git.oblomov.eu Git - 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  *      $Id: addrconf.c,v 1.69 2001/10/31 21:55:54 davem Exp $
  10  *
  11  *      This program is free software; you can redistribute it and/or
  12  *      modify it under the terms of the GNU General Public License
  13  *      as published by the Free Software Foundation; either version
  14  *      2 of the License, or (at your option) any later version.
  15  */
  16
  17 /*
  18  *      Changes:
  19  *
  20  *      Janos Farkas                    :       delete timer on ifdown
  21  *      <chexum@bankinf.banki.hu>
  22  *      Andi Kleen                      :       kill double kfree on module
  23  *                                              unload.
  24  *      Maciej W. Rozycki               :       FDDI support
  25  *      sekiya@USAGI                    :       Don't send too many RS
  26  *                                              packets.
  27  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  28  *                                              packets.
  29  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  30  *                                              address validation timer.
  31  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  32  *                                              support.
  33  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  34  *                                              address on a same interface.
  35  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  36  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  37  *                                              seq_file.
  38  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  39  *                                              selection; consider scope,
  40  *                                              status etc.
  41  */
  42
  43 #include <linux/errno.h>
  44 #include <linux/types.h>
  45 #include <linux/socket.h>
  46 #include <linux/sockios.h>
  47 #include <linux/net.h>
  48 #include <linux/in6.h>
  49 #include <linux/netdevice.h>
  50 #include <linux/if_addr.h>
  51 #include <linux/if_arp.h>
  52 #include <linux/if_arcnet.h>
  53 #include <linux/if_infiniband.h>
  54 #include <linux/route.h>
  55 #include <linux/inetdevice.h>
  56 #include <linux/init.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <asm/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90
  91 /* Set to 3 to get tracing... */
  92 #define ACONF_DEBUG 2
  93
  94 #if ACONF_DEBUG >= 3
  95 #define ADBG(x) printk x
  96 #else
  97 #define ADBG(x)
  98 #endif
  99
 100 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 101 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 102
 103 #ifdef CONFIG_SYSCTL
 104 static void addrconf_sysctl_register(struct inet6_dev *idev);
 105 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 106 #else
 107 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 108 {
 109 }
 110
 111 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 112 {
 113 }
 114 #endif
 115
 116 #ifdef CONFIG_IPV6_PRIVACY
 117 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 118 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 119 static void ipv6_regen_rndid(unsigned long data);
 120
 121 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 122 #endif
 123
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188 };
 189
 190 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 191         .forwarding             = 0,
 192         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 193         .mtu6                   = IPV6_MIN_MTU,
 194         .accept_ra              = 1,
 195         .accept_redirects       = 1,
 196         .autoconf               = 1,
 197         .dad_transmits          = 1,
 198         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 199         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 200         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 201 #ifdef CONFIG_IPV6_PRIVACY
 202         .use_tempaddr           = 0,
 203         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 204         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 205         .regen_max_retry        = REGEN_MAX_RETRY,
 206         .max_desync_factor      = MAX_DESYNC_FACTOR,
 207 #endif
 208         .max_addresses          = IPV6_MAX_ADDRESSES,
 209         .accept_ra_defrtr       = 1,
 210         .accept_ra_pinfo        = 1,
 211 #ifdef CONFIG_IPV6_ROUTER_PREF
 212         .accept_ra_rtr_pref     = 1,
 213         .rtr_probe_interval     = 60 * HZ,
 214 #ifdef CONFIG_IPV6_ROUTE_INFO
 215         .accept_ra_rt_info_max_plen = 0,
 216 #endif
 217 #endif
 218         .proxy_ndp              = 0,
 219         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 220 };
 221
 222 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 223 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 224 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 225 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 226 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 227
 228 /* Check if a valid qdisc is available */
 229 static inline int addrconf_qdisc_ok(struct net_device *dev)
 230 {
 231         return (dev->qdisc != &noop_qdisc);
 232 }
 233
 234 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 235 {
 236         if (del_timer(&ifp->timer))
 237                 __in6_ifa_put(ifp);
 238 }
 239
 240 enum addrconf_timer_t
 241 {
 242         AC_NONE,
 243         AC_DAD,
 244         AC_RS,
 245 };
 246
 247 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 248                                enum addrconf_timer_t what,
 249                                unsigned long when)
 250 {
 251         if (!del_timer(&ifp->timer))
 252                 in6_ifa_hold(ifp);
 253
 254         switch (what) {
 255         case AC_DAD:
 256                 ifp->timer.function = addrconf_dad_timer;
 257                 break;
 258         case AC_RS:
 259                 ifp->timer.function = addrconf_rs_timer;
 260                 break;
 261         default:;
 262         }
 263         ifp->timer.expires = jiffies + when;
 264         add_timer(&ifp->timer);
 265 }
 266
 267 static int snmp6_alloc_dev(struct inet6_dev *idev)
 268 {
 269         if (snmp_mib_init((void **)idev->stats.ipv6,
 270                           sizeof(struct ipstats_mib)) < 0)
 271                 goto err_ip;
 272         if (snmp_mib_init((void **)idev->stats.icmpv6,
 273                           sizeof(struct icmpv6_mib)) < 0)
 274                 goto err_icmp;
 275         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 276                           sizeof(struct icmpv6msg_mib)) < 0)
 277                 goto err_icmpmsg;
 278
 279         return 0;
 280
 281 err_icmpmsg:
 282         snmp_mib_free((void **)idev->stats.icmpv6);
 283 err_icmp:
 284         snmp_mib_free((void **)idev->stats.ipv6);
 285 err_ip:
 286         return -ENOMEM;
 287 }
 288
 289 static void snmp6_free_dev(struct inet6_dev *idev)
 290 {
 291         snmp_mib_free((void **)idev->stats.icmpv6msg);
 292         snmp_mib_free((void **)idev->stats.icmpv6);
 293         snmp_mib_free((void **)idev->stats.ipv6);
 294 }
 295
 296 /* Nobody refers to this device, we may destroy it. */
 297
 298 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 299 {
 300         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 301         kfree(idev);
 302 }
 303
 304 void in6_dev_finish_destroy(struct inet6_dev *idev)
 305 {
 306         struct net_device *dev = idev->dev;
 307         BUG_TRAP(idev->addr_list==NULL);
 308         BUG_TRAP(idev->mc_list==NULL);
 309 #ifdef NET_REFCNT_DEBUG
 310         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 311 #endif
 312         dev_put(dev);
 313         if (!idev->dead) {
 314                 printk("Freeing alive inet6 device %p\n", idev);
 315                 return;
 316         }
 317         snmp6_free_dev(idev);
 318         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 319 }
 320
 321 EXPORT_SYMBOL(in6_dev_finish_destroy);
 322
 323 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 324 {
 325         struct inet6_dev *ndev;
 326
 327         ASSERT_RTNL();
 328
 329         if (dev->mtu < IPV6_MIN_MTU)
 330                 return NULL;
 331
 332         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 333
 334         if (ndev == NULL)
 335                 return NULL;
 336
 337         rwlock_init(&ndev->lock);
 338         ndev->dev = dev;
 339         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 340         ndev->cnf.mtu6 = dev->mtu;
 341         ndev->cnf.sysctl = NULL;
 342         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 343         if (ndev->nd_parms == NULL) {
 344                 kfree(ndev);
 345                 return NULL;
 346         }
 347         /* We refer to the device */
 348         dev_hold(dev);
 349
 350         if (snmp6_alloc_dev(ndev) < 0) {
 351                 ADBG((KERN_WARNING
 352                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 353                         __func__, dev->name));
 354                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 355                 ndev->dead = 1;
 356                 in6_dev_finish_destroy(ndev);
 357                 return NULL;
 358         }
 359
 360         if (snmp6_register_dev(ndev) < 0) {
 361                 ADBG((KERN_WARNING
 362                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 363                         __func__, dev->name));
 364                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 365                 ndev->dead = 1;
 366                 in6_dev_finish_destroy(ndev);
 367                 return NULL;
 368         }
 369
 370         /* One reference from device.  We must do this before
 371          * we invoke __ipv6_regen_rndid().
 372          */
 373         in6_dev_hold(ndev);
 374
 375 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 376         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 377                 printk(KERN_INFO
 378                        "%s: Disabled Multicast RS\n",
 379                        dev->name);
 380                 ndev->cnf.rtr_solicits = 0;
 381         }
 382 #endif
 383
 384 #ifdef CONFIG_IPV6_PRIVACY
 385         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 386         if ((dev->flags&IFF_LOOPBACK) ||
 387             dev->type == ARPHRD_TUNNEL ||
 388             dev->type == ARPHRD_TUNNEL6 ||
 389             dev->type == ARPHRD_SIT ||
 390             dev->type == ARPHRD_NONE) {
 391                 printk(KERN_INFO
 392                        "%s: Disabled Privacy Extensions\n",
 393                        dev->name);
 394                 ndev->cnf.use_tempaddr = -1;
 395         } else {
 396                 in6_dev_hold(ndev);
 397                 ipv6_regen_rndid((unsigned long) ndev);
 398         }
 399 #endif
 400
 401         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 402                 ndev->if_flags |= IF_READY;
 403
 404         ipv6_mc_init_dev(ndev);
 405         ndev->tstamp = jiffies;
 406         addrconf_sysctl_register(ndev);
 407         /* protected by rtnl_lock */
 408         rcu_assign_pointer(dev->ip6_ptr, ndev);
 409
 410         /* Join all-node multicast group */
 411         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 412
 413         return ndev;
 414 }
 415
 416 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 417 {
 418         struct inet6_dev *idev;
 419
 420         ASSERT_RTNL();
 421
 422         if ((idev = __in6_dev_get(dev)) == NULL) {
 423                 if ((idev = ipv6_add_dev(dev)) == NULL)
 424                         return NULL;
 425         }
 426
 427         if (dev->flags&IFF_UP)
 428                 ipv6_mc_up(idev);
 429         return idev;
 430 }
 431
 432 #ifdef CONFIG_SYSCTL
 433 static void dev_forward_change(struct inet6_dev *idev)
 434 {
 435         struct net_device *dev;
 436         struct inet6_ifaddr *ifa;
 437
 438         if (!idev)
 439                 return;
 440         dev = idev->dev;
 441         if (dev && (dev->flags & IFF_MULTICAST)) {
 442                 if (idev->cnf.forwarding)
 443                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 444                 else
 445                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 446         }
 447         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 448                 if (ifa->flags&IFA_F_TENTATIVE)
 449                         continue;
 450                 if (idev->cnf.forwarding)
 451                         addrconf_join_anycast(ifa);
 452                 else
 453                         addrconf_leave_anycast(ifa);
 454         }
 455 }
 456
 457
 458 static void addrconf_forward_change(struct net *net, __s32 newf)
 459 {
 460         struct net_device *dev;
 461         struct inet6_dev *idev;
 462
 463         read_lock(&dev_base_lock);
 464         for_each_netdev(net, dev) {
 465                 rcu_read_lock();
 466                 idev = __in6_dev_get(dev);
 467                 if (idev) {
 468                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 469                         idev->cnf.forwarding = newf;
 470                         if (changed)
 471                                 dev_forward_change(idev);
 472                 }
 473                 rcu_read_unlock();
 474         }
 475         read_unlock(&dev_base_lock);
 476 }
 477
 478 static void addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 479 {
 480         struct net *net;
 481
 482         net = (struct net *)table->extra2;
 483         if (p == &net->ipv6.devconf_dflt->forwarding)
 484                 return;
 485
 486         if (p == &net->ipv6.devconf_all->forwarding) {
 487                 __s32 newf = net->ipv6.devconf_all->forwarding;
 488                 net->ipv6.devconf_dflt->forwarding = newf;
 489                 addrconf_forward_change(net, newf);
 490         } else if ((!*p) ^ (!old))
 491                 dev_forward_change((struct inet6_dev *)table->extra1);
 492
 493         if (*p)
 494                 rt6_purge_dflt_routers(net);
 495 }
 496 #endif
 497
 498 /* Nobody refers to this ifaddr, destroy it */
 499
 500 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 501 {
 502         BUG_TRAP(ifp->if_next==NULL);
 503         BUG_TRAP(ifp->lst_next==NULL);
 504 #ifdef NET_REFCNT_DEBUG
 505         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 506 #endif
 507
 508         in6_dev_put(ifp->idev);
 509
 510         if (del_timer(&ifp->timer))
 511                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 512
 513         if (!ifp->dead) {
 514                 printk("Freeing alive inet6 address %p\n", ifp);
 515                 return;
 516         }
 517         dst_release(&ifp->rt->u.dst);
 518
 519         kfree(ifp);
 520 }
 521
 522 static void
 523 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 524 {
 525         struct inet6_ifaddr *ifa, **ifap;
 526         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 527
 528         /*
 529          * Each device address list is sorted in order of scope -
 530          * global before linklocal.
 531          */
 532         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 533              ifap = &ifa->if_next) {
 534                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 535                         break;
 536         }
 537
 538         ifp->if_next = *ifap;
 539         *ifap = ifp;
 540 }
 541
 542 /*
 543  *      Hash function taken from net_alias.c
 544  */
 545 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 546 {
 547         __u32 word;
 548
 549         /*
 550          * We perform the hash function over the last 64 bits of the address
 551          * This will include the IEEE address token on links that support it.
 552          */
 553
 554         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 555         word ^= (word >> 16);
 556         word ^= (word >> 8);
 557
 558         return ((word ^ (word >> 4)) & 0x0f);
 559 }
 560
 561 /* On success it returns ifp with increased reference count */
 562
 563 static struct inet6_ifaddr *
 564 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 565               int scope, u32 flags)
 566 {
 567         struct inet6_ifaddr *ifa = NULL;
 568         struct rt6_info *rt;
 569         int hash;
 570         int err = 0;
 571
 572         rcu_read_lock_bh();
 573         if (idev->dead) {
 574                 err = -ENODEV;                  /*XXX*/
 575                 goto out2;
 576         }
 577
 578         write_lock(&addrconf_hash_lock);
 579
 580         /* Ignore adding duplicate addresses on an interface */
 581         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 582                 ADBG(("ipv6_add_addr: already assigned\n"));
 583                 err = -EEXIST;
 584                 goto out;
 585         }
 586
 587         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 588
 589         if (ifa == NULL) {
 590                 ADBG(("ipv6_add_addr: malloc failed\n"));
 591                 err = -ENOBUFS;
 592                 goto out;
 593         }
 594
 595         rt = addrconf_dst_alloc(idev, addr, 0);
 596         if (IS_ERR(rt)) {
 597                 err = PTR_ERR(rt);
 598                 goto out;
 599         }
 600
 601         ipv6_addr_copy(&ifa->addr, addr);
 602
 603         spin_lock_init(&ifa->lock);
 604         init_timer(&ifa->timer);
 605         ifa->timer.data = (unsigned long) ifa;
 606         ifa->scope = scope;
 607         ifa->prefix_len = pfxlen;
 608         ifa->flags = flags | IFA_F_TENTATIVE;
 609         ifa->cstamp = ifa->tstamp = jiffies;
 610
 611         ifa->rt = rt;
 612
 613         /*
 614          * part one of RFC 4429, section 3.3
 615          * We should not configure an address as
 616          * optimistic if we do not yet know the link
 617          * layer address of our nexhop router
 618          */
 619
 620         if (rt->rt6i_nexthop == NULL)
 621                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 622
 623         ifa->idev = idev;
 624         in6_dev_hold(idev);
 625         /* For caller */
 626         in6_ifa_hold(ifa);
 627
 628         /* Add to big hash table */
 629         hash = ipv6_addr_hash(addr);
 630
 631         ifa->lst_next = inet6_addr_lst[hash];
 632         inet6_addr_lst[hash] = ifa;
 633         in6_ifa_hold(ifa);
 634         write_unlock(&addrconf_hash_lock);
 635
 636         write_lock(&idev->lock);
 637         /* Add to inet6_dev unicast addr list. */
 638         ipv6_link_dev_addr(idev, ifa);
 639
 640 #ifdef CONFIG_IPV6_PRIVACY
 641         if (ifa->flags&IFA_F_TEMPORARY) {
 642                 ifa->tmp_next = idev->tempaddr_list;
 643                 idev->tempaddr_list = ifa;
 644                 in6_ifa_hold(ifa);
 645         }
 646 #endif
 647
 648         in6_ifa_hold(ifa);
 649         write_unlock(&idev->lock);
 650 out2:
 651         rcu_read_unlock_bh();
 652
 653         if (likely(err == 0))
 654                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 655         else {
 656                 kfree(ifa);
 657                 ifa = ERR_PTR(err);
 658         }
 659
 660         return ifa;
 661 out:
 662         write_unlock(&addrconf_hash_lock);
 663         goto out2;
 664 }
 665
 666 /* This function wants to get referenced ifp and releases it before return */
 667
 668 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 669 {
 670         struct inet6_ifaddr *ifa, **ifap;
 671         struct inet6_dev *idev = ifp->idev;
 672         int hash;
 673         int deleted = 0, onlink = 0;
 674         unsigned long expires = jiffies;
 675
 676         hash = ipv6_addr_hash(&ifp->addr);
 677
 678         ifp->dead = 1;
 679
 680         write_lock_bh(&addrconf_hash_lock);
 681         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 682              ifap = &ifa->lst_next) {
 683                 if (ifa == ifp) {
 684                         *ifap = ifa->lst_next;
 685                         __in6_ifa_put(ifp);
 686                         ifa->lst_next = NULL;
 687                         break;
 688                 }
 689         }
 690         write_unlock_bh(&addrconf_hash_lock);
 691
 692         write_lock_bh(&idev->lock);
 693 #ifdef CONFIG_IPV6_PRIVACY
 694         if (ifp->flags&IFA_F_TEMPORARY) {
 695                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 696                      ifap = &ifa->tmp_next) {
 697                         if (ifa == ifp) {
 698                                 *ifap = ifa->tmp_next;
 699                                 if (ifp->ifpub) {
 700                                         in6_ifa_put(ifp->ifpub);
 701                                         ifp->ifpub = NULL;
 702                                 }
 703                                 __in6_ifa_put(ifp);
 704                                 ifa->tmp_next = NULL;
 705                                 break;
 706                         }
 707                 }
 708         }
 709 #endif
 710
 711         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 712                 if (ifa == ifp) {
 713                         *ifap = ifa->if_next;
 714                         __in6_ifa_put(ifp);
 715                         ifa->if_next = NULL;
 716                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 717                                 break;
 718                         deleted = 1;
 719                         continue;
 720                 } else if (ifp->flags & IFA_F_PERMANENT) {
 721                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 722                                               ifp->prefix_len)) {
 723                                 if (ifa->flags & IFA_F_PERMANENT) {
 724                                         onlink = 1;
 725                                         if (deleted)
 726                                                 break;
 727                                 } else {
 728                                         unsigned long lifetime;
 729
 730                                         if (!onlink)
 731                                                 onlink = -1;
 732
 733                                         spin_lock(&ifa->lock);
 734
 735                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 736                                         /*
 737                                          * Note: Because this address is
 738                                          * not permanent, lifetime <
 739                                          * LONG_MAX / HZ here.
