SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/addrconf.c

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/socket.h>
  44 #include <linux/sockios.h>
  45 #include <linux/net.h>
  46 #include <linux/in6.h>
  47 #include <linux/netdevice.h>
  48 #include <linux/if_addr.h>
  49 #include <linux/if_arp.h>
  50 #include <linux/if_arcnet.h>
  51 #include <linux/if_infiniband.h>
  52 #include <linux/route.h>
  53 #include <linux/inetdevice.h>
  54 #include <linux/init.h>
  55 #ifdef CONFIG_SYSCTL
  56 #include <linux/sysctl.h>
  57 #endif
  58 #include <linux/capability.h>
  59 #include <linux/delay.h>
  60 #include <linux/notifier.h>
  61 #include <linux/string.h>
  62
  63 #include <net/net_namespace.h>
  64 #include <net/sock.h>
  65 #include <net/snmp.h>
  66
  67 #include <net/ipv6.h>
  68 #include <net/protocol.h>
  69 #include <net/ndisc.h>
  70 #include <net/ip6_route.h>
  71 #include <net/addrconf.h>
  72 #include <net/tcp.h>
  73 #include <net/ip.h>
  74 #include <net/netlink.h>
  75 #include <net/pkt_sched.h>
  76 #include <linux/if_tunnel.h>
  77 #include <linux/rtnetlink.h>
  78
  79 #ifdef CONFIG_IPV6_PRIVACY
  80 #include <linux/random.h>
  81 #endif
  82
  83 #include <asm/uaccess.h>
  84 #include <asm/unaligned.h>
  85
  86 #include <linux/proc_fs.h>
  87 #include <linux/seq_file.h>
  88
  89 /* Set to 3 to get tracing... */
  90 #define ACONF_DEBUG 2
  91
  92 #if ACONF_DEBUG >= 3
  93 #define ADBG(x) printk x
  94 #else
  95 #define ADBG(x)
  96 #endif
  97
  98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 100
 101 #ifdef CONFIG_SYSCTL
 102 static void addrconf_sysctl_register(struct inet6_dev *idev);
 103 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 104 #else
 105 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 106 {
 107 }
 108
 109 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 110 {
 111 }
 112 #endif
 113
 114 #ifdef CONFIG_IPV6_PRIVACY
 115 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 116 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 117 static void ipv6_regen_rndid(unsigned long data);
 118
 119 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 120 #endif
 121
 122 static int ipv6_count_addresses(struct inet6_dev *idev);
 123
 124 /*
 125  *      Configured unicast address hash table
 126  */
 127 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 128 static DEFINE_RWLOCK(addrconf_hash_lock);
 129
 130 static void addrconf_verify(unsigned long);
 131
 132 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 133 static DEFINE_SPINLOCK(addrconf_verify_lock);
 134
 135 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 136 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 137
 138 static int addrconf_ifdown(struct net_device *dev, int how);
 139
 140 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 141 static void addrconf_dad_timer(unsigned long data);
 142 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 143 static void addrconf_dad_run(struct inet6_dev *idev);
 144 static void addrconf_rs_timer(unsigned long data);
 145 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 146 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 147
 148 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 149                                 struct prefix_info *pinfo);
 150 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 151                               struct net_device *dev);
 152
 153 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 154
 155 struct ipv6_devconf ipv6_devconf __read_mostly = {
 156         .forwarding             = 0,
 157         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 158         .mtu6                   = IPV6_MIN_MTU,
 159         .accept_ra              = 1,
 160         .accept_redirects       = 1,
 161         .autoconf               = 1,
 162         .force_mld_version      = 0,
 163         .dad_transmits          = 1,
 164         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 165         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 166         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 167 #ifdef CONFIG_IPV6_PRIVACY
 168         .use_tempaddr           = 0,
 169         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 170         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 171         .regen_max_retry        = REGEN_MAX_RETRY,
 172         .max_desync_factor      = MAX_DESYNC_FACTOR,
 173 #endif
 174         .max_addresses          = IPV6_MAX_ADDRESSES,
 175         .accept_ra_defrtr       = 1,
 176         .accept_ra_pinfo        = 1,
 177 #ifdef CONFIG_IPV6_ROUTER_PREF
 178         .accept_ra_rtr_pref     = 1,
 179         .rtr_probe_interval     = 60 * HZ,
 180 #ifdef CONFIG_IPV6_ROUTE_INFO
 181         .accept_ra_rt_info_max_plen = 0,
 182 #endif
 183 #endif
 184         .proxy_ndp              = 0,
 185         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 186 };
 187
 188 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 189         .forwarding             = 0,
 190         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 191         .mtu6                   = IPV6_MIN_MTU,
 192         .accept_ra              = 1,
 193         .accept_redirects       = 1,
 194         .autoconf               = 1,
 195         .dad_transmits          = 1,
 196         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 197         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 198         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 199 #ifdef CONFIG_IPV6_PRIVACY
 200         .use_tempaddr           = 0,
 201         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 202         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 203         .regen_max_retry        = REGEN_MAX_RETRY,
 204         .max_desync_factor      = MAX_DESYNC_FACTOR,
 205 #endif
 206         .max_addresses          = IPV6_MAX_ADDRESSES,
 207         .accept_ra_defrtr       = 1,
 208         .accept_ra_pinfo        = 1,
 209 #ifdef CONFIG_IPV6_ROUTER_PREF
 210         .accept_ra_rtr_pref     = 1,
 211         .rtr_probe_interval     = 60 * HZ,
 212 #ifdef CONFIG_IPV6_ROUTE_INFO
 213         .accept_ra_rt_info_max_plen = 0,
 214 #endif
 215 #endif
 216         .proxy_ndp              = 0,
 217         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 218 };
 219
 220 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 221 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 222 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 223 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 224 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 225
 226 /* Check if a valid qdisc is available */
 227 static inline int addrconf_qdisc_ok(struct net_device *dev)
 228 {
 229         return (dev->qdisc != &noop_qdisc);
 230 }
 231
 232 /* Check if a route is valid prefix route */
 233 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 234 {
 235         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 236 }
 237
 238 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 239 {
 240         if (del_timer(&ifp->timer))
 241                 __in6_ifa_put(ifp);
 242 }
 243
 244 enum addrconf_timer_t
 245 {
 246         AC_NONE,
 247         AC_DAD,
 248         AC_RS,
 249 };
 250
 251 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 252                                enum addrconf_timer_t what,
 253                                unsigned long when)
 254 {
 255         if (!del_timer(&ifp->timer))
 256                 in6_ifa_hold(ifp);
 257
 258         switch (what) {
 259         case AC_DAD:
 260                 ifp->timer.function = addrconf_dad_timer;
 261                 break;
 262         case AC_RS:
 263                 ifp->timer.function = addrconf_rs_timer;
 264                 break;
 265         default:;
 266         }
 267         ifp->timer.expires = jiffies + when;
 268         add_timer(&ifp->timer);
 269 }
 270
 271 static int snmp6_alloc_dev(struct inet6_dev *idev)
 272 {
 273         if (snmp_mib_init((void **)idev->stats.ipv6,
 274                           sizeof(struct ipstats_mib)) < 0)
 275                 goto err_ip;
 276         if (snmp_mib_init((void **)idev->stats.icmpv6,
 277                           sizeof(struct icmpv6_mib)) < 0)
 278                 goto err_icmp;
 279         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 280                           sizeof(struct icmpv6msg_mib)) < 0)
 281                 goto err_icmpmsg;
 282
 283         return 0;
 284
 285 err_icmpmsg:
 286         snmp_mib_free((void **)idev->stats.icmpv6);
 287 err_icmp:
 288         snmp_mib_free((void **)idev->stats.ipv6);
 289 err_ip:
 290         return -ENOMEM;
 291 }
 292
 293 static void snmp6_free_dev(struct inet6_dev *idev)
 294 {
 295         snmp_mib_free((void **)idev->stats.icmpv6msg);
 296         snmp_mib_free((void **)idev->stats.icmpv6);
 297         snmp_mib_free((void **)idev->stats.ipv6);
 298 }
 299
 300 /* Nobody refers to this device, we may destroy it. */
 301
 302 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 303 {
 304         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 305         kfree(idev);
 306 }
 307
 308 void in6_dev_finish_destroy(struct inet6_dev *idev)
 309 {
 310         struct net_device *dev = idev->dev;
 311         BUG_TRAP(idev->addr_list==NULL);
 312         BUG_TRAP(idev->mc_list==NULL);
 313 #ifdef NET_REFCNT_DEBUG
 314         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 315 #endif
 316         dev_put(dev);
 317         if (!idev->dead) {
 318                 printk("Freeing alive inet6 device %p\n", idev);
 319                 return;
 320         }
 321         snmp6_free_dev(idev);
 322         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 323 }
 324
 325 EXPORT_SYMBOL(in6_dev_finish_destroy);
 326
 327 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 328 {
 329         struct inet6_dev *ndev;
 330
 331         ASSERT_RTNL();
 332
 333         if (dev->mtu < IPV6_MIN_MTU)
 334                 return NULL;
 335
 336         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 337
 338         if (ndev == NULL)
 339                 return NULL;
 340
 341         rwlock_init(&ndev->lock);
 342         ndev->dev = dev;
 343         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 344         ndev->cnf.mtu6 = dev->mtu;
 345         ndev->cnf.sysctl = NULL;
 346         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 347         if (ndev->nd_parms == NULL) {
 348                 kfree(ndev);
 349                 return NULL;
 350         }
 351         if (ndev->cnf.forwarding)
 352                 dev_disable_lro(dev);
 353         /* We refer to the device */
 354         dev_hold(dev);
 355
 356         if (snmp6_alloc_dev(ndev) < 0) {
 357                 ADBG((KERN_WARNING
 358                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 359                         __func__, dev->name));
 360                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 361                 ndev->dead = 1;
 362                 in6_dev_finish_destroy(ndev);
 363                 return NULL;
 364         }
 365
 366         if (snmp6_register_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         /* One reference from device.  We must do this before
 377          * we invoke __ipv6_regen_rndid().
 378          */
 379         in6_dev_hold(ndev);
 380
 381 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 382         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 383                 printk(KERN_INFO
 384                        "%s: Disabled Multicast RS\n",
 385                        dev->name);
 386                 ndev->cnf.rtr_solicits = 0;
 387         }
 388 #endif
 389
 390 #ifdef CONFIG_IPV6_PRIVACY
 391         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 392         if ((dev->flags&IFF_LOOPBACK) ||
 393             dev->type == ARPHRD_TUNNEL ||
 394             dev->type == ARPHRD_TUNNEL6 ||
 395             dev->type == ARPHRD_SIT ||
 396             dev->type == ARPHRD_NONE) {
 397                 printk(KERN_INFO
 398                        "%s: Disabled Privacy Extensions\n",
 399                        dev->name);
 400                 ndev->cnf.use_tempaddr = -1;
 401         } else {
 402                 in6_dev_hold(ndev);
 403                 ipv6_regen_rndid((unsigned long) ndev);
 404         }
 405 #endif
 406
 407         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 408                 ndev->if_flags |= IF_READY;
 409
 410         ipv6_mc_init_dev(ndev);
 411         ndev->tstamp = jiffies;
 412         addrconf_sysctl_register(ndev);
 413         /* protected by rtnl_lock */
 414         rcu_assign_pointer(dev->ip6_ptr, ndev);
 415
 416         /* Join all-node multicast group */
 417         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 418
 419         return ndev;
 420 }
 421
 422 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 423 {
 424         struct inet6_dev *idev;
 425
 426         ASSERT_RTNL();
 427
 428         if ((idev = __in6_dev_get(dev)) == NULL) {
 429                 if ((idev = ipv6_add_dev(dev)) == NULL)
 430                         return NULL;
 431         }
 432
 433         if (dev->flags&IFF_UP)
 434                 ipv6_mc_up(idev);
 435         return idev;
 436 }
 437
 438 #ifdef CONFIG_SYSCTL
 439 static void dev_forward_change(struct inet6_dev *idev)
 440 {
 441         struct net_device *dev;
 442         struct inet6_ifaddr *ifa;
 443
 444         if (!idev)
 445                 return;
 446         dev = idev->dev;
 447         if (idev->cnf.forwarding)
 448                 dev_disable_lro(dev);
 449         if (dev && (dev->flags & IFF_MULTICAST)) {
 450                 if (idev->cnf.forwarding)
 451                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 452                 else
 453                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 454         }
 455         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 456                 if (ifa->flags&IFA_F_TENTATIVE)
 457                         continue;
 458                 if (idev->cnf.forwarding)
 459                         addrconf_join_anycast(ifa);
 460                 else
 461                         addrconf_leave_anycast(ifa);
 462         }
 463 }
 464
 465
 466 static void addrconf_forward_change(struct net *net, __s32 newf)
 467 {
 468         struct net_device *dev;
 469         struct inet6_dev *idev;
 470
 471         read_lock(&dev_base_lock);
 472         for_each_netdev(net, dev) {
 473                 rcu_read_lock();
 474                 idev = __in6_dev_get(dev);
 475                 if (idev) {
 476                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 477                         idev->cnf.forwarding = newf;
 478                         if (changed)
 479                                 dev_forward_change(idev);
 480                 }
 481                 rcu_read_unlock();
 482         }
 483         read_unlock(&dev_base_lock);
 484 }
 485
 486 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 487 {
 488         struct net *net;
 489
 490         net = (struct net *)table->extra2;
 491         if (p == &net->ipv6.devconf_dflt->forwarding)
 492                 return;
 493
 494         rtnl_lock();
 495         if (p == &net->ipv6.devconf_all->forwarding) {
 496                 __s32 newf = net->ipv6.devconf_all->forwarding;
 497                 net->ipv6.devconf_dflt->forwarding = newf;
 498                 addrconf_forward_change(net, newf);
 499         } else if ((!*p) ^ (!old))
 500                 dev_forward_change((struct inet6_dev *)table->extra1);
 501         rtnl_unlock();
 502
 503         if (*p)
 504                 rt6_purge_dflt_routers(net);
 505 }
 506 #endif
 507
 508 /* Nobody refers to this ifaddr, destroy it */
 509
 510 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 511 {
 512         BUG_TRAP(ifp->if_next==NULL);
 513         BUG_TRAP(ifp->lst_next==NULL);
 514 #ifdef NET_REFCNT_DEBUG
 515         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 516 #endif
 517
 518         in6_dev_put(ifp->idev);
 519
 520         if (del_timer(&ifp->timer))
 521                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 522
 523         if (!ifp->dead) {
 524                 printk("Freeing alive inet6 address %p\n", ifp);
 525                 return;
 526         }
 527         dst_release(&ifp->rt->u.dst);
 528
 529         kfree(ifp);
 530 }
 531
 532 static void
 533 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 534 {
 535         struct inet6_ifaddr *ifa, **ifap;
 536         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 537
 538         /*
 539          * Each device address list is sorted in order of scope -
 540          * global before linklocal.
 541          */
 542         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 543              ifap = &ifa->if_next) {
 544                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 545                         break;
 546         }
 547
 548         ifp->if_next = *ifap;
 549         *ifap = ifp;
 550 }
 551
 552 /*
 553  *      Hash function taken from net_alias.c
 554  */
 555 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 556 {
 557         __u32 word;
 558
 559         /*
 560          * We perform the hash function over the last 64 bits of the address
 561          * This will include the IEEE address token on links that support it.
