net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      $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  *      <chexum@bankinf.banki.hu>
  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 <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
 104 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
 105 #endif
 106
 107 #ifdef CONFIG_IPV6_PRIVACY
 108 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 109 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 110 static void ipv6_regen_rndid(unsigned long data);
 111
 112 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 113 #endif
 114
 115 static int ipv6_count_addresses(struct inet6_dev *idev);
 116
 117 /*
 118  *      Configured unicast address hash table
 119  */
 120 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 121 static DEFINE_RWLOCK(addrconf_hash_lock);
 122
 123 static void addrconf_verify(unsigned long);
 124
 125 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 126 static DEFINE_SPINLOCK(addrconf_verify_lock);
 127
 128 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 129 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 130
 131 static int addrconf_ifdown(struct net_device *dev, int how);
 132
 133 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 134 static void addrconf_dad_timer(unsigned long data);
 135 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 136 static void addrconf_dad_run(struct inet6_dev *idev);
 137 static void addrconf_rs_timer(unsigned long data);
 138 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 139 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 140
 141 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 142                                 struct prefix_info *pinfo);
 143 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
 144
 145 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 146
 147 struct ipv6_devconf ipv6_devconf __read_mostly = {
 148         .forwarding             = 0,
 149         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 150         .mtu6                   = IPV6_MIN_MTU,
 151         .accept_ra              = 1,
 152         .accept_redirects       = 1,
 153         .autoconf               = 1,
 154         .force_mld_version      = 0,
 155         .dad_transmits          = 1,
 156         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 157         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 158         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 159 #ifdef CONFIG_IPV6_PRIVACY
 160         .use_tempaddr           = 0,
 161         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 162         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 163         .regen_max_retry        = REGEN_MAX_RETRY,
 164         .max_desync_factor      = MAX_DESYNC_FACTOR,
 165 #endif
 166         .max_addresses          = IPV6_MAX_ADDRESSES,
 167         .accept_ra_defrtr       = 1,
 168         .accept_ra_pinfo        = 1,
 169 #ifdef CONFIG_IPV6_ROUTER_PREF
 170         .accept_ra_rtr_pref     = 1,
 171         .rtr_probe_interval     = 60 * HZ,
 172 #ifdef CONFIG_IPV6_ROUTE_INFO
 173         .accept_ra_rt_info_max_plen = 0,
 174 #endif
 175 #endif
 176         .proxy_ndp              = 0,
 177         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 178 };
 179
 180 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 181         .forwarding             = 0,
 182         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 183         .mtu6                   = IPV6_MIN_MTU,
 184         .accept_ra              = 1,
 185         .accept_redirects       = 1,
 186         .autoconf               = 1,
 187         .dad_transmits          = 1,
 188         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 189         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 190         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 191 #ifdef CONFIG_IPV6_PRIVACY
 192         .use_tempaddr           = 0,
 193         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 194         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 195         .regen_max_retry        = REGEN_MAX_RETRY,
 196         .max_desync_factor      = MAX_DESYNC_FACTOR,
 197 #endif
 198         .max_addresses          = IPV6_MAX_ADDRESSES,
 199         .accept_ra_defrtr       = 1,
 200         .accept_ra_pinfo        = 1,
 201 #ifdef CONFIG_IPV6_ROUTER_PREF
 202         .accept_ra_rtr_pref     = 1,
 203         .rtr_probe_interval     = 60 * HZ,
 204 #ifdef CONFIG_IPV6_ROUTE_INFO
 205         .accept_ra_rt_info_max_plen = 0,
 206 #endif
 207 #endif
 208         .proxy_ndp              = 0,
 209         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 210 };
 211
 212 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 213 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 214 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 215
 216 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 217 {
 218         if (del_timer(&ifp->timer))
 219                 __in6_ifa_put(ifp);
 220 }
 221
 222 enum addrconf_timer_t
 223 {
 224         AC_NONE,
 225         AC_DAD,
 226         AC_RS,
 227 };
 228
 229 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 230                                enum addrconf_timer_t what,
 231                                unsigned long when)
 232 {
 233         if (!del_timer(&ifp->timer))
 234                 in6_ifa_hold(ifp);
 235
 236         switch (what) {
 237         case AC_DAD:
 238                 ifp->timer.function = addrconf_dad_timer;
 239                 break;
 240         case AC_RS:
 241                 ifp->timer.function = addrconf_rs_timer;
 242                 break;
 243         default:;
 244         }
 245         ifp->timer.expires = jiffies + when;
 246         add_timer(&ifp->timer);
 247 }
 248
 249 static int snmp6_alloc_dev(struct inet6_dev *idev)
 250 {
 251         int err = -ENOMEM;
 252
 253         if (!idev || !idev->dev)
 254                 return -EINVAL;
 255
 256         if (snmp_mib_init((void **)idev->stats.ipv6,
 257                           sizeof(struct ipstats_mib),
 258                           __alignof__(struct ipstats_mib)) < 0)
 259                 goto err_ip;
 260         if (snmp_mib_init((void **)idev->stats.icmpv6,
 261                           sizeof(struct icmpv6_mib),
 262                           __alignof__(struct icmpv6_mib)) < 0)
 263                 goto err_icmp;
 264         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 265                           sizeof(struct icmpv6msg_mib),
 266                           __alignof__(struct icmpv6msg_mib)) < 0)
 267                 goto err_icmpmsg;
 268
 269         return 0;
 270
 271 err_icmpmsg:
 272         snmp_mib_free((void **)idev->stats.icmpv6);
 273 err_icmp:
 274         snmp_mib_free((void **)idev->stats.ipv6);
 275 err_ip:
 276         return err;
 277 }
 278
 279 static int snmp6_free_dev(struct inet6_dev *idev)
 280 {
 281         snmp_mib_free((void **)idev->stats.icmpv6msg);
 282         snmp_mib_free((void **)idev->stats.icmpv6);
 283         snmp_mib_free((void **)idev->stats.ipv6);
 284         return 0;
 285 }
 286
 287 /* Nobody refers to this device, we may destroy it. */
 288
 289 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 290 {
 291         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 292         kfree(idev);
 293 }
 294
 295 void in6_dev_finish_destroy(struct inet6_dev *idev)
 296 {
 297         struct net_device *dev = idev->dev;
 298         BUG_TRAP(idev->addr_list==NULL);
 299         BUG_TRAP(idev->mc_list==NULL);
 300 #ifdef NET_REFCNT_DEBUG
 301         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 302 #endif
 303         dev_put(dev);
 304         if (!idev->dead) {
 305                 printk("Freeing alive inet6 device %p\n", idev);
 306                 return;
 307         }
 308         snmp6_free_dev(idev);
 309         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 310 }
 311
 312 EXPORT_SYMBOL(in6_dev_finish_destroy);
 313
 314 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 315 {
 316         struct inet6_dev *ndev;
 317         struct in6_addr maddr;
 318
 319         ASSERT_RTNL();
 320
 321         if (dev->mtu < IPV6_MIN_MTU)
 322                 return NULL;
 323
 324         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 325
 326         if (ndev == NULL)
 327                 return NULL;
 328
 329         rwlock_init(&ndev->lock);
 330         ndev->dev = dev;
 331         memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
 332         ndev->cnf.mtu6 = dev->mtu;
 333         ndev->cnf.sysctl = NULL;
 334         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 335         if (ndev->nd_parms == NULL) {
 336                 kfree(ndev);
 337                 return NULL;
 338         }
 339         /* We refer to the device */
 340         dev_hold(dev);
 341
 342         if (snmp6_alloc_dev(ndev) < 0) {
 343                 ADBG((KERN_WARNING
 344                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 345                         __FUNCTION__, dev->name));
 346                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 347                 ndev->dead = 1;
 348                 in6_dev_finish_destroy(ndev);
 349                 return NULL;
 350         }
 351
 352         if (snmp6_register_dev(ndev) < 0) {
 353                 ADBG((KERN_WARNING
 354                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 355                         __FUNCTION__, dev->name));
 356                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 357                 ndev->dead = 1;
 358                 in6_dev_finish_destroy(ndev);
 359                 return NULL;
 360         }
 361
 362         /* One reference from device.  We must do this before
 363          * we invoke __ipv6_regen_rndid().
 364          */
 365         in6_dev_hold(ndev);
 366
 367 #ifdef CONFIG_IPV6_PRIVACY
 368         init_timer(&ndev->regen_timer);
 369         ndev->regen_timer.function = ipv6_regen_rndid;
 370         ndev->regen_timer.data = (unsigned long) ndev;
 371         if ((dev->flags&IFF_LOOPBACK) ||
 372             dev->type == ARPHRD_TUNNEL ||
 373 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 374             dev->type == ARPHRD_SIT ||
 375 #endif
 376             dev->type == ARPHRD_NONE) {
 377                 printk(KERN_INFO
 378                        "%s: Disabled Privacy Extensions\n",
 379                        dev->name);
 380                 ndev->cnf.use_tempaddr = -1;
 381         } else {
 382                 in6_dev_hold(ndev);
 383                 ipv6_regen_rndid((unsigned long) ndev);
 384         }
 385 #endif
 386
 387         if (netif_running(dev) && netif_carrier_ok(dev))
 388                 ndev->if_flags |= IF_READY;
 389
 390         ipv6_mc_init_dev(ndev);
 391         ndev->tstamp = jiffies;
 392 #ifdef CONFIG_SYSCTL
 393         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
 394                               NET_IPV6_NEIGH, "ipv6",
 395                               &ndisc_ifinfo_sysctl_change,
 396                               NULL);
 397         addrconf_sysctl_register(ndev, &ndev->cnf);
 398 #endif
 399         /* protected by rtnl_lock */
 400         rcu_assign_pointer(dev->ip6_ptr, ndev);
 401
 402         /* Join all-node multicast group */
 403         ipv6_addr_all_nodes(&maddr);
 404         ipv6_dev_mc_inc(dev, &maddr);
 405
 406         return ndev;
 407 }
 408
 409 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 410 {
 411         struct inet6_dev *idev;
 412
 413         ASSERT_RTNL();
 414
 415         if ((idev = __in6_dev_get(dev)) == NULL) {
 416                 if ((idev = ipv6_add_dev(dev)) == NULL)
 417                         return NULL;
 418         }
 419
 420         if (dev->flags&IFF_UP)
 421                 ipv6_mc_up(idev);
 422         return idev;
 423 }
 424
 425 #ifdef CONFIG_SYSCTL
 426 static void dev_forward_change(struct inet6_dev *idev)
 427 {
 428         struct net_device *dev;
 429         struct inet6_ifaddr *ifa;
 430         struct in6_addr addr;
 431
 432         if (!idev)
 433                 return;
 434         dev = idev->dev;
 435         if (dev && (dev->flags & IFF_MULTICAST)) {
 436                 ipv6_addr_all_routers(&addr);
 437
 438                 if (idev->cnf.forwarding)
 439                         ipv6_dev_mc_inc(dev, &addr);
 440                 else
 441                         ipv6_dev_mc_dec(dev, &addr);
 442         }
 443         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 444                 if (ifa->flags&IFA_F_TENTATIVE)
 445                         continue;
 446                 if (idev->cnf.forwarding)
 447                         addrconf_join_anycast(ifa);
 448                 else
 449                         addrconf_leave_anycast(ifa);
 450         }
 451 }
 452
 453
 454 static void addrconf_forward_change(void)
 455 {
 456         struct net_device *dev;
 457         struct inet6_dev *idev;
 458
 459         read_lock(&dev_base_lock);
 460         for_each_netdev(&init_net, dev) {
 461                 rcu_read_lock();
 462                 idev = __in6_dev_get(dev);
 463                 if (idev) {
 464                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
 465                         idev->cnf.forwarding = ipv6_devconf.forwarding;
 466                         if (changed)
 467                                 dev_forward_change(idev);
 468                 }
 469                 rcu_read_unlock();
 470         }
 471         read_unlock(&dev_base_lock);
 472 }
 473 #endif
 474
 475 /* Nobody refers to this ifaddr, destroy it */
 476
 477 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 478 {
 479         BUG_TRAP(ifp->if_next==NULL);
 480         BUG_TRAP(ifp->lst_next==NULL);
 481 #ifdef NET_REFCNT_DEBUG
 482         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 483 #endif
 484
 485         in6_dev_put(ifp->idev);
 486
 487         if (del_timer(&ifp->timer))
 488                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 489
 490         if (!ifp->dead) {
 491                 printk("Freeing alive inet6 address %p\n", ifp);
 492                 return;
 493         }
 494         dst_release(&ifp->rt->u.dst);
 495
 496         kfree(ifp);
 497 }
 498
 499 static void
 500 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 501 {
 502         struct inet6_ifaddr *ifa, **ifap;
 503         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 504
 505         /*
 506          * Each device address list is sorted in order of scope -
 507          * global before linklocal.
