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 <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <asm/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #ifdef CONFIG_SYSCTL
 104 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
 105 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
 106 #endif
 107
 108 #ifdef CONFIG_IPV6_PRIVACY
 109 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 110 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 111 static void ipv6_regen_rndid(unsigned long data);
 112
 113 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 114 #endif
 115
 116 static int ipv6_count_addresses(struct inet6_dev *idev);
 117
 118 /*
 119  *      Configured unicast address hash table
 120  */
 121 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 122 static DEFINE_RWLOCK(addrconf_hash_lock);
 123
 124 static void addrconf_verify(unsigned long);
 125
 126 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 127 static DEFINE_SPINLOCK(addrconf_verify_lock);
 128
 129 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 130 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 131
 132 static int addrconf_ifdown(struct net_device *dev, int how);
 133
 134 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 135 static void addrconf_dad_timer(unsigned long data);
 136 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 137 static void addrconf_dad_run(struct inet6_dev *idev);
 138 static void addrconf_rs_timer(unsigned long data);
 139 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 140 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 141
 142 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 143                                 struct prefix_info *pinfo);
 144 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
 145
 146 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 147
 148 struct ipv6_devconf ipv6_devconf __read_mostly = {
 149         .forwarding             = 0,
 150         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 151         .mtu6                   = IPV6_MIN_MTU,
 152         .accept_ra              = 1,
 153         .accept_redirects       = 1,
 154         .autoconf               = 1,
 155         .force_mld_version      = 0,
 156         .dad_transmits          = 1,
 157         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 158         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 159         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 160 #ifdef CONFIG_IPV6_PRIVACY
 161         .use_tempaddr           = 0,
 162         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 163         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 164         .regen_max_retry        = REGEN_MAX_RETRY,
 165         .max_desync_factor      = MAX_DESYNC_FACTOR,
 166 #endif
 167         .max_addresses          = IPV6_MAX_ADDRESSES,
 168         .accept_ra_defrtr       = 1,
 169         .accept_ra_pinfo        = 1,
 170 #ifdef CONFIG_IPV6_ROUTER_PREF
 171         .accept_ra_rtr_pref     = 1,
 172         .rtr_probe_interval     = 60 * HZ,
 173 #ifdef CONFIG_IPV6_ROUTE_INFO
 174         .accept_ra_rt_info_max_plen = 0,
 175 #endif
 176 #endif
 177         .proxy_ndp              = 0,
 178         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 179 };
 180
 181 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 182         .forwarding             = 0,
 183         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 184         .mtu6                   = IPV6_MIN_MTU,
 185         .accept_ra              = 1,
 186         .accept_redirects       = 1,
 187         .autoconf               = 1,
 188         .dad_transmits          = 1,
 189         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 190         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 191         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 192 #ifdef CONFIG_IPV6_PRIVACY
 193         .use_tempaddr           = 0,
 194         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 195         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 196         .regen_max_retry        = REGEN_MAX_RETRY,
 197         .max_desync_factor      = MAX_DESYNC_FACTOR,
 198 #endif
 199         .max_addresses          = IPV6_MAX_ADDRESSES,
 200         .accept_ra_defrtr       = 1,
 201         .accept_ra_pinfo        = 1,
 202 #ifdef CONFIG_IPV6_ROUTER_PREF
 203         .accept_ra_rtr_pref     = 1,
 204         .rtr_probe_interval     = 60 * HZ,
 205 #ifdef CONFIG_IPV6_ROUTE_INFO
 206         .accept_ra_rt_info_max_plen = 0,
 207 #endif
 208 #endif
 209         .proxy_ndp              = 0,
 210         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 211 };
 212
 213 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 214 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 215 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 216
 217 /* Check if a valid qdisc is available */
 218 static inline int addrconf_qdisc_ok(struct net_device *dev)
 219 {
 220         return (dev->qdisc != &noop_qdisc);
 221 }
 222
 223 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 224 {
 225         if (del_timer(&ifp->timer))
 226                 __in6_ifa_put(ifp);
 227 }
 228
 229 enum addrconf_timer_t
 230 {
 231         AC_NONE,
 232         AC_DAD,
 233         AC_RS,
 234 };
 235
 236 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 237                                enum addrconf_timer_t what,
 238                                unsigned long when)
 239 {
 240         if (!del_timer(&ifp->timer))
 241                 in6_ifa_hold(ifp);
 242
 243         switch (what) {
 244         case AC_DAD:
 245                 ifp->timer.function = addrconf_dad_timer;
 246                 break;
 247         case AC_RS:
 248                 ifp->timer.function = addrconf_rs_timer;
 249                 break;
 250         default:;
 251         }
 252         ifp->timer.expires = jiffies + when;
 253         add_timer(&ifp->timer);
 254 }
 255
 256 static int snmp6_alloc_dev(struct inet6_dev *idev)
 257 {
 258         if (snmp_mib_init((void **)idev->stats.ipv6,
 259                           sizeof(struct ipstats_mib),
 260                           __alignof__(struct ipstats_mib)) < 0)
 261                 goto err_ip;
 262         if (snmp_mib_init((void **)idev->stats.icmpv6,
 263                           sizeof(struct icmpv6_mib),
 264                           __alignof__(struct icmpv6_mib)) < 0)
 265                 goto err_icmp;
 266         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 267                           sizeof(struct icmpv6msg_mib),
 268                           __alignof__(struct icmpv6msg_mib)) < 0)
 269                 goto err_icmpmsg;
 270
 271         return 0;
 272
 273 err_icmpmsg:
 274         snmp_mib_free((void **)idev->stats.icmpv6);
 275 err_icmp:
 276         snmp_mib_free((void **)idev->stats.ipv6);
 277 err_ip:
 278         return -ENOMEM;
 279 }
 280
 281 static void snmp6_free_dev(struct inet6_dev *idev)
 282 {
 283         snmp_mib_free((void **)idev->stats.icmpv6msg);
 284         snmp_mib_free((void **)idev->stats.icmpv6);
 285         snmp_mib_free((void **)idev->stats.ipv6);
 286 }
 287
 288 /* Nobody refers to this device, we may destroy it. */
 289
 290 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 291 {
 292         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 293         kfree(idev);
 294 }
 295
 296 void in6_dev_finish_destroy(struct inet6_dev *idev)
 297 {
 298         struct net_device *dev = idev->dev;
 299         BUG_TRAP(idev->addr_list==NULL);
 300         BUG_TRAP(idev->mc_list==NULL);
 301 #ifdef NET_REFCNT_DEBUG
 302         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 303 #endif
 304         dev_put(dev);
 305         if (!idev->dead) {
 306                 printk("Freeing alive inet6 device %p\n", idev);
 307                 return;
 308         }
 309         snmp6_free_dev(idev);
 310         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 311 }
 312
 313 EXPORT_SYMBOL(in6_dev_finish_destroy);
 314
 315 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 316 {
 317         struct inet6_dev *ndev;
 318         struct in6_addr maddr;
 319
 320         ASSERT_RTNL();
 321
 322         if (dev->mtu < IPV6_MIN_MTU)
 323                 return NULL;
 324
 325         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 326
 327         if (ndev == NULL)
 328                 return NULL;
 329
 330         rwlock_init(&ndev->lock);
 331         ndev->dev = dev;
 332         memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
 333         ndev->cnf.mtu6 = dev->mtu;
 334         ndev->cnf.sysctl = NULL;
 335         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 336         if (ndev->nd_parms == NULL) {
 337                 kfree(ndev);
 338                 return NULL;
 339         }
 340         /* We refer to the device */
 341         dev_hold(dev);
 342
 343         if (snmp6_alloc_dev(ndev) < 0) {
 344                 ADBG((KERN_WARNING
 345                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 346                         __FUNCTION__, dev->name));
 347                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 348                 ndev->dead = 1;
 349                 in6_dev_finish_destroy(ndev);
 350                 return NULL;
 351         }
 352
 353         if (snmp6_register_dev(ndev) < 0) {
 354                 ADBG((KERN_WARNING
 355                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 356                         __FUNCTION__, dev->name));
 357                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 358                 ndev->dead = 1;
 359                 in6_dev_finish_destroy(ndev);
 360                 return NULL;
 361         }
 362
 363         /* One reference from device.  We must do this before
 364          * we invoke __ipv6_regen_rndid().
 365          */
 366         in6_dev_hold(ndev);
 367
 368 #ifdef CONFIG_IPV6_PRIVACY
 369         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 370         if ((dev->flags&IFF_LOOPBACK) ||
 371             dev->type == ARPHRD_TUNNEL ||
 372 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 373             dev->type == ARPHRD_SIT ||
 374 #endif
 375             dev->type == ARPHRD_NONE) {
 376                 printk(KERN_INFO
 377                        "%s: Disabled Privacy Extensions\n",
 378                        dev->name);
 379                 ndev->cnf.use_tempaddr = -1;
 380         } else {
 381                 in6_dev_hold(ndev);
 382                 ipv6_regen_rndid((unsigned long) ndev);
 383         }
 384 #endif
 385
 386         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 387                 ndev->if_flags |= IF_READY;
 388
 389         ipv6_mc_init_dev(ndev);
 390         ndev->tstamp = jiffies;
 391 #ifdef CONFIG_SYSCTL
 392         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
 393                               NET_IPV6_NEIGH, "ipv6",
 394                               &ndisc_ifinfo_sysctl_change,
 395                               NULL);
 396         addrconf_sysctl_register(ndev, &ndev->cnf);
 397 #endif
 398         /* protected by rtnl_lock */
 399         rcu_assign_pointer(dev->ip6_ptr, ndev);
 400
 401         /* Join all-node multicast group */
 402         ipv6_addr_all_nodes(&maddr);
 403         ipv6_dev_mc_inc(dev, &maddr);
 404
 405         return ndev;
 406 }
 407
 408 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 409 {
 410         struct inet6_dev *idev;
 411
 412         ASSERT_RTNL();
 413
 414         if ((idev = __in6_dev_get(dev)) == NULL) {
 415                 if ((idev = ipv6_add_dev(dev)) == NULL)
 416                         return NULL;
 417         }
 418
 419         if (dev->flags&IFF_UP)
 420                 ipv6_mc_up(idev);
 421         return idev;
 422 }
 423
 424 #ifdef CONFIG_SYSCTL
 425 static void dev_forward_change(struct inet6_dev *idev)
 426 {
 427         struct net_device *dev;
 428         struct inet6_ifaddr *ifa;
 429         struct in6_addr addr;
 430
 431         if (!idev)
 432                 return;
 433         dev = idev->dev;
 434         if (dev && (dev->flags & IFF_MULTICAST)) {
 435                 ipv6_addr_all_routers(&addr);
 436
 437                 if (idev->cnf.forwarding)
 438                         ipv6_dev_mc_inc(dev, &addr);
 439                 else
 440                         ipv6_dev_mc_dec(dev, &addr);
 441         }
 442         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 443                 if (ifa->flags&IFA_F_TENTATIVE)
 444                         continue;
 445                 if (idev->cnf.forwarding)
 446                         addrconf_join_anycast(ifa);
 447                 else
 448                         addrconf_leave_anycast(ifa);
 449         }
 450 }
 451
 452
 453 static void addrconf_forward_change(void)
 454 {
 455         struct net_device *dev;
 456         struct inet6_dev *idev;
 457
 458         read_lock(&dev_base_lock);
 459         for_each_netdev(&init_net, dev) {
 460                 rcu_read_lock();
 461                 idev = __in6_dev_get(dev);
 462                 if (idev) {
 463                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
 464                         idev->cnf.forwarding = ipv6_devconf.forwarding;
 465                         if (changed)
 466                                 dev_forward_change(idev);
 467                 }
 468                 rcu_read_unlock();
 469         }
 470         read_unlock(&dev_base_lock);
 471 }
 472 #endif
 473
 474 /* Nobody refers to this ifaddr, destroy it */
 475
 476 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 477 {
 478         BUG_TRAP(ifp->if_next==NULL);
 479         BUG_TRAP(ifp->lst_next==NULL);
 480 #ifdef NET_REFCNT_DEBUG
 481         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 482 #endif
 483
 484         in6_dev_put(ifp->idev);
 485
 486         if (del_timer(&ifp->timer))
 487                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 488
 489         if (!ifp->dead) {
 490                 printk("Freeing alive inet6 address %p\n", ifp);
 491                 return;
 492         }
 493         dst_release(&ifp->rt->u.dst);
 494
 495         kfree(ifp);
 496 }
 497
 498 static void
 499 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 500 {
 501         struct inet6_ifaddr *ifa, **ifap;
 502         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 503
 504         /*
 505          * Each device address list is sorted in order of scope -
 506          * global before linklocal.