 740                                          */
 741                                         if (time_before(expires,
 742                                                         ifa->tstamp + lifetime * HZ))
 743                                                 expires = ifa->tstamp + lifetime * HZ;
 744                                         spin_unlock(&ifa->lock);
 745                                 }
 746                         }
 747                 }
 748                 ifap = &ifa->if_next;
 749         }
 750         write_unlock_bh(&idev->lock);
 751
 752         ipv6_ifa_notify(RTM_DELADDR, ifp);
 753
 754         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 755
 756         addrconf_del_timer(ifp);
 757
 758         /*
 759          * Purge or update corresponding prefix
 760          *
 761          * 1) we don't purge prefix here if address was not permanent.
 762          *    prefix is managed by its own lifetime.
 763          * 2) if there're no addresses, delete prefix.
 764          * 3) if there're still other permanent address(es),
 765          *    corresponding prefix is still permanent.
 766          * 4) otherwise, update prefix lifetime to the
 767          *    longest valid lifetime among the corresponding
 768          *    addresses on the device.
 769          *    Note: subsequent RA will update lifetime.
 770          *
 771          * --yoshfuji
 772          */
 773         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 774                 struct in6_addr prefix;
 775                 struct rt6_info *rt;
 776                 struct net *net = dev_net(ifp->idev->dev);
 777                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 778                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 779
 780                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 781                         if (onlink == 0) {
 782                                 ip6_del_rt(rt);
 783                                 rt = NULL;
 784                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 785                                 rt->rt6i_expires = expires;
 786                                 rt->rt6i_flags |= RTF_EXPIRES;
 787                         }
 788                 }
 789                 dst_release(&rt->u.dst);
 790         }
 791
 792         in6_ifa_put(ifp);
 793 }
 794
 795 #ifdef CONFIG_IPV6_PRIVACY
 796 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 797 {
 798         struct inet6_dev *idev = ifp->idev;
 799         struct in6_addr addr, *tmpaddr;
 800         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 801         unsigned long regen_advance;
 802         int tmp_plen;
 803         int ret = 0;
 804         int max_addresses;
 805         u32 addr_flags;
 806
 807         write_lock(&idev->lock);
 808         if (ift) {
 809                 spin_lock_bh(&ift->lock);
 810                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 811                 spin_unlock_bh(&ift->lock);
 812                 tmpaddr = &addr;
 813         } else {
 814                 tmpaddr = NULL;
 815         }
 816 retry:
 817         in6_dev_hold(idev);
 818         if (idev->cnf.use_tempaddr <= 0) {
 819                 write_unlock(&idev->lock);
 820                 printk(KERN_INFO
 821                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 822                 in6_dev_put(idev);
 823                 ret = -1;
 824                 goto out;
 825         }
 826         spin_lock_bh(&ifp->lock);
 827         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 828                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 829                 spin_unlock_bh(&ifp->lock);
 830                 write_unlock(&idev->lock);
 831                 printk(KERN_WARNING
 832                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 833                 in6_dev_put(idev);
 834                 ret = -1;
 835                 goto out;
 836         }
 837         in6_ifa_hold(ifp);
 838         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 839         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 840                 spin_unlock_bh(&ifp->lock);
 841                 write_unlock(&idev->lock);
 842                 printk(KERN_WARNING
 843                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 844                 in6_ifa_put(ifp);
 845                 in6_dev_put(idev);
 846                 ret = -1;
 847                 goto out;
 848         }
 849         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 850         tmp_valid_lft = min_t(__u32,
 851                               ifp->valid_lft,
 852                               idev->cnf.temp_valid_lft);
 853         tmp_prefered_lft = min_t(__u32,
 854                                  ifp->prefered_lft,
 855                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 856         tmp_plen = ifp->prefix_len;
 857         max_addresses = idev->cnf.max_addresses;
 858         tmp_cstamp = ifp->cstamp;
 859         tmp_tstamp = ifp->tstamp;
 860         spin_unlock_bh(&ifp->lock);
 861
 862         regen_advance = idev->cnf.regen_max_retry *
 863                         idev->cnf.dad_transmits *
 864                         idev->nd_parms->retrans_time / HZ;
 865         write_unlock(&idev->lock);
 866
 867         /* A temporary address is created only if this calculated Preferred
 868          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 869          * an implementation must not create a temporary address with a zero
 870          * Preferred Lifetime.
 871          */
 872         if (tmp_prefered_lft <= regen_advance) {
 873                 in6_ifa_put(ifp);
 874                 in6_dev_put(idev);
 875                 ret = -1;
 876                 goto out;
 877         }
 878
 879         addr_flags = IFA_F_TEMPORARY;
 880         /* set in addrconf_prefix_rcv() */
 881         if (ifp->flags & IFA_F_OPTIMISTIC)
 882                 addr_flags |= IFA_F_OPTIMISTIC;
 883
 884         ift = !max_addresses ||
 885               ipv6_count_addresses(idev) < max_addresses ?
 886                 ipv6_add_addr(idev, &addr, tmp_plen,
 887                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 888                               addr_flags) : NULL;
 889         if (!ift || IS_ERR(ift)) {
 890                 in6_ifa_put(ifp);
 891                 in6_dev_put(idev);
 892                 printk(KERN_INFO
 893                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 894                 tmpaddr = &addr;
 895                 write_lock(&idev->lock);
 896                 goto retry;
 897         }
 898
 899         spin_lock_bh(&ift->lock);
 900         ift->ifpub = ifp;
 901         ift->valid_lft = tmp_valid_lft;
 902         ift->prefered_lft = tmp_prefered_lft;
 903         ift->cstamp = tmp_cstamp;
 904         ift->tstamp = tmp_tstamp;
 905         spin_unlock_bh(&ift->lock);
 906
 907         addrconf_dad_start(ift, 0);
 908         in6_ifa_put(ift);
 909         in6_dev_put(idev);
 910 out:
 911         return ret;
 912 }
 913 #endif
 914
 915 /*
 916  *      Choose an appropriate source address (RFC3484)
 917  */
 918 enum {
 919         IPV6_SADDR_RULE_INIT = 0,
 920         IPV6_SADDR_RULE_LOCAL,
 921         IPV6_SADDR_RULE_SCOPE,
 922         IPV6_SADDR_RULE_PREFERRED,
 923 #ifdef CONFIG_IPV6_MIP6
 924         IPV6_SADDR_RULE_HOA,
 925 #endif
 926         IPV6_SADDR_RULE_OIF,
 927         IPV6_SADDR_RULE_LABEL,
 928 #ifdef CONFIG_IPV6_PRIVACY
 929         IPV6_SADDR_RULE_PRIVACY,
 930 #endif
 931         IPV6_SADDR_RULE_ORCHID,
 932         IPV6_SADDR_RULE_PREFIX,
 933         IPV6_SADDR_RULE_MAX
 934 };
 935
 936 struct ipv6_saddr_score {
 937         int                     rule;
 938         int                     addr_type;
 939         struct inet6_ifaddr     *ifa;
 940         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 941         int                     scopedist;
 942         int                     matchlen;
 943 };
 944
 945 struct ipv6_saddr_dst {
 946         const struct in6_addr *addr;
 947         int ifindex;
 948         int scope;
 949         int label;
 950         unsigned int prefs;
 951 };
 952
 953 static inline int ipv6_saddr_preferred(int type)
 954 {
 955         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 956                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 957                 return 1;
 958         return 0;
 959 }
 960
 961 static int ipv6_get_saddr_eval(struct ipv6_saddr_score *score,
 962                                struct ipv6_saddr_dst *dst,
 963                                int i)
 964 {
 965         int ret;
 966
 967         if (i <= score->rule) {
 968                 switch (i) {
 969                 case IPV6_SADDR_RULE_SCOPE:
 970                         ret = score->scopedist;
 971                         break;
 972                 case IPV6_SADDR_RULE_PREFIX:
 973                         ret = score->matchlen;
 974                         break;
 975                 default:
 976                         ret = !!test_bit(i, score->scorebits);
 977                 }
 978                 goto out;
 979         }
 980
 981         switch (i) {
 982         case IPV6_SADDR_RULE_INIT:
 983                 /* Rule 0: remember if hiscore is not ready yet */
 984                 ret = !!score->ifa;
 985                 break;
 986         case IPV6_SADDR_RULE_LOCAL:
 987                 /* Rule 1: Prefer same address */
 988                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
 989                 break;
 990         case IPV6_SADDR_RULE_SCOPE:
 991                 /* Rule 2: Prefer appropriate scope
 992                  *
 993                  *      ret
 994                  *       ^
 995                  *    -1 |  d 15
 996                  *    ---+--+-+---> scope
 997                  *       |
 998                  *       |             d is scope of the destination.
 999                  *  B-d  |  \
1000                  *       |   \      <- smaller scope is better if
1001                  *  B-15 |    \        if scope is enough for destinaion.
1002                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1003                  * d-C-1 | /
1004                  *       |/         <- greater is better
1005                  *   -C  /             if scope is not enough for destination.
1006                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1007                  *
1008                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1009                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1010                  * Assume B = 0 and we get C > 29.
1011                  */
1012                 ret = __ipv6_addr_src_scope(score->addr_type);
1013                 if (ret >= dst->scope)
1014                         ret = -ret;
1015                 else
1016                         ret -= 128;     /* 30 is enough */
1017                 score->scopedist = ret;
1018                 break;
1019         case IPV6_SADDR_RULE_PREFERRED:
1020                 /* Rule 3: Avoid deprecated and optimistic addresses */
1021                 ret = ipv6_saddr_preferred(score->addr_type) ||
1022                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1023                 break;
1024 #ifdef CONFIG_IPV6_MIP6
1025         case IPV6_SADDR_RULE_HOA:
1026             {
1027                 /* Rule 4: Prefer home address */
1028                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1029                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1030                 break;
1031             }
1032 #endif
1033         case IPV6_SADDR_RULE_OIF:
1034                 /* Rule 5: Prefer outgoing interface */
1035                 ret = (!dst->ifindex ||
1036                        dst->ifindex == score->ifa->idev->dev->ifindex);
1037                 break;
1038         case IPV6_SADDR_RULE_LABEL:
1039                 /* Rule 6: Prefer matching label */
1040                 ret = ipv6_addr_label(&score->ifa->addr, score->addr_type,
1041                                       score->ifa->idev->dev->ifindex) == dst->label;
1042                 break;
1043 #ifdef CONFIG_IPV6_PRIVACY
1044         case IPV6_SADDR_RULE_PRIVACY:
1045             {
1046                 /* Rule 7: Prefer public address
1047                  * Note: prefer temprary address if use_tempaddr >= 2
1048                  */
1049                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1050                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1051                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1052                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1053                 break;
1054             }
1055 #endif
1056         case IPV6_SADDR_RULE_ORCHID:
1057                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1058                  *          non-ORCHID vs non-ORCHID
1059                  */
1060                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1061                         ipv6_addr_orchid(dst->addr));
1062                 break;
1063         case IPV6_SADDR_RULE_PREFIX:
1064                 /* Rule 8: Use longest matching prefix */
1065                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1066                                                        dst->addr);
1067                 break;
1068         default:
1069                 ret = 0;
1070         }
1071
1072         if (ret)
1073                 __set_bit(i, score->scorebits);
1074         score->rule = i;
1075 out:
1076         return ret;
1077 }
1078
1079 int ipv6_dev_get_saddr(struct net_device *dst_dev,
1080                        const struct in6_addr *daddr, unsigned int prefs,
1081                        struct in6_addr *saddr)
1082 {
1083         struct ipv6_saddr_score scores[2],
1084                                 *score = &scores[0], *hiscore = &scores[1];
1085         struct net *net = dev_net(dst_dev);
1086         struct ipv6_saddr_dst dst;
1087         struct net_device *dev;
1088         int dst_type;
1089
1090         dst_type = __ipv6_addr_type(daddr);
1091         dst.addr = daddr;
1092         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1093         dst.scope = __ipv6_addr_src_scope(dst_type);
1094         dst.label = ipv6_addr_label(daddr, dst_type, dst.ifindex);
1095         dst.prefs = prefs;
1096
1097         hiscore->rule = -1;
1098         hiscore->ifa = NULL;
1099
1100         read_lock(&dev_base_lock);
1101         rcu_read_lock();
1102
1103         for_each_netdev(net, dev) {
1104                 struct inet6_dev *idev;
1105
1106                 /* Candidate Source Address (section 4)
1107                  *  - multicast and link-local destination address,
1108                  *    the set of candidate source address MUST only
1109                  *    include addresses assigned to interfaces
1110                  *    belonging to the same link as the outgoing
1111                  *    interface.
1112                  * (- For site-local destination addresses, the
1113                  *    set of candidate source addresses MUST only
1114                  *    include addresses assigned to interfaces
1115                  *    belonging to the same site as the outgoing
1116                  *    interface.)
1117                  */
1118                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1119                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1120                     dst.ifindex && dev->ifindex != dst.ifindex)
1121                         continue;
1122
1123                 idev = __in6_dev_get(dev);
1124                 if (!idev)
1125                         continue;
1126
1127                 read_lock_bh(&idev->lock);
1128                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1129                         int i;
1130
1131                         /*
1132                          * - Tentative Address (RFC2462 section 5.4)
1133                          *  - A tentative address is not considered
1134                          *    "assigned to an interface" in the traditional
1135                          *    sense, unless it is also flagged as optimistic.