 562          */
 563
 564         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 565         word ^= (word >> 16);
 566         word ^= (word >> 8);
 567
 568         return ((word ^ (word >> 4)) & 0x0f);
 569 }
 570
 571 /* On success it returns ifp with increased reference count */
 572
 573 static struct inet6_ifaddr *
 574 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 575               int scope, u32 flags)
 576 {
 577         struct inet6_ifaddr *ifa = NULL;
 578         struct rt6_info *rt;
 579         int hash;
 580         int err = 0;
 581         int addr_type = ipv6_addr_type(addr);
 582
 583         if (addr_type == IPV6_ADDR_ANY ||
 584             addr_type & IPV6_ADDR_MULTICAST ||
 585             (!(idev->dev->flags & IFF_LOOPBACK) &&
 586              addr_type & IPV6_ADDR_LOOPBACK))
 587                 return ERR_PTR(-EADDRNOTAVAIL);
 588
 589         rcu_read_lock_bh();
 590         if (idev->dead) {
 591                 err = -ENODEV;                  /*XXX*/
 592                 goto out2;
 593         }
 594
 595         write_lock(&addrconf_hash_lock);
 596
 597         /* Ignore adding duplicate addresses on an interface */
 598         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 599                 ADBG(("ipv6_add_addr: already assigned\n"));
 600                 err = -EEXIST;
 601                 goto out;
 602         }
 603
 604         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 605
 606         if (ifa == NULL) {
 607                 ADBG(("ipv6_add_addr: malloc failed\n"));
 608                 err = -ENOBUFS;
 609                 goto out;
 610         }
 611
 612         rt = addrconf_dst_alloc(idev, addr, 0);
 613         if (IS_ERR(rt)) {
 614                 err = PTR_ERR(rt);
 615                 goto out;
 616         }
 617
 618         ipv6_addr_copy(&ifa->addr, addr);
 619
 620         spin_lock_init(&ifa->lock);
 621         init_timer(&ifa->timer);
 622         ifa->timer.data = (unsigned long) ifa;
 623         ifa->scope = scope;
 624         ifa->prefix_len = pfxlen;
 625         ifa->flags = flags | IFA_F_TENTATIVE;
 626         ifa->cstamp = ifa->tstamp = jiffies;
 627
 628         ifa->rt = rt;
 629
 630         /*
 631          * part one of RFC 4429, section 3.3
 632          * We should not configure an address as
 633          * optimistic if we do not yet know the link
 634          * layer address of our nexhop router
 635          */
 636
 637         if (rt->rt6i_nexthop == NULL)
 638                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 639
 640         ifa->idev = idev;
 641         in6_dev_hold(idev);
 642         /* For caller */
 643         in6_ifa_hold(ifa);
 644
 645         /* Add to big hash table */
 646         hash = ipv6_addr_hash(addr);
 647
 648         ifa->lst_next = inet6_addr_lst[hash];
 649         inet6_addr_lst[hash] = ifa;
 650         in6_ifa_hold(ifa);
 651         write_unlock(&addrconf_hash_lock);
 652
 653         write_lock(&idev->lock);
 654         /* Add to inet6_dev unicast addr list. */
 655         ipv6_link_dev_addr(idev, ifa);
 656
 657 #ifdef CONFIG_IPV6_PRIVACY
 658         if (ifa->flags&IFA_F_TEMPORARY) {
 659                 ifa->tmp_next = idev->tempaddr_list;
 660                 idev->tempaddr_list = ifa;
 661                 in6_ifa_hold(ifa);
 662         }
 663 #endif
 664
 665         in6_ifa_hold(ifa);
 666         write_unlock(&idev->lock);
 667 out2:
 668         rcu_read_unlock_bh();
 669
 670         if (likely(err == 0))
 671                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 672         else {
 673                 kfree(ifa);
 674                 ifa = ERR_PTR(err);
 675         }
 676
 677         return ifa;
 678 out:
 679         write_unlock(&addrconf_hash_lock);
 680         goto out2;
 681 }
 682
 683 /* This function wants to get referenced ifp and releases it before return */
 684
 685 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 686 {
 687         struct inet6_ifaddr *ifa, **ifap;
 688         struct inet6_dev *idev = ifp->idev;
 689         int hash;
 690         int deleted = 0, onlink = 0;
 691         unsigned long expires = jiffies;
 692
 693         hash = ipv6_addr_hash(&ifp->addr);
 694
 695         ifp->dead = 1;
 696
 697         write_lock_bh(&addrconf_hash_lock);
 698         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 699              ifap = &ifa->lst_next) {
 700                 if (ifa == ifp) {
 701                         *ifap = ifa->lst_next;
 702                         __in6_ifa_put(ifp);
 703                         ifa->lst_next = NULL;
 704                         break;
 705                 }
 706         }
 707         write_unlock_bh(&addrconf_hash_lock);
 708
 709         write_lock_bh(&idev->lock);
 710 #ifdef CONFIG_IPV6_PRIVACY
 711         if (ifp->flags&IFA_F_TEMPORARY) {
 712                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 713                      ifap = &ifa->tmp_next) {
 714                         if (ifa == ifp) {
 715                                 *ifap = ifa->tmp_next;
 716                                 if (ifp->ifpub) {
 717                                         in6_ifa_put(ifp->ifpub);
 718                                         ifp->ifpub = NULL;
 719                                 }
 720                                 __in6_ifa_put(ifp);
 721                                 ifa->tmp_next = NULL;
 722                                 break;
 723                         }
 724                 }
 725         }
 726 #endif
 727
 728         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 729                 if (ifa == ifp) {
 730                         *ifap = ifa->if_next;
 731                         __in6_ifa_put(ifp);
 732                         ifa->if_next = NULL;
 733                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 734                                 break;
 735                         deleted = 1;
 736                         continue;
 737                 } else if (ifp->flags & IFA_F_PERMANENT) {
 738                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 739                                               ifp->prefix_len)) {
 740                                 if (ifa->flags & IFA_F_PERMANENT) {
 741                                         onlink = 1;
 742                                         if (deleted)
 743                                                 break;
 744                                 } else {
 745                                         unsigned long lifetime;
 746
 747                                         if (!onlink)
 748                                                 onlink = -1;
 749
 750                                         spin_lock(&ifa->lock);
 751
 752                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 753                                         /*
 754                                          * Note: Because this address is
 755                                          * not permanent, lifetime <
 756                                          * LONG_MAX / HZ here.
 757                                          */
 758                                         if (time_before(expires,
 759                                                         ifa->tstamp + lifetime * HZ))
 760                                                 expires = ifa->tstamp + lifetime * HZ;
 761                                         spin_unlock(&ifa->lock);
 762                                 }
 763                         }
 764                 }
 765                 ifap = &ifa->if_next;
 766         }
 767         write_unlock_bh(&idev->lock);
 768
 769         ipv6_ifa_notify(RTM_DELADDR, ifp);
 770
 771         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 772
 773         addrconf_del_timer(ifp);
 774
 775         /*
 776          * Purge or update corresponding prefix
 777          *
 778          * 1) we don't purge prefix here if address was not permanent.
 779          *    prefix is managed by its own lifetime.
 780          * 2) if there're no addresses, delete prefix.
 781          * 3) if there're still other permanent address(es),
 782          *    corresponding prefix is still permanent.
 783          * 4) otherwise, update prefix lifetime to the
 784          *    longest valid lifetime among the corresponding
 785          *    addresses on the device.
 786          *    Note: subsequent RA will update lifetime.
 787          *
 788          * --yoshfuji
 789          */
 790         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 791                 struct in6_addr prefix;
 792                 struct rt6_info *rt;
 793                 struct net *net = dev_net(ifp->idev->dev);
 794                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 795                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 796
 797                 if (rt && addrconf_is_prefix_route(rt)) {
 798                         if (onlink == 0) {
 799                                 ip6_del_rt(rt);
 800                                 rt = NULL;
 801                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 802                                 rt->rt6i_expires = expires;
 803                                 rt->rt6i_flags |= RTF_EXPIRES;
 804                         }
 805                 }
 806                 dst_release(&rt->u.dst);
 807         }
 808
 809         in6_ifa_put(ifp);
 810 }
 811
 812 #ifdef CONFIG_IPV6_PRIVACY
 813 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 814 {
 815         struct inet6_dev *idev = ifp->idev;
 816         struct in6_addr addr, *tmpaddr;
 817         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 818         unsigned long regen_advance;
 819         int tmp_plen;
 820         int ret = 0;
 821         int max_addresses;
 822         u32 addr_flags;
 823
 824         write_lock(&idev->lock);
 825         if (ift) {
 826                 spin_lock_bh(&ift->lock);
 827                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 828                 spin_unlock_bh(&ift->lock);
 829                 tmpaddr = &addr;
 830         } else {
 831                 tmpaddr = NULL;
 832         }
 833 retry:
 834         in6_dev_hold(idev);
 835         if (idev->cnf.use_tempaddr <= 0) {
 836                 write_unlock(&idev->lock);
 837                 printk(KERN_INFO
 838                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 839                 in6_dev_put(idev);
 840                 ret = -1;
 841                 goto out;
 842         }
 843         spin_lock_bh(&ifp->lock);
 844         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 845                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 846                 spin_unlock_bh(&ifp->lock);
 847                 write_unlock(&idev->lock);
 848                 printk(KERN_WARNING
 849                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 850                 in6_dev_put(idev);
 851                 ret = -1;
 852                 goto out;
 853         }
 854         in6_ifa_hold(ifp);
 855         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 856         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 857                 spin_unlock_bh(&ifp->lock);
 858                 write_unlock(&idev->lock);
 859                 printk(KERN_WARNING
 860                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 861                 in6_ifa_put(ifp);
 862                 in6_dev_put(idev);
 863                 ret = -1;
 864                 goto out;
 865         }
 866         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 867         tmp_valid_lft = min_t(__u32,
 868                               ifp->valid_lft,
 869                               idev->cnf.temp_valid_lft);
 870         tmp_prefered_lft = min_t(__u32,
 871                                  ifp->prefered_lft,
 872                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 873         tmp_plen = ifp->prefix_len;
 874         max_addresses = idev->cnf.max_addresses;
 875         tmp_cstamp = ifp->cstamp;
 876         tmp_tstamp = ifp->tstamp;
 877         spin_unlock_bh(&ifp->lock);
 878
 879         regen_advance = idev->cnf.regen_max_retry *
 880                         idev->cnf.dad_transmits *
 881                         idev->nd_parms->retrans_time / HZ;
 882         write_unlock(&idev->lock);
 883
 884         /* A temporary address is created only if this calculated Preferred
 885          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 886          * an implementation must not create a temporary address with a zero
 887          * Preferred Lifetime.
 888          */
 889         if (tmp_prefered_lft <= regen_advance) {
 890                 in6_ifa_put(ifp);
 891                 in6_dev_put(idev);
 892                 ret = -1;
 893                 goto out;
 894         }
 895
 896         addr_flags = IFA_F_TEMPORARY;
 897         /* set in addrconf_prefix_rcv() */
 898         if (ifp->flags & IFA_F_OPTIMISTIC)
 899                 addr_flags |= IFA_F_OPTIMISTIC;
 900
 901         ift = !max_addresses ||
 902               ipv6_count_addresses(idev) < max_addresses ?
 903                 ipv6_add_addr(idev, &addr, tmp_plen,
 904                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 905                               addr_flags) : NULL;
 906         if (!ift || IS_ERR(ift)) {
 907                 in6_ifa_put(ifp);
 908                 in6_dev_put(idev);
 909                 printk(KERN_INFO
 910                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 911                 tmpaddr = &addr;
 912                 write_lock(&idev->lock);
 913                 goto retry;
 914         }
 915
 916         spin_lock_bh(&ift->lock);
 917         ift->ifpub = ifp;
 918         ift->valid_lft = tmp_valid_lft;
 919         ift->prefered_lft = tmp_prefered_lft;
 920         ift->cstamp = tmp_cstamp;
 921         ift->tstamp = tmp_tstamp;
 922         spin_unlock_bh(&ift->lock);
 923
 924         addrconf_dad_start(ift, 0);
 925         in6_ifa_put(ift);
 926         in6_dev_put(idev);
 927 out:
 928         return ret;
 929 }
 930 #endif
 931
 932 /*
 933  *      Choose an appropriate source address (RFC3484)
 934  */
 935 enum {
 936         IPV6_SADDR_RULE_INIT = 0,
 937         IPV6_SADDR_RULE_LOCAL,
 938         IPV6_SADDR_RULE_SCOPE,
 939         IPV6_SADDR_RULE_PREFERRED,
 940 #ifdef CONFIG_IPV6_MIP6
 941         IPV6_SADDR_RULE_HOA,
 942 #endif
 943         IPV6_SADDR_RULE_OIF,
 944         IPV6_SADDR_RULE_LABEL,
 945 #ifdef CONFIG_IPV6_PRIVACY
 946         IPV6_SADDR_RULE_PRIVACY,
 947 #endif
 948         IPV6_SADDR_RULE_ORCHID,
 949         IPV6_SADDR_RULE_PREFIX,
 950         IPV6_SADDR_RULE_MAX
 951 };
 952
 953 struct ipv6_saddr_score {
 954         int                     rule;
 955         int                     addr_type;
 956         struct inet6_ifaddr     *ifa;
 957         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 958         int                     scopedist;
 959         int                     matchlen;
 960 };
 961
 962 struct ipv6_saddr_dst {
 963         const struct in6_addr *addr;
 964         int ifindex;
 965         int scope;
 966         int label;
 967         unsigned int prefs;
 968 };
 969
 970 static inline int ipv6_saddr_preferred(int type)
 971 {
 972         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 973                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 974                 return 1;
 975         return 0;
 976 }
 977
 978 static int ipv6_get_saddr_eval(struct net *net,
 979                                struct ipv6_saddr_score *score,
 980                                struct ipv6_saddr_dst *dst,
 981                                int i)
 982 {
 983         int ret;
 984
 985         if (i <= score->rule) {
 986                 switch (i) {
 987                 case IPV6_SADDR_RULE_SCOPE:
 988                         ret = score->scopedist;
 989                         break;
 990                 case IPV6_SADDR_RULE_PREFIX:
 991                         ret = score->matchlen;
 992                         break;
 993                 default:
 994                         ret = !!test_bit(i, score->scorebits);
 995                 }
 996                 goto out;
 997         }
 998
 999         switch (i) {
1000         case IPV6_SADDR_RULE_INIT:
1001                 /* Rule 0: remember if hiscore is not ready yet */
1002                 ret = !!score->ifa;
1003                 break;
1004         case IPV6_SADDR_RULE_LOCAL:
1005                 /* Rule 1: Prefer same address */
1006                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1007                 break;
1008         case IPV6_SADDR_RULE_SCOPE:
1009                 /* Rule 2: Prefer appropriate scope
1010                  *
1011                  *      ret
1012                  *       ^
1013                  *    -1 |  d 15
1014                  *    ---+--+-+---> scope
1015                  *       |
1016                  *       |             d is scope of the destination.
1017                  *  B-d  |  \
1018                  *       |   \      <- smaller scope is better if
1019                  *  B-15 |    \        if scope is enough for destinaion.
1020                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1021                  * d-C-1 | /
1022                  *       |/         <- greater is better
1023                  *   -C  /             if scope is not enough for destination.
1024                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1025                  *
1026                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1027                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1028                  * Assume B = 0 and we get C > 29.
1029                  */
1030                 ret = __ipv6_addr_src_scope(score->addr_type);
1031                 if (ret >= dst->scope)
1032                         ret = -ret;
1033                 else
1034                         ret -= 128;     /* 30 is enough */
1035                 score->scopedist = ret;
1036                 break;
1037         case IPV6_SADDR_RULE_PREFERRED:
1038                 /* Rule 3: Avoid deprecated and optimistic addresses */
1039                 ret = ipv6_saddr_preferred(score->addr_type) ||
1040                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1041                 break;
1042 #ifdef CONFIG_IPV6_MIP6
1043         case IPV6_SADDR_RULE_HOA:
1044             {
1045                 /* Rule 4: Prefer home address */
1046                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1047                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1048                 break;
1049             }
1050 #endif
1051         case IPV6_SADDR_RULE_OIF:
1052                 /* Rule 5: Prefer outgoing interface */
1053                 ret = (!dst->ifindex ||
1054                        dst->ifindex == score->ifa->idev->dev->ifindex);
1055                 break;
1056         case IPV6_SADDR_RULE_LABEL:
1057                 /* Rule 6: Prefer matching label */
1058                 ret = ipv6_addr_label(net,
1059                                       &score->ifa->addr, score->addr_type,
1060                                       score->ifa->idev->dev->ifindex) == dst->label;
1061                 break;
1062 #ifdef CONFIG_IPV6_PRIVACY
1063         case IPV6_SADDR_RULE_PRIVACY:
1064             {
1065                 /* Rule 7: Prefer public address
1066                  * Note: prefer temprary address if use_tempaddr >= 2
1067                  */
1068                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1069                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1070                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1071                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1072                 break;
1073             }
1074 #endif
1075         case IPV6_SADDR_RULE_ORCHID:
1076                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1077                  *          non-ORCHID vs non-ORCHID
1078                  */
1079                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1080                         ipv6_addr_orchid(dst->addr));
1081                 break;
1082         case IPV6_SADDR_RULE_PREFIX:
1083                 /* Rule 8: Use longest matching prefix */
1084                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1085                                                        dst->addr);
1086                 break;
1087         default:
1088                 ret = 0;
1089         }
1090
1091         if (ret)
1092                 __set_bit(i, score->scorebits);
1093         score->rule = i;
1094 out:
1095         return ret;
1096 }
1097
1098 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1099                        const struct in6_addr *daddr, unsigned int prefs,
1100                        struct in6_addr *saddr)
1101 {
1102         struct ipv6_saddr_score scores[2],
1103                                 *score = &scores[0], *hiscore = &scores[1];
1104         struct net *net = dev_net(dst_dev);
1105         struct ipv6_saddr_dst dst;
1106         struct net_device *dev;
1107         int dst_type;
1108
1109         dst_type = __ipv6_addr_type(daddr);
1110         dst.addr = daddr;
1111         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1112         dst.scope = __ipv6_addr_src_scope(dst_type);
1113         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1114         dst.prefs = prefs;
1115
1116         hiscore->rule = -1;
1117         hiscore->ifa = NULL;
1118
1119         read_lock(&dev_base_lock);
1120         rcu_read_lock();
1121
1122         for_each_netdev(net, dev) {
1123                 struct inet6_dev *idev;
1124
1125                 /* Candidate Source Address (section 4)
1126                  *  - multicast and link-local destination address,
1127                  *    the set of candidate source address MUST only
1128                  *    include addresses assigned to interfaces
1129                  *    belonging to the same link as the outgoing
1130                  *    interface.