 508          */
 509         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 510              ifap = &ifa->if_next) {
 511                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 512                         break;
 513         }
 514
 515         ifp->if_next = *ifap;
 516         *ifap = ifp;
 517 }
 518
 519 /* On success it returns ifp with increased reference count */
 520
 521 static struct inet6_ifaddr *
 522 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 523               int scope, u32 flags)
 524 {
 525         struct inet6_ifaddr *ifa = NULL;
 526         struct rt6_info *rt;
 527         int hash;
 528         int err = 0;
 529
 530         rcu_read_lock_bh();
 531         if (idev->dead) {
 532                 err = -ENODEV;                  /*XXX*/
 533                 goto out2;
 534         }
 535
 536         write_lock(&addrconf_hash_lock);
 537
 538         /* Ignore adding duplicate addresses on an interface */
 539         if (ipv6_chk_same_addr(addr, idev->dev)) {
 540                 ADBG(("ipv6_add_addr: already assigned\n"));
 541                 err = -EEXIST;
 542                 goto out;
 543         }
 544
 545         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 546
 547         if (ifa == NULL) {
 548                 ADBG(("ipv6_add_addr: malloc failed\n"));
 549                 err = -ENOBUFS;
 550                 goto out;
 551         }
 552
 553         rt = addrconf_dst_alloc(idev, addr, 0);
 554         if (IS_ERR(rt)) {
 555                 err = PTR_ERR(rt);
 556                 goto out;
 557         }
 558
 559         ipv6_addr_copy(&ifa->addr, addr);
 560
 561         spin_lock_init(&ifa->lock);
 562         init_timer(&ifa->timer);
 563         ifa->timer.data = (unsigned long) ifa;
 564         ifa->scope = scope;
 565         ifa->prefix_len = pfxlen;
 566         ifa->flags = flags | IFA_F_TENTATIVE;
 567         ifa->cstamp = ifa->tstamp = jiffies;
 568
 569         ifa->rt = rt;
 570
 571         /*
 572          * part one of RFC 4429, section 3.3
 573          * We should not configure an address as
 574          * optimistic if we do not yet know the link
 575          * layer address of our nexhop router
 576          */
 577
 578         if (rt->rt6i_nexthop == NULL)
 579                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 580
 581         ifa->idev = idev;
 582         in6_dev_hold(idev);
 583         /* For caller */
 584         in6_ifa_hold(ifa);
 585
 586         /* Add to big hash table */
 587         hash = ipv6_addr_hash(addr);
 588
 589         ifa->lst_next = inet6_addr_lst[hash];
 590         inet6_addr_lst[hash] = ifa;
 591         in6_ifa_hold(ifa);
 592         write_unlock(&addrconf_hash_lock);
 593
 594         write_lock(&idev->lock);
 595         /* Add to inet6_dev unicast addr list. */
 596         ipv6_link_dev_addr(idev, ifa);
 597
 598 #ifdef CONFIG_IPV6_PRIVACY
 599         if (ifa->flags&IFA_F_TEMPORARY) {
 600                 ifa->tmp_next = idev->tempaddr_list;
 601                 idev->tempaddr_list = ifa;
 602                 in6_ifa_hold(ifa);
 603         }
 604 #endif
 605
 606         in6_ifa_hold(ifa);
 607         write_unlock(&idev->lock);
 608 out2:
 609         rcu_read_unlock_bh();
 610
 611         if (likely(err == 0))
 612                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 613         else {
 614                 kfree(ifa);
 615                 ifa = ERR_PTR(err);
 616         }
 617
 618         return ifa;
 619 out:
 620         write_unlock(&addrconf_hash_lock);
 621         goto out2;
 622 }
 623
 624 /* This function wants to get referenced ifp and releases it before return */
 625
 626 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 627 {
 628         struct inet6_ifaddr *ifa, **ifap;
 629         struct inet6_dev *idev = ifp->idev;
 630         int hash;
 631         int deleted = 0, onlink = 0;
 632         unsigned long expires = jiffies;
 633
 634         hash = ipv6_addr_hash(&ifp->addr);
 635
 636         ifp->dead = 1;
 637
 638         write_lock_bh(&addrconf_hash_lock);
 639         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 640              ifap = &ifa->lst_next) {
 641                 if (ifa == ifp) {
 642                         *ifap = ifa->lst_next;
 643                         __in6_ifa_put(ifp);
 644                         ifa->lst_next = NULL;
 645                         break;
 646                 }
 647         }
 648         write_unlock_bh(&addrconf_hash_lock);
 649
 650         write_lock_bh(&idev->lock);
 651 #ifdef CONFIG_IPV6_PRIVACY
 652         if (ifp->flags&IFA_F_TEMPORARY) {
 653                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 654                      ifap = &ifa->tmp_next) {
 655                         if (ifa == ifp) {
 656                                 *ifap = ifa->tmp_next;
 657                                 if (ifp->ifpub) {
 658                                         in6_ifa_put(ifp->ifpub);
 659                                         ifp->ifpub = NULL;
 660                                 }
 661                                 __in6_ifa_put(ifp);
 662                                 ifa->tmp_next = NULL;
 663                                 break;
 664                         }
 665                 }
 666         }
 667 #endif
 668
 669         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 670                 if (ifa == ifp) {
 671                         *ifap = ifa->if_next;
 672                         __in6_ifa_put(ifp);
 673                         ifa->if_next = NULL;
 674                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 675                                 break;
 676                         deleted = 1;
 677                         continue;
 678                 } else if (ifp->flags & IFA_F_PERMANENT) {
 679                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 680                                               ifp->prefix_len)) {
 681                                 if (ifa->flags & IFA_F_PERMANENT) {
 682                                         onlink = 1;
 683                                         if (deleted)
 684                                                 break;
 685                                 } else {
 686                                         unsigned long lifetime;
 687
 688                                         if (!onlink)
 689                                                 onlink = -1;
 690
 691                                         spin_lock(&ifa->lock);
 692                                         lifetime = min_t(unsigned long,
 693                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 694                                         if (time_before(expires,
 695                                                         ifa->tstamp + lifetime * HZ))
 696                                                 expires = ifa->tstamp + lifetime * HZ;
 697                                         spin_unlock(&ifa->lock);
 698                                 }
 699                         }
 700                 }
 701                 ifap = &ifa->if_next;
 702         }
 703         write_unlock_bh(&idev->lock);
 704
 705         ipv6_ifa_notify(RTM_DELADDR, ifp);
 706
 707         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 708
 709         addrconf_del_timer(ifp);
 710
 711         /*
 712          * Purge or update corresponding prefix
 713          *
 714          * 1) we don't purge prefix here if address was not permanent.
 715          *    prefix is managed by its own lifetime.
 716          * 2) if there're no addresses, delete prefix.
 717          * 3) if there're still other permanent address(es),
 718          *    corresponding prefix is still permanent.
 719          * 4) otherwise, update prefix lifetime to the
 720          *    longest valid lifetime among the corresponding
 721          *    addresses on the device.
 722          *    Note: subsequent RA will update lifetime.
 723          *
 724          * --yoshfuji
 725          */
 726         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 727                 struct in6_addr prefix;
 728                 struct rt6_info *rt;
 729
 730                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 731                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
 732
 733                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 734                         if (onlink == 0) {
 735                                 ip6_del_rt(rt);
 736                                 rt = NULL;
 737                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 738                                 rt->rt6i_expires = expires;
 739                                 rt->rt6i_flags |= RTF_EXPIRES;
 740                         }
 741                 }
 742                 dst_release(&rt->u.dst);
 743         }
 744
 745         in6_ifa_put(ifp);
 746 }
 747
 748 #ifdef CONFIG_IPV6_PRIVACY
 749 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 750 {
 751         struct inet6_dev *idev = ifp->idev;
 752         struct in6_addr addr, *tmpaddr;
 753         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 754         int tmp_plen;
 755         int ret = 0;
 756         int max_addresses;
 757         u32 addr_flags;
 758
 759         write_lock(&idev->lock);
 760         if (ift) {
 761                 spin_lock_bh(&ift->lock);
 762                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 763                 spin_unlock_bh(&ift->lock);
 764                 tmpaddr = &addr;
 765         } else {
 766                 tmpaddr = NULL;
 767         }
 768 retry:
 769         in6_dev_hold(idev);
 770         if (idev->cnf.use_tempaddr <= 0) {
 771                 write_unlock(&idev->lock);
 772                 printk(KERN_INFO
 773                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 774                 in6_dev_put(idev);
 775                 ret = -1;
 776                 goto out;
 777         }
 778         spin_lock_bh(&ifp->lock);
 779         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 780                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 781                 spin_unlock_bh(&ifp->lock);
 782                 write_unlock(&idev->lock);
 783                 printk(KERN_WARNING
 784                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 785                 in6_dev_put(idev);
 786                 ret = -1;
 787                 goto out;
 788         }
 789         in6_ifa_hold(ifp);
 790         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 791         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 792                 spin_unlock_bh(&ifp->lock);
 793                 write_unlock(&idev->lock);
 794                 printk(KERN_WARNING
 795                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 796                 in6_ifa_put(ifp);
 797                 in6_dev_put(idev);
 798                 ret = -1;
 799                 goto out;
 800         }
 801         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 802         tmp_valid_lft = min_t(__u32,
 803                               ifp->valid_lft,
 804                               idev->cnf.temp_valid_lft);
 805         tmp_prefered_lft = min_t(__u32,
 806                                  ifp->prefered_lft,
 807                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 808         tmp_plen = ifp->prefix_len;
 809         max_addresses = idev->cnf.max_addresses;
 810         tmp_cstamp = ifp->cstamp;
 811         tmp_tstamp = ifp->tstamp;
 812         spin_unlock_bh(&ifp->lock);
 813
 814         write_unlock(&idev->lock);
 815
 816         addr_flags = IFA_F_TEMPORARY;
 817         /* set in addrconf_prefix_rcv() */
 818         if (ifp->flags & IFA_F_OPTIMISTIC)
 819                 addr_flags |= IFA_F_OPTIMISTIC;
 820
 821         ift = !max_addresses ||
 822               ipv6_count_addresses(idev) < max_addresses ?
 823                 ipv6_add_addr(idev, &addr, tmp_plen,
 824                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 825                               addr_flags) : NULL;
 826         if (!ift || IS_ERR(ift)) {
 827                 in6_ifa_put(ifp);
 828                 in6_dev_put(idev);
 829                 printk(KERN_INFO
 830                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 831                 tmpaddr = &addr;
 832                 write_lock(&idev->lock);
 833                 goto retry;
 834         }
 835
 836         spin_lock_bh(&ift->lock);
 837         ift->ifpub = ifp;
 838         ift->valid_lft = tmp_valid_lft;
 839         ift->prefered_lft = tmp_prefered_lft;
 840         ift->cstamp = tmp_cstamp;
 841         ift->tstamp = tmp_tstamp;
 842         spin_unlock_bh(&ift->lock);
 843
 844         addrconf_dad_start(ift, 0);
 845         in6_ifa_put(ift);
 846         in6_dev_put(idev);
 847 out:
 848         return ret;
 849 }
 850 #endif
 851
 852 /*
 853  *      Choose an appropriate source address (RFC3484)
 854  */
 855 struct ipv6_saddr_score {
 856         int             addr_type;
 857         unsigned int    attrs;
 858         int             matchlen;
 859         int             scope;
 860         unsigned int    rule;
 861 };
 862
 863 #define IPV6_SADDR_SCORE_LOCAL          0x0001
 864 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
 865 #define IPV6_SADDR_SCORE_HOA            0x0008
 866 #define IPV6_SADDR_SCORE_OIF            0x0010
 867 #define IPV6_SADDR_SCORE_LABEL          0x0020
 868 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
 869
 870 static inline int ipv6_saddr_preferred(int type)
 871 {
 872         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 873                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 874                 return 1;
 875         return 0;
 876 }
 877
 878 /* static matching label */
 879 static inline int ipv6_saddr_label(const struct in6_addr *addr, int type)
 880 {
 881  /*
 882   *     prefix (longest match)  label
 883   *     -----------------------------
 884   *     ::1/128                 0
 885   *     ::/0                    1
 886   *     2002::/16               2
 887   *     ::/96                   3
 888   *     ::ffff:0:0/96           4
 889   *     fc00::/7                5
 890   *     2001::/32               6
 891   */
 892         if (type & IPV6_ADDR_LOOPBACK)
 893                 return 0;
 894         else if (type & IPV6_ADDR_COMPATv4)
 895                 return 3;
 896         else if (type & IPV6_ADDR_MAPPED)
 897                 return 4;
 898         else if (addr->s6_addr32[0] == htonl(0x20010000))
 899                 return 6;
 900         else if (addr->s6_addr16[0] == htons(0x2002))
 901                 return 2;
 902         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
 903                 return 5;
 904         return 1;
 905 }
 906
 907 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
 908                        struct in6_addr *daddr, struct in6_addr *saddr)
 909 {
 910         struct ipv6_saddr_score hiscore;
 911         struct inet6_ifaddr *ifa_result = NULL;
 912         int daddr_type = __ipv6_addr_type(daddr);
 913         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
 914         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
 915         struct net_device *dev;
 916
 917         memset(&hiscore, 0, sizeof(hiscore));
 918
 919         read_lock(&dev_base_lock);
 920         rcu_read_lock();
 921
 922         for_each_netdev(&init_net, dev) {
 923                 struct inet6_dev *idev;
 924                 struct inet6_ifaddr *ifa;
 925
 926                 /* Rule 0: Candidate Source Address (section 4)
 927                  *  - multicast and link-local destination address,
 928                  *    the set of candidate source address MUST only
 929                  *    include addresses assigned to interfaces
 930                  *    belonging to the same link as the outgoing
 931                  *    interface.
 932                  * (- For site-local destination addresses, the
 933                  *    set of candidate source addresses MUST only
 934                  *    include addresses assigned to interfaces
 935                  *    belonging to the same site as the outgoing
 936                  *    interface.)
 937                  */
 938                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
 939                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
 940                     daddr_dev && dev != daddr_dev)
 941                         continue;
 942
 943                 idev = __in6_dev_get(dev);
 944                 if (!idev)
 945                         continue;
 946
 947                 read_lock_bh(&idev->lock);
 948                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
 949                         struct ipv6_saddr_score score;
 950
 951                         score.addr_type = __ipv6_addr_type(&ifa->addr);
 952
 953                         /* Rule 0:
 954                          * - Tentative Address (RFC2462 section 5.4)
 955                          *  - A tentative address is not considered
 956                          *    "assigned to an interface" in the traditional
 957                          *    sense, unless it is also flagged as optimistic.
 958                          * - Candidate Source Address (section 4)
 959                          *  - In any case, anycast addresses, multicast
 960                          *    addresses, and the unspecified address MUST
 961                          *    NOT be included in a candidate set.
