660         }
 661
 662         return ifa;
 663 out:
 664         write_unlock(&addrconf_hash_lock);
 665         goto out2;
 666 }
 667
 668 /* This function wants to get referenced ifp and releases it before return */
 669
 670 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 671 {
 672         struct inet6_ifaddr *ifa, **ifap;
 673         struct inet6_dev *idev = ifp->idev;
 674         int hash;
 675         int deleted = 0, onlink = 0;
 676         unsigned long expires = jiffies;
 677
 678         hash = ipv6_addr_hash(&ifp->addr);
 679
 680         ifp->dead = 1;
 681
 682         write_lock_bh(&addrconf_hash_lock);
 683         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 684              ifap = &ifa->lst_next) {
 685                 if (ifa == ifp) {
 686                         *ifap = ifa->lst_next;
 687                         __in6_ifa_put(ifp);
 688                         ifa->lst_next = NULL;
 689                         break;
 690                 }
 691         }
 692         write_unlock_bh(&addrconf_hash_lock);
 693
 694         write_lock_bh(&idev->lock);
 695 #ifdef CONFIG_IPV6_PRIVACY
 696         if (ifp->flags&IFA_F_TEMPORARY) {
 697                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 698                      ifap = &ifa->tmp_next) {
 699                         if (ifa == ifp) {
 700                                 *ifap = ifa->tmp_next;
 701                                 if (ifp->ifpub) {
 702                                         in6_ifa_put(ifp->ifpub);
 703                                         ifp->ifpub = NULL;
 704                                 }
 705                                 __in6_ifa_put(ifp);
 706                                 ifa->tmp_next = NULL;
 707                                 break;
 708                         }
 709                 }
 710         }
 711 #endif
 712
 713         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 714                 if (ifa == ifp) {
 715                         *ifap = ifa->if_next;
 716                         __in6_ifa_put(ifp);
 717                         ifa->if_next = NULL;
 718                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 719                                 break;
 720                         deleted = 1;
 721                         continue;
 722                 } else if (ifp->flags & IFA_F_PERMANENT) {
 723                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 724                                               ifp->prefix_len)) {
 725                                 if (ifa->flags & IFA_F_PERMANENT) {
 726                                         onlink = 1;
 727                                         if (deleted)
 728                                                 break;
 729                                 } else {
 730                                         unsigned long lifetime;
 731
 732                                         if (!onlink)
 733                                                 onlink = -1;
 734
 735                                         spin_lock(&ifa->lock);
 736                                         lifetime = min_t(unsigned long,
 737                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 738                                         if (time_before(expires,
 739                                                         ifa->tstamp + lifetime * HZ))
 740                                                 expires = ifa->tstamp + lifetime * HZ;
 741                                         spin_unlock(&ifa->lock);
 742                                 }
 743                         }
 744                 }
 745                 ifap = &ifa->if_next;
 746         }
 747         write_unlock_bh(&idev->lock);
 748
 749         ipv6_ifa_notify(RTM_DELADDR, ifp);
 750
 751         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 752
 753         addrconf_del_timer(ifp);
 754
 755         /*
 756          * Purge or update corresponding prefix
 757          *
 758          * 1) we don't purge prefix here if address was not permanent.
 759          *    prefix is managed by its own lifetime.
 760          * 2) if there're no addresses, delete prefix.
 761          * 3) if there're still other permanent address(es),
 762          *    corresponding prefix is still permanent.
 763          * 4) otherwise, update prefix lifetime to the
 764          *    longest valid lifetime among the corresponding
 765          *    addresses on the device.
 766          *    Note: subsequent RA will update lifetime.
 767          *
 768          * --yoshfuji
 769          */
 770         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 771                 struct in6_addr prefix;
 772                 struct rt6_info *rt;
 773                 struct net *net = dev_net(ifp->idev->dev);
 774                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 775                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 776
 777                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 778                         if (onlink == 0) {
 779                                 ip6_del_rt(rt);
 780                                 rt = NULL;
 781                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 782                                 rt->rt6i_expires = expires;
 783                                 rt->rt6i_flags |= RTF_EXPIRES;
 784                         }
 785                 }
 786                 dst_release(&rt->u.dst);
 787         }
 788
 789         in6_ifa_put(ifp);
 790 }
 791
 792 #ifdef CONFIG_IPV6_PRIVACY
 793 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 794 {
 795         struct inet6_dev *idev = ifp->idev;
 796         struct in6_addr addr, *tmpaddr;
 797         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 798         unsigned long regen_advance;
 799         int tmp_plen;
 800         int ret = 0;
 801         int max_addresses;
 802         u32 addr_flags;
 803
 804         write_lock(&idev->lock);
 805         if (ift) {
 806                 spin_lock_bh(&ift->lock);
 807                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 808                 spin_unlock_bh(&ift->lock);
 809                 tmpaddr = &addr;
 810         } else {
 811                 tmpaddr = NULL;
 812         }
 813 retry:
 814         in6_dev_hold(idev);
 815         if (idev->cnf.use_tempaddr <= 0) {
 816                 write_unlock(&idev->lock);
 817                 printk(KERN_INFO
 818                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 819                 in6_dev_put(idev);
 820                 ret = -1;
 821                 goto out;
 822         }
 823         spin_lock_bh(&ifp->lock);
 824         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 825                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 826                 spin_unlock_bh(&ifp->lock);
 827                 write_unlock(&idev->lock);
 828                 printk(KERN_WARNING
 829                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 830                 in6_dev_put(idev);
 831                 ret = -1;
 832                 goto out;
 833         }
 834         in6_ifa_hold(ifp);
 835         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 836         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 837                 spin_unlock_bh(&ifp->lock);
 838                 write_unlock(&idev->lock);
 839                 printk(KERN_WARNING
 840                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 841                 in6_ifa_put(ifp);
 842                 in6_dev_put(idev);
 843                 ret = -1;
 844                 goto out;
 845         }
 846         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 847         tmp_valid_lft = min_t(__u32,
 848                               ifp->valid_lft,
 849                               idev->cnf.temp_valid_lft);
 850         tmp_prefered_lft = min_t(__u32,
 851                                  ifp->prefered_lft,
 852                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 853         tmp_plen = ifp->prefix_len;
 854         max_addresses = idev->cnf.max_addresses;
 855         tmp_cstamp = ifp->cstamp;
 856         tmp_tstamp = ifp->tstamp;
 857         spin_unlock_bh(&ifp->lock);
 858
 859         regen_advance = idev->cnf.regen_max_retry *
 860                         idev->cnf.dad_transmits *
 861                         idev->nd_parms->retrans_time / HZ;
 862         write_unlock(&idev->lock);
 863
 864         /* A temporary address is created only if this calculated Preferred
 865          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 866          * an implementation must not create a temporary address with a zero
 867          * Preferred Lifetime.
 868          */
 869         if (tmp_prefered_lft <= regen_advance) {
 870                 in6_ifa_put(ifp);
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875
 876         addr_flags = IFA_F_TEMPORARY;
 877         /* set in addrconf_prefix_rcv() */
 878         if (ifp->flags & IFA_F_OPTIMISTIC)
 879                 addr_flags |= IFA_F_OPTIMISTIC;
 880
 881         ift = !max_addresses ||
 882               ipv6_count_addresses(idev) < max_addresses ?
 883                 ipv6_add_addr(idev, &addr, tmp_plen,
 884                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 885                               addr_flags) : NULL;
 886         if (!ift || IS_ERR(ift)) {
 887                 in6_ifa_put(ifp);
 888                 in6_dev_put(idev);
 889                 printk(KERN_INFO
 890                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 891                 tmpaddr = &addr;
 892                 write_lock(&idev->lock);
 893                 goto retry;
 894         }
 895
 896         spin_lock_bh(&ift->lock);
 897         ift->ifpub = ifp;
 898         ift->valid_lft = tmp_valid_lft;
 899         ift->prefered_lft = tmp_prefered_lft;
 900         ift->cstamp = tmp_cstamp;
 901         ift->tstamp = tmp_tstamp;
 902         spin_unlock_bh(&ift->lock);
 903
 904         addrconf_dad_start(ift, 0);
 905         in6_ifa_put(ift);
 906         in6_dev_put(idev);
 907 out:
 908         return ret;
 909 }
 910 #endif
 911
 912 /*
 913  *      Choose an appropriate source address (RFC3484)
 914  */
 915 enum {
 916         IPV6_SADDR_RULE_INIT = 0,
 917         IPV6_SADDR_RULE_LOCAL,
 918         IPV6_SADDR_RULE_SCOPE,
 919         IPV6_SADDR_RULE_PREFERRED,
 920 #ifdef CONFIG_IPV6_MIP6
 921         IPV6_SADDR_RULE_HOA,
 922 #endif
 923         IPV6_SADDR_RULE_OIF,
 924         IPV6_SADDR_RULE_LABEL,
 925 #ifdef CONFIG_IPV6_PRIVACY
 926         IPV6_SADDR_RULE_PRIVACY,
 927 #endif
 928         IPV6_SADDR_RULE_ORCHID,
 929         IPV6_SADDR_RULE_PREFIX,
 930         IPV6_SADDR_RULE_MAX
 931 };
 932
 933 struct ipv6_saddr_score {
 934         int                     rule;
 935         int                     addr_type;
 936         struct inet6_ifaddr     *ifa;
 937         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 938         int                     scopedist;
 939         int                     matchlen;
 940 };
 941
 942 struct ipv6_saddr_dst {
 943         struct in6_addr *addr;
 944         int ifindex;
 945         int scope;
 946         int label;
 947         unsigned int prefs;
 948 };
 949
 950 static inline int ipv6_saddr_preferred(int type)
 951 {
 952         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 953                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 954                 return 1;
 955         return 0;
 956 }
 957
 958 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
 959                                struct ipv6_saddr_dst *dst,
 960                                int i)
 961 {
 962         int ret;
 963
 964         if (i <= score->rule) {
 965                 switch (i) {
 966                 case IPV6_SADDR_RULE_SCOPE:
 967                         ret = score->scopedist;
 968                         break;
 969                 case IPV6_SADDR_RULE_PREFIX:
 970                         ret = score->matchlen;
 971                         break;
 972                 default:
 973                         ret = !!test_bit(i, score->scorebits);
 974                 }
 975                 goto out;
 976         }
 977
 978         switch (i) {
 979         case IPV6_SADDR_RULE_INIT:
 980                 /* Rule 0: remember if hiscore is not ready yet */
 981                 ret = !!score->ifa;
 982                 break;
 983         case IPV6_SADDR_RULE_LOCAL:
 984                 /* Rule 1: Prefer same address */
 985                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
 986                 break;
 987         case IPV6_SADDR_RULE_SCOPE:
 988                 /* Rule 2: Prefer appropriate scope
 989                  *
 990                  *      ret
 991                  *       ^
 992                  *    -1 |  d 15
 993                  *    ---+--+-+---> scope
 994                  *       |
 995                  *       |             d is scope of the destination.
 996                  *  B-d  |  \
 997                  *       |   \      <- smaller scope is better if
 998                  *  B-15 |    \        if scope is enough for destinaion.
 999                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1000                  * d-C-1 | /
1001                  *       |/         <- greater is better
1002                  *   -C  /             if scope is not enough for destination.
1003                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1004                  *
1005                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1006                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1007                  * Assume B = 0 and we get C > 29.
1008                  */
1009                 ret = __ipv6_addr_src_scope(score->addr_type);
1010                 if (ret >= dst->scope)
1011                         ret = -ret;
1012                 else
1013                         ret -= 128;     /* 30 is enough */
1014                 score->scopedist = ret;
1015                 break;
1016         case IPV6_SADDR_RULE_PREFERRED:
1017                 /* Rule 3: Avoid deprecated and optimistic addresses */
1018                 ret = ipv6_saddr_preferred(score->addr_type) ||
1019                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1020                 break;
1021 #ifdef CONFIG_IPV6_MIP6
1022         case IPV6_SADDR_RULE_HOA:
1023             {
1024                 /* Rule 4: Prefer home address */
1025                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1026                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1027                 break;
1028             }
1029 #endif
1030         case IPV6_SADDR_RULE_OIF:
1031                 /* Rule 5: Prefer outgoing interface */
1032                 ret = (!dst->ifindex ||
1033                        dst->ifindex == score->ifa->idev->dev->ifindex);
1034                 break;
1035         case IPV6_SADDR_RULE_LABEL:
1036                 /* Rule 6: Prefer matching label */
1037                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1038                                       score->ifa->idev->dev->ifindex) == dst->label;
1039                 break;
1040 #ifdef CONFIG_IPV6_PRIVACY
1041         case IPV6_SADDR_RULE_PRIVACY:
1042             {
1043                 /* Rule 7: Prefer public address
1044                  * Note: prefer temprary address if use_tempaddr >= 2
1045                  */
1046                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1047                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1048                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1049                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1050                 break;
1051             }
1052 #endif
1053         case IPV6_SADDR_RULE_ORCHID:
1054                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1055                  *          non-ORCHID vs non-ORCHID
1056                  */
1057                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1058                         ipv6_addr_orchid(dst->addr));
1059                 break;
1060         case IPV6_SADDR_RULE_PREFIX:
1061                 /* Rule 8: Use longest matching prefix */
1062                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1063                                                        dst->addr);
1064                 break;
1065         default:
1066                 ret = 0;
1067         }
1068
1069         if (ret)
1070                 __set_bit(i, score->scorebits);
1071         score->rule = i;
1072 out:
1073         return ret;
1074 }
1075
1076 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1077                        struct in6_addr *daddr, unsigned int prefs,
1078                        struct in6_addr *saddr)
1079 {
1080         struct ipv6_saddr_score scores[2],
1081                                 *score = &scores[0], *hiscore = &scores[1];
1082         struct net *net = dev_net(dst_dev);
1083         struct ipv6_saddr_dst dst;
1084         struct net_device *dev;
1085         int dst_type;
1086
1087         dst_type = __ipv6_addr_type(daddr);
1088         dst.addr = daddr;
1089         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1090         dst.scope = __ipv6_addr_src_scope(dst_type);
1091         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1092         dst.prefs = prefs;
1093
1094         hiscore->rule = -1;
1095         hiscore->ifa = NULL;
1096
1097         read_lock(&dev_base_lock);
1098         rcu_read_lock();
1099
1100         for_each_netdev(net, dev) {
1101                 struct inet6_dev *idev;
1102
1103                 /* Candidate Source Address (section 4)
1104                  *  - multicast and link-local destination address,
1105                  *    the set of candidate source address MUST only
1106                  *    include addresses assigned to interfaces
1107                  *    belonging to the same link as the outgoing
1108                  *    interface.