 507          */
 508         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 509              ifap = &ifa->if_next) {
 510                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 511                         break;
 512         }
 513
 514         ifp->if_next = *ifap;
 515         *ifap = ifp;
 516 }
 517
 518 /* On success it returns ifp with increased reference count */
 519
 520 static struct inet6_ifaddr *
 521 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 522               int scope, u32 flags)
 523 {
 524         struct inet6_ifaddr *ifa = NULL;
 525         struct rt6_info *rt;
 526         int hash;
 527         int err = 0;
 528
 529         rcu_read_lock_bh();
 530         if (idev->dead) {
 531                 err = -ENODEV;                  /*XXX*/
 532                 goto out2;
 533         }
 534
 535         write_lock(&addrconf_hash_lock);
 536
 537         /* Ignore adding duplicate addresses on an interface */
 538         if (ipv6_chk_same_addr(addr, idev->dev)) {
 539                 ADBG(("ipv6_add_addr: already assigned\n"));
 540                 err = -EEXIST;
 541                 goto out;
 542         }
 543
 544         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 545
 546         if (ifa == NULL) {
 547                 ADBG(("ipv6_add_addr: malloc failed\n"));
 548                 err = -ENOBUFS;
 549                 goto out;
 550         }
 551
 552         rt = addrconf_dst_alloc(idev, addr, 0);
 553         if (IS_ERR(rt)) {
 554                 err = PTR_ERR(rt);
 555                 goto out;
 556         }
 557
 558         ipv6_addr_copy(&ifa->addr, addr);
 559
 560         spin_lock_init(&ifa->lock);
 561         init_timer(&ifa->timer);
 562         ifa->timer.data = (unsigned long) ifa;
 563         ifa->scope = scope;
 564         ifa->prefix_len = pfxlen;
 565         ifa->flags = flags | IFA_F_TENTATIVE;
 566         ifa->cstamp = ifa->tstamp = jiffies;
 567
 568         ifa->rt = rt;
 569
 570         /*
 571          * part one of RFC 4429, section 3.3
 572          * We should not configure an address as
 573          * optimistic if we do not yet know the link
 574          * layer address of our nexhop router
 575          */
 576
 577         if (rt->rt6i_nexthop == NULL)
 578                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 579
 580         ifa->idev = idev;
 581         in6_dev_hold(idev);
 582         /* For caller */
 583         in6_ifa_hold(ifa);
 584
 585         /* Add to big hash table */
 586         hash = ipv6_addr_hash(addr);
 587
 588         ifa->lst_next = inet6_addr_lst[hash];
 589         inet6_addr_lst[hash] = ifa;
 590         in6_ifa_hold(ifa);
 591         write_unlock(&addrconf_hash_lock);
 592
 593         write_lock(&idev->lock);
 594         /* Add to inet6_dev unicast addr list. */
 595         ipv6_link_dev_addr(idev, ifa);
 596
 597 #ifdef CONFIG_IPV6_PRIVACY
 598         if (ifa->flags&IFA_F_TEMPORARY) {
 599                 ifa->tmp_next = idev->tempaddr_list;
 600                 idev->tempaddr_list = ifa;
 601                 in6_ifa_hold(ifa);
 602         }
 603 #endif
 604
 605         in6_ifa_hold(ifa);
 606         write_unlock(&idev->lock);
 607 out2:
 608         rcu_read_unlock_bh();
 609
 610         if (likely(err == 0))
 611                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 612         else {
 613                 kfree(ifa);
 614                 ifa = ERR_PTR(err);
 615         }
 616
 617         return ifa;
 618 out:
 619         write_unlock(&addrconf_hash_lock);
 620         goto out2;
 621 }
 622
 623 /* This function wants to get referenced ifp and releases it before return */
 624
 625 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 626 {
 627         struct inet6_ifaddr *ifa, **ifap;
 628         struct inet6_dev *idev = ifp->idev;
 629         int hash;
 630         int deleted = 0, onlink = 0;
 631         unsigned long expires = jiffies;
 632
 633         hash = ipv6_addr_hash(&ifp->addr);
 634
 635         ifp->dead = 1;
 636
 637         write_lock_bh(&addrconf_hash_lock);
 638         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 639              ifap = &ifa->lst_next) {
 640                 if (ifa == ifp) {
 641                         *ifap = ifa->lst_next;
 642                         __in6_ifa_put(ifp);
 643                         ifa->lst_next = NULL;
 644                         break;
 645                 }
 646         }
 647         write_unlock_bh(&addrconf_hash_lock);
 648
 649         write_lock_bh(&idev->lock);
 650 #ifdef CONFIG_IPV6_PRIVACY
 651         if (ifp->flags&IFA_F_TEMPORARY) {
 652                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 653                      ifap = &ifa->tmp_next) {
 654                         if (ifa == ifp) {
 655                                 *ifap = ifa->tmp_next;
 656                                 if (ifp->ifpub) {
 657                                         in6_ifa_put(ifp->ifpub);
 658                                         ifp->ifpub = NULL;
 659                                 }
 660                                 __in6_ifa_put(ifp);
 661                                 ifa->tmp_next = NULL;
 662                                 break;
 663                         }
 664                 }
 665         }
 666 #endif
 667
 668         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 669                 if (ifa == ifp) {
 670                         *ifap = ifa->if_next;
 671                         __in6_ifa_put(ifp);
 672                         ifa->if_next = NULL;
 673                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 674                                 break;
 675                         deleted = 1;
 676                         continue;
 677                 } else if (ifp->flags & IFA_F_PERMANENT) {
 678                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 679                                               ifp->prefix_len)) {
 680                                 if (ifa->flags & IFA_F_PERMANENT) {
 681                                         onlink = 1;
 682                                         if (deleted)
 683                                                 break;
 684                                 } else {
 685                                         unsigned long lifetime;
 686
 687                                         if (!onlink)
 688                                                 onlink = -1;
 689
 690                                         spin_lock(&ifa->lock);
 691                                         lifetime = min_t(unsigned long,
 692                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 693                                         if (time_before(expires,
 694                                                         ifa->tstamp + lifetime * HZ))
 695                                                 expires = ifa->tstamp + lifetime * HZ;
 696                                         spin_unlock(&ifa->lock);
 697                                 }
 698                         }
 699                 }
 700                 ifap = &ifa->if_next;
 701         }
 702         write_unlock_bh(&idev->lock);
 703
 704         ipv6_ifa_notify(RTM_DELADDR, ifp);
 705
 706         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 707
 708         addrconf_del_timer(ifp);
 709
 710         /*
 711          * Purge or update corresponding prefix
 712          *
 713          * 1) we don't purge prefix here if address was not permanent.
 714          *    prefix is managed by its own lifetime.
 715          * 2) if there're no addresses, delete prefix.
 716          * 3) if there're still other permanent address(es),
 717          *    corresponding prefix is still permanent.
 718          * 4) otherwise, update prefix lifetime to the
 719          *    longest valid lifetime among the corresponding
 720          *    addresses on the device.
 721          *    Note: subsequent RA will update lifetime.
 722          *
 723          * --yoshfuji
 724          */
 725         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 726                 struct in6_addr prefix;
 727                 struct rt6_info *rt;
 728
 729                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 730                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
 731
 732                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 733                         if (onlink == 0) {
 734                                 ip6_del_rt(rt);
 735                                 rt = NULL;
 736                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 737                                 rt->rt6i_expires = expires;
 738                                 rt->rt6i_flags |= RTF_EXPIRES;
 739                         }
 740                 }
 741                 dst_release(&rt->u.dst);
 742         }
 743
 744         in6_ifa_put(ifp);
 745 }
 746
 747 #ifdef CONFIG_IPV6_PRIVACY
 748 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 749 {
 750         struct inet6_dev *idev = ifp->idev;
 751         struct in6_addr addr, *tmpaddr;
 752         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 753         int tmp_plen;
 754         int ret = 0;
 755         int max_addresses;
 756         u32 addr_flags;
 757
 758         write_lock(&idev->lock);
 759         if (ift) {
 760                 spin_lock_bh(&ift->lock);
 761                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 762                 spin_unlock_bh(&ift->lock);
 763                 tmpaddr = &addr;
 764         } else {
 765                 tmpaddr = NULL;
 766         }
 767 retry:
 768         in6_dev_hold(idev);
 769         if (idev->cnf.use_tempaddr <= 0) {
 770                 write_unlock(&idev->lock);
 771                 printk(KERN_INFO
 772                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 773                 in6_dev_put(idev);
 774                 ret = -1;
 775                 goto out;
 776         }
 777         spin_lock_bh(&ifp->lock);
 778         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 779                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 780                 spin_unlock_bh(&ifp->lock);
 781                 write_unlock(&idev->lock);
 782                 printk(KERN_WARNING
 783                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 784                 in6_dev_put(idev);
 785                 ret = -1;
 786                 goto out;
 787         }
 788         in6_ifa_hold(ifp);
 789         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 790         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 791                 spin_unlock_bh(&ifp->lock);
 792                 write_unlock(&idev->lock);
 793                 printk(KERN_WARNING
 794                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 795                 in6_ifa_put(ifp);
 796                 in6_dev_put(idev);
 797                 ret = -1;
 798                 goto out;
 799         }
 800         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 801         tmp_valid_lft = min_t(__u32,
 802                               ifp->valid_lft,
 803                               idev->cnf.temp_valid_lft);
 804         tmp_prefered_lft = min_t(__u32,
 805                                  ifp->prefered_lft,
 806                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 807         tmp_plen = ifp->prefix_len;
 808         max_addresses = idev->cnf.max_addresses;
 809         tmp_cstamp = ifp->cstamp;
 810         tmp_tstamp = ifp->tstamp;
 811         spin_unlock_bh(&ifp->lock);
 812
 813         write_unlock(&idev->lock);
 814
 815         addr_flags = IFA_F_TEMPORARY;
 816         /* set in addrconf_prefix_rcv() */
 817         if (ifp->flags & IFA_F_OPTIMISTIC)
 818                 addr_flags |= IFA_F_OPTIMISTIC;
 819
 820         ift = !max_addresses ||
 821               ipv6_count_addresses(idev) < max_addresses ?
 822                 ipv6_add_addr(idev, &addr, tmp_plen,
 823                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 824                               addr_flags) : NULL;
 825         if (!ift || IS_ERR(ift)) {
 826                 in6_ifa_put(ifp);
 827                 in6_dev_put(idev);
 828                 printk(KERN_INFO
 829                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 830                 tmpaddr = &addr;
 831                 write_lock(&idev->lock);
 832                 goto retry;
 833         }
 834
 835         spin_lock_bh(&ift->lock);
 836         ift->ifpub = ifp;
 837         ift->valid_lft = tmp_valid_lft;
 838         ift->prefered_lft = tmp_prefered_lft;
 839         ift->cstamp = tmp_cstamp;
 840         ift->tstamp = tmp_tstamp;
 841         spin_unlock_bh(&ift->lock);
 842
 843         addrconf_dad_start(ift, 0);
 844         in6_ifa_put(ift);
 845         in6_dev_put(idev);
 846 out:
 847         return ret;
 848 }
 849 #endif
 850
 851 /*
 852  *      Choose an appropriate source address (RFC3484)
 853  */
 854 struct ipv6_saddr_score {
 855         int             addr_type;
 856         unsigned int    attrs;
 857         int             matchlen;
 858         int             scope;
 859         unsigned int    rule;
 860 };
 861
 862 #define IPV6_SADDR_SCORE_LOCAL          0x0001
 863 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
 864 #define IPV6_SADDR_SCORE_HOA            0x0008
 865 #define IPV6_SADDR_SCORE_OIF            0x0010
 866 #define IPV6_SADDR_SCORE_LABEL          0x0020
 867 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
 868
 869 static inline int ipv6_saddr_preferred(int type)
 870 {
 871         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 872                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 873                 return 1;
 874         return 0;
 875 }
 876
 877 /* static matching label */
 878 static inline int ipv6_addr_label(const struct in6_addr *addr, int type,
 879                                   int ifindex)
 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         int daddr_ifindex = daddr_dev ? daddr_dev->ifindex : 0;
 915         u32 daddr_label = ipv6_addr_label(daddr, daddr_type, daddr_ifindex);
 916         struct net_device *dev;
 917
 918         memset(&hiscore, 0, sizeof(hiscore));
 919
 920         read_lock(&dev_base_lock);
 921         rcu_read_lock();
 922
 923         for_each_netdev(&init_net, dev) {
 924                 struct inet6_dev *idev;
 925                 struct inet6_ifaddr *ifa;
 926
 927                 /* Rule 0: Candidate Source Address (section 4)
 928                  *  - multicast and link-local destination address,
 929                  *    the set of candidate source address MUST only
 930                  *    include addresses assigned to interfaces
 931                  *    belonging to the same link as the outgoing
 932                  *    interface.
 933                  * (- For site-local destination addresses, the
 934                  *    set of candidate source addresses MUST only
 935                  *    include addresses assigned to interfaces
 936                  *    belonging to the same site as the outgoing
 937                  *    interface.)
 938                  */
 939                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
 940                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
 941                     daddr_dev && dev != daddr_dev)
 942                         continue;
 943
 944                 idev = __in6_dev_get(dev);
 945                 if (!idev)
 946                         continue;
 947
 948                 read_lock_bh(&idev->lock);
 949                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
 950                         struct ipv6_saddr_score score;
 951
 952                         score.addr_type = __ipv6_addr_type(&ifa->addr);
 953
 954                         /* Rule 0:
 955                          * - Tentative Address (RFC2462 section 5.4)
 956                          *  - A tentative address is not considered
 957                          *    "assigned to an interface" in the traditional
 958                          *    sense, unless it is also flagged as optimistic.
 959                          * - Candidate Source Address (section 4)
 960                          *  - In any case, anycast addresses, multicast
 961                          *    addresses, and the unspecified address MUST
 962                          *    NOT be included in a candidate set.
 963                          */
 964                         if ((ifa->flags & IFA_F_TENTATIVE) &&
 965                             (!(ifa->flags & IFA_F_OPTIMISTIC)))
 966                                 continue;
 967                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
 968                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
 969                                 LIMIT_NETDEBUG(KERN_DEBUG
 970                                                "ADDRCONF: unspecified / multicast address "
 971                                                "assigned as unicast address on %s",
 972                                                dev->name);
 973                                 continue;
 974                         }
 975
 976                         score.attrs = 0;
 977                         score.matchlen = 0;
 978                         score.scope = 0;
 979                         score.rule = 0;
 980
 981                         if (ifa_result == NULL) {
 982                                 /* record it if the first available entry */
 983                                 goto record_it;
 984                         }
 985
 986                         /* Rule 1: Prefer same address */
 987                         if (hiscore.rule < 1) {
 988                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
 989                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
 990                                 hiscore.rule++;
 991                         }
 992                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
 993                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
 994                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
 995                                         score.rule = 1;
 996                                         goto record_it;
 997                                 }
 998                         } else {
 999                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
1000                                         continue;
1001                         }
1002
1003                         /* Rule 2: Prefer appropriate scope */
1004                         if (hiscore.rule < 2) {
1005                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1006                                 hiscore.rule++;
1007                         }
1008                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1009                         if (hiscore.scope < score.scope) {
1010                                 if (hiscore.scope < daddr_scope) {
1011                                         score.rule = 2;
1012                                         goto record_it;
1013                                 } else
1014                                         continue;
1015                         } else if (score.scope < hiscore.scope) {
1016                                 if (score.scope < daddr_scope)
1017                                         break; /* addresses sorted by scope */
1018                                 else {
1019                                         score.rule = 2;
1020                                         goto record_it;
1021                                 }
1022                         }
1023
1024                         /* Rule 3: Avoid deprecated and optimistic addresses */
1025                         if (hiscore.rule < 3) {
1026                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1027                                    (((ifa_result->flags &
1028                                     (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0)))
1029                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1030                                 hiscore.rule++;
1031                         }
1032                         if (ipv6_saddr_preferred(score.addr_type) ||
1033                            (((ifa->flags &
1034                             (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)) == 0))) {
1035                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1036                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1037                                         score.rule = 3;
1038                                         goto record_it;
1039                                 }
1040                         } else {
1041                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1042                                         continue;
1043                         }
1044
1045                         /* Rule 4: Prefer home address */
1046 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
1047                         if (hiscore.rule < 4) {
1048                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1049                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1050                                 hiscore.rule++;
1051                         }
1052                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1053                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1054                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1055                                         score.rule = 4;
1056                                         goto record_it;
1057                                 }
1058                         } else {
1059                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1060                                         continue;
1061                         }
1062 #else
1063                         if (hiscore.rule < 4)
1064                                 hiscore.rule++;
1065 #endif
1066
1067                         /* Rule 5: Prefer outgoing interface */
1068                         if (hiscore.rule < 5) {
1069                                 if (daddr_dev == NULL ||
1070                                     daddr_dev == ifa_result->idev->dev)
1071                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1072                                 hiscore.rule++;
1073                         }