1136                          * - Candidate Source Address (section 4)
1137                          *  - In any case, anycast addresses, multicast
1138                          *    addresses, and the unspecified address MUST
1139                          *    NOT be included in a candidate set.
1140                          */
1141                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1142                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1143                                 continue;
1144
1145                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1146
1147                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1148                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1149                                 LIMIT_NETDEBUG(KERN_DEBUG
1150                                                "ADDRCONF: unspecified / multicast address "
1151                                                "assigned as unicast address on %s",
1152                                                dev->name);
1153                                 continue;
1154                         }
1155
1156                         score->rule = -1;
1157                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1158
1159                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1160                                 int minihiscore, miniscore;
1161
1162                                 minihiscore = ipv6_get_saddr_eval(hiscore, &dst, i);
1163                                 miniscore = ipv6_get_saddr_eval(score, &dst, i);
1164
1165                                 if (minihiscore > miniscore) {
1166                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1167                                             score->scopedist > 0) {
1168                                                 /*
1169                                                  * special case:
1170                                                  * each remaining entry
1171                                                  * has too small (not enough)
1172                                                  * scope, because ifa entries
1173                                                  * are sorted by their scope
1174                                                  * values.
1175                                                  */
1176                                                 goto try_nextdev;
1177                                         }
1178                                         break;
1179                                 } else if (minihiscore < miniscore) {
1180                                         struct ipv6_saddr_score *tmp;
1181
1182                                         if (hiscore->ifa)
1183                                                 in6_ifa_put(hiscore->ifa);
1184
1185                                         in6_ifa_hold(score->ifa);
1186
1187                                         tmp = hiscore;
1188                                         hiscore = score;
1189                                         score = tmp;
1190
1191                                         /* restore our iterator */
1192                                         score->ifa = hiscore->ifa;
1193
1194                                         break;
1195                                 }
1196                         }
1197                 }
1198 try_nextdev:
1199                 read_unlock_bh(&idev->lock);
1200         }
1201         rcu_read_unlock();
1202         read_unlock(&dev_base_lock);
1203
1204         if (!hiscore->ifa)
1205                 return -EADDRNOTAVAIL;
1206
1207         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1208         in6_ifa_put(hiscore->ifa);
1209         return 0;
1210 }
1211
1212 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1213
1214 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1215                     unsigned char banned_flags)
1216 {
1217         struct inet6_dev *idev;
1218         int err = -EADDRNOTAVAIL;
1219
1220         rcu_read_lock();
1221         if ((idev = __in6_dev_get(dev)) != NULL) {
1222                 struct inet6_ifaddr *ifp;
1223
1224                 read_lock_bh(&idev->lock);
1225                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1226                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1227                                 ipv6_addr_copy(addr, &ifp->addr);
1228                                 err = 0;
1229                                 break;
1230                         }
1231                 }
1232                 read_unlock_bh(&idev->lock);
1233         }
1234         rcu_read_unlock();
1235         return err;
1236 }
1237
1238 static int ipv6_count_addresses(struct inet6_dev *idev)
1239 {
1240         int cnt = 0;
1241         struct inet6_ifaddr *ifp;
1242
1243         read_lock_bh(&idev->lock);
1244         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1245                 cnt++;
1246         read_unlock_bh(&idev->lock);
1247         return cnt;
1248 }
1249
1250 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1251                   struct net_device *dev, int strict)
1252 {
1253         struct inet6_ifaddr * ifp;
1254         u8 hash = ipv6_addr_hash(addr);
1255
1256         read_lock_bh(&addrconf_hash_lock);
1257         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1258                 if (!net_eq(dev_net(ifp->idev->dev), net))
1259                         continue;
1260                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1261                     !(ifp->flags&IFA_F_TENTATIVE)) {
1262                         if (dev == NULL || ifp->idev->dev == dev ||
1263                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1264                                 break;
1265                 }
1266         }
1267         read_unlock_bh(&addrconf_hash_lock);
1268         return ifp != NULL;
1269 }
1270 EXPORT_SYMBOL(ipv6_chk_addr);
1271
1272 static
1273 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1274                        struct net_device *dev)
1275 {
1276         struct inet6_ifaddr * ifp;
1277         u8 hash = ipv6_addr_hash(addr);
1278
1279         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1280                 if (!net_eq(dev_net(ifp->idev->dev), net))
1281                         continue;
1282                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1283                         if (dev == NULL || ifp->idev->dev == dev)
1284                                 break;
1285                 }
1286         }
1287         return ifp != NULL;
1288 }
1289
1290 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1291 {
1292         struct inet6_dev *idev;
1293         struct inet6_ifaddr *ifa;
1294         int     onlink;
1295
1296         onlink = 0;
1297         rcu_read_lock();
1298         idev = __in6_dev_get(dev);
1299         if (idev) {
1300                 read_lock_bh(&idev->lock);
1301                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1302                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1303                                                    ifa->prefix_len);
1304                         if (onlink)
1305                                 break;
1306                 }
1307                 read_unlock_bh(&idev->lock);
1308         }
1309         rcu_read_unlock();
1310         return onlink;
1311 }
1312
1313 EXPORT_SYMBOL(ipv6_chk_prefix);
1314
1315 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1316                                      struct net_device *dev, int strict)
1317 {
1318         struct inet6_ifaddr * ifp;
1319         u8 hash = ipv6_addr_hash(addr);
1320
1321         read_lock_bh(&addrconf_hash_lock);
1322         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1323                 if (!net_eq(dev_net(ifp->idev->dev), net))
1324                         continue;
1325                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1326                         if (dev == NULL || ifp->idev->dev == dev ||
1327                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1328                                 in6_ifa_hold(ifp);
1329                                 break;
1330                         }
1331                 }
1332         }
1333         read_unlock_bh(&addrconf_hash_lock);
1334
1335         return ifp;
1336 }
1337
1338 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1339 {
1340         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1341         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1342         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1343         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1344         int sk_ipv6only = ipv6_only_sock(sk);
1345         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1346         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1347         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1348
1349         if (!sk2_rcv_saddr && !sk_ipv6only)
1350                 return 1;
1351
1352         if (addr_type2 == IPV6_ADDR_ANY &&
1353             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1354                 return 1;
1355
1356         if (addr_type == IPV6_ADDR_ANY &&
1357             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1358                 return 1;
1359
1360         if (sk2_rcv_saddr6 &&
1361             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1362                 return 1;
1363
1364         if (addr_type == IPV6_ADDR_MAPPED &&
1365             !sk2_ipv6only &&
1366             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1367                 return 1;
1368
1369         return 0;
1370 }
1371
1372 /* Gets referenced address, destroys ifaddr */
1373
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1375 {
1376         if (ifp->flags&IFA_F_PERMANENT) {
1377                 spin_lock_bh(&ifp->lock);
1378                 addrconf_del_timer(ifp);
1379                 ifp->flags |= IFA_F_TENTATIVE;
1380                 spin_unlock_bh(&ifp->lock);
1381                 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383         } else if (ifp->flags&IFA_F_TEMPORARY) {
1384                 struct inet6_ifaddr *ifpub;
1385                 spin_lock_bh(&ifp->lock);
1386                 ifpub = ifp->ifpub;
1387                 if (ifpub) {
1388                         in6_ifa_hold(ifpub);
1389                         spin_unlock_bh(&ifp->lock);
1390                         ipv6_create_tempaddr(ifpub, ifp);
1391                         in6_ifa_put(ifpub);
1392                 } else {
1393                         spin_unlock_bh(&ifp->lock);
1394                 }
1395                 ipv6_del_addr(ifp);
1396 #endif
1397         } else
1398                 ipv6_del_addr(ifp);
1399 }
1400
1401 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1402 {
1403         if (net_ratelimit())
1404                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1405         addrconf_dad_stop(ifp);
1406 }
1407
1408 /* Join to solicited addr multicast group. */
1409
1410 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1411 {
1412         struct in6_addr maddr;
1413
1414         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1415                 return;
1416
1417         addrconf_addr_solict_mult(addr, &maddr);
1418         ipv6_dev_mc_inc(dev, &maddr);
1419 }
1420
1421 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1422 {
1423         struct in6_addr maddr;
1424
1425         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1426                 return;
1427
1428         addrconf_addr_solict_mult(addr, &maddr);
1429         __ipv6_dev_mc_dec(idev, &maddr);
1430 }
1431
1432 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1433 {
1434         struct in6_addr addr;
1435         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1436         if (ipv6_addr_any(&addr))
1437                 return;
1438         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1439 }
1440
1441 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1442 {
1443         struct in6_addr addr;
1444         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1445         if (ipv6_addr_any(&addr))
1446                 return;
1447         __ipv6_dev_ac_dec(ifp->idev, &addr);
1448 }
1449
1450 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1451 {
1452         if (dev->addr_len != ETH_ALEN)
1453                 return -1;
1454         memcpy(eui, dev->dev_addr, 3);
1455         memcpy(eui + 5, dev->dev_addr + 3, 3);
1456
1457         /*
1458          * The zSeries OSA network cards can be shared among various
1459          * OS instances, but the OSA cards have only one MAC address.
1460          * This leads to duplicate address conflicts in conjunction
1461          * with IPv6 if more than one instance uses the same card.
1462          *
1463          * The driver for these cards can deliver a unique 16-bit
1464          * identifier for each instance sharing the same card.  It is
1465          * placed instead of 0xFFFE in the interface identifier.  The
1466          * "u" bit of the interface identifier is not inverted in this
1467          * case.  Hence the resulting interface identifier has local
1468          * scope according to RFC2373.
1469          */
1470         if (dev->dev_id) {
1471                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1472                 eui[4] = dev->dev_id & 0xFF;
1473         } else {
1474                 eui[3] = 0xFF;
1475                 eui[4] = 0xFE;
1476                 eui[0] ^= 2;
1477         }
1478         return 0;
1479 }
1480
1481 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1482 {
1483         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1484         if (dev->addr_len != ARCNET_ALEN)
1485                 return -1;
1486         memset(eui, 0, 7);
1487         eui[7] = *(u8*)dev->dev_addr;
1488         return 0;
1489 }
1490
1491 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1492 {
1493         if (dev->addr_len != INFINIBAND_ALEN)
1494                 return -1;
1495         memcpy(eui, dev->dev_addr + 12, 8);
1496         eui[0] |= 2;
1497         return 0;
1498 }
1499
1500 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1501 {
1502         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1503                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1504                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1505                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1506                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1507                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1508         eui[1] = 0;
1509         eui[2] = 0x5E;
1510         eui[3] = 0xFE;
1511         memcpy(eui + 4, &addr, 4);
1512         return 0;
1513 }
1514 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1515
1516 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1517 {
1518         if (dev->priv_flags & IFF_ISATAP)
1519                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1520         return -1;
1521 }
1522
1523 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1524 {
1525         switch (dev->type) {
1526         case ARPHRD_ETHER:
1527         case ARPHRD_FDDI:
1528         case ARPHRD_IEEE802_TR:
1529                 return addrconf_ifid_eui48(eui, dev);
1530         case ARPHRD_ARCNET:
1531                 return addrconf_ifid_arcnet(eui, dev);
1532         case ARPHRD_INFINIBAND:
1533                 return addrconf_ifid_infiniband(eui, dev);
1534         case ARPHRD_SIT:
1535                 return addrconf_ifid_sit(eui, dev);
1536         }
1537         return -1;
1538 }
1539
1540 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1541 {
1542         int err = -1;
1543         struct inet6_ifaddr *ifp;
1544
1545         read_lock_bh(&idev->lock);
1546         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1547                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1548                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1549                         err = 0;
1550                         break;
1551                 }
1552         }
1553         read_unlock_bh(&idev->lock);
1554         return err;
1555 }
1556
1557 #ifdef CONFIG_IPV6_PRIVACY
1558 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1559 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1560 {
1561 regen:
1562         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1563         idev->rndid[0] &= ~0x02;
1564
1565         /*
1566          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1567          * check if generated address is not inappropriate
1568          *
1569          *  - Reserved subnet anycast (RFC 2526)
1570          *      11111101 11....11 1xxxxxxx
1571          *  - ISATAP (RFC4214) 6.1
1572          *      00-00-5E-FE-xx-xx-xx-xx
1573          *  - value 0
1574          *  - XXX: already assigned to an address on the device
1575          */
1576         if (idev->rndid[0] == 0xfd &&
1577             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1578             (idev->rndid[7]&0x80))
1579                 goto regen;
1580         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1581                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1582                         goto regen;
1583                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1584                         goto regen;
1585         }
1586
1587         return 0;
1588 }
1589
1590 static void ipv6_regen_rndid(unsigned long data)
1591 {
1592         struct inet6_dev *idev = (struct inet6_dev *) data;
1593         unsigned long expires;
1594
1595         rcu_read_lock_bh();
1596         write_lock_bh(&idev->lock);
1597
1598         if (idev->dead)
1599                 goto out;
1600
1601         if (__ipv6_regen_rndid(idev) < 0)
1602                 goto out;
1603
1604         expires = jiffies +
1605                 idev->cnf.temp_prefered_lft * HZ -
1606                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1607         if (time_before(expires, jiffies)) {
1608                 printk(KERN_WARNING
1609                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1610                         idev->dev->name);
1611                 goto out;
1612         }
1613
1614         if (!mod_timer(&idev->regen_timer, expires))
1615                 in6_dev_hold(idev);
1616
1617 out:
1618         write_unlock_bh(&idev->lock);
1619         rcu_read_unlock_bh();
1620         in6_dev_put(idev);
1621 }
1622
1623 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1624         int ret = 0;
1625
1626         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1627                 ret = __ipv6_regen_rndid(idev);
1628         return ret;
1629 }
1630 #endif
1631
1632 /*
1633  *      Add prefix route.
1634  */
1635
1636 static void
1637 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1638                       unsigned long expires, u32 flags)
1639 {
1640         struct fib6_config cfg = {
1641                 .fc_table = RT6_TABLE_PREFIX,
1642                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1643                 .fc_ifindex = dev->ifindex,
1644                 .fc_expires = expires,
1645                 .fc_dst_len = plen,
1646                 .fc_flags = RTF_UP | flags,
1647                 .fc_nlinfo.nl_net = dev_net(dev),
1648         };
1649
1650         ipv6_addr_copy(&cfg.fc_dst, pfx);
1651
1652         /* Prevent useless cloning on PtP SIT.
1653            This thing is done here expecting that the whole
1654            class of non-broadcast devices need not cloning.
1655          */
1656 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1657         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1658                 cfg.fc_flags |= RTF_NONEXTHOP;
1659 #endif
1660
1661         ip6_route_add(&cfg);
1662 }
1663
1664 /* Create "default" multicast route to the interface */
1665
1666 static void addrconf_add_mroute(struct net_device *dev)
1667 {
1668         struct fib6_config cfg = {
1669                 .fc_table = RT6_TABLE_LOCAL,
1670                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1671                 .fc_ifindex = dev->ifindex,
1672                 .fc_dst_len = 8,
1673                 .fc_flags = RTF_UP,
1674                 .fc_nlinfo.nl_net = dev_net(dev),
1675         };
1676
1677         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1678
1679         ip6_route_add(&cfg);
1680 }
1681
1682 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1683 static void sit_route_add(struct net_device *dev)
1684 {
1685         struct fib6_config cfg = {
1686                 .fc_table = RT6_TABLE_MAIN,
1687                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1688                 .fc_ifindex = dev->ifindex,
1689                 .fc_dst_len = 96,
1690                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1691                 .fc_nlinfo.nl_net = dev_net(dev),
1692         };
1693
1694         /* prefix length - 96 bits "::d.d.d.d" */
1695         ip6_route_add(&cfg);
1696 }
1697 #endif
1698
1699 static void addrconf_add_lroute(struct net_device *dev)
1700 {
1701         struct in6_addr addr;
1702
1703         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1704         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1705 }
1706
1707 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1708 {
1709         struct inet6_dev *idev;
1710
1711         ASSERT_RTNL();
1712
1713         if ((idev = ipv6_find_idev(dev)) == NULL)
1714                 return NULL;
1715
1716         /* Add default multicast route */
1717         addrconf_add_mroute(dev);
1718
1719         /* Add link local route */
1720         addrconf_add_lroute(dev);
1721         return idev;
1722 }
1723
1724 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1725 {
1726         struct prefix_info *pinfo;
1727         __u32 valid_lft;
1728         __u32 prefered_lft;
1729         int addr_type;
1730         struct inet6_dev *in6_dev;
1731
1732         pinfo = (struct prefix_info *) opt;
1733
1734         if (len < sizeof(struct prefix_info)) {
1735                 ADBG(("addrconf: prefix option too short\n"));
1736                 return;
1737         }
1738
1739         /*
1740          *      Validation checks ([ADDRCONF], page 19)
1741          */
1742
1743         addr_type = ipv6_addr_type(&pinfo->prefix);
1744
1745         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1746                 return;
1747
1748         valid_lft = ntohl(pinfo->valid);
1749         prefered_lft = ntohl(pinfo->prefered);
1750
1751         if (prefered_lft > valid_lft) {
1752                 if (net_ratelimit())
1753                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1754                 return;
1755         }
1756
1757         in6_dev = in6_dev_get(dev);
1758
1759         if (in6_dev == NULL) {
1760                 if (net_ratelimit())
1761                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1762                 return;
1763         }
1764
1765         /*
1766          *      Two things going on here:
1767          *      1) Add routes for on-link prefixes
1768          *      2) Configure prefixes with the auto flag set
1769          */
1770
1771         if (pinfo->onlink) {
1772                 struct rt6_info *rt;
1773                 unsigned long rt_expires;
1774
1775                 /* Avoid arithmetic overflow. Really, we could
1776                  * save rt_expires in seconds, likely valid_lft,
1777                  * but it would require division in fib gc, that it
1778                  * not good.
