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/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <linux/if_tunnel.h>
  77 #include <linux/rtnetlink.h>
  78
  79 #ifdef CONFIG_IPV6_PRIVACY
  80 #include <linux/random.h>
  81 #endif
  82
  83 #include <asm/uaccess.h>
  84
  85 #include <linux/proc_fs.h>
  86 #include <linux/seq_file.h>
  87
  88 /* Set to 3 to get tracing... */
  89 #define ACONF_DEBUG 2
  90
  91 #if ACONF_DEBUG >= 3
  92 #define ADBG(x) printk x
  93 #else
  94 #define ADBG(x)
  95 #endif
  96
  97 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  98 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
  99
 100 #ifdef CONFIG_SYSCTL
 101 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
 102 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
 103 #endif
 104
 105 #ifdef CONFIG_IPV6_PRIVACY
 106 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 107 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 108 static void ipv6_regen_rndid(unsigned long data);
 109
 110 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 111 #endif
 112
 113 static int ipv6_count_addresses(struct inet6_dev *idev);
 114
 115 /*
 116  *      Configured unicast address hash table
 117  */
 118 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 119 static DEFINE_RWLOCK(addrconf_hash_lock);
 120
 121 static void addrconf_verify(unsigned long);
 122
 123 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 124 static DEFINE_SPINLOCK(addrconf_verify_lock);
 125
 126 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 127 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 128
 129 static int addrconf_ifdown(struct net_device *dev, int how);
 130
 131 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 132 static void addrconf_dad_timer(unsigned long data);
 133 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 134 static void addrconf_dad_run(struct inet6_dev *idev);
 135 static void addrconf_rs_timer(unsigned long data);
 136 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 137 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 138
 139 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 140                                 struct prefix_info *pinfo);
 141 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
 142
 143 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 144
 145 struct ipv6_devconf ipv6_devconf __read_mostly = {
 146         .forwarding             = 0,
 147         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 148         .mtu6                   = IPV6_MIN_MTU,
 149         .accept_ra              = 1,
 150         .accept_redirects       = 1,
 151         .autoconf               = 1,
 152         .force_mld_version      = 0,
 153         .dad_transmits          = 1,
 154         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 155         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 156         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 157 #ifdef CONFIG_IPV6_PRIVACY
 158         .use_tempaddr           = 0,
 159         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 160         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 161         .regen_max_retry        = REGEN_MAX_RETRY,
 162         .max_desync_factor      = MAX_DESYNC_FACTOR,
 163 #endif
 164         .max_addresses          = IPV6_MAX_ADDRESSES,
 165         .accept_ra_defrtr       = 1,
 166         .accept_ra_pinfo        = 1,
 167 #ifdef CONFIG_IPV6_ROUTER_PREF
 168         .accept_ra_rtr_pref     = 1,
 169         .rtr_probe_interval     = 60 * HZ,
 170 #ifdef CONFIG_IPV6_ROUTE_INFO
 171         .accept_ra_rt_info_max_plen = 0,
 172 #endif
 173 #endif
 174         .proxy_ndp              = 0,
 175 };
 176
 177 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 178         .forwarding             = 0,
 179         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 180         .mtu6                   = IPV6_MIN_MTU,
 181         .accept_ra              = 1,
 182         .accept_redirects       = 1,
 183         .autoconf               = 1,
 184         .dad_transmits          = 1,
 185         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 186         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 187         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 188 #ifdef CONFIG_IPV6_PRIVACY
 189         .use_tempaddr           = 0,
 190         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 191         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 192         .regen_max_retry        = REGEN_MAX_RETRY,
 193         .max_desync_factor      = MAX_DESYNC_FACTOR,
 194 #endif
 195         .max_addresses          = IPV6_MAX_ADDRESSES,
 196         .accept_ra_defrtr       = 1,
 197         .accept_ra_pinfo        = 1,
 198 #ifdef CONFIG_IPV6_ROUTER_PREF
 199         .accept_ra_rtr_pref     = 1,
 200         .rtr_probe_interval     = 60 * HZ,
 201 #ifdef CONFIG_IPV6_ROUTE_INFO
 202         .accept_ra_rt_info_max_plen = 0,
 203 #endif
 204 #endif
 205         .proxy_ndp              = 0,
 206 };
 207
 208 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 209 #if 0
 210 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 211 #endif
 212 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 213
 214 #define IPV6_ADDR_SCOPE_TYPE(scope)     ((scope) << 16)
 215
 216 static inline unsigned ipv6_addr_scope2type(unsigned scope)
 217 {
 218         switch(scope) {
 219         case IPV6_ADDR_SCOPE_NODELOCAL:
 220                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_NODELOCAL) |
 221                         IPV6_ADDR_LOOPBACK);
 222         case IPV6_ADDR_SCOPE_LINKLOCAL:
 223                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL) |
 224                         IPV6_ADDR_LINKLOCAL);
 225         case IPV6_ADDR_SCOPE_SITELOCAL:
 226                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL) |
 227                         IPV6_ADDR_SITELOCAL);
 228         }
 229         return IPV6_ADDR_SCOPE_TYPE(scope);
 230 }
 231
 232 int __ipv6_addr_type(const struct in6_addr *addr)
 233 {
 234         __be32 st;
 235
 236         st = addr->s6_addr32[0];
 237
 238         /* Consider all addresses with the first three bits different of
 239            000 and 111 as unicasts.