1109                  * (- For site-local destination addresses, the
1110                  *    set of candidate source addresses MUST only
1111                  *    include addresses assigned to interfaces
1112                  *    belonging to the same site as the outgoing
1113                  *    interface.)
1114                  */
1115                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1116                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1117                     dst.ifindex && dev->ifindex != dst.ifindex)
1118                         continue;
1119
1120                 idev = __in6_dev_get(dev);
1121                 if (!idev)
1122                         continue;
1123
1124                 read_lock_bh(&idev->lock);
1125                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1126                         int i;
1127
1128                         /*
1129                          * - Tentative Address (RFC2462 section 5.4)
1130                          *  - A tentative address is not considered
1131                          *    "assigned to an interface" in the traditional
1132                          *    sense, unless it is also flagged as optimistic.
1133                          * - Candidate Source Address (section 4)
1134                          *  - In any case, anycast addresses, multicast
1135                          *    addresses, and the unspecified address MUST
1136                          *    NOT be included in a candidate set.
1137                          */
1138                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1139                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1140                                 continue;
1141
1142                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1143
1144                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1145                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1146                                 LIMIT_NETDEBUG(KERN_DEBUG
1147                                                "ADDRCONF: unspecified / multicast address "
1148                                                "assigned as unicast address on %s",
1149                                                dev->name);
1150                                 continue;
1151                         }
1152
1153                         score->rule = -1;
1154                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1155
1156                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1157                                 int minihiscore, miniscore;
1158
1159                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1160                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1161
1162                                 if (minihiscore > miniscore) {
1163                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1164                                             score->scopedist > 0) {
1165                                                 /*
1166                                                  * special case:
1167                                                  * each remaining entry
1168                                                  * has too small (not enough)
1169                                                  * scope, because ifa entries
1170                                                  * are sorted by their scope
1171                                                  * values.
1172                                                  */
1173                                                 goto try_nextdev;
1174                                         }
1175                                         break;
1176                                 } else if (minihiscore < miniscore) {
1177                                         struct ipv6_saddr_score *tmp;
1178
1179                                         if (hiscore->ifa)
1180                                                 in6_ifa_put(hiscore->ifa);
1181
1182                                         in6_ifa_hold(score->ifa);
1183
1184                                         tmp = hiscore;
1185                                         hiscore = score;
1186                                         score = tmp;
1187
1188                                         /* restore our iterator */
1189                                         score->ifa = hiscore->ifa;
1190
1191                                         break;
1192                                 }
1193                         }
1194                 }
1195 try_nextdev:
1196                 read_unlock_bh(&idev->lock);
1197         }
1198         rcu_read_unlock();
1199         read_unlock(&dev_base_lock);
1200
1201         if (!hiscore->ifa)
1202                 return -EADDRNOTAVAIL;
1203
1204         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1205         in6_ifa_put(hiscore->ifa);
1206         return 0;
1207 }
1208
1209 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1210
1211 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1212                     unsigned char banned_flags)
1213 {
1214         struct inet6_dev *idev;
1215         int err = -EADDRNOTAVAIL;
1216
1217         rcu_read_lock();
1218         if ((idev = __in6_dev_get(dev)) != NULL) {
1219                 struct inet6_ifaddr *ifp;
1220
1221                 read_lock_bh(&idev->lock);
1222                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1223                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1224                                 ipv6_addr_copy(addr, &ifp->addr);
1225                                 err = 0;
1226                                 break;
1227                         }
1228                 }
1229                 read_unlock_bh(&idev->lock);
1230         }
1231         rcu_read_unlock();
1232         return err;
1233 }
1234
1235 static int ipv6_count_addresses(struct inet6_dev *idev)
1236 {
1237         int cnt = 0;
1238         struct inet6_ifaddr *ifp;
1239
1240         read_lock_bh(&idev->lock);
1241         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1242                 cnt++;
1243         read_unlock_bh(&idev->lock);
1244         return cnt;
1245 }
1246
1247 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1248                   struct net_device *dev, int strict)
1249 {
1250         struct inet6_ifaddr * ifp;
1251         u8 hash = ipv6_addr_hash(addr);
1252
1253         read_lock_bh(&addrconf_hash_lock);
1254         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1255                 if (!net_eq(dev_net(ifp->idev->dev), net))
1256                         continue;
1257                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1258                     !(ifp->flags&IFA_F_TENTATIVE)) {
1259                         if (dev == NULL || ifp->idev->dev == dev ||
1260                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1261                                 break;
1262                 }
1263         }
1264         read_unlock_bh(&addrconf_hash_lock);
1265         return ifp != NULL;
1266 }
1267 EXPORT_SYMBOL(ipv6_chk_addr);
1268
1269 static
1270 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1271                        struct net_device *dev)
1272 {
1273         struct inet6_ifaddr * ifp;
1274         u8 hash = ipv6_addr_hash(addr);
1275
1276         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1277                 if (!net_eq(dev_net(ifp->idev->dev), net))
1278                         continue;
1279                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1280                         if (dev == NULL || ifp->idev->dev == dev)
1281                                 break;
1282                 }
1283         }
1284         return ifp != NULL;
1285 }
1286
1287 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1288 {
1289         struct inet6_dev *idev;
1290         struct inet6_ifaddr *ifa;
1291         int     onlink;
1292
1293         onlink = 0;
1294         rcu_read_lock();
1295         idev = __in6_dev_get(dev);
1296         if (idev) {
1297                 read_lock_bh(&idev->lock);
1298                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1299                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1300                                                    ifa->prefix_len);
1301                         if (onlink)
1302                                 break;
1303                 }
1304                 read_unlock_bh(&idev->lock);
1305         }
1306         rcu_read_unlock();
1307         return onlink;
1308 }
1309
1310 EXPORT_SYMBOL(ipv6_chk_prefix);
1311
1312 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, struct in6_addr *addr,
1313                                      struct net_device *dev, int strict)
1314 {
1315         struct inet6_ifaddr * ifp;
1316         u8 hash = ipv6_addr_hash(addr);
1317
1318         read_lock_bh(&addrconf_hash_lock);
1319         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1320                 if (!net_eq(dev_net(ifp->idev->dev), net))
1321                         continue;
1322                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1323                         if (dev == NULL || ifp->idev->dev == dev ||
1324                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1325                                 in6_ifa_hold(ifp);
1326                                 break;
1327                         }
1328                 }
1329         }
1330         read_unlock_bh(&addrconf_hash_lock);
1331
1332         return ifp;
1333 }
1334
1335 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1336 {
1337         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1338         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1339         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1340         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1341         int sk_ipv6only = ipv6_only_sock(sk);
1342         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1343         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1344         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1345
1346         if (!sk2_rcv_saddr && !sk_ipv6only)
1347                 return 1;
1348
1349         if (addr_type2 == IPV6_ADDR_ANY &&
1350             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1351                 return 1;
1352
1353         if (addr_type == IPV6_ADDR_ANY &&
1354             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1355                 return 1;
1356
1357         if (sk2_rcv_saddr6 &&
1358             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1359                 return 1;
1360
1361         if (addr_type == IPV6_ADDR_MAPPED &&
1362             !sk2_ipv6only &&
1363             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1364                 return 1;
1365
1366         return 0;
1367 }
1368
1369 /* Gets referenced address, destroys ifaddr */
1370
1371 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1372 {
1373         if (ifp->flags&IFA_F_PERMANENT) {
1374                 spin_lock_bh(&ifp->lock);
1375                 addrconf_del_timer(ifp);
1376                 ifp->flags |= IFA_F_TENTATIVE;
1377                 spin_unlock_bh(&ifp->lock);
1378                 in6_ifa_put(ifp);
1379 #ifdef CONFIG_IPV6_PRIVACY
1380         } else if (ifp->flags&IFA_F_TEMPORARY) {
1381                 struct inet6_ifaddr *ifpub;
1382                 spin_lock_bh(&ifp->lock);
1383                 ifpub = ifp->ifpub;
1384                 if (ifpub) {
1385                         in6_ifa_hold(ifpub);
1386                         spin_unlock_bh(&ifp->lock);
1387                         ipv6_create_tempaddr(ifpub, ifp);
1388                         in6_ifa_put(ifpub);
1389                 } else {
1390                         spin_unlock_bh(&ifp->lock);
1391                 }
1392                 ipv6_del_addr(ifp);
1393 #endif
1394         } else
1395                 ipv6_del_addr(ifp);
1396 }
1397
1398 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1399 {
1400         if (net_ratelimit())
1401                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1402         addrconf_dad_stop(ifp);
1403 }
1404
1405 /* Join to solicited addr multicast group. */
1406
1407 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1408 {
1409         struct in6_addr maddr;
1410
1411         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1412                 return;
1413
1414         addrconf_addr_solict_mult(addr, &maddr);
1415         ipv6_dev_mc_inc(dev, &maddr);
1416 }
1417
1418 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1419 {
1420         struct in6_addr maddr;
1421
1422         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1423                 return;
1424
1425         addrconf_addr_solict_mult(addr, &maddr);
1426         __ipv6_dev_mc_dec(idev, &maddr);
1427 }
1428
1429 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1430 {
1431         struct in6_addr addr;
1432         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1433         if (ipv6_addr_any(&addr))
1434                 return;
1435         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1436 }
1437
1438 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1439 {
1440         struct in6_addr addr;
1441         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1442         if (ipv6_addr_any(&addr))
1443                 return;
1444         __ipv6_dev_ac_dec(ifp->idev, &addr);
1445 }
1446
1447 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1448 {
1449         if (dev->addr_len != ETH_ALEN)
1450                 return -1;
1451         memcpy(eui, dev->dev_addr, 3);
1452         memcpy(eui + 5, dev->dev_addr + 3, 3);
1453
1454         /*
1455          * The zSeries OSA network cards can be shared among various
1456          * OS instances, but the OSA cards have only one MAC address.
1457          * This leads to duplicate address conflicts in conjunction
1458          * with IPv6 if more than one instance uses the same card.
1459          *
1460          * The driver for these cards can deliver a unique 16-bit
1461          * identifier for each instance sharing the same card.  It is
1462          * placed instead of 0xFFFE in the interface identifier.  The
1463          * "u" bit of the interface identifier is not inverted in this
1464          * case.  Hence the resulting interface identifier has local
1465          * scope according to RFC2373.
1466          */
1467         if (dev->dev_id) {
1468                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1469                 eui[4] = dev->dev_id & 0xFF;
1470         } else {
1471                 eui[3] = 0xFF;
1472                 eui[4] = 0xFE;
1473                 eui[0] ^= 2;
1474         }
1475         return 0;
1476 }
1477
1478 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1479 {
1480         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1481         if (dev->addr_len != ARCNET_ALEN)
1482                 return -1;
1483         memset(eui, 0, 7);
1484         eui[7] = *(u8*)dev->dev_addr;
1485         return 0;
1486 }
1487
1488 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1489 {
1490         if (dev->addr_len != INFINIBAND_ALEN)
1491                 return -1;
1492         memcpy(eui, dev->dev_addr + 12, 8);
1493         eui[0] |= 2;
1494         return 0;
1495 }
1496
1497 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1498 {
1499         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1500                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1501                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1502                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1503                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1504                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1505         eui[1] = 0;
1506         eui[2] = 0x5E;
1507         eui[3] = 0xFE;
1508         memcpy(eui + 4, &addr, 4);
1509         return 0;
1510 }
1511 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1512
1513 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1514 {
1515         if (dev->priv_flags & IFF_ISATAP)
1516                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1517         return -1;
1518 }
1519
1520 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1521 {
1522         switch (dev->type) {
1523         case ARPHRD_ETHER:
1524         case ARPHRD_FDDI:
1525         case ARPHRD_IEEE802_TR:
1526                 return addrconf_ifid_eui48(eui, dev);
1527         case ARPHRD_ARCNET:
1528                 return addrconf_ifid_arcnet(eui, dev);
1529         case ARPHRD_INFINIBAND:
1530                 return addrconf_ifid_infiniband(eui, dev);
1531         case ARPHRD_SIT:
1532                 return addrconf_ifid_sit(eui, dev);
1533         }
1534         return -1;
1535 }
1536
1537 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1538 {
1539         int err = -1;
1540         struct inet6_ifaddr *ifp;
1541
1542         read_lock_bh(&idev->lock);
1543         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1544                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1545                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1546                         err = 0;
1547                         break;
1548                 }
1549         }
1550         read_unlock_bh(&idev->lock);
1551         return err;
1552 }
1553
1554 #ifdef CONFIG_IPV6_PRIVACY
1555 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1556 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1557 {
1558 regen:
1559         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1560         idev->rndid[0] &= ~0x02;
1561
1562         /*
1563          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1564          * check if generated address is not inappropriate
1565          *
1566          *  - Reserved subnet anycast (RFC 2526)
1567          *      11111101 11....11 1xxxxxxx
1568          *  - ISATAP (RFC4214) 6.1
1569          *      00-00-5E-FE-xx-xx-xx-xx
1570          *  - value 0
1571          *  - XXX: already assigned to an address on the device
1572          */
1573         if (idev->rndid[0] == 0xfd &&
1574             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1575             (idev->rndid[7]&0x80))
1576                 goto regen;
1577         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1578                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1579                         goto regen;
1580                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1581                         goto regen;
1582         }
1583
1584         return 0;
1585 }
1586
1587 static void ipv6_regen_rndid(unsigned long data)
1588 {
1589         struct inet6_dev *idev = (struct inet6_dev *) data;
1590         unsigned long expires;
1591
1592         rcu_read_lock_bh();
1593         write_lock_bh(&idev->lock);
1594
1595         if (idev->dead)
1596                 goto out;
1597
1598         if (__ipv6_regen_rndid(idev) < 0)
1599                 goto out;
1600
1601         expires = jiffies +
1602                 idev->cnf.temp_prefered_lft * HZ -
1603                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1604         if (time_before(expires, jiffies)) {
1605                 printk(KERN_WARNING
1606                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1607                         idev->dev->name);
1608                 goto out;
1609         }
1610
1611         if (!mod_timer(&idev->regen_timer, expires))
1612                 in6_dev_hold(idev);
1613
1614 out:
1615         write_unlock_bh(&idev->lock);
1616         rcu_read_unlock_bh();
1617         in6_dev_put(idev);
1618 }
1619
1620 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1621         int ret = 0;
1622
1623         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1624                 ret = __ipv6_regen_rndid(idev);
1625         return ret;
1626 }
1627 #endif
1628
1629 /*
1630  *      Add prefix route.