 962                          */
 963                         if ((ifa->flags & IFA_F_TENTATIVE) &&
 964                             (!(ifa->flags & IFA_F_OPTIMISTIC)))
 965                                 continue;
 966                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
 967                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
 968                                 LIMIT_NETDEBUG(KERN_DEBUG
 969                                                "ADDRCONF: unspecified / multicast address"
 970                                                "assigned as unicast address on %s",
 971                                                dev->name);
 972                                 continue;
 973                         }
 974
 975                         score.attrs = 0;
 976                         score.matchlen = 0;
 977                         score.scope = 0;
 978                         score.rule = 0;
 979
 980                         if (ifa_result == NULL) {
 981                                 /* record it if the first available entry */
 982                                 goto record_it;
 983                         }
 984
 985                         /* Rule 1: Prefer same address */
 986                         if (hiscore.rule < 1) {
 987                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
 988                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
 989                                 hiscore.rule++;
 990                         }
 991                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
 992                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
 993                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
 994                                         score.rule = 1;
 995                                         goto record_it;
 996                                 }
 997                         } else {
 998                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
 999                                         continue;
1000                         }
1001
1002                         /* Rule 2: Prefer appropriate scope */
1003                         if (hiscore.rule < 2) {
1004                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1005                                 hiscore.rule++;
1006                         }
1007                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1008                         if (hiscore.scope < score.scope) {
1009                                 if (hiscore.scope < daddr_scope) {
1010                                         score.rule = 2;
1011                                         goto record_it;
1012                                 } else
1013                                         continue;
1014                         } else if (score.scope < hiscore.scope) {
1015                                 if (score.scope < daddr_scope)
1016                                         break; /* addresses sorted by scope */
1017                                 else {
1018                                         score.rule = 2;
1019                                         goto record_it;
1020                                 }
1021                         }
1022
1023                         /* Rule 3: Avoid deprecated and optimistic addresses */
1024                         if (hiscore.rule < 3) {
1025                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1026                                    (((ifa_result->flags &
1027                                     (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1028                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1029                                 hiscore.rule++;
1030                         }
1031                         if (ipv6_saddr_preferred(score.addr_type) ||
1032                            (((ifa->flags &
1033                             (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1034                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1035                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1036                                         score.rule = 3;
1037                                         goto record_it;
1038                                 }
1039                         } else {
1040                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1041                                         continue;
1042                         }
1043
1044                         /* Rule 4: Prefer home address */
1045 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1046                         if (hiscore.rule < 4) {
1047                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1048                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1049                                 hiscore.rule++;
1050                         }
1051                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1052                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1053                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1054                                         score.rule = 4;
1055                                         goto record_it;
1056                                 }
1057                         } else {
1058                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1059                                         continue;
1060                         }
1061 #else
1062                         if (hiscore.rule < 4)
1063                                 hiscore.rule++;
1064 #endif
1065
1066                         /* Rule 5: Prefer outgoing interface */
1067                         if (hiscore.rule < 5) {
1068                                 if (daddr_dev == NULL ||
1069                                     daddr_dev == ifa_result->idev->dev)
1070                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1071                                 hiscore.rule++;
1072                         }
1073                         if (daddr_dev == NULL ||
1074                             daddr_dev == ifa->idev->dev) {
1075                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1076                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1077                                         score.rule = 5;
1078                                         goto record_it;
1079                                 }
1080                         } else {
1081                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1082                                         continue;
1083                         }
1084
1085                         /* Rule 6: Prefer matching label */
1086                         if (hiscore.rule < 6) {
1087                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1088                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1089                                 hiscore.rule++;
1090                         }
1091                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1092                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1093                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1094                                         score.rule = 6;
1095                                         goto record_it;
1096                                 }
1097                         } else {
1098                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1099                                         continue;
1100                         }
1101
1102 #ifdef CONFIG_IPV6_PRIVACY
1103                         /* Rule 7: Prefer public address
1104                          * Note: prefer temprary address if use_tempaddr >= 2
1105                          */
1106                         if (hiscore.rule < 7) {
1107                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1108                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1109                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1110                                 hiscore.rule++;
1111                         }
1112                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1113                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1114                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1115                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1116                                         score.rule = 7;
1117                                         goto record_it;
1118                                 }
1119                         } else {
1120                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1121                                         continue;
1122                         }
1123 #else
1124                         if (hiscore.rule < 7)
1125                                 hiscore.rule++;
1126 #endif
1127                         /* Rule 8: Use longest matching prefix */
1128                         if (hiscore.rule < 8) {
1129                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1130                                 hiscore.rule++;
1131                         }
1132                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1133                         if (score.matchlen > hiscore.matchlen) {
1134                                 score.rule = 8;
1135                                 goto record_it;
1136                         }
1137 #if 0
1138                         else if (score.matchlen < hiscore.matchlen)
1139                                 continue;
1140 #endif
1141
1142                         /* Final Rule: choose first available one */
1143                         continue;
1144 record_it:
1145                         if (ifa_result)
1146                                 in6_ifa_put(ifa_result);
1147                         in6_ifa_hold(ifa);
1148                         ifa_result = ifa;
1149                         hiscore = score;
1150                 }
1151                 read_unlock_bh(&idev->lock);
1152         }
1153         rcu_read_unlock();
1154         read_unlock(&dev_base_lock);
1155
1156         if (!ifa_result)
1157                 return -EADDRNOTAVAIL;
1158
1159         ipv6_addr_copy(saddr, &ifa_result->addr);
1160         in6_ifa_put(ifa_result);
1161         return 0;
1162 }
1163
1164
1165 int ipv6_get_saddr(struct dst_entry *dst,
1166                    struct in6_addr *daddr, struct in6_addr *saddr)
1167 {
1168         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1169 }
1170
1171 EXPORT_SYMBOL(ipv6_get_saddr);
1172
1173 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1174                     unsigned char banned_flags)
1175 {
1176         struct inet6_dev *idev;
1177         int err = -EADDRNOTAVAIL;
1178
1179         rcu_read_lock();
1180         if ((idev = __in6_dev_get(dev)) != NULL) {
1181                 struct inet6_ifaddr *ifp;
1182
1183                 read_lock_bh(&idev->lock);
1184                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1185                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1186                                 ipv6_addr_copy(addr, &ifp->addr);
1187                                 err = 0;
1188                                 break;
1189                         }
1190                 }
1191                 read_unlock_bh(&idev->lock);
1192         }
1193         rcu_read_unlock();
1194         return err;
1195 }
1196
1197 static int ipv6_count_addresses(struct inet6_dev *idev)
1198 {
1199         int cnt = 0;
1200         struct inet6_ifaddr *ifp;
1201
1202         read_lock_bh(&idev->lock);
1203         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1204                 cnt++;
1205         read_unlock_bh(&idev->lock);
1206         return cnt;
1207 }
1208
1209 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1210 {
1211         struct inet6_ifaddr * ifp;
1212         u8 hash = ipv6_addr_hash(addr);
1213
1214         read_lock_bh(&addrconf_hash_lock);
1215         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1216                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1217                     !(ifp->flags&IFA_F_TENTATIVE)) {
1218                         if (dev == NULL || ifp->idev->dev == dev ||
1219                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1220                                 break;
1221                 }
1222         }
1223         read_unlock_bh(&addrconf_hash_lock);
1224         return ifp != NULL;
1225 }
1226
1227 EXPORT_SYMBOL(ipv6_chk_addr);
1228
1229 static
1230 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1231 {
1232         struct inet6_ifaddr * ifp;
1233         u8 hash = ipv6_addr_hash(addr);
1234
1235         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1236                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1237                         if (dev == NULL || ifp->idev->dev == dev)
1238                                 break;
1239                 }
1240         }
1241         return ifp != NULL;
1242 }
1243
1244 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1245 {
1246         struct inet6_ifaddr * ifp;
1247         u8 hash = ipv6_addr_hash(addr);
1248
1249         read_lock_bh(&addrconf_hash_lock);
1250         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1251                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1252                         if (dev == NULL || ifp->idev->dev == dev ||
1253                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1254                                 in6_ifa_hold(ifp);
1255                                 break;
1256                         }
1257                 }
1258         }
1259         read_unlock_bh(&addrconf_hash_lock);
1260
1261         return ifp;
1262 }
1263
1264 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1265 {
1266         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1267         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1268         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1269         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1270         int sk_ipv6only = ipv6_only_sock(sk);
1271         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1272         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1273         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1274
1275         if (!sk2_rcv_saddr && !sk_ipv6only)
1276                 return 1;
1277
1278         if (addr_type2 == IPV6_ADDR_ANY &&
1279             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1280                 return 1;
1281
1282         if (addr_type == IPV6_ADDR_ANY &&
1283             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1284                 return 1;
1285
1286         if (sk2_rcv_saddr6 &&
1287             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1288                 return 1;
1289
1290         if (addr_type == IPV6_ADDR_MAPPED &&
1291             !sk2_ipv6only &&
1292             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1293                 return 1;
1294
1295         return 0;
1296 }
1297
1298 /* Gets referenced address, destroys ifaddr */
1299
1300 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1301 {
1302         if (ifp->flags&IFA_F_PERMANENT) {
1303                 spin_lock_bh(&ifp->lock);
1304                 addrconf_del_timer(ifp);
1305                 ifp->flags |= IFA_F_TENTATIVE;
1306                 spin_unlock_bh(&ifp->lock);
1307                 in6_ifa_put(ifp);
1308 #ifdef CONFIG_IPV6_PRIVACY
1309         } else if (ifp->flags&IFA_F_TEMPORARY) {
1310                 struct inet6_ifaddr *ifpub;
1311                 spin_lock_bh(&ifp->lock);
1312                 ifpub = ifp->ifpub;
1313                 if (ifpub) {
1314                         in6_ifa_hold(ifpub);
1315                         spin_unlock_bh(&ifp->lock);
1316                         ipv6_create_tempaddr(ifpub, ifp);
1317                         in6_ifa_put(ifpub);
1318                 } else {
1319                         spin_unlock_bh(&ifp->lock);
1320                 }
1321                 ipv6_del_addr(ifp);
1322 #endif
1323         } else
1324                 ipv6_del_addr(ifp);
1325 }
1326
1327 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1328 {
1329         if (net_ratelimit())
1330                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1331         addrconf_dad_stop(ifp);
1332 }
1333
1334 /* Join to solicited addr multicast group. */
1335
1336 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1337 {
1338         struct in6_addr maddr;
1339
1340         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1341                 return;
1342
1343         addrconf_addr_solict_mult(addr, &maddr);
1344         ipv6_dev_mc_inc(dev, &maddr);
1345 }
1346
1347 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1348 {
1349         struct in6_addr maddr;
1350
1351         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1352                 return;
1353
1354         addrconf_addr_solict_mult(addr, &maddr);
1355         __ipv6_dev_mc_dec(idev, &maddr);
1356 }
1357
1358 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1359 {
1360         struct in6_addr addr;
1361         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1362         if (ipv6_addr_any(&addr))
1363                 return;
1364         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1365 }
1366
1367 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1368 {
1369         struct in6_addr addr;
1370         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1371         if (ipv6_addr_any(&addr))
1372                 return;
1373         __ipv6_dev_ac_dec(ifp->idev, &addr);
1374 }
1375
1376 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1377 {
1378         if (dev->addr_len != ETH_ALEN)
1379                 return -1;
1380         memcpy(eui, dev->dev_addr, 3);
1381         memcpy(eui + 5, dev->dev_addr + 3, 3);
1382
1383         /*
1384          * The zSeries OSA network cards can be shared among various
1385          * OS instances, but the OSA cards have only one MAC address.
1386          * This leads to duplicate address conflicts in conjunction
1387          * with IPv6 if more than one instance uses the same card.
1388          *
1389          * The driver for these cards can deliver a unique 16-bit
1390          * identifier for each instance sharing the same card.  It is
1391          * placed instead of 0xFFFE in the interface identifier.  The
1392          * "u" bit of the interface identifier is not inverted in this
1393          * case.  Hence the resulting interface identifier has local
1394          * scope according to RFC2373.
1395          */
1396         if (dev->dev_id) {
1397                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1398                 eui[4] = dev->dev_id & 0xFF;
1399         } else {
1400                 eui[3] = 0xFF;
1401                 eui[4] = 0xFE;
1402                 eui[0] ^= 2;
1403         }
1404         return 0;
1405 }
1406
1407 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1408 {
1409         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1410         if (dev->addr_len != ARCNET_ALEN)
1411                 return -1;
1412         memset(eui, 0, 7);
1413         eui[7] = *(u8*)dev->dev_addr;
1414         return 0;
1415 }
1416
1417 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1418 {
1419         if (dev->addr_len != INFINIBAND_ALEN)
1420                 return -1;
1421         memcpy(eui, dev->dev_addr + 12, 8);
1422         eui[0] |= 2;
1423         return 0;
1424 }
1425
1426 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1427 {
1428         switch (dev->type) {
1429         case ARPHRD_ETHER:
1430         case ARPHRD_FDDI:
1431         case ARPHRD_IEEE802_TR:
1432                 return addrconf_ifid_eui48(eui, dev);
1433         case ARPHRD_ARCNET:
1434                 return addrconf_ifid_arcnet(eui, dev);
1435         case ARPHRD_INFINIBAND:
1436                 return addrconf_ifid_infiniband(eui, dev);
1437         }
1438         return -1;
1439 }
1440
1441 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1442 {
1443         int err = -1;
1444         struct inet6_ifaddr *ifp;
1445
1446         read_lock_bh(&idev->lock);
1447         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1448                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1449                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1450                         err = 0;
1451                         break;
1452                 }
1453         }
1454         read_unlock_bh(&idev->lock);
1455         return err;
1456 }
1457
1458 #ifdef CONFIG_IPV6_PRIVACY
1459 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1460 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1461 {
1462 regen:
1463         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1464         idev->rndid[0] &= ~0x02;
1465
1466         /*
1467          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1468          * check if generated address is not inappropriate
1469          *
1470          *  - Reserved subnet anycast (RFC 2526)
1471          *      11111101 11....11 1xxxxxxx
1472          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1473          *      00-00-5E-FE-xx-xx-xx-xx
1474          *  - value 0
1475          *  - XXX: already assigned to an address on the device
1476          */
1477         if (idev->rndid[0] == 0xfd &&
1478             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1479             (idev->rndid[7]&0x80))
1480                 goto regen;
1481         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1482                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1483                         goto regen;
1484                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1485                         goto regen;
1486         }
1487
1488         return 0;
1489 }
1490
1491 static void ipv6_regen_rndid(unsigned long data)
1492 {
1493         struct inet6_dev *idev = (struct inet6_dev *) data;
1494         unsigned long expires;
1495
1496         rcu_read_lock_bh();
1497         write_lock_bh(&idev->lock);
1498
1499         if (idev->dead)
1500                 goto out;
1501
1502         if (__ipv6_regen_rndid(idev) < 0)
1503                 goto out;
1504
1505         expires = jiffies +
1506                 idev->cnf.temp_prefered_lft * HZ -
1507                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1508         if (time_before(expires, jiffies)) {
1509                 printk(KERN_WARNING
1510                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1511                         idev->dev->name);
1512                 goto out;
1513         }
1514
1515         if (!mod_timer(&idev->regen_timer, expires))
1516                 in6_dev_hold(idev);
1517
1518 out:
1519         write_unlock_bh(&idev->lock);
1520         rcu_read_unlock_bh();
1521         in6_dev_put(idev);
1522 }
1523
1524 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1525         int ret = 0;
1526
1527         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1528                 ret = __ipv6_regen_rndid(idev);
1529         return ret;
1530 }
1531 #endif
1532
1533 /*
1534  *      Add prefix route.
1535  */
1536
1537 static void
1538 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1539                       unsigned long expires, u32 flags)
1540 {
1541         struct fib6_config cfg = {
1542                 .fc_table = RT6_TABLE_PREFIX,
1543                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1544                 .fc_ifindex = dev->ifindex,
1545                 .fc_expires = expires,
1546                 .fc_dst_len = plen,
1547                 .fc_flags = RTF_UP | flags,
1548         };
1549
1550         ipv6_addr_copy(&cfg.fc_dst, pfx);
1551
1552         /* Prevent useless cloning on PtP SIT.
1553            This thing is done here expecting that the whole
1554            class of non-broadcast devices need not cloning.
1555          */
1556 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1557         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1558                 cfg.fc_flags |= RTF_NONEXTHOP;
1559 #endif
1560
1561         ip6_route_add(&cfg);
1562 }
1563
1564 /* Create "default" multicast route to the interface */
1565
1566 static void addrconf_add_mroute(struct net_device *dev)
1567 {
1568         struct fib6_config cfg = {
1569                 .fc_table = RT6_TABLE_LOCAL,
1570                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1571                 .fc_ifindex = dev->ifindex,
1572                 .fc_dst_len = 8,
1573                 .fc_flags = RTF_UP,
1574         };
1575
1576         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1577
1578         ip6_route_add(&cfg);
1579 }
1580
1581 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1582 static void sit_route_add(struct net_device *dev)
1583 {
1584         struct fib6_config cfg = {
1585                 .fc_table = RT6_TABLE_MAIN,
1586                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1587                 .fc_ifindex = dev->ifindex,
1588                 .fc_dst_len = 96,
1589                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1590         };
1591
1592         /* prefix length - 96 bits "::d.d.d.d" */
1593         ip6_route_add(&cfg);
1594 }
1595 #endif
1596
1597 static void addrconf_add_lroute(struct net_device *dev)
1598 {
1599         struct in6_addr addr;
1600
1601         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1602         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1603 }
1604
1605 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1606 {
1607         struct inet6_dev *idev;
1608
1609         ASSERT_RTNL();
1610
1611         if ((idev = ipv6_find_idev(dev)) == NULL)
1612                 return NULL;
1613
1614         /* Add default multicast route */
1615         addrconf_add_mroute(dev);
1616
1617         /* Add link local route */
1618         addrconf_add_lroute(dev);
1619         return idev;
1620 }
1621
1622 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1623 {
1624         struct prefix_info *pinfo;
1625         __u32 valid_lft;
1626         __u32 prefered_lft;
1627         int addr_type;
1628         unsigned long rt_expires;
1629         struct inet6_dev *in6_dev;
1630
1631         pinfo = (struct prefix_info *) opt;
1632
1633         if (len < sizeof(struct prefix_info)) {
1634                 ADBG(("addrconf: prefix option too short\n"));
1635                 return;
1636         }
1637
1638         /*
1639          *      Validation checks ([ADDRCONF], page 19)
1640          */
1641
1642         addr_type = ipv6_addr_type(&pinfo->prefix);
1643
1644         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1645                 return;
1646
1647         valid_lft = ntohl(pinfo->valid);
1648         prefered_lft = ntohl(pinfo->prefered);
1649
1650         if (prefered_lft > valid_lft) {
1651                 if (net_ratelimit())
1652                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1653                 return;
1654         }
1655
1656         in6_dev = in6_dev_get(dev);
1657
1658         if (in6_dev == NULL) {
1659                 if (net_ratelimit())
1660                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1661                 return;
1662         }
1663
1664         /*
1665          *      Two things going on here:
1666          *      1) Add routes for on-link prefixes
1667          *      2) Configure prefixes with the auto flag set
1668          */
1669
1670         /* Avoid arithmetic overflow. Really, we could
1671            save rt_expires in seconds, likely valid_lft,
1672            but it would require division in fib gc, that it
1673            not good.