1074                         if (daddr_dev == NULL ||
1075                             daddr_dev == ifa->idev->dev) {
1076                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1077                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1078                                         score.rule = 5;
1079                                         goto record_it;
1080                                 }
1081                         } else {
1082                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1083                                         continue;
1084                         }
1085
1086                         /* Rule 6: Prefer matching label */
1087                         if (hiscore.rule < 6) {
1088                                 if (ipv6_addr_label(&ifa_result->addr,
1089                                                     hiscore.addr_type,
1090                                                     ifa_result->idev->dev->ifindex) == daddr_label)
1091                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1092                                 hiscore.rule++;
1093                         }
1094                         if (ipv6_addr_label(&ifa->addr,
1095                                             score.addr_type,
1096                                             ifa->idev->dev->ifindex) == daddr_label) {
1097                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1098                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1099                                         score.rule = 6;
1100                                         goto record_it;
1101                                 }
1102                         } else {
1103                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1104                                         continue;
1105                         }
1106
1107 #ifdef CONFIG_IPV6_PRIVACY
1108                         /* Rule 7: Prefer public address
1109                          * Note: prefer temprary address if use_tempaddr >= 2
1110                          */
1111                         if (hiscore.rule < 7) {
1112                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1113                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1114                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1115                                 hiscore.rule++;
1116                         }
1117                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1118                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1119                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1120                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1121                                         score.rule = 7;
1122                                         goto record_it;
1123                                 }
1124                         } else {
1125                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1126                                         continue;
1127                         }
1128 #else
1129                         if (hiscore.rule < 7)
1130                                 hiscore.rule++;
1131 #endif
1132                         /* Rule 8: Use longest matching prefix */
1133                         if (hiscore.rule < 8) {
1134                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1135                                 hiscore.rule++;
1136                         }
1137                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1138                         if (score.matchlen > hiscore.matchlen) {
1139                                 score.rule = 8;
1140                                 goto record_it;
1141                         }
1142 #if 0
1143                         else if (score.matchlen < hiscore.matchlen)
1144                                 continue;
1145 #endif
1146
1147                         /* Final Rule: choose first available one */
1148                         continue;
1149 record_it:
1150                         if (ifa_result)
1151                                 in6_ifa_put(ifa_result);
1152                         in6_ifa_hold(ifa);
1153                         ifa_result = ifa;
1154                         hiscore = score;
1155                 }
1156                 read_unlock_bh(&idev->lock);
1157         }
1158         rcu_read_unlock();
1159         read_unlock(&dev_base_lock);
1160
1161         if (!ifa_result)
1162                 return -EADDRNOTAVAIL;
1163
1164         ipv6_addr_copy(saddr, &ifa_result->addr);
1165         in6_ifa_put(ifa_result);
1166         return 0;
1167 }
1168
1169
1170 int ipv6_get_saddr(struct dst_entry *dst,
1171                    struct in6_addr *daddr, struct in6_addr *saddr)
1172 {
1173         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1174 }
1175
1176 EXPORT_SYMBOL(ipv6_get_saddr);
1177
1178 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1179                     unsigned char banned_flags)
1180 {
1181         struct inet6_dev *idev;
1182         int err = -EADDRNOTAVAIL;
1183
1184         rcu_read_lock();
1185         if ((idev = __in6_dev_get(dev)) != NULL) {
1186                 struct inet6_ifaddr *ifp;
1187
1188                 read_lock_bh(&idev->lock);
1189                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1190                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1191                                 ipv6_addr_copy(addr, &ifp->addr);
1192                                 err = 0;
1193                                 break;
1194                         }
1195                 }
1196                 read_unlock_bh(&idev->lock);
1197         }
1198         rcu_read_unlock();
1199         return err;
1200 }
1201
1202 static int ipv6_count_addresses(struct inet6_dev *idev)
1203 {
1204         int cnt = 0;
1205         struct inet6_ifaddr *ifp;
1206
1207         read_lock_bh(&idev->lock);
1208         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1209                 cnt++;
1210         read_unlock_bh(&idev->lock);
1211         return cnt;
1212 }
1213
1214 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1215 {
1216         struct inet6_ifaddr * ifp;
1217         u8 hash = ipv6_addr_hash(addr);
1218
1219         read_lock_bh(&addrconf_hash_lock);
1220         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1221                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1222                     !(ifp->flags&IFA_F_TENTATIVE)) {
1223                         if (dev == NULL || ifp->idev->dev == dev ||
1224                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1225                                 break;
1226                 }
1227         }
1228         read_unlock_bh(&addrconf_hash_lock);
1229         return ifp != NULL;
1230 }
1231
1232 EXPORT_SYMBOL(ipv6_chk_addr);
1233
1234 static
1235 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1236 {
1237         struct inet6_ifaddr * ifp;
1238         u8 hash = ipv6_addr_hash(addr);
1239
1240         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1241                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1242                         if (dev == NULL || ifp->idev->dev == dev)
1243                                 break;
1244                 }
1245         }
1246         return ifp != NULL;
1247 }
1248
1249 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1250 {
1251         struct inet6_ifaddr * ifp;
1252         u8 hash = ipv6_addr_hash(addr);
1253
1254         read_lock_bh(&addrconf_hash_lock);
1255         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1256                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1257                         if (dev == NULL || ifp->idev->dev == dev ||
1258                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1259                                 in6_ifa_hold(ifp);
1260                                 break;
1261                         }
1262                 }
1263         }
1264         read_unlock_bh(&addrconf_hash_lock);
1265
1266         return ifp;
1267 }
1268
1269 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1270 {
1271         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1272         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1273         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1274         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1275         int sk_ipv6only = ipv6_only_sock(sk);
1276         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1277         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1278         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1279
1280         if (!sk2_rcv_saddr && !sk_ipv6only)
1281                 return 1;
1282
1283         if (addr_type2 == IPV6_ADDR_ANY &&
1284             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1285                 return 1;
1286
1287         if (addr_type == IPV6_ADDR_ANY &&
1288             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1289                 return 1;
1290
1291         if (sk2_rcv_saddr6 &&
1292             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1293                 return 1;
1294
1295         if (addr_type == IPV6_ADDR_MAPPED &&
1296             !sk2_ipv6only &&
1297             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1298                 return 1;
1299
1300         return 0;
1301 }
1302
1303 /* Gets referenced address, destroys ifaddr */
1304
1305 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1306 {
1307         if (ifp->flags&IFA_F_PERMANENT) {
1308                 spin_lock_bh(&ifp->lock);
1309                 addrconf_del_timer(ifp);
1310                 ifp->flags |= IFA_F_TENTATIVE;
1311                 spin_unlock_bh(&ifp->lock);
1312                 in6_ifa_put(ifp);
1313 #ifdef CONFIG_IPV6_PRIVACY
1314         } else if (ifp->flags&IFA_F_TEMPORARY) {
1315                 struct inet6_ifaddr *ifpub;
1316                 spin_lock_bh(&ifp->lock);
1317                 ifpub = ifp->ifpub;
1318                 if (ifpub) {
1319                         in6_ifa_hold(ifpub);
1320                         spin_unlock_bh(&ifp->lock);
1321                         ipv6_create_tempaddr(ifpub, ifp);
1322                         in6_ifa_put(ifpub);
1323                 } else {
1324                         spin_unlock_bh(&ifp->lock);
1325                 }
1326                 ipv6_del_addr(ifp);
1327 #endif
1328         } else
1329                 ipv6_del_addr(ifp);
1330 }
1331
1332 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1333 {
1334         if (net_ratelimit())
1335                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1336         addrconf_dad_stop(ifp);
1337 }
1338
1339 /* Join to solicited addr multicast group. */
1340
1341 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1342 {
1343         struct in6_addr maddr;
1344
1345         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1346                 return;
1347
1348         addrconf_addr_solict_mult(addr, &maddr);
1349         ipv6_dev_mc_inc(dev, &maddr);
1350 }
1351
1352 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1353 {
1354         struct in6_addr maddr;
1355
1356         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1357                 return;
1358
1359         addrconf_addr_solict_mult(addr, &maddr);
1360         __ipv6_dev_mc_dec(idev, &maddr);
1361 }
1362
1363 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1364 {
1365         struct in6_addr addr;
1366         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1367         if (ipv6_addr_any(&addr))
1368                 return;
1369         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1370 }
1371
1372 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1373 {
1374         struct in6_addr addr;
1375         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1376         if (ipv6_addr_any(&addr))
1377                 return;
1378         __ipv6_dev_ac_dec(ifp->idev, &addr);
1379 }
1380
1381 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1382 {
1383         if (dev->addr_len != ETH_ALEN)
1384                 return -1;
1385         memcpy(eui, dev->dev_addr, 3);
1386         memcpy(eui + 5, dev->dev_addr + 3, 3);
1387
1388         /*
1389          * The zSeries OSA network cards can be shared among various
1390          * OS instances, but the OSA cards have only one MAC address.
1391          * This leads to duplicate address conflicts in conjunction
1392          * with IPv6 if more than one instance uses the same card.
1393          *
1394          * The driver for these cards can deliver a unique 16-bit
1395          * identifier for each instance sharing the same card.  It is
1396          * placed instead of 0xFFFE in the interface identifier.  The
1397          * "u" bit of the interface identifier is not inverted in this
1398          * case.  Hence the resulting interface identifier has local
1399          * scope according to RFC2373.
1400          */
1401         if (dev->dev_id) {
1402                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1403                 eui[4] = dev->dev_id & 0xFF;
1404         } else {
1405                 eui[3] = 0xFF;
1406                 eui[4] = 0xFE;
1407                 eui[0] ^= 2;
1408         }
1409         return 0;
1410 }
1411
1412 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1413 {
1414         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1415         if (dev->addr_len != ARCNET_ALEN)
1416                 return -1;
1417         memset(eui, 0, 7);
1418         eui[7] = *(u8*)dev->dev_addr;
1419         return 0;
1420 }
1421
1422 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1423 {
1424         if (dev->addr_len != INFINIBAND_ALEN)
1425                 return -1;
1426         memcpy(eui, dev->dev_addr + 12, 8);
1427         eui[0] |= 2;
1428         return 0;
1429 }
1430
1431 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1432 {
1433         switch (dev->type) {
1434         case ARPHRD_ETHER:
1435         case ARPHRD_FDDI:
1436         case ARPHRD_IEEE802_TR:
1437                 return addrconf_ifid_eui48(eui, dev);
1438         case ARPHRD_ARCNET:
1439                 return addrconf_ifid_arcnet(eui, dev);
1440         case ARPHRD_INFINIBAND:
1441                 return addrconf_ifid_infiniband(eui, dev);
1442         }
1443         return -1;
1444 }
1445
1446 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1447 {
1448         int err = -1;
1449         struct inet6_ifaddr *ifp;
1450
1451         read_lock_bh(&idev->lock);
1452         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1453                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1454                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1455                         err = 0;
1456                         break;
1457                 }
1458         }
1459         read_unlock_bh(&idev->lock);
1460         return err;
1461 }
1462
1463 #ifdef CONFIG_IPV6_PRIVACY
1464 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1465 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1466 {
1467 regen:
1468         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1469         idev->rndid[0] &= ~0x02;
1470
1471         /*
1472          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1473          * check if generated address is not inappropriate
1474          *
1475          *  - Reserved subnet anycast (RFC 2526)
1476          *      11111101 11....11 1xxxxxxx
1477          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1478          *      00-00-5E-FE-xx-xx-xx-xx
1479          *  - value 0
1480          *  - XXX: already assigned to an address on the device
1481          */
1482         if (idev->rndid[0] == 0xfd &&
1483             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1484             (idev->rndid[7]&0x80))
1485                 goto regen;
1486         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1487                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1488                         goto regen;
1489                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1490                         goto regen;
1491         }
1492
1493         return 0;
1494 }
1495
1496 static void ipv6_regen_rndid(unsigned long data)
1497 {
1498         struct inet6_dev *idev = (struct inet6_dev *) data;
1499         unsigned long expires;
1500
1501         rcu_read_lock_bh();
1502         write_lock_bh(&idev->lock);
1503
1504         if (idev->dead)
1505                 goto out;
1506
1507         if (__ipv6_regen_rndid(idev) < 0)
1508                 goto out;
1509
1510         expires = jiffies +
1511                 idev->cnf.temp_prefered_lft * HZ -
1512                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1513         if (time_before(expires, jiffies)) {
1514                 printk(KERN_WARNING
1515                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1516                         idev->dev->name);
1517                 goto out;
1518         }
1519
1520         if (!mod_timer(&idev->regen_timer, expires))
1521                 in6_dev_hold(idev);
1522
1523 out:
1524         write_unlock_bh(&idev->lock);
1525         rcu_read_unlock_bh();
1526         in6_dev_put(idev);
1527 }
1528
1529 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1530         int ret = 0;
1531
1532         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1533                 ret = __ipv6_regen_rndid(idev);
1534         return ret;
1535 }
1536 #endif
1537
1538 /*
1539  *      Add prefix route.
1540  */
1541
1542 static void
1543 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1544                       unsigned long expires, u32 flags)
1545 {
1546         struct fib6_config cfg = {
1547                 .fc_table = RT6_TABLE_PREFIX,
1548                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1549                 .fc_ifindex = dev->ifindex,
1550                 .fc_expires = expires,
1551                 .fc_dst_len = plen,
1552                 .fc_flags = RTF_UP | flags,
1553         };
1554
1555         ipv6_addr_copy(&cfg.fc_dst, pfx);
1556
1557         /* Prevent useless cloning on PtP SIT.
1558            This thing is done here expecting that the whole
1559            class of non-broadcast devices need not cloning.
1560          */
1561 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1562         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1563                 cfg.fc_flags |= RTF_NONEXTHOP;
1564 #endif
1565
1566         ip6_route_add(&cfg);
1567 }
1568
1569 /* Create "default" multicast route to the interface */
1570
1571 static void addrconf_add_mroute(struct net_device *dev)
1572 {
1573         struct fib6_config cfg = {
1574                 .fc_table = RT6_TABLE_LOCAL,
1575                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1576                 .fc_ifindex = dev->ifindex,
1577                 .fc_dst_len = 8,
1578                 .fc_flags = RTF_UP,
1579         };
1580
1581         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1582
1583         ip6_route_add(&cfg);
1584 }
1585
1586 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1587 static void sit_route_add(struct net_device *dev)
1588 {
1589         struct fib6_config cfg = {
1590                 .fc_table = RT6_TABLE_MAIN,
1591                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1592                 .fc_ifindex = dev->ifindex,
1593                 .fc_dst_len = 96,
1594                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1595         };
1596
1597         /* prefix length - 96 bits "::d.d.d.d" */
1598         ip6_route_add(&cfg);
1599 }
1600 #endif
1601
1602 static void addrconf_add_lroute(struct net_device *dev)
1603 {
1604         struct in6_addr addr;
1605
1606         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1607         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1608 }
1609
1610 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1611 {
1612         struct inet6_dev *idev;
1613
1614         ASSERT_RTNL();
1615
1616         if ((idev = ipv6_find_idev(dev)) == NULL)
1617                 return NULL;
1618
1619         /* Add default multicast route */
1620         addrconf_add_mroute(dev);
1621
1622         /* Add link local route */
1623         addrconf_add_lroute(dev);
1624         return idev;
1625 }
1626
1627 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1628 {
1629         struct prefix_info *pinfo;
1630         __u32 valid_lft;
1631         __u32 prefered_lft;
1632         int addr_type;
1633         unsigned long rt_expires;
1634         struct inet6_dev *in6_dev;
1635
1636         pinfo = (struct prefix_info *) opt;
1637
1638         if (len < sizeof(struct prefix_info)) {
1639                 ADBG(("addrconf: prefix option too short\n"));
1640                 return;
1641         }
1642
1643         /*
1644          *      Validation checks ([ADDRCONF], page 19)
1645          */
1646
1647         addr_type = ipv6_addr_type(&pinfo->prefix);
1648
1649         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1650                 return;
1651
1652         valid_lft = ntohl(pinfo->valid);
1653         prefered_lft = ntohl(pinfo->prefered);
1654
1655         if (prefered_lft > valid_lft) {
1656                 if (net_ratelimit())
1657                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1658                 return;
1659         }
1660
1661         in6_dev = in6_dev_get(dev);
1662
1663         if (in6_dev == NULL) {
1664                 if (net_ratelimit())
1665                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1666                 return;
1667         }
1668
1669         /*
1670          *      Two things going on here:
1671          *      1) Add routes for on-link prefixes
1672          *      2) Configure prefixes with the auto flag set
1673          */
1674
1675         /* Avoid arithmetic overflow. Really, we could
1676            save rt_expires in seconds, likely valid_lft,
1677            but it would require division in fib gc, that it
1678            not good.
1679          */
1680         if (valid_lft >= 0x7FFFFFFF/HZ)
1681                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1682         else
1683                 rt_expires = valid_lft * HZ;
1684
1685         /*
1686          * We convert this (in jiffies) to clock_t later.
1687          * Avoid arithmetic overflow there as well.
1688          * Overflow can happen only if HZ < USER_HZ.