1131                  * (- For site-local destination addresses, the
1132                  *    set of candidate source addresses MUST only
1133                  *    include addresses assigned to interfaces
1134                  *    belonging to the same site as the outgoing
1135                  *    interface.)
1136                  */
1137                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1138                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1139                     dst.ifindex && dev->ifindex != dst.ifindex)
1140                         continue;
1141
1142                 idev = __in6_dev_get(dev);
1143                 if (!idev)
1144                         continue;
1145
1146                 read_lock_bh(&idev->lock);
1147                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1148                         int i;
1149
1150                         /*
1151                          * - Tentative Address (RFC2462 section 5.4)
1152                          *  - A tentative address is not considered
1153                          *    "assigned to an interface" in the traditional
1154                          *    sense, unless it is also flagged as optimistic.
1155                          * - Candidate Source Address (section 4)
1156                          *  - In any case, anycast addresses, multicast
1157                          *    addresses, and the unspecified address MUST
1158                          *    NOT be included in a candidate set.
1159                          */
1160                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1161                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1162                                 continue;
1163
1164                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1165
1166                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1167                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1168                                 LIMIT_NETDEBUG(KERN_DEBUG
1169                                                "ADDRCONF: unspecified / multicast address "
1170                                                "assigned as unicast address on %s",
1171                                                dev->name);
1172                                 continue;
1173                         }
1174
1175                         score->rule = -1;
1176                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1177
1178                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1179                                 int minihiscore, miniscore;
1180
1181                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1182                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1183
1184                                 if (minihiscore > miniscore) {
1185                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1186                                             score->scopedist > 0) {
1187                                                 /*
1188                                                  * special case:
1189                                                  * each remaining entry
1190                                                  * has too small (not enough)
1191                                                  * scope, because ifa entries
1192                                                  * are sorted by their scope
1193                                                  * values.
1194                                                  */
1195                                                 goto try_nextdev;
1196                                         }
1197                                         break;
1198                                 } else if (minihiscore < miniscore) {
1199                                         struct ipv6_saddr_score *tmp;
1200
1201                                         if (hiscore->ifa)
1202                                                 in6_ifa_put(hiscore->ifa);
1203
1204                                         in6_ifa_hold(score->ifa);
1205
1206                                         tmp = hiscore;
1207                                         hiscore = score;
1208                                         score = tmp;
1209
1210                                         /* restore our iterator */
1211                                         score->ifa = hiscore->ifa;
1212
1213                                         break;
1214                                 }
1215                         }
1216                 }
1217 try_nextdev:
1218                 read_unlock_bh(&idev->lock);
1219         }
1220         rcu_read_unlock();
1221         read_unlock(&dev_base_lock);
1222
1223         if (!hiscore->ifa)
1224                 return -EADDRNOTAVAIL;
1225
1226         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1227         in6_ifa_put(hiscore->ifa);
1228         return 0;
1229 }
1230
1231 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1232
1233 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1234                     unsigned char banned_flags)
1235 {
1236         struct inet6_dev *idev;
1237         int err = -EADDRNOTAVAIL;
1238
1239         rcu_read_lock();
1240         if ((idev = __in6_dev_get(dev)) != NULL) {
1241                 struct inet6_ifaddr *ifp;
1242
1243                 read_lock_bh(&idev->lock);
1244                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1245                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1246                                 ipv6_addr_copy(addr, &ifp->addr);
1247                                 err = 0;
1248                                 break;
1249                         }
1250                 }
1251                 read_unlock_bh(&idev->lock);
1252         }
1253         rcu_read_unlock();
1254         return err;
1255 }
1256
1257 static int ipv6_count_addresses(struct inet6_dev *idev)
1258 {
1259         int cnt = 0;
1260         struct inet6_ifaddr *ifp;
1261
1262         read_lock_bh(&idev->lock);
1263         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1264                 cnt++;
1265         read_unlock_bh(&idev->lock);
1266         return cnt;
1267 }
1268
1269 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1270                   struct net_device *dev, int strict)
1271 {
1272         struct inet6_ifaddr * ifp;
1273         u8 hash = ipv6_addr_hash(addr);
1274
1275         read_lock_bh(&addrconf_hash_lock);
1276         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1277                 if (!net_eq(dev_net(ifp->idev->dev), net))
1278                         continue;
1279                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1280                     !(ifp->flags&IFA_F_TENTATIVE)) {
1281                         if (dev == NULL || ifp->idev->dev == dev ||
1282                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1283                                 break;
1284                 }
1285         }
1286         read_unlock_bh(&addrconf_hash_lock);
1287         return ifp != NULL;
1288 }
1289 EXPORT_SYMBOL(ipv6_chk_addr);
1290
1291 static
1292 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1293                        struct net_device *dev)
1294 {
1295         struct inet6_ifaddr * ifp;
1296         u8 hash = ipv6_addr_hash(addr);
1297
1298         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1299                 if (!net_eq(dev_net(ifp->idev->dev), net))
1300                         continue;
1301                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1302                         if (dev == NULL || ifp->idev->dev == dev)
1303                                 break;
1304                 }
1305         }
1306         return ifp != NULL;
1307 }
1308
1309 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1310 {
1311         struct inet6_dev *idev;
1312         struct inet6_ifaddr *ifa;
1313         int     onlink;
1314
1315         onlink = 0;
1316         rcu_read_lock();
1317         idev = __in6_dev_get(dev);
1318         if (idev) {
1319                 read_lock_bh(&idev->lock);
1320                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1321                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1322                                                    ifa->prefix_len);
1323                         if (onlink)
1324                                 break;
1325                 }
1326                 read_unlock_bh(&idev->lock);
1327         }
1328         rcu_read_unlock();
1329         return onlink;
1330 }
1331
1332 EXPORT_SYMBOL(ipv6_chk_prefix);
1333
1334 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1335                                      struct net_device *dev, int strict)
1336 {
1337         struct inet6_ifaddr * ifp;
1338         u8 hash = ipv6_addr_hash(addr);
1339
1340         read_lock_bh(&addrconf_hash_lock);
1341         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1342                 if (!net_eq(dev_net(ifp->idev->dev), net))
1343                         continue;
1344                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1345                         if (dev == NULL || ifp->idev->dev == dev ||
1346                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1347                                 in6_ifa_hold(ifp);
1348                                 break;
1349                         }
1350                 }
1351         }
1352         read_unlock_bh(&addrconf_hash_lock);
1353
1354         return ifp;
1355 }
1356
1357 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1358 {
1359         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1360         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1361         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1362         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1363         int sk_ipv6only = ipv6_only_sock(sk);
1364         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1365         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1366         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1367
1368         if (!sk2_rcv_saddr && !sk_ipv6only)
1369                 return 1;
1370
1371         if (addr_type2 == IPV6_ADDR_ANY &&
1372             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1373                 return 1;
1374
1375         if (addr_type == IPV6_ADDR_ANY &&
1376             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1377                 return 1;
1378
1379         if (sk2_rcv_saddr6 &&
1380             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1381                 return 1;
1382
1383         if (addr_type == IPV6_ADDR_MAPPED &&
1384             !sk2_ipv6only &&
1385             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1386                 return 1;
1387
1388         return 0;
1389 }
1390
1391 /* Gets referenced address, destroys ifaddr */
1392
1393 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1394 {
1395         if (ifp->flags&IFA_F_PERMANENT) {
1396                 spin_lock_bh(&ifp->lock);
1397                 addrconf_del_timer(ifp);
1398                 ifp->flags |= IFA_F_TENTATIVE;
1399                 spin_unlock_bh(&ifp->lock);
1400                 in6_ifa_put(ifp);
1401 #ifdef CONFIG_IPV6_PRIVACY
1402         } else if (ifp->flags&IFA_F_TEMPORARY) {
1403                 struct inet6_ifaddr *ifpub;
1404                 spin_lock_bh(&ifp->lock);
1405                 ifpub = ifp->ifpub;
1406                 if (ifpub) {
1407                         in6_ifa_hold(ifpub);
1408                         spin_unlock_bh(&ifp->lock);
1409                         ipv6_create_tempaddr(ifpub, ifp);
1410                         in6_ifa_put(ifpub);
1411                 } else {
1412                         spin_unlock_bh(&ifp->lock);
1413                 }
1414                 ipv6_del_addr(ifp);
1415 #endif
1416         } else
1417                 ipv6_del_addr(ifp);
1418 }
1419
1420 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1421 {
1422         if (net_ratelimit())
1423                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1424         addrconf_dad_stop(ifp);
1425 }
1426
1427 /* Join to solicited addr multicast group. */
1428
1429 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1430 {
1431         struct in6_addr maddr;
1432
1433         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1434                 return;
1435
1436         addrconf_addr_solict_mult(addr, &maddr);
1437         ipv6_dev_mc_inc(dev, &maddr);
1438 }
1439
1440 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1441 {
1442         struct in6_addr maddr;
1443
1444         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1445                 return;
1446
1447         addrconf_addr_solict_mult(addr, &maddr);
1448         __ipv6_dev_mc_dec(idev, &maddr);
1449 }
1450
1451 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1452 {
1453         struct in6_addr addr;
1454         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1455         if (ipv6_addr_any(&addr))
1456                 return;
1457         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1458 }
1459
1460 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1461 {
1462         struct in6_addr addr;
1463         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1464         if (ipv6_addr_any(&addr))
1465                 return;
1466         __ipv6_dev_ac_dec(ifp->idev, &addr);
1467 }
1468
1469 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1470 {
1471         if (dev->addr_len != ETH_ALEN)
1472                 return -1;
1473         memcpy(eui, dev->dev_addr, 3);
1474         memcpy(eui + 5, dev->dev_addr + 3, 3);
1475
1476         /*
1477          * The zSeries OSA network cards can be shared among various
1478          * OS instances, but the OSA cards have only one MAC address.
1479          * This leads to duplicate address conflicts in conjunction
1480          * with IPv6 if more than one instance uses the same card.
1481          *
1482          * The driver for these cards can deliver a unique 16-bit
1483          * identifier for each instance sharing the same card.  It is
1484          * placed instead of 0xFFFE in the interface identifier.  The
1485          * "u" bit of the interface identifier is not inverted in this
1486          * case.  Hence the resulting interface identifier has local
1487          * scope according to RFC2373.
1488          */
1489         if (dev->dev_id) {
1490                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1491                 eui[4] = dev->dev_id & 0xFF;
1492         } else {
1493                 eui[3] = 0xFF;
1494                 eui[4] = 0xFE;
1495                 eui[0] ^= 2;
1496         }
1497         return 0;
1498 }
1499
1500 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1501 {
1502         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1503         if (dev->addr_len != ARCNET_ALEN)
1504                 return -1;
1505         memset(eui, 0, 7);
1506         eui[7] = *(u8*)dev->dev_addr;
1507         return 0;
1508 }
1509
1510 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1511 {
1512         if (dev->addr_len != INFINIBAND_ALEN)
1513                 return -1;
1514         memcpy(eui, dev->dev_addr + 12, 8);
1515         eui[0] |= 2;
1516         return 0;
1517 }
1518
1519 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1520 {
1521         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1522                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1523                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1524                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1525                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1526                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1527         eui[1] = 0;
1528         eui[2] = 0x5E;
1529         eui[3] = 0xFE;
1530         memcpy(eui + 4, &addr, 4);
1531         return 0;
1532 }
1533 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1534
1535 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1536 {
1537         if (dev->priv_flags & IFF_ISATAP)
1538                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1539         return -1;
1540 }
1541
1542 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1543 {
1544         switch (dev->type) {
1545         case ARPHRD_ETHER:
1546         case ARPHRD_FDDI:
1547         case ARPHRD_IEEE802_TR:
1548                 return addrconf_ifid_eui48(eui, dev);
1549         case ARPHRD_ARCNET:
1550                 return addrconf_ifid_arcnet(eui, dev);
1551         case ARPHRD_INFINIBAND:
1552                 return addrconf_ifid_infiniband(eui, dev);
1553         case ARPHRD_SIT:
1554                 return addrconf_ifid_sit(eui, dev);
1555         }
1556         return -1;
1557 }
1558
1559 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1560 {
1561         int err = -1;
1562         struct inet6_ifaddr *ifp;
1563
1564         read_lock_bh(&idev->lock);
1565         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1566                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1567                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1568                         err = 0;
1569                         break;
1570                 }
1571         }
1572         read_unlock_bh(&idev->lock);
1573         return err;
1574 }
1575
1576 #ifdef CONFIG_IPV6_PRIVACY
1577 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1578 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1579 {
1580 regen:
1581         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1582         idev->rndid[0] &= ~0x02;
1583
1584         /*
1585          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1586          * check if generated address is not inappropriate
1587          *
1588          *  - Reserved subnet anycast (RFC 2526)
1589          *      11111101 11....11 1xxxxxxx
1590          *  - ISATAP (RFC4214) 6.1
1591          *      00-00-5E-FE-xx-xx-xx-xx
1592          *  - value 0
1593          *  - XXX: already assigned to an address on the device
1594          */
1595         if (idev->rndid[0] == 0xfd &&
1596             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1597             (idev->rndid[7]&0x80))
1598                 goto regen;
1599         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1600                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1601                         goto regen;
1602                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1603                         goto regen;
1604         }
1605
1606         return 0;
1607 }
1608
1609 static void ipv6_regen_rndid(unsigned long data)
1610 {
1611         struct inet6_dev *idev = (struct inet6_dev *) data;
1612         unsigned long expires;
1613
1614         rcu_read_lock_bh();
1615         write_lock_bh(&idev->lock);
1616
1617         if (idev->dead)
1618                 goto out;
1619
1620         if (__ipv6_regen_rndid(idev) < 0)
1621                 goto out;
1622
1623         expires = jiffies +
1624                 idev->cnf.temp_prefered_lft * HZ -
1625                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1626         if (time_before(expires, jiffies)) {
1627                 printk(KERN_WARNING
1628                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1629                         idev->dev->name);
1630                 goto out;
1631         }
1632
1633         if (!mod_timer(&idev->regen_timer, expires))
1634                 in6_dev_hold(idev);
1635
1636 out:
1637         write_unlock_bh(&idev->lock);
1638         rcu_read_unlock_bh();
1639         in6_dev_put(idev);
1640 }
1641
1642 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1643         int ret = 0;
1644
1645         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1646                 ret = __ipv6_regen_rndid(idev);
1647         return ret;
1648 }
1649 #endif
1650
1651 /*
1652  *      Add prefix route.
1653  */
1654
1655 static void
1656 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1657                       unsigned long expires, u32 flags)
1658 {
1659         struct fib6_config cfg = {
1660                 .fc_table = RT6_TABLE_PREFIX,
1661                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1662                 .fc_ifindex = dev->ifindex,
1663                 .fc_expires = expires,
1664                 .fc_dst_len = plen,
1665                 .fc_flags = RTF_UP | flags,
1666                 .fc_nlinfo.nl_net = dev_net(dev),
1667         };
1668
1669         ipv6_addr_copy(&cfg.fc_dst, pfx);
1670
1671         /* Prevent useless cloning on PtP SIT.
1672            This thing is done here expecting that the whole
1673            class of non-broadcast devices need not cloning.