1674          */
1675         if (valid_lft >= 0x7FFFFFFF/HZ)
1676                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1677         else
1678                 rt_expires = valid_lft * HZ;
1679
1680         /*
1681          * We convert this (in jiffies) to clock_t later.
1682          * Avoid arithmetic overflow there as well.
1683          * Overflow can happen only if HZ < USER_HZ.
1684          */
1685         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1686                 rt_expires = 0x7FFFFFFF / USER_HZ;
1687
1688         if (pinfo->onlink) {
1689                 struct rt6_info *rt;
1690                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1691
1692                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1693                         if (rt->rt6i_flags&RTF_EXPIRES) {
1694                                 if (valid_lft == 0) {
1695                                         ip6_del_rt(rt);
1696                                         rt = NULL;
1697                                 } else {
1698                                         rt->rt6i_expires = jiffies + rt_expires;
1699                                 }
1700                         }
1701                 } else if (valid_lft) {
1702                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1703                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1704                 }
1705                 if (rt)
1706                         dst_release(&rt->u.dst);
1707         }
1708
1709         /* Try to figure out our local address for this prefix */
1710
1711         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1712                 struct inet6_ifaddr * ifp;
1713                 struct in6_addr addr;
1714                 int create = 0, update_lft = 0;
1715
1716                 if (pinfo->prefix_len == 64) {
1717                         memcpy(&addr, &pinfo->prefix, 8);
1718                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1719                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1720                                 in6_dev_put(in6_dev);
1721                                 return;
1722                         }
1723                         goto ok;
1724                 }
1725                 if (net_ratelimit())
1726                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1727                                pinfo->prefix_len);
1728                 in6_dev_put(in6_dev);
1729                 return;
1730
1731 ok:
1732
1733                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1734
1735                 if (ifp == NULL && valid_lft) {
1736                         int max_addresses = in6_dev->cnf.max_addresses;
1737                         u32 addr_flags = 0;
1738
1739 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1740                         if (in6_dev->cnf.optimistic_dad &&
1741                             !ipv6_devconf.forwarding)
1742                                 addr_flags = IFA_F_OPTIMISTIC;
1743 #endif
1744
1745                         /* Do not allow to create too much of autoconfigured
1746                          * addresses; this would be too easy way to crash kernel.
1747                          */
1748                         if (!max_addresses ||
1749                             ipv6_count_addresses(in6_dev) < max_addresses)
1750                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1751                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1752                                                     addr_flags);
1753
1754                         if (!ifp || IS_ERR(ifp)) {
1755                                 in6_dev_put(in6_dev);
1756                                 return;
1757                         }
1758
1759                         update_lft = create = 1;
1760                         ifp->cstamp = jiffies;
1761                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1762                 }
1763
1764                 if (ifp) {
1765                         int flags;
1766                         unsigned long now;
1767 #ifdef CONFIG_IPV6_PRIVACY
1768                         struct inet6_ifaddr *ift;
1769 #endif
1770                         u32 stored_lft;
1771
1772                         /* update lifetime (RFC2462 5.5.3 e) */
1773                         spin_lock(&ifp->lock);
1774                         now = jiffies;
1775                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1776                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1777                         else
1778                                 stored_lft = 0;
1779                         if (!update_lft && stored_lft) {
1780                                 if (valid_lft > MIN_VALID_LIFETIME ||
1781                                     valid_lft > stored_lft)
1782                                         update_lft = 1;
1783                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1784                                         /* valid_lft <= stored_lft is always true */
1785                                         /* XXX: IPsec */
1786                                         update_lft = 0;
1787                                 } else {
1788                                         valid_lft = MIN_VALID_LIFETIME;
1789                                         if (valid_lft < prefered_lft)
1790                                                 prefered_lft = valid_lft;
1791                                         update_lft = 1;
1792                                 }
1793                         }
1794
1795                         if (update_lft) {
1796                                 ifp->valid_lft = valid_lft;
1797                                 ifp->prefered_lft = prefered_lft;
1798                                 ifp->tstamp = now;
1799                                 flags = ifp->flags;
1800                                 ifp->flags &= ~IFA_F_DEPRECATED;
1801                                 spin_unlock(&ifp->lock);
1802
1803                                 if (!(flags&IFA_F_TENTATIVE))
1804                                         ipv6_ifa_notify(0, ifp);
1805                         } else
1806                                 spin_unlock(&ifp->lock);
1807
1808 #ifdef CONFIG_IPV6_PRIVACY
1809                         read_lock_bh(&in6_dev->lock);
1810                         /* update all temporary addresses in the list */
1811                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1812                                 /*
1813                                  * When adjusting the lifetimes of an existing
1814                                  * temporary address, only lower the lifetimes.
1815                                  * Implementations must not increase the
1816                                  * lifetimes of an existing temporary address
1817                                  * when processing a Prefix Information Option.
1818                                  */
1819                                 spin_lock(&ift->lock);
1820                                 flags = ift->flags;
1821                                 if (ift->valid_lft > valid_lft &&
1822                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1823                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1824                                 if (ift->prefered_lft > prefered_lft &&
1825                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1826                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1827                                 spin_unlock(&ift->lock);
1828                                 if (!(flags&IFA_F_TENTATIVE))
1829                                         ipv6_ifa_notify(0, ift);
1830                         }
1831
1832                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1833                                 /*
1834                                  * When a new public address is created as described in [ADDRCONF],
1835                                  * also create a new temporary address.
1836                                  */
1837                                 read_unlock_bh(&in6_dev->lock);
1838                                 ipv6_create_tempaddr(ifp, NULL);
1839                         } else {
1840                                 read_unlock_bh(&in6_dev->lock);
1841                         }
1842 #endif
1843                         in6_ifa_put(ifp);
1844                         addrconf_verify(0);
1845                 }
1846         }
1847         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1848         in6_dev_put(in6_dev);
1849 }
1850
1851 /*
1852  *      Set destination address.
1853  *      Special case for SIT interfaces where we create a new "virtual"
1854  *      device.
1855  */
1856 int addrconf_set_dstaddr(void __user *arg)
1857 {
1858         struct in6_ifreq ireq;
1859         struct net_device *dev;
1860         int err = -EINVAL;
1861
1862         rtnl_lock();
1863
1864         err = -EFAULT;
1865         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1866                 goto err_exit;
1867
1868         dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1869
1870         err = -ENODEV;
1871         if (dev == NULL)
1872                 goto err_exit;
1873
1874 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1875         if (dev->type == ARPHRD_SIT) {
1876                 struct ifreq ifr;
1877                 mm_segment_t    oldfs;
1878                 struct ip_tunnel_parm p;
1879
1880                 err = -EADDRNOTAVAIL;
1881                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1882                         goto err_exit;
1883
1884                 memset(&p, 0, sizeof(p));
1885                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1886                 p.iph.saddr = 0;
1887                 p.iph.version = 4;
1888                 p.iph.ihl = 5;
1889                 p.iph.protocol = IPPROTO_IPV6;
1890                 p.iph.ttl = 64;
1891                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1892
1893                 oldfs = get_fs(); set_fs(KERNEL_DS);
1894                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1895                 set_fs(oldfs);
1896
1897                 if (err == 0) {
1898                         err = -ENOBUFS;
1899                         if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1900                                 goto err_exit;
1901                         err = dev_open(dev);
1902                 }
1903         }
1904 #endif
1905
1906 err_exit:
1907         rtnl_unlock();
1908         return err;
1909 }
1910
1911 /*
1912  *      Manual configuration of address on an interface
1913  */
1914 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1915                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1916 {
1917         struct inet6_ifaddr *ifp;
1918         struct inet6_dev *idev;
1919         struct net_device *dev;
1920         int scope;
1921         u32 flags = RTF_EXPIRES;
1922
1923         ASSERT_RTNL();
1924
1925         /* check the lifetime */
1926         if (!valid_lft || prefered_lft > valid_lft)
1927                 return -EINVAL;
1928
1929         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1930                 return -ENODEV;
1931
1932         if ((idev = addrconf_add_dev(dev)) == NULL)
1933                 return -ENOBUFS;
1934
1935         scope = ipv6_addr_scope(pfx);
1936
1937         if (valid_lft == INFINITY_LIFE_TIME) {
1938                 ifa_flags |= IFA_F_PERMANENT;
1939                 flags = 0;
1940         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1941                 valid_lft = 0x7FFFFFFF/HZ;
1942
1943         if (prefered_lft == 0)
1944                 ifa_flags |= IFA_F_DEPRECATED;
1945         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1946                  (prefered_lft != INFINITY_LIFE_TIME))
1947                 prefered_lft = 0x7FFFFFFF/HZ;
1948
1949         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1950
1951         if (!IS_ERR(ifp)) {
1952                 spin_lock_bh(&ifp->lock);
1953                 ifp->valid_lft = valid_lft;
1954                 ifp->prefered_lft = prefered_lft;
1955                 ifp->tstamp = jiffies;
1956                 spin_unlock_bh(&ifp->lock);
1957
1958                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1959                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1960                 /*
1961                  * Note that section 3.1 of RFC 4429 indicates
1962                  * that the Optimistic flag should not be set for
1963                  * manually configured addresses
1964                  */
1965                 addrconf_dad_start(ifp, 0);
1966                 in6_ifa_put(ifp);
1967                 addrconf_verify(0);
1968                 return 0;
1969         }
1970
1971         return PTR_ERR(ifp);
1972 }
1973
1974 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1975 {
1976         struct inet6_ifaddr *ifp;
1977         struct inet6_dev *idev;
1978         struct net_device *dev;
1979
1980         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1981                 return -ENODEV;
1982
1983         if ((idev = __in6_dev_get(dev)) == NULL)
1984                 return -ENXIO;
1985
1986         read_lock_bh(&idev->lock);
1987         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1988                 if (ifp->prefix_len == plen &&
1989                     ipv6_addr_equal(pfx, &ifp->addr)) {
1990                         in6_ifa_hold(ifp);
1991                         read_unlock_bh(&idev->lock);
1992
1993                         ipv6_del_addr(ifp);
1994
1995                         /* If the last address is deleted administratively,
1996                            disable IPv6 on this interface.
1997                          */
1998                         if (idev->addr_list == NULL)
1999                                 addrconf_ifdown(idev->dev, 1);
2000                         return 0;
2001                 }
2002         }
2003         read_unlock_bh(&idev->lock);
2004         return -EADDRNOTAVAIL;
2005 }
2006
2007
2008 int addrconf_add_ifaddr(void __user *arg)
2009 {
2010         struct in6_ifreq ireq;
2011         int err;
2012
2013         if (!capable(CAP_NET_ADMIN))
2014                 return -EPERM;
2015
2016         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2017                 return -EFAULT;
2018
2019         rtnl_lock();
2020         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2021                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2022         rtnl_unlock();
2023         return err;
2024 }
2025
2026 int addrconf_del_ifaddr(void __user *arg)
2027 {
2028         struct in6_ifreq ireq;
2029         int err;
2030
2031         if (!capable(CAP_NET_ADMIN))
2032                 return -EPERM;
2033
2034         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2035                 return -EFAULT;
2036
2037         rtnl_lock();
2038         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2039         rtnl_unlock();
2040         return err;
2041 }
2042
2043 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2044 static void sit_add_v4_addrs(struct inet6_dev *idev)
2045 {
2046         struct inet6_ifaddr * ifp;
2047         struct in6_addr addr;
2048         struct net_device *dev;
2049         int scope;
2050
2051         ASSERT_RTNL();
2052
2053         memset(&addr, 0, sizeof(struct in6_addr));
2054         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2055
2056         if (idev->dev->flags&IFF_POINTOPOINT) {
2057                 addr.s6_addr32[0] = htonl(0xfe800000);
2058                 scope = IFA_LINK;
2059         } else {
2060                 scope = IPV6_ADDR_COMPATv4;
2061         }
2062
2063         if (addr.s6_addr32[3]) {
2064                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2065                 if (!IS_ERR(ifp)) {
2066                         spin_lock_bh(&ifp->lock);
2067                         ifp->flags &= ~IFA_F_TENTATIVE;
2068                         spin_unlock_bh(&ifp->lock);
2069                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2070                         in6_ifa_put(ifp);
2071                 }
2072                 return;
2073         }
2074
2075         for_each_netdev(&init_net, dev) {
2076                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2077                 if (in_dev && (dev->flags & IFF_UP)) {
2078                         struct in_ifaddr * ifa;
2079
2080                         int flag = scope;
2081
2082                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2083                                 int plen;
2084
2085                                 addr.s6_addr32[3] = ifa->ifa_local;
2086
2087                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2088                                         continue;
2089                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2090                                         if (idev->dev->flags&IFF_POINTOPOINT)
2091                                                 continue;
2092                                         flag |= IFA_HOST;
2093                                 }
2094                                 if (idev->dev->flags&IFF_POINTOPOINT)
2095                                         plen = 64;
2096                                 else
2097                                         plen = 96;
2098
2099                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2100                                                     IFA_F_PERMANENT);
2101                                 if (!IS_ERR(ifp)) {
2102                                         spin_lock_bh(&ifp->lock);
2103                                         ifp->flags &= ~IFA_F_TENTATIVE;
2104                                         spin_unlock_bh(&ifp->lock);
2105                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2106                                         in6_ifa_put(ifp);
2107                                 }
2108                         }
2109                 }
2110         }
2111 }
2112 #endif
2113
2114 static void init_loopback(struct net_device *dev)
2115 {
2116         struct inet6_dev  *idev;
2117         struct inet6_ifaddr * ifp;
2118
2119         /* ::1 */
2120
2121         ASSERT_RTNL();
2122
2123         if ((idev = ipv6_find_idev(dev)) == NULL) {
2124                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2125                 return;
2126         }
2127
2128         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2129         if (!IS_ERR(ifp)) {
2130                 spin_lock_bh(&ifp->lock);
2131                 ifp->flags &= ~IFA_F_TENTATIVE;
2132                 spin_unlock_bh(&ifp->lock);
2133                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2134                 in6_ifa_put(ifp);
2135         }
2136 }
2137
2138 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2139 {
2140         struct inet6_ifaddr * ifp;
2141         u32 addr_flags = IFA_F_PERMANENT;
2142
2143 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2144         if (idev->cnf.optimistic_dad &&
2145             !ipv6_devconf.forwarding)
2146                 addr_flags |= IFA_F_OPTIMISTIC;
2147 #endif
2148
2149
2150         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2151         if (!IS_ERR(ifp)) {
2152                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2153                 addrconf_dad_start(ifp, 0);
2154                 in6_ifa_put(ifp);
2155         }
2156 }
2157
2158 static void addrconf_dev_config(struct net_device *dev)
2159 {
2160         struct in6_addr addr;
2161         struct inet6_dev    * idev;
2162
2163         ASSERT_RTNL();
2164
2165         if ((dev->type != ARPHRD_ETHER) &&
2166             (dev->type != ARPHRD_FDDI) &&
2167             (dev->type != ARPHRD_IEEE802_TR) &&
2168             (dev->type != ARPHRD_ARCNET) &&
2169             (dev->type != ARPHRD_INFINIBAND)) {
2170                 /* Alas, we support only Ethernet autoconfiguration. */
2171                 return;
2172         }
2173
2174         idev = addrconf_add_dev(dev);
2175         if (idev == NULL)
2176                 return;
2177
2178         memset(&addr, 0, sizeof(struct in6_addr));
2179         addr.s6_addr32[0] = htonl(0xFE800000);
2180
2181         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2182                 addrconf_add_linklocal(idev, &addr);
2183 }
2184
2185 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2186 static void addrconf_sit_config(struct net_device *dev)
2187 {
2188         struct inet6_dev *idev;
2189
2190         ASSERT_RTNL();
2191
2192         /*
2193          * Configure the tunnel with one of our IPv4
2194          * addresses... we should configure all of
2195          * our v4 addrs in the tunnel
2196          */
2197
2198         if ((idev = ipv6_find_idev(dev)) == NULL) {
2199                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2200                 return;
2201         }
2202
2203         sit_add_v4_addrs(idev);
2204
2205         if (dev->flags&IFF_POINTOPOINT) {
2206                 addrconf_add_mroute(dev);
2207                 addrconf_add_lroute(dev);
2208         } else
2209                 sit_route_add(dev);
2210 }
2211 #endif
2212
2213 static inline int
2214 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2215 {
2216         struct in6_addr lladdr;
2217
2218         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2219                 addrconf_add_linklocal(idev, &lladdr);
2220                 return 0;
2221         }
2222         return -1;
2223 }
2224
2225 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2226 {
2227         struct net_device *link_dev;
2228
2229         /* first try to inherit the link-local address from the link device */
2230         if (idev->dev->iflink &&
2231             (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2232                 if (!ipv6_inherit_linklocal(idev, link_dev))
2233                         return;
2234         }
2235         /* then try to inherit it from any device */
2236         for_each_netdev(&init_net, link_dev) {
2237                 if (!ipv6_inherit_linklocal(idev, link_dev))
2238                         return;
2239         }
2240         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2241 }
2242
2243 /*
2244  * Autoconfigure tunnel with a link-local address so routing protocols,
2245  * DHCPv6, MLD etc. can be run over the virtual link
2246  */
2247
2248 static void addrconf_ip6_tnl_config(struct net_device *dev)
2249 {
2250         struct inet6_dev *idev;
2251
2252         ASSERT_RTNL();
2253
2254         if ((idev = addrconf_add_dev(dev)) == NULL) {
2255                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2256                 return;
2257         }
2258         ip6_tnl_add_linklocal(idev);
2259 }
2260
2261 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2262                            void * data)
2263 {
2264         struct net_device *dev = (struct net_device *) data;
2265         struct inet6_dev *idev = __in6_dev_get(dev);
2266         int run_pending = 0;
2267         int err;
2268
2269         if (dev->nd_net != &init_net)
2270                 return NOTIFY_DONE;
2271
2272         switch(event) {
2273         case NETDEV_REGISTER:
2274                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2275                         idev = ipv6_add_dev(dev);
2276                         if (!idev)
2277                                 return notifier_from_errno(-ENOMEM);
2278                 }
2279                 break;
2280         case NETDEV_UP:
2281         case NETDEV_CHANGE:
2282                 if (dev->flags & IFF_SLAVE)
2283                         break;
2284
2285                 if (event == NETDEV_UP) {
2286                         if (!netif_carrier_ok(dev)) {
2287                                 /* device is not ready yet. */
2288                                 printk(KERN_INFO
2289                                         "ADDRCONF(NETDEV_UP): %s: "
2290                                         "link is not ready\n",
2291                                         dev->name);
2292                                 break;
2293                         }
2294
2295                         if (idev)
2296                                 idev->if_flags |= IF_READY;
2297                 } else {
2298                         if (!netif_carrier_ok(dev)) {
2299                                 /* device is still not ready. */
2300                                 break;
2301                         }
2302
2303                         if (idev) {
2304                                 if (idev->if_flags & IF_READY) {
2305                                         /* device is already configured. */
2306                                         break;
2307                                 }
2308                                 idev->if_flags |= IF_READY;
2309                         }
2310
2311                         printk(KERN_INFO
2312                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2313                                         "link becomes ready\n",
2314                                         dev->name);
2315
2316                         run_pending = 1;
2317                 }
2318
2319                 switch(dev->type) {
2320 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2321                 case ARPHRD_SIT:
2322                         addrconf_sit_config(dev);
2323                         break;
2324 #endif
2325                 case ARPHRD_TUNNEL6:
2326                         addrconf_ip6_tnl_config(dev);
2327                         break;
2328                 case ARPHRD_LOOPBACK:
2329                         init_loopback(dev);
2330                         break;
2331
2332                 default:
2333                         addrconf_dev_config(dev);
2334                         break;
2335                 }
2336                 if (idev) {
2337                         if (run_pending)
2338                                 addrconf_dad_run(idev);
2339
2340                         /* If the MTU changed during the interface down, when the
2341                            interface up, the changed MTU must be reflected in the
2342                            idev as well as routers.