1689          */
1690         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1691                 rt_expires = 0x7FFFFFFF / USER_HZ;
1692
1693         if (pinfo->onlink) {
1694                 struct rt6_info *rt;
1695                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1696
1697                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1698                         if (rt->rt6i_flags&RTF_EXPIRES) {
1699                                 if (valid_lft == 0) {
1700                                         ip6_del_rt(rt);
1701                                         rt = NULL;
1702                                 } else {
1703                                         rt->rt6i_expires = jiffies + rt_expires;
1704                                 }
1705                         }
1706                 } else if (valid_lft) {
1707                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1708                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1709                 }
1710                 if (rt)
1711                         dst_release(&rt->u.dst);
1712         }
1713
1714         /* Try to figure out our local address for this prefix */
1715
1716         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1717                 struct inet6_ifaddr * ifp;
1718                 struct in6_addr addr;
1719                 int create = 0, update_lft = 0;
1720
1721                 if (pinfo->prefix_len == 64) {
1722                         memcpy(&addr, &pinfo->prefix, 8);
1723                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1724                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1725                                 in6_dev_put(in6_dev);
1726                                 return;
1727                         }
1728                         goto ok;
1729                 }
1730                 if (net_ratelimit())
1731                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1732                                pinfo->prefix_len);
1733                 in6_dev_put(in6_dev);
1734                 return;
1735
1736 ok:
1737
1738                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1739
1740                 if (ifp == NULL && valid_lft) {
1741                         int max_addresses = in6_dev->cnf.max_addresses;
1742                         u32 addr_flags = 0;
1743
1744 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1745                         if (in6_dev->cnf.optimistic_dad &&
1746                             !ipv6_devconf.forwarding)
1747                                 addr_flags = IFA_F_OPTIMISTIC;
1748 #endif
1749
1750                         /* Do not allow to create too much of autoconfigured
1751                          * addresses; this would be too easy way to crash kernel.
1752                          */
1753                         if (!max_addresses ||
1754                             ipv6_count_addresses(in6_dev) < max_addresses)
1755                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1756                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1757                                                     addr_flags);
1758
1759                         if (!ifp || IS_ERR(ifp)) {
1760                                 in6_dev_put(in6_dev);
1761                                 return;
1762                         }
1763
1764                         update_lft = create = 1;
1765                         ifp->cstamp = jiffies;
1766                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1767                 }
1768
1769                 if (ifp) {
1770                         int flags;
1771                         unsigned long now;
1772 #ifdef CONFIG_IPV6_PRIVACY
1773                         struct inet6_ifaddr *ift;
1774 #endif
1775                         u32 stored_lft;
1776
1777                         /* update lifetime (RFC2462 5.5.3 e) */
1778                         spin_lock(&ifp->lock);
1779                         now = jiffies;
1780                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1781                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1782                         else
1783                                 stored_lft = 0;
1784                         if (!update_lft && stored_lft) {
1785                                 if (valid_lft > MIN_VALID_LIFETIME ||
1786                                     valid_lft > stored_lft)
1787                                         update_lft = 1;
1788                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1789                                         /* valid_lft <= stored_lft is always true */
1790                                         /* XXX: IPsec */
1791                                         update_lft = 0;
1792                                 } else {
1793                                         valid_lft = MIN_VALID_LIFETIME;
1794                                         if (valid_lft < prefered_lft)
1795                                                 prefered_lft = valid_lft;
1796                                         update_lft = 1;
1797                                 }
1798                         }
1799
1800                         if (update_lft) {
1801                                 ifp->valid_lft = valid_lft;
1802                                 ifp->prefered_lft = prefered_lft;
1803                                 ifp->tstamp = now;
1804                                 flags = ifp->flags;
1805                                 ifp->flags &= ~IFA_F_DEPRECATED;
1806                                 spin_unlock(&ifp->lock);
1807
1808                                 if (!(flags&IFA_F_TENTATIVE))
1809                                         ipv6_ifa_notify(0, ifp);
1810                         } else
1811                                 spin_unlock(&ifp->lock);
1812
1813 #ifdef CONFIG_IPV6_PRIVACY
1814                         read_lock_bh(&in6_dev->lock);
1815                         /* update all temporary addresses in the list */
1816                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1817                                 /*
1818                                  * When adjusting the lifetimes of an existing
1819                                  * temporary address, only lower the lifetimes.
1820                                  * Implementations must not increase the
1821                                  * lifetimes of an existing temporary address
1822                                  * when processing a Prefix Information Option.
1823                                  */
1824                                 spin_lock(&ift->lock);
1825                                 flags = ift->flags;
1826                                 if (ift->valid_lft > valid_lft &&
1827                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1828                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1829                                 if (ift->prefered_lft > prefered_lft &&
1830                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1831                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1832                                 spin_unlock(&ift->lock);
1833                                 if (!(flags&IFA_F_TENTATIVE))
1834                                         ipv6_ifa_notify(0, ift);
1835                         }
1836
1837                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1838                                 /*
1839                                  * When a new public address is created as described in [ADDRCONF],
1840                                  * also create a new temporary address.
1841                                  */
1842                                 read_unlock_bh(&in6_dev->lock);
1843                                 ipv6_create_tempaddr(ifp, NULL);
1844                         } else {
1845                                 read_unlock_bh(&in6_dev->lock);
1846                         }
1847 #endif
1848                         in6_ifa_put(ifp);
1849                         addrconf_verify(0);
1850                 }
1851         }
1852         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1853         in6_dev_put(in6_dev);
1854 }
1855
1856 /*
1857  *      Set destination address.
1858  *      Special case for SIT interfaces where we create a new "virtual"
1859  *      device.
1860  */
1861 int addrconf_set_dstaddr(void __user *arg)
1862 {
1863         struct in6_ifreq ireq;
1864         struct net_device *dev;
1865         int err = -EINVAL;
1866
1867         rtnl_lock();
1868
1869         err = -EFAULT;
1870         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1871                 goto err_exit;
1872
1873         dev = __dev_get_by_index(&init_net, ireq.ifr6_ifindex);
1874
1875         err = -ENODEV;
1876         if (dev == NULL)
1877                 goto err_exit;
1878
1879 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1880         if (dev->type == ARPHRD_SIT) {
1881                 struct ifreq ifr;
1882                 mm_segment_t    oldfs;
1883                 struct ip_tunnel_parm p;
1884
1885                 err = -EADDRNOTAVAIL;
1886                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1887                         goto err_exit;
1888
1889                 memset(&p, 0, sizeof(p));
1890                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1891                 p.iph.saddr = 0;
1892                 p.iph.version = 4;
1893                 p.iph.ihl = 5;
1894                 p.iph.protocol = IPPROTO_IPV6;
1895                 p.iph.ttl = 64;
1896                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1897
1898                 oldfs = get_fs(); set_fs(KERNEL_DS);
1899                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1900                 set_fs(oldfs);
1901
1902                 if (err == 0) {
1903                         err = -ENOBUFS;
1904                         if ((dev = __dev_get_by_name(&init_net, p.name)) == NULL)
1905                                 goto err_exit;
1906                         err = dev_open(dev);
1907                 }
1908         }
1909 #endif
1910
1911 err_exit:
1912         rtnl_unlock();
1913         return err;
1914 }
1915
1916 /*
1917  *      Manual configuration of address on an interface
1918  */
1919 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1920                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1921 {
1922         struct inet6_ifaddr *ifp;
1923         struct inet6_dev *idev;
1924         struct net_device *dev;
1925         int scope;
1926         u32 flags = RTF_EXPIRES;
1927
1928         ASSERT_RTNL();
1929
1930         /* check the lifetime */
1931         if (!valid_lft || prefered_lft > valid_lft)
1932                 return -EINVAL;
1933
1934         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1935                 return -ENODEV;
1936
1937         if ((idev = addrconf_add_dev(dev)) == NULL)
1938                 return -ENOBUFS;
1939
1940         scope = ipv6_addr_scope(pfx);
1941
1942         if (valid_lft == INFINITY_LIFE_TIME) {
1943                 ifa_flags |= IFA_F_PERMANENT;
1944                 flags = 0;
1945         } else if (valid_lft >= 0x7FFFFFFF/HZ)
1946                 valid_lft = 0x7FFFFFFF/HZ;
1947
1948         if (prefered_lft == 0)
1949                 ifa_flags |= IFA_F_DEPRECATED;
1950         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1951                  (prefered_lft != INFINITY_LIFE_TIME))
1952                 prefered_lft = 0x7FFFFFFF/HZ;
1953
1954         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1955
1956         if (!IS_ERR(ifp)) {
1957                 spin_lock_bh(&ifp->lock);
1958                 ifp->valid_lft = valid_lft;
1959                 ifp->prefered_lft = prefered_lft;
1960                 ifp->tstamp = jiffies;
1961                 spin_unlock_bh(&ifp->lock);
1962
1963                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
1964                                       jiffies_to_clock_t(valid_lft * HZ), flags);
1965                 /*
1966                  * Note that section 3.1 of RFC 4429 indicates
1967                  * that the Optimistic flag should not be set for
1968                  * manually configured addresses
1969                  */
1970                 addrconf_dad_start(ifp, 0);
1971                 in6_ifa_put(ifp);
1972                 addrconf_verify(0);
1973                 return 0;
1974         }
1975
1976         return PTR_ERR(ifp);
1977 }
1978
1979 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1980 {
1981         struct inet6_ifaddr *ifp;
1982         struct inet6_dev *idev;
1983         struct net_device *dev;
1984
1985         if ((dev = __dev_get_by_index(&init_net, ifindex)) == NULL)
1986                 return -ENODEV;
1987
1988         if ((idev = __in6_dev_get(dev)) == NULL)
1989                 return -ENXIO;
1990
1991         read_lock_bh(&idev->lock);
1992         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1993                 if (ifp->prefix_len == plen &&
1994                     ipv6_addr_equal(pfx, &ifp->addr)) {
1995                         in6_ifa_hold(ifp);
1996                         read_unlock_bh(&idev->lock);
1997
1998                         ipv6_del_addr(ifp);
1999
2000                         /* If the last address is deleted administratively,
2001                            disable IPv6 on this interface.
2002                          */
2003                         if (idev->addr_list == NULL)
2004                                 addrconf_ifdown(idev->dev, 1);
2005                         return 0;
2006                 }
2007         }
2008         read_unlock_bh(&idev->lock);
2009         return -EADDRNOTAVAIL;
2010 }
2011
2012
2013 int addrconf_add_ifaddr(void __user *arg)
2014 {
2015         struct in6_ifreq ireq;
2016         int err;
2017
2018         if (!capable(CAP_NET_ADMIN))
2019                 return -EPERM;
2020
2021         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2022                 return -EFAULT;
2023
2024         rtnl_lock();
2025         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2026                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2027         rtnl_unlock();
2028         return err;
2029 }
2030
2031 int addrconf_del_ifaddr(void __user *arg)
2032 {
2033         struct in6_ifreq ireq;
2034         int err;
2035
2036         if (!capable(CAP_NET_ADMIN))
2037                 return -EPERM;
2038
2039         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2040                 return -EFAULT;
2041
2042         rtnl_lock();
2043         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2044         rtnl_unlock();
2045         return err;
2046 }
2047
2048 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2049 static void sit_add_v4_addrs(struct inet6_dev *idev)
2050 {
2051         struct inet6_ifaddr * ifp;
2052         struct in6_addr addr;
2053         struct net_device *dev;
2054         int scope;
2055
2056         ASSERT_RTNL();
2057
2058         memset(&addr, 0, sizeof(struct in6_addr));
2059         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2060
2061         if (idev->dev->flags&IFF_POINTOPOINT) {
2062                 addr.s6_addr32[0] = htonl(0xfe800000);
2063                 scope = IFA_LINK;
2064         } else {
2065                 scope = IPV6_ADDR_COMPATv4;
2066         }
2067
2068         if (addr.s6_addr32[3]) {
2069                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2070                 if (!IS_ERR(ifp)) {
2071                         spin_lock_bh(&ifp->lock);
2072                         ifp->flags &= ~IFA_F_TENTATIVE;
2073                         spin_unlock_bh(&ifp->lock);
2074                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2075                         in6_ifa_put(ifp);
2076                 }
2077                 return;
2078         }
2079
2080         for_each_netdev(&init_net, dev) {
2081                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2082                 if (in_dev && (dev->flags & IFF_UP)) {
2083                         struct in_ifaddr * ifa;
2084
2085                         int flag = scope;
2086
2087                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2088                                 int plen;
2089
2090                                 addr.s6_addr32[3] = ifa->ifa_local;
2091
2092                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2093                                         continue;
2094                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2095                                         if (idev->dev->flags&IFF_POINTOPOINT)
2096                                                 continue;
2097                                         flag |= IFA_HOST;
2098                                 }
2099                                 if (idev->dev->flags&IFF_POINTOPOINT)
2100                                         plen = 64;
2101                                 else
2102                                         plen = 96;
2103
2104                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2105                                                     IFA_F_PERMANENT);
2106                                 if (!IS_ERR(ifp)) {
2107                                         spin_lock_bh(&ifp->lock);
2108                                         ifp->flags &= ~IFA_F_TENTATIVE;
2109                                         spin_unlock_bh(&ifp->lock);
2110                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2111                                         in6_ifa_put(ifp);
2112                                 }
2113                         }
2114                 }
2115         }
2116 }
2117 #endif
2118
2119 static void init_loopback(struct net_device *dev)
2120 {
2121         struct inet6_dev  *idev;
2122         struct inet6_ifaddr * ifp;
2123
2124         /* ::1 */
2125
2126         ASSERT_RTNL();
2127
2128         if ((idev = ipv6_find_idev(dev)) == NULL) {
2129                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2130                 return;
2131         }
2132
2133         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2134         if (!IS_ERR(ifp)) {
2135                 spin_lock_bh(&ifp->lock);
2136                 ifp->flags &= ~IFA_F_TENTATIVE;
2137                 spin_unlock_bh(&ifp->lock);
2138                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2139                 in6_ifa_put(ifp);
2140         }
2141 }
2142
2143 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2144 {
2145         struct inet6_ifaddr * ifp;
2146         u32 addr_flags = IFA_F_PERMANENT;
2147
2148 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2149         if (idev->cnf.optimistic_dad &&
2150             !ipv6_devconf.forwarding)
2151                 addr_flags |= IFA_F_OPTIMISTIC;
2152 #endif
2153
2154
2155         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2156         if (!IS_ERR(ifp)) {
2157                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2158                 addrconf_dad_start(ifp, 0);
2159                 in6_ifa_put(ifp);
2160         }
2161 }
2162
2163 static void addrconf_dev_config(struct net_device *dev)
2164 {
2165         struct in6_addr addr;
2166         struct inet6_dev    * idev;
2167
2168         ASSERT_RTNL();
2169
2170         if ((dev->type != ARPHRD_ETHER) &&
2171             (dev->type != ARPHRD_FDDI) &&
2172             (dev->type != ARPHRD_IEEE802_TR) &&
2173             (dev->type != ARPHRD_ARCNET) &&
2174             (dev->type != ARPHRD_INFINIBAND)) {
2175                 /* Alas, we support only Ethernet autoconfiguration. */
2176                 return;
2177         }
2178
2179         idev = addrconf_add_dev(dev);
2180         if (idev == NULL)
2181                 return;
2182
2183         memset(&addr, 0, sizeof(struct in6_addr));
2184         addr.s6_addr32[0] = htonl(0xFE800000);
2185
2186         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2187                 addrconf_add_linklocal(idev, &addr);
2188 }
2189
2190 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2191 static void addrconf_sit_config(struct net_device *dev)
2192 {
2193         struct inet6_dev *idev;
2194
2195         ASSERT_RTNL();
2196
2197         /*
2198          * Configure the tunnel with one of our IPv4
2199          * addresses... we should configure all of
2200          * our v4 addrs in the tunnel
2201          */
2202
2203         if ((idev = ipv6_find_idev(dev)) == NULL) {
2204                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2205                 return;
2206         }
2207
2208         sit_add_v4_addrs(idev);
2209
2210         if (dev->flags&IFF_POINTOPOINT) {
2211                 addrconf_add_mroute(dev);
2212                 addrconf_add_lroute(dev);
2213         } else
2214                 sit_route_add(dev);
2215 }
2216 #endif
2217
2218 static inline int
2219 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2220 {
2221         struct in6_addr lladdr;
2222
2223         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2224                 addrconf_add_linklocal(idev, &lladdr);
2225                 return 0;
2226         }
2227         return -1;
2228 }
2229
2230 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2231 {
2232         struct net_device *link_dev;
2233
2234         /* first try to inherit the link-local address from the link device */
2235         if (idev->dev->iflink &&
2236             (link_dev = __dev_get_by_index(&init_net, idev->dev->iflink))) {
2237                 if (!ipv6_inherit_linklocal(idev, link_dev))
2238                         return;
2239         }
2240         /* then try to inherit it from any device */
2241         for_each_netdev(&init_net, link_dev) {
2242                 if (!ipv6_inherit_linklocal(idev, link_dev))
2243                         return;
2244         }
2245         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2246 }
2247
2248 /*
2249  * Autoconfigure tunnel with a link-local address so routing protocols,
2250  * DHCPv6, MLD etc. can be run over the virtual link
2251  */
2252
2253 static void addrconf_ip6_tnl_config(struct net_device *dev)
2254 {
2255         struct inet6_dev *idev;
2256
2257         ASSERT_RTNL();
2258
2259         if ((idev = addrconf_add_dev(dev)) == NULL) {
2260                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2261                 return;
2262         }
2263         ip6_tnl_add_linklocal(idev);
2264 }
2265
2266 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2267                            void * data)
2268 {
2269         struct net_device *dev = (struct net_device *) data;
2270         struct inet6_dev *idev = __in6_dev_get(dev);
2271         int run_pending = 0;
2272         int err;
2273
2274         if (dev->nd_net != &init_net)
2275                 return NOTIFY_DONE;
2276
2277         switch(event) {
2278         case NETDEV_REGISTER:
2279                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2280                         idev = ipv6_add_dev(dev);
2281                         if (!idev)
2282                                 return notifier_from_errno(-ENOMEM);
2283                 }
2284                 break;
2285         case NETDEV_UP:
2286         case NETDEV_CHANGE:
2287                 if (dev->flags & IFF_SLAVE)
2288                         break;
2289
2290                 if (event == NETDEV_UP) {
2291                         if (!addrconf_qdisc_ok(dev)) {
2292                                 /* device is not ready yet. */
2293                                 printk(KERN_INFO
2294                                         "ADDRCONF(NETDEV_UP): %s: "
2295                                         "link is not ready\n",
2296                                         dev->name);
2297                                 break;
2298                         }
2299
2300                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2301                                 idev = ipv6_add_dev(dev);
2302
2303                         if (idev)
2304                                 idev->if_flags |= IF_READY;
2305                 } else {
2306                         if (!addrconf_qdisc_ok(dev)) {
2307                                 /* device is still not ready. */
2308                                 break;
2309                         }
2310
2311                         if (idev) {
2312                                 if (idev->if_flags & IF_READY) {
2313                                         /* device is already configured. */
2314                                         break;
2315                                 }
2316                                 idev->if_flags |= IF_READY;
2317                         }
2318
2319                         printk(KERN_INFO
2320                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2321                                         "link becomes ready\n",
2322                                         dev->name);
2323
2324                         run_pending = 1;
2325                 }
2326
2327                 switch(dev->type) {
2328 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2329                 case ARPHRD_SIT:
2330                         addrconf_sit_config(dev);
2331                         break;
2332 #endif
2333                 case ARPHRD_TUNNEL6:
2334                         addrconf_ip6_tnl_config(dev);
2335                         break;
2336                 case ARPHRD_LOOPBACK:
2337                         init_loopback(dev);
2338                         break;
2339
2340                 default:
2341                         addrconf_dev_config(dev);
2342                         break;
2343                 }
2344                 if (idev) {
2345                         if (run_pending)
2346                                 addrconf_dad_run(idev);
2347
2348                         /* If the MTU changed during the interface down, when the
2349                            interface up, the changed MTU must be reflected in the
2350                            idev as well as routers.