1631  */
1632
1633 static void
1634 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1635                       unsigned long expires, u32 flags)
1636 {
1637         struct fib6_config cfg = {
1638                 .fc_table = RT6_TABLE_PREFIX,
1639                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1640                 .fc_ifindex = dev->ifindex,
1641                 .fc_expires = expires,
1642                 .fc_dst_len = plen,
1643                 .fc_flags = RTF_UP | flags,
1644                 .fc_nlinfo.nl_net = dev_net(dev),
1645         };
1646
1647         ipv6_addr_copy(&cfg.fc_dst, pfx);
1648
1649         /* Prevent useless cloning on PtP SIT.
1650            This thing is done here expecting that the whole
1651            class of non-broadcast devices need not cloning.
1652          */
1653 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1654         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1655                 cfg.fc_flags |= RTF_NONEXTHOP;
1656 #endif
1657
1658         ip6_route_add(&cfg);
1659 }
1660
1661 /* Create "default" multicast route to the interface */
1662
1663 static void addrconf_add_mroute(struct net_device *dev)
1664 {
1665         struct fib6_config cfg = {
1666                 .fc_table = RT6_TABLE_LOCAL,
1667                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1668                 .fc_ifindex = dev->ifindex,
1669                 .fc_dst_len = 8,
1670                 .fc_flags = RTF_UP,
1671                 .fc_nlinfo.nl_net = dev_net(dev),
1672         };
1673
1674         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1675
1676         ip6_route_add(&cfg);
1677 }
1678
1679 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1680 static void sit_route_add(struct net_device *dev)
1681 {
1682         struct fib6_config cfg = {
1683                 .fc_table = RT6_TABLE_MAIN,
1684                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1685                 .fc_ifindex = dev->ifindex,
1686                 .fc_dst_len = 96,
1687                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1688                 .fc_nlinfo.nl_net = dev_net(dev),
1689         };
1690
1691         /* prefix length - 96 bits "::d.d.d.d" */
1692         ip6_route_add(&cfg);
1693 }
1694 #endif
1695
1696 static void addrconf_add_lroute(struct net_device *dev)
1697 {
1698         struct in6_addr addr;
1699
1700         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1701         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1702 }
1703
1704 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1705 {
1706         struct inet6_dev *idev;
1707
1708         ASSERT_RTNL();
1709
1710         if ((idev = ipv6_find_idev(dev)) == NULL)
1711                 return NULL;
1712
1713         /* Add default multicast route */
1714         addrconf_add_mroute(dev);
1715
1716         /* Add link local route */
1717         addrconf_add_lroute(dev);
1718         return idev;
1719 }
1720
1721 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1722 {
1723         struct prefix_info *pinfo;
1724         __u32 valid_lft;
1725         __u32 prefered_lft;
1726         int addr_type;
1727         unsigned long rt_expires;
1728         struct inet6_dev *in6_dev;
1729
1730         pinfo = (struct prefix_info *) opt;
1731
1732         if (len < sizeof(struct prefix_info)) {
1733                 ADBG(("addrconf: prefix option too short\n"));
1734                 return;
1735         }
1736
1737         /*
1738          *      Validation checks ([ADDRCONF], page 19)
1739          */
1740
1741         addr_type = ipv6_addr_type(&pinfo->prefix);
1742
1743         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1744                 return;
1745
1746         valid_lft = ntohl(pinfo->valid);
1747         prefered_lft = ntohl(pinfo->prefered);
1748
1749         if (prefered_lft > valid_lft) {
1750                 if (net_ratelimit())
1751                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1752                 return;
1753         }
1754
1755         in6_dev = in6_dev_get(dev);
1756
1757         if (in6_dev == NULL) {
1758                 if (net_ratelimit())
1759                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1760                 return;
1761         }
1762
1763         /*
1764          *      Two things going on here:
1765          *      1) Add routes for on-link prefixes
1766          *      2) Configure prefixes with the auto flag set
1767          */
1768
1769         /* Avoid arithmetic overflow. Really, we could
1770            save rt_expires in seconds, likely valid_lft,
1771            but it would require division in fib gc, that it
1772            not good.
1773          */
1774         if (valid_lft >= 0x7FFFFFFF/HZ)
1775                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1776         else
1777                 rt_expires = valid_lft * HZ;
1778
1779         /*
1780          * We convert this (in jiffies) to clock_t later.
1781          * Avoid arithmetic overflow there as well.
1782          * Overflow can happen only if HZ < USER_HZ.
1783          */
1784         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1785                 rt_expires = 0x7FFFFFFF / USER_HZ;
1786
1787         if (pinfo->onlink) {
1788                 struct rt6_info *rt;
1789                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1790                                 dev->ifindex, 1);
1791
1792                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1793                         if (rt->rt6i_flags&RTF_EXPIRES) {
1794                                 if (valid_lft == 0) {
1795                                         ip6_del_rt(rt);
1796                                         rt = NULL;
1797                                 } else {
1798                                         rt->rt6i_expires = jiffies + rt_expires;
1799                                 }
1800                         }
1801                 } else if (valid_lft) {
1802                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1803                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1804                 }
1805                 if (rt)
1806                         dst_release(&rt->u.dst);
1807         }
1808
1809         /* Try to figure out our local address for this prefix */
1810
1811         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1812                 struct inet6_ifaddr * ifp;
1813                 struct in6_addr addr;
1814                 int create = 0, update_lft = 0;
1815
1816                 if (pinfo->prefix_len == 64) {
1817                         memcpy(&addr, &pinfo->prefix, 8);
1818                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1819                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1820                                 in6_dev_put(in6_dev);
1821                                 return;
1822                         }
1823                         goto ok;
1824                 }
1825                 if (net_ratelimit())
1826                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1827                                pinfo->prefix_len);
1828                 in6_dev_put(in6_dev);
1829                 return;
1830
1831 ok:
1832
1833                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1834
1835                 if (ifp == NULL && valid_lft) {
1836                         int max_addresses = in6_dev->cnf.max_addresses;
1837                         u32 addr_flags = 0;
1838
1839 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1840                         if (in6_dev->cnf.optimistic_dad &&
1841                             !ipv6_devconf.forwarding)
1842                                 addr_flags = IFA_F_OPTIMISTIC;
1843 #endif
1844
1845                         /* Do not allow to create too much of autoconfigured
1846                          * addresses; this would be too easy way to crash kernel.
1847                          */
1848                         if (!max_addresses ||
1849                             ipv6_count_addresses(in6_dev) < max_addresses)
1850                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1851                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1852                                                     addr_flags);
1853
1854                         if (!ifp || IS_ERR(ifp)) {
1855                                 in6_dev_put(in6_dev);
1856                                 return;
1857                         }
1858
1859                         update_lft = create = 1;
1860                         ifp->cstamp = jiffies;
1861                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1862                 }
1863
1864                 if (ifp) {
1865                         int flags;
1866                         unsigned long now;
1867 #ifdef CONFIG_IPV6_PRIVACY
1868                         struct inet6_ifaddr *ift;
1869 #endif
1870                         u32 stored_lft;
1871
1872                         /* update lifetime (RFC2462 5.5.3 e) */
1873                         spin_lock(&ifp->lock);
1874                         now = jiffies;
1875                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1876                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1877                         else
1878                                 stored_lft = 0;
1879                         if (!update_lft && stored_lft) {
1880                                 if (valid_lft > MIN_VALID_LIFETIME ||
1881                                     valid_lft > stored_lft)
1882                                         update_lft = 1;
1883                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1884                                         /* valid_lft <= stored_lft is always true */
1885                                         /* XXX: IPsec */
1886                                         update_lft = 0;
1887                                 } else {
1888                                         valid_lft = MIN_VALID_LIFETIME;
1889                                         if (valid_lft < prefered_lft)
1890                                                 prefered_lft = valid_lft;
1891                                         update_lft = 1;
1892                                 }
1893                         }
1894
1895                         if (update_lft) {
1896                                 ifp->valid_lft = valid_lft;
1897                                 ifp->prefered_lft = prefered_lft;
1898                                 ifp->tstamp = now;
1899                                 flags = ifp->flags;
1900                                 ifp->flags &= ~IFA_F_DEPRECATED;
1901                                 spin_unlock(&ifp->lock);
1902
1903                                 if (!(flags&IFA_F_TENTATIVE))
1904                                         ipv6_ifa_notify(0, ifp);
1905                         } else
1906                                 spin_unlock(&ifp->lock);
1907
1908 #ifdef CONFIG_IPV6_PRIVACY
1909                         read_lock_bh(&in6_dev->lock);
1910                         /* update all temporary addresses in the list */
1911                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1912                                 /*
1913                                  * When adjusting the lifetimes of an existing
1914                                  * temporary address, only lower the lifetimes.
1915                                  * Implementations must not increase the
1916                                  * lifetimes of an existing temporary address
1917                                  * when processing a Prefix Information Option.
1918                                  */
1919                                 if (ifp != ift->ifpub)
1920                                         continue;
1921
1922                                 spin_lock(&ift->lock);
1923                                 flags = ift->flags;
1924                                 if (ift->valid_lft > valid_lft &&
1925                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1926                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1927                                 if (ift->prefered_lft > prefered_lft &&
1928                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1929                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1930                                 spin_unlock(&ift->lock);
1931                                 if (!(flags&IFA_F_TENTATIVE))
1932                                         ipv6_ifa_notify(0, ift);
1933                         }
1934
1935                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1936                                 /*
1937                                  * When a new public address is created as described in [ADDRCONF],
1938                                  * also create a new temporary address.
1939                                  */
1940                                 read_unlock_bh(&in6_dev->lock);
1941                                 ipv6_create_tempaddr(ifp, NULL);
1942                         } else {
1943                                 read_unlock_bh(&in6_dev->lock);
1944                         }
1945 #endif
1946                         in6_ifa_put(ifp);
1947                         addrconf_verify(0);
1948                 }
1949         }
1950         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1951         in6_dev_put(in6_dev);
1952 }
1953
1954 /*
1955  *      Set destination address.
1956  *      Special case for SIT interfaces where we create a new "virtual"
1957  *      device.
1958  */
1959 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1960 {
1961         struct in6_ifreq ireq;
1962         struct net_device *dev;
1963         int err = -EINVAL;
1964
1965         rtnl_lock();
1966
1967         err = -EFAULT;
1968         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1969                 goto err_exit;
1970
1971         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1972
1973         err = -ENODEV;
1974         if (dev == NULL)
1975                 goto err_exit;
1976
1977 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1978         if (dev->type == ARPHRD_SIT) {
1979                 struct ifreq ifr;
1980                 mm_segment_t    oldfs;
1981                 struct ip_tunnel_parm p;
1982
1983                 err = -EADDRNOTAVAIL;
1984                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1985                         goto err_exit;
1986
1987                 memset(&p, 0, sizeof(p));
1988                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1989                 p.iph.saddr = 0;
1990                 p.iph.version = 4;
1991                 p.iph.ihl = 5;
1992                 p.iph.protocol = IPPROTO_IPV6;
1993                 p.iph.ttl = 64;
1994                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
1995
1996                 oldfs = get_fs(); set_fs(KERNEL_DS);
1997                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1998                 set_fs(oldfs);
1999
2000                 if (err == 0) {
2001                         err = -ENOBUFS;
2002                         dev = __dev_get_by_name(net, p.name);
2003                         if (!dev)
2004                                 goto err_exit;
2005                         err = dev_open(dev);
2006                 }
2007         }
2008 #endif
2009
2010 err_exit:
2011         rtnl_unlock();
2012         return err;
2013 }
2014
2015 /*
2016  *      Manual configuration of address on an interface
2017  */
2018 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2019                           int plen, __u8 ifa_flags, __u32 prefered_lft,
2020                           __u32 valid_lft)
2021 {
2022         struct inet6_ifaddr *ifp;
2023         struct inet6_dev *idev;
2024         struct net_device *dev;
2025         int scope;
2026         u32 flags = RTF_EXPIRES;
2027
2028         ASSERT_RTNL();
2029
2030         /* check the lifetime */
2031         if (!valid_lft || prefered_lft > valid_lft)
2032                 return -EINVAL;
2033
2034         dev = __dev_get_by_index(net, ifindex);
2035         if (!dev)
2036                 return -ENODEV;
2037
2038         if ((idev = addrconf_add_dev(dev)) == NULL)
2039                 return -ENOBUFS;
2040
2041         scope = ipv6_addr_scope(pfx);
2042
2043         if (valid_lft == INFINITY_LIFE_TIME) {
2044                 ifa_flags |= IFA_F_PERMANENT;
2045                 flags = 0;
2046         } else if (valid_lft >= 0x7FFFFFFF/HZ)
2047                 valid_lft = 0x7FFFFFFF/HZ;
2048
2049         if (prefered_lft == 0)
2050                 ifa_flags |= IFA_F_DEPRECATED;
2051         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2052                  (prefered_lft != INFINITY_LIFE_TIME))
2053                 prefered_lft = 0x7FFFFFFF/HZ;
2054
2055         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2056
2057         if (!IS_ERR(ifp)) {
2058                 spin_lock_bh(&ifp->lock);
2059                 ifp->valid_lft = valid_lft;
2060                 ifp->prefered_lft = prefered_lft;
2061                 ifp->tstamp = jiffies;
2062                 spin_unlock_bh(&ifp->lock);
2063
2064                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2065                                       jiffies_to_clock_t(valid_lft * HZ), flags);
2066                 /*
2067                  * Note that section 3.1 of RFC 4429 indicates
2068                  * that the Optimistic flag should not be set for
2069                  * manually configured addresses
2070                  */
2071                 addrconf_dad_start(ifp, 0);
2072                 in6_ifa_put(ifp);
2073                 addrconf_verify(0);
2074                 return 0;
2075         }
2076
2077         return PTR_ERR(ifp);
2078 }
2079
2080 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2081                           int plen)
2082 {
2083         struct inet6_ifaddr *ifp;
2084         struct inet6_dev *idev;
2085         struct net_device *dev;
2086
2087         dev = __dev_get_by_index(net, ifindex);
2088         if (!dev)
2089                 return -ENODEV;
2090
2091         if ((idev = __in6_dev_get(dev)) == NULL)
2092                 return -ENXIO;
2093
2094         read_lock_bh(&idev->lock);
2095         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2096                 if (ifp->prefix_len == plen &&
2097                     ipv6_addr_equal(pfx, &ifp->addr)) {
2098                         in6_ifa_hold(ifp);
2099                         read_unlock_bh(&idev->lock);
2100
2101                         ipv6_del_addr(ifp);
2102
2103                         /* If the last address is deleted administratively,
2104                            disable IPv6 on this interface.