1674          */
1675 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1676         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1677                 cfg.fc_flags |= RTF_NONEXTHOP;
1678 #endif
1679
1680         ip6_route_add(&cfg);
1681 }
1682
1683 /* Create "default" multicast route to the interface */
1684
1685 static void addrconf_add_mroute(struct net_device *dev)
1686 {
1687         struct fib6_config cfg = {
1688                 .fc_table = RT6_TABLE_LOCAL,
1689                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1690                 .fc_ifindex = dev->ifindex,
1691                 .fc_dst_len = 8,
1692                 .fc_flags = RTF_UP,
1693                 .fc_nlinfo.nl_net = dev_net(dev),
1694         };
1695
1696         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1697
1698         ip6_route_add(&cfg);
1699 }
1700
1701 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1702 static void sit_route_add(struct net_device *dev)
1703 {
1704         struct fib6_config cfg = {
1705                 .fc_table = RT6_TABLE_MAIN,
1706                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1707                 .fc_ifindex = dev->ifindex,
1708                 .fc_dst_len = 96,
1709                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1710                 .fc_nlinfo.nl_net = dev_net(dev),
1711         };
1712
1713         /* prefix length - 96 bits "::d.d.d.d" */
1714         ip6_route_add(&cfg);
1715 }
1716 #endif
1717
1718 static void addrconf_add_lroute(struct net_device *dev)
1719 {
1720         struct in6_addr addr;
1721
1722         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1723         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1724 }
1725
1726 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1727 {
1728         struct inet6_dev *idev;
1729
1730         ASSERT_RTNL();
1731
1732         if ((idev = ipv6_find_idev(dev)) == NULL)
1733                 return NULL;
1734
1735         /* Add default multicast route */
1736         addrconf_add_mroute(dev);
1737
1738         /* Add link local route */
1739         addrconf_add_lroute(dev);
1740         return idev;
1741 }
1742
1743 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1744 {
1745         struct prefix_info *pinfo;
1746         __u32 valid_lft;
1747         __u32 prefered_lft;
1748         int addr_type;
1749         struct inet6_dev *in6_dev;
1750
1751         pinfo = (struct prefix_info *) opt;
1752
1753         if (len < sizeof(struct prefix_info)) {
1754                 ADBG(("addrconf: prefix option too short\n"));
1755                 return;
1756         }
1757
1758         /*
1759          *      Validation checks ([ADDRCONF], page 19)
1760          */
1761
1762         addr_type = ipv6_addr_type(&pinfo->prefix);
1763
1764         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1765                 return;
1766
1767         valid_lft = ntohl(pinfo->valid);
1768         prefered_lft = ntohl(pinfo->prefered);
1769
1770         if (prefered_lft > valid_lft) {
1771                 if (net_ratelimit())
1772                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1773                 return;
1774         }
1775
1776         in6_dev = in6_dev_get(dev);
1777
1778         if (in6_dev == NULL) {
1779                 if (net_ratelimit())
1780                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1781                 return;
1782         }
1783
1784         /*
1785          *      Two things going on here:
1786          *      1) Add routes for on-link prefixes
1787          *      2) Configure prefixes with the auto flag set
1788          */
1789
1790         if (pinfo->onlink) {
1791                 struct rt6_info *rt;
1792                 unsigned long rt_expires;
1793
1794                 /* Avoid arithmetic overflow. Really, we could
1795                  * save rt_expires in seconds, likely valid_lft,
1796                  * but it would require division in fib gc, that it
1797                  * not good.
1798                  */
1799                 if (HZ > USER_HZ)
1800                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1801                 else
1802                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1803
1804                 if (addrconf_finite_timeout(rt_expires))
1805                         rt_expires *= HZ;
1806
1807                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1808                                 dev->ifindex, 1);
1809
1810                 if (rt && addrconf_is_prefix_route(rt)) {
1811                         /* Autoconf prefix route */
1812                         if (valid_lft == 0) {
1813                                 ip6_del_rt(rt);
1814                                 rt = NULL;
1815                         } else if (addrconf_finite_timeout(rt_expires)) {
1816                                 /* not infinity */
1817                                 rt->rt6i_expires = jiffies + rt_expires;
1818                                 rt->rt6i_flags |= RTF_EXPIRES;
1819                         } else {
1820                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1821                                 rt->rt6i_expires = 0;
1822                         }
1823                 } else if (valid_lft) {
1824                         clock_t expires = 0;
1825                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1826                         if (addrconf_finite_timeout(rt_expires)) {
1827                                 /* not infinity */
1828                                 flags |= RTF_EXPIRES;
1829                                 expires = jiffies_to_clock_t(rt_expires);
1830                         }
1831                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1832                                               dev, expires, flags);
1833                 }
1834                 if (rt)
1835                         dst_release(&rt->u.dst);
1836         }
1837
1838         /* Try to figure out our local address for this prefix */
1839
1840         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1841                 struct inet6_ifaddr * ifp;
1842                 struct in6_addr addr;
1843                 int create = 0, update_lft = 0;
1844
1845                 if (pinfo->prefix_len == 64) {
1846                         memcpy(&addr, &pinfo->prefix, 8);
1847                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1848                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1849                                 in6_dev_put(in6_dev);
1850                                 return;
1851                         }
1852                         goto ok;
1853                 }
1854                 if (net_ratelimit())
1855                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1856                                pinfo->prefix_len);
1857                 in6_dev_put(in6_dev);
1858                 return;
1859
1860 ok:
1861
1862                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1863
1864                 if (ifp == NULL && valid_lft) {
1865                         int max_addresses = in6_dev->cnf.max_addresses;
1866                         u32 addr_flags = 0;
1867
1868 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1869                         if (in6_dev->cnf.optimistic_dad &&
1870                             !ipv6_devconf.forwarding)
1871                                 addr_flags = IFA_F_OPTIMISTIC;
1872 #endif
1873
1874                         /* Do not allow to create too much of autoconfigured
1875                          * addresses; this would be too easy way to crash kernel.
1876                          */
1877                         if (!max_addresses ||
1878                             ipv6_count_addresses(in6_dev) < max_addresses)
1879                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1880                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1881                                                     addr_flags);
1882
1883                         if (!ifp || IS_ERR(ifp)) {
1884                                 in6_dev_put(in6_dev);
1885                                 return;
1886                         }
1887
1888                         update_lft = create = 1;
1889                         ifp->cstamp = jiffies;
1890                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1891                 }
1892
1893                 if (ifp) {
1894                         int flags;
1895                         unsigned long now;
1896 #ifdef CONFIG_IPV6_PRIVACY
1897                         struct inet6_ifaddr *ift;
1898 #endif
1899                         u32 stored_lft;
1900
1901                         /* update lifetime (RFC2462 5.5.3 e) */
1902                         spin_lock(&ifp->lock);
1903                         now = jiffies;
1904                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1905                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1906                         else
1907                                 stored_lft = 0;
1908                         if (!update_lft && stored_lft) {
1909                                 if (valid_lft > MIN_VALID_LIFETIME ||
1910                                     valid_lft > stored_lft)
1911                                         update_lft = 1;
1912                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1913                                         /* valid_lft <= stored_lft is always true */
1914                                         /* XXX: IPsec */
1915                                         update_lft = 0;
1916                                 } else {
1917                                         valid_lft = MIN_VALID_LIFETIME;
1918                                         if (valid_lft < prefered_lft)
1919                                                 prefered_lft = valid_lft;
1920                                         update_lft = 1;
1921                                 }
1922                         }
1923
1924                         if (update_lft) {
1925                                 ifp->valid_lft = valid_lft;
1926                                 ifp->prefered_lft = prefered_lft;
1927                                 ifp->tstamp = now;
1928                                 flags = ifp->flags;
1929                                 ifp->flags &= ~IFA_F_DEPRECATED;
1930                                 spin_unlock(&ifp->lock);
1931
1932                                 if (!(flags&IFA_F_TENTATIVE))
1933                                         ipv6_ifa_notify(0, ifp);
1934                         } else
1935                                 spin_unlock(&ifp->lock);
1936
1937 #ifdef CONFIG_IPV6_PRIVACY
1938                         read_lock_bh(&in6_dev->lock);
1939                         /* update all temporary addresses in the list */
1940                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1941                                 /*
1942                                  * When adjusting the lifetimes of an existing
1943                                  * temporary address, only lower the lifetimes.
1944                                  * Implementations must not increase the
1945                                  * lifetimes of an existing temporary address
1946                                  * when processing a Prefix Information Option.
1947                                  */
1948                                 if (ifp != ift->ifpub)
1949                                         continue;
1950
1951                                 spin_lock(&ift->lock);
1952                                 flags = ift->flags;
1953                                 if (ift->valid_lft > valid_lft &&
1954                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1955                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1956                                 if (ift->prefered_lft > prefered_lft &&
1957                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1958                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1959                                 spin_unlock(&ift->lock);
1960                                 if (!(flags&IFA_F_TENTATIVE))
1961                                         ipv6_ifa_notify(0, ift);
1962                         }
1963
1964                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1965                                 /*
1966                                  * When a new public address is created as described in [ADDRCONF],
1967                                  * also create a new temporary address.
1968                                  */
1969                                 read_unlock_bh(&in6_dev->lock);
1970                                 ipv6_create_tempaddr(ifp, NULL);
1971                         } else {
1972                                 read_unlock_bh(&in6_dev->lock);
1973                         }
1974 #endif
1975                         in6_ifa_put(ifp);
1976                         addrconf_verify(0);
1977                 }
1978         }
1979         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1980         in6_dev_put(in6_dev);
1981 }
1982
1983 /*
1984  *      Set destination address.
1985  *      Special case for SIT interfaces where we create a new "virtual"
1986  *      device.
1987  */
1988 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1989 {
1990         struct in6_ifreq ireq;
1991         struct net_device *dev;
1992         int err = -EINVAL;
1993
1994         rtnl_lock();
1995
1996         err = -EFAULT;
1997         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1998                 goto err_exit;
1999
2000         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2001
2002         err = -ENODEV;
2003         if (dev == NULL)
2004                 goto err_exit;
2005
2006 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2007         if (dev->type == ARPHRD_SIT) {
2008                 struct ifreq ifr;
2009                 mm_segment_t    oldfs;
2010                 struct ip_tunnel_parm p;
2011
2012                 err = -EADDRNOTAVAIL;
2013                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2014                         goto err_exit;
2015
2016                 memset(&p, 0, sizeof(p));
2017                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2018                 p.iph.saddr = 0;
2019                 p.iph.version = 4;
2020                 p.iph.ihl = 5;
2021                 p.iph.protocol = IPPROTO_IPV6;
2022                 p.iph.ttl = 64;
2023                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2024
2025                 oldfs = get_fs(); set_fs(KERNEL_DS);
2026                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2027                 set_fs(oldfs);
2028
2029                 if (err == 0) {
2030                         err = -ENOBUFS;
2031                         dev = __dev_get_by_name(net, p.name);
2032                         if (!dev)
2033                                 goto err_exit;
2034                         err = dev_open(dev);
2035                 }
2036         }
2037 #endif
2038
2039 err_exit:
2040         rtnl_unlock();
2041         return err;
2042 }
2043
2044 /*
2045  *      Manual configuration of address on an interface
2046  */
2047 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2048                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2049                           __u32 valid_lft)
2050 {
2051         struct inet6_ifaddr *ifp;
2052         struct inet6_dev *idev;
2053         struct net_device *dev;
2054         int scope;
2055         u32 flags;
2056         clock_t expires;
2057         unsigned long timeout;
2058
2059         ASSERT_RTNL();
2060
2061         if (plen > 128)
2062                 return -EINVAL;
2063
2064         /* check the lifetime */
2065         if (!valid_lft || prefered_lft > valid_lft)
2066                 return -EINVAL;
2067
2068         dev = __dev_get_by_index(net, ifindex);
2069         if (!dev)
2070                 return -ENODEV;
2071
2072         if ((idev = addrconf_add_dev(dev)) == NULL)
2073                 return -ENOBUFS;
2074
2075         scope = ipv6_addr_scope(pfx);
2076
2077         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2078         if (addrconf_finite_timeout(timeout)) {
2079                 expires = jiffies_to_clock_t(timeout * HZ);
2080                 valid_lft = timeout;
2081                 flags = RTF_EXPIRES;
2082         } else {
2083                 expires = 0;
2084                 flags = 0;
2085                 ifa_flags |= IFA_F_PERMANENT;
2086         }
2087
2088         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2089         if (addrconf_finite_timeout(timeout)) {
2090                 if (timeout == 0)
2091                         ifa_flags |= IFA_F_DEPRECATED;
2092                 prefered_lft = timeout;
2093         }
2094
2095         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2096
2097         if (!IS_ERR(ifp)) {
2098                 spin_lock_bh(&ifp->lock);
2099                 ifp->valid_lft = valid_lft;
2100                 ifp->prefered_lft = prefered_lft;
2101                 ifp->tstamp = jiffies;
2102                 spin_unlock_bh(&ifp->lock);
2103
2104                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2105                                       expires, flags);
2106                 /*
2107                  * Note that section 3.1 of RFC 4429 indicates
2108                  * that the Optimistic flag should not be set for
2109                  * manually configured addresses
2110                  */
2111                 addrconf_dad_start(ifp, 0);
2112                 in6_ifa_put(ifp);
2113                 addrconf_verify(0);
2114                 return 0;
2115         }
2116
2117         return PTR_ERR(ifp);
2118 }
2119
2120 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2121                           unsigned int plen)
2122 {
2123         struct inet6_ifaddr *ifp;
2124         struct inet6_dev *idev;
2125         struct net_device *dev;
2126
2127         if (plen > 128)
2128                 return -EINVAL;
2129
2130         dev = __dev_get_by_index(net, ifindex);
2131         if (!dev)
2132                 return -ENODEV;
2133
2134         if ((idev = __in6_dev_get(dev)) == NULL)
2135                 return -ENXIO;
2136
2137         read_lock_bh(&idev->lock);
2138         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2139                 if (ifp->prefix_len == plen &&
2140                     ipv6_addr_equal(pfx, &ifp->addr)) {
2141                         in6_ifa_hold(ifp);
2142                         read_unlock_bh(&idev->lock);
2143
2144                         ipv6_del_addr(ifp);
2145
2146                         /* If the last address is deleted administratively,
2147                            disable IPv6 on this interface.