2343                          */
2344                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2345                                 rt6_mtu_change(dev, dev->mtu);
2346                                 idev->cnf.mtu6 = dev->mtu;
2347                         }
2348                         idev->tstamp = jiffies;
2349                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2350                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2351                            stop IPv6 on this interface.
2352                          */
2353                         if (dev->mtu < IPV6_MIN_MTU)
2354                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2355                 }
2356                 break;
2357
2358         case NETDEV_CHANGEMTU:
2359                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2360                         rt6_mtu_change(dev, dev->mtu);
2361                         idev->cnf.mtu6 = dev->mtu;
2362                         break;
2363                 }
2364
2365                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2366
2367         case NETDEV_DOWN:
2368         case NETDEV_UNREGISTER:
2369                 /*
2370                  *      Remove all addresses from this interface.
2371                  */
2372                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2373                 break;
2374
2375         case NETDEV_CHANGENAME:
2376                 if (idev) {
2377                         snmp6_unregister_dev(idev);
2378 #ifdef CONFIG_SYSCTL
2379                         addrconf_sysctl_unregister(&idev->cnf);
2380                         neigh_sysctl_unregister(idev->nd_parms);
2381                         neigh_sysctl_register(dev, idev->nd_parms,
2382                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2383                                               &ndisc_ifinfo_sysctl_change,
2384                                               NULL);
2385                         addrconf_sysctl_register(idev, &idev->cnf);
2386 #endif
2387                         err = snmp6_register_dev(idev);
2388                         if (err)
2389                                 return notifier_from_errno(err);
2390                 }
2391                 break;
2392         }
2393
2394         return NOTIFY_OK;
2395 }
2396
2397 /*
2398  *      addrconf module should be notified of a device going up
2399  */
2400 static struct notifier_block ipv6_dev_notf = {
2401         .notifier_call = addrconf_notify,
2402         .priority = 0
2403 };
2404
2405 static int addrconf_ifdown(struct net_device *dev, int how)
2406 {
2407         struct inet6_dev *idev;
2408         struct inet6_ifaddr *ifa, **bifa;
2409         int i;
2410
2411         ASSERT_RTNL();
2412
2413         if (dev == init_net.loopback_dev && how == 1)
2414                 how = 0;
2415
2416         rt6_ifdown(dev);
2417         neigh_ifdown(&nd_tbl, dev);
2418
2419         idev = __in6_dev_get(dev);
2420         if (idev == NULL)
2421                 return -ENODEV;
2422
2423         /* Step 1: remove reference to ipv6 device from parent device.
2424                    Do not dev_put!
2425          */
2426         if (how == 1) {
2427                 idev->dead = 1;
2428
2429                 /* protected by rtnl_lock */
2430                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2431
2432                 /* Step 1.5: remove snmp6 entry */
2433                 snmp6_unregister_dev(idev);
2434
2435         }
2436
2437         /* Step 2: clear hash table */
2438         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2439                 bifa = &inet6_addr_lst[i];
2440
2441                 write_lock_bh(&addrconf_hash_lock);
2442                 while ((ifa = *bifa) != NULL) {
2443                         if (ifa->idev == idev) {
2444                                 *bifa = ifa->lst_next;
2445                                 ifa->lst_next = NULL;
2446                                 addrconf_del_timer(ifa);
2447                                 in6_ifa_put(ifa);
2448                                 continue;
2449                         }
2450                         bifa = &ifa->lst_next;
2451                 }
2452                 write_unlock_bh(&addrconf_hash_lock);
2453         }
2454
2455         write_lock_bh(&idev->lock);
2456
2457         /* Step 3: clear flags for stateless addrconf */
2458         if (how != 1)
2459                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2460
2461         /* Step 4: clear address list */
2462 #ifdef CONFIG_IPV6_PRIVACY
2463         if (how == 1 && del_timer(&idev->regen_timer))
2464                 in6_dev_put(idev);
2465
2466         /* clear tempaddr list */
2467         while ((ifa = idev->tempaddr_list) != NULL) {
2468                 idev->tempaddr_list = ifa->tmp_next;
2469                 ifa->tmp_next = NULL;
2470                 ifa->dead = 1;
2471                 write_unlock_bh(&idev->lock);
2472                 spin_lock_bh(&ifa->lock);
2473
2474                 if (ifa->ifpub) {
2475                         in6_ifa_put(ifa->ifpub);
2476                         ifa->ifpub = NULL;
2477                 }
2478                 spin_unlock_bh(&ifa->lock);
2479                 in6_ifa_put(ifa);
2480                 write_lock_bh(&idev->lock);
2481         }
2482 #endif
2483         while ((ifa = idev->addr_list) != NULL) {
2484                 idev->addr_list = ifa->if_next;
2485                 ifa->if_next = NULL;
2486                 ifa->dead = 1;
2487                 addrconf_del_timer(ifa);
2488                 write_unlock_bh(&idev->lock);
2489
2490                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2491                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2492                 in6_ifa_put(ifa);
2493
2494                 write_lock_bh(&idev->lock);
2495         }
2496         write_unlock_bh(&idev->lock);
2497
2498         /* Step 5: Discard multicast list */
2499
2500         if (how == 1)
2501                 ipv6_mc_destroy_dev(idev);
2502         else
2503                 ipv6_mc_down(idev);
2504
2505         idev->tstamp = jiffies;
2506
2507         /* Shot the device (if unregistered) */
2508
2509         if (how == 1) {
2510 #ifdef CONFIG_SYSCTL
2511                 addrconf_sysctl_unregister(&idev->cnf);
2512                 neigh_sysctl_unregister(idev->nd_parms);
2513 #endif
2514                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2515                 neigh_ifdown(&nd_tbl, dev);
2516                 in6_dev_put(idev);
2517         }
2518         return 0;
2519 }
2520
2521 static void addrconf_rs_timer(unsigned long data)
2522 {
2523         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2524
2525         if (ifp->idev->cnf.forwarding)
2526                 goto out;
2527
2528         if (ifp->idev->if_flags & IF_RA_RCVD) {
2529                 /*
2530                  *      Announcement received after solicitation
2531                  *      was sent
2532                  */
2533                 goto out;
2534         }
2535
2536         spin_lock(&ifp->lock);
2537         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2538                 struct in6_addr all_routers;
2539
2540                 /* The wait after the last probe can be shorter */
2541                 addrconf_mod_timer(ifp, AC_RS,
2542                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2543                                    ifp->idev->cnf.rtr_solicit_delay :
2544                                    ifp->idev->cnf.rtr_solicit_interval);
2545                 spin_unlock(&ifp->lock);
2546
2547                 ipv6_addr_all_routers(&all_routers);
2548
2549                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2550         } else {
2551                 spin_unlock(&ifp->lock);
2552                 /*
2553                  * Note: we do not support deprecated "all on-link"
2554                  * assumption any longer.
2555                  */
2556                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2557                        ifp->idev->dev->name);
2558         }
2559
2560 out:
2561         in6_ifa_put(ifp);
2562 }
2563
2564 /*
2565  *      Duplicate Address Detection
2566  */
2567 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2568 {
2569         unsigned long rand_num;
2570         struct inet6_dev *idev = ifp->idev;
2571
2572         if (ifp->flags & IFA_F_OPTIMISTIC)
2573                 rand_num = 0;
2574         else
2575                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2576
2577         ifp->probes = idev->cnf.dad_transmits;
2578         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2579 }
2580
2581 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2582 {
2583         struct inet6_dev *idev = ifp->idev;
2584         struct net_device *dev = idev->dev;
2585
2586         addrconf_join_solict(dev, &ifp->addr);
2587
2588         net_srandom(ifp->addr.s6_addr32[3]);
2589
2590         read_lock_bh(&idev->lock);
2591         if (ifp->dead)
2592                 goto out;
2593         spin_lock_bh(&ifp->lock);
2594
2595         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2596             !(ifp->flags&IFA_F_TENTATIVE) ||
2597             ifp->flags & IFA_F_NODAD) {
2598                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2599                 spin_unlock_bh(&ifp->lock);
2600                 read_unlock_bh(&idev->lock);
2601
2602                 addrconf_dad_completed(ifp);
2603                 return;
2604         }
2605
2606         if (!(idev->if_flags & IF_READY)) {
2607                 spin_unlock_bh(&ifp->lock);
2608                 read_unlock_bh(&idev->lock);
2609                 /*
2610                  * If the defice is not ready:
2611                  * - keep it tentative if it is a permanent address.
2612                  * - otherwise, kill it.
2613                  */
2614                 in6_ifa_hold(ifp);
2615                 addrconf_dad_stop(ifp);
2616                 return;
2617         }
2618
2619         /*
2620          * Optimistic nodes can start receiving
2621          * Frames right away
2622          */
2623         if(ifp->flags & IFA_F_OPTIMISTIC)
2624                 ip6_ins_rt(ifp->rt);
2625
2626         addrconf_dad_kick(ifp);
2627         spin_unlock_bh(&ifp->lock);
2628 out:
2629         read_unlock_bh(&idev->lock);
2630 }
2631
2632 static void addrconf_dad_timer(unsigned long data)
2633 {
2634         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2635         struct inet6_dev *idev = ifp->idev;
2636         struct in6_addr unspec;
2637         struct in6_addr mcaddr;
2638
2639         read_lock_bh(&idev->lock);
2640         if (idev->dead) {
2641                 read_unlock_bh(&idev->lock);
2642                 goto out;
2643         }
2644         spin_lock_bh(&ifp->lock);
2645         if (ifp->probes == 0) {
2646                 /*
2647                  * DAD was successful
2648                  */
2649
2650                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2651                 spin_unlock_bh(&ifp->lock);
2652                 read_unlock_bh(&idev->lock);
2653
2654                 addrconf_dad_completed(ifp);
2655
2656                 goto out;
2657         }
2658
2659         ifp->probes--;
2660         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2661         spin_unlock_bh(&ifp->lock);
2662         read_unlock_bh(&idev->lock);
2663
2664         /* send a neighbour solicitation for our addr */
2665         memset(&unspec, 0, sizeof(unspec));
2666         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2667         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2668 out:
2669         in6_ifa_put(ifp);
2670 }
2671
2672 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2673 {
2674         struct net_device *     dev = ifp->idev->dev;
2675
2676         /*
2677          *      Configure the address for reception. Now it is valid.
2678          */
2679
2680         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2681
2682         /* If added prefix is link local and forwarding is off,
2683            start sending router solicitations.