 240          */
 241         if ((st & htonl(0xE0000000)) != htonl(0x00000000) &&
 242             (st & htonl(0xE0000000)) != htonl(0xE0000000))
 243                 return (IPV6_ADDR_UNICAST |
 244                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));
 245
 246         if ((st & htonl(0xFF000000)) == htonl(0xFF000000)) {
 247                 /* multicast */
 248                 /* addr-select 3.1 */
 249                 return (IPV6_ADDR_MULTICAST |
 250                         ipv6_addr_scope2type(IPV6_ADDR_MC_SCOPE(addr)));
 251         }
 252
 253         if ((st & htonl(0xFFC00000)) == htonl(0xFE800000))
 254                 return (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST |
 255                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));               /* addr-select 3.1 */
 256         if ((st & htonl(0xFFC00000)) == htonl(0xFEC00000))
 257                 return (IPV6_ADDR_SITELOCAL | IPV6_ADDR_UNICAST |
 258                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL));               /* addr-select 3.1 */
 259
 260         if ((addr->s6_addr32[0] | addr->s6_addr32[1]) == 0) {
 261                 if (addr->s6_addr32[2] == 0) {
 262                         if (addr->s6_addr32[3] == 0)
 263                                 return IPV6_ADDR_ANY;
 264
 265                         if (addr->s6_addr32[3] == htonl(0x00000001))
 266                                 return (IPV6_ADDR_LOOPBACK | IPV6_ADDR_UNICAST |
 267                                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));       /* addr-select 3.4 */
 268
 269                         return (IPV6_ADDR_COMPATv4 | IPV6_ADDR_UNICAST |
 270                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
 271                 }
 272
 273                 if (addr->s6_addr32[2] == htonl(0x0000ffff))
 274                         return (IPV6_ADDR_MAPPED |
 275                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
 276         }
 277
 278         return (IPV6_ADDR_RESERVED |
 279                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.4 */
 280 }
 281
 282 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 283 {
 284         if (del_timer(&ifp->timer))
 285                 __in6_ifa_put(ifp);
 286 }
 287
 288 enum addrconf_timer_t
 289 {
 290         AC_NONE,
 291         AC_DAD,
 292         AC_RS,
 293 };
 294
 295 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 296                                enum addrconf_timer_t what,
 297                                unsigned long when)
 298 {
 299         if (!del_timer(&ifp->timer))
 300                 in6_ifa_hold(ifp);
 301
 302         switch (what) {
 303         case AC_DAD:
 304                 ifp->timer.function = addrconf_dad_timer;
 305                 break;
 306         case AC_RS:
 307                 ifp->timer.function = addrconf_rs_timer;
 308                 break;
 309         default:;
 310         }
 311         ifp->timer.expires = jiffies + when;
 312         add_timer(&ifp->timer);
 313 }
 314
 315 /* Nobody refers to this device, we may destroy it. */
 316
 317 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 318 {
 319         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 320         kfree(idev);
 321 }
 322
 323 void in6_dev_finish_destroy(struct inet6_dev *idev)
 324 {
 325         struct net_device *dev = idev->dev;
 326         BUG_TRAP(idev->addr_list==NULL);
 327         BUG_TRAP(idev->mc_list==NULL);
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 printk("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 338 }
 339
 340 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 341 {
 342         struct inet6_dev *ndev;
 343         struct in6_addr maddr;
 344
 345         ASSERT_RTNL();
 346
 347         if (dev->mtu < IPV6_MIN_MTU)
 348                 return NULL;
 349
 350         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 351
 352         if (ndev == NULL)
 353                 return NULL;
 354
 355         rwlock_init(&ndev->lock);
 356         ndev->dev = dev;
 357         memcpy(&ndev->cnf, &ipv6_devconf_dflt, sizeof(ndev->cnf));
 358         ndev->cnf.mtu6 = dev->mtu;
 359         ndev->cnf.sysctl = NULL;
 360         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 361         if (ndev->nd_parms == NULL) {
 362                 kfree(ndev);
 363                 return NULL;
 364         }
 365         /* We refer to the device */
 366         dev_hold(dev);
 367
 368         if (snmp6_alloc_dev(ndev) < 0) {
 369                 ADBG((KERN_WARNING
 370                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 371                         __FUNCTION__, dev->name));
 372                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 373                 ndev->dead = 1;
 374                 in6_dev_finish_destroy(ndev);
 375                 return NULL;
 376         }
 377
 378         if (snmp6_register_dev(ndev) < 0) {
 379                 ADBG((KERN_WARNING
 380                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 381                         __FUNCTION__, dev->name));
 382                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 383                 ndev->dead = 1;
 384                 in6_dev_finish_destroy(ndev);
 385                 return NULL;
 386         }
 387
 388         /* One reference from device.  We must do this before
 389          * we invoke __ipv6_regen_rndid().