1779                  */
1780                 if (HZ > USER_HZ)
1781                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1782                 else
1783                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1784
1785                 if (addrconf_finite_timeout(rt_expires))
1786                         rt_expires *= HZ;
1787
1788                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1789                                 dev->ifindex, 1);
1790
1791                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1792                         /* Autoconf prefix route */
1793                         if (valid_lft == 0) {
1794                                 ip6_del_rt(rt);
1795                                 rt = NULL;
1796                         } else if (addrconf_finite_timeout(rt_expires)) {
1797                                 /* not infinity */
1798                                 rt->rt6i_expires = jiffies + rt_expires;
1799                                 rt->rt6i_flags |= RTF_EXPIRES;
1800                         } else {
1801                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1802                                 rt->rt6i_expires = 0;
1803                         }
1804                 } else if (valid_lft) {
1805                         clock_t expires = 0;
1806                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1807                         if (addrconf_finite_timeout(rt_expires)) {
1808                                 /* not infinity */
1809                                 flags |= RTF_EXPIRES;
1810                                 expires = jiffies_to_clock_t(rt_expires);
1811                         }
1812                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1813                                               dev, expires, flags);
1814                 }
1815                 if (rt)
1816                         dst_release(&rt->u.dst);
1817         }
1818
1819         /* Try to figure out our local address for this prefix */
1820
1821         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1822                 struct inet6_ifaddr * ifp;
1823                 struct in6_addr addr;
1824                 int create = 0, update_lft = 0;
1825
1826                 if (pinfo->prefix_len == 64) {
1827                         memcpy(&addr, &pinfo->prefix, 8);
1828                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1829                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1830                                 in6_dev_put(in6_dev);
1831                                 return;
1832                         }
1833                         goto ok;
1834                 }
1835                 if (net_ratelimit())
1836                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1837                                pinfo->prefix_len);
1838                 in6_dev_put(in6_dev);
1839                 return;
1840
1841 ok:
1842
1843                 ifp = ipv6_get_ifaddr(dev_net(dev), &addr, dev, 1);
1844
1845                 if (ifp == NULL && valid_lft) {
1846                         int max_addresses = in6_dev->cnf.max_addresses;
1847                         u32 addr_flags = 0;
1848
1849 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1850                         if (in6_dev->cnf.optimistic_dad &&
1851                             !ipv6_devconf.forwarding)
1852                                 addr_flags = IFA_F_OPTIMISTIC;
1853 #endif
1854
1855                         /* Do not allow to create too much of autoconfigured
1856                          * addresses; this would be too easy way to crash kernel.
1857                          */
1858                         if (!max_addresses ||
1859                             ipv6_count_addresses(in6_dev) < max_addresses)
1860                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1861                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1862                                                     addr_flags);
1863
1864                         if (!ifp || IS_ERR(ifp)) {
1865                                 in6_dev_put(in6_dev);
1866                                 return;
1867                         }
1868
1869                         update_lft = create = 1;
1870                         ifp->cstamp = jiffies;
1871                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1872                 }
1873
1874                 if (ifp) {
1875                         int flags;
1876                         unsigned long now;
1877 #ifdef CONFIG_IPV6_PRIVACY
1878                         struct inet6_ifaddr *ift;
1879 #endif
1880                         u32 stored_lft;
1881
1882                         /* update lifetime (RFC2462 5.5.3 e) */
1883                         spin_lock(&ifp->lock);
1884                         now = jiffies;
1885                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1886                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1887                         else
1888                                 stored_lft = 0;
1889                         if (!update_lft && stored_lft) {
1890                                 if (valid_lft > MIN_VALID_LIFETIME ||
1891                                     valid_lft > stored_lft)
1892                                         update_lft = 1;
1893                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1894                                         /* valid_lft <= stored_lft is always true */
1895                                         /* XXX: IPsec */
1896                                         update_lft = 0;
1897                                 } else {
1898                                         valid_lft = MIN_VALID_LIFETIME;
1899                                         if (valid_lft < prefered_lft)
1900                                                 prefered_lft = valid_lft;
1901                                         update_lft = 1;
1902                                 }
1903                         }
1904
1905                         if (update_lft) {
1906                                 ifp->valid_lft = valid_lft;
1907                                 ifp->prefered_lft = prefered_lft;
1908                                 ifp->tstamp = now;
1909                                 flags = ifp->flags;
1910                                 ifp->flags &= ~IFA_F_DEPRECATED;
1911                                 spin_unlock(&ifp->lock);
1912
1913                                 if (!(flags&IFA_F_TENTATIVE))
1914                                         ipv6_ifa_notify(0, ifp);
1915                         } else
1916                                 spin_unlock(&ifp->lock);
1917
1918 #ifdef CONFIG_IPV6_PRIVACY
1919                         read_lock_bh(&in6_dev->lock);
1920                         /* update all temporary addresses in the list */
1921                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1922                                 /*
1923                                  * When adjusting the lifetimes of an existing
1924                                  * temporary address, only lower the lifetimes.
1925                                  * Implementations must not increase the
1926                                  * lifetimes of an existing temporary address
1927                                  * when processing a Prefix Information Option.
1928                                  */
1929                                 if (ifp != ift->ifpub)
1930                                         continue;
1931
1932                                 spin_lock(&ift->lock);
1933                                 flags = ift->flags;
1934                                 if (ift->valid_lft > valid_lft &&
1935                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1936                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1937                                 if (ift->prefered_lft > prefered_lft &&
1938                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1939                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1940                                 spin_unlock(&ift->lock);
1941                                 if (!(flags&IFA_F_TENTATIVE))
1942                                         ipv6_ifa_notify(0, ift);
1943                         }
1944
1945                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1946                                 /*
1947                                  * When a new public address is created as described in [ADDRCONF],
1948                                  * also create a new temporary address.
1949                                  */
1950                                 read_unlock_bh(&in6_dev->lock);
1951                                 ipv6_create_tempaddr(ifp, NULL);
1952                         } else {
1953                                 read_unlock_bh(&in6_dev->lock);
1954                         }
1955 #endif
1956                         in6_ifa_put(ifp);
1957                         addrconf_verify(0);
1958                 }
1959         }
1960         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1961         in6_dev_put(in6_dev);
1962 }
1963
1964 /*
1965  *      Set destination address.
1966  *      Special case for SIT interfaces where we create a new "virtual"
1967  *      device.
1968  */
1969 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1970 {
1971         struct in6_ifreq ireq;
1972         struct net_device *dev;
1973         int err = -EINVAL;
1974
1975         rtnl_lock();
1976
1977         err = -EFAULT;
1978         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1979                 goto err_exit;
1980
1981         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
1982
1983         err = -ENODEV;
1984         if (dev == NULL)
1985                 goto err_exit;
1986
1987 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1988         if (dev->type == ARPHRD_SIT) {
1989                 struct ifreq ifr;
1990                 mm_segment_t    oldfs;
1991                 struct ip_tunnel_parm p;
1992
1993                 err = -EADDRNOTAVAIL;
1994                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1995                         goto err_exit;
1996
1997                 memset(&p, 0, sizeof(p));
1998                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1999                 p.iph.saddr = 0;
2000                 p.iph.version = 4;
2001                 p.iph.ihl = 5;
2002                 p.iph.protocol = IPPROTO_IPV6;
2003                 p.iph.ttl = 64;
2004                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2005
2006                 oldfs = get_fs(); set_fs(KERNEL_DS);
2007                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2008                 set_fs(oldfs);
2009
2010                 if (err == 0) {
2011                         err = -ENOBUFS;
2012                         dev = __dev_get_by_name(net, p.name);
2013                         if (!dev)
2014                                 goto err_exit;
2015                         err = dev_open(dev);
2016                 }
2017         }
2018 #endif
2019
2020 err_exit:
2021         rtnl_unlock();
2022         return err;
2023 }
2024
2025 /*
2026  *      Manual configuration of address on an interface
2027  */
2028 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2029                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2030                           __u32 valid_lft)
2031 {
2032         struct inet6_ifaddr *ifp;
2033         struct inet6_dev *idev;
2034         struct net_device *dev;
2035         int scope;
2036         u32 flags;
2037         clock_t expires;
2038         unsigned long timeout;
2039
2040         ASSERT_RTNL();
2041
2042         if (plen > 128)
2043                 return -EINVAL;
2044
2045         /* check the lifetime */
2046         if (!valid_lft || prefered_lft > valid_lft)
2047                 return -EINVAL;
2048
2049         dev = __dev_get_by_index(net, ifindex);
2050         if (!dev)
2051                 return -ENODEV;
2052
2053         if ((idev = addrconf_add_dev(dev)) == NULL)
2054                 return -ENOBUFS;
2055
2056         scope = ipv6_addr_scope(pfx);
2057
2058         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2059         if (addrconf_finite_timeout(timeout)) {
2060                 expires = jiffies_to_clock_t(timeout * HZ);
2061                 valid_lft = timeout;
2062                 flags = RTF_EXPIRES;
2063         } else {
2064                 expires = 0;
2065                 flags = 0;
2066                 ifa_flags |= IFA_F_PERMANENT;
2067         }
2068
2069         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2070         if (addrconf_finite_timeout(timeout)) {
2071                 if (timeout == 0)
2072                         ifa_flags |= IFA_F_DEPRECATED;
2073                 prefered_lft = timeout;
2074         }
2075
2076         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2077
2078         if (!IS_ERR(ifp)) {
2079                 spin_lock_bh(&ifp->lock);
2080                 ifp->valid_lft = valid_lft;
2081                 ifp->prefered_lft = prefered_lft;
2082                 ifp->tstamp = jiffies;
2083                 spin_unlock_bh(&ifp->lock);
2084
2085                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2086                                       expires, flags);
2087                 /*
2088                  * Note that section 3.1 of RFC 4429 indicates
2089                  * that the Optimistic flag should not be set for
2090                  * manually configured addresses
2091                  */
2092                 addrconf_dad_start(ifp, 0);
2093                 in6_ifa_put(ifp);
2094                 addrconf_verify(0);
2095                 return 0;
2096         }
2097
2098         return PTR_ERR(ifp);
2099 }
2100
2101 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2102                           unsigned int plen)
2103 {
2104         struct inet6_ifaddr *ifp;
2105         struct inet6_dev *idev;
2106         struct net_device *dev;
2107
2108         if (plen > 128)
2109                 return -EINVAL;
2110
2111         dev = __dev_get_by_index(net, ifindex);
2112         if (!dev)
2113                 return -ENODEV;
2114
2115         if ((idev = __in6_dev_get(dev)) == NULL)
2116                 return -ENXIO;
2117
2118         read_lock_bh(&idev->lock);
2119         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2120                 if (ifp->prefix_len == plen &&
2121                     ipv6_addr_equal(pfx, &ifp->addr)) {
2122                         in6_ifa_hold(ifp);
2123                         read_unlock_bh(&idev->lock);
2124
2125                         ipv6_del_addr(ifp);
2126
2127                         /* If the last address is deleted administratively,
2128                            disable IPv6 on this interface.