2351                          */
2352                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2353                                 rt6_mtu_change(dev, dev->mtu);
2354                                 idev->cnf.mtu6 = dev->mtu;
2355                         }
2356                         idev->tstamp = jiffies;
2357                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2358                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2359                            stop IPv6 on this interface.
2360                          */
2361                         if (dev->mtu < IPV6_MIN_MTU)
2362                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2363                 }
2364                 break;
2365
2366         case NETDEV_CHANGEMTU:
2367                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2368                         rt6_mtu_change(dev, dev->mtu);
2369                         idev->cnf.mtu6 = dev->mtu;
2370                         break;
2371                 }
2372
2373                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2374                         idev = ipv6_add_dev(dev);
2375                         if (idev)
2376                                 break;
2377                 }
2378
2379                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2380
2381         case NETDEV_DOWN:
2382         case NETDEV_UNREGISTER:
2383                 /*
2384                  *      Remove all addresses from this interface.
2385                  */
2386                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2387                 break;
2388
2389         case NETDEV_CHANGENAME:
2390                 if (idev) {
2391                         snmp6_unregister_dev(idev);
2392 #ifdef CONFIG_SYSCTL
2393                         addrconf_sysctl_unregister(&idev->cnf);
2394                         neigh_sysctl_unregister(idev->nd_parms);
2395                         neigh_sysctl_register(dev, idev->nd_parms,
2396                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2397                                               &ndisc_ifinfo_sysctl_change,
2398                                               NULL);
2399                         addrconf_sysctl_register(idev, &idev->cnf);
2400 #endif
2401                         err = snmp6_register_dev(idev);
2402                         if (err)
2403                                 return notifier_from_errno(err);
2404                 }
2405                 break;
2406         }
2407
2408         return NOTIFY_OK;
2409 }
2410
2411 /*
2412  *      addrconf module should be notified of a device going up
2413  */
2414 static struct notifier_block ipv6_dev_notf = {
2415         .notifier_call = addrconf_notify,
2416         .priority = 0
2417 };
2418
2419 static int addrconf_ifdown(struct net_device *dev, int how)
2420 {
2421         struct inet6_dev *idev;
2422         struct inet6_ifaddr *ifa, **bifa;
2423         int i;
2424
2425         ASSERT_RTNL();
2426
2427         if (dev == init_net.loopback_dev && how == 1)
2428                 how = 0;
2429
2430         rt6_ifdown(dev);
2431         neigh_ifdown(&nd_tbl, dev);
2432
2433         idev = __in6_dev_get(dev);
2434         if (idev == NULL)
2435                 return -ENODEV;
2436
2437         /* Step 1: remove reference to ipv6 device from parent device.
2438                    Do not dev_put!
2439          */
2440         if (how == 1) {
2441                 idev->dead = 1;
2442
2443                 /* protected by rtnl_lock */
2444                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2445
2446                 /* Step 1.5: remove snmp6 entry */
2447                 snmp6_unregister_dev(idev);
2448
2449         }
2450
2451         /* Step 2: clear hash table */
2452         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2453                 bifa = &inet6_addr_lst[i];
2454
2455                 write_lock_bh(&addrconf_hash_lock);
2456                 while ((ifa = *bifa) != NULL) {
2457                         if (ifa->idev == idev) {
2458                                 *bifa = ifa->lst_next;
2459                                 ifa->lst_next = NULL;
2460                                 addrconf_del_timer(ifa);
2461                                 in6_ifa_put(ifa);
2462                                 continue;
2463                         }
2464                         bifa = &ifa->lst_next;
2465                 }
2466                 write_unlock_bh(&addrconf_hash_lock);
2467         }
2468
2469         write_lock_bh(&idev->lock);
2470
2471         /* Step 3: clear flags for stateless addrconf */
2472         if (how != 1)
2473                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2474
2475         /* Step 4: clear address list */
2476 #ifdef CONFIG_IPV6_PRIVACY
2477         if (how == 1 && del_timer(&idev->regen_timer))
2478                 in6_dev_put(idev);
2479
2480         /* clear tempaddr list */
2481         while ((ifa = idev->tempaddr_list) != NULL) {
2482                 idev->tempaddr_list = ifa->tmp_next;
2483                 ifa->tmp_next = NULL;
2484                 ifa->dead = 1;
2485                 write_unlock_bh(&idev->lock);
2486                 spin_lock_bh(&ifa->lock);
2487
2488                 if (ifa->ifpub) {
2489                         in6_ifa_put(ifa->ifpub);
2490                         ifa->ifpub = NULL;
2491                 }
2492                 spin_unlock_bh(&ifa->lock);
2493                 in6_ifa_put(ifa);
2494                 write_lock_bh(&idev->lock);
2495         }
2496 #endif
2497         while ((ifa = idev->addr_list) != NULL) {
2498                 idev->addr_list = ifa->if_next;
2499                 ifa->if_next = NULL;
2500                 ifa->dead = 1;
2501                 addrconf_del_timer(ifa);
2502                 write_unlock_bh(&idev->lock);
2503
2504                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2505                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2506                 in6_ifa_put(ifa);
2507
2508                 write_lock_bh(&idev->lock);
2509         }
2510         write_unlock_bh(&idev->lock);
2511
2512         /* Step 5: Discard multicast list */
2513
2514         if (how == 1)
2515                 ipv6_mc_destroy_dev(idev);
2516         else
2517                 ipv6_mc_down(idev);
2518
2519         idev->tstamp = jiffies;
2520
2521         /* Shot the device (if unregistered) */
2522
2523         if (how == 1) {
2524 #ifdef CONFIG_SYSCTL
2525                 addrconf_sysctl_unregister(&idev->cnf);
2526                 neigh_sysctl_unregister(idev->nd_parms);
2527 #endif
2528                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2529                 neigh_ifdown(&nd_tbl, dev);
2530                 in6_dev_put(idev);
2531         }
2532         return 0;
2533 }
2534
2535 static void addrconf_rs_timer(unsigned long data)
2536 {
2537         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2538
2539         if (ifp->idev->cnf.forwarding)
2540                 goto out;
2541
2542         if (ifp->idev->if_flags & IF_RA_RCVD) {
2543                 /*
2544                  *      Announcement received after solicitation
2545                  *      was sent
2546                  */
2547                 goto out;
2548         }
2549
2550         spin_lock(&ifp->lock);
2551         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2552                 struct in6_addr all_routers;
2553
2554                 /* The wait after the last probe can be shorter */
2555                 addrconf_mod_timer(ifp, AC_RS,
2556                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2557                                    ifp->idev->cnf.rtr_solicit_delay :
2558                                    ifp->idev->cnf.rtr_solicit_interval);
2559                 spin_unlock(&ifp->lock);
2560
2561                 ipv6_addr_all_routers(&all_routers);
2562
2563                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2564         } else {
2565                 spin_unlock(&ifp->lock);
2566                 /*
2567                  * Note: we do not support deprecated "all on-link"
2568                  * assumption any longer.
2569                  */
2570                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2571                        ifp->idev->dev->name);
2572         }
2573
2574 out:
2575         in6_ifa_put(ifp);
2576 }
2577
2578 /*
2579  *      Duplicate Address Detection
2580  */
2581 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2582 {
2583         unsigned long rand_num;
2584         struct inet6_dev *idev = ifp->idev;
2585
2586         if (ifp->flags & IFA_F_OPTIMISTIC)
2587                 rand_num = 0;
2588         else
2589                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2590
2591         ifp->probes = idev->cnf.dad_transmits;
2592         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2593 }
2594
2595 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2596 {
2597         struct inet6_dev *idev = ifp->idev;
2598         struct net_device *dev = idev->dev;
2599
2600         addrconf_join_solict(dev, &ifp->addr);
2601
2602         net_srandom(ifp->addr.s6_addr32[3]);
2603
2604         read_lock_bh(&idev->lock);
2605         if (ifp->dead)
2606                 goto out;
2607         spin_lock_bh(&ifp->lock);
2608
2609         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2610             !(ifp->flags&IFA_F_TENTATIVE) ||
2611             ifp->flags & IFA_F_NODAD) {
2612                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2613                 spin_unlock_bh(&ifp->lock);
2614                 read_unlock_bh(&idev->lock);
2615
2616                 addrconf_dad_completed(ifp);
2617                 return;
2618         }
2619
2620         if (!(idev->if_flags & IF_READY)) {
2621                 spin_unlock_bh(&ifp->lock);
2622                 read_unlock_bh(&idev->lock);
2623                 /*
2624                  * If the defice is not ready:
2625                  * - keep it tentative if it is a permanent address.
2626                  * - otherwise, kill it.
2627                  */
2628                 in6_ifa_hold(ifp);
2629                 addrconf_dad_stop(ifp);
2630                 return;
2631         }
2632
2633         /*
2634          * Optimistic nodes can start receiving
2635          * Frames right away
2636          */
2637         if(ifp->flags & IFA_F_OPTIMISTIC)
2638                 ip6_ins_rt(ifp->rt);
2639
2640         addrconf_dad_kick(ifp);
2641         spin_unlock_bh(&ifp->lock);
2642 out:
2643         read_unlock_bh(&idev->lock);
2644 }
2645
2646 static void addrconf_dad_timer(unsigned long data)
2647 {
2648         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2649         struct inet6_dev *idev = ifp->idev;
2650         struct in6_addr unspec;
2651         struct in6_addr mcaddr;
2652
2653         read_lock_bh(&idev->lock);
2654         if (idev->dead) {
2655                 read_unlock_bh(&idev->lock);
2656                 goto out;
2657         }
2658         spin_lock_bh(&ifp->lock);
2659         if (ifp->probes == 0) {
2660                 /*
2661                  * DAD was successful
2662                  */
2663
2664                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2665                 spin_unlock_bh(&ifp->lock);
2666                 read_unlock_bh(&idev->lock);
2667
2668                 addrconf_dad_completed(ifp);
2669
2670                 goto out;
2671         }
2672
2673         ifp->probes--;
2674         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2675         spin_unlock_bh(&ifp->lock);
2676         read_unlock_bh(&idev->lock);
2677
2678         /* send a neighbour solicitation for our addr */
2679         memset(&unspec, 0, sizeof(unspec));
2680         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2681         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2682 out:
2683         in6_ifa_put(ifp);
2684 }
2685
2686 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2687 {
2688         struct net_device *     dev = ifp->idev->dev;
2689
2690         /*
2691          *      Configure the address for reception. Now it is valid.
2692          */
2693
2694         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2695
2696         /* If added prefix is link local and forwarding is off,
2697            start sending router solicitations.