2148                          */
2149                         if (idev->addr_list == NULL)
2150                                 addrconf_ifdown(idev->dev, 1);
2151                         return 0;
2152                 }
2153         }
2154         read_unlock_bh(&idev->lock);
2155         return -EADDRNOTAVAIL;
2156 }
2157
2158
2159 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2160 {
2161         struct in6_ifreq ireq;
2162         int err;
2163
2164         if (!capable(CAP_NET_ADMIN))
2165                 return -EPERM;
2166
2167         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2168                 return -EFAULT;
2169
2170         rtnl_lock();
2171         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2172                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2173                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2174         rtnl_unlock();
2175         return err;
2176 }
2177
2178 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2179 {
2180         struct in6_ifreq ireq;
2181         int err;
2182
2183         if (!capable(CAP_NET_ADMIN))
2184                 return -EPERM;
2185
2186         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2187                 return -EFAULT;
2188
2189         rtnl_lock();
2190         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2191                              ireq.ifr6_prefixlen);
2192         rtnl_unlock();
2193         return err;
2194 }
2195
2196 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2197 static void sit_add_v4_addrs(struct inet6_dev *idev)
2198 {
2199         struct inet6_ifaddr * ifp;
2200         struct in6_addr addr;
2201         struct net_device *dev;
2202         struct net *net = dev_net(idev->dev);
2203         int scope;
2204
2205         ASSERT_RTNL();
2206
2207         memset(&addr, 0, sizeof(struct in6_addr));
2208         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2209
2210         if (idev->dev->flags&IFF_POINTOPOINT) {
2211                 addr.s6_addr32[0] = htonl(0xfe800000);
2212                 scope = IFA_LINK;
2213         } else {
2214                 scope = IPV6_ADDR_COMPATv4;
2215         }
2216
2217         if (addr.s6_addr32[3]) {
2218                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2219                 if (!IS_ERR(ifp)) {
2220                         spin_lock_bh(&ifp->lock);
2221                         ifp->flags &= ~IFA_F_TENTATIVE;
2222                         spin_unlock_bh(&ifp->lock);
2223                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2224                         in6_ifa_put(ifp);
2225                 }
2226                 return;
2227         }
2228
2229         for_each_netdev(net, dev) {
2230                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2231                 if (in_dev && (dev->flags & IFF_UP)) {
2232                         struct in_ifaddr * ifa;
2233
2234                         int flag = scope;
2235
2236                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2237                                 int plen;
2238
2239                                 addr.s6_addr32[3] = ifa->ifa_local;
2240
2241                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2242                                         continue;
2243                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2244                                         if (idev->dev->flags&IFF_POINTOPOINT)
2245                                                 continue;
2246                                         flag |= IFA_HOST;
2247                                 }
2248                                 if (idev->dev->flags&IFF_POINTOPOINT)
2249                                         plen = 64;
2250                                 else
2251                                         plen = 96;
2252
2253                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2254                                                     IFA_F_PERMANENT);
2255                                 if (!IS_ERR(ifp)) {
2256                                         spin_lock_bh(&ifp->lock);
2257                                         ifp->flags &= ~IFA_F_TENTATIVE;
2258                                         spin_unlock_bh(&ifp->lock);
2259                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2260                                         in6_ifa_put(ifp);
2261                                 }
2262                         }
2263                 }
2264         }
2265 }
2266 #endif
2267
2268 static void init_loopback(struct net_device *dev)
2269 {
2270         struct inet6_dev  *idev;
2271         struct inet6_ifaddr * ifp;
2272
2273         /* ::1 */
2274
2275         ASSERT_RTNL();
2276
2277         if ((idev = ipv6_find_idev(dev)) == NULL) {
2278                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2279                 return;
2280         }
2281
2282         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2283         if (!IS_ERR(ifp)) {
2284                 spin_lock_bh(&ifp->lock);
2285                 ifp->flags &= ~IFA_F_TENTATIVE;
2286                 spin_unlock_bh(&ifp->lock);
2287                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2288                 in6_ifa_put(ifp);
2289         }
2290 }
2291
2292 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2293 {
2294         struct inet6_ifaddr * ifp;
2295         u32 addr_flags = IFA_F_PERMANENT;
2296
2297 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2298         if (idev->cnf.optimistic_dad &&
2299             !ipv6_devconf.forwarding)
2300                 addr_flags |= IFA_F_OPTIMISTIC;
2301 #endif
2302
2303
2304         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2305         if (!IS_ERR(ifp)) {
2306                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2307                 addrconf_dad_start(ifp, 0);
2308                 in6_ifa_put(ifp);
2309         }
2310 }
2311
2312 static void addrconf_dev_config(struct net_device *dev)
2313 {
2314         struct in6_addr addr;
2315         struct inet6_dev    * idev;
2316
2317         ASSERT_RTNL();
2318
2319         if ((dev->type != ARPHRD_ETHER) &&
2320             (dev->type != ARPHRD_FDDI) &&
2321             (dev->type != ARPHRD_IEEE802_TR) &&
2322             (dev->type != ARPHRD_ARCNET) &&
2323             (dev->type != ARPHRD_INFINIBAND)) {
2324                 /* Alas, we support only Ethernet autoconfiguration. */
2325                 return;
2326         }
2327
2328         idev = addrconf_add_dev(dev);
2329         if (idev == NULL)
2330                 return;
2331
2332         memset(&addr, 0, sizeof(struct in6_addr));
2333         addr.s6_addr32[0] = htonl(0xFE800000);
2334
2335         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2336                 addrconf_add_linklocal(idev, &addr);
2337 }
2338
2339 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2340 static void addrconf_sit_config(struct net_device *dev)
2341 {
2342         struct inet6_dev *idev;
2343
2344         ASSERT_RTNL();
2345
2346         /*
2347          * Configure the tunnel with one of our IPv4
2348          * addresses... we should configure all of
2349          * our v4 addrs in the tunnel
2350          */
2351
2352         if ((idev = ipv6_find_idev(dev)) == NULL) {
2353                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2354                 return;
2355         }
2356
2357         if (dev->priv_flags & IFF_ISATAP) {
2358                 struct in6_addr addr;
2359
2360                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2361                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2362                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2363                         addrconf_add_linklocal(idev, &addr);
2364                 return;
2365         }
2366
2367         sit_add_v4_addrs(idev);
2368
2369         if (dev->flags&IFF_POINTOPOINT) {
2370                 addrconf_add_mroute(dev);
2371                 addrconf_add_lroute(dev);
2372         } else
2373                 sit_route_add(dev);
2374 }
2375 #endif
2376
2377 static inline int
2378 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2379 {
2380         struct in6_addr lladdr;
2381
2382         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2383                 addrconf_add_linklocal(idev, &lladdr);
2384                 return 0;
2385         }
2386         return -1;
2387 }
2388
2389 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2390 {
2391         struct net_device *link_dev;
2392         struct net *net = dev_net(idev->dev);
2393
2394         /* first try to inherit the link-local address from the link device */
2395         if (idev->dev->iflink &&
2396             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2397                 if (!ipv6_inherit_linklocal(idev, link_dev))
2398                         return;
2399         }
2400         /* then try to inherit it from any device */
2401         for_each_netdev(net, link_dev) {
2402                 if (!ipv6_inherit_linklocal(idev, link_dev))
2403                         return;
2404         }
2405         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2406 }
2407
2408 /*
2409  * Autoconfigure tunnel with a link-local address so routing protocols,
2410  * DHCPv6, MLD etc. can be run over the virtual link
2411  */
2412
2413 static void addrconf_ip6_tnl_config(struct net_device *dev)
2414 {
2415         struct inet6_dev *idev;
2416
2417         ASSERT_RTNL();
2418
2419         if ((idev = addrconf_add_dev(dev)) == NULL) {
2420                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2421                 return;
2422         }
2423         ip6_tnl_add_linklocal(idev);
2424 }
2425
2426 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2427                            void * data)
2428 {
2429         struct net_device *dev = (struct net_device *) data;
2430         struct inet6_dev *idev = __in6_dev_get(dev);
2431         int run_pending = 0;
2432         int err;
2433
2434         switch(event) {
2435         case NETDEV_REGISTER:
2436                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2437                         idev = ipv6_add_dev(dev);
2438                         if (!idev)
2439                                 return notifier_from_errno(-ENOMEM);
2440                 }
2441                 break;
2442         case NETDEV_UP:
2443         case NETDEV_CHANGE:
2444                 if (dev->flags & IFF_SLAVE)
2445                         break;
2446
2447                 if (event == NETDEV_UP) {
2448                         if (!addrconf_qdisc_ok(dev)) {
2449                                 /* device is not ready yet. */
2450                                 printk(KERN_INFO
2451                                         "ADDRCONF(NETDEV_UP): %s: "
2452                                         "link is not ready\n",
2453                                         dev->name);
2454                                 break;
2455                         }
2456
2457                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2458                                 idev = ipv6_add_dev(dev);
2459
2460                         if (idev)
2461                                 idev->if_flags |= IF_READY;
2462                 } else {
2463                         if (!addrconf_qdisc_ok(dev)) {
2464                                 /* device is still not ready. */
2465                                 break;
2466                         }
2467
2468                         if (idev) {
2469                                 if (idev->if_flags & IF_READY) {
2470                                         /* device is already configured. */
2471                                         break;
2472                                 }
2473                                 idev->if_flags |= IF_READY;
2474                         }
2475
2476                         printk(KERN_INFO
2477                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2478                                         "link becomes ready\n",
2479                                         dev->name);
2480
2481                         run_pending = 1;
2482                 }
2483
2484                 switch(dev->type) {
2485 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2486                 case ARPHRD_SIT:
2487                         addrconf_sit_config(dev);
2488                         break;
2489 #endif
2490                 case ARPHRD_TUNNEL6:
2491                         addrconf_ip6_tnl_config(dev);
2492                         break;
2493                 case ARPHRD_LOOPBACK:
2494                         init_loopback(dev);
2495                         break;
2496
2497                 default:
2498                         addrconf_dev_config(dev);
2499                         break;
2500                 }
2501                 if (idev) {
2502                         if (run_pending)
2503                                 addrconf_dad_run(idev);
2504
2505                         /* If the MTU changed during the interface down, when the
2506                            interface up, the changed MTU must be reflected in the
2507                            idev as well as routers.
2508                          */
2509                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2510                                 rt6_mtu_change(dev, dev->mtu);
2511                                 idev->cnf.mtu6 = dev->mtu;
2512                         }
2513                         idev->tstamp = jiffies;
2514                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2515                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2516                            stop IPv6 on this interface.
2517                          */
2518                         if (dev->mtu < IPV6_MIN_MTU)
2519                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2520                 }
2521                 break;
2522
2523         case NETDEV_CHANGEMTU:
2524                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2525                         rt6_mtu_change(dev, dev->mtu);
2526                         idev->cnf.mtu6 = dev->mtu;
2527                         break;
2528                 }
2529
2530                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2531                         idev = ipv6_add_dev(dev);
2532                         if (idev)
2533                                 break;
2534                 }
2535
2536                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2537
2538         case NETDEV_DOWN:
2539         case NETDEV_UNREGISTER:
2540                 /*
2541                  *      Remove all addresses from this interface.
2542                  */
2543                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2544                 break;
2545
2546         case NETDEV_CHANGENAME:
2547                 if (idev) {
2548                         snmp6_unregister_dev(idev);
2549                         addrconf_sysctl_unregister(idev);
2550                         addrconf_sysctl_register(idev);
2551                         err = snmp6_register_dev(idev);
2552                         if (err)
2553                                 return notifier_from_errno(err);
2554                 }
2555                 break;
2556         }
2557
2558         return NOTIFY_OK;
2559 }
2560
2561 /*
2562  *      addrconf module should be notified of a device going up
2563  */
2564 static struct notifier_block ipv6_dev_notf = {
2565         .notifier_call = addrconf_notify,
2566         .priority = 0
2567 };
2568
2569 static int addrconf_ifdown(struct net_device *dev, int how)
2570 {
2571         struct inet6_dev *idev;
2572         struct inet6_ifaddr *ifa, **bifa;
2573         struct net *net = dev_net(dev);
2574         int i;
2575
2576         ASSERT_RTNL();
2577
2578         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2579                 how = 0;
2580
2581         rt6_ifdown(net, dev);
2582         neigh_ifdown(&nd_tbl, dev);
2583
2584         idev = __in6_dev_get(dev);
2585         if (idev == NULL)
2586                 return -ENODEV;
2587
2588         /* Step 1: remove reference to ipv6 device from parent device.
2589                    Do not dev_put!
2590          */
2591         if (how) {
2592                 idev->dead = 1;
2593
2594                 /* protected by rtnl_lock */
2595                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2596
2597                 /* Step 1.5: remove snmp6 entry */
2598                 snmp6_unregister_dev(idev);
2599
2600         }
2601
2602         /* Step 2: clear hash table */
2603         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2604                 bifa = &inet6_addr_lst[i];
2605
2606                 write_lock_bh(&addrconf_hash_lock);
2607                 while ((ifa = *bifa) != NULL) {
2608                         if (ifa->idev == idev) {
2609                                 *bifa = ifa->lst_next;
2610                                 ifa->lst_next = NULL;
2611                                 addrconf_del_timer(ifa);
2612                                 in6_ifa_put(ifa);
2613                                 continue;
2614                         }
2615                         bifa = &ifa->lst_next;
2616                 }
2617                 write_unlock_bh(&addrconf_hash_lock);
2618         }
2619
2620         write_lock_bh(&idev->lock);
2621
2622         /* Step 3: clear flags for stateless addrconf */
2623         if (!how)
2624                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2625
2626         /* Step 4: clear address list */
2627 #ifdef CONFIG_IPV6_PRIVACY
2628         if (how && del_timer(&idev->regen_timer))
2629                 in6_dev_put(idev);
2630
2631         /* clear tempaddr list */
2632         while ((ifa = idev->tempaddr_list) != NULL) {
2633                 idev->tempaddr_list = ifa->tmp_next;
2634                 ifa->tmp_next = NULL;
2635                 ifa->dead = 1;
2636                 write_unlock_bh(&idev->lock);
2637                 spin_lock_bh(&ifa->lock);
2638
2639                 if (ifa->ifpub) {
2640                         in6_ifa_put(ifa->ifpub);
2641                         ifa->ifpub = NULL;
2642                 }
2643                 spin_unlock_bh(&ifa->lock);
2644                 in6_ifa_put(ifa);
2645                 write_lock_bh(&idev->lock);
2646         }
2647 #endif
2648         while ((ifa = idev->addr_list) != NULL) {
2649                 idev->addr_list = ifa->if_next;
2650                 ifa->if_next = NULL;
2651                 ifa->dead = 1;
2652                 addrconf_del_timer(ifa);
2653                 write_unlock_bh(&idev->lock);
2654
2655                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2656                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2657                 in6_ifa_put(ifa);
2658
2659                 write_lock_bh(&idev->lock);
2660         }
2661         write_unlock_bh(&idev->lock);
2662
2663         /* Step 5: Discard multicast list */
2664
2665         if (how)
2666                 ipv6_mc_destroy_dev(idev);
2667         else
2668                 ipv6_mc_down(idev);
2669
2670         idev->tstamp = jiffies;
2671
2672         /* Shot the device (if unregistered) */
2673
2674         if (how) {
2675                 addrconf_sysctl_unregister(idev);
2676                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2677                 neigh_ifdown(&nd_tbl, dev);
2678                 in6_dev_put(idev);
2679         }
2680         return 0;
2681 }
2682
2683 static void addrconf_rs_timer(unsigned long data)
2684 {
2685         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2686
2687         if (ifp->idev->cnf.forwarding)
2688                 goto out;
2689
2690         if (ifp->idev->if_flags & IF_RA_RCVD) {
2691                 /*
2692                  *      Announcement received after solicitation
2693                  *      was sent
2694                  */
2695                 goto out;
2696         }
2697
2698         spin_lock(&ifp->lock);
2699         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2700                 /* The wait after the last probe can be shorter */
2701                 addrconf_mod_timer(ifp, AC_RS,
2702                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2703                                    ifp->idev->cnf.rtr_solicit_delay :
2704                                    ifp->idev->cnf.rtr_solicit_interval);
2705                 spin_unlock(&ifp->lock);
2706
2707                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2708         } else {
2709                 spin_unlock(&ifp->lock);
2710                 /*
2711                  * Note: we do not support deprecated "all on-link"
2712                  * assumption any longer.
2713                  */
2714                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2715                        ifp->idev->dev->name);
2716         }
2717
2718 out:
2719         in6_ifa_put(ifp);
2720 }
2721
2722 /*
2723  *      Duplicate Address Detection
2724  */
2725 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2726 {
2727         unsigned long rand_num;
2728         struct inet6_dev *idev = ifp->idev;
2729
2730         if (ifp->flags & IFA_F_OPTIMISTIC)
2731                 rand_num = 0;
2732         else
2733                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2734
2735         ifp->probes = idev->cnf.dad_transmits;
2736         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2737 }
2738
2739 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2740 {
2741         struct inet6_dev *idev = ifp->idev;
2742         struct net_device *dev = idev->dev;
2743
2744         addrconf_join_solict(dev, &ifp->addr);
2745
2746         net_srandom(ifp->addr.s6_addr32[3]);
2747
2748         read_lock_bh(&idev->lock);
2749         if (ifp->dead)
2750                 goto out;
2751         spin_lock_bh(&ifp->lock);
2752
2753         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2754             !(ifp->flags&IFA_F_TENTATIVE) ||
2755             ifp->flags & IFA_F_NODAD) {
2756                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2757                 spin_unlock_bh(&ifp->lock);
2758                 read_unlock_bh(&idev->lock);
2759
2760                 addrconf_dad_completed(ifp);
2761                 return;
2762         }
2763
2764         if (!(idev->if_flags & IF_READY)) {
2765                 spin_unlock_bh(&ifp->lock);
2766                 read_unlock_bh(&idev->lock);
2767                 /*
2768                  * If the defice is not ready:
2769                  * - keep it tentative if it is a permanent address.
2770                  * - otherwise, kill it.
2771                  */
2772                 in6_ifa_hold(ifp);
2773                 addrconf_dad_stop(ifp);
2774                 return;
2775         }
2776
2777         /*
2778          * Optimistic nodes can start receiving
2779          * Frames right away
2780          */
2781         if(ifp->flags & IFA_F_OPTIMISTIC)
2782                 ip6_ins_rt(ifp->rt);
2783
2784         addrconf_dad_kick(ifp);
2785         spin_unlock_bh(&ifp->lock);
2786 out:
2787         read_unlock_bh(&idev->lock);
2788 }
2789
2790 static void addrconf_dad_timer(unsigned long data)
2791 {
2792         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2793         struct inet6_dev *idev = ifp->idev;
2794         struct in6_addr mcaddr;
2795
2796         read_lock_bh(&idev->lock);
2797         if (idev->dead) {
2798                 read_unlock_bh(&idev->lock);
2799                 goto out;
2800         }
2801         spin_lock_bh(&ifp->lock);
2802         if (ifp->probes == 0) {
2803                 /*
2804                  * DAD was successful
2805                  */
2806
2807                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2808                 spin_unlock_bh(&ifp->lock);
2809                 read_unlock_bh(&idev->lock);
2810
2811                 addrconf_dad_completed(ifp);
2812
2813                 goto out;
2814         }
2815
2816         ifp->probes--;
2817         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2818         spin_unlock_bh(&ifp->lock);
2819         read_unlock_bh(&idev->lock);
2820
2821         /* send a neighbour solicitation for our addr */
2822         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2823         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2824 out:
2825         in6_ifa_put(ifp);
2826 }
2827
2828 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2829 {
2830         struct net_device *     dev = ifp->idev->dev;
2831
2832         /*
2833          *      Configure the address for reception. Now it is valid.