2129                          */
2130                         if (idev->addr_list == NULL)
2131                                 addrconf_ifdown(idev->dev, 1);
2132                         return 0;
2133                 }
2134         }
2135         read_unlock_bh(&idev->lock);
2136         return -EADDRNOTAVAIL;
2137 }
2138
2139
2140 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2141 {
2142         struct in6_ifreq ireq;
2143         int err;
2144
2145         if (!capable(CAP_NET_ADMIN))
2146                 return -EPERM;
2147
2148         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2149                 return -EFAULT;
2150
2151         rtnl_lock();
2152         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2153                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2154                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2155         rtnl_unlock();
2156         return err;
2157 }
2158
2159 int addrconf_del_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_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2172                              ireq.ifr6_prefixlen);
2173         rtnl_unlock();
2174         return err;
2175 }
2176
2177 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2178 static void sit_add_v4_addrs(struct inet6_dev *idev)
2179 {
2180         struct inet6_ifaddr * ifp;
2181         struct in6_addr addr;
2182         struct net_device *dev;
2183         struct net *net = dev_net(idev->dev);
2184         int scope;
2185
2186         ASSERT_RTNL();
2187
2188         memset(&addr, 0, sizeof(struct in6_addr));
2189         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2190
2191         if (idev->dev->flags&IFF_POINTOPOINT) {
2192                 addr.s6_addr32[0] = htonl(0xfe800000);
2193                 scope = IFA_LINK;
2194         } else {
2195                 scope = IPV6_ADDR_COMPATv4;
2196         }
2197
2198         if (addr.s6_addr32[3]) {
2199                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2200                 if (!IS_ERR(ifp)) {
2201                         spin_lock_bh(&ifp->lock);
2202                         ifp->flags &= ~IFA_F_TENTATIVE;
2203                         spin_unlock_bh(&ifp->lock);
2204                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2205                         in6_ifa_put(ifp);
2206                 }
2207                 return;
2208         }
2209
2210         for_each_netdev(net, dev) {
2211                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2212                 if (in_dev && (dev->flags & IFF_UP)) {
2213                         struct in_ifaddr * ifa;
2214
2215                         int flag = scope;
2216
2217                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2218                                 int plen;
2219
2220                                 addr.s6_addr32[3] = ifa->ifa_local;
2221
2222                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2223                                         continue;
2224                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2225                                         if (idev->dev->flags&IFF_POINTOPOINT)
2226                                                 continue;
2227                                         flag |= IFA_HOST;
2228                                 }
2229                                 if (idev->dev->flags&IFF_POINTOPOINT)
2230                                         plen = 64;
2231                                 else
2232                                         plen = 96;
2233
2234                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2235                                                     IFA_F_PERMANENT);
2236                                 if (!IS_ERR(ifp)) {
2237                                         spin_lock_bh(&ifp->lock);
2238                                         ifp->flags &= ~IFA_F_TENTATIVE;
2239                                         spin_unlock_bh(&ifp->lock);
2240                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2241                                         in6_ifa_put(ifp);
2242                                 }
2243                         }
2244                 }
2245         }
2246 }
2247 #endif
2248
2249 static void init_loopback(struct net_device *dev)
2250 {
2251         struct inet6_dev  *idev;
2252         struct inet6_ifaddr * ifp;
2253
2254         /* ::1 */
2255
2256         ASSERT_RTNL();
2257
2258         if ((idev = ipv6_find_idev(dev)) == NULL) {
2259                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2260                 return;
2261         }
2262
2263         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2264         if (!IS_ERR(ifp)) {
2265                 spin_lock_bh(&ifp->lock);
2266                 ifp->flags &= ~IFA_F_TENTATIVE;
2267                 spin_unlock_bh(&ifp->lock);
2268                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2269                 in6_ifa_put(ifp);
2270         }
2271 }
2272
2273 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2274 {
2275         struct inet6_ifaddr * ifp;
2276         u32 addr_flags = IFA_F_PERMANENT;
2277
2278 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2279         if (idev->cnf.optimistic_dad &&
2280             !ipv6_devconf.forwarding)
2281                 addr_flags |= IFA_F_OPTIMISTIC;
2282 #endif
2283
2284
2285         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2286         if (!IS_ERR(ifp)) {
2287                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2288                 addrconf_dad_start(ifp, 0);
2289                 in6_ifa_put(ifp);
2290         }
2291 }
2292
2293 static void addrconf_dev_config(struct net_device *dev)
2294 {
2295         struct in6_addr addr;
2296         struct inet6_dev    * idev;
2297
2298         ASSERT_RTNL();
2299
2300         if ((dev->type != ARPHRD_ETHER) &&
2301             (dev->type != ARPHRD_FDDI) &&
2302             (dev->type != ARPHRD_IEEE802_TR) &&
2303             (dev->type != ARPHRD_ARCNET) &&
2304             (dev->type != ARPHRD_INFINIBAND)) {
2305                 /* Alas, we support only Ethernet autoconfiguration. */
2306                 return;
2307         }
2308
2309         idev = addrconf_add_dev(dev);
2310         if (idev == NULL)
2311                 return;
2312
2313         memset(&addr, 0, sizeof(struct in6_addr));
2314         addr.s6_addr32[0] = htonl(0xFE800000);
2315
2316         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2317                 addrconf_add_linklocal(idev, &addr);
2318 }
2319
2320 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2321 static void addrconf_sit_config(struct net_device *dev)
2322 {
2323         struct inet6_dev *idev;
2324
2325         ASSERT_RTNL();
2326
2327         /*
2328          * Configure the tunnel with one of our IPv4
2329          * addresses... we should configure all of
2330          * our v4 addrs in the tunnel
2331          */
2332
2333         if ((idev = ipv6_find_idev(dev)) == NULL) {
2334                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2335                 return;
2336         }
2337
2338         if (dev->priv_flags & IFF_ISATAP) {
2339                 struct in6_addr addr;
2340
2341                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2342                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2343                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2344                         addrconf_add_linklocal(idev, &addr);
2345                 return;
2346         }
2347
2348         sit_add_v4_addrs(idev);
2349
2350         if (dev->flags&IFF_POINTOPOINT) {
2351                 addrconf_add_mroute(dev);
2352                 addrconf_add_lroute(dev);
2353         } else
2354                 sit_route_add(dev);
2355 }
2356 #endif
2357
2358 static inline int
2359 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2360 {
2361         struct in6_addr lladdr;
2362
2363         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2364                 addrconf_add_linklocal(idev, &lladdr);
2365                 return 0;
2366         }
2367         return -1;
2368 }
2369
2370 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2371 {
2372         struct net_device *link_dev;
2373         struct net *net = dev_net(idev->dev);
2374
2375         /* first try to inherit the link-local address from the link device */
2376         if (idev->dev->iflink &&
2377             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2378                 if (!ipv6_inherit_linklocal(idev, link_dev))
2379                         return;
2380         }
2381         /* then try to inherit it from any device */
2382         for_each_netdev(net, link_dev) {
2383                 if (!ipv6_inherit_linklocal(idev, link_dev))
2384                         return;
2385         }
2386         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2387 }
2388
2389 /*
2390  * Autoconfigure tunnel with a link-local address so routing protocols,
2391  * DHCPv6, MLD etc. can be run over the virtual link
2392  */
2393
2394 static void addrconf_ip6_tnl_config(struct net_device *dev)
2395 {
2396         struct inet6_dev *idev;
2397
2398         ASSERT_RTNL();
2399
2400         if ((idev = addrconf_add_dev(dev)) == NULL) {
2401                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2402                 return;
2403         }
2404         ip6_tnl_add_linklocal(idev);
2405 }
2406
2407 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2408                            void * data)
2409 {
2410         struct net_device *dev = (struct net_device *) data;
2411         struct inet6_dev *idev = __in6_dev_get(dev);
2412         int run_pending = 0;
2413         int err;
2414
2415         switch(event) {
2416         case NETDEV_REGISTER:
2417                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2418                         idev = ipv6_add_dev(dev);
2419                         if (!idev)
2420                                 return notifier_from_errno(-ENOMEM);
2421                 }
2422                 break;
2423         case NETDEV_UP:
2424         case NETDEV_CHANGE:
2425                 if (dev->flags & IFF_SLAVE)
2426                         break;
2427
2428                 if (event == NETDEV_UP) {
2429                         if (!addrconf_qdisc_ok(dev)) {
2430                                 /* device is not ready yet. */
2431                                 printk(KERN_INFO
2432                                         "ADDRCONF(NETDEV_UP): %s: "
2433                                         "link is not ready\n",
2434                                         dev->name);
2435                                 break;
2436                         }
2437
2438                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2439                                 idev = ipv6_add_dev(dev);
2440
2441                         if (idev)
2442                                 idev->if_flags |= IF_READY;
2443                 } else {
2444                         if (!addrconf_qdisc_ok(dev)) {
2445                                 /* device is still not ready. */
2446                                 break;
2447                         }
2448
2449                         if (idev) {
2450                                 if (idev->if_flags & IF_READY) {
2451                                         /* device is already configured. */
2452                                         break;
2453                                 }
2454                                 idev->if_flags |= IF_READY;
2455                         }
2456
2457                         printk(KERN_INFO
2458                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2459                                         "link becomes ready\n",
2460                                         dev->name);
2461
2462                         run_pending = 1;
2463                 }
2464
2465                 switch(dev->type) {
2466 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2467                 case ARPHRD_SIT:
2468                         addrconf_sit_config(dev);
2469                         break;
2470 #endif
2471                 case ARPHRD_TUNNEL6:
2472                         addrconf_ip6_tnl_config(dev);
2473                         break;
2474                 case ARPHRD_LOOPBACK:
2475                         init_loopback(dev);
2476                         break;
2477
2478                 default:
2479                         addrconf_dev_config(dev);
2480                         break;
2481                 }
2482                 if (idev) {
2483                         if (run_pending)
2484                                 addrconf_dad_run(idev);
2485
2486                         /* If the MTU changed during the interface down, when the
2487                            interface up, the changed MTU must be reflected in the
2488                            idev as well as routers.
2489                          */
2490                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2491                                 rt6_mtu_change(dev, dev->mtu);
2492                                 idev->cnf.mtu6 = dev->mtu;
2493                         }
2494                         idev->tstamp = jiffies;
2495                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2496                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2497                            stop IPv6 on this interface.
2498                          */
2499                         if (dev->mtu < IPV6_MIN_MTU)
2500                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2501                 }
2502                 break;
2503
2504         case NETDEV_CHANGEMTU:
2505                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2506                         rt6_mtu_change(dev, dev->mtu);
2507                         idev->cnf.mtu6 = dev->mtu;
2508                         break;
2509                 }
2510
2511                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2512                         idev = ipv6_add_dev(dev);
2513                         if (idev)
2514                                 break;
2515                 }
2516
2517                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2518
2519         case NETDEV_DOWN:
2520         case NETDEV_UNREGISTER:
2521                 /*
2522                  *      Remove all addresses from this interface.
2523                  */
2524                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2525                 break;
2526
2527         case NETDEV_CHANGENAME:
2528                 if (idev) {
2529                         snmp6_unregister_dev(idev);
2530                         addrconf_sysctl_unregister(idev);
2531                         addrconf_sysctl_register(idev);
2532                         err = snmp6_register_dev(idev);
2533                         if (err)
2534                                 return notifier_from_errno(err);
2535                 }
2536                 break;
2537         }
2538
2539         return NOTIFY_OK;
2540 }
2541
2542 /*
2543  *      addrconf module should be notified of a device going up
2544  */
2545 static struct notifier_block ipv6_dev_notf = {
2546         .notifier_call = addrconf_notify,
2547         .priority = 0
2548 };
2549
2550 static int addrconf_ifdown(struct net_device *dev, int how)
2551 {
2552         struct inet6_dev *idev;
2553         struct inet6_ifaddr *ifa, **bifa;
2554         struct net *net = dev_net(dev);
2555         int i;
2556
2557         ASSERT_RTNL();
2558
2559         if ((dev->flags & IFF_LOOPBACK) && how == 1)
2560                 how = 0;
2561
2562         rt6_ifdown(net, dev);
2563         neigh_ifdown(&nd_tbl, dev);
2564
2565         idev = __in6_dev_get(dev);
2566         if (idev == NULL)
2567                 return -ENODEV;
2568
2569         /* Step 1: remove reference to ipv6 device from parent device.
2570                    Do not dev_put!
2571          */
2572         if (how) {
2573                 idev->dead = 1;
2574
2575                 /* protected by rtnl_lock */
2576                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2577
2578                 /* Step 1.5: remove snmp6 entry */
2579                 snmp6_unregister_dev(idev);
2580
2581         }
2582
2583         /* Step 2: clear hash table */
2584         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2585                 bifa = &inet6_addr_lst[i];
2586
2587                 write_lock_bh(&addrconf_hash_lock);
2588                 while ((ifa = *bifa) != NULL) {
2589                         if (ifa->idev == idev) {
2590                                 *bifa = ifa->lst_next;
2591                                 ifa->lst_next = NULL;
2592                                 addrconf_del_timer(ifa);
2593                                 in6_ifa_put(ifa);
2594                                 continue;
2595                         }
2596                         bifa = &ifa->lst_next;
2597                 }
2598                 write_unlock_bh(&addrconf_hash_lock);
2599         }
2600
2601         write_lock_bh(&idev->lock);
2602
2603         /* Step 3: clear flags for stateless addrconf */
2604         if (!how)
2605                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2606
2607         /* Step 4: clear address list */
2608 #ifdef CONFIG_IPV6_PRIVACY
2609         if (how && del_timer(&idev->regen_timer))
2610                 in6_dev_put(idev);
2611
2612         /* clear tempaddr list */
2613         while ((ifa = idev->tempaddr_list) != NULL) {
2614                 idev->tempaddr_list = ifa->tmp_next;
2615                 ifa->tmp_next = NULL;
2616                 ifa->dead = 1;
2617                 write_unlock_bh(&idev->lock);
2618                 spin_lock_bh(&ifa->lock);
2619
2620                 if (ifa->ifpub) {
2621                         in6_ifa_put(ifa->ifpub);
2622                         ifa->ifpub = NULL;
2623                 }
2624                 spin_unlock_bh(&ifa->lock);
2625                 in6_ifa_put(ifa);
2626                 write_lock_bh(&idev->lock);
2627         }
2628 #endif
2629         while ((ifa = idev->addr_list) != NULL) {
2630                 idev->addr_list = ifa->if_next;
2631                 ifa->if_next = NULL;
2632                 ifa->dead = 1;
2633                 addrconf_del_timer(ifa);
2634                 write_unlock_bh(&idev->lock);
2635
2636                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2637                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2638                 in6_ifa_put(ifa);
2639
2640                 write_lock_bh(&idev->lock);
2641         }
2642         write_unlock_bh(&idev->lock);
2643
2644         /* Step 5: Discard multicast list */
2645
2646         if (how)
2647                 ipv6_mc_destroy_dev(idev);
2648         else
2649                 ipv6_mc_down(idev);
2650
2651         idev->tstamp = jiffies;
2652
2653         /* Shot the device (if unregistered) */
2654
2655         if (how) {
2656                 addrconf_sysctl_unregister(idev);
2657                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2658                 neigh_ifdown(&nd_tbl, dev);
2659                 in6_dev_put(idev);
2660         }
2661         return 0;
2662 }
2663
2664 static void addrconf_rs_timer(unsigned long data)
2665 {
2666         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2667
2668         if (ifp->idev->cnf.forwarding)
2669                 goto out;
2670
2671         if (ifp->idev->if_flags & IF_RA_RCVD) {
2672                 /*
2673                  *      Announcement received after solicitation
2674                  *      was sent
2675                  */
2676                 goto out;
2677         }
2678
2679         spin_lock(&ifp->lock);
2680         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2681                 /* The wait after the last probe can be shorter */
2682                 addrconf_mod_timer(ifp, AC_RS,
2683                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2684                                    ifp->idev->cnf.rtr_solicit_delay :
2685                                    ifp->idev->cnf.rtr_solicit_interval);
2686                 spin_unlock(&ifp->lock);
2687
2688                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2689         } else {
2690                 spin_unlock(&ifp->lock);
2691                 /*
2692                  * Note: we do not support deprecated "all on-link"
2693                  * assumption any longer.
2694                  */
2695                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2696                        ifp->idev->dev->name);
2697         }
2698
2699 out:
2700         in6_ifa_put(ifp);
2701 }
2702
2703 /*
2704  *      Duplicate Address Detection
2705  */
2706 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2707 {
2708         unsigned long rand_num;
2709         struct inet6_dev *idev = ifp->idev;
2710
2711         if (ifp->flags & IFA_F_OPTIMISTIC)
2712                 rand_num = 0;
2713         else
2714                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2715
2716         ifp->probes = idev->cnf.dad_transmits;
2717         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2718 }
2719
2720 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2721 {
2722         struct inet6_dev *idev = ifp->idev;
2723         struct net_device *dev = idev->dev;
2724
2725         addrconf_join_solict(dev, &ifp->addr);
2726
2727         net_srandom(ifp->addr.s6_addr32[3]);
2728
2729         read_lock_bh(&idev->lock);
2730         if (ifp->dead)
2731                 goto out;
2732         spin_lock_bh(&ifp->lock);
2733
2734         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2735             !(ifp->flags&IFA_F_TENTATIVE) ||
2736             ifp->flags & IFA_F_NODAD) {
2737                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2738                 spin_unlock_bh(&ifp->lock);
2739                 read_unlock_bh(&idev->lock);
2740
2741                 addrconf_dad_completed(ifp);
2742                 return;
2743         }
2744
2745         if (!(idev->if_flags & IF_READY)) {
2746                 spin_unlock_bh(&ifp->lock);
2747                 read_unlock_bh(&idev->lock);
2748                 /*
2749                  * If the defice is not ready:
2750                  * - keep it tentative if it is a permanent address.
2751                  * - otherwise, kill it.
2752                  */
2753                 in6_ifa_hold(ifp);
2754                 addrconf_dad_stop(ifp);
2755                 return;
2756         }
2757
2758         /*
2759          * Optimistic nodes can start receiving
2760          * Frames right away
2761          */
2762         if(ifp->flags & IFA_F_OPTIMISTIC)
2763                 ip6_ins_rt(ifp->rt);
2764
2765         addrconf_dad_kick(ifp);
2766         spin_unlock_bh(&ifp->lock);
2767 out:
2768         read_unlock_bh(&idev->lock);
2769 }
2770
2771 static void addrconf_dad_timer(unsigned long data)
2772 {
2773         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2774         struct inet6_dev *idev = ifp->idev;
2775         struct in6_addr mcaddr;
2776
2777         read_lock_bh(&idev->lock);
2778         if (idev->dead) {
2779                 read_unlock_bh(&idev->lock);
2780                 goto out;
2781         }
2782         spin_lock_bh(&ifp->lock);
2783         if (ifp->probes == 0) {
2784                 /*
2785                  * DAD was successful
2786                  */
2787
2788                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2789                 spin_unlock_bh(&ifp->lock);
2790                 read_unlock_bh(&idev->lock);
2791
2792                 addrconf_dad_completed(ifp);
2793
2794                 goto out;
2795         }
2796
2797         ifp->probes--;
2798         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2799         spin_unlock_bh(&ifp->lock);
2800         read_unlock_bh(&idev->lock);
2801
2802         /* send a neighbour solicitation for our addr */
2803         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2804         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2805 out:
2806         in6_ifa_put(ifp);
2807 }
2808
2809 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2810 {
2811         struct net_device *     dev = ifp->idev->dev;
2812
2813         /*
2814          *      Configure the address for reception. Now it is valid.
2815          */
2816
2817         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2818
2819         /* If added prefix is link local and forwarding is off,
2820            start sending router solicitations.