2698          */
2699
2700         if (ifp->idev->cnf.forwarding == 0 &&
2701             ifp->idev->cnf.rtr_solicits > 0 &&
2702             (dev->flags&IFF_LOOPBACK) == 0 &&
2703             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2704                 struct in6_addr all_routers;
2705
2706                 ipv6_addr_all_routers(&all_routers);
2707
2708                 /*
2709                  *      If a host as already performed a random delay
2710                  *      [...] as part of DAD [...] there is no need
2711                  *      to delay again before sending the first RS
2712                  */
2713                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2714
2715                 spin_lock_bh(&ifp->lock);
2716                 ifp->probes = 1;
2717                 ifp->idev->if_flags |= IF_RS_SENT;
2718                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2719                 spin_unlock_bh(&ifp->lock);
2720         }
2721 }
2722
2723 static void addrconf_dad_run(struct inet6_dev *idev) {
2724         struct inet6_ifaddr *ifp;
2725
2726         read_lock_bh(&idev->lock);
2727         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2728                 spin_lock_bh(&ifp->lock);
2729                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2730                         spin_unlock_bh(&ifp->lock);
2731                         continue;
2732                 }
2733                 spin_unlock_bh(&ifp->lock);
2734                 addrconf_dad_kick(ifp);
2735         }
2736         read_unlock_bh(&idev->lock);
2737 }
2738
2739 #ifdef CONFIG_PROC_FS
2740 struct if6_iter_state {
2741         int bucket;
2742 };
2743
2744 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2745 {
2746         struct inet6_ifaddr *ifa = NULL;
2747         struct if6_iter_state *state = seq->private;
2748
2749         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2750                 ifa = inet6_addr_lst[state->bucket];
2751                 if (ifa)
2752                         break;
2753         }
2754         return ifa;
2755 }
2756
2757 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2758 {
2759         struct if6_iter_state *state = seq->private;
2760
2761         ifa = ifa->lst_next;
2762 try_again:
2763         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2764                 ifa = inet6_addr_lst[state->bucket];
2765                 goto try_again;
2766         }
2767         return ifa;
2768 }
2769
2770 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2771 {
2772         struct inet6_ifaddr *ifa = if6_get_first(seq);
2773
2774         if (ifa)
2775                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2776                         --pos;
2777         return pos ? NULL : ifa;
2778 }
2779
2780 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2781 {
2782         read_lock_bh(&addrconf_hash_lock);
2783         return if6_get_idx(seq, *pos);
2784 }
2785
2786 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2787 {
2788         struct inet6_ifaddr *ifa;
2789
2790         ifa = if6_get_next(seq, v);
2791         ++*pos;
2792         return ifa;
2793 }
2794
2795 static void if6_seq_stop(struct seq_file *seq, void *v)
2796 {
2797         read_unlock_bh(&addrconf_hash_lock);
2798 }
2799
2800 static int if6_seq_show(struct seq_file *seq, void *v)
2801 {
2802         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2803         seq_printf(seq,
2804                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2805                    NIP6(ifp->addr),
2806                    ifp->idev->dev->ifindex,
2807                    ifp->prefix_len,
2808                    ifp->scope,
2809                    ifp->flags,
2810                    ifp->idev->dev->name);
2811         return 0;
2812 }
2813
2814 static const struct seq_operations if6_seq_ops = {
2815         .start  = if6_seq_start,
2816         .next   = if6_seq_next,
2817         .show   = if6_seq_show,
2818         .stop   = if6_seq_stop,
2819 };
2820
2821 static int if6_seq_open(struct inode *inode, struct file *file)
2822 {
2823         return seq_open_private(file, &if6_seq_ops,
2824                         sizeof(struct if6_iter_state));
2825 }
2826
2827 static const struct file_operations if6_fops = {
2828         .owner          = THIS_MODULE,
2829         .open           = if6_seq_open,
2830         .read           = seq_read,
2831         .llseek         = seq_lseek,
2832         .release        = seq_release_private,
2833 };
2834
2835 int __init if6_proc_init(void)
2836 {
2837         if (!proc_net_fops_create(&init_net, "if_inet6", S_IRUGO, &if6_fops))
2838                 return -ENOMEM;
2839         return 0;
2840 }
2841
2842 void if6_proc_exit(void)
2843 {
2844         proc_net_remove(&init_net, "if_inet6");
2845 }
2846 #endif  /* CONFIG_PROC_FS */
2847
2848 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2849 /* Check if address is a home address configured on any interface. */
2850 int ipv6_chk_home_addr(struct in6_addr *addr)
2851 {
2852         int ret = 0;
2853         struct inet6_ifaddr * ifp;
2854         u8 hash = ipv6_addr_hash(addr);
2855         read_lock_bh(&addrconf_hash_lock);
2856         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2857                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2858                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2859                         ret = 1;
2860                         break;
2861                 }
2862         }
2863         read_unlock_bh(&addrconf_hash_lock);
2864         return ret;
2865 }
2866 #endif
2867
2868 /*
2869  *      Periodic address status verification
2870  */
2871
2872 static void addrconf_verify(unsigned long foo)
2873 {
2874         struct inet6_ifaddr *ifp;
2875         unsigned long now, next;
2876         int i;
2877
2878         spin_lock_bh(&addrconf_verify_lock);
2879         now = jiffies;
2880         next = now + ADDR_CHECK_FREQUENCY;
2881
2882         del_timer(&addr_chk_timer);
2883
2884         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2885
2886 restart:
2887                 read_lock(&addrconf_hash_lock);
2888                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2889                         unsigned long age;
2890 #ifdef CONFIG_IPV6_PRIVACY
2891                         unsigned long regen_advance;
2892 #endif
2893
2894                         if (ifp->flags & IFA_F_PERMANENT)
2895                                 continue;
2896
2897                         spin_lock(&ifp->lock);
2898                         age = (now - ifp->tstamp) / HZ;
2899
2900 #ifdef CONFIG_IPV6_PRIVACY
2901                         regen_advance = ifp->idev->cnf.regen_max_retry *
2902                                         ifp->idev->cnf.dad_transmits *
2903                                         ifp->idev->nd_parms->retrans_time / HZ;
2904 #endif
2905
2906                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2907                             age >= ifp->valid_lft) {
2908                                 spin_unlock(&ifp->lock);
2909                                 in6_ifa_hold(ifp);
2910                                 read_unlock(&addrconf_hash_lock);
2911                                 ipv6_del_addr(ifp);
2912                                 goto restart;
2913                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2914                                 spin_unlock(&ifp->lock);
2915                                 continue;
2916                         } else if (age >= ifp->prefered_lft) {
2917                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2918                                 int deprecate = 0;
2919
2920                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2921                                         deprecate = 1;
2922                                         ifp->flags |= IFA_F_DEPRECATED;
2923                                 }
2924
2925                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2926                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2927
2928                                 spin_unlock(&ifp->lock);
2929
2930                                 if (deprecate) {
2931                                         in6_ifa_hold(ifp);
2932                                         read_unlock(&addrconf_hash_lock);
2933
2934                                         ipv6_ifa_notify(0, ifp);
2935                                         in6_ifa_put(ifp);
2936                                         goto restart;
2937                                 }
2938 #ifdef CONFIG_IPV6_PRIVACY
2939                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2940                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2941                                 if (age >= ifp->prefered_lft - regen_advance) {
2942                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2943                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2944                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2945                                         if (!ifp->regen_count && ifpub) {
2946                                                 ifp->regen_count++;
2947                                                 in6_ifa_hold(ifp);
2948                                                 in6_ifa_hold(ifpub);
2949                                                 spin_unlock(&ifp->lock);
2950                                                 read_unlock(&addrconf_hash_lock);
2951                                                 spin_lock(&ifpub->lock);
2952                                                 ifpub->regen_count = 0;
2953                                                 spin_unlock(&ifpub->lock);
2954                                                 ipv6_create_tempaddr(ifpub, ifp);
2955                                                 in6_ifa_put(ifpub);
2956                                                 in6_ifa_put(ifp);
2957                                                 goto restart;
2958                                         }
2959                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2960                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2961                                 spin_unlock(&ifp->lock);
2962 #endif
2963                         } else {
2964                                 /* ifp->prefered_lft <= ifp->valid_lft */
2965                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2966                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2967                                 spin_unlock(&ifp->lock);
2968                         }
2969                 }
2970                 read_unlock(&addrconf_hash_lock);
2971         }
2972
2973         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2974         add_timer(&addr_chk_timer);
2975         spin_unlock_bh(&addrconf_verify_lock);
2976 }
2977
2978 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2979 {
2980         struct in6_addr *pfx = NULL;
2981
2982         if (addr)
2983                 pfx = nla_data(addr);
2984
2985         if (local) {
2986                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2987                         pfx = NULL;
2988                 else
2989                         pfx = nla_data(local);
2990         }
2991
2992         return pfx;
2993 }
2994
2995 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
2996         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2997         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2998         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2999 };
3000
3001 static int
3002 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3003 {
3004         struct ifaddrmsg *ifm;
3005         struct nlattr *tb[IFA_MAX+1];
3006         struct in6_addr *pfx;
3007         int err;
3008
3009         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3010         if (err < 0)
3011                 return err;
3012
3013         ifm = nlmsg_data(nlh);
3014         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3015         if (pfx == NULL)
3016                 return -EINVAL;
3017
3018         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3019 }
3020
3021 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3022                              u32 prefered_lft, u32 valid_lft)
3023 {
3024         u32 flags = RTF_EXPIRES;
3025
3026         if (!valid_lft || (prefered_lft > valid_lft))
3027                 return -EINVAL;
3028
3029         if (valid_lft == INFINITY_LIFE_TIME) {
3030                 ifa_flags |= IFA_F_PERMANENT;
3031                 flags = 0;
3032         } else if (valid_lft >= 0x7FFFFFFF/HZ)
3033                 valid_lft = 0x7FFFFFFF/HZ;
3034
3035         if (prefered_lft == 0)
3036                 ifa_flags |= IFA_F_DEPRECATED;
3037         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
3038                  (prefered_lft != INFINITY_LIFE_TIME))
3039                 prefered_lft = 0x7FFFFFFF/HZ;
3040
3041         spin_lock_bh(&ifp->lock);
3042         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3043         ifp->tstamp = jiffies;
3044         ifp->valid_lft = valid_lft;
3045         ifp->prefered_lft = prefered_lft;
3046
3047         spin_unlock_bh(&ifp->lock);
3048         if (!(ifp->flags&IFA_F_TENTATIVE))
3049                 ipv6_ifa_notify(0, ifp);
3050
3051         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3052                               jiffies_to_clock_t(valid_lft * HZ), flags);
3053         addrconf_verify(0);
3054
3055         return 0;
3056 }
3057
3058 static int
3059 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3060 {
3061         struct ifaddrmsg *ifm;
3062         struct nlattr *tb[IFA_MAX+1];
3063         struct in6_addr *pfx;
3064         struct inet6_ifaddr *ifa;
3065         struct net_device *dev;
3066         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3067         u8 ifa_flags;
3068         int err;
3069
3070         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3071         if (err < 0)
3072                 return err;
3073
3074         ifm = nlmsg_data(nlh);
3075         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3076         if (pfx == NULL)
3077                 return -EINVAL;
3078
3079         if (tb[IFA_CACHEINFO]) {
3080                 struct ifa_cacheinfo *ci;
3081
3082                 ci = nla_data(tb[IFA_CACHEINFO]);
3083                 valid_lft = ci->ifa_valid;
3084                 preferred_lft = ci->ifa_prefered;
3085         } else {
3086                 preferred_lft = INFINITY_LIFE_TIME;
3087                 valid_lft = INFINITY_LIFE_TIME;
3088         }
3089
3090         dev =  __dev_get_by_index(&init_net, ifm->ifa_index);
3091         if (dev == NULL)
3092                 return -ENODEV;
3093
3094         /* We ignore other flags so far. */
3095         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3096
3097         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3098         if (ifa == NULL) {
3099                 /*
3100                  * It would be best to check for !NLM_F_CREATE here but
3101                  * userspace alreay relies on not having to provide this.