2684          */
2685
2686         if (ifp->idev->cnf.forwarding == 0 &&
2687             ifp->idev->cnf.rtr_solicits > 0 &&
2688             (dev->flags&IFF_LOOPBACK) == 0 &&
2689             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2690                 struct in6_addr all_routers;
2691
2692                 ipv6_addr_all_routers(&all_routers);
2693
2694                 /*
2695                  *      If a host as already performed a random delay
2696                  *      [...] as part of DAD [...] there is no need
2697                  *      to delay again before sending the first RS
2698                  */
2699                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2700
2701                 spin_lock_bh(&ifp->lock);
2702                 ifp->probes = 1;
2703                 ifp->idev->if_flags |= IF_RS_SENT;
2704                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2705                 spin_unlock_bh(&ifp->lock);
2706         }
2707 }
2708
2709 static void addrconf_dad_run(struct inet6_dev *idev) {
2710         struct inet6_ifaddr *ifp;
2711
2712         read_lock_bh(&idev->lock);
2713         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2714                 spin_lock_bh(&ifp->lock);
2715                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2716                         spin_unlock_bh(&ifp->lock);
2717                         continue;
2718                 }
2719                 spin_unlock_bh(&ifp->lock);
2720                 addrconf_dad_kick(ifp);
2721         }
2722         read_unlock_bh(&idev->lock);
2723 }
2724
2725 #ifdef CONFIG_PROC_FS
2726 struct if6_iter_state {
2727         int bucket;
2728 };
2729
2730 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2731 {
2732         struct inet6_ifaddr *ifa = NULL;
2733         struct if6_iter_state *state = seq->private;
2734
2735         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2736                 ifa = inet6_addr_lst[state->bucket];
2737                 if (ifa)
2738                         break;
2739         }
2740         return ifa;
2741 }
2742
2743 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2744 {
2745         struct if6_iter_state *state = seq->private;
2746
2747         ifa = ifa->lst_next;
2748 try_again:
2749         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2750                 ifa = inet6_addr_lst[state->bucket];
2751                 goto try_again;
2752         }
2753         return ifa;
2754 }
2755
2756 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2757 {
2758         struct inet6_ifaddr *ifa = if6_get_first(seq);
2759
2760         if (ifa)
2761                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2762                         --pos;
2763         return pos ? NULL : ifa;
2764 }
2765
2766 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2767 {
2768         read_lock_bh(&addrconf_hash_lock);
2769         return if6_get_idx(seq, *pos);
2770 }
2771
2772 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2773 {
2774         struct inet6_ifaddr *ifa;
2775
2776         ifa = if6_get_next(seq, v);
2777         ++*pos;
2778         return ifa;
2779 }
2780
2781 static void if6_seq_stop(struct seq_file *seq, void *v)
2782 {
2783         read_unlock_bh(&addrconf_hash_lock);
2784 }
2785
2786 static int if6_seq_show(struct seq_file *seq, void *v)
2787 {
2788         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2789         seq_printf(seq,
2790                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2791                    NIP6(ifp->addr),
2792                    ifp->idev->dev->ifindex,
2793                    ifp->prefix_len,
2794                    ifp->scope,
2795                    ifp->flags,
2796                    ifp->idev->dev->name);
2797         return 0;
2798 }
2799
2800 static const struct seq_operations if6_seq_ops = {
2801         .start  = if6_seq_start,
2802         .next   = if6_seq_next,
2803         .show   = if6_seq_show,
2804         .stop   = if6_seq_stop,
2805 };
2806
2807 static int if6_seq_open(struct inode *inode, struct file *file)
2808 {
2809         struct seq_file *seq;
2810         int rc = -ENOMEM;
2811         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2812
2813         if (!s)
2814                 goto out;
2815
2816         rc = seq_open(file, &if6_seq_ops);
2817         if (rc)
2818                 goto out_kfree;
2819
2820         seq = file->private_data;
2821         seq->private = s;
2822 out:
2823         return rc;
2824 out_kfree:
2825         kfree(s);
2826         goto out;
2827 }
2828
2829 static const struct file_operations if6_fops = {
2830         .owner          = THIS_MODULE,
2831         .open           = if6_seq_open,
2832         .read           = seq_read,
2833         .llseek         = seq_lseek,
2834         .release        = seq_release_private,
2835 };
2836
2837 int __init if6_proc_init(void)
2838 {
2839         if (!proc_net_fops_create(&init_net, "if_inet6", S_IRUGO, &if6_fops))
2840                 return -ENOMEM;
2841         return 0;
2842 }
2843
2844 void if6_proc_exit(void)
2845 {
2846         proc_net_remove(&init_net, "if_inet6");
2847 }
2848 #endif  /* CONFIG_PROC_FS */
2849
2850 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2851 /* Check if address is a home address configured on any interface. */
2852 int ipv6_chk_home_addr(struct in6_addr *addr)
2853 {
2854         int ret = 0;
2855         struct inet6_ifaddr * ifp;
2856         u8 hash = ipv6_addr_hash(addr);
2857         read_lock_bh(&addrconf_hash_lock);
2858         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2859                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2860                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2861                         ret = 1;
2862                         break;
2863                 }
2864         }
2865         read_unlock_bh(&addrconf_hash_lock);
2866         return ret;
2867 }
2868 #endif
2869
2870 /*
2871  *      Periodic address status verification
2872  */
2873
2874 static void addrconf_verify(unsigned long foo)
2875 {
2876         struct inet6_ifaddr *ifp;
2877         unsigned long now, next;
2878         int i;
2879
2880         spin_lock_bh(&addrconf_verify_lock);
2881         now = jiffies;
2882         next = now + ADDR_CHECK_FREQUENCY;
2883
2884         del_timer(&addr_chk_timer);
2885
2886         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2887
2888 restart:
2889                 read_lock(&addrconf_hash_lock);
2890                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2891                         unsigned long age;
2892 #ifdef CONFIG_IPV6_PRIVACY
2893                         unsigned long regen_advance;
2894 #endif
2895
2896                         if (ifp->flags & IFA_F_PERMANENT)
2897                                 continue;
2898
2899                         spin_lock(&ifp->lock);
2900                         age = (now - ifp->tstamp) / HZ;
2901
2902 #ifdef CONFIG_IPV6_PRIVACY
2903                         regen_advance = ifp->idev->cnf.regen_max_retry *
2904                                         ifp->idev->cnf.dad_transmits *
2905                                         ifp->idev->nd_parms->retrans_time / HZ;
2906 #endif
2907
2908                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2909                             age >= ifp->valid_lft) {
2910                                 spin_unlock(&ifp->lock);
2911                                 in6_ifa_hold(ifp);
2912                                 read_unlock(&addrconf_hash_lock);
2913                                 ipv6_del_addr(ifp);
2914                                 goto restart;
2915                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2916                                 spin_unlock(&ifp->lock);
2917                                 continue;
2918                         } else if (age >= ifp->prefered_lft) {
2919                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2920                                 int deprecate = 0;
2921
2922                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2923                                         deprecate = 1;
2924                                         ifp->flags |= IFA_F_DEPRECATED;
2925                                 }
2926
2927                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2928                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2929
2930                                 spin_unlock(&ifp->lock);
2931
2932                                 if (deprecate) {
2933                                         in6_ifa_hold(ifp);
2934                                         read_unlock(&addrconf_hash_lock);
2935
2936                                         ipv6_ifa_notify(0, ifp);
2937                                         in6_ifa_put(ifp);
2938                                         goto restart;
2939                                 }
2940 #ifdef CONFIG_IPV6_PRIVACY
2941                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2942                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2943                                 if (age >= ifp->prefered_lft - regen_advance) {
2944                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2945                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2946                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2947                                         if (!ifp->regen_count && ifpub) {
2948                                                 ifp->regen_count++;
2949                                                 in6_ifa_hold(ifp);
2950                                                 in6_ifa_hold(ifpub);
2951                                                 spin_unlock(&ifp->lock);
2952                                                 read_unlock(&addrconf_hash_lock);
2953                                                 spin_lock(&ifpub->lock);
2954                                                 ifpub->regen_count = 0;
2955                                                 spin_unlock(&ifpub->lock);
2956                                                 ipv6_create_tempaddr(ifpub, ifp);
2957                                                 in6_ifa_put(ifpub);
2958                                                 in6_ifa_put(ifp);
2959                                                 goto restart;
2960                                         }
2961                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2962                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2963                                 spin_unlock(&ifp->lock);
2964 #endif
2965                         } else {
2966                                 /* ifp->prefered_lft <= ifp->valid_lft */
2967                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2968                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2969                                 spin_unlock(&ifp->lock);
2970                         }
2971                 }
2972                 read_unlock(&addrconf_hash_lock);
2973         }
2974
2975         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2976         add_timer(&addr_chk_timer);
2977         spin_unlock_bh(&addrconf_verify_lock);
2978 }
2979
2980 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2981 {
2982         struct in6_addr *pfx = NULL;
2983
2984         if (addr)
2985                 pfx = nla_data(addr);
2986
2987         if (local) {
2988                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2989                         pfx = NULL;
2990                 else
2991                         pfx = nla_data(local);
2992         }
2993
2994         return pfx;
2995 }
2996
2997 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
2998         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2999         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3000         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3001 };
3002
3003 static int
3004 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3005 {
3006         struct ifaddrmsg *ifm;
3007         struct nlattr *tb[IFA_MAX+1];
3008         struct in6_addr *pfx;
3009         int err;
3010
3011         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3012         if (err < 0)
3013                 return err;
3014
3015         ifm = nlmsg_data(nlh);
3016         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3017         if (pfx == NULL)
3018                 return -EINVAL;
3019
3020         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3021 }
3022
3023 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3024                              u32 prefered_lft, u32 valid_lft)
3025 {
3026         u32 flags = RTF_EXPIRES;
3027
3028         if (!valid_lft || (prefered_lft > valid_lft))
3029                 return -EINVAL;
3030
3031         if (valid_lft == INFINITY_LIFE_TIME) {
3032                 ifa_flags |= IFA_F_PERMANENT;
3033                 flags = 0;
3034         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3035                 valid_lft = 0x7FFFFFFF/HZ;
3036
3037         if (prefered_lft == 0)
3038                 ifa_flags |= IFA_F_DEPRECATED;
3039         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3040                  (prefered_lft != INFINITY_LIFE_TIME))
3041                 prefered_lft = 0x7FFFFFFF/HZ;
3042
3043         spin_lock_bh(&ifp->lock);
3044         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3045         ifp->tstamp = jiffies;
3046         ifp->valid_lft = valid_lft;
3047         ifp->prefered_lft = prefered_lft;
3048
3049         spin_unlock_bh(&ifp->lock);
3050         if (!(ifp->flags&IFA_F_TENTATIVE))
3051                 ipv6_ifa_notify(0, ifp);
3052
3053         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3054                               jiffies_to_clock_t(valid_lft * HZ), flags);
3055         addrconf_verify(0);
3056
3057         return 0;
3058 }
3059
3060 static int
3061 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3062 {
3063         struct ifaddrmsg *ifm;
3064         struct nlattr *tb[IFA_MAX+1];
3065         struct in6_addr *pfx;
3066         struct inet6_ifaddr *ifa;
3067         struct net_device *dev;
3068         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3069         u8 ifa_flags;
3070         int err;
3071
3072         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3073         if (err < 0)
3074                 return err;
3075
3076         ifm = nlmsg_data(nlh);
3077         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3078         if (pfx == NULL)
3079                 return -EINVAL;
3080
3081         if (tb[IFA_CACHEINFO]) {
3082                 struct ifa_cacheinfo *ci;
3083
3084                 ci = nla_data(tb[IFA_CACHEINFO]);
3085                 valid_lft = ci->ifa_valid;
3086                 preferred_lft = ci->ifa_prefered;
3087         } else {
3088                 preferred_lft = INFINITY_LIFE_TIME;
3089                 valid_lft = INFINITY_LIFE_TIME;
3090         }
3091
3092         dev =  __dev_get_by_index(&init_net, ifm->ifa_index);
3093         if (dev == NULL)
3094                 return -ENODEV;
3095
3096         /* We ignore other flags so far. */
3097         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3098
3099         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3100         if (ifa == NULL) {
3101                 /*
3102                  * It would be best to check for !NLM_F_CREATE here but
3103                  * userspace alreay relies on not having to provide this.