2821          */
2822
2823         if (ifp->idev->cnf.forwarding == 0 &&
2824             ifp->idev->cnf.rtr_solicits > 0 &&
2825             (dev->flags&IFF_LOOPBACK) == 0 &&
2826             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2827                 /*
2828                  *      If a host as already performed a random delay
2829                  *      [...] as part of DAD [...] there is no need
2830                  *      to delay again before sending the first RS
2831                  */
2832                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2833
2834                 spin_lock_bh(&ifp->lock);
2835                 ifp->probes = 1;
2836                 ifp->idev->if_flags |= IF_RS_SENT;
2837                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2838                 spin_unlock_bh(&ifp->lock);
2839         }
2840 }
2841
2842 static void addrconf_dad_run(struct inet6_dev *idev) {
2843         struct inet6_ifaddr *ifp;
2844
2845         read_lock_bh(&idev->lock);
2846         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2847                 spin_lock_bh(&ifp->lock);
2848                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2849                         spin_unlock_bh(&ifp->lock);
2850                         continue;
2851                 }
2852                 spin_unlock_bh(&ifp->lock);
2853                 addrconf_dad_kick(ifp);
2854         }
2855         read_unlock_bh(&idev->lock);
2856 }
2857
2858 #ifdef CONFIG_PROC_FS
2859 struct if6_iter_state {
2860         struct seq_net_private p;
2861         int bucket;
2862 };
2863
2864 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2865 {
2866         struct inet6_ifaddr *ifa = NULL;
2867         struct if6_iter_state *state = seq->private;
2868         struct net *net = seq_file_net(seq);
2869
2870         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2871                 ifa = inet6_addr_lst[state->bucket];
2872
2873                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2874                         ifa = ifa->lst_next;
2875                 if (ifa)
2876                         break;
2877         }
2878         return ifa;
2879 }
2880
2881 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2882 {
2883         struct if6_iter_state *state = seq->private;
2884         struct net *net = seq_file_net(seq);
2885
2886         ifa = ifa->lst_next;
2887 try_again:
2888         if (ifa) {
2889                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2890                         ifa = ifa->lst_next;
2891                         goto try_again;
2892                 }
2893         }
2894
2895         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2896                 ifa = inet6_addr_lst[state->bucket];
2897                 goto try_again;
2898         }
2899
2900         return ifa;
2901 }
2902
2903 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2904 {
2905         struct inet6_ifaddr *ifa = if6_get_first(seq);
2906
2907         if (ifa)
2908                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2909                         --pos;
2910         return pos ? NULL : ifa;
2911 }
2912
2913 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2914         __acquires(addrconf_hash_lock)
2915 {
2916         read_lock_bh(&addrconf_hash_lock);
2917         return if6_get_idx(seq, *pos);
2918 }
2919
2920 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2921 {
2922         struct inet6_ifaddr *ifa;
2923
2924         ifa = if6_get_next(seq, v);
2925         ++*pos;
2926         return ifa;
2927 }
2928
2929 static void if6_seq_stop(struct seq_file *seq, void *v)
2930         __releases(addrconf_hash_lock)
2931 {
2932         read_unlock_bh(&addrconf_hash_lock);
2933 }
2934
2935 static int if6_seq_show(struct seq_file *seq, void *v)
2936 {
2937         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2938         seq_printf(seq,
2939                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2940                    NIP6(ifp->addr),
2941                    ifp->idev->dev->ifindex,
2942                    ifp->prefix_len,
2943                    ifp->scope,
2944                    ifp->flags,
2945                    ifp->idev->dev->name);
2946         return 0;
2947 }
2948
2949 static const struct seq_operations if6_seq_ops = {
2950         .start  = if6_seq_start,
2951         .next   = if6_seq_next,
2952         .show   = if6_seq_show,
2953         .stop   = if6_seq_stop,
2954 };
2955
2956 static int if6_seq_open(struct inode *inode, struct file *file)
2957 {
2958         return seq_open_net(inode, file, &if6_seq_ops,
2959                             sizeof(struct if6_iter_state));
2960 }
2961
2962 static const struct file_operations if6_fops = {
2963         .owner          = THIS_MODULE,
2964         .open           = if6_seq_open,
2965         .read           = seq_read,
2966         .llseek         = seq_lseek,
2967         .release        = seq_release_net,
2968 };
2969
2970 static int if6_proc_net_init(struct net *net)
2971 {
2972         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2973                 return -ENOMEM;
2974         return 0;
2975 }
2976
2977 static void if6_proc_net_exit(struct net *net)
2978 {
2979        proc_net_remove(net, "if_inet6");
2980 }
2981
2982 static struct pernet_operations if6_proc_net_ops = {
2983        .init = if6_proc_net_init,
2984        .exit = if6_proc_net_exit,
2985 };
2986
2987 int __init if6_proc_init(void)
2988 {
2989         return register_pernet_subsys(&if6_proc_net_ops);
2990 }
2991
2992 void if6_proc_exit(void)
2993 {
2994         unregister_pernet_subsys(&if6_proc_net_ops);
2995 }
2996 #endif  /* CONFIG_PROC_FS */
2997
2998 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
2999 /* Check if address is a home address configured on any interface. */
3000 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3001 {
3002         int ret = 0;
3003         struct inet6_ifaddr * ifp;
3004         u8 hash = ipv6_addr_hash(addr);
3005         read_lock_bh(&addrconf_hash_lock);
3006         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3007                 if (!net_eq(dev_net(ifp->idev->dev), net))
3008                         continue;
3009                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3010                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3011                         ret = 1;
3012                         break;
3013                 }
3014         }
3015         read_unlock_bh(&addrconf_hash_lock);
3016         return ret;
3017 }
3018 #endif
3019
3020 /*
3021  *      Periodic address status verification
3022  */
3023
3024 static void addrconf_verify(unsigned long foo)
3025 {
3026         struct inet6_ifaddr *ifp;
3027         unsigned long now, next;
3028         int i;
3029
3030         spin_lock_bh(&addrconf_verify_lock);
3031         now = jiffies;
3032         next = now + ADDR_CHECK_FREQUENCY;
3033
3034         del_timer(&addr_chk_timer);
3035
3036         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3037
3038 restart:
3039                 read_lock(&addrconf_hash_lock);
3040                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3041                         unsigned long age;
3042 #ifdef CONFIG_IPV6_PRIVACY
3043                         unsigned long regen_advance;
3044 #endif
3045
3046                         if (ifp->flags & IFA_F_PERMANENT)
3047                                 continue;
3048
3049                         spin_lock(&ifp->lock);
3050                         age = (now - ifp->tstamp) / HZ;
3051
3052 #ifdef CONFIG_IPV6_PRIVACY
3053                         regen_advance = ifp->idev->cnf.regen_max_retry *
3054                                         ifp->idev->cnf.dad_transmits *
3055                                         ifp->idev->nd_parms->retrans_time / HZ;
3056 #endif
3057
3058                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3059                             age >= ifp->valid_lft) {
3060                                 spin_unlock(&ifp->lock);
3061                                 in6_ifa_hold(ifp);
3062                                 read_unlock(&addrconf_hash_lock);
3063                                 ipv6_del_addr(ifp);
3064                                 goto restart;
3065                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3066                                 spin_unlock(&ifp->lock);
3067                                 continue;
3068                         } else if (age >= ifp->prefered_lft) {
3069                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3070                                 int deprecate = 0;
3071
3072                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3073                                         deprecate = 1;
3074                                         ifp->flags |= IFA_F_DEPRECATED;
3075                                 }
3076
3077                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3078                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3079
3080                                 spin_unlock(&ifp->lock);
3081
3082                                 if (deprecate) {
3083                                         in6_ifa_hold(ifp);
3084                                         read_unlock(&addrconf_hash_lock);
3085
3086                                         ipv6_ifa_notify(0, ifp);
3087                                         in6_ifa_put(ifp);
3088                                         goto restart;
3089                                 }
3090 #ifdef CONFIG_IPV6_PRIVACY
3091                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3092                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3093                                 if (age >= ifp->prefered_lft - regen_advance) {
3094                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3095                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3096                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3097                                         if (!ifp->regen_count && ifpub) {
3098                                                 ifp->regen_count++;
3099                                                 in6_ifa_hold(ifp);
3100                                                 in6_ifa_hold(ifpub);
3101                                                 spin_unlock(&ifp->lock);
3102                                                 read_unlock(&addrconf_hash_lock);
3103                                                 spin_lock(&ifpub->lock);
3104                                                 ifpub->regen_count = 0;
3105                                                 spin_unlock(&ifpub->lock);
3106                                                 ipv6_create_tempaddr(ifpub, ifp);
3107                                                 in6_ifa_put(ifpub);
3108                                                 in6_ifa_put(ifp);
3109                                                 goto restart;
3110                                         }
3111                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3112                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3113                                 spin_unlock(&ifp->lock);
3114 #endif
3115                         } else {
3116                                 /* ifp->prefered_lft <= ifp->valid_lft */
3117                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3118                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3119                                 spin_unlock(&ifp->lock);
3120                         }
3121                 }
3122                 read_unlock(&addrconf_hash_lock);
3123         }
3124
3125         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3126         add_timer(&addr_chk_timer);
3127         spin_unlock_bh(&addrconf_verify_lock);
3128 }
3129
3130 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3131 {
3132         struct in6_addr *pfx = NULL;
3133
3134         if (addr)
3135                 pfx = nla_data(addr);
3136
3137         if (local) {
3138                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3139                         pfx = NULL;
3140                 else
3141                         pfx = nla_data(local);
3142         }
3143
3144         return pfx;
3145 }
3146
3147 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3148         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3149         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3150         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3151 };
3152
3153 static int
3154 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3155 {
3156         struct net *net = sock_net(skb->sk);
3157         struct ifaddrmsg *ifm;
3158         struct nlattr *tb[IFA_MAX+1];
3159         struct in6_addr *pfx;
3160         int err;
3161
3162         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3163         if (err < 0)
3164                 return err;
3165
3166         ifm = nlmsg_data(nlh);
3167         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3168         if (pfx == NULL)
3169                 return -EINVAL;
3170
3171         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3172 }
3173
3174 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3175                              u32 prefered_lft, u32 valid_lft)
3176 {
3177         u32 flags;
3178         clock_t expires;
3179         unsigned long timeout;
3180
3181         if (!valid_lft || (prefered_lft > valid_lft))
3182                 return -EINVAL;
3183
3184         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3185         if (addrconf_finite_timeout(timeout)) {
3186                 expires = jiffies_to_clock_t(timeout * HZ);
3187                 valid_lft = timeout;
3188                 flags = RTF_EXPIRES;
3189         } else {
3190                 expires = 0;
3191                 flags = 0;
3192                 ifa_flags |= IFA_F_PERMANENT;
3193         }
3194
3195         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3196         if (addrconf_finite_timeout(timeout)) {
3197                 if (timeout == 0)
3198                         ifa_flags |= IFA_F_DEPRECATED;
3199                 prefered_lft = timeout;
3200         }
3201
3202         spin_lock_bh(&ifp->lock);
3203         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3204         ifp->tstamp = jiffies;
3205         ifp->valid_lft = valid_lft;
3206         ifp->prefered_lft = prefered_lft;
3207
3208         spin_unlock_bh(&ifp->lock);
3209         if (!(ifp->flags&IFA_F_TENTATIVE))
3210                 ipv6_ifa_notify(0, ifp);
3211
3212         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3213                               expires, flags);
3214         addrconf_verify(0);
3215
3216         return 0;
3217 }
3218
3219 static int
3220 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3221 {
3222         struct net *net = sock_net(skb->sk);
3223         struct ifaddrmsg *ifm;
3224         struct nlattr *tb[IFA_MAX+1];
3225         struct in6_addr *pfx;
3226         struct inet6_ifaddr *ifa;
3227         struct net_device *dev;
3228         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3229         u8 ifa_flags;
3230         int err;
3231
3232         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3233         if (err < 0)
3234                 return err;
3235
3236         ifm = nlmsg_data(nlh);
3237         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3238         if (pfx == NULL)
3239                 return -EINVAL;
3240
3241         if (tb[IFA_CACHEINFO]) {
3242                 struct ifa_cacheinfo *ci;
3243
3244                 ci = nla_data(tb[IFA_CACHEINFO]);
3245                 valid_lft = ci->ifa_valid;
3246                 preferred_lft = ci->ifa_prefered;
3247         } else {
3248                 preferred_lft = INFINITY_LIFE_TIME;
3249                 valid_lft = INFINITY_LIFE_TIME;
3250         }
3251
3252         dev =  __dev_get_by_index(net, ifm->ifa_index);
3253         if (dev == NULL)
3254                 return -ENODEV;
3255
3256         /* We ignore other flags so far. */
3257         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3258
3259         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3260         if (ifa == NULL) {
3261                 /*
3262                  * It would be best to check for !NLM_F_CREATE here but
3263                  * userspace alreay relies on not having to provide this.