2105                          */
2106                         if (idev->addr_list == NULL)
2107                                 addrconf_ifdown(idev->dev, 1);
2108                         return 0;
2109                 }
2110         }
2111         read_unlock_bh(&idev->lock);
2112         return -EADDRNOTAVAIL;
2113 }
2114
2115
2116 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2117 {
2118         struct in6_ifreq ireq;
2119         int err;
2120
2121         if (!capable(CAP_NET_ADMIN))
2122                 return -EPERM;
2123
2124         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2125                 return -EFAULT;
2126
2127         rtnl_lock();
2128         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2129                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2130                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2131         rtnl_unlock();
2132         return err;
2133 }
2134
2135 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2136 {
2137         struct in6_ifreq ireq;
2138         int err;
2139
2140         if (!capable(CAP_NET_ADMIN))
2141                 return -EPERM;
2142
2143         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2144                 return -EFAULT;
2145
2146         rtnl_lock();
2147         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2148                              ireq.ifr6_prefixlen);
2149         rtnl_unlock();
2150         return err;
2151 }
2152
2153 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2154 static void sit_add_v4_addrs(struct inet6_dev *idev)
2155 {
2156         struct inet6_ifaddr * ifp;
2157         struct in6_addr addr;
2158         struct net_device *dev;
2159         struct net *net = dev_net(idev->dev);
2160         int scope;
2161
2162         ASSERT_RTNL();
2163
2164         memset(&addr, 0, sizeof(struct in6_addr));
2165         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2166
2167         if (idev->dev->flags&IFF_POINTOPOINT) {
2168                 addr.s6_addr32[0] = htonl(0xfe800000);
2169                 scope = IFA_LINK;
2170         } else {
2171                 scope = IPV6_ADDR_COMPATv4;
2172         }
2173
2174         if (addr.s6_addr32[3]) {
2175                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2176                 if (!IS_ERR(ifp)) {
2177                         spin_lock_bh(&ifp->lock);
2178                         ifp->flags &= ~IFA_F_TENTATIVE;
2179                         spin_unlock_bh(&ifp->lock);
2180                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2181                         in6_ifa_put(ifp);
2182                 }
2183                 return;
2184         }
2185
2186         for_each_netdev(net, dev) {
2187                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2188                 if (in_dev && (dev->flags & IFF_UP)) {
2189                         struct in_ifaddr * ifa;
2190
2191                         int flag = scope;
2192
2193                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2194                                 int plen;
2195
2196                                 addr.s6_addr32[3] = ifa->ifa_local;
2197
2198                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2199                                         continue;
2200                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2201                                         if (idev->dev->flags&IFF_POINTOPOINT)
2202                                                 continue;
2203                                         flag |= IFA_HOST;
2204                                 }
2205                                 if (idev->dev->flags&IFF_POINTOPOINT)
2206                                         plen = 64;
2207                                 else
2208                                         plen = 96;
2209
2210                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2211                                                     IFA_F_PERMANENT);
2212                                 if (!IS_ERR(ifp)) {
2213                                         spin_lock_bh(&ifp->lock);
2214                                         ifp->flags &= ~IFA_F_TENTATIVE;
2215                                         spin_unlock_bh(&ifp->lock);
2216                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2217                                         in6_ifa_put(ifp);
2218                                 }
2219                         }
2220                 }
2221         }
2222 }
2223 #endif
2224
2225 static void init_loopback(struct net_device *dev)
2226 {
2227         struct inet6_dev  *idev;
2228         struct inet6_ifaddr * ifp;
2229
2230         /* ::1 */
2231
2232         ASSERT_RTNL();
2233
2234         if ((idev = ipv6_find_idev(dev)) == NULL) {
2235                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2236                 return;
2237         }
2238
2239         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2240         if (!IS_ERR(ifp)) {
2241                 spin_lock_bh(&ifp->lock);
2242                 ifp->flags &= ~IFA_F_TENTATIVE;
2243                 spin_unlock_bh(&ifp->lock);
2244                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2245                 in6_ifa_put(ifp);
2246         }
2247 }
2248
2249 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2250 {
2251         struct inet6_ifaddr * ifp;
2252         u32 addr_flags = IFA_F_PERMANENT;
2253
2254 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2255         if (idev->cnf.optimistic_dad &&
2256             !ipv6_devconf.forwarding)
2257                 addr_flags |= IFA_F_OPTIMISTIC;
2258 #endif
2259
2260
2261         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2262         if (!IS_ERR(ifp)) {
2263                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2264                 addrconf_dad_start(ifp, 0);
2265                 in6_ifa_put(ifp);
2266         }
2267 }
2268
2269 static void addrconf_dev_config(struct net_device *dev)
2270 {
2271         struct in6_addr addr;
2272         struct inet6_dev    * idev;
2273
2274         ASSERT_RTNL();
2275
2276         if ((dev->type != ARPHRD_ETHER) &&
2277             (dev->type != ARPHRD_FDDI) &&
2278             (dev->type != ARPHRD_IEEE802_TR) &&
2279             (dev->type != ARPHRD_ARCNET) &&
2280             (dev->type != ARPHRD_INFINIBAND)) {
2281                 /* Alas, we support only Ethernet autoconfiguration. */
2282                 return;
2283         }
2284
2285         idev = addrconf_add_dev(dev);
2286         if (idev == NULL)
2287                 return;
2288
2289         memset(&addr, 0, sizeof(struct in6_addr));
2290         addr.s6_addr32[0] = htonl(0xFE800000);
2291
2292         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2293                 addrconf_add_linklocal(idev, &addr);
2294 }
2295
2296 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2297 static void addrconf_sit_config(struct net_device *dev)
2298 {
2299         struct inet6_dev *idev;
2300
2301         ASSERT_RTNL();
2302
2303         /*
2304          * Configure the tunnel with one of our IPv4
2305          * addresses... we should configure all of
2306          * our v4 addrs in the tunnel
2307          */
2308
2309         if ((idev = ipv6_find_idev(dev)) == NULL) {
2310                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2311                 return;
2312         }
2313
2314         if (dev->priv_flags & IFF_ISATAP) {
2315                 struct in6_addr addr;
2316
2317                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2318                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2319                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2320                         addrconf_add_linklocal(idev, &addr);
2321                 return;
2322         }
2323
2324         sit_add_v4_addrs(idev);
2325
2326         if (dev->flags&IFF_POINTOPOINT) {
2327                 addrconf_add_mroute(dev);
2328                 addrconf_add_lroute(dev);
2329         } else
2330                 sit_route_add(dev);
2331 }
2332 #endif
2333
2334 static inline int
2335 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2336 {
2337         struct in6_addr lladdr;
2338
2339         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2340                 addrconf_add_linklocal(idev, &lladdr);
2341                 return 0;
2342         }
2343         return -1;
2344 }
2345
2346 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2347 {
2348         struct net_device *link_dev;
2349         struct net *net = dev_net(idev->dev);
2350
2351         /* first try to inherit the link-local address from the link device */
2352         if (idev->dev->iflink &&
2353             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2354                 if (!ipv6_inherit_linklocal(idev, link_dev))
2355                         return;
2356         }
2357         /* then try to inherit it from any device */
2358         for_each_netdev(net, link_dev) {
2359                 if (!ipv6_inherit_linklocal(idev, link_dev))
2360                         return;
2361         }
2362         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2363 }
2364
2365 /*
2366  * Autoconfigure tunnel with a link-local address so routing protocols,
2367  * DHCPv6, MLD etc. can be run over the virtual link
2368  */
2369
2370 static void addrconf_ip6_tnl_config(struct net_device *dev)
2371 {
2372         struct inet6_dev *idev;
2373
2374         ASSERT_RTNL();
2375
2376         if ((idev = addrconf_add_dev(dev)) == NULL) {
2377                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2378                 return;
2379         }
2380         ip6_tnl_add_linklocal(idev);
2381 }
2382
2383 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2384                            void * data)
2385 {
2386         struct net_device *dev = (struct net_device *) data;
2387         struct inet6_dev *idev = __in6_dev_get(dev);
2388         int run_pending = 0;
2389         int err;
2390
2391         switch(event) {
2392         case NETDEV_REGISTER:
2393                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2394                         idev = ipv6_add_dev(dev);
2395                         if (!idev)
2396                                 return notifier_from_errno(-ENOMEM);
2397                 }
2398                 break;
2399         case NETDEV_UP:
2400         case NETDEV_CHANGE:
2401                 if (dev->flags & IFF_SLAVE)
2402                         break;
2403
2404                 if (event == NETDEV_UP) {
2405                         if (!addrconf_qdisc_ok(dev)) {
2406                                 /* device is not ready yet. */
2407                                 printk(KERN_INFO
2408                                         "ADDRCONF(NETDEV_UP): %s: "
2409                                         "link is not ready\n",
2410                                         dev->name);
2411                                 break;
2412                         }
2413
2414                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2415                                 idev = ipv6_add_dev(dev);
2416
2417                         if (idev)
2418                                 idev->if_flags |= IF_READY;
2419                 } else {
2420                         if (!addrconf_qdisc_ok(dev)) {
2421                                 /* device is still not ready. */
2422                                 break;
2423                         }
2424
2425                         if (idev) {
2426                                 if (idev->if_flags & IF_READY) {
2427                                         /* device is already configured. */
2428                                         break;
2429                                 }
2430                                 idev->if_flags |= IF_READY;
2431                         }
2432
2433                         printk(KERN_INFO
2434                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2435                                         "link becomes ready\n",
2436                                         dev->name);
2437
2438                         run_pending = 1;
2439                 }
2440
2441                 switch(dev->type) {
2442 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2443                 case ARPHRD_SIT:
2444                         addrconf_sit_config(dev);
2445                         break;
2446 #endif
2447                 case ARPHRD_TUNNEL6:
2448                         addrconf_ip6_tnl_config(dev);
2449                         break;
2450                 case ARPHRD_LOOPBACK:
2451                         init_loopback(dev);
2452                         break;
2453
2454                 default:
2455                         addrconf_dev_config(dev);
2456                         break;
2457                 }
2458                 if (idev) {
2459                         if (run_pending)
2460                                 addrconf_dad_run(idev);
2461
2462                         /* If the MTU changed during the interface down, when the
2463                            interface up, the changed MTU must be reflected in the
2464                            idev as well as routers.
2465                          */
2466                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2467                                 rt6_mtu_change(dev, dev->mtu);
2468                                 idev->cnf.mtu6 = dev->mtu;
2469                         }
2470                         idev->tstamp = jiffies;
2471                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2472                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2473                            stop IPv6 on this interface.
2474                          */
2475                         if (dev->mtu < IPV6_MIN_MTU)
2476                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2477                 }
2478                 break;
2479
2480         case NETDEV_CHANGEMTU:
2481                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2482                         rt6_mtu_change(dev, dev->mtu);
2483                         idev->cnf.mtu6 = dev->mtu;
2484                         break;
2485                 }
2486
2487                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2488                         idev = ipv6_add_dev(dev);
2489                         if (idev)
2490                                 break;
2491                 }
2492
2493                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2494
2495         case NETDEV_DOWN:
2496         case NETDEV_UNREGISTER:
2497                 /*
2498                  *      Remove all addresses from this interface.
2499                  */
2500                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2501                 break;
2502
2503         case NETDEV_CHANGENAME:
2504                 if (idev) {
2505                         snmp6_unregister_dev(idev);
2506                         addrconf_sysctl_unregister(idev);
2507                         addrconf_sysctl_register(idev);
2508                         err = snmp6_register_dev(idev);
2509                         if (err)
2510                                 return notifier_from_errno(err);
2511                 }
2512                 break;
2513         }
2514
2515         return NOTIFY_OK;
2516 }
2517
2518 /*
2519  *      addrconf module should be notified of a device going up
2520  */
2521 static struct notifier_block ipv6_dev_notf = {
2522         .notifier_call = addrconf_notify,
2523         .priority = 0
2524 };
2525
2526 static int addrconf_ifdown(struct net_device *dev, int how)
2527 {
2528         struct inet6_dev *idev;
2529         struct inet6_ifaddr *ifa, **bifa;
2530         struct net *net = dev_net(dev);
2531         int i;
2532
2533         ASSERT_RTNL();
2534
2535         if (dev == init_net.loopback_dev && how == 1)
2536                 how = 0;
2537
2538         rt6_ifdown(net, dev);
2539         neigh_ifdown(&nd_tbl, dev);
2540
2541         idev = __in6_dev_get(dev);
2542         if (idev == NULL)
2543                 return -ENODEV;
2544
2545         /* Step 1: remove reference to ipv6 device from parent device.