3104                  */
3105                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3106                                       ifa_flags, preferred_lft, valid_lft);
3107         }
3108
3109         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3110             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3111                 err = -EEXIST;
3112         else
3113                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3114
3115         in6_ifa_put(ifa);
3116
3117         return err;
3118 }
3119
3120 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3121                           u8 scope, int ifindex)
3122 {
3123         struct ifaddrmsg *ifm;
3124
3125         ifm = nlmsg_data(nlh);
3126         ifm->ifa_family = AF_INET6;
3127         ifm->ifa_prefixlen = prefixlen;
3128         ifm->ifa_flags = flags;
3129         ifm->ifa_scope = scope;
3130         ifm->ifa_index = ifindex;
3131 }
3132
3133 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3134                          unsigned long tstamp, u32 preferred, u32 valid)
3135 {
3136         struct ifa_cacheinfo ci;
3137
3138         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3139                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3140         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3141                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3142         ci.ifa_prefered = preferred;
3143         ci.ifa_valid = valid;
3144
3145         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3146 }
3147
3148 static inline int rt_scope(int ifa_scope)
3149 {
3150         if (ifa_scope & IFA_HOST)
3151                 return RT_SCOPE_HOST;
3152         else if (ifa_scope & IFA_LINK)
3153                 return RT_SCOPE_LINK;
3154         else if (ifa_scope & IFA_SITE)
3155                 return RT_SCOPE_SITE;
3156         else
3157                 return RT_SCOPE_UNIVERSE;
3158 }
3159
3160 static inline int inet6_ifaddr_msgsize(void)
3161 {
3162         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3163                + nla_total_size(16) /* IFA_ADDRESS */
3164                + nla_total_size(sizeof(struct ifa_cacheinfo));
3165 }
3166
3167 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3168                              u32 pid, u32 seq, int event, unsigned int flags)
3169 {
3170         struct nlmsghdr  *nlh;
3171         u32 preferred, valid;
3172
3173         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3174         if (nlh == NULL)
3175                 return -EMSGSIZE;
3176
3177         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3178                       ifa->idev->dev->ifindex);
3179
3180         if (!(ifa->flags&IFA_F_PERMANENT)) {
3181                 preferred = ifa->prefered_lft;
3182                 valid = ifa->valid_lft;
3183                 if (preferred != INFINITY_LIFE_TIME) {
3184                         long tval = (jiffies - ifa->tstamp)/HZ;
3185                         preferred -= tval;
3186                         if (valid != INFINITY_LIFE_TIME)
3187                                 valid -= tval;
3188                 }
3189         } else {
3190                 preferred = INFINITY_LIFE_TIME;
3191                 valid = INFINITY_LIFE_TIME;
3192         }
3193
3194         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3195             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3196                 nlmsg_cancel(skb, nlh);
3197                 return -EMSGSIZE;
3198         }
3199
3200         return nlmsg_end(skb, nlh);
3201 }
3202
3203 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3204                                 u32 pid, u32 seq, int event, u16 flags)
3205 {
3206         struct nlmsghdr  *nlh;
3207         u8 scope = RT_SCOPE_UNIVERSE;
3208         int ifindex = ifmca->idev->dev->ifindex;
3209
3210         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3211                 scope = RT_SCOPE_SITE;
3212
3213         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3214         if (nlh == NULL)
3215                 return -EMSGSIZE;
3216
3217         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3218         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3219             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3220                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3221                 nlmsg_cancel(skb, nlh);
3222                 return -EMSGSIZE;
3223         }
3224
3225         return nlmsg_end(skb, nlh);
3226 }
3227
3228 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3229                                 u32 pid, u32 seq, int event, unsigned int flags)
3230 {
3231         struct nlmsghdr  *nlh;
3232         u8 scope = RT_SCOPE_UNIVERSE;
3233         int ifindex = ifaca->aca_idev->dev->ifindex;
3234
3235         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3236                 scope = RT_SCOPE_SITE;
3237
3238         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3239         if (nlh == NULL)
3240                 return -EMSGSIZE;
3241
3242         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3243         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3244             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3245                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3246                 nlmsg_cancel(skb, nlh);
3247                 return -EMSGSIZE;
3248         }
3249
3250         return nlmsg_end(skb, nlh);
3251 }
3252
3253 enum addr_type_t
3254 {
3255         UNICAST_ADDR,
3256         MULTICAST_ADDR,
3257         ANYCAST_ADDR,
3258 };
3259
3260 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3261                            enum addr_type_t type)
3262 {
3263         int idx, ip_idx;
3264         int s_idx, s_ip_idx;
3265         int err = 1;
3266         struct net_device *dev;
3267         struct inet6_dev *idev = NULL;
3268         struct inet6_ifaddr *ifa;
3269         struct ifmcaddr6 *ifmca;
3270         struct ifacaddr6 *ifaca;
3271
3272         s_idx = cb->args[0];
3273         s_ip_idx = ip_idx = cb->args[1];
3274
3275         idx = 0;
3276         for_each_netdev(&init_net, dev) {
3277                 if (idx < s_idx)
3278                         goto cont;
3279                 if (idx > s_idx)
3280                         s_ip_idx = 0;
3281                 ip_idx = 0;
3282                 if ((idev = in6_dev_get(dev)) == NULL)
3283                         goto cont;
3284                 read_lock_bh(&idev->lock);
3285                 switch (type) {
3286                 case UNICAST_ADDR:
3287                         /* unicast address incl. temp addr */
3288                         for (ifa = idev->addr_list; ifa;
3289                              ifa = ifa->if_next, ip_idx++) {
3290                                 if (ip_idx < s_ip_idx)
3291                                         continue;
3292                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3293                                     NETLINK_CB(cb->skb).pid,
3294                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3295                                     NLM_F_MULTI)) <= 0)
3296                                         goto done;
3297                         }
3298                         break;
3299                 case MULTICAST_ADDR:
3300                         /* multicast address */
3301                         for (ifmca = idev->mc_list; ifmca;
3302                              ifmca = ifmca->next, ip_idx++) {
3303                                 if (ip_idx < s_ip_idx)
3304                                         continue;
3305                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3306                                     NETLINK_CB(cb->skb).pid,
3307                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3308                                     NLM_F_MULTI)) <= 0)
3309                                         goto done;
3310                         }
3311                         break;
3312                 case ANYCAST_ADDR:
3313                         /* anycast address */
3314                         for (ifaca = idev->ac_list; ifaca;
3315                              ifaca = ifaca->aca_next, ip_idx++) {
3316                                 if (ip_idx < s_ip_idx)
3317                                         continue;
3318                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3319                                     NETLINK_CB(cb->skb).pid,
3320                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3321                                     NLM_F_MULTI)) <= 0)
3322                                         goto done;
3323                         }
3324                         break;
3325                 default:
3326                         break;
3327                 }
3328                 read_unlock_bh(&idev->lock);
3329                 in6_dev_put(idev);
3330 cont:
3331                 idx++;
3332         }
3333 done:
3334         if (err <= 0) {
3335                 read_unlock_bh(&idev->lock);
3336                 in6_dev_put(idev);
3337         }
3338         cb->args[0] = idx;
3339         cb->args[1] = ip_idx;
3340         return skb->len;
3341 }
3342
3343 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3344 {
3345         enum addr_type_t type = UNICAST_ADDR;
3346         return inet6_dump_addr(skb, cb, type);
3347 }
3348
3349 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3350 {
3351         enum addr_type_t type = MULTICAST_ADDR;
3352         return inet6_dump_addr(skb, cb, type);
3353 }
3354
3355
3356 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3357 {
3358         enum addr_type_t type = ANYCAST_ADDR;
3359         return inet6_dump_addr(skb, cb, type);
3360 }
3361
3362 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3363                              void *arg)
3364 {
3365         struct ifaddrmsg *ifm;
3366         struct nlattr *tb[IFA_MAX+1];
3367         struct in6_addr *addr = NULL;
3368         struct net_device *dev = NULL;
3369         struct inet6_ifaddr *ifa;
3370         struct sk_buff *skb;
3371         int err;
3372
3373         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3374         if (err < 0)
3375                 goto errout;
3376
3377         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3378         if (addr == NULL) {
3379                 err = -EINVAL;
3380                 goto errout;
3381         }
3382
3383         ifm = nlmsg_data(nlh);
3384         if (ifm->ifa_index)
3385                 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3386
3387         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3388                 err = -EADDRNOTAVAIL;
3389                 goto errout;
3390         }
3391
3392         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3393                 err = -ENOBUFS;
3394                 goto errout_ifa;
3395         }
3396
3397         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3398                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3399         if (err < 0) {
3400                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3401                 WARN_ON(err == -EMSGSIZE);
3402                 kfree_skb(skb);
3403                 goto errout_ifa;
3404         }
3405         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3406 errout_ifa:
3407         in6_ifa_put(ifa);
3408 errout:
3409         return err;
3410 }
3411
3412 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3413 {
3414         struct sk_buff *skb;
3415         int err = -ENOBUFS;
3416
3417         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3418         if (skb == NULL)
3419                 goto errout;
3420
3421         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3422         if (err < 0) {
3423                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3424                 WARN_ON(err == -EMSGSIZE);
3425                 kfree_skb(skb);
3426                 goto errout;
3427         }
3428         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3429 errout:
3430         if (err < 0)
3431                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3432 }
3433
3434 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3435                                 __s32 *array, int bytes)
3436 {
3437         BUG_ON(bytes < (DEVCONF_MAX * 4));
3438
3439         memset(array, 0, bytes);
3440         array[DEVCONF_FORWARDING] = cnf->forwarding;
3441         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3442         array[DEVCONF_MTU6] = cnf->mtu6;
3443         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3444         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3445         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3446         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3447         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3448         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3449         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3450         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3451 #ifdef CONFIG_IPV6_PRIVACY
3452         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3453         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3454         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3455         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3456         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3457 #endif
3458         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3459         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3460         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3461 #ifdef CONFIG_IPV6_ROUTER_PREF
3462         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3463         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3464 #ifdef CONFIG_IPV6_ROUTE_INFO
3465         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3466 #endif
3467 #endif
3468         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3469         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3470 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3471         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3472 #endif
3473 }
3474
3475 static inline size_t inet6_if_nlmsg_size(void)
3476 {
3477         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3478                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3479                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3480                + nla_total_size(4) /* IFLA_MTU */
3481                + nla_total_size(4) /* IFLA_LINK */
3482                + nla_total_size( /* IFLA_PROTINFO */
3483                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3484                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3485                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3486                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3487                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3488                  );
3489 }
3490
3491 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3492                                       int bytes)
3493 {
3494         int i;
3495         int pad = bytes - sizeof(u64) * items;
3496         BUG_ON(pad < 0);
3497
3498         /* Use put_unaligned() because stats may not be aligned for u64. */
3499         put_unaligned(items, &stats[0]);
3500         for (i = 1; i < items; i++)
3501                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3502
3503         memset(&stats[items], 0, pad);
3504 }
3505
3506 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3507                              int bytes)
3508 {
3509         switch(attrtype) {
3510         case IFLA_INET6_STATS:
3511                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3512                 break;
3513         case IFLA_INET6_ICMP6STATS:
3514                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3515                 break;
3516         }
3517 }
3518
3519 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3520                              u32 pid, u32 seq, int event, unsigned int flags)
3521 {
3522         struct net_device *dev = idev->dev;
3523         struct nlattr *nla;
3524         struct ifinfomsg *hdr;
3525         struct nlmsghdr *nlh;
3526         void *protoinfo;
3527         struct ifla_cacheinfo ci;
3528
3529         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3530         if (nlh == NULL)
3531                 return -EMSGSIZE;
3532
3533         hdr = nlmsg_data(nlh);
3534         hdr->ifi_family = AF_INET6;
3535         hdr->__ifi_pad = 0;
3536         hdr->ifi_type = dev->type;
3537         hdr->ifi_index = dev->ifindex;
3538         hdr->ifi_flags = dev_get_flags(dev);
3539         hdr->ifi_change = 0;
3540
3541         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3542
3543         if (dev->addr_len)
3544                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3545
3546         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3547         if (dev->ifindex != dev->iflink)
3548                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3549
3550         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3551         if (protoinfo == NULL)
3552                 goto nla_put_failure;
3553
3554         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3555
3556         ci.max_reasm_len = IPV6_MAXPLEN;
3557         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3558                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3559         ci.reachable_time = idev->nd_parms->reachable_time;
3560         ci.retrans_time = idev->nd_parms->retrans_time;
3561         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3562
3563         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3564         if (nla == NULL)
3565                 goto nla_put_failure;
3566         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3567
3568         /* XXX - MC not implemented */
3569
3570         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3571         if (nla == NULL)
3572                 goto nla_put_failure;
3573         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3574
3575         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3576         if (nla == NULL)
3577                 goto nla_put_failure;
3578         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3579
3580         nla_nest_end(skb, protoinfo);
3581         return nlmsg_end(skb, nlh);
3582
3583 nla_put_failure:
3584         nlmsg_cancel(skb, nlh);
3585         return -EMSGSIZE;
3586 }
3587
3588 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3589 {
3590         int idx, err;
3591         int s_idx = cb->args[0];
3592         struct net_device *dev;
3593         struct inet6_dev *idev;
3594
3595         read_lock(&dev_base_lock);
3596         idx = 0;
3597         for_each_netdev(&init_net, dev) {
3598                 if (idx < s_idx)
3599                         goto cont;
3600                 if ((idev = in6_dev_get(dev)) == NULL)
3601                         goto cont;
3602                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3603                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3604                 in6_dev_put(idev);
3605                 if (err <= 0)
3606                         break;
3607 cont:
3608                 idx++;
3609         }
3610         read_unlock(&dev_base_lock);
3611         cb->args[0] = idx;
3612
3613         return skb->len;
3614 }
3615
3616 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3617 {
3618         struct sk_buff *skb;
3619         int err = -ENOBUFS;
3620
3621         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3622         if (skb == NULL)
3623                 goto errout;
3624
3625         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3626         if (err < 0) {
3627                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3628                 WARN_ON(err == -EMSGSIZE);
3629                 kfree_skb(skb);
3630                 goto errout;
3631         }
3632         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3633 errout:
3634         if (err < 0)
3635                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3636 }
3637
3638 static inline size_t inet6_prefix_nlmsg_size(void)
3639 {
3640         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3641                + nla_total_size(sizeof(struct in6_addr))
3642                + nla_total_size(sizeof(struct prefix_cacheinfo));
3643 }
3644
3645 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3646                              struct prefix_info *pinfo, u32 pid, u32 seq,
3647                              int event, unsigned int flags)
3648 {
3649         struct prefixmsg *pmsg;
3650         struct nlmsghdr *nlh;
3651         struct prefix_cacheinfo ci;
3652
3653         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3654         if (nlh == NULL)
3655                 return -EMSGSIZE;
3656
3657         pmsg = nlmsg_data(nlh);
3658         pmsg->prefix_family = AF_INET6;
3659         pmsg->prefix_pad1 = 0;
3660         pmsg->prefix_pad2 = 0;
3661         pmsg->prefix_ifindex = idev->dev->ifindex;
3662         pmsg->prefix_len = pinfo->prefix_len;
3663         pmsg->prefix_type = pinfo->type;
3664         pmsg->prefix_pad3 = 0;
3665         pmsg->prefix_flags = 0;
3666         if (pinfo->onlink)
3667                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3668         if (pinfo->autoconf)
3669                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3670
3671         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3672
3673         ci.preferred_time = ntohl(pinfo->prefered);
3674         ci.valid_time = ntohl(pinfo->valid);
3675         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3676
3677         return nlmsg_end(skb, nlh);
3678
3679 nla_put_failure:
3680         nlmsg_cancel(skb, nlh);
3681         return -EMSGSIZE;
3682 }
3683
3684 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3685                          struct prefix_info *pinfo)
3686 {
3687         struct sk_buff *skb;
3688         int err = -ENOBUFS;
3689
3690         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3691         if (skb == NULL)
3692                 goto errout;
3693
3694         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3695         if (err < 0) {
3696                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3697                 WARN_ON(err == -EMSGSIZE);
3698                 kfree_skb(skb);
3699                 goto errout;
3700         }
3701         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3702 errout:
3703         if (err < 0)
3704                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3705 }
3706
3707 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3708 {
3709         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3710
3711         switch (event) {
3712         case RTM_NEWADDR:
3713                 /*
3714                  * If the address was optimistic
3715                  * we inserted the route at the start of
3716                  * our DAD process, so we don't need
3717                  * to do it again
3718                  */
3719                 if (!(ifp->rt->rt6i_node))
3720                         ip6_ins_rt(ifp->rt);
3721                 if (ifp->idev->cnf.forwarding)
3722                         addrconf_join_anycast(ifp);
3723                 break;
3724         case RTM_DELADDR:
3725                 if (ifp->idev->cnf.forwarding)
3726                         addrconf_leave_anycast(ifp);
3727                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3728                 dst_hold(&ifp->rt->u.dst);
3729                 if (ip6_del_rt(ifp->rt))
3730                         dst_free(&ifp->rt->u.dst);
3731                 break;
3732         }
3733 }
3734
3735 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3736 {
3737         rcu_read_lock_bh();
3738         if (likely(ifp->idev->dead == 0))
3739                 __ipv6_ifa_notify(event, ifp);
3740         rcu_read_unlock_bh();
3741 }
3742
3743 #ifdef CONFIG_SYSCTL
3744
3745 static
3746 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3747                            void __user *buffer, size_t *lenp, loff_t *ppos)
3748 {
3749         int *valp = ctl->data;
3750         int val = *valp;
3751         int ret;
3752
3753         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3754
3755         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3756                 if (valp != &ipv6_devconf.forwarding) {