3264                  */
3265                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3266                                       ifm->ifa_prefixlen, ifa_flags,
3267                                       preferred_lft, valid_lft);
3268         }
3269
3270         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3271             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3272                 err = -EEXIST;
3273         else
3274                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3275
3276         in6_ifa_put(ifa);
3277
3278         return err;
3279 }
3280
3281 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3282                           u8 scope, int ifindex)
3283 {
3284         struct ifaddrmsg *ifm;
3285
3286         ifm = nlmsg_data(nlh);
3287         ifm->ifa_family = AF_INET6;
3288         ifm->ifa_prefixlen = prefixlen;
3289         ifm->ifa_flags = flags;
3290         ifm->ifa_scope = scope;
3291         ifm->ifa_index = ifindex;
3292 }
3293
3294 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3295                          unsigned long tstamp, u32 preferred, u32 valid)
3296 {
3297         struct ifa_cacheinfo ci;
3298
3299         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3300                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3301         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3302                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3303         ci.ifa_prefered = preferred;
3304         ci.ifa_valid = valid;
3305
3306         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3307 }
3308
3309 static inline int rt_scope(int ifa_scope)
3310 {
3311         if (ifa_scope & IFA_HOST)
3312                 return RT_SCOPE_HOST;
3313         else if (ifa_scope & IFA_LINK)
3314                 return RT_SCOPE_LINK;
3315         else if (ifa_scope & IFA_SITE)
3316                 return RT_SCOPE_SITE;
3317         else
3318                 return RT_SCOPE_UNIVERSE;
3319 }
3320
3321 static inline int inet6_ifaddr_msgsize(void)
3322 {
3323         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3324                + nla_total_size(16) /* IFA_ADDRESS */
3325                + nla_total_size(sizeof(struct ifa_cacheinfo));
3326 }
3327
3328 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3329                              u32 pid, u32 seq, int event, unsigned int flags)
3330 {
3331         struct nlmsghdr  *nlh;
3332         u32 preferred, valid;
3333
3334         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3335         if (nlh == NULL)
3336                 return -EMSGSIZE;
3337
3338         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3339                       ifa->idev->dev->ifindex);
3340
3341         if (!(ifa->flags&IFA_F_PERMANENT)) {
3342                 preferred = ifa->prefered_lft;
3343                 valid = ifa->valid_lft;
3344                 if (preferred != INFINITY_LIFE_TIME) {
3345                         long tval = (jiffies - ifa->tstamp)/HZ;
3346                         preferred -= tval;
3347                         if (valid != INFINITY_LIFE_TIME)
3348                                 valid -= tval;
3349                 }
3350         } else {
3351                 preferred = INFINITY_LIFE_TIME;
3352                 valid = INFINITY_LIFE_TIME;
3353         }
3354
3355         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3356             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3357                 nlmsg_cancel(skb, nlh);
3358                 return -EMSGSIZE;
3359         }
3360
3361         return nlmsg_end(skb, nlh);
3362 }
3363
3364 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3365                                 u32 pid, u32 seq, int event, u16 flags)
3366 {
3367         struct nlmsghdr  *nlh;
3368         u8 scope = RT_SCOPE_UNIVERSE;
3369         int ifindex = ifmca->idev->dev->ifindex;
3370
3371         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3372                 scope = RT_SCOPE_SITE;
3373
3374         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3375         if (nlh == NULL)
3376                 return -EMSGSIZE;
3377
3378         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3379         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3380             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3381                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3382                 nlmsg_cancel(skb, nlh);
3383                 return -EMSGSIZE;
3384         }
3385
3386         return nlmsg_end(skb, nlh);
3387 }
3388
3389 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3390                                 u32 pid, u32 seq, int event, unsigned int flags)
3391 {
3392         struct nlmsghdr  *nlh;
3393         u8 scope = RT_SCOPE_UNIVERSE;
3394         int ifindex = ifaca->aca_idev->dev->ifindex;
3395
3396         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3397                 scope = RT_SCOPE_SITE;
3398
3399         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3400         if (nlh == NULL)
3401                 return -EMSGSIZE;
3402
3403         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3404         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3405             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3406                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3407                 nlmsg_cancel(skb, nlh);
3408                 return -EMSGSIZE;
3409         }
3410
3411         return nlmsg_end(skb, nlh);
3412 }
3413
3414 enum addr_type_t
3415 {
3416         UNICAST_ADDR,
3417         MULTICAST_ADDR,
3418         ANYCAST_ADDR,
3419 };
3420
3421 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3422                            enum addr_type_t type)
3423 {
3424         int idx, ip_idx;
3425         int s_idx, s_ip_idx;
3426         int err = 1;
3427         struct net_device *dev;
3428         struct inet6_dev *idev = NULL;
3429         struct inet6_ifaddr *ifa;
3430         struct ifmcaddr6 *ifmca;
3431         struct ifacaddr6 *ifaca;
3432         struct net *net = sock_net(skb->sk);
3433
3434         s_idx = cb->args[0];
3435         s_ip_idx = ip_idx = cb->args[1];
3436
3437         idx = 0;
3438         for_each_netdev(net, dev) {
3439                 if (idx < s_idx)
3440                         goto cont;
3441                 if (idx > s_idx)
3442                         s_ip_idx = 0;
3443                 ip_idx = 0;
3444                 if ((idev = in6_dev_get(dev)) == NULL)
3445                         goto cont;
3446                 read_lock_bh(&idev->lock);
3447                 switch (type) {
3448                 case UNICAST_ADDR:
3449                         /* unicast address incl. temp addr */
3450                         for (ifa = idev->addr_list; ifa;
3451                              ifa = ifa->if_next, ip_idx++) {
3452                                 if (ip_idx < s_ip_idx)
3453                                         continue;
3454                                 err = inet6_fill_ifaddr(skb, ifa,
3455                                                         NETLINK_CB(cb->skb).pid,
3456                                                         cb->nlh->nlmsg_seq,
3457                                                         RTM_NEWADDR,
3458                                                         NLM_F_MULTI);
3459                         }
3460                         break;
3461                 case MULTICAST_ADDR:
3462                         /* multicast address */
3463                         for (ifmca = idev->mc_list; ifmca;
3464                              ifmca = ifmca->next, ip_idx++) {
3465                                 if (ip_idx < s_ip_idx)
3466                                         continue;
3467                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3468                                                           NETLINK_CB(cb->skb).pid,
3469                                                           cb->nlh->nlmsg_seq,
3470                                                           RTM_GETMULTICAST,
3471                                                           NLM_F_MULTI);
3472                         }
3473                         break;
3474                 case ANYCAST_ADDR:
3475                         /* anycast address */
3476                         for (ifaca = idev->ac_list; ifaca;
3477                              ifaca = ifaca->aca_next, ip_idx++) {
3478                                 if (ip_idx < s_ip_idx)
3479                                         continue;
3480                                 err = inet6_fill_ifacaddr(skb, ifaca,
3481                                                           NETLINK_CB(cb->skb).pid,
3482                                                           cb->nlh->nlmsg_seq,
3483                                                           RTM_GETANYCAST,
3484                                                           NLM_F_MULTI);
3485                         }
3486                         break;
3487                 default:
3488                         break;
3489                 }
3490                 read_unlock_bh(&idev->lock);
3491                 in6_dev_put(idev);
3492
3493                 if (err <= 0)
3494                         break;
3495 cont:
3496                 idx++;
3497         }
3498         cb->args[0] = idx;
3499         cb->args[1] = ip_idx;
3500         return skb->len;
3501 }
3502
3503 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3504 {
3505         enum addr_type_t type = UNICAST_ADDR;
3506
3507         return inet6_dump_addr(skb, cb, type);
3508 }
3509
3510 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3511 {
3512         enum addr_type_t type = MULTICAST_ADDR;
3513
3514         return inet6_dump_addr(skb, cb, type);
3515 }
3516
3517
3518 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3519 {
3520         enum addr_type_t type = ANYCAST_ADDR;
3521
3522         return inet6_dump_addr(skb, cb, type);
3523 }
3524
3525 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3526                              void *arg)
3527 {
3528         struct net *net = sock_net(in_skb->sk);
3529         struct ifaddrmsg *ifm;
3530         struct nlattr *tb[IFA_MAX+1];
3531         struct in6_addr *addr = NULL;
3532         struct net_device *dev = NULL;
3533         struct inet6_ifaddr *ifa;
3534         struct sk_buff *skb;
3535         int err;
3536
3537         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3538         if (err < 0)
3539                 goto errout;
3540
3541         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3542         if (addr == NULL) {
3543                 err = -EINVAL;
3544                 goto errout;
3545         }
3546
3547         ifm = nlmsg_data(nlh);
3548         if (ifm->ifa_index)
3549                 dev = __dev_get_by_index(net, ifm->ifa_index);
3550
3551         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3552                 err = -EADDRNOTAVAIL;
3553                 goto errout;
3554         }
3555
3556         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3557                 err = -ENOBUFS;
3558                 goto errout_ifa;
3559         }
3560
3561         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3562                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3563         if (err < 0) {
3564                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3565                 WARN_ON(err == -EMSGSIZE);
3566                 kfree_skb(skb);
3567                 goto errout_ifa;
3568         }
3569         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3570 errout_ifa:
3571         in6_ifa_put(ifa);
3572 errout:
3573         return err;
3574 }
3575
3576 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3577 {
3578         struct sk_buff *skb;
3579         struct net *net = dev_net(ifa->idev->dev);
3580         int err = -ENOBUFS;
3581
3582         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3583         if (skb == NULL)
3584                 goto errout;
3585
3586         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3587         if (err < 0) {
3588                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3589                 WARN_ON(err == -EMSGSIZE);
3590                 kfree_skb(skb);
3591                 goto errout;
3592         }
3593         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3594 errout:
3595         if (err < 0)
3596                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3597 }
3598
3599 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3600                                 __s32 *array, int bytes)
3601 {
3602         BUG_ON(bytes < (DEVCONF_MAX * 4));
3603
3604         memset(array, 0, bytes);
3605         array[DEVCONF_FORWARDING] = cnf->forwarding;
3606         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3607         array[DEVCONF_MTU6] = cnf->mtu6;
3608         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3609         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3610         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3611         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3612         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3613         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3614         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3615         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3616 #ifdef CONFIG_IPV6_PRIVACY
3617         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3618         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3619         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3620         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3621         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3622 #endif
3623         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3624         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3625         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3626 #ifdef CONFIG_IPV6_ROUTER_PREF
3627         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3628         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3629 #ifdef CONFIG_IPV6_ROUTE_INFO
3630         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3631 #endif
3632 #endif
3633         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3634         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3635 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3636         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3637 #endif
3638 #ifdef CONFIG_IPV6_MROUTE
3639         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3640 #endif
3641 }
3642
3643 static inline size_t inet6_if_nlmsg_size(void)
3644 {
3645         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3646                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3647                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3648                + nla_total_size(4) /* IFLA_MTU */
3649                + nla_total_size(4) /* IFLA_LINK */
3650                + nla_total_size( /* IFLA_PROTINFO */
3651                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3652                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3653                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3654                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3655                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3656                  );
3657 }
3658
3659 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3660                                       int bytes)
3661 {
3662         int i;
3663         int pad = bytes - sizeof(u64) * items;
3664         BUG_ON(pad < 0);
3665
3666         /* Use put_unaligned() because stats may not be aligned for u64. */
3667         put_unaligned(items, &stats[0]);
3668         for (i = 1; i < items; i++)
3669                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3670
3671         memset(&stats[items], 0, pad);
3672 }
3673
3674 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3675                              int bytes)
3676 {
3677         switch(attrtype) {
3678         case IFLA_INET6_STATS:
3679                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3680                 break;
3681         case IFLA_INET6_ICMP6STATS:
3682                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3683                 break;
3684         }
3685 }
3686
3687 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3688                              u32 pid, u32 seq, int event, unsigned int flags)
3689 {
3690         struct net_device *dev = idev->dev;
3691         struct nlattr *nla;
3692         struct ifinfomsg *hdr;
3693         struct nlmsghdr *nlh;
3694         void *protoinfo;
3695         struct ifla_cacheinfo ci;
3696
3697         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3698         if (nlh == NULL)
3699                 return -EMSGSIZE;
3700
3701         hdr = nlmsg_data(nlh);
3702         hdr->ifi_family = AF_INET6;
3703         hdr->__ifi_pad = 0;
3704         hdr->ifi_type = dev->type;
3705         hdr->ifi_index = dev->ifindex;
3706         hdr->ifi_flags = dev_get_flags(dev);
3707         hdr->ifi_change = 0;
3708
3709         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3710
3711         if (dev->addr_len)
3712                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3713
3714         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3715         if (dev->ifindex != dev->iflink)
3716                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3717
3718         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3719         if (protoinfo == NULL)
3720                 goto nla_put_failure;
3721
3722         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3723
3724         ci.max_reasm_len = IPV6_MAXPLEN;
3725         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3726                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3727         ci.reachable_time = idev->nd_parms->reachable_time;
3728         ci.retrans_time = idev->nd_parms->retrans_time;
3729         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3730
3731         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3732         if (nla == NULL)
3733                 goto nla_put_failure;
3734         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3735
3736         /* XXX - MC not implemented */
3737
3738         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3739         if (nla == NULL)
3740                 goto nla_put_failure;
3741         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3742
3743         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3744         if (nla == NULL)
3745                 goto nla_put_failure;
3746         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3747
3748         nla_nest_end(skb, protoinfo);
3749         return nlmsg_end(skb, nlh);
3750
3751 nla_put_failure:
3752         nlmsg_cancel(skb, nlh);
3753         return -EMSGSIZE;
3754 }
3755
3756 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3757 {
3758         struct net *net = sock_net(skb->sk);
3759         int idx, err;
3760         int s_idx = cb->args[0];
3761         struct net_device *dev;
3762         struct inet6_dev *idev;
3763
3764         read_lock(&dev_base_lock);
3765         idx = 0;
3766         for_each_netdev(net, dev) {
3767                 if (idx < s_idx)
3768                         goto cont;
3769                 if ((idev = in6_dev_get(dev)) == NULL)
3770                         goto cont;
3771                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3772                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3773                 in6_dev_put(idev);
3774                 if (err <= 0)
3775                         break;
3776 cont:
3777                 idx++;
3778         }
3779         read_unlock(&dev_base_lock);
3780         cb->args[0] = idx;
3781
3782         return skb->len;
3783 }
3784
3785 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3786 {
3787         struct sk_buff *skb;
3788         struct net *net = dev_net(idev->dev);
3789         int err = -ENOBUFS;
3790
3791         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3792         if (skb == NULL)
3793                 goto errout;
3794
3795         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3796         if (err < 0) {
3797                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3798                 WARN_ON(err == -EMSGSIZE);
3799                 kfree_skb(skb);
3800                 goto errout;
3801         }
3802         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3803 errout:
3804         if (err < 0)
3805                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3806 }
3807
3808 static inline size_t inet6_prefix_nlmsg_size(void)
3809 {
3810         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3811                + nla_total_size(sizeof(struct in6_addr))
3812                + nla_total_size(sizeof(struct prefix_cacheinfo));
3813 }
3814
3815 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3816                              struct prefix_info *pinfo, u32 pid, u32 seq,
3817                              int event, unsigned int flags)
3818 {
3819         struct prefixmsg *pmsg;
3820         struct nlmsghdr *nlh;
3821         struct prefix_cacheinfo ci;
3822
3823         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3824         if (nlh == NULL)
3825                 return -EMSGSIZE;
3826
3827         pmsg = nlmsg_data(nlh);
3828         pmsg->prefix_family = AF_INET6;
3829         pmsg->prefix_pad1 = 0;
3830         pmsg->prefix_pad2 = 0;
3831         pmsg->prefix_ifindex = idev->dev->ifindex;
3832         pmsg->prefix_len = pinfo->prefix_len;
3833         pmsg->prefix_type = pinfo->type;
3834         pmsg->prefix_pad3 = 0;
3835         pmsg->prefix_flags = 0;
3836         if (pinfo->onlink)
3837                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3838         if (pinfo->autoconf)
3839                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3840
3841         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3842
3843         ci.preferred_time = ntohl(pinfo->prefered);
3844         ci.valid_time = ntohl(pinfo->valid);
3845         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3846
3847         return nlmsg_end(skb, nlh);
3848
3849 nla_put_failure:
3850         nlmsg_cancel(skb, nlh);
3851         return -EMSGSIZE;
3852 }
3853
3854 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3855                          struct prefix_info *pinfo)
3856 {
3857         struct sk_buff *skb;
3858         struct net *net = dev_net(idev->dev);
3859         int err = -ENOBUFS;
3860
3861         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3862         if (skb == NULL)
3863                 goto errout;
3864
3865         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3866         if (err < 0) {
3867                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3868                 WARN_ON(err == -EMSGSIZE);
3869                 kfree_skb(skb);
3870                 goto errout;
3871         }
3872         err = rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3873 errout:
3874         if (err < 0)
3875                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3876 }
3877
3878 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3879 {
3880         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3881
3882         switch (event) {
3883         case RTM_NEWADDR:
3884                 /*
3885                  * If the address was optimistic
3886                  * we inserted the route at the start of
3887                  * our DAD process, so we don't need
3888                  * to do it again
3889                  */
3890                 if (!(ifp->rt->rt6i_node))
3891                         ip6_ins_rt(ifp->rt);
3892                 if (ifp->idev->cnf.forwarding)
3893                         addrconf_join_anycast(ifp);
3894                 break;
3895         case RTM_DELADDR:
3896                 if (ifp->idev->cnf.forwarding)
3897                         addrconf_leave_anycast(ifp);
3898                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3899                 dst_hold(&ifp->rt->u.dst);
3900                 if (ip6_del_rt(ifp->rt))
3901                         dst_free(&ifp->rt->u.dst);
3902                 break;
3903         }
3904 }
3905
3906 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3907 {
3908         rcu_read_lock_bh();
3909         if (likely(ifp->idev->dead == 0))