2546                    Do not dev_put!
2547          */
2548         if (how == 1) {
2549                 idev->dead = 1;
2550
2551                 /* protected by rtnl_lock */
2552                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2553
2554                 /* Step 1.5: remove snmp6 entry */
2555                 snmp6_unregister_dev(idev);
2556
2557         }
2558
2559         /* Step 2: clear hash table */
2560         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2561                 bifa = &inet6_addr_lst[i];
2562
2563                 write_lock_bh(&addrconf_hash_lock);
2564                 while ((ifa = *bifa) != NULL) {
2565                         if (ifa->idev == idev) {
2566                                 *bifa = ifa->lst_next;
2567                                 ifa->lst_next = NULL;
2568                                 addrconf_del_timer(ifa);
2569                                 in6_ifa_put(ifa);
2570                                 continue;
2571                         }
2572                         bifa = &ifa->lst_next;
2573                 }
2574                 write_unlock_bh(&addrconf_hash_lock);
2575         }
2576
2577         write_lock_bh(&idev->lock);
2578
2579         /* Step 3: clear flags for stateless addrconf */
2580         if (how != 1)
2581                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2582
2583         /* Step 4: clear address list */
2584 #ifdef CONFIG_IPV6_PRIVACY
2585         if (how == 1 && del_timer(&idev->regen_timer))
2586                 in6_dev_put(idev);
2587
2588         /* clear tempaddr list */
2589         while ((ifa = idev->tempaddr_list) != NULL) {
2590                 idev->tempaddr_list = ifa->tmp_next;
2591                 ifa->tmp_next = NULL;
2592                 ifa->dead = 1;
2593                 write_unlock_bh(&idev->lock);
2594                 spin_lock_bh(&ifa->lock);
2595
2596                 if (ifa->ifpub) {
2597                         in6_ifa_put(ifa->ifpub);
2598                         ifa->ifpub = NULL;
2599                 }
2600                 spin_unlock_bh(&ifa->lock);
2601                 in6_ifa_put(ifa);
2602                 write_lock_bh(&idev->lock);
2603         }
2604 #endif
2605         while ((ifa = idev->addr_list) != NULL) {
2606                 idev->addr_list = ifa->if_next;
2607                 ifa->if_next = NULL;
2608                 ifa->dead = 1;
2609                 addrconf_del_timer(ifa);
2610                 write_unlock_bh(&idev->lock);
2611
2612                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2613                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2614                 in6_ifa_put(ifa);
2615
2616                 write_lock_bh(&idev->lock);
2617         }
2618         write_unlock_bh(&idev->lock);
2619
2620         /* Step 5: Discard multicast list */
2621
2622         if (how == 1)
2623                 ipv6_mc_destroy_dev(idev);
2624         else
2625                 ipv6_mc_down(idev);
2626
2627         idev->tstamp = jiffies;
2628
2629         /* Shot the device (if unregistered) */
2630
2631         if (how == 1) {
2632                 addrconf_sysctl_unregister(idev);
2633                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2634                 neigh_ifdown(&nd_tbl, dev);
2635                 in6_dev_put(idev);
2636         }
2637         return 0;
2638 }
2639
2640 static void addrconf_rs_timer(unsigned long data)
2641 {
2642         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2643
2644         if (ifp->idev->cnf.forwarding)
2645                 goto out;
2646
2647         if (ifp->idev->if_flags & IF_RA_RCVD) {
2648                 /*
2649                  *      Announcement received after solicitation
2650                  *      was sent
2651                  */
2652                 goto out;
2653         }
2654
2655         spin_lock(&ifp->lock);
2656         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2657                 struct in6_addr all_routers;
2658
2659                 /* The wait after the last probe can be shorter */
2660                 addrconf_mod_timer(ifp, AC_RS,
2661                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2662                                    ifp->idev->cnf.rtr_solicit_delay :
2663                                    ifp->idev->cnf.rtr_solicit_interval);
2664                 spin_unlock(&ifp->lock);
2665
2666                 ipv6_addr_all_routers(&all_routers);
2667
2668                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2669         } else {
2670                 spin_unlock(&ifp->lock);
2671                 /*
2672                  * Note: we do not support deprecated "all on-link"
2673                  * assumption any longer.
2674                  */
2675                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2676                        ifp->idev->dev->name);
2677         }
2678
2679 out:
2680         in6_ifa_put(ifp);
2681 }
2682
2683 /*
2684  *      Duplicate Address Detection
2685  */
2686 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2687 {
2688         unsigned long rand_num;
2689         struct inet6_dev *idev = ifp->idev;
2690
2691         if (ifp->flags & IFA_F_OPTIMISTIC)
2692                 rand_num = 0;
2693         else
2694                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2695
2696         ifp->probes = idev->cnf.dad_transmits;
2697         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2698 }
2699
2700 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2701 {
2702         struct inet6_dev *idev = ifp->idev;
2703         struct net_device *dev = idev->dev;
2704
2705         addrconf_join_solict(dev, &ifp->addr);
2706
2707         net_srandom(ifp->addr.s6_addr32[3]);
2708
2709         read_lock_bh(&idev->lock);
2710         if (ifp->dead)
2711                 goto out;
2712         spin_lock_bh(&ifp->lock);
2713
2714         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2715             !(ifp->flags&IFA_F_TENTATIVE) ||
2716             ifp->flags & IFA_F_NODAD) {
2717                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2718                 spin_unlock_bh(&ifp->lock);
2719                 read_unlock_bh(&idev->lock);
2720
2721                 addrconf_dad_completed(ifp);
2722                 return;
2723         }
2724
2725         if (!(idev->if_flags & IF_READY)) {
2726                 spin_unlock_bh(&ifp->lock);
2727                 read_unlock_bh(&idev->lock);
2728                 /*
2729                  * If the defice is not ready:
2730                  * - keep it tentative if it is a permanent address.
2731                  * - otherwise, kill it.
2732                  */
2733                 in6_ifa_hold(ifp);
2734                 addrconf_dad_stop(ifp);
2735                 return;
2736         }
2737
2738         /*
2739          * Optimistic nodes can start receiving
2740          * Frames right away
2741          */
2742         if(ifp->flags & IFA_F_OPTIMISTIC)
2743                 ip6_ins_rt(ifp->rt);
2744
2745         addrconf_dad_kick(ifp);
2746         spin_unlock_bh(&ifp->lock);
2747 out:
2748         read_unlock_bh(&idev->lock);
2749 }
2750
2751 static void addrconf_dad_timer(unsigned long data)
2752 {
2753         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2754         struct inet6_dev *idev = ifp->idev;
2755         struct in6_addr unspec;
2756         struct in6_addr mcaddr;
2757
2758         read_lock_bh(&idev->lock);
2759         if (idev->dead) {
2760                 read_unlock_bh(&idev->lock);
2761                 goto out;
2762         }
2763         spin_lock_bh(&ifp->lock);
2764         if (ifp->probes == 0) {
2765                 /*
2766                  * DAD was successful
2767                  */
2768
2769                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2770                 spin_unlock_bh(&ifp->lock);
2771                 read_unlock_bh(&idev->lock);
2772
2773                 addrconf_dad_completed(ifp);
2774
2775                 goto out;
2776         }
2777
2778         ifp->probes--;
2779         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2780         spin_unlock_bh(&ifp->lock);
2781         read_unlock_bh(&idev->lock);
2782
2783         /* send a neighbour solicitation for our addr */
2784         memset(&unspec, 0, sizeof(unspec));
2785         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2786         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2787 out:
2788         in6_ifa_put(ifp);
2789 }
2790
2791 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2792 {
2793         struct net_device *     dev = ifp->idev->dev;
2794
2795         /*
2796          *      Configure the address for reception. Now it is valid.
2797          */
2798
2799         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2800
2801         /* If added prefix is link local and forwarding is off,
2802            start sending router solicitations.
2803          */
2804
2805         if (ifp->idev->cnf.forwarding == 0 &&
2806             ifp->idev->cnf.rtr_solicits > 0 &&
2807             (dev->flags&IFF_LOOPBACK) == 0 &&
2808             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2809                 struct in6_addr all_routers;
2810
2811                 ipv6_addr_all_routers(&all_routers);
2812
2813                 /*
2814                  *      If a host as already performed a random delay
2815                  *      [...] as part of DAD [...] there is no need
2816                  *      to delay again before sending the first RS
2817                  */
2818                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2819
2820                 spin_lock_bh(&ifp->lock);
2821                 ifp->probes = 1;
2822                 ifp->idev->if_flags |= IF_RS_SENT;
2823                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2824                 spin_unlock_bh(&ifp->lock);
2825         }
2826 }
2827
2828 static void addrconf_dad_run(struct inet6_dev *idev) {
2829         struct inet6_ifaddr *ifp;
2830
2831         read_lock_bh(&idev->lock);
2832         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2833                 spin_lock_bh(&ifp->lock);
2834                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2835                         spin_unlock_bh(&ifp->lock);
2836                         continue;
2837                 }
2838                 spin_unlock_bh(&ifp->lock);
2839                 addrconf_dad_kick(ifp);
2840         }
2841         read_unlock_bh(&idev->lock);
2842 }
2843
2844 #ifdef CONFIG_PROC_FS
2845 struct if6_iter_state {
2846         struct seq_net_private p;
2847         int bucket;
2848 };
2849
2850 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2851 {
2852         struct inet6_ifaddr *ifa = NULL;
2853         struct if6_iter_state *state = seq->private;
2854         struct net *net = seq_file_net(seq);
2855
2856         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2857                 ifa = inet6_addr_lst[state->bucket];
2858
2859                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2860                         ifa = ifa->lst_next;
2861                 if (ifa)
2862                         break;
2863         }
2864         return ifa;
2865 }
2866
2867 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2868 {
2869         struct if6_iter_state *state = seq->private;
2870         struct net *net = seq_file_net(seq);
2871
2872         ifa = ifa->lst_next;
2873 try_again:
2874         if (ifa) {
2875                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2876                         ifa = ifa->lst_next;
2877                         goto try_again;
2878                 }
2879         }
2880
2881         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2882                 ifa = inet6_addr_lst[state->bucket];
2883                 goto try_again;
2884         }
2885
2886         return ifa;
2887 }
2888
2889 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2890 {
2891         struct inet6_ifaddr *ifa = if6_get_first(seq);
2892
2893         if (ifa)
2894                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2895                         --pos;
2896         return pos ? NULL : ifa;
2897 }
2898
2899 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2900         __acquires(addrconf_hash_lock)
2901 {
2902         read_lock_bh(&addrconf_hash_lock);
2903         return if6_get_idx(seq, *pos);
2904 }
2905
2906 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2907 {
2908         struct inet6_ifaddr *ifa;
2909
2910         ifa = if6_get_next(seq, v);
2911         ++*pos;
2912         return ifa;
2913 }
2914
2915 static void if6_seq_stop(struct seq_file *seq, void *v)
2916         __releases(addrconf_hash_lock)
2917 {
2918         read_unlock_bh(&addrconf_hash_lock);
2919 }
2920
2921 static int if6_seq_show(struct seq_file *seq, void *v)
2922 {
2923         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2924         seq_printf(seq,
2925                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2926                    NIP6(ifp->addr),
2927                    ifp->idev->dev->ifindex,
2928                    ifp->prefix_len,
2929                    ifp->scope,
2930                    ifp->flags,
2931                    ifp->idev->dev->name);
2932         return 0;
2933 }
2934
2935 static const struct seq_operations if6_seq_ops = {
2936         .start  = if6_seq_start,
2937         .next   = if6_seq_next,
2938         .show   = if6_seq_show,
2939         .stop   = if6_seq_stop,
2940 };
2941
2942 static int if6_seq_open(struct inode *inode, struct file *file)
2943 {
2944         return seq_open_net(inode, file, &if6_seq_ops,
2945                             sizeof(struct if6_iter_state));
2946 }
2947
2948 static const struct file_operations if6_fops = {
2949         .owner          = THIS_MODULE,
2950         .open           = if6_seq_open,
2951         .read           = seq_read,
2952         .llseek         = seq_lseek,
2953         .release        = seq_release_net,
2954 };
2955
2956 static int if6_proc_net_init(struct net *net)
2957 {
2958         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2959                 return -ENOMEM;
2960         return 0;
2961 }
2962
2963 static void if6_proc_net_exit(struct net *net)
2964 {
2965        proc_net_remove(net, "if_inet6");
2966 }
2967
2968 static struct pernet_operations if6_proc_net_ops = {
2969        .init = if6_proc_net_init,
2970        .exit = if6_proc_net_exit,
2971 };
2972
2973 int __init if6_proc_init(void)
2974 {
2975         return register_pernet_subsys(&if6_proc_net_ops);
2976 }
2977
2978 void if6_proc_exit(void)
2979 {
2980         unregister_pernet_subsys(&if6_proc_net_ops);
2981 }
2982 #endif  /* CONFIG_PROC_FS */
2983
2984 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2985 /* Check if address is a home address configured on any interface. */
2986 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
2987 {
2988         int ret = 0;
2989         struct inet6_ifaddr * ifp;
2990         u8 hash = ipv6_addr_hash(addr);
2991         read_lock_bh(&addrconf_hash_lock);
2992         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2993                 if (!net_eq(dev_net(ifp->idev->dev), net))
2994                         continue;
2995                 if (ipv6_addr_equal(&ifp->addr, addr) &&
2996                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2997                         ret = 1;
2998                         break;
2999                 }
3000         }
3001         read_unlock_bh(&addrconf_hash_lock);
3002         return ret;
3003 }
3004 #endif
3005
3006 /*
3007  *      Periodic address status verification
3008  */
3009
3010 static void addrconf_verify(unsigned long foo)
3011 {
3012         struct inet6_ifaddr *ifp;
3013         unsigned long now, next;
3014         int i;
3015
3016         spin_lock_bh(&addrconf_verify_lock);
3017         now = jiffies;
3018         next = now + ADDR_CHECK_FREQUENCY;
3019
3020         del_timer(&addr_chk_timer);
3021
3022         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3023
3024 restart:
3025                 read_lock(&addrconf_hash_lock);
3026                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3027                         unsigned long age;
3028 #ifdef CONFIG_IPV6_PRIVACY
3029                         unsigned long regen_advance;
3030 #endif
3031
3032                         if (ifp->flags & IFA_F_PERMANENT)
3033                                 continue;
3034
3035                         spin_lock(&ifp->lock);
3036                         age = (now - ifp->tstamp) / HZ;
3037
3038 #ifdef CONFIG_IPV6_PRIVACY
3039                         regen_advance = ifp->idev->cnf.regen_max_retry *
3040                                         ifp->idev->cnf.dad_transmits *
3041                                         ifp->idev->nd_parms->retrans_time / HZ;
3042 #endif
3043
3044                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3045                             age >= ifp->valid_lft) {
3046                                 spin_unlock(&ifp->lock);
3047                                 in6_ifa_hold(ifp);
3048                                 read_unlock(&addrconf_hash_lock);
3049                                 ipv6_del_addr(ifp);
3050                                 goto restart;
3051                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3052                                 spin_unlock(&ifp->lock);
3053                                 continue;
3054                         } else if (age >= ifp->prefered_lft) {
3055                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3056                                 int deprecate = 0;
3057
3058                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3059                                         deprecate = 1;
3060                                         ifp->flags |= IFA_F_DEPRECATED;
3061                                 }
3062
3063                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3064                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3065
3066                                 spin_unlock(&ifp->lock);
3067
3068                                 if (deprecate) {
3069                                         in6_ifa_hold(ifp);
3070                                         read_unlock(&addrconf_hash_lock);
3071
3072                                         ipv6_ifa_notify(0, ifp);
3073                                         in6_ifa_put(ifp);
3074                                         goto restart;
3075                                 }
3076 #ifdef CONFIG_IPV6_PRIVACY