3757                         if ((!*valp) ^ (!val)) {
3758                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3759                                 if (idev == NULL)
3760                                         return ret;
3761                                 dev_forward_change(idev);
3762                         }
3763                 } else {
3764                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3765                         addrconf_forward_change();
3766                 }
3767                 if (*valp)
3768                         rt6_purge_dflt_routers();
3769         }
3770
3771         return ret;
3772 }
3773
3774 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3775                                             int __user *name, int nlen,
3776                                             void __user *oldval,
3777                                             size_t __user *oldlenp,
3778                                             void __user *newval, size_t newlen)
3779 {
3780         int *valp = table->data;
3781         int new;
3782
3783         if (!newval || !newlen)
3784                 return 0;
3785         if (newlen != sizeof(int))
3786                 return -EINVAL;
3787         if (get_user(new, (int __user *)newval))
3788                 return -EFAULT;
3789         if (new == *valp)
3790                 return 0;
3791         if (oldval && oldlenp) {
3792                 size_t len;
3793                 if (get_user(len, oldlenp))
3794                         return -EFAULT;
3795                 if (len) {
3796                         if (len > table->maxlen)
3797                                 len = table->maxlen;
3798                         if (copy_to_user(oldval, valp, len))
3799                                 return -EFAULT;
3800                         if (put_user(len, oldlenp))
3801                                 return -EFAULT;
3802                 }
3803         }
3804
3805         if (valp != &ipv6_devconf_dflt.forwarding) {
3806                 if (valp != &ipv6_devconf.forwarding) {
3807                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3808                         int changed;
3809                         if (unlikely(idev == NULL))
3810                                 return -ENODEV;
3811                         changed = (!*valp) ^ (!new);
3812                         *valp = new;
3813                         if (changed)
3814                                 dev_forward_change(idev);
3815                 } else {
3816                         *valp = new;
3817                         addrconf_forward_change();
3818                 }
3819
3820                 if (*valp)
3821                         rt6_purge_dflt_routers();
3822         } else
3823                 *valp = new;
3824
3825         return 1;
3826 }
3827
3828 static struct addrconf_sysctl_table
3829 {
3830         struct ctl_table_header *sysctl_header;
3831         ctl_table addrconf_vars[__NET_IPV6_MAX];
3832         ctl_table addrconf_dev[2];
3833         ctl_table addrconf_conf_dir[2];
3834         ctl_table addrconf_proto_dir[2];
3835         ctl_table addrconf_root_dir[2];
3836 } addrconf_sysctl __read_mostly = {
3837         .sysctl_header = NULL,
3838         .addrconf_vars = {
3839                 {
3840                         .ctl_name       =       NET_IPV6_FORWARDING,
3841                         .procname       =       "forwarding",
3842                         .data           =       &ipv6_devconf.forwarding,
3843                         .maxlen         =       sizeof(int),
3844                         .mode           =       0644,
3845                         .proc_handler   =       &addrconf_sysctl_forward,
3846                         .strategy       =       &addrconf_sysctl_forward_strategy,
3847                 },
3848                 {
3849                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3850                         .procname       =       "hop_limit",
3851                         .data           =       &ipv6_devconf.hop_limit,
3852                         .maxlen         =       sizeof(int),
3853                         .mode           =       0644,
3854                         .proc_handler   =       proc_dointvec,
3855                 },
3856                 {
3857                         .ctl_name       =       NET_IPV6_MTU,
3858                         .procname       =       "mtu",
3859                         .data           =       &ipv6_devconf.mtu6,
3860                         .maxlen         =       sizeof(int),
3861                         .mode           =       0644,
3862                         .proc_handler   =       &proc_dointvec,
3863                 },
3864                 {
3865                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3866                         .procname       =       "accept_ra",
3867                         .data           =       &ipv6_devconf.accept_ra,
3868                         .maxlen         =       sizeof(int),
3869                         .mode           =       0644,
3870                         .proc_handler   =       &proc_dointvec,
3871                 },
3872                 {
3873                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3874                         .procname       =       "accept_redirects",
3875                         .data           =       &ipv6_devconf.accept_redirects,
3876                         .maxlen         =       sizeof(int),
3877                         .mode           =       0644,
3878                         .proc_handler   =       &proc_dointvec,
3879                 },
3880                 {
3881                         .ctl_name       =       NET_IPV6_AUTOCONF,
3882                         .procname       =       "autoconf",
3883                         .data           =       &ipv6_devconf.autoconf,
3884                         .maxlen         =       sizeof(int),
3885                         .mode           =       0644,
3886                         .proc_handler   =       &proc_dointvec,
3887                 },
3888                 {
3889                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3890                         .procname       =       "dad_transmits",
3891                         .data           =       &ipv6_devconf.dad_transmits,
3892                         .maxlen         =       sizeof(int),
3893                         .mode           =       0644,
3894                         .proc_handler   =       &proc_dointvec,
3895                 },
3896                 {
3897                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3898                         .procname       =       "router_solicitations",
3899                         .data           =       &ipv6_devconf.rtr_solicits,
3900                         .maxlen         =       sizeof(int),
3901                         .mode           =       0644,
3902                         .proc_handler   =       &proc_dointvec,
3903                 },
3904                 {
3905                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3906                         .procname       =       "router_solicitation_interval",
3907                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3908                         .maxlen         =       sizeof(int),
3909                         .mode           =       0644,
3910                         .proc_handler   =       &proc_dointvec_jiffies,
3911                         .strategy       =       &sysctl_jiffies,
3912                 },
3913                 {
3914                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3915                         .procname       =       "router_solicitation_delay",
3916                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3917                         .maxlen         =       sizeof(int),
3918                         .mode           =       0644,
3919                         .proc_handler   =       &proc_dointvec_jiffies,
3920                         .strategy       =       &sysctl_jiffies,
3921                 },
3922                 {
3923                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3924                         .procname       =       "force_mld_version",
3925                         .data           =       &ipv6_devconf.force_mld_version,
3926                         .maxlen         =       sizeof(int),
3927                         .mode           =       0644,
3928                         .proc_handler   =       &proc_dointvec,
3929                 },
3930 #ifdef CONFIG_IPV6_PRIVACY
3931                 {
3932                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3933                         .procname       =       "use_tempaddr",
3934                         .data           =       &ipv6_devconf.use_tempaddr,
3935                         .maxlen         =       sizeof(int),
3936                         .mode           =       0644,
3937                         .proc_handler   =       &proc_dointvec,
3938                 },
3939                 {
3940                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3941                         .procname       =       "temp_valid_lft",
3942                         .data           =       &ipv6_devconf.temp_valid_lft,
3943                         .maxlen         =       sizeof(int),
3944                         .mode           =       0644,
3945                         .proc_handler   =       &proc_dointvec,
3946                 },
3947                 {
3948                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3949                         .procname       =       "temp_prefered_lft",
3950                         .data           =       &ipv6_devconf.temp_prefered_lft,
3951                         .maxlen         =       sizeof(int),
3952                         .mode           =       0644,
3953                         .proc_handler   =       &proc_dointvec,
3954                 },
3955                 {
3956                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3957                         .procname       =       "regen_max_retry",
3958                         .data           =       &ipv6_devconf.regen_max_retry,
3959                         .maxlen         =       sizeof(int),
3960                         .mode           =       0644,
3961                         .proc_handler   =       &proc_dointvec,
3962                 },
3963                 {
3964                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3965                         .procname       =       "max_desync_factor",
3966                         .data           =       &ipv6_devconf.max_desync_factor,
3967                         .maxlen         =       sizeof(int),
3968                         .mode           =       0644,
3969                         .proc_handler   =       &proc_dointvec,
3970                 },
3971 #endif
3972                 {
3973                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3974                         .procname       =       "max_addresses",
3975                         .data           =       &ipv6_devconf.max_addresses,
3976                         .maxlen         =       sizeof(int),
3977                         .mode           =       0644,
3978                         .proc_handler   =       &proc_dointvec,
3979                 },
3980                 {
3981                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3982                         .procname       =       "accept_ra_defrtr",
3983                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3984                         .maxlen         =       sizeof(int),
3985                         .mode           =       0644,
3986                         .proc_handler   =       &proc_dointvec,
3987                 },
3988                 {
3989                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3990                         .procname       =       "accept_ra_pinfo",
3991                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3992                         .maxlen         =       sizeof(int),
3993                         .mode           =       0644,
3994                         .proc_handler   =       &proc_dointvec,
3995                 },
3996 #ifdef CONFIG_IPV6_ROUTER_PREF
3997                 {
3998                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3999                         .procname       =       "accept_ra_rtr_pref",
4000                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4001                         .maxlen         =       sizeof(int),
4002                         .mode           =       0644,
4003                         .proc_handler   =       &proc_dointvec,
4004                 },
4005                 {
4006                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4007                         .procname       =       "router_probe_interval",
4008                         .data           =       &ipv6_devconf.rtr_probe_interval,
4009                         .maxlen         =       sizeof(int),
4010                         .mode           =       0644,
4011                         .proc_handler   =       &proc_dointvec_jiffies,
4012                         .strategy       =       &sysctl_jiffies,
4013                 },
4014 #ifdef CONFIG_IPV6_ROUTE_INFO
4015                 {
4016                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4017                         .procname       =       "accept_ra_rt_info_max_plen",
4018                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4019                         .maxlen         =       sizeof(int),
4020                         .mode           =       0644,
4021                         .proc_handler   =       &proc_dointvec,
4022                 },
4023 #endif
4024 #endif
4025                 {
4026                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4027                         .procname       =       "proxy_ndp",
4028                         .data           =       &ipv6_devconf.proxy_ndp,
4029                         .maxlen         =       sizeof(int),
4030                         .mode           =       0644,
4031                         .proc_handler   =       &proc_dointvec,
4032                 },
4033                 {
4034                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4035                         .procname       =       "accept_source_route",
4036                         .data           =       &ipv6_devconf.accept_source_route,
4037                         .maxlen         =       sizeof(int),
4038                         .mode           =       0644,
4039                         .proc_handler   =       &proc_dointvec,
4040                 },
4041 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4042                 {
4043                         .ctl_name       =       CTL_UNNUMBERED,
4044                         .procname       =       "optimistic_dad",
4045                         .data           =       &ipv6_devconf.optimistic_dad,
4046                         .maxlen         =       sizeof(int),
4047                         .mode           =       0644,
4048                         .proc_handler   =       &proc_dointvec,
4049
4050                 },
4051 #endif
4052                 {
4053                         .ctl_name       =       0,      /* sentinel */
4054                 }
4055         },
4056         .addrconf_dev = {
4057                 {
4058                         .ctl_name       =       NET_PROTO_CONF_ALL,
4059                         .procname       =       "all",
4060                         .mode           =       0555,
4061                         .child          =       addrconf_sysctl.addrconf_vars,
4062                 },
4063                 {
4064                         .ctl_name       =       0,      /* sentinel */
4065                 }
4066         },
4067         .addrconf_conf_dir = {
4068                 {
4069                         .ctl_name       =       NET_IPV6_CONF,
4070                         .procname       =       "conf",
4071                         .mode           =       0555,
4072                         .child          =       addrconf_sysctl.addrconf_dev,
4073                 },
4074                 {
4075                         .ctl_name       =       0,      /* sentinel */
4076                 }
4077         },
4078         .addrconf_proto_dir = {
4079                 {
4080                         .ctl_name       =       NET_IPV6,
4081                         .procname       =       "ipv6",
4082                         .mode           =       0555,
4083                         .child          =       addrconf_sysctl.addrconf_conf_dir,
4084                 },
4085                 {
4086                         .ctl_name       =       0,      /* sentinel */
4087                 }
4088         },
4089         .addrconf_root_dir = {
4090                 {
4091                         .ctl_name       =       CTL_NET,
4092                         .procname       =       "net",
4093                         .mode           =       0555,
4094                         .child          =       addrconf_sysctl.addrconf_proto_dir,
4095                 },
4096                 {
4097                         .ctl_name       =       0,      /* sentinel */
4098                 }
4099         },
4100 };
4101
4102 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
4103 {
4104         int i;
4105         struct net_device *dev = idev ? idev->dev : NULL;
4106         struct addrconf_sysctl_table *t;
4107         char *dev_name = NULL;
4108
4109         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4110         if (t == NULL)
4111                 return;
4112         for (i=0; t->addrconf_vars[i].data; i++) {
4113                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4114                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4115         }
4116         if (dev) {
4117                 dev_name = dev->name;
4118                 t->addrconf_dev[0].ctl_name = dev->ifindex;
4119         } else {
4120                 dev_name = "default";
4121                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
4122         }
4123
4124         /*
4125          * Make a copy of dev_name, because '.procname' is regarded as const
4126          * by sysctl and we wouldn't want anyone to change it under our feet
4127          * (see SIOCSIFNAME).
4128          */
4129         dev_name = kstrdup(dev_name, GFP_KERNEL);
4130         if (!dev_name)
4131             goto free;
4132
4133         t->addrconf_dev[0].procname = dev_name;
4134
4135         t->addrconf_dev[0].child = t->addrconf_vars;
4136         t->addrconf_conf_dir[0].child = t->addrconf_dev;
4137         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4138         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4139
4140         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir);
4141         if (t->sysctl_header == NULL)
4142                 goto free_procname;
4143         else
4144                 p->sysctl = t;
4145         return;
4146
4147         /* error path */
4148  free_procname:
4149         kfree(dev_name);
4150  free:
4151         kfree(t);
4152
4153         return;
4154 }
4155
4156 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4157 {
4158         if (p->sysctl) {
4159                 struct addrconf_sysctl_table *t = p->sysctl;
4160                 p->sysctl = NULL;
4161                 unregister_sysctl_table(t->sysctl_header);
4162                 kfree(t->addrconf_dev[0].procname);
4163                 kfree(t);
4164         }
4165 }
4166
4167
4168 #endif
4169
4170 /*
4171  *      Device notifier
4172  */
4173
4174 int register_inet6addr_notifier(struct notifier_block *nb)
4175 {
4176         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4177 }
4178
4179 EXPORT_SYMBOL(register_inet6addr_notifier);
4180
4181 int unregister_inet6addr_notifier(struct notifier_block *nb)
4182 {
4183         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4184 }
4185
4186 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4187
4188 /*
4189  *      Init / cleanup code
4190  */
4191
4192 int __init addrconf_init(void)
4193 {
4194         int err = 0;
4195
4196         /* The addrconf netdev notifier requires that loopback_dev
4197          * has it's ipv6 private information allocated and setup
4198          * before it can bring up and give link-local addresses
4199          * to other devices which are up.
4200          *
4201          * Unfortunately, loopback_dev is not necessarily the first
4202          * entry in the global dev_base list of net devices.  In fact,
4203          * it is likely to be the very last entry on that list.
4204          * So this causes the notifier registry below to try and
4205          * give link-local addresses to all devices besides loopback_dev
4206          * first, then loopback_dev, which cases all the non-loopback_dev
4207          * devices to fail to get a link-local address.
4208          *
4209          * So, as a temporary fix, allocate the ipv6 structure for
4210          * loopback_dev first by hand.
4211          * Longer term, all of the dependencies ipv6 has upon the loopback
4212          * device and it being up should be removed.
4213          */
4214         rtnl_lock();
4215         if (!ipv6_add_dev(init_net.loopback_dev))
4216                 err = -ENOMEM;
4217         rtnl_unlock();
4218         if (err)
4219                 return err;
4220
4221         ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4222         ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4223 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4224         ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4225         ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4226         ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4227         ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4228 #endif
4229
4230         register_netdevice_notifier(&ipv6_dev_notf);
4231
4232         addrconf_verify(0);
4233
4234         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4235         if (err < 0)
4236                 goto errout;
4237
4238         /* Only the first call to __rtnl_register can fail */
4239         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4240         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4241         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4242         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4243         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4244
4245 #ifdef CONFIG_SYSCTL
4246         addrconf_sysctl.sysctl_header =
4247                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir);
4248         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4249 #endif
4250
4251         return 0;
4252 errout:
4253         unregister_netdevice_notifier(&ipv6_dev_notf);
4254
4255         return err;
4256 }
4257
4258 void __exit addrconf_cleanup(void)
4259 {
4260         struct net_device *dev;
4261         struct inet6_ifaddr *ifa;
4262         int i;
4263
4264         unregister_netdevice_notifier(&ipv6_dev_notf);
4265
4266 #ifdef CONFIG_SYSCTL
4267         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4268         addrconf_sysctl_unregister(&ipv6_devconf);
4269 #endif
4270
4271         rtnl_lock();
4272
4273         /*
4274          *      clean dev list.
4275          */
4276
4277         for_each_netdev(&init_net, dev) {
4278                 if (__in6_dev_get(dev) == NULL)
4279                         continue;
4280                 addrconf_ifdown(dev, 1);
4281         }
4282         addrconf_ifdown(init_net.loopback_dev, 2);
4283
4284         /*
4285          *      Check hash table.
4286          */
4287
4288         write_lock_bh(&addrconf_hash_lock);
4289         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4290                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4291                         struct inet6_ifaddr *bifa;
4292
4293                         bifa = ifa;
4294                         ifa = ifa->lst_next;
4295                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4296                         /* Do not free it; something is wrong.
4297                            Now we can investigate it with debugger.
4298                          */
4299                 }
4300         }
4301         write_unlock_bh(&addrconf_hash_lock);
4302
4303         del_timer(&addr_chk_timer);
4304
4305         rtnl_unlock();
4306
4307 #ifdef CONFIG_PROC_FS
4308         proc_net_remove(&init_net, "if_inet6");
4309 #endif
4310 }