3102                  */
3103                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3104                                       ifa_flags, preferred_lft, valid_lft);
3105         }
3106
3107         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3108             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3109                 err = -EEXIST;
3110         else
3111                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3112
3113         in6_ifa_put(ifa);
3114
3115         return err;
3116 }
3117
3118 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3119                           u8 scope, int ifindex)
3120 {
3121         struct ifaddrmsg *ifm;
3122
3123         ifm = nlmsg_data(nlh);
3124         ifm->ifa_family = AF_INET6;
3125         ifm->ifa_prefixlen = prefixlen;
3126         ifm->ifa_flags = flags;
3127         ifm->ifa_scope = scope;
3128         ifm->ifa_index = ifindex;
3129 }
3130
3131 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3132                          unsigned long tstamp, u32 preferred, u32 valid)
3133 {
3134         struct ifa_cacheinfo ci;
3135
3136         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3137                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3138         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3139                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3140         ci.ifa_prefered = preferred;
3141         ci.ifa_valid = valid;
3142
3143         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3144 }
3145
3146 static inline int rt_scope(int ifa_scope)
3147 {
3148         if (ifa_scope & IFA_HOST)
3149                 return RT_SCOPE_HOST;
3150         else if (ifa_scope & IFA_LINK)
3151                 return RT_SCOPE_LINK;
3152         else if (ifa_scope & IFA_SITE)
3153                 return RT_SCOPE_SITE;
3154         else
3155                 return RT_SCOPE_UNIVERSE;
3156 }
3157
3158 static inline int inet6_ifaddr_msgsize(void)
3159 {
3160         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3161                + nla_total_size(16) /* IFA_ADDRESS */
3162                + nla_total_size(sizeof(struct ifa_cacheinfo));
3163 }
3164
3165 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3166                              u32 pid, u32 seq, int event, unsigned int flags)
3167 {
3168         struct nlmsghdr  *nlh;
3169         u32 preferred, valid;
3170
3171         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3172         if (nlh == NULL)
3173                 return -EMSGSIZE;
3174
3175         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3176                       ifa->idev->dev->ifindex);
3177
3178         if (!(ifa->flags&IFA_F_PERMANENT)) {
3179                 preferred = ifa->prefered_lft;
3180                 valid = ifa->valid_lft;
3181                 if (preferred != INFINITY_LIFE_TIME) {
3182                         long tval = (jiffies - ifa->tstamp)/HZ;
3183                         preferred -= tval;
3184                         if (valid != INFINITY_LIFE_TIME)
3185                                 valid -= tval;
3186                 }
3187         } else {
3188                 preferred = INFINITY_LIFE_TIME;
3189                 valid = INFINITY_LIFE_TIME;
3190         }
3191
3192         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3193             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3194                 nlmsg_cancel(skb, nlh);
3195                 return -EMSGSIZE;
3196         }
3197
3198         return nlmsg_end(skb, nlh);
3199 }
3200
3201 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3202                                 u32 pid, u32 seq, int event, u16 flags)
3203 {
3204         struct nlmsghdr  *nlh;
3205         u8 scope = RT_SCOPE_UNIVERSE;
3206         int ifindex = ifmca->idev->dev->ifindex;
3207
3208         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3209                 scope = RT_SCOPE_SITE;
3210
3211         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3212         if (nlh == NULL)
3213                 return -EMSGSIZE;
3214
3215         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3216         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3217             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3218                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3219                 nlmsg_cancel(skb, nlh);
3220                 return -EMSGSIZE;
3221         }
3222
3223         return nlmsg_end(skb, nlh);
3224 }
3225
3226 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3227                                 u32 pid, u32 seq, int event, unsigned int flags)
3228 {
3229         struct nlmsghdr  *nlh;
3230         u8 scope = RT_SCOPE_UNIVERSE;
3231         int ifindex = ifaca->aca_idev->dev->ifindex;
3232
3233         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3234                 scope = RT_SCOPE_SITE;
3235
3236         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3237         if (nlh == NULL)
3238                 return -EMSGSIZE;
3239
3240         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3241         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3242             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3243                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3244                 nlmsg_cancel(skb, nlh);
3245                 return -EMSGSIZE;
3246         }
3247
3248         return nlmsg_end(skb, nlh);
3249 }
3250
3251 enum addr_type_t
3252 {
3253         UNICAST_ADDR,
3254         MULTICAST_ADDR,
3255         ANYCAST_ADDR,
3256 };
3257
3258 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3259                            enum addr_type_t type)
3260 {
3261         int idx, ip_idx;
3262         int s_idx, s_ip_idx;
3263         int err = 1;
3264         struct net_device *dev;
3265         struct inet6_dev *idev = NULL;
3266         struct inet6_ifaddr *ifa;
3267         struct ifmcaddr6 *ifmca;
3268         struct ifacaddr6 *ifaca;
3269
3270         s_idx = cb->args[0];
3271         s_ip_idx = ip_idx = cb->args[1];
3272
3273         idx = 0;
3274         for_each_netdev(&init_net, dev) {
3275                 if (idx < s_idx)
3276                         goto cont;
3277                 if (idx > s_idx)
3278                         s_ip_idx = 0;
3279                 ip_idx = 0;
3280                 if ((idev = in6_dev_get(dev)) == NULL)
3281                         goto cont;
3282                 read_lock_bh(&idev->lock);
3283                 switch (type) {
3284                 case UNICAST_ADDR:
3285                         /* unicast address incl. temp addr */
3286                         for (ifa = idev->addr_list; ifa;
3287                              ifa = ifa->if_next, ip_idx++) {
3288                                 if (ip_idx < s_ip_idx)
3289                                         continue;
3290                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3291                                     NETLINK_CB(cb->skb).pid,
3292                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3293                                     NLM_F_MULTI)) <= 0)
3294                                         goto done;
3295                         }
3296                         break;
3297                 case MULTICAST_ADDR:
3298                         /* multicast address */
3299                         for (ifmca = idev->mc_list; ifmca;
3300                              ifmca = ifmca->next, ip_idx++) {
3301                                 if (ip_idx < s_ip_idx)
3302                                         continue;
3303                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3304                                     NETLINK_CB(cb->skb).pid,
3305                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3306                                     NLM_F_MULTI)) <= 0)
3307                                         goto done;
3308                         }
3309                         break;
3310                 case ANYCAST_ADDR:
3311                         /* anycast address */
3312                         for (ifaca = idev->ac_list; ifaca;
3313                              ifaca = ifaca->aca_next, ip_idx++) {
3314                                 if (ip_idx < s_ip_idx)
3315                                         continue;
3316                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3317                                     NETLINK_CB(cb->skb).pid,
3318                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3319                                     NLM_F_MULTI)) <= 0)
3320                                         goto done;
3321                         }
3322                         break;
3323                 default:
3324                         break;
3325                 }
3326                 read_unlock_bh(&idev->lock);
3327                 in6_dev_put(idev);
3328 cont:
3329                 idx++;
3330         }
3331 done:
3332         if (err <= 0) {
3333                 read_unlock_bh(&idev->lock);
3334                 in6_dev_put(idev);
3335         }
3336         cb->args[0] = idx;
3337         cb->args[1] = ip_idx;
3338         return skb->len;
3339 }
3340
3341 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3342 {
3343         enum addr_type_t type = UNICAST_ADDR;
3344         return inet6_dump_addr(skb, cb, type);
3345 }
3346
3347 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3348 {
3349         enum addr_type_t type = MULTICAST_ADDR;
3350         return inet6_dump_addr(skb, cb, type);
3351 }
3352
3353
3354 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3355 {
3356         enum addr_type_t type = ANYCAST_ADDR;
3357         return inet6_dump_addr(skb, cb, type);
3358 }
3359
3360 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3361                              void *arg)
3362 {
3363         struct ifaddrmsg *ifm;
3364         struct nlattr *tb[IFA_MAX+1];
3365         struct in6_addr *addr = NULL;
3366         struct net_device *dev = NULL;
3367         struct inet6_ifaddr *ifa;
3368         struct sk_buff *skb;
3369         int err;
3370
3371         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3372         if (err < 0)
3373                 goto errout;
3374
3375         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3376         if (addr == NULL) {
3377                 err = -EINVAL;
3378                 goto errout;
3379         }
3380
3381         ifm = nlmsg_data(nlh);
3382         if (ifm->ifa_index)
3383                 dev = __dev_get_by_index(&init_net, ifm->ifa_index);
3384
3385         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3386                 err = -EADDRNOTAVAIL;
3387                 goto errout;
3388         }
3389
3390         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3391                 err = -ENOBUFS;
3392                 goto errout_ifa;
3393         }
3394
3395         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3396                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3397         if (err < 0) {
3398                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3399                 WARN_ON(err == -EMSGSIZE);
3400                 kfree_skb(skb);
3401                 goto errout_ifa;
3402         }
3403         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3404 errout_ifa:
3405         in6_ifa_put(ifa);
3406 errout:
3407         return err;
3408 }
3409
3410 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3411 {
3412         struct sk_buff *skb;
3413         int err = -ENOBUFS;
3414
3415         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3416         if (skb == NULL)
3417                 goto errout;
3418
3419         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3420         if (err < 0) {
3421                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3422                 WARN_ON(err == -EMSGSIZE);
3423                 kfree_skb(skb);
3424                 goto errout;
3425         }
3426         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3427 errout:
3428         if (err < 0)
3429                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3430 }
3431
3432 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3433                                 __s32 *array, int bytes)
3434 {
3435         BUG_ON(bytes < (DEVCONF_MAX * 4));
3436
3437         memset(array, 0, bytes);
3438         array[DEVCONF_FORWARDING] = cnf->forwarding;
3439         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3440         array[DEVCONF_MTU6] = cnf->mtu6;
3441         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3442         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3443         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3444         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3445         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3446         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3447         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3448         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3449 #ifdef CONFIG_IPV6_PRIVACY
3450         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3451         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3452         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3453         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3454         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3455 #endif
3456         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3457         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3458         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3459 #ifdef CONFIG_IPV6_ROUTER_PREF
3460         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3461         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3462 #ifdef CONFIG_IPV6_ROUTE_INFO
3463         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3464 #endif
3465 #endif
3466         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3467         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3468 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3469         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3470 #endif
3471 }
3472
3473 static inline size_t inet6_if_nlmsg_size(void)
3474 {
3475         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3476                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3477                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3478                + nla_total_size(4) /* IFLA_MTU */
3479                + nla_total_size(4) /* IFLA_LINK */
3480                + nla_total_size( /* IFLA_PROTINFO */
3481                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3482                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3483                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3484                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3485                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3486                  );
3487 }
3488
3489 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3490                                       int bytes)
3491 {
3492         int i;
3493         int pad = bytes - sizeof(u64) * items;
3494         BUG_ON(pad < 0);
3495
3496         /* Use put_unaligned() because stats may not be aligned for u64. */
3497         put_unaligned(items, &stats[0]);
3498         for (i = 1; i < items; i++)
3499                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3500
3501         memset(&stats[items], 0, pad);
3502 }
3503
3504 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3505                              int bytes)
3506 {
3507         switch(attrtype) {
3508         case IFLA_INET6_STATS:
3509                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3510                 break;
3511         case IFLA_INET6_ICMP6STATS:
3512                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3513                 break;
3514         }
3515 }
3516
3517 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3518                              u32 pid, u32 seq, int event, unsigned int flags)
3519 {
3520         struct net_device *dev = idev->dev;
3521         struct nlattr *nla;
3522         struct ifinfomsg *hdr;
3523         struct nlmsghdr *nlh;
3524         void *protoinfo;
3525         struct ifla_cacheinfo ci;
3526
3527         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3528         if (nlh == NULL)
3529                 return -EMSGSIZE;
3530
3531         hdr = nlmsg_data(nlh);
3532         hdr->ifi_family = AF_INET6;
3533         hdr->__ifi_pad = 0;
3534         hdr->ifi_type = dev->type;
3535         hdr->ifi_index = dev->ifindex;
3536         hdr->ifi_flags = dev_get_flags(dev);
3537         hdr->ifi_change = 0;
3538
3539         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3540
3541         if (dev->addr_len)
3542                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3543
3544         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3545         if (dev->ifindex != dev->iflink)
3546                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3547
3548         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3549         if (protoinfo == NULL)
3550                 goto nla_put_failure;
3551
3552         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3553
3554         ci.max_reasm_len = IPV6_MAXPLEN;
3555         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3556                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3557         ci.reachable_time = idev->nd_parms->reachable_time;
3558         ci.retrans_time = idev->nd_parms->retrans_time;
3559         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3560
3561         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3562         if (nla == NULL)
3563                 goto nla_put_failure;
3564         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3565
3566         /* XXX - MC not implemented */
3567
3568         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3569         if (nla == NULL)
3570                 goto nla_put_failure;
3571         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3572
3573         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3574         if (nla == NULL)
3575                 goto nla_put_failure;
3576         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3577
3578         nla_nest_end(skb, protoinfo);
3579         return nlmsg_end(skb, nlh);
3580
3581 nla_put_failure:
3582         nlmsg_cancel(skb, nlh);
3583         return -EMSGSIZE;
3584 }
3585
3586 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3587 {
3588         int idx, err;
3589         int s_idx = cb->args[0];
3590         struct net_device *dev;
3591         struct inet6_dev *idev;
3592
3593         read_lock(&dev_base_lock);
3594         idx = 0;
3595         for_each_netdev(&init_net, dev) {
3596                 if (idx < s_idx)
3597                         goto cont;
3598                 if ((idev = in6_dev_get(dev)) == NULL)
3599                         goto cont;
3600                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3601                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3602                 in6_dev_put(idev);
3603                 if (err <= 0)
3604                         break;
3605 cont:
3606                 idx++;
3607         }
3608         read_unlock(&dev_base_lock);
3609         cb->args[0] = idx;
3610
3611         return skb->len;
3612 }
3613
3614 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3615 {
3616         struct sk_buff *skb;
3617         int err = -ENOBUFS;
3618
3619         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3620         if (skb == NULL)
3621                 goto errout;
3622
3623         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3624         if (err < 0) {
3625                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3626                 WARN_ON(err == -EMSGSIZE);
3627                 kfree_skb(skb);
3628                 goto errout;
3629         }
3630         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3631 errout:
3632         if (err < 0)
3633                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3634 }
3635
3636 static inline size_t inet6_prefix_nlmsg_size(void)
3637 {
3638         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3639                + nla_total_size(sizeof(struct in6_addr))
3640                + nla_total_size(sizeof(struct prefix_cacheinfo));
3641 }
3642
3643 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3644                              struct prefix_info *pinfo, u32 pid, u32 seq,
3645                              int event, unsigned int flags)
3646 {
3647         struct prefixmsg *pmsg;
3648         struct nlmsghdr *nlh;
3649         struct prefix_cacheinfo ci;
3650
3651         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3652         if (nlh == NULL)
3653                 return -EMSGSIZE;
3654
3655         pmsg = nlmsg_data(nlh);
3656         pmsg->prefix_family = AF_INET6;
3657         pmsg->prefix_pad1 = 0;
3658         pmsg->prefix_pad2 = 0;
3659         pmsg->prefix_ifindex = idev->dev->ifindex;
3660         pmsg->prefix_len = pinfo->prefix_len;
3661         pmsg->prefix_type = pinfo->type;
3662         pmsg->prefix_pad3 = 0;
3663         pmsg->prefix_flags = 0;
3664         if (pinfo->onlink)
3665                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3666         if (pinfo->autoconf)
3667                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3668
3669         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3670
3671         ci.preferred_time = ntohl(pinfo->prefered);
3672         ci.valid_time = ntohl(pinfo->valid);
3673         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3674
3675         return nlmsg_end(skb, nlh);
3676
3677 nla_put_failure:
3678         nlmsg_cancel(skb, nlh);
3679         return -EMSGSIZE;
3680 }
3681
3682 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3683                          struct prefix_info *pinfo)
3684 {
3685         struct sk_buff *skb;
3686         int err = -ENOBUFS;
3687
3688         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3689         if (skb == NULL)
3690                 goto errout;
3691
3692         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3693         if (err < 0) {
3694                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3695                 WARN_ON(err == -EMSGSIZE);
3696                 kfree_skb(skb);
3697                 goto errout;
3698         }
3699         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3700 errout:
3701         if (err < 0)
3702                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3703 }
3704
3705 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3706 {
3707         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3708
3709         switch (event) {
3710         case RTM_NEWADDR:
3711                 /*
3712                  * If the address was optimistic
3713                  * we inserted the route at the start of
3714                  * our DAD process, so we don't need
3715                  * to do it again
3716                  */
3717                 if (!(ifp->rt->rt6i_node))
3718                         ip6_ins_rt(ifp->rt);
3719                 if (ifp->idev->cnf.forwarding)
3720                         addrconf_join_anycast(ifp);
3721                 break;
3722         case RTM_DELADDR:
3723                 if (ifp->idev->cnf.forwarding)
3724                         addrconf_leave_anycast(ifp);
3725                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3726                 dst_hold(&ifp->rt->u.dst);
3727                 if (ip6_del_rt(ifp->rt))
3728                         dst_free(&ifp->rt->u.dst);
3729                 break;
3730         }
3731 }
3732
3733 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3734 {
3735         rcu_read_lock_bh();
3736         if (likely(ifp->idev->dead == 0))
3737                 __ipv6_ifa_notify(event, ifp);
3738         rcu_read_unlock_bh();
3739 }
3740
3741 #ifdef CONFIG_SYSCTL
3742
3743 static
3744 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3745                            void __user *buffer, size_t *lenp, loff_t *ppos)
3746 {
3747         int *valp = ctl->data;
3748         int val = *valp;
3749         int ret;
3750
3751         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3752
3753         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3754                 if (valp != &ipv6_devconf.forwarding) {