3077                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3078                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3079                                 if (age >= ifp->prefered_lft - regen_advance) {
3080                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3081                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3082                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3083                                         if (!ifp->regen_count && ifpub) {
3084                                                 ifp->regen_count++;
3085                                                 in6_ifa_hold(ifp);
3086                                                 in6_ifa_hold(ifpub);
3087                                                 spin_unlock(&ifp->lock);
3088                                                 read_unlock(&addrconf_hash_lock);
3089                                                 spin_lock(&ifpub->lock);
3090                                                 ifpub->regen_count = 0;
3091                                                 spin_unlock(&ifpub->lock);
3092                                                 ipv6_create_tempaddr(ifpub, ifp);
3093                                                 in6_ifa_put(ifpub);
3094                                                 in6_ifa_put(ifp);
3095                                                 goto restart;
3096                                         }
3097                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3098                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3099                                 spin_unlock(&ifp->lock);
3100 #endif
3101                         } else {
3102                                 /* ifp->prefered_lft <= ifp->valid_lft */
3103                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3104                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3105                                 spin_unlock(&ifp->lock);
3106                         }
3107                 }
3108                 read_unlock(&addrconf_hash_lock);
3109         }
3110
3111         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3112         add_timer(&addr_chk_timer);
3113         spin_unlock_bh(&addrconf_verify_lock);
3114 }
3115
3116 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3117 {
3118         struct in6_addr *pfx = NULL;
3119
3120         if (addr)
3121                 pfx = nla_data(addr);
3122
3123         if (local) {
3124                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3125                         pfx = NULL;
3126                 else
3127                         pfx = nla_data(local);
3128         }
3129
3130         return pfx;
3131 }
3132
3133 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3134         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3135         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3136         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3137 };
3138
3139 static int
3140 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3141 {
3142         struct net *net = sock_net(skb->sk);
3143         struct ifaddrmsg *ifm;
3144         struct nlattr *tb[IFA_MAX+1];
3145         struct in6_addr *pfx;
3146         int err;
3147
3148         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3149         if (err < 0)
3150                 return err;
3151
3152         ifm = nlmsg_data(nlh);
3153         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3154         if (pfx == NULL)
3155                 return -EINVAL;
3156
3157         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3158 }
3159
3160 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3161                              u32 prefered_lft, u32 valid_lft)
3162 {
3163         u32 flags = RTF_EXPIRES;
3164
3165         if (!valid_lft || (prefered_lft > valid_lft))
3166                 return -EINVAL;
3167
3168         if (valid_lft == INFINITY_LIFE_TIME) {
3169                 ifa_flags |= IFA_F_PERMANENT;
3170                 flags = 0;
3171         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3172                 valid_lft = 0x7FFFFFFF/HZ;
3173
3174         if (prefered_lft == 0)
3175                 ifa_flags |= IFA_F_DEPRECATED;
3176         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3177                  (prefered_lft != INFINITY_LIFE_TIME))
3178                 prefered_lft = 0x7FFFFFFF/HZ;
3179
3180         spin_lock_bh(&ifp->lock);
3181         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3182         ifp->tstamp = jiffies;
3183         ifp->valid_lft = valid_lft;
3184         ifp->prefered_lft = prefered_lft;
3185
3186         spin_unlock_bh(&ifp->lock);
3187         if (!(ifp->flags&IFA_F_TENTATIVE))
3188                 ipv6_ifa_notify(0, ifp);
3189
3190         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3191                               jiffies_to_clock_t(valid_lft * HZ), flags);
3192         addrconf_verify(0);
3193
3194         return 0;
3195 }
3196
3197 static int
3198 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3199 {
3200         struct net *net = sock_net(skb->sk);
3201         struct ifaddrmsg *ifm;
3202         struct nlattr *tb[IFA_MAX+1];
3203         struct in6_addr *pfx;
3204         struct inet6_ifaddr *ifa;
3205         struct net_device *dev;
3206         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3207         u8 ifa_flags;
3208         int err;
3209
3210         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3211         if (err < 0)
3212                 return err;
3213
3214         ifm = nlmsg_data(nlh);
3215         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3216         if (pfx == NULL)
3217                 return -EINVAL;
3218
3219         if (tb[IFA_CACHEINFO]) {
3220                 struct ifa_cacheinfo *ci;
3221
3222                 ci = nla_data(tb[IFA_CACHEINFO]);
3223                 valid_lft = ci->ifa_valid;
3224                 preferred_lft = ci->ifa_prefered;
3225         } else {
3226                 preferred_lft = INFINITY_LIFE_TIME;
3227                 valid_lft = INFINITY_LIFE_TIME;
3228         }
3229
3230         dev =  __dev_get_by_index(net, ifm->ifa_index);
3231         if (dev == NULL)
3232                 return -ENODEV;
3233
3234         /* We ignore other flags so far. */
3235         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3236
3237         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3238         if (ifa == NULL) {
3239                 /*
3240                  * It would be best to check for !NLM_F_CREATE here but
3241                  * userspace alreay relies on not having to provide this.
3242                  */
3243                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3244                                       ifm->ifa_prefixlen, ifa_flags,
3245                                       preferred_lft, valid_lft);
3246         }
3247
3248         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3249             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3250                 err = -EEXIST;
3251         else
3252                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3253
3254         in6_ifa_put(ifa);
3255
3256         return err;
3257 }
3258
3259 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3260                           u8 scope, int ifindex)
3261 {
3262         struct ifaddrmsg *ifm;
3263
3264         ifm = nlmsg_data(nlh);
3265         ifm->ifa_family = AF_INET6;
3266         ifm->ifa_prefixlen = prefixlen;
3267         ifm->ifa_flags = flags;
3268         ifm->ifa_scope = scope;
3269         ifm->ifa_index = ifindex;
3270 }
3271
3272 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3273                          unsigned long tstamp, u32 preferred, u32 valid)
3274 {
3275         struct ifa_cacheinfo ci;
3276
3277         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3278                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3279         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3280                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3281         ci.ifa_prefered = preferred;
3282         ci.ifa_valid = valid;
3283
3284         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3285 }
3286
3287 static inline int rt_scope(int ifa_scope)
3288 {
3289         if (ifa_scope & IFA_HOST)
3290                 return RT_SCOPE_HOST;
3291         else if (ifa_scope & IFA_LINK)
3292                 return RT_SCOPE_LINK;
3293         else if (ifa_scope & IFA_SITE)
3294                 return RT_SCOPE_SITE;
3295         else
3296                 return RT_SCOPE_UNIVERSE;
3297 }
3298
3299 static inline int inet6_ifaddr_msgsize(void)
3300 {
3301         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3302                + nla_total_size(16) /* IFA_ADDRESS */
3303                + nla_total_size(sizeof(struct ifa_cacheinfo));
3304 }
3305
3306 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3307                              u32 pid, u32 seq, int event, unsigned int flags)
3308 {
3309         struct nlmsghdr  *nlh;
3310         u32 preferred, valid;
3311
3312         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3313         if (nlh == NULL)
3314                 return -EMSGSIZE;
3315
3316         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3317                       ifa->idev->dev->ifindex);
3318
3319         if (!(ifa->flags&IFA_F_PERMANENT)) {
3320                 preferred = ifa->prefered_lft;
3321                 valid = ifa->valid_lft;
3322                 if (preferred != INFINITY_LIFE_TIME) {
3323                         long tval = (jiffies - ifa->tstamp)/HZ;
3324                         preferred -= tval;
3325                         if (valid != INFINITY_LIFE_TIME)
3326                                 valid -= tval;
3327                 }
3328         } else {
3329                 preferred = INFINITY_LIFE_TIME;
3330                 valid = INFINITY_LIFE_TIME;
3331         }
3332
3333         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3334             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3335                 nlmsg_cancel(skb, nlh);
3336                 return -EMSGSIZE;
3337         }
3338
3339         return nlmsg_end(skb, nlh);
3340 }
3341
3342 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3343                                 u32 pid, u32 seq, int event, u16 flags)
3344 {
3345         struct nlmsghdr  *nlh;
3346         u8 scope = RT_SCOPE_UNIVERSE;
3347         int ifindex = ifmca->idev->dev->ifindex;
3348
3349         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3350                 scope = RT_SCOPE_SITE;
3351
3352         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3353         if (nlh == NULL)
3354                 return -EMSGSIZE;
3355
3356         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3357         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3358             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3359                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3360                 nlmsg_cancel(skb, nlh);
3361                 return -EMSGSIZE;
3362         }
3363
3364         return nlmsg_end(skb, nlh);
3365 }
3366
3367 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3368                                 u32 pid, u32 seq, int event, unsigned int flags)
3369 {
3370         struct nlmsghdr  *nlh;
3371         u8 scope = RT_SCOPE_UNIVERSE;
3372         int ifindex = ifaca->aca_idev->dev->ifindex;
3373
3374         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3375                 scope = RT_SCOPE_SITE;
3376
3377         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3378         if (nlh == NULL)
3379                 return -EMSGSIZE;
3380
3381         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3382         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3383             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3384                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3385                 nlmsg_cancel(skb, nlh);
3386                 return -EMSGSIZE;
3387         }
3388
3389         return nlmsg_end(skb, nlh);
3390 }
3391
3392 enum addr_type_t
3393 {
3394         UNICAST_ADDR,
3395         MULTICAST_ADDR,
3396         ANYCAST_ADDR,
3397 };
3398
3399 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3400                            enum addr_type_t type)
3401 {
3402         int idx, ip_idx;
3403         int s_idx, s_ip_idx;
3404         int err = 1;
3405         struct net_device *dev;
3406         struct inet6_dev *idev = NULL;
3407         struct inet6_ifaddr *ifa;
3408         struct ifmcaddr6 *ifmca;
3409         struct ifacaddr6 *ifaca;
3410         struct net *net = sock_net(skb->sk);
3411
3412         s_idx = cb->args[0];
3413         s_ip_idx = ip_idx = cb->args[1];
3414
3415         idx = 0;
3416         for_each_netdev(net, dev) {
3417                 if (idx < s_idx)
3418                         goto cont;
3419                 if (idx > s_idx)
3420                         s_ip_idx = 0;
3421                 ip_idx = 0;
3422                 if ((idev = in6_dev_get(dev)) == NULL)
3423                         goto cont;
3424                 read_lock_bh(&idev->lock);
3425                 switch (type) {
3426                 case UNICAST_ADDR:
3427                         /* unicast address incl. temp addr */
3428                         for (ifa = idev->addr_list; ifa;
3429                              ifa = ifa->if_next, ip_idx++) {
3430                                 if (ip_idx < s_ip_idx)
3431                                         continue;
3432                                 err = inet6_fill_ifaddr(skb, ifa,
3433                                                         NETLINK_CB(cb->skb).pid,
3434                                                         cb->nlh->nlmsg_seq,
3435                                                         RTM_NEWADDR,
3436                                                         NLM_F_MULTI);
3437                         }
3438                         break;
3439                 case MULTICAST_ADDR:
3440                         /* multicast address */
3441                         for (ifmca = idev->mc_list; ifmca;
3442                              ifmca = ifmca->next, ip_idx++) {
3443                                 if (ip_idx < s_ip_idx)
3444                                         continue;
3445                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3446                                                           NETLINK_CB(cb->skb).pid,
3447                                                           cb->nlh->nlmsg_seq,
3448                                                           RTM_GETMULTICAST,
3449                                                           NLM_F_MULTI);
3450                         }
3451                         break;
3452                 case ANYCAST_ADDR:
3453                         /* anycast address */
3454                         for (ifaca = idev->ac_list; ifaca;
3455                              ifaca = ifaca->aca_next, ip_idx++) {
3456                                 if (ip_idx < s_ip_idx)
3457                                         continue;
3458                                 err = inet6_fill_ifacaddr(skb, ifaca,
3459                                                           NETLINK_CB(cb->skb).pid,
3460                                                           cb->nlh->nlmsg_seq,
3461                                                           RTM_GETANYCAST,
3462                                                           NLM_F_MULTI);
3463                         }
3464                         break;
3465                 default:
3466                         break;
3467                 }
3468                 read_unlock_bh(&idev->lock);
3469                 in6_dev_put(idev);
3470
3471                 if (err <= 0)
3472                         break;
3473 cont:
3474                 idx++;
3475         }
3476         cb->args[0] = idx;
3477         cb->args[1] = ip_idx;
3478         return skb->len;
3479 }
3480
3481 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3482 {
3483         enum addr_type_t type = UNICAST_ADDR;
3484
3485         return inet6_dump_addr(skb, cb, type);
3486 }
3487
3488 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3489 {
3490         enum addr_type_t type = MULTICAST_ADDR;
3491
3492         return inet6_dump_addr(skb, cb, type);
3493 }
3494
3495
3496 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3497 {
3498         enum addr_type_t type = ANYCAST_ADDR;
3499
3500         return inet6_dump_addr(skb, cb, type);
3501 }
3502
3503 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3504                              void *arg)
3505 {
3506         struct net *net = sock_net(in_skb->sk);
3507         struct ifaddrmsg *ifm;
3508         struct nlattr *tb[IFA_MAX+1];
3509         struct in6_addr *addr = NULL;
3510         struct net_device *dev = NULL;
3511         struct inet6_ifaddr *ifa;
3512         struct sk_buff *skb;
3513         int err;
3514
3515         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3516         if (err < 0)
3517                 goto errout;
3518
3519         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3520         if (addr == NULL) {
3521                 err = -EINVAL;
3522                 goto errout;
3523         }
3524
3525         ifm = nlmsg_data(nlh);
3526         if (ifm->ifa_index)
3527                 dev = __dev_get_by_index(net, ifm->ifa_index);
3528
3529         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3530                 err = -EADDRNOTAVAIL;
3531                 goto errout;
3532         }
3533
3534         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3535                 err = -ENOBUFS;
3536                 goto errout_ifa;
3537         }
3538
3539         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3540                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3541         if (err < 0) {
3542                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3543                 WARN_ON(err == -EMSGSIZE);
3544                 kfree_skb(skb);
3545                 goto errout_ifa;
3546         }
3547         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3548 errout_ifa:
3549         in6_ifa_put(ifa);
3550 errout:
3551         return err;
3552 }
3553
3554 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3555 {
3556         struct sk_buff *skb;
3557         struct net *net = dev_net(ifa->idev->dev);
3558         int err = -ENOBUFS;
3559
3560         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3561         if (skb == NULL)
3562                 goto errout;
3563