3910                 __ipv6_ifa_notify(event, ifp);
3911         rcu_read_unlock_bh();
3912 }
3913
3914 #ifdef CONFIG_SYSCTL
3915
3916 static
3917 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3918                            void __user *buffer, size_t *lenp, loff_t *ppos)
3919 {
3920         int *valp = ctl->data;
3921         int val = *valp;
3922         int ret;
3923
3924         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3925
3926         if (write)
3927                 addrconf_fixup_forwarding(ctl, valp, val);
3928         return ret;
3929 }
3930
3931 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3932                                             int __user *name, int nlen,
3933                                             void __user *oldval,
3934                                             size_t __user *oldlenp,
3935                                             void __user *newval, size_t newlen)
3936 {
3937         int *valp = table->data;
3938         int val = *valp;
3939         int new;
3940
3941         if (!newval || !newlen)
3942                 return 0;
3943         if (newlen != sizeof(int))
3944                 return -EINVAL;
3945         if (get_user(new, (int __user *)newval))
3946                 return -EFAULT;
3947         if (new == *valp)
3948                 return 0;
3949         if (oldval && oldlenp) {
3950                 size_t len;
3951                 if (get_user(len, oldlenp))
3952                         return -EFAULT;
3953                 if (len) {
3954                         if (len > table->maxlen)
3955                                 len = table->maxlen;
3956                         if (copy_to_user(oldval, valp, len))
3957                                 return -EFAULT;
3958                         if (put_user(len, oldlenp))
3959                                 return -EFAULT;
3960                 }
3961         }
3962
3963         *valp = new;
3964         addrconf_fixup_forwarding(table, valp, val);
3965         return 1;
3966 }
3967
3968 static struct addrconf_sysctl_table
3969 {
3970         struct ctl_table_header *sysctl_header;
3971         ctl_table addrconf_vars[DEVCONF_MAX+1];
3972         char *dev_name;
3973 } addrconf_sysctl __read_mostly = {
3974         .sysctl_header = NULL,
3975         .addrconf_vars = {
3976                 {
3977                         .ctl_name       =       NET_IPV6_FORWARDING,
3978                         .procname       =       "forwarding",
3979                         .data           =       &ipv6_devconf.forwarding,
3980                         .maxlen         =       sizeof(int),
3981                         .mode           =       0644,
3982                         .proc_handler   =       &addrconf_sysctl_forward,
3983                         .strategy       =       &addrconf_sysctl_forward_strategy,
3984                 },
3985                 {
3986                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3987                         .procname       =       "hop_limit",
3988                         .data           =       &ipv6_devconf.hop_limit,
3989                         .maxlen         =       sizeof(int),
3990                         .mode           =       0644,
3991                         .proc_handler   =       proc_dointvec,
3992                 },
3993                 {
3994                         .ctl_name       =       NET_IPV6_MTU,
3995                         .procname       =       "mtu",
3996                         .data           =       &ipv6_devconf.mtu6,
3997                         .maxlen         =       sizeof(int),
3998                         .mode           =       0644,
3999                         .proc_handler   =       &proc_dointvec,
4000                 },
4001                 {
4002                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4003                         .procname       =       "accept_ra",
4004                         .data           =       &ipv6_devconf.accept_ra,
4005                         .maxlen         =       sizeof(int),
4006                         .mode           =       0644,
4007                         .proc_handler   =       &proc_dointvec,
4008                 },
4009                 {
4010                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4011                         .procname       =       "accept_redirects",
4012                         .data           =       &ipv6_devconf.accept_redirects,
4013                         .maxlen         =       sizeof(int),
4014                         .mode           =       0644,
4015                         .proc_handler   =       &proc_dointvec,
4016                 },
4017                 {
4018                         .ctl_name       =       NET_IPV6_AUTOCONF,
4019                         .procname       =       "autoconf",
4020                         .data           =       &ipv6_devconf.autoconf,
4021                         .maxlen         =       sizeof(int),
4022                         .mode           =       0644,
4023                         .proc_handler   =       &proc_dointvec,
4024                 },
4025                 {
4026                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4027                         .procname       =       "dad_transmits",
4028                         .data           =       &ipv6_devconf.dad_transmits,
4029                         .maxlen         =       sizeof(int),
4030                         .mode           =       0644,
4031                         .proc_handler   =       &proc_dointvec,
4032                 },
4033                 {
4034                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4035                         .procname       =       "router_solicitations",
4036                         .data           =       &ipv6_devconf.rtr_solicits,
4037                         .maxlen         =       sizeof(int),
4038                         .mode           =       0644,
4039                         .proc_handler   =       &proc_dointvec,
4040                 },
4041                 {
4042                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4043                         .procname       =       "router_solicitation_interval",
4044                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4045                         .maxlen         =       sizeof(int),
4046                         .mode           =       0644,
4047                         .proc_handler   =       &proc_dointvec_jiffies,
4048                         .strategy       =       &sysctl_jiffies,
4049                 },
4050                 {
4051                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4052                         .procname       =       "router_solicitation_delay",
4053                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4054                         .maxlen         =       sizeof(int),
4055                         .mode           =       0644,
4056                         .proc_handler   =       &proc_dointvec_jiffies,
4057                         .strategy       =       &sysctl_jiffies,
4058                 },
4059                 {
4060                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4061                         .procname       =       "force_mld_version",
4062                         .data           =       &ipv6_devconf.force_mld_version,
4063                         .maxlen         =       sizeof(int),
4064                         .mode           =       0644,
4065                         .proc_handler   =       &proc_dointvec,
4066                 },
4067 #ifdef CONFIG_IPV6_PRIVACY
4068                 {
4069                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4070                         .procname       =       "use_tempaddr",
4071                         .data           =       &ipv6_devconf.use_tempaddr,
4072                         .maxlen         =       sizeof(int),
4073                         .mode           =       0644,
4074                         .proc_handler   =       &proc_dointvec,
4075                 },
4076                 {
4077                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4078                         .procname       =       "temp_valid_lft",
4079                         .data           =       &ipv6_devconf.temp_valid_lft,
4080                         .maxlen         =       sizeof(int),
4081                         .mode           =       0644,
4082                         .proc_handler   =       &proc_dointvec,
4083                 },
4084                 {
4085                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4086                         .procname       =       "temp_prefered_lft",
4087                         .data           =       &ipv6_devconf.temp_prefered_lft,
4088                         .maxlen         =       sizeof(int),
4089                         .mode           =       0644,
4090                         .proc_handler   =       &proc_dointvec,
4091                 },
4092                 {
4093                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4094                         .procname       =       "regen_max_retry",
4095                         .data           =       &ipv6_devconf.regen_max_retry,
4096                         .maxlen         =       sizeof(int),
4097                         .mode           =       0644,
4098                         .proc_handler   =       &proc_dointvec,
4099                 },
4100                 {
4101                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4102                         .procname       =       "max_desync_factor",
4103                         .data           =       &ipv6_devconf.max_desync_factor,
4104                         .maxlen         =       sizeof(int),
4105                         .mode           =       0644,
4106                         .proc_handler   =       &proc_dointvec,
4107                 },
4108 #endif
4109                 {
4110                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4111                         .procname       =       "max_addresses",
4112                         .data           =       &ipv6_devconf.max_addresses,
4113                         .maxlen         =       sizeof(int),
4114                         .mode           =       0644,
4115                         .proc_handler   =       &proc_dointvec,
4116                 },
4117                 {
4118                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4119                         .procname       =       "accept_ra_defrtr",
4120                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4121                         .maxlen         =       sizeof(int),
4122                         .mode           =       0644,
4123                         .proc_handler   =       &proc_dointvec,
4124                 },
4125                 {
4126                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4127                         .procname       =       "accept_ra_pinfo",
4128                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4129                         .maxlen         =       sizeof(int),
4130                         .mode           =       0644,
4131                         .proc_handler   =       &proc_dointvec,
4132                 },
4133 #ifdef CONFIG_IPV6_ROUTER_PREF
4134                 {
4135                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4136                         .procname       =       "accept_ra_rtr_pref",
4137                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4138                         .maxlen         =       sizeof(int),
4139                         .mode           =       0644,
4140                         .proc_handler   =       &proc_dointvec,
4141                 },
4142                 {
4143                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4144                         .procname       =       "router_probe_interval",
4145                         .data           =       &ipv6_devconf.rtr_probe_interval,
4146                         .maxlen         =       sizeof(int),
4147                         .mode           =       0644,
4148                         .proc_handler   =       &proc_dointvec_jiffies,
4149                         .strategy       =       &sysctl_jiffies,
4150                 },
4151 #ifdef CONFIG_IPV6_ROUTE_INFO
4152                 {
4153                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4154                         .procname       =       "accept_ra_rt_info_max_plen",
4155                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4156                         .maxlen         =       sizeof(int),
4157                         .mode           =       0644,
4158                         .proc_handler   =       &proc_dointvec,
4159                 },
4160 #endif
4161 #endif
4162                 {
4163                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4164                         .procname       =       "proxy_ndp",
4165                         .data           =       &ipv6_devconf.proxy_ndp,
4166                         .maxlen         =       sizeof(int),
4167                         .mode           =       0644,
4168                         .proc_handler   =       &proc_dointvec,
4169                 },
4170                 {
4171                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4172                         .procname       =       "accept_source_route",
4173                         .data           =       &ipv6_devconf.accept_source_route,
4174                         .maxlen         =       sizeof(int),
4175                         .mode           =       0644,
4176                         .proc_handler   =       &proc_dointvec,
4177                 },
4178 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4179                 {
4180                         .ctl_name       =       CTL_UNNUMBERED,
4181                         .procname       =       "optimistic_dad",
4182                         .data           =       &ipv6_devconf.optimistic_dad,
4183                         .maxlen         =       sizeof(int),
4184                         .mode           =       0644,
4185                         .proc_handler   =       &proc_dointvec,
4186
4187                 },
4188 #endif
4189 #ifdef CONFIG_IPV6_MROUTE
4190                 {
4191                         .ctl_name       =       CTL_UNNUMBERED,
4192                         .procname       =       "mc_forwarding",
4193                         .data           =       &ipv6_devconf.mc_forwarding,
4194                         .maxlen         =       sizeof(int),
4195                         .mode           =       0644,
4196                         .proc_handler   =       &proc_dointvec,
4197                 },
4198 #endif
4199                 {
4200                         .ctl_name       =       0,      /* sentinel */
4201                 }
4202         },
4203 };
4204
4205 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4206                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4207 {
4208         int i;
4209         struct addrconf_sysctl_table *t;
4210
4211 #define ADDRCONF_CTL_PATH_DEV   3
4212
4213         struct ctl_path addrconf_ctl_path[] = {
4214                 { .procname = "net", .ctl_name = CTL_NET, },
4215                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4216                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4217                 { /* to be set */ },
4218                 { },
4219         };
4220
4221
4222         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4223         if (t == NULL)
4224                 goto out;
4225
4226         for (i=0; t->addrconf_vars[i].data; i++) {
4227                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4228                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4229                 t->addrconf_vars[i].extra2 = net;
4230         }
4231
4232         /*
4233          * Make a copy of dev_name, because '.procname' is regarded as const
4234          * by sysctl and we wouldn't want anyone to change it under our feet
4235          * (see SIOCSIFNAME).
4236          */
4237         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4238         if (!t->dev_name)
4239                 goto free;
4240
4241         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4242         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4243
4244         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4245                         t->addrconf_vars);
4246         if (t->sysctl_header == NULL)
4247                 goto free_procname;
4248
4249         p->sysctl = t;
4250         return 0;
4251
4252 free_procname:
4253         kfree(t->dev_name);
4254 free:
4255         kfree(t);
4256 out:
4257         return -ENOBUFS;
4258 }
4259
4260 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4261 {
4262         struct addrconf_sysctl_table *t;
4263
4264         if (p->sysctl == NULL)
4265                 return;
4266
4267         t = p->sysctl;
4268         p->sysctl = NULL;
4269         unregister_sysctl_table(t->sysctl_header);
4270         kfree(t->dev_name);
4271         kfree(t);
4272 }
4273
4274 static void addrconf_sysctl_register(struct inet6_dev *idev)
4275 {
4276         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4277                               NET_IPV6_NEIGH, "ipv6",
4278                               &ndisc_ifinfo_sysctl_change,
4279                               ndisc_ifinfo_sysctl_strategy);
4280         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4281                         idev->dev->ifindex, idev, &idev->cnf);
4282 }
4283
4284 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4285 {
4286         __addrconf_sysctl_unregister(&idev->cnf);
4287         neigh_sysctl_unregister(idev->nd_parms);
4288 }
4289
4290
4291 #endif
4292
4293 static int addrconf_init_net(struct net *net)
4294 {
4295         int err;
4296         struct ipv6_devconf *all, *dflt;
4297
4298         err = -ENOMEM;
4299         all = &ipv6_devconf;
4300         dflt = &ipv6_devconf_dflt;
4301
4302         if (net != &init_net) {
4303                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4304                 if (all == NULL)
4305                         goto err_alloc_all;
4306
4307                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4308                 if (dflt == NULL)
4309                         goto err_alloc_dflt;
4310         }
4311
4312         net->ipv6.devconf_all = all;
4313         net->ipv6.devconf_dflt = dflt;
4314
4315 #ifdef CONFIG_SYSCTL
4316         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4317                         NULL, all);
4318         if (err < 0)
4319                 goto err_reg_all;
4320
4321         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4322                         NULL, dflt);
4323         if (err < 0)
4324                 goto err_reg_dflt;
4325 #endif
4326         return 0;
4327
4328 #ifdef CONFIG_SYSCTL
4329 err_reg_dflt:
4330         __addrconf_sysctl_unregister(all);
4331 err_reg_all:
4332         kfree(dflt);
4333 #endif
4334 err_alloc_dflt:
4335         kfree(all);
4336 err_alloc_all:
4337         return err;
4338 }
4339
4340 static void addrconf_exit_net(struct net *net)
4341 {
4342 #ifdef CONFIG_SYSCTL
4343         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4344         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4345 #endif
4346         if (net != &init_net) {
4347                 kfree(net->ipv6.devconf_dflt);
4348                 kfree(net->ipv6.devconf_all);
4349         }
4350 }
4351
4352 static struct pernet_operations addrconf_ops = {
4353         .init = addrconf_init_net,
4354         .exit = addrconf_exit_net,
4355 };
4356
4357 /*
4358  *      Device notifier
4359  */
4360
4361 int register_inet6addr_notifier(struct notifier_block *nb)
4362 {
4363         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4364 }
4365
4366 EXPORT_SYMBOL(register_inet6addr_notifier);
4367
4368 int unregister_inet6addr_notifier(struct notifier_block *nb)
4369 {
4370         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4371 }
4372
4373 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4374
4375 static void addrconf_net_exit(struct net *net)
4376 {
4377         struct net_device *dev;
4378
4379         rtnl_lock();
4380         /* clean dev list */
4381         for_each_netdev(net, dev) {
4382                 if (__in6_dev_get(dev) == NULL)
4383                         continue;
4384                 addrconf_ifdown(dev, 1);
4385         }
4386         addrconf_ifdown(net->loopback_dev, 2);
4387         rtnl_unlock();
4388 }
4389
4390 static struct pernet_operations addrconf_net_ops = {
4391         .exit = addrconf_net_exit,
4392 };
4393
4394 /*
4395  *      Init / cleanup code
4396  */
4397
4398 int __init addrconf_init(void)
4399 {
4400         int err;
4401
4402         if ((err = ipv6_addr_label_init()) < 0) {
4403                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4404                         err);
4405                 return err;
4406         }
4407
4408         register_pernet_subsys(&addrconf_ops);
4409
4410         /* The addrconf netdev notifier requires that loopback_dev
4411          * has it's ipv6 private information allocated and setup
4412          * before it can bring up and give link-local addresses
4413          * to other devices which are up.
4414          *
4415          * Unfortunately, loopback_dev is not necessarily the first
4416          * entry in the global dev_base list of net devices.  In fact,
4417          * it is likely to be the very last entry on that list.
4418          * So this causes the notifier registry below to try and
4419          * give link-local addresses to all devices besides loopback_dev
4420          * first, then loopback_dev, which cases all the non-loopback_dev
4421          * devices to fail to get a link-local address.
4422          *
4423          * So, as a temporary fix, allocate the ipv6 structure for
4424          * loopback_dev first by hand.
4425          * Longer term, all of the dependencies ipv6 has upon the loopback
4426          * device and it being up should be removed.
4427          */
4428         rtnl_lock();
4429         if (!ipv6_add_dev(init_net.loopback_dev))
4430                 err = -ENOMEM;
4431         rtnl_unlock();
4432         if (err)
4433                 goto errlo;
4434
4435         err = register_pernet_device(&addrconf_net_ops);
4436         if (err)
4437                 return err;
4438
4439         register_netdevice_notifier(&ipv6_dev_notf);
4440
4441         addrconf_verify(0);
4442
4443         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4444         if (err < 0)
4445                 goto errout;
4446
4447         /* Only the first call to __rtnl_register can fail */
4448         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4449         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4450         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4451         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4452         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4453
4454         ipv6_addr_label_rtnl_register();
4455
4456         return 0;
4457 errout:
4458         unregister_netdevice_notifier(&ipv6_dev_notf);
4459 errlo:
4460         unregister_pernet_subsys(&addrconf_ops);
4461
4462         return err;
4463 }
4464
4465 void addrconf_cleanup(void)
4466 {
4467         struct inet6_ifaddr *ifa;
4468         int i;
4469
4470         unregister_netdevice_notifier(&ipv6_dev_notf);
4471         unregister_pernet_device(&addrconf_net_ops);
4472
4473         unregister_pernet_subsys(&addrconf_ops);
4474
4475         rtnl_lock();
4476
4477         /*
4478          *      Check hash table.
4479          */
4480         write_lock_bh(&addrconf_hash_lock);
4481         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4482                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4483                         struct inet6_ifaddr *bifa;
4484
4485                         bifa = ifa;
4486                         ifa = ifa->lst_next;
4487                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4488                         /* Do not free it; something is wrong.
4489                            Now we can investigate it with debugger.
4490                          */
4491                 }
4492         }
4493         write_unlock_bh(&addrconf_hash_lock);
4494
4495         del_timer(&addr_chk_timer);
4496         rtnl_unlock();
4497
4498         unregister_pernet_subsys(&addrconf_net_ops);
4499 }