2834          */
2835
2836         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2837
2838         /* If added prefix is link local and forwarding is off,
2839            start sending router solicitations.
2840          */
2841
2842         if (ifp->idev->cnf.forwarding == 0 &&
2843             ifp->idev->cnf.rtr_solicits > 0 &&
2844             (dev->flags&IFF_LOOPBACK) == 0 &&
2845             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2846                 /*
2847                  *      If a host as already performed a random delay
2848                  *      [...] as part of DAD [...] there is no need
2849                  *      to delay again before sending the first RS
2850                  */
2851                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2852
2853                 spin_lock_bh(&ifp->lock);
2854                 ifp->probes = 1;
2855                 ifp->idev->if_flags |= IF_RS_SENT;
2856                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2857                 spin_unlock_bh(&ifp->lock);
2858         }
2859 }
2860
2861 static void addrconf_dad_run(struct inet6_dev *idev) {
2862         struct inet6_ifaddr *ifp;
2863
2864         read_lock_bh(&idev->lock);
2865         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2866                 spin_lock_bh(&ifp->lock);
2867                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2868                         spin_unlock_bh(&ifp->lock);
2869                         continue;
2870                 }
2871                 spin_unlock_bh(&ifp->lock);
2872                 addrconf_dad_kick(ifp);
2873         }
2874         read_unlock_bh(&idev->lock);
2875 }
2876
2877 #ifdef CONFIG_PROC_FS
2878 struct if6_iter_state {
2879         struct seq_net_private p;
2880         int bucket;
2881 };
2882
2883 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2884 {
2885         struct inet6_ifaddr *ifa = NULL;
2886         struct if6_iter_state *state = seq->private;
2887         struct net *net = seq_file_net(seq);
2888
2889         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2890                 ifa = inet6_addr_lst[state->bucket];
2891
2892                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2893                         ifa = ifa->lst_next;
2894                 if (ifa)
2895                         break;
2896         }
2897         return ifa;
2898 }
2899
2900 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2901 {
2902         struct if6_iter_state *state = seq->private;
2903         struct net *net = seq_file_net(seq);
2904
2905         ifa = ifa->lst_next;
2906 try_again:
2907         if (ifa) {
2908                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2909                         ifa = ifa->lst_next;
2910                         goto try_again;
2911                 }
2912         }
2913
2914         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2915                 ifa = inet6_addr_lst[state->bucket];
2916                 goto try_again;
2917         }
2918
2919         return ifa;
2920 }
2921
2922 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2923 {
2924         struct inet6_ifaddr *ifa = if6_get_first(seq);
2925
2926         if (ifa)
2927                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2928                         --pos;
2929         return pos ? NULL : ifa;
2930 }
2931
2932 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2933         __acquires(addrconf_hash_lock)
2934 {
2935         read_lock_bh(&addrconf_hash_lock);
2936         return if6_get_idx(seq, *pos);
2937 }
2938
2939 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2940 {
2941         struct inet6_ifaddr *ifa;
2942
2943         ifa = if6_get_next(seq, v);
2944         ++*pos;
2945         return ifa;
2946 }
2947
2948 static void if6_seq_stop(struct seq_file *seq, void *v)
2949         __releases(addrconf_hash_lock)
2950 {
2951         read_unlock_bh(&addrconf_hash_lock);
2952 }
2953
2954 static int if6_seq_show(struct seq_file *seq, void *v)
2955 {
2956         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2957         seq_printf(seq,
2958                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2959                    NIP6(ifp->addr),
2960                    ifp->idev->dev->ifindex,
2961                    ifp->prefix_len,
2962                    ifp->scope,
2963                    ifp->flags,
2964                    ifp->idev->dev->name);
2965         return 0;
2966 }
2967
2968 static const struct seq_operations if6_seq_ops = {
2969         .start  = if6_seq_start,
2970         .next   = if6_seq_next,
2971         .show   = if6_seq_show,
2972         .stop   = if6_seq_stop,
2973 };
2974
2975 static int if6_seq_open(struct inode *inode, struct file *file)
2976 {
2977         return seq_open_net(inode, file, &if6_seq_ops,
2978                             sizeof(struct if6_iter_state));
2979 }
2980
2981 static const struct file_operations if6_fops = {
2982         .owner          = THIS_MODULE,
2983         .open           = if6_seq_open,
2984         .read           = seq_read,
2985         .llseek         = seq_lseek,
2986         .release        = seq_release_net,
2987 };
2988
2989 static int if6_proc_net_init(struct net *net)
2990 {
2991         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2992                 return -ENOMEM;
2993         return 0;
2994 }
2995
2996 static void if6_proc_net_exit(struct net *net)
2997 {
2998        proc_net_remove(net, "if_inet6");
2999 }
3000
3001 static struct pernet_operations if6_proc_net_ops = {
3002        .init = if6_proc_net_init,
3003        .exit = if6_proc_net_exit,
3004 };
3005
3006 int __init if6_proc_init(void)
3007 {
3008         return register_pernet_subsys(&if6_proc_net_ops);
3009 }
3010
3011 void if6_proc_exit(void)
3012 {
3013         unregister_pernet_subsys(&if6_proc_net_ops);
3014 }
3015 #endif  /* CONFIG_PROC_FS */
3016
3017 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3018 /* Check if address is a home address configured on any interface. */
3019 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3020 {
3021         int ret = 0;
3022         struct inet6_ifaddr * ifp;
3023         u8 hash = ipv6_addr_hash(addr);
3024         read_lock_bh(&addrconf_hash_lock);
3025         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3026                 if (!net_eq(dev_net(ifp->idev->dev), net))
3027                         continue;
3028                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3029                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3030                         ret = 1;
3031                         break;
3032                 }
3033         }
3034         read_unlock_bh(&addrconf_hash_lock);
3035         return ret;
3036 }
3037 #endif
3038
3039 /*
3040  *      Periodic address status verification
3041  */
3042
3043 static void addrconf_verify(unsigned long foo)
3044 {
3045         struct inet6_ifaddr *ifp;
3046         unsigned long now, next;
3047         int i;
3048
3049         spin_lock_bh(&addrconf_verify_lock);
3050         now = jiffies;
3051         next = now + ADDR_CHECK_FREQUENCY;
3052
3053         del_timer(&addr_chk_timer);
3054
3055         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3056
3057 restart:
3058                 read_lock(&addrconf_hash_lock);
3059                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3060                         unsigned long age;
3061 #ifdef CONFIG_IPV6_PRIVACY
3062                         unsigned long regen_advance;
3063 #endif
3064
3065                         if (ifp->flags & IFA_F_PERMANENT)
3066                                 continue;
3067
3068                         spin_lock(&ifp->lock);
3069                         age = (now - ifp->tstamp) / HZ;
3070
3071 #ifdef CONFIG_IPV6_PRIVACY
3072                         regen_advance = ifp->idev->cnf.regen_max_retry *
3073                                         ifp->idev->cnf.dad_transmits *
3074                                         ifp->idev->nd_parms->retrans_time / HZ;
3075 #endif
3076
3077                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3078                             age >= ifp->valid_lft) {
3079                                 spin_unlock(&ifp->lock);
3080                                 in6_ifa_hold(ifp);
3081                                 read_unlock(&addrconf_hash_lock);
3082                                 ipv6_del_addr(ifp);
3083                                 goto restart;
3084                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3085                                 spin_unlock(&ifp->lock);
3086                                 continue;
3087                         } else if (age >= ifp->prefered_lft) {
3088                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3089                                 int deprecate = 0;
3090
3091                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3092                                         deprecate = 1;
3093                                         ifp->flags |= IFA_F_DEPRECATED;
3094                                 }
3095
3096                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3097                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3098
3099                                 spin_unlock(&ifp->lock);
3100
3101                                 if (deprecate) {
3102                                         in6_ifa_hold(ifp);
3103                                         read_unlock(&addrconf_hash_lock);
3104
3105                                         ipv6_ifa_notify(0, ifp);
3106                                         in6_ifa_put(ifp);
3107                                         goto restart;
3108                                 }
3109 #ifdef CONFIG_IPV6_PRIVACY
3110                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3111                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3112                                 if (age >= ifp->prefered_lft - regen_advance) {
3113                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3114                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3115                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3116                                         if (!ifp->regen_count && ifpub) {
3117                                                 ifp->regen_count++;
3118                                                 in6_ifa_hold(ifp);
3119                                                 in6_ifa_hold(ifpub);
3120                                                 spin_unlock(&ifp->lock);
3121                                                 read_unlock(&addrconf_hash_lock);
3122                                                 spin_lock(&ifpub->lock);
3123                                                 ifpub->regen_count = 0;
3124                                                 spin_unlock(&ifpub->lock);
3125                                                 ipv6_create_tempaddr(ifpub, ifp);
3126                                                 in6_ifa_put(ifpub);
3127                                                 in6_ifa_put(ifp);
3128                                                 goto restart;
3129                                         }
3130                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3131                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3132                                 spin_unlock(&ifp->lock);
3133 #endif
3134                         } else {
3135                                 /* ifp->prefered_lft <= ifp->valid_lft */
3136                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3137                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3138                                 spin_unlock(&ifp->lock);
3139                         }
3140                 }
3141                 read_unlock(&addrconf_hash_lock);
3142         }
3143
3144         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3145         add_timer(&addr_chk_timer);
3146         spin_unlock_bh(&addrconf_verify_lock);
3147 }
3148
3149 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3150 {
3151         struct in6_addr *pfx = NULL;
3152
3153         if (addr)
3154                 pfx = nla_data(addr);
3155
3156         if (local) {
3157                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3158                         pfx = NULL;
3159                 else
3160                         pfx = nla_data(local);
3161         }
3162
3163         return pfx;
3164 }
3165
3166 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3167         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3168         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3169         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3170 };
3171
3172 static int
3173 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3174 {
3175         struct net *net = sock_net(skb->sk);
3176         struct ifaddrmsg *ifm;
3177         struct nlattr *tb[IFA_MAX+1];
3178         struct in6_addr *pfx;
3179         int err;
3180
3181         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3182         if (err < 0)
3183                 return err;
3184
3185         ifm = nlmsg_data(nlh);
3186         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3187         if (pfx == NULL)
3188                 return -EINVAL;
3189
3190         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3191 }
3192
3193 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3194                              u32 prefered_lft, u32 valid_lft)
3195 {
3196         u32 flags;
3197         clock_t expires;
3198         unsigned long timeout;
3199
3200         if (!valid_lft || (prefered_lft > valid_lft))
3201                 return -EINVAL;
3202
3203         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3204         if (addrconf_finite_timeout(timeout)) {
3205                 expires = jiffies_to_clock_t(timeout * HZ);
3206                 valid_lft = timeout;
3207                 flags = RTF_EXPIRES;
3208         } else {
3209                 expires = 0;
3210                 flags = 0;
3211                 ifa_flags |= IFA_F_PERMANENT;
3212         }
3213
3214         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3215         if (addrconf_finite_timeout(timeout)) {
3216                 if (timeout == 0)
3217                         ifa_flags |= IFA_F_DEPRECATED;
3218                 prefered_lft = timeout;
3219         }
3220
3221         spin_lock_bh(&ifp->lock);
3222         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3223         ifp->tstamp = jiffies;
3224         ifp->valid_lft = valid_lft;
3225         ifp->prefered_lft = prefered_lft;
3226
3227         spin_unlock_bh(&ifp->lock);
3228         if (!(ifp->flags&IFA_F_TENTATIVE))
3229                 ipv6_ifa_notify(0, ifp);
3230
3231         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3232                               expires, flags);
3233         addrconf_verify(0);
3234
3235         return 0;
3236 }
3237
3238 static int
3239 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3240 {
3241         struct net *net = sock_net(skb->sk);
3242         struct ifaddrmsg *ifm;
3243         struct nlattr *tb[IFA_MAX+1];
3244         struct in6_addr *pfx;
3245         struct inet6_ifaddr *ifa;
3246         struct net_device *dev;
3247         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3248         u8 ifa_flags;
3249         int err;
3250
3251         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3252         if (err < 0)
3253                 return err;
3254
3255         ifm = nlmsg_data(nlh);
3256         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3257         if (pfx == NULL)
3258                 return -EINVAL;
3259
3260         if (tb[IFA_CACHEINFO]) {
3261                 struct ifa_cacheinfo *ci;
3262
3263                 ci = nla_data(tb[IFA_CACHEINFO]);
3264                 valid_lft = ci->ifa_valid;
3265                 preferred_lft = ci->ifa_prefered;
3266         } else {
3267                 preferred_lft = INFINITY_LIFE_TIME;
3268                 valid_lft = INFINITY_LIFE_TIME;
3269         }
3270
3271         dev =  __dev_get_by_index(net, ifm->ifa_index);
3272         if (dev == NULL)
3273                 return -ENODEV;
3274
3275         /* We ignore other flags so far. */
3276         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3277
3278         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3279         if (ifa == NULL) {
3280                 /*
3281                  * It would be best to check for !NLM_F_CREATE here but
3282                  * userspace alreay relies on not having to provide this.