3564         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3565         if (err < 0) {
3566                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3567                 WARN_ON(err == -EMSGSIZE);
3568                 kfree_skb(skb);
3569                 goto errout;
3570         }
3571         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3572 errout:
3573         if (err < 0)
3574                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3575 }
3576
3577 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3578                                 __s32 *array, int bytes)
3579 {
3580         BUG_ON(bytes < (DEVCONF_MAX * 4));
3581
3582         memset(array, 0, bytes);
3583         array[DEVCONF_FORWARDING] = cnf->forwarding;
3584         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3585         array[DEVCONF_MTU6] = cnf->mtu6;
3586         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3587         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3588         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3589         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3590         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3591         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3592         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3593         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3594 #ifdef CONFIG_IPV6_PRIVACY
3595         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3596         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3597         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3598         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3599         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3600 #endif
3601         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3602         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3603         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3604 #ifdef CONFIG_IPV6_ROUTER_PREF
3605         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3606         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3607 #ifdef CONFIG_IPV6_ROUTE_INFO
3608         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3609 #endif
3610 #endif
3611         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3612         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3613 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3614         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3615 #endif
3616 #ifdef CONFIG_IPV6_MROUTE
3617         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3618 #endif
3619 }
3620
3621 static inline size_t inet6_if_nlmsg_size(void)
3622 {
3623         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3624                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3625                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3626                + nla_total_size(4) /* IFLA_MTU */
3627                + nla_total_size(4) /* IFLA_LINK */
3628                + nla_total_size( /* IFLA_PROTINFO */
3629                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3630                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3631                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3632                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3633                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3634                  );
3635 }
3636
3637 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3638                                       int bytes)
3639 {
3640         int i;
3641         int pad = bytes - sizeof(u64) * items;
3642         BUG_ON(pad < 0);
3643
3644         /* Use put_unaligned() because stats may not be aligned for u64. */
3645         put_unaligned(items, &stats[0]);
3646         for (i = 1; i < items; i++)
3647                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3648
3649         memset(&stats[items], 0, pad);
3650 }
3651
3652 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3653                              int bytes)
3654 {
3655         switch(attrtype) {
3656         case IFLA_INET6_STATS:
3657                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3658                 break;
3659         case IFLA_INET6_ICMP6STATS:
3660                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3661                 break;
3662         }
3663 }
3664
3665 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3666                              u32 pid, u32 seq, int event, unsigned int flags)
3667 {
3668         struct net_device *dev = idev->dev;
3669         struct nlattr *nla;
3670         struct ifinfomsg *hdr;
3671         struct nlmsghdr *nlh;
3672         void *protoinfo;
3673         struct ifla_cacheinfo ci;
3674
3675         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3676         if (nlh == NULL)
3677                 return -EMSGSIZE;
3678
3679         hdr = nlmsg_data(nlh);
3680         hdr->ifi_family = AF_INET6;
3681         hdr->__ifi_pad = 0;
3682         hdr->ifi_type = dev->type;
3683         hdr->ifi_index = dev->ifindex;
3684         hdr->ifi_flags = dev_get_flags(dev);
3685         hdr->ifi_change = 0;
3686
3687         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3688
3689         if (dev->addr_len)
3690                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3691
3692         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3693         if (dev->ifindex != dev->iflink)
3694                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3695
3696         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3697         if (protoinfo == NULL)
3698                 goto nla_put_failure;
3699
3700         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3701
3702         ci.max_reasm_len = IPV6_MAXPLEN;
3703         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3704                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3705         ci.reachable_time = idev->nd_parms->reachable_time;
3706         ci.retrans_time = idev->nd_parms->retrans_time;
3707         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3708
3709         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3710         if (nla == NULL)
3711                 goto nla_put_failure;
3712         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3713
3714         /* XXX - MC not implemented */
3715
3716         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3717         if (nla == NULL)
3718                 goto nla_put_failure;
3719         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3720
3721         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3722         if (nla == NULL)
3723                 goto nla_put_failure;
3724         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3725
3726         nla_nest_end(skb, protoinfo);
3727         return nlmsg_end(skb, nlh);
3728
3729 nla_put_failure:
3730         nlmsg_cancel(skb, nlh);
3731         return -EMSGSIZE;
3732 }
3733
3734 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3735 {
3736         struct net *net = sock_net(skb->sk);
3737         int idx, err;
3738         int s_idx = cb->args[0];
3739         struct net_device *dev;
3740         struct inet6_dev *idev;
3741
3742         read_lock(&dev_base_lock);
3743         idx = 0;
3744         for_each_netdev(net, dev) {
3745                 if (idx < s_idx)
3746                         goto cont;
3747                 if ((idev = in6_dev_get(dev)) == NULL)
3748                         goto cont;
3749                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3750                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3751                 in6_dev_put(idev);
3752                 if (err <= 0)
3753                         break;
3754 cont:
3755                 idx++;
3756         }
3757         read_unlock(&dev_base_lock);
3758         cb->args[0] = idx;
3759
3760         return skb->len;
3761 }
3762
3763 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3764 {
3765         struct sk_buff *skb;
3766         struct net *net = dev_net(idev->dev);
3767         int err = -ENOBUFS;
3768
3769         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3770         if (skb == NULL)
3771                 goto errout;
3772
3773         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3774         if (err < 0) {
3775                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3776                 WARN_ON(err == -EMSGSIZE);
3777                 kfree_skb(skb);
3778                 goto errout;
3779         }
3780         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3781 errout:
3782         if (err < 0)
3783                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3784 }
3785
3786 static inline size_t inet6_prefix_nlmsg_size(void)
3787 {
3788         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3789                + nla_total_size(sizeof(struct in6_addr))
3790                + nla_total_size(sizeof(struct prefix_cacheinfo));
3791 }
3792
3793 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3794                              struct prefix_info *pinfo, u32 pid, u32 seq,
3795                              int event, unsigned int flags)
3796 {
3797         struct prefixmsg *pmsg;
3798         struct nlmsghdr *nlh;
3799         struct prefix_cacheinfo ci;
3800
3801         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3802         if (nlh == NULL)
3803                 return -EMSGSIZE;
3804
3805         pmsg = nlmsg_data(nlh);
3806         pmsg->prefix_family = AF_INET6;
3807         pmsg->prefix_pad1 = 0;
3808         pmsg->prefix_pad2 = 0;
3809         pmsg->prefix_ifindex = idev->dev->ifindex;
3810         pmsg->prefix_len = pinfo->prefix_len;
3811         pmsg->prefix_type = pinfo->type;
3812         pmsg->prefix_pad3 = 0;
3813         pmsg->prefix_flags = 0;
3814         if (pinfo->onlink)
3815                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3816         if (pinfo->autoconf)
3817                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3818
3819         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3820
3821         ci.preferred_time = ntohl(pinfo->prefered);
3822         ci.valid_time = ntohl(pinfo->valid);
3823         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3824
3825         return nlmsg_end(skb, nlh);
3826
3827 nla_put_failure:
3828         nlmsg_cancel(skb, nlh);
3829         return -EMSGSIZE;
3830 }
3831
3832 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3833                          struct prefix_info *pinfo)
3834 {
3835         struct sk_buff *skb;
3836         struct net *net = dev_net(idev->dev);
3837         int err = -ENOBUFS;
3838
3839         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3840         if (skb == NULL)
3841                 goto errout;
3842
3843         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3844         if (err < 0) {
3845                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3846                 WARN_ON(err == -EMSGSIZE);
3847                 kfree_skb(skb);
3848                 goto errout;
3849         }
3850         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3851 errout:
3852         if (err < 0)
3853                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3854 }
3855
3856 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3857 {
3858         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3859
3860         switch (event) {
3861         case RTM_NEWADDR:
3862                 /*
3863                  * If the address was optimistic
3864                  * we inserted the route at the start of
3865                  * our DAD process, so we don't need
3866                  * to do it again
3867                  */
3868                 if (!(ifp->rt->rt6i_node))
3869                         ip6_ins_rt(ifp->rt);
3870                 if (ifp->idev->cnf.forwarding)
3871                         addrconf_join_anycast(ifp);
3872                 break;
3873         case RTM_DELADDR:
3874                 if (ifp->idev->cnf.forwarding)
3875                         addrconf_leave_anycast(ifp);
3876                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3877                 dst_hold(&ifp->rt->u.dst);
3878                 if (ip6_del_rt(ifp->rt))
3879                         dst_free(&ifp->rt->u.dst);
3880                 break;
3881         }
3882 }
3883
3884 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3885 {
3886         rcu_read_lock_bh();
3887         if (likely(ifp->idev->dead == 0))
3888                 __ipv6_ifa_notify(event, ifp);
3889         rcu_read_unlock_bh();
3890 }
3891
3892 #ifdef CONFIG_SYSCTL
3893
3894 static
3895 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3896                            void __user *buffer, size_t *lenp, loff_t *ppos)
3897 {
3898         int *valp = ctl->data;
3899         int val = *valp;
3900         int ret;
3901
3902         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3903
3904         if (write)
3905                 addrconf_fixup_forwarding(ctl, valp, val);
3906         return ret;
3907 }
3908
3909 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3910                                             int __user *name, int nlen,
3911                                             void __user *oldval,
3912                                             size_t __user *oldlenp,
3913                                             void __user *newval, size_t newlen)
3914 {
3915         int *valp = table->data;
3916         int val = *valp;
3917         int new;
3918
3919         if (!newval || !newlen)
3920                 return 0;
3921         if (newlen != sizeof(int))
3922                 return -EINVAL;
3923         if (get_user(new, (int __user *)newval))
3924                 return -EFAULT;
3925         if (new == *valp)
3926                 return 0;
3927         if (oldval && oldlenp) {
3928                 size_t len;
3929                 if (get_user(len, oldlenp))
3930                         return -EFAULT;
3931                 if (len) {
3932                         if (len > table->maxlen)
3933                                 len = table->maxlen;
3934                         if (copy_to_user(oldval, valp, len))
3935                                 return -EFAULT;
3936                         if (put_user(len, oldlenp))
3937                                 return -EFAULT;
3938                 }
3939         }
3940
3941         *valp = new;
3942         addrconf_fixup_forwarding(table, valp, val);
3943         return 1;
3944 }
3945
3946 static struct addrconf_sysctl_table
3947 {
3948         struct ctl_table_header *sysctl_header;
3949         ctl_table addrconf_vars[DEVCONF_MAX+1];
3950         char *dev_name;
3951 } addrconf_sysctl __read_mostly = {
3952         .sysctl_header = NULL,
3953         .addrconf_vars = {
3954                 {
3955                         .ctl_name       =       NET_IPV6_FORWARDING,
3956                         .procname       =       "forwarding",
3957                         .data           =       &ipv6_devconf.forwarding,
3958                         .maxlen         =       sizeof(int),
3959                         .mode           =       0644,
3960                         .proc_handler   =       &addrconf_sysctl_forward,
3961                         .strategy       =       &addrconf_sysctl_forward_strategy,
3962                 },
3963                 {
3964                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3965                         .procname       =       "hop_limit",
3966                         .data           =       &ipv6_devconf.hop_limit,
3967                         .maxlen         =       sizeof(int),
3968                         .mode           =       0644,
3969                         .proc_handler   =       proc_dointvec,
3970                 },
3971                 {
3972                         .ctl_name       =       NET_IPV6_MTU,
3973                         .procname       =       "mtu",
3974                         .data           =       &ipv6_devconf.mtu6,
3975                         .maxlen         =       sizeof(int),
3976                         .mode           =       0644,
3977                         .proc_handler   =       &proc_dointvec,
3978                 },
3979                 {
3980                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3981                         .procname       =       "accept_ra",
3982                         .data           =       &ipv6_devconf.accept_ra,
3983                         .maxlen         =       sizeof(int),
3984                         .mode           =       0644,
3985                         .proc_handler   =       &proc_dointvec,
3986                 },
3987                 {
3988                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3989                         .procname       =       "accept_redirects",
3990                         .data           =       &ipv6_devconf.accept_redirects,
3991                         .maxlen         =       sizeof(int),
3992                         .mode           =       0644,
3993                         .proc_handler   =       &proc_dointvec,
3994                 },
3995                 {
3996                         .ctl_name       =       NET_IPV6_AUTOCONF,
3997                         .procname       =       "autoconf",
3998                         .data           =       &ipv6_devconf.autoconf,
3999                         .maxlen         =       sizeof(int),
4000                         .mode           =       0644,
4001                         .proc_handler   =       &proc_dointvec,
4002                 },
4003                 {
4004                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4005                         .procname       =       "dad_transmits",
4006                         .data           =       &ipv6_devconf.dad_transmits,
4007                         .maxlen         =       sizeof(int),
4008                         .mode           =       0644,
4009                         .proc_handler   =       &proc_dointvec,
4010                 },
4011                 {
4012                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4013                         .procname       =       "router_solicitations",
4014                         .data           =       &ipv6_devconf.rtr_solicits,
4015                         .maxlen         =       sizeof(int),
4016                         .mode           =       0644,
4017                         .proc_handler   =       &proc_dointvec,
4018                 },
4019                 {