3755                         if ((!*valp) ^ (!val)) {
3756                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3757                                 if (idev == NULL)
3758                                         return ret;
3759                                 dev_forward_change(idev);
3760                         }
3761                 } else {
3762                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3763                         addrconf_forward_change();
3764                 }
3765                 if (*valp)
3766                         rt6_purge_dflt_routers();
3767         }
3768
3769         return ret;
3770 }
3771
3772 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3773                                             int __user *name, int nlen,
3774                                             void __user *oldval,
3775                                             size_t __user *oldlenp,
3776                                             void __user *newval, size_t newlen)
3777 {
3778         int *valp = table->data;
3779         int new;
3780
3781         if (!newval || !newlen)
3782                 return 0;
3783         if (newlen != sizeof(int))
3784                 return -EINVAL;
3785         if (get_user(new, (int __user *)newval))
3786                 return -EFAULT;
3787         if (new == *valp)
3788                 return 0;
3789         if (oldval && oldlenp) {
3790                 size_t len;
3791                 if (get_user(len, oldlenp))
3792                         return -EFAULT;
3793                 if (len) {
3794                         if (len > table->maxlen)
3795                                 len = table->maxlen;
3796                         if (copy_to_user(oldval, valp, len))
3797                                 return -EFAULT;
3798                         if (put_user(len, oldlenp))
3799                                 return -EFAULT;
3800                 }
3801         }
3802
3803         if (valp != &ipv6_devconf_dflt.forwarding) {
3804                 if (valp != &ipv6_devconf.forwarding) {
3805                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3806                         int changed;
3807                         if (unlikely(idev == NULL))
3808                                 return -ENODEV;
3809                         changed = (!*valp) ^ (!new);
3810                         *valp = new;
3811                         if (changed)
3812                                 dev_forward_change(idev);
3813                 } else {
3814                         *valp = new;
3815                         addrconf_forward_change();
3816                 }
3817
3818                 if (*valp)
3819                         rt6_purge_dflt_routers();
3820         } else
3821                 *valp = new;
3822
3823         return 1;
3824 }
3825
3826 static struct addrconf_sysctl_table
3827 {
3828         struct ctl_table_header *sysctl_header;
3829         ctl_table addrconf_vars[__NET_IPV6_MAX];
3830         ctl_table addrconf_dev[2];
3831         ctl_table addrconf_conf_dir[2];
3832         ctl_table addrconf_proto_dir[2];
3833         ctl_table addrconf_root_dir[2];
3834 } addrconf_sysctl __read_mostly = {
3835         .sysctl_header = NULL,
3836         .addrconf_vars = {
3837                 {
3838                         .ctl_name       =       NET_IPV6_FORWARDING,
3839                         .procname       =       "forwarding",
3840                         .data           =       &ipv6_devconf.forwarding,
3841                         .maxlen         =       sizeof(int),
3842                         .mode           =       0644,
3843                         .proc_handler   =       &addrconf_sysctl_forward,
3844                         .strategy       =       &addrconf_sysctl_forward_strategy,
3845                 },
3846                 {
3847                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3848                         .procname       =       "hop_limit",
3849                         .data           =       &ipv6_devconf.hop_limit,
3850                         .maxlen         =       sizeof(int),
3851                         .mode           =       0644,
3852                         .proc_handler   =       proc_dointvec,
3853                 },
3854                 {
3855                         .ctl_name       =       NET_IPV6_MTU,
3856                         .procname       =       "mtu",
3857                         .data           =       &ipv6_devconf.mtu6,
3858                         .maxlen         =       sizeof(int),
3859                         .mode           =       0644,
3860                         .proc_handler   =       &proc_dointvec,
3861                 },
3862                 {
3863                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3864                         .procname       =       "accept_ra",
3865                         .data           =       &ipv6_devconf.accept_ra,
3866                         .maxlen         =       sizeof(int),
3867                         .mode           =       0644,
3868                         .proc_handler   =       &proc_dointvec,
3869                 },
3870                 {
3871                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3872                         .procname       =       "accept_redirects",
3873                         .data           =       &ipv6_devconf.accept_redirects,
3874                         .maxlen         =       sizeof(int),
3875                         .mode           =       0644,
3876                         .proc_handler   =       &proc_dointvec,
3877                 },
3878                 {
3879                         .ctl_name       =       NET_IPV6_AUTOCONF,
3880                         .procname       =       "autoconf",
3881                         .data           =       &ipv6_devconf.autoconf,
3882                         .maxlen         =       sizeof(int),
3883                         .mode           =       0644,
3884                         .proc_handler   =       &proc_dointvec,
3885                 },
3886                 {
3887                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3888                         .procname       =       "dad_transmits",
3889                         .data           =       &ipv6_devconf.dad_transmits,
3890                         .maxlen         =       sizeof(int),
3891                         .mode           =       0644,
3892                         .proc_handler   =       &proc_dointvec,
3893                 },
3894                 {
3895                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3896                         .procname       =       "router_solicitations",
3897                         .data           =       &ipv6_devconf.rtr_solicits,
3898                         .maxlen         =       sizeof(int),
3899                         .mode           =       0644,
3900                         .proc_handler   =       &proc_dointvec,
3901                 },
3902                 {
3903                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3904                         .procname       =       "router_solicitation_interval",
3905                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3906                         .maxlen         =       sizeof(int),
3907                         .mode           =       0644,
3908                         .proc_handler   =       &proc_dointvec_jiffies,
3909                         .strategy       =       &sysctl_jiffies,
3910                 },
3911                 {
3912                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3913                         .procname       =       "router_solicitation_delay",
3914                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3915                         .maxlen         =       sizeof(int),
3916                         .mode           =       0644,
3917                         .proc_handler   =       &proc_dointvec_jiffies,
3918                         .strategy       =       &sysctl_jiffies,
3919                 },
3920                 {
3921                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3922                         .procname       =       "force_mld_version",
3923                         .data           =       &ipv6_devconf.force_mld_version,
3924                         .maxlen         =       sizeof(int),
3925                         .mode           =       0644,
3926                         .proc_handler   =       &proc_dointvec,
3927                 },
3928 #ifdef CONFIG_IPV6_PRIVACY
3929                 {
3930                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3931                         .procname       =       "use_tempaddr",
3932                         .data           =       &ipv6_devconf.use_tempaddr,
3933                         .maxlen         =       sizeof(int),
3934                         .mode           =       0644,
3935                         .proc_handler   =       &proc_dointvec,
3936                 },
3937                 {
3938                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3939                         .procname       =       "temp_valid_lft",
3940                         .data           =       &ipv6_devconf.temp_valid_lft,
3941                         .maxlen         =       sizeof(int),
3942                         .mode           =       0644,
3943                         .proc_handler   =       &proc_dointvec,
3944                 },
3945                 {
3946                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3947                         .procname       =       "temp_prefered_lft",
3948                         .data           =       &ipv6_devconf.temp_prefered_lft,
3949                         .maxlen         =       sizeof(int),
3950                         .mode           =       0644,
3951                         .proc_handler   =       &proc_dointvec,
3952                 },
3953                 {
3954                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3955                         .procname       =       "regen_max_retry",
3956                         .data           =       &ipv6_devconf.regen_max_retry,
3957                         .maxlen         =       sizeof(int),
3958                         .mode           =       0644,
3959                         .proc_handler   =       &proc_dointvec,
3960                 },
3961                 {
3962                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3963                         .procname       =       "max_desync_factor",
3964                         .data           =       &ipv6_devconf.max_desync_factor,
3965                         .maxlen         =       sizeof(int),
3966                         .mode           =       0644,
3967                         .proc_handler   =       &proc_dointvec,
3968                 },
3969 #endif
3970                 {
3971                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3972                         .procname       =       "max_addresses",
3973                         .data           =       &ipv6_devconf.max_addresses,
3974                         .maxlen         =       sizeof(int),
3975                         .mode           =       0644,
3976                         .proc_handler   =       &proc_dointvec,
3977                 },
3978                 {
3979                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3980                         .procname       =       "accept_ra_defrtr",
3981                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3982                         .maxlen         =       sizeof(int),
3983                         .mode           =       0644,
3984                         .proc_handler   =       &proc_dointvec,
3985                 },
3986                 {
3987                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3988                         .procname       =       "accept_ra_pinfo",
3989                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3990                         .maxlen         =       sizeof(int),
3991                         .mode           =       0644,
3992                         .proc_handler   =       &proc_dointvec,
3993                 },
3994 #ifdef CONFIG_IPV6_ROUTER_PREF
3995                 {
3996                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3997                         .procname       =       "accept_ra_rtr_pref",
3998                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3999                         .maxlen         =       sizeof(int),
4000                         .mode           =       0644,
4001                         .proc_handler   =       &proc_dointvec,
4002                 },
4003                 {
4004                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4005                         .procname       =       "router_probe_interval",
4006                         .data           =       &ipv6_devconf.rtr_probe_interval,
4007                         .maxlen         =       sizeof(int),
4008                         .mode           =       0644,
4009                         .proc_handler   =       &proc_dointvec_jiffies,
4010                         .strategy       =       &sysctl_jiffies,
4011                 },
4012 #ifdef CONFIG_IPV6_ROUTE_INFO
4013                 {
4014                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4015                         .procname       =       "accept_ra_rt_info_max_plen",
4016                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4017                         .maxlen         =       sizeof(int),
4018                         .mode           =       0644,
4019                         .proc_handler   =       &proc_dointvec,
4020                 },
4021 #endif
4022 #endif
4023                 {
4024                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4025                         .procname       =       "proxy_ndp",
4026                         .data           =       &ipv6_devconf.proxy_ndp,
4027                         .maxlen         =       sizeof(int),
4028                         .mode           =       0644,
4029                         .proc_handler   =       &proc_dointvec,
4030                 },
4031                 {
4032                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4033                         .procname       =       "accept_source_route",
4034                         .data           =       &ipv6_devconf.accept_source_route,
4035                         .maxlen         =       sizeof(int),
4036                         .mode           =       0644,
4037                         .proc_handler   =       &proc_dointvec,
4038                 },
4039 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4040                 {
4041                         .ctl_name       =       CTL_UNNUMBERED,
4042                         .procname       =       "optimistic_dad",
4043                         .data           =       &ipv6_devconf.optimistic_dad,
4044                         .maxlen         =       sizeof(int),
4045                         .mode           =       0644,
4046                         .proc_handler   =       &proc_dointvec,
4047
4048                 },
4049 #endif
4050                 {
4051                         .ctl_name       =       0,      /* sentinel */
4052                 }
4053         },
4054         .addrconf_dev = {
4055                 {
4056                         .ctl_name       =       NET_PROTO_CONF_ALL,
4057                         .procname       =       "all",
4058                         .mode           =       0555,
4059                         .child          =       addrconf_sysctl.addrconf_vars,
4060                 },
4061                 {
4062                         .ctl_name       =       0,      /* sentinel */
4063                 }
4064         },
4065         .addrconf_conf_dir = {
4066                 {
4067                         .ctl_name       =       NET_IPV6_CONF,
4068                         .procname       =       "conf",
4069                         .mode           =       0555,
4070                         .child          =       addrconf_sysctl.addrconf_dev,
4071                 },
4072                 {
4073                         .ctl_name       =       0,      /* sentinel */
4074                 }
4075         },
4076         .addrconf_proto_dir = {
4077                 {
4078                         .ctl_name       =       NET_IPV6,
4079                         .procname       =       "ipv6",
4080                         .mode           =       0555,
4081                         .child          =       addrconf_sysctl.addrconf_conf_dir,
4082                 },
4083                 {
4084                         .ctl_name       =       0,      /* sentinel */
4085                 }
4086         },
4087         .addrconf_root_dir = {
4088                 {
4089                         .ctl_name       =       CTL_NET,
4090                         .procname       =       "net",
4091                         .mode           =       0555,
4092                         .child          =       addrconf_sysctl.addrconf_proto_dir,
4093                 },
4094                 {
4095                         .ctl_name       =       0,      /* sentinel */
4096                 }
4097         },
4098 };
4099
4100 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
4101 {
4102         int i;
4103         struct net_device *dev = idev ? idev->dev : NULL;
4104         struct addrconf_sysctl_table *t;
4105         char *dev_name = NULL;
4106
4107         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4108         if (t == NULL)
4109                 return;
4110         for (i=0; t->addrconf_vars[i].data; i++) {
4111                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4112                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4113         }
4114         if (dev) {
4115                 dev_name = dev->name;
4116                 t->addrconf_dev[0].ctl_name = dev->ifindex;
4117         } else {
4118                 dev_name = "default";
4119                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
4120         }
4121
4122         /*
4123          * Make a copy of dev_name, because '.procname' is regarded as const
4124          * by sysctl and we wouldn't want anyone to change it under our feet
4125          * (see SIOCSIFNAME).
4126          */
4127         dev_name = kstrdup(dev_name, GFP_KERNEL);
4128         if (!dev_name)
4129             goto free;
4130
4131         t->addrconf_dev[0].procname = dev_name;
4132
4133         t->addrconf_dev[0].child = t->addrconf_vars;
4134         t->addrconf_conf_dir[0].child = t->addrconf_dev;
4135         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4136         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4137
4138         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir);
4139         if (t->sysctl_header == NULL)
4140                 goto free_procname;
4141         else
4142                 p->sysctl = t;
4143         return;
4144
4145         /* error path */
4146  free_procname:
4147         kfree(dev_name);
4148  free:
4149         kfree(t);
4150
4151         return;
4152 }
4153
4154 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4155 {
4156         if (p->sysctl) {
4157                 struct addrconf_sysctl_table *t = p->sysctl;
4158                 p->sysctl = NULL;
4159                 unregister_sysctl_table(t->sysctl_header);
4160                 kfree(t->addrconf_dev[0].procname);
4161                 kfree(t);
4162         }
4163 }
4164
4165
4166 #endif
4167
4168 /*
4169  *      Device notifier
4170  */
4171
4172 int register_inet6addr_notifier(struct notifier_block *nb)
4173 {
4174         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4175 }
4176
4177 EXPORT_SYMBOL(register_inet6addr_notifier);
4178
4179 int unregister_inet6addr_notifier(struct notifier_block *nb)
4180 {
4181         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4182 }
4183
4184 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4185
4186 /*
4187  *      Init / cleanup code
4188  */
4189
4190 int __init addrconf_init(void)
4191 {
4192         int err = 0;
4193
4194         /* The addrconf netdev notifier requires that loopback_dev
4195          * has it's ipv6 private information allocated and setup
4196          * before it can bring up and give link-local addresses
4197          * to other devices which are up.
4198          *
4199          * Unfortunately, loopback_dev is not necessarily the first
4200          * entry in the global dev_base list of net devices.  In fact,
4201          * it is likely to be the very last entry on that list.
4202          * So this causes the notifier registry below to try and
4203          * give link-local addresses to all devices besides loopback_dev
4204          * first, then loopback_dev, which cases all the non-loopback_dev
4205          * devices to fail to get a link-local address.
4206          *
4207          * So, as a temporary fix, allocate the ipv6 structure for
4208          * loopback_dev first by hand.
4209          * Longer term, all of the dependencies ipv6 has upon the loopback
4210          * device and it being up should be removed.
4211          */
4212         rtnl_lock();
4213         if (!ipv6_add_dev(init_net.loopback_dev))
4214                 err = -ENOMEM;
4215         rtnl_unlock();
4216         if (err)
4217                 return err;
4218
4219         ip6_null_entry.u.dst.dev = init_net.loopback_dev;
4220         ip6_null_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4221 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4222         ip6_prohibit_entry.u.dst.dev = init_net.loopback_dev;
4223         ip6_prohibit_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4224         ip6_blk_hole_entry.u.dst.dev = init_net.loopback_dev;
4225         ip6_blk_hole_entry.rt6i_idev = in6_dev_get(init_net.loopback_dev);
4226 #endif
4227
4228         register_netdevice_notifier(&ipv6_dev_notf);
4229
4230         addrconf_verify(0);
4231
4232         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4233         if (err < 0)
4234                 goto errout;
4235
4236         /* Only the first call to __rtnl_register can fail */
4237         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4238         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4239         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4240         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4241         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4242
4243 #ifdef CONFIG_SYSCTL
4244         addrconf_sysctl.sysctl_header =
4245                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir);
4246         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4247 #endif
4248
4249         return 0;
4250 errout:
4251         unregister_netdevice_notifier(&ipv6_dev_notf);
4252
4253         return err;
4254 }
4255
4256 void __exit addrconf_cleanup(void)
4257 {
4258         struct net_device *dev;
4259         struct inet6_ifaddr *ifa;
4260         int i;
4261
4262         unregister_netdevice_notifier(&ipv6_dev_notf);
4263
4264 #ifdef CONFIG_SYSCTL
4265         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4266         addrconf_sysctl_unregister(&ipv6_devconf);
4267 #endif
4268
4269         rtnl_lock();
4270
4271         /*
4272          *      clean dev list.
4273          */
4274
4275         for_each_netdev(&init_net, dev) {
4276                 if (__in6_dev_get(dev) == NULL)
4277                         continue;
4278                 addrconf_ifdown(dev, 1);
4279         }
4280         addrconf_ifdown(init_net.loopback_dev, 2);
4281
4282         /*
4283          *      Check hash table.
4284          */
4285
4286         write_lock_bh(&addrconf_hash_lock);
4287         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4288                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4289                         struct inet6_ifaddr *bifa;
4290
4291                         bifa = ifa;
4292                         ifa = ifa->lst_next;
4293                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4294                         /* Do not free it; something is wrong.
4295                            Now we can investigate it with debugger.
4296                          */
4297                 }
4298         }
4299         write_unlock_bh(&addrconf_hash_lock);
4300
4301         del_timer(&addr_chk_timer);
4302
4303         rtnl_unlock();
4304 }