3283                  */
3284                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3285                                       ifm->ifa_prefixlen, ifa_flags,
3286                                       preferred_lft, valid_lft);
3287         }
3288
3289         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3290             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3291                 err = -EEXIST;
3292         else
3293                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3294
3295         in6_ifa_put(ifa);
3296
3297         return err;
3298 }
3299
3300 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3301                           u8 scope, int ifindex)
3302 {
3303         struct ifaddrmsg *ifm;
3304
3305         ifm = nlmsg_data(nlh);
3306         ifm->ifa_family = AF_INET6;
3307         ifm->ifa_prefixlen = prefixlen;
3308         ifm->ifa_flags = flags;
3309         ifm->ifa_scope = scope;
3310         ifm->ifa_index = ifindex;
3311 }
3312
3313 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3314                          unsigned long tstamp, u32 preferred, u32 valid)
3315 {
3316         struct ifa_cacheinfo ci;
3317
3318         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3319                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3320         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3321                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3322         ci.ifa_prefered = preferred;
3323         ci.ifa_valid = valid;
3324
3325         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3326 }
3327
3328 static inline int rt_scope(int ifa_scope)
3329 {
3330         if (ifa_scope & IFA_HOST)
3331                 return RT_SCOPE_HOST;
3332         else if (ifa_scope & IFA_LINK)
3333                 return RT_SCOPE_LINK;
3334         else if (ifa_scope & IFA_SITE)
3335                 return RT_SCOPE_SITE;
3336         else
3337                 return RT_SCOPE_UNIVERSE;
3338 }
3339
3340 static inline int inet6_ifaddr_msgsize(void)
3341 {
3342         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3343                + nla_total_size(16) /* IFA_ADDRESS */
3344                + nla_total_size(sizeof(struct ifa_cacheinfo));
3345 }
3346
3347 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3348                              u32 pid, u32 seq, int event, unsigned int flags)
3349 {
3350         struct nlmsghdr  *nlh;
3351         u32 preferred, valid;
3352
3353         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3354         if (nlh == NULL)
3355                 return -EMSGSIZE;
3356
3357         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3358                       ifa->idev->dev->ifindex);
3359
3360         if (!(ifa->flags&IFA_F_PERMANENT)) {
3361                 preferred = ifa->prefered_lft;
3362                 valid = ifa->valid_lft;
3363                 if (preferred != INFINITY_LIFE_TIME) {
3364                         long tval = (jiffies - ifa->tstamp)/HZ;
3365                         preferred -= tval;
3366                         if (valid != INFINITY_LIFE_TIME)
3367                                 valid -= tval;
3368                 }
3369         } else {
3370                 preferred = INFINITY_LIFE_TIME;
3371                 valid = INFINITY_LIFE_TIME;
3372         }
3373
3374         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3375             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3376                 nlmsg_cancel(skb, nlh);
3377                 return -EMSGSIZE;
3378         }
3379
3380         return nlmsg_end(skb, nlh);
3381 }
3382
3383 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3384                                 u32 pid, u32 seq, int event, u16 flags)
3385 {
3386         struct nlmsghdr  *nlh;
3387         u8 scope = RT_SCOPE_UNIVERSE;
3388         int ifindex = ifmca->idev->dev->ifindex;
3389
3390         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3391                 scope = RT_SCOPE_SITE;
3392
3393         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3394         if (nlh == NULL)
3395                 return -EMSGSIZE;
3396
3397         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3398         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3399             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3400                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3401                 nlmsg_cancel(skb, nlh);
3402                 return -EMSGSIZE;
3403         }
3404
3405         return nlmsg_end(skb, nlh);
3406 }
3407
3408 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3409                                 u32 pid, u32 seq, int event, unsigned int flags)
3410 {
3411         struct nlmsghdr  *nlh;
3412         u8 scope = RT_SCOPE_UNIVERSE;
3413         int ifindex = ifaca->aca_idev->dev->ifindex;
3414
3415         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3416                 scope = RT_SCOPE_SITE;
3417
3418         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3419         if (nlh == NULL)
3420                 return -EMSGSIZE;
3421
3422         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3423         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3424             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3425                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3426                 nlmsg_cancel(skb, nlh);
3427                 return -EMSGSIZE;
3428         }
3429
3430         return nlmsg_end(skb, nlh);
3431 }
3432
3433 enum addr_type_t
3434 {
3435         UNICAST_ADDR,
3436         MULTICAST_ADDR,
3437         ANYCAST_ADDR,
3438 };
3439
3440 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3441                            enum addr_type_t type)
3442 {
3443         int idx, ip_idx;
3444         int s_idx, s_ip_idx;
3445         int err = 1;
3446         struct net_device *dev;
3447         struct inet6_dev *idev = NULL;
3448         struct inet6_ifaddr *ifa;
3449         struct ifmcaddr6 *ifmca;
3450         struct ifacaddr6 *ifaca;
3451         struct net *net = sock_net(skb->sk);
3452
3453         s_idx = cb->args[0];
3454         s_ip_idx = ip_idx = cb->args[1];
3455
3456         idx = 0;
3457         for_each_netdev(net, dev) {
3458                 if (idx < s_idx)
3459                         goto cont;
3460                 if (idx > s_idx)
3461                         s_ip_idx = 0;
3462                 ip_idx = 0;
3463                 if ((idev = in6_dev_get(dev)) == NULL)
3464                         goto cont;
3465                 read_lock_bh(&idev->lock);
3466                 switch (type) {
3467                 case UNICAST_ADDR:
3468                         /* unicast address incl. temp addr */
3469                         for (ifa = idev->addr_list; ifa;
3470                              ifa = ifa->if_next, ip_idx++) {
3471                                 if (ip_idx < s_ip_idx)
3472                                         continue;
3473                                 err = inet6_fill_ifaddr(skb, ifa,
3474                                                         NETLINK_CB(cb->skb).pid,
3475                                                         cb->nlh->nlmsg_seq,
3476                                                         RTM_NEWADDR,
3477                                                         NLM_F_MULTI);
3478                         }
3479                         break;
3480                 case MULTICAST_ADDR:
3481                         /* multicast address */
3482                         for (ifmca = idev->mc_list; ifmca;
3483                              ifmca = ifmca->next, ip_idx++) {
3484                                 if (ip_idx < s_ip_idx)
3485                                         continue;
3486                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3487                                                           NETLINK_CB(cb->skb).pid,
3488                                                           cb->nlh->nlmsg_seq,
3489                                                           RTM_GETMULTICAST,
3490                                                           NLM_F_MULTI);
3491                         }
3492                         break;
3493                 case ANYCAST_ADDR:
3494                         /* anycast address */
3495                         for (ifaca = idev->ac_list; ifaca;
3496                              ifaca = ifaca->aca_next, ip_idx++) {
3497                                 if (ip_idx < s_ip_idx)
3498                                         continue;
3499                                 err = inet6_fill_ifacaddr(skb, ifaca,
3500                                                           NETLINK_CB(cb->skb).pid,
3501                                                           cb->nlh->nlmsg_seq,
3502                                                           RTM_GETANYCAST,
3503                                                           NLM_F_MULTI);
3504                         }
3505                         break;
3506                 default:
3507                         break;
3508                 }
3509                 read_unlock_bh(&idev->lock);
3510                 in6_dev_put(idev);
3511
3512                 if (err <= 0)
3513                         break;
3514 cont:
3515                 idx++;
3516         }
3517         cb->args[0] = idx;
3518         cb->args[1] = ip_idx;
3519         return skb->len;
3520 }
3521
3522 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3523 {
3524         enum addr_type_t type = UNICAST_ADDR;
3525
3526         return inet6_dump_addr(skb, cb, type);
3527 }
3528
3529 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3530 {
3531         enum addr_type_t type = MULTICAST_ADDR;
3532
3533         return inet6_dump_addr(skb, cb, type);
3534 }
3535
3536
3537 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3538 {
3539         enum addr_type_t type = ANYCAST_ADDR;
3540
3541         return inet6_dump_addr(skb, cb, type);
3542 }
3543
3544 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3545                              void *arg)
3546 {
3547         struct net *net = sock_net(in_skb->sk);
3548         struct ifaddrmsg *ifm;
3549         struct nlattr *tb[IFA_MAX+1];
3550         struct in6_addr *addr = NULL;
3551         struct net_device *dev = NULL;
3552         struct inet6_ifaddr *ifa;
3553         struct sk_buff *skb;
3554         int err;
3555
3556         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3557         if (err < 0)
3558                 goto errout;
3559
3560         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3561         if (addr == NULL) {
3562                 err = -EINVAL;
3563                 goto errout;
3564         }
3565
3566         ifm = nlmsg_data(nlh);
3567         if (ifm->ifa_index)
3568                 dev = __dev_get_by_index(net, ifm->ifa_index);
3569
3570         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3571                 err = -EADDRNOTAVAIL;
3572                 goto errout;
3573         }
3574
3575         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3576                 err = -ENOBUFS;
3577                 goto errout_ifa;
3578         }
3579
3580         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3581                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3582         if (err < 0) {
3583                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3584                 WARN_ON(err == -EMSGSIZE);
3585                 kfree_skb(skb);
3586                 goto errout_ifa;
3587         }
3588         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3589 errout_ifa:
3590         in6_ifa_put(ifa);
3591 errout:
3592         return err;
3593 }
3594
3595 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3596 {
3597         struct sk_buff *skb;
3598         struct net *net = dev_net(ifa->idev->dev);
3599         int err = -ENOBUFS;
3600
3601         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3602         if (skb == NULL)
3603                 goto errout;
3604
3605         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3606         if (err < 0) {
3607                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3608                 WARN_ON(err == -EMSGSIZE);
3609                 kfree_skb(skb);
3610                 goto errout;
3611         }
3612         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3613 errout:
3614         if (err < 0)
3615                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3616 }
3617
3618 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3619                                 __s32 *array, int bytes)
3620 {
3621         BUG_ON(bytes < (DEVCONF_MAX * 4));
3622
3623         memset(array, 0, bytes);
3624         array[DEVCONF_FORWARDING] = cnf->forwarding;
3625         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3626         array[DEVCONF_MTU6] = cnf->mtu6;
3627         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3628         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3629         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3630         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3631         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3632         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3633         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3634         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3635 #ifdef CONFIG_IPV6_PRIVACY
3636         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3637         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3638         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3639         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3640         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3641 #endif
3642         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3643         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3644         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3645 #ifdef CONFIG_IPV6_ROUTER_PREF
3646         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3647         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3648 #ifdef CONFIG_IPV6_ROUTE_INFO
3649         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3650 #endif
3651 #endif
3652         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3653         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3654 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3655         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3656 #endif
3657 #ifdef CONFIG_IPV6_MROUTE
3658         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3659 #endif
3660 }
3661
3662 static inline size_t inet6_if_nlmsg_size(void)
3663 {
3664         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3665                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3666                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3667                + nla_total_size(4) /* IFLA_MTU */
3668                + nla_total_size(4) /* IFLA_LINK */
3669                + nla_total_size( /* IFLA_PROTINFO */
3670                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3671                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3672                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3673                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3674                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3675                  );
3676 }
3677
3678 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3679                                       int bytes)
3680 {
3681         int i;
3682         int pad = bytes - sizeof(u64) * items;
3683         BUG_ON(pad < 0);
3684
3685         /* Use put_unaligned() because stats may not be aligned for u64. */
3686         put_unaligned(items, &stats[0]);
3687         for (i = 1; i < items; i++)
3688                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3689
3690         memset(&stats[items], 0, pad);
3691 }
3692
3693 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3694                              int bytes)
3695 {
3696         switch(attrtype) {
3697         case IFLA_INET6_STATS:
3698                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3699                 break;
3700         case IFLA_INET6_ICMP6STATS:
3701                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3702                 break;
3703         }
3704 }
3705
3706 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3707                              u32 pid, u32 seq, int event, unsigned int flags)
3708 {
3709         struct net_device *dev = idev->dev;
3710         struct nlattr *nla;
3711         struct ifinfomsg *hdr;
3712         struct nlmsghdr *nlh;
3713         void *protoinfo;
3714         struct ifla_cacheinfo ci;
3715
3716         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3717         if (nlh == NULL)
3718                 return -EMSGSIZE;
3719
3720         hdr = nlmsg_data(nlh);
3721         hdr->ifi_family = AF_INET6;
3722         hdr->__ifi_pad = 0;
3723         hdr->ifi_type = dev->type;
3724         hdr->ifi_index = dev->ifindex;
3725         hdr->ifi_flags = dev_get_flags(dev);
3726         hdr->ifi_change = 0;
3727
3728         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3729
3730         if (dev->addr_len)
3731                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3732
3733         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3734         if (dev->ifindex != dev->iflink)
3735                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3736
3737         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3738         if (protoinfo == NULL)
3739                 goto nla_put_failure;
3740
3741         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3742
3743         ci.max_reasm_len = IPV6_MAXPLEN;
3744         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3745                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3746         ci.reachable_time = idev->nd_parms->reachable_time;
3747         ci.retrans_time = idev->nd_parms->retrans_time;
3748         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3749
3750         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3751         if (nla == NULL)
3752                 goto nla_put_failure;
3753         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3754
3755         /* XXX - MC not implemented */
3756
3757         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3758         if (nla == NULL)
3759                 goto nla_put_failure;
3760         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3761
3762         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3763         if (nla == NULL)
3764                 goto nla_put_failure;
3765         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3766
3767         nla_nest_end(skb, protoinfo);
3768         return nlmsg_end(skb, nlh);
3769
3770 nla_put_failure:
3771         nlmsg_cancel(skb, nlh);
3772         return -EMSGSIZE;
3773 }
3774
3775 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3776 {
3777         struct net *net = sock_net(skb->sk);
3778         int idx, err;
3779         int s_idx = cb->args[0];
3780         struct net_device *dev;
3781         struct inet6_dev *idev;
3782
3783         read_lock(&dev_base_lock);
3784         idx = 0;
3785         for_each_netdev(net, dev) {
3786                 if (idx < s_idx)
3787                         goto cont;
3788                 if ((idev = in6_dev_get(dev)) == NULL)
3789                         goto cont;
3790                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3791                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3792                 in6_dev_put(idev);
3793                 if (err <= 0)
3794                         break;
3795 cont:
3796                 idx++;
3797         }
3798         read_unlock(&dev_base_lock);
3799         cb->args[0] = idx;
3800
3801         return skb->len;
3802 }
3803
3804 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3805 {
3806         struct sk_buff *skb;
3807         struct net *net = dev_net(idev->dev);
3808         int err = -ENOBUFS;
3809
3810         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3811         if (skb == NULL)
3812                 goto errout;
3813
3814         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3815         if (err < 0) {
3816                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3817                 WARN_ON(err == -EMSGSIZE);
3818                 kfree_skb(skb);
3819                 goto errout;
3820         }
3821         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3822 errout:
3823         if (err < 0)
3824                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3825 }
3826
3827 static inline size_t inet6_prefix_nlmsg_size(void)
3828 {
3829         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3830                + nla_total_size(sizeof(struct in6_addr))
3831                + nla_total_size(sizeof(struct prefix_cacheinfo));
3832 }
3833
3834 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3835                              struct prefix_info *pinfo, u32 pid, u32 seq,
3836                              int event, unsigned int flags)
3837 {
3838         struct prefixmsg *pmsg;
3839         struct nlmsghdr *nlh;
3840         struct prefix_cacheinfo ci;
3841
3842         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3843         if (nlh == NULL)
3844                 return -EMSGSIZE;
3845
3846         pmsg = nlmsg_data(nlh);
3847         pmsg->prefix_family = AF_INET6;
3848         pmsg->prefix_pad1 = 0;
3849         pmsg->prefix_pad2 = 0;
3850         pmsg->prefix_ifindex = idev->dev->ifindex;
3851         pmsg->prefix_len = pinfo->prefix_len;
3852         pmsg->prefix_type = pinfo->type;
3853         pmsg->prefix_pad3 = 0;
3854         pmsg->prefix_flags = 0;
3855         if (pinfo->onlink)
3856                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3857         if (pinfo->autoconf)
3858                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3859
3860         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3861
3862         ci.preferred_time = ntohl(pinfo->prefered);
3863         ci.valid_time = ntohl(pinfo->valid);
3864         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3865
3866         return nlmsg_end(skb, nlh);
3867
3868 nla_put_failure:
3869         nlmsg_cancel(skb, nlh);
3870         return -EMSGSIZE;
3871 }
3872
3873 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3874                          struct prefix_info *pinfo)
3875 {
3876         struct sk_buff *skb;
3877         struct net *net = dev_net(idev->dev);
3878         int err = -ENOBUFS;
3879
3880         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3881         if (skb == NULL)
3882                 goto errout;
3883
3884         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3885         if (err < 0) {
3886                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3887                 WARN_ON(err == -EMSGSIZE);
3888                 kfree_skb(skb);
3889                 goto errout;
3890         }
3891         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3892 errout:
3893         if (err < 0)
3894                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3895 }
3896
3897 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3898 {
3899         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3900
3901         switch (event) {
3902         case RTM_NEWADDR:
3903                 /*
3904                  * If the address was optimistic
3905                  * we inserted the route at the start of
3906                  * our DAD process, so we don't need
3907                  * to do it again
3908                  */
3909                 if (!(ifp->rt->rt6i_node))
3910                         ip6_ins_rt(ifp->rt);
3911                 if (ifp->idev->cnf.forwarding)
3912                         addrconf_join_anycast(ifp);
3913                 break;
3914         case RTM_DELADDR:
3915                 if (ifp->idev->cnf.forwarding)
3916                         addrconf_leave_anycast(ifp);
3917                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3918                 dst_hold(&ifp->rt->u.dst);
3919                 if (ip6_del_rt(ifp->rt))
3920                         dst_free(&ifp->rt->u.dst);
3921                 break;
3922         }
3923 }
3924
3925 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3926 {
3927         rcu_read_lock_bh();
3928         if (likely(ifp->idev->dead == 0))