4020                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4021                         .procname       =       "router_solicitation_interval",
4022                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4023                         .maxlen         =       sizeof(int),
4024                         .mode           =       0644,
4025                         .proc_handler   =       &proc_dointvec_jiffies,
4026                         .strategy       =       &sysctl_jiffies,
4027                 },
4028                 {
4029                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4030                         .procname       =       "router_solicitation_delay",
4031                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4032                         .maxlen         =       sizeof(int),
4033                         .mode           =       0644,
4034                         .proc_handler   =       &proc_dointvec_jiffies,
4035                         .strategy       =       &sysctl_jiffies,
4036                 },
4037                 {
4038                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4039                         .procname       =       "force_mld_version",
4040                         .data           =       &ipv6_devconf.force_mld_version,
4041                         .maxlen         =       sizeof(int),
4042                         .mode           =       0644,
4043                         .proc_handler   =       &proc_dointvec,
4044                 },
4045 #ifdef CONFIG_IPV6_PRIVACY
4046                 {
4047                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4048                         .procname       =       "use_tempaddr",
4049                         .data           =       &ipv6_devconf.use_tempaddr,
4050                         .maxlen         =       sizeof(int),
4051                         .mode           =       0644,
4052                         .proc_handler   =       &proc_dointvec,
4053                 },
4054                 {
4055                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4056                         .procname       =       "temp_valid_lft",
4057                         .data           =       &ipv6_devconf.temp_valid_lft,
4058                         .maxlen         =       sizeof(int),
4059                         .mode           =       0644,
4060                         .proc_handler   =       &proc_dointvec,
4061                 },
4062                 {
4063                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4064                         .procname       =       "temp_prefered_lft",
4065                         .data           =       &ipv6_devconf.temp_prefered_lft,
4066                         .maxlen         =       sizeof(int),
4067                         .mode           =       0644,
4068                         .proc_handler   =       &proc_dointvec,
4069                 },
4070                 {
4071                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4072                         .procname       =       "regen_max_retry",
4073                         .data           =       &ipv6_devconf.regen_max_retry,
4074                         .maxlen         =       sizeof(int),
4075                         .mode           =       0644,
4076                         .proc_handler   =       &proc_dointvec,
4077                 },
4078                 {
4079                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4080                         .procname       =       "max_desync_factor",
4081                         .data           =       &ipv6_devconf.max_desync_factor,
4082                         .maxlen         =       sizeof(int),
4083                         .mode           =       0644,
4084                         .proc_handler   =       &proc_dointvec,
4085                 },
4086 #endif
4087                 {
4088                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4089                         .procname       =       "max_addresses",
4090                         .data           =       &ipv6_devconf.max_addresses,
4091                         .maxlen         =       sizeof(int),
4092                         .mode           =       0644,
4093                         .proc_handler   =       &proc_dointvec,
4094                 },
4095                 {
4096                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4097                         .procname       =       "accept_ra_defrtr",
4098                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4099                         .maxlen         =       sizeof(int),
4100                         .mode           =       0644,
4101                         .proc_handler   =       &proc_dointvec,
4102                 },
4103                 {
4104                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4105                         .procname       =       "accept_ra_pinfo",
4106                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4107                         .maxlen         =       sizeof(int),
4108                         .mode           =       0644,
4109                         .proc_handler   =       &proc_dointvec,
4110                 },
4111 #ifdef CONFIG_IPV6_ROUTER_PREF
4112                 {
4113                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4114                         .procname       =       "accept_ra_rtr_pref",
4115                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4116                         .maxlen         =       sizeof(int),
4117                         .mode           =       0644,
4118                         .proc_handler   =       &proc_dointvec,
4119                 },
4120                 {
4121                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4122                         .procname       =       "router_probe_interval",
4123                         .data           =       &ipv6_devconf.rtr_probe_interval,
4124                         .maxlen         =       sizeof(int),
4125                         .mode           =       0644,
4126                         .proc_handler   =       &proc_dointvec_jiffies,
4127                         .strategy       =       &sysctl_jiffies,
4128                 },
4129 #ifdef CONFIG_IPV6_ROUTE_INFO
4130                 {
4131                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4132                         .procname       =       "accept_ra_rt_info_max_plen",
4133                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4134                         .maxlen         =       sizeof(int),
4135                         .mode           =       0644,
4136                         .proc_handler   =       &proc_dointvec,
4137                 },
4138 #endif
4139 #endif
4140                 {
4141                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4142                         .procname       =       "proxy_ndp",
4143                         .data           =       &ipv6_devconf.proxy_ndp,
4144                         .maxlen         =       sizeof(int),
4145                         .mode           =       0644,
4146                         .proc_handler   =       &proc_dointvec,
4147                 },
4148                 {
4149                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4150                         .procname       =       "accept_source_route",
4151                         .data           =       &ipv6_devconf.accept_source_route,
4152                         .maxlen         =       sizeof(int),
4153                         .mode           =       0644,
4154                         .proc_handler   =       &proc_dointvec,
4155                 },
4156 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4157                 {
4158                         .ctl_name       =       CTL_UNNUMBERED,
4159                         .procname       =       "optimistic_dad",
4160                         .data           =       &ipv6_devconf.optimistic_dad,
4161                         .maxlen         =       sizeof(int),
4162                         .mode           =       0644,
4163                         .proc_handler   =       &proc_dointvec,
4164
4165                 },
4166 #endif
4167 #ifdef CONFIG_IPV6_MROUTE
4168                 {
4169                         .ctl_name       =       CTL_UNNUMBERED,
4170                         .procname       =       "mc_forwarding",
4171                         .data           =       &ipv6_devconf.mc_forwarding,
4172                         .maxlen         =       sizeof(int),
4173                         .mode           =       0644,
4174                         .proc_handler   =       &proc_dointvec,
4175                 },
4176 #endif
4177                 {
4178                         .ctl_name       =       0,      /* sentinel */
4179                 }
4180         },
4181 };
4182
4183 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4184                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4185 {
4186         int i;
4187         struct addrconf_sysctl_table *t;
4188
4189 #define ADDRCONF_CTL_PATH_DEV   3
4190
4191         struct ctl_path addrconf_ctl_path[] = {
4192                 { .procname = "net", .ctl_name = CTL_NET, },
4193                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4194                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4195                 { /* to be set */ },
4196                 { },
4197         };
4198
4199
4200         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4201         if (t == NULL)
4202                 goto out;
4203
4204         for (i=0; t->addrconf_vars[i].data; i++) {
4205                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4206                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4207                 t->addrconf_vars[i].extra2 = net;
4208         }
4209
4210         /*
4211          * Make a copy of dev_name, because '.procname' is regarded as const
4212          * by sysctl and we wouldn't want anyone to change it under our feet
4213          * (see SIOCSIFNAME).
4214          */
4215         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4216         if (!t->dev_name)
4217                 goto free;
4218
4219         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4220         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4221
4222         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4223                         t->addrconf_vars);
4224         if (t->sysctl_header == NULL)
4225                 goto free_procname;
4226
4227         p->sysctl = t;
4228         return 0;
4229
4230 free_procname:
4231         kfree(t->dev_name);
4232 free:
4233         kfree(t);
4234 out:
4235         return -ENOBUFS;
4236 }
4237
4238 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4239 {
4240         struct addrconf_sysctl_table *t;
4241
4242         if (p->sysctl == NULL)
4243                 return;
4244
4245         t = p->sysctl;
4246         p->sysctl = NULL;
4247         unregister_sysctl_table(t->sysctl_header);
4248         kfree(t->dev_name);
4249         kfree(t);
4250 }
4251
4252 static void addrconf_sysctl_register(struct inet6_dev *idev)
4253 {
4254         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4255                               NET_IPV6_NEIGH, "ipv6",
4256                               &ndisc_ifinfo_sysctl_change,
4257                               NULL);
4258         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4259                         idev->dev->ifindex, idev, &idev->cnf);
4260 }
4261
4262 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4263 {
4264         __addrconf_sysctl_unregister(&idev->cnf);
4265         neigh_sysctl_unregister(idev->nd_parms);
4266 }
4267
4268
4269 #endif
4270
4271 static int addrconf_init_net(struct net *net)
4272 {
4273         int err;
4274         struct ipv6_devconf *all, *dflt;
4275
4276         err = -ENOMEM;
4277         all = &ipv6_devconf;
4278         dflt = &ipv6_devconf_dflt;
4279
4280         if (net != &init_net) {
4281                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4282                 if (all == NULL)
4283                         goto err_alloc_all;
4284
4285                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4286                 if (dflt == NULL)
4287                         goto err_alloc_dflt;
4288         }
4289
4290         net->ipv6.devconf_all = all;
4291         net->ipv6.devconf_dflt = dflt;
4292
4293 #ifdef CONFIG_SYSCTL
4294         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4295                         NULL, all);
4296         if (err < 0)
4297                 goto err_reg_all;
4298
4299         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4300                         NULL, dflt);
4301         if (err < 0)
4302                 goto err_reg_dflt;
4303 #endif
4304         return 0;
4305
4306 #ifdef CONFIG_SYSCTL
4307 err_reg_dflt:
4308         __addrconf_sysctl_unregister(all);
4309 err_reg_all:
4310         kfree(dflt);
4311 #endif
4312 err_alloc_dflt:
4313         kfree(all);
4314 err_alloc_all:
4315         return err;
4316 }
4317
4318 static void addrconf_exit_net(struct net *net)
4319 {
4320 #ifdef CONFIG_SYSCTL
4321         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4322         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4323 #endif
4324         if (net != &init_net) {
4325                 kfree(net->ipv6.devconf_dflt);
4326                 kfree(net->ipv6.devconf_all);
4327         }
4328 }
4329
4330 static struct pernet_operations addrconf_ops = {
4331         .init = addrconf_init_net,
4332         .exit = addrconf_exit_net,
4333 };
4334
4335 /*
4336  *      Device notifier
4337  */
4338
4339 int register_inet6addr_notifier(struct notifier_block *nb)
4340 {
4341         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4342 }
4343
4344 EXPORT_SYMBOL(register_inet6addr_notifier);
4345
4346 int unregister_inet6addr_notifier(struct notifier_block *nb)
4347 {
4348         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4349 }
4350
4351 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4352
4353
4354 static int addrconf_net_init(struct net *net)
4355 {
4356         return 0;
4357 }
4358
4359 static void addrconf_net_exit(struct net *net)
4360 {
4361         struct net_device *dev;
4362
4363         rtnl_lock();
4364         /* clean dev list */
4365         for_each_netdev(net, dev) {
4366                 if (__in6_dev_get(dev) == NULL)
4367                         continue;
4368                 addrconf_ifdown(dev, 1);
4369         }
4370         addrconf_ifdown(net->loopback_dev, 2);
4371         rtnl_unlock();
4372 }
4373
4374 static struct pernet_operations addrconf_net_ops = {
4375         .init = addrconf_net_init,
4376         .exit = addrconf_net_exit,
4377 };
4378
4379 /*
4380  *      Init / cleanup code
4381  */
4382
4383 int __init addrconf_init(void)
4384 {
4385         int err;
4386
4387         if ((err = ipv6_addr_label_init()) < 0) {
4388                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4389                         err);
4390                 return err;
4391         }
4392
4393         register_pernet_subsys(&addrconf_ops);
4394
4395         /* The addrconf netdev notifier requires that loopback_dev
4396          * has it's ipv6 private information allocated and setup
4397          * before it can bring up and give link-local addresses
4398          * to other devices which are up.
4399          *
4400          * Unfortunately, loopback_dev is not necessarily the first
4401          * entry in the global dev_base list of net devices.  In fact,
4402          * it is likely to be the very last entry on that list.
4403          * So this causes the notifier registry below to try and
4404          * give link-local addresses to all devices besides loopback_dev
4405          * first, then loopback_dev, which cases all the non-loopback_dev
4406          * devices to fail to get a link-local address.
4407          *
4408          * So, as a temporary fix, allocate the ipv6 structure for
4409          * loopback_dev first by hand.
4410          * Longer term, all of the dependencies ipv6 has upon the loopback
4411          * device and it being up should be removed.
4412          */
4413         rtnl_lock();
4414         if (!ipv6_add_dev(init_net.loopback_dev))
4415                 err = -ENOMEM;
4416         rtnl_unlock();
4417         if (err)
4418                 goto errlo;
4419
4420         err = register_pernet_device(&addrconf_net_ops);
4421         if (err)
4422                 return err;
4423
4424         register_netdevice_notifier(&ipv6_dev_notf);
4425
4426         addrconf_verify(0);
4427
4428         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4429         if (err < 0)
4430                 goto errout;
4431
4432         /* Only the first call to __rtnl_register can fail */
4433         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4434         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4435         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4436         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4437         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4438
4439         ipv6_addr_label_rtnl_register();
4440
4441         return 0;
4442 errout:
4443         unregister_netdevice_notifier(&ipv6_dev_notf);
4444 errlo:
4445         unregister_pernet_subsys(&addrconf_ops);
4446
4447         return err;
4448 }
4449
4450 void addrconf_cleanup(void)
4451 {
4452         struct inet6_ifaddr *ifa;
4453         int i;
4454
4455         unregister_netdevice_notifier(&ipv6_dev_notf);
4456         unregister_pernet_device(&addrconf_net_ops);
4457
4458         unregister_pernet_subsys(&addrconf_ops);
4459
4460         rtnl_lock();
4461
4462         /*
4463          *      Check hash table.
4464          */
4465         write_lock_bh(&addrconf_hash_lock);
4466         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4467                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4468                         struct inet6_ifaddr *bifa;
4469
4470                         bifa = ifa;
4471                         ifa = ifa->lst_next;
4472                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4473                         /* Do not free it; something is wrong.
4474                            Now we can investigate it with debugger.
4475                          */
4476                 }
4477         }
4478         write_unlock_bh(&addrconf_hash_lock);
4479
4480         del_timer(&addr_chk_timer);
4481         rtnl_unlock();
4482
4483         unregister_pernet_subsys(&addrconf_net_ops);
4484 }