3929                 __ipv6_ifa_notify(event, ifp);
3930         rcu_read_unlock_bh();
3931 }
3932
3933 #ifdef CONFIG_SYSCTL
3934
3935 static
3936 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3937                            void __user *buffer, size_t *lenp, loff_t *ppos)
3938 {
3939         int *valp = ctl->data;
3940         int val = *valp;
3941         int ret;
3942
3943         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3944
3945         if (write)
3946                 addrconf_fixup_forwarding(ctl, valp, val);
3947         return ret;
3948 }
3949
3950 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3951                                             int __user *name, int nlen,
3952                                             void __user *oldval,
3953                                             size_t __user *oldlenp,
3954                                             void __user *newval, size_t newlen)
3955 {
3956         int *valp = table->data;
3957         int val = *valp;
3958         int new;
3959
3960         if (!newval || !newlen)
3961                 return 0;
3962         if (newlen != sizeof(int))
3963                 return -EINVAL;
3964         if (get_user(new, (int __user *)newval))
3965                 return -EFAULT;
3966         if (new == *valp)
3967                 return 0;
3968         if (oldval && oldlenp) {
3969                 size_t len;
3970                 if (get_user(len, oldlenp))
3971                         return -EFAULT;
3972                 if (len) {
3973                         if (len > table->maxlen)
3974                                 len = table->maxlen;
3975                         if (copy_to_user(oldval, valp, len))
3976                                 return -EFAULT;
3977                         if (put_user(len, oldlenp))
3978                                 return -EFAULT;
3979                 }
3980         }
3981
3982         *valp = new;
3983         addrconf_fixup_forwarding(table, valp, val);
3984         return 1;
3985 }
3986
3987 static struct addrconf_sysctl_table
3988 {
3989         struct ctl_table_header *sysctl_header;
3990         ctl_table addrconf_vars[DEVCONF_MAX+1];
3991         char *dev_name;
3992 } addrconf_sysctl __read_mostly = {
3993         .sysctl_header = NULL,
3994         .addrconf_vars = {
3995                 {
3996                         .ctl_name       =       NET_IPV6_FORWARDING,
3997                         .procname       =       "forwarding",
3998                         .data           =       &ipv6_devconf.forwarding,
3999                         .maxlen         =       sizeof(int),
4000                         .mode           =       0644,
4001                         .proc_handler   =       &addrconf_sysctl_forward,
4002                         .strategy       =       &addrconf_sysctl_forward_strategy,
4003                 },
4004                 {
4005                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4006                         .procname       =       "hop_limit",
4007                         .data           =       &ipv6_devconf.hop_limit,
4008                         .maxlen         =       sizeof(int),
4009                         .mode           =       0644,
4010                         .proc_handler   =       proc_dointvec,
4011                 },
4012                 {
4013                         .ctl_name       =       NET_IPV6_MTU,
4014                         .procname       =       "mtu",
4015                         .data           =       &ipv6_devconf.mtu6,
4016                         .maxlen         =       sizeof(int),
4017                         .mode           =       0644,
4018                         .proc_handler   =       &proc_dointvec,
4019                 },
4020                 {
4021                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4022                         .procname       =       "accept_ra",
4023                         .data           =       &ipv6_devconf.accept_ra,
4024                         .maxlen         =       sizeof(int),
4025                         .mode           =       0644,
4026                         .proc_handler   =       &proc_dointvec,
4027                 },
4028                 {
4029                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4030                         .procname       =       "accept_redirects",
4031                         .data           =       &ipv6_devconf.accept_redirects,
4032                         .maxlen         =       sizeof(int),
4033                         .mode           =       0644,
4034                         .proc_handler   =       &proc_dointvec,
4035                 },
4036                 {
4037                         .ctl_name       =       NET_IPV6_AUTOCONF,
4038                         .procname       =       "autoconf",
4039                         .data           =       &ipv6_devconf.autoconf,
4040                         .maxlen         =       sizeof(int),
4041                         .mode           =       0644,
4042                         .proc_handler   =       &proc_dointvec,
4043                 },
4044                 {
4045                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4046                         .procname       =       "dad_transmits",
4047                         .data           =       &ipv6_devconf.dad_transmits,
4048                         .maxlen         =       sizeof(int),
4049                         .mode           =       0644,
4050                         .proc_handler   =       &proc_dointvec,
4051                 },
4052                 {
4053                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4054                         .procname       =       "router_solicitations",
4055                         .data           =       &ipv6_devconf.rtr_solicits,
4056                         .maxlen         =       sizeof(int),
4057                         .mode           =       0644,
4058                         .proc_handler   =       &proc_dointvec,
4059                 },
4060                 {
4061                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4062                         .procname       =       "router_solicitation_interval",
4063                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4064                         .maxlen         =       sizeof(int),
4065                         .mode           =       0644,
4066                         .proc_handler   =       &proc_dointvec_jiffies,
4067                         .strategy       =       &sysctl_jiffies,
4068                 },
4069                 {
4070                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4071                         .procname       =       "router_solicitation_delay",
4072                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4073                         .maxlen         =       sizeof(int),
4074                         .mode           =       0644,
4075                         .proc_handler   =       &proc_dointvec_jiffies,
4076                         .strategy       =       &sysctl_jiffies,
4077                 },
4078                 {
4079                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4080                         .procname       =       "force_mld_version",
4081                         .data           =       &ipv6_devconf.force_mld_version,
4082                         .maxlen         =       sizeof(int),
4083                         .mode           =       0644,
4084                         .proc_handler   =       &proc_dointvec,
4085                 },
4086 #ifdef CONFIG_IPV6_PRIVACY
4087                 {
4088                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4089                         .procname       =       "use_tempaddr",
4090                         .data           =       &ipv6_devconf.use_tempaddr,
4091                         .maxlen         =       sizeof(int),
4092                         .mode           =       0644,
4093                         .proc_handler   =       &proc_dointvec,
4094                 },
4095                 {
4096                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4097                         .procname       =       "temp_valid_lft",
4098                         .data           =       &ipv6_devconf.temp_valid_lft,
4099                         .maxlen         =       sizeof(int),
4100                         .mode           =       0644,
4101                         .proc_handler   =       &proc_dointvec,
4102                 },
4103                 {
4104                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4105                         .procname       =       "temp_prefered_lft",
4106                         .data           =       &ipv6_devconf.temp_prefered_lft,
4107                         .maxlen         =       sizeof(int),
4108                         .mode           =       0644,
4109                         .proc_handler   =       &proc_dointvec,
4110                 },
4111                 {
4112                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4113                         .procname       =       "regen_max_retry",
4114                         .data           =       &ipv6_devconf.regen_max_retry,
4115                         .maxlen         =       sizeof(int),
4116                         .mode           =       0644,
4117                         .proc_handler   =       &proc_dointvec,
4118                 },
4119                 {
4120                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4121                         .procname       =       "max_desync_factor",
4122                         .data           =       &ipv6_devconf.max_desync_factor,
4123                         .maxlen         =       sizeof(int),
4124                         .mode           =       0644,
4125                         .proc_handler   =       &proc_dointvec,
4126                 },
4127 #endif
4128                 {
4129                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4130                         .procname       =       "max_addresses",
4131                         .data           =       &ipv6_devconf.max_addresses,
4132                         .maxlen         =       sizeof(int),
4133                         .mode           =       0644,
4134                         .proc_handler   =       &proc_dointvec,
4135                 },
4136                 {
4137                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4138                         .procname       =       "accept_ra_defrtr",
4139                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4140                         .maxlen         =       sizeof(int),
4141                         .mode           =       0644,
4142                         .proc_handler   =       &proc_dointvec,
4143                 },
4144                 {
4145                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4146                         .procname       =       "accept_ra_pinfo",
4147                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4148                         .maxlen         =       sizeof(int),
4149                         .mode           =       0644,
4150                         .proc_handler   =       &proc_dointvec,
4151                 },
4152 #ifdef CONFIG_IPV6_ROUTER_PREF
4153                 {
4154                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4155                         .procname       =       "accept_ra_rtr_pref",
4156                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4157                         .maxlen         =       sizeof(int),
4158                         .mode           =       0644,
4159                         .proc_handler   =       &proc_dointvec,
4160                 },
4161                 {
4162                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4163                         .procname       =       "router_probe_interval",
4164                         .data           =       &ipv6_devconf.rtr_probe_interval,
4165                         .maxlen         =       sizeof(int),
4166                         .mode           =       0644,
4167                         .proc_handler   =       &proc_dointvec_jiffies,
4168                         .strategy       =       &sysctl_jiffies,
4169                 },
4170 #ifdef CONFIG_IPV6_ROUTE_INFO
4171                 {
4172                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4173                         .procname       =       "accept_ra_rt_info_max_plen",
4174                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4175                         .maxlen         =       sizeof(int),
4176                         .mode           =       0644,
4177                         .proc_handler   =       &proc_dointvec,
4178                 },
4179 #endif
4180 #endif
4181                 {
4182                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4183                         .procname       =       "proxy_ndp",
4184                         .data           =       &ipv6_devconf.proxy_ndp,
4185                         .maxlen         =       sizeof(int),
4186                         .mode           =       0644,
4187                         .proc_handler   =       &proc_dointvec,
4188                 },
4189                 {
4190                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4191                         .procname       =       "accept_source_route",
4192                         .data           =       &ipv6_devconf.accept_source_route,
4193                         .maxlen         =       sizeof(int),
4194                         .mode           =       0644,
4195                         .proc_handler   =       &proc_dointvec,
4196                 },
4197 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4198                 {
4199                         .ctl_name       =       CTL_UNNUMBERED,
4200                         .procname       =       "optimistic_dad",
4201                         .data           =       &ipv6_devconf.optimistic_dad,
4202                         .maxlen         =       sizeof(int),
4203                         .mode           =       0644,
4204                         .proc_handler   =       &proc_dointvec,
4205
4206                 },
4207 #endif
4208 #ifdef CONFIG_IPV6_MROUTE
4209                 {
4210                         .ctl_name       =       CTL_UNNUMBERED,
4211                         .procname       =       "mc_forwarding",
4212                         .data           =       &ipv6_devconf.mc_forwarding,
4213                         .maxlen         =       sizeof(int),
4214                         .mode           =       0644,
4215                         .proc_handler   =       &proc_dointvec,
4216                 },
4217 #endif
4218                 {
4219                         .ctl_name       =       0,      /* sentinel */
4220                 }
4221         },
4222 };
4223
4224 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4225                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4226 {
4227         int i;
4228         struct addrconf_sysctl_table *t;
4229
4230 #define ADDRCONF_CTL_PATH_DEV   3
4231
4232         struct ctl_path addrconf_ctl_path[] = {
4233                 { .procname = "net", .ctl_name = CTL_NET, },
4234                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4235                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4236                 { /* to be set */ },
4237                 { },
4238         };
4239
4240
4241         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4242         if (t == NULL)
4243                 goto out;
4244
4245         for (i=0; t->addrconf_vars[i].data; i++) {
4246                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4247                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4248                 t->addrconf_vars[i].extra2 = net;
4249         }
4250
4251         /*
4252          * Make a copy of dev_name, because '.procname' is regarded as const
4253          * by sysctl and we wouldn't want anyone to change it under our feet
4254          * (see SIOCSIFNAME).
4255          */
4256         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4257         if (!t->dev_name)
4258                 goto free;
4259
4260         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4261         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4262
4263         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4264                         t->addrconf_vars);
4265         if (t->sysctl_header == NULL)
4266                 goto free_procname;
4267
4268         p->sysctl = t;
4269         return 0;
4270
4271 free_procname:
4272         kfree(t->dev_name);
4273 free:
4274         kfree(t);
4275 out:
4276         return -ENOBUFS;
4277 }
4278
4279 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4280 {
4281         struct addrconf_sysctl_table *t;
4282
4283         if (p->sysctl == NULL)
4284                 return;
4285
4286         t = p->sysctl;
4287         p->sysctl = NULL;
4288         unregister_sysctl_table(t->sysctl_header);
4289         kfree(t->dev_name);
4290         kfree(t);
4291 }
4292
4293 static void addrconf_sysctl_register(struct inet6_dev *idev)
4294 {
4295         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4296                               NET_IPV6_NEIGH, "ipv6",
4297                               &ndisc_ifinfo_sysctl_change,
4298                               ndisc_ifinfo_sysctl_strategy);
4299         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4300                         idev->dev->ifindex, idev, &idev->cnf);
4301 }
4302
4303 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4304 {
4305         __addrconf_sysctl_unregister(&idev->cnf);
4306         neigh_sysctl_unregister(idev->nd_parms);
4307 }
4308
4309
4310 #endif
4311
4312 static int addrconf_init_net(struct net *net)
4313 {
4314         int err;
4315         struct ipv6_devconf *all, *dflt;
4316
4317         err = -ENOMEM;
4318         all = &ipv6_devconf;
4319         dflt = &ipv6_devconf_dflt;
4320
4321         if (net != &init_net) {
4322                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4323                 if (all == NULL)
4324                         goto err_alloc_all;
4325
4326                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4327                 if (dflt == NULL)
4328                         goto err_alloc_dflt;
4329         }
4330
4331         net->ipv6.devconf_all = all;
4332         net->ipv6.devconf_dflt = dflt;
4333
4334 #ifdef CONFIG_SYSCTL
4335         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4336                         NULL, all);
4337         if (err < 0)
4338                 goto err_reg_all;
4339
4340         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4341                         NULL, dflt);
4342         if (err < 0)
4343                 goto err_reg_dflt;
4344 #endif
4345         return 0;
4346
4347 #ifdef CONFIG_SYSCTL
4348 err_reg_dflt:
4349         __addrconf_sysctl_unregister(all);
4350 err_reg_all:
4351         kfree(dflt);
4352 #endif
4353 err_alloc_dflt:
4354         kfree(all);
4355 err_alloc_all:
4356         return err;
4357 }
4358
4359 static void addrconf_exit_net(struct net *net)
4360 {
4361 #ifdef CONFIG_SYSCTL
4362         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4363         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4364 #endif
4365         if (net != &init_net) {
4366                 kfree(net->ipv6.devconf_dflt);
4367                 kfree(net->ipv6.devconf_all);
4368         }
4369 }
4370
4371 static struct pernet_operations addrconf_ops = {
4372         .init = addrconf_init_net,
4373         .exit = addrconf_exit_net,
4374 };
4375
4376 /*
4377  *      Device notifier
4378  */
4379
4380 int register_inet6addr_notifier(struct notifier_block *nb)
4381 {
4382         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4383 }
4384
4385 EXPORT_SYMBOL(register_inet6addr_notifier);
4386
4387 int unregister_inet6addr_notifier(struct notifier_block *nb)
4388 {
4389         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4390 }
4391
4392 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4393
4394 static void addrconf_net_exit(struct net *net)
4395 {
4396         struct net_device *dev;
4397
4398         rtnl_lock();
4399         /* clean dev list */
4400         for_each_netdev(net, dev) {
4401                 if (__in6_dev_get(dev) == NULL)
4402                         continue;
4403                 addrconf_ifdown(dev, 1);
4404         }
4405         addrconf_ifdown(net->loopback_dev, 2);
4406         rtnl_unlock();
4407 }
4408
4409 static struct pernet_operations addrconf_net_ops = {
4410         .exit = addrconf_net_exit,
4411 };
4412
4413 /*
4414  *      Init / cleanup code
4415  */
4416
4417 int __init addrconf_init(void)
4418 {
4419         int err;
4420
4421         if ((err = ipv6_addr_label_init()) < 0) {
4422                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4423                         err);
4424                 return err;
4425         }
4426
4427         register_pernet_subsys(&addrconf_ops);
4428
4429         /* The addrconf netdev notifier requires that loopback_dev
4430          * has it's ipv6 private information allocated and setup
4431          * before it can bring up and give link-local addresses
4432          * to other devices which are up.
4433          *
4434          * Unfortunately, loopback_dev is not necessarily the first
4435          * entry in the global dev_base list of net devices.  In fact,
4436          * it is likely to be the very last entry on that list.
4437          * So this causes the notifier registry below to try and
4438          * give link-local addresses to all devices besides loopback_dev
4439          * first, then loopback_dev, which cases all the non-loopback_dev
4440          * devices to fail to get a link-local address.
4441          *
4442          * So, as a temporary fix, allocate the ipv6 structure for
4443          * loopback_dev first by hand.
4444          * Longer term, all of the dependencies ipv6 has upon the loopback
4445          * device and it being up should be removed.
4446          */
4447         rtnl_lock();
4448         if (!ipv6_add_dev(init_net.loopback_dev))
4449                 err = -ENOMEM;
4450         rtnl_unlock();
4451         if (err)
4452                 goto errlo;
4453
4454         err = register_pernet_device(&addrconf_net_ops);
4455         if (err)
4456                 return err;
4457
4458         register_netdevice_notifier(&ipv6_dev_notf);
4459
4460         addrconf_verify(0);
4461
4462         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4463         if (err < 0)
4464                 goto errout;
4465
4466         /* Only the first call to __rtnl_register can fail */
4467         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4468         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4469         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4470         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4471         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4472
4473         ipv6_addr_label_rtnl_register();
4474
4475         return 0;
4476 errout:
4477         unregister_netdevice_notifier(&ipv6_dev_notf);
4478 errlo:
4479         unregister_pernet_subsys(&addrconf_ops);
4480
4481         return err;
4482 }
4483
4484 void addrconf_cleanup(void)
4485 {
4486         struct inet6_ifaddr *ifa;
4487         int i;
4488
4489         unregister_netdevice_notifier(&ipv6_dev_notf);
4490         unregister_pernet_device(&addrconf_net_ops);
4491
4492         unregister_pernet_subsys(&addrconf_ops);
4493
4494         rtnl_lock();
4495
4496         /*
4497          *      Check hash table.
4498          */
4499         write_lock_bh(&addrconf_hash_lock);
4500         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4501                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4502                         struct inet6_ifaddr *bifa;
4503
4504                         bifa = ifa;
4505                         ifa = ifa->lst_next;
4506                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4507                         /* Do not free it; something is wrong.
4508                            Now we can investigate it with debugger.
4509                          */
4510                 }
4511         }
4512         write_unlock_bh(&addrconf_hash_lock);
4513
4514         del_timer(&addr_chk_timer);
4515         rtnl_unlock();
4516
4517         unregister_pernet_subsys(&addrconf_net_ops);
4518 }