 390          */
 391         in6_dev_hold(ndev);
 392
 393 #ifdef CONFIG_IPV6_PRIVACY
 394         init_timer(&ndev->regen_timer);
 395         ndev->regen_timer.function = ipv6_regen_rndid;
 396         ndev->regen_timer.data = (unsigned long) ndev;
 397         if ((dev->flags&IFF_LOOPBACK) ||
 398             dev->type == ARPHRD_TUNNEL ||
 399 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 400             dev->type == ARPHRD_SIT ||
 401 #endif
 402             dev->type == ARPHRD_NONE) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Privacy Extensions\n",
 405                        dev->name);
 406                 ndev->cnf.use_tempaddr = -1;
 407         } else {
 408                 in6_dev_hold(ndev);
 409                 ipv6_regen_rndid((unsigned long) ndev);
 410         }
 411 #endif
 412
 413         if (netif_carrier_ok(dev))
 414                 ndev->if_flags |= IF_READY;
 415
 416
 417         ipv6_mc_init_dev(ndev);
 418         ndev->tstamp = jiffies;
 419 #ifdef CONFIG_SYSCTL
 420         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
 421                               NET_IPV6_NEIGH, "ipv6",
 422                               &ndisc_ifinfo_sysctl_change,
 423                               NULL);
 424         addrconf_sysctl_register(ndev, &ndev->cnf);
 425 #endif
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join all-node multicast group */
 430         ipv6_addr_all_nodes(&maddr);
 431         ipv6_dev_mc_inc(dev, &maddr);
 432
 433         return ndev;
 434 }
 435
 436 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 437 {
 438         struct inet6_dev *idev;
 439
 440         ASSERT_RTNL();
 441
 442         if ((idev = __in6_dev_get(dev)) == NULL) {
 443                 if ((idev = ipv6_add_dev(dev)) == NULL)
 444                         return NULL;
 445         }
 446
 447         if (dev->flags&IFF_UP)
 448                 ipv6_mc_up(idev);
 449         return idev;
 450 }
 451
 452 #ifdef CONFIG_SYSCTL
 453 static void dev_forward_change(struct inet6_dev *idev)
 454 {
 455         struct net_device *dev;
 456         struct inet6_ifaddr *ifa;
 457         struct in6_addr addr;
 458
 459         if (!idev)
 460                 return;
 461         dev = idev->dev;
 462         if (dev && (dev->flags & IFF_MULTICAST)) {
 463                 ipv6_addr_all_routers(&addr);
 464
 465                 if (idev->cnf.forwarding)
 466                         ipv6_dev_mc_inc(dev, &addr);
 467                 else
 468                         ipv6_dev_mc_dec(dev, &addr);
 469         }
 470         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 471                 if (idev->cnf.forwarding)
 472                         addrconf_join_anycast(ifa);
 473                 else
 474                         addrconf_leave_anycast(ifa);
 475         }
 476 }
 477
 478
 479 static void addrconf_forward_change(void)
 480 {
 481         struct net_device *dev;
 482         struct inet6_dev *idev;
 483
 484         read_lock(&dev_base_lock);
 485         for (dev=dev_base; dev; dev=dev->next) {
 486                 rcu_read_lock();
 487                 idev = __in6_dev_get(dev);
 488                 if (idev) {
 489                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
 490                         idev->cnf.forwarding = ipv6_devconf.forwarding;
 491                         if (changed)
 492                                 dev_forward_change(idev);
 493                 }
 494                 rcu_read_unlock();
 495         }
 496         read_unlock(&dev_base_lock);
 497 }
 498 #endif
 499
 500 /* Nobody refers to this ifaddr, destroy it */
 501
 502 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 503 {
 504         BUG_TRAP(ifp->if_next==NULL);
 505         BUG_TRAP(ifp->lst_next==NULL);
 506 #ifdef NET_REFCNT_DEBUG
 507         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 508 #endif
 509
 510         in6_dev_put(ifp->idev);
 511
 512         if (del_timer(&ifp->timer))
 513                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 514
 515         if (!ifp->dead) {
 516                 printk("Freeing alive inet6 address %p\n", ifp);
 517                 return;
 518         }
 519         dst_release(&ifp->rt->u.dst);
 520
 521         kfree(ifp);
 522 }
 523
 524 static void
 525 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 526 {
 527         struct inet6_ifaddr *ifa, **ifap;
 528         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 529
 530         /*
 531          * Each device address list is sorted in order of scope -
 532          * global before linklocal.
 533          */
 534         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 535              ifap = &ifa->if_next) {
 536                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 537                         break;
 538         }
 539
 540         ifp->if_next = *ifap;
 541         *ifap = ifp;
 542 }
 543
 544 /* On success it returns ifp with increased reference count */
 545
 546 static struct inet6_ifaddr *
 547 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 548               int scope, u32 flags)
 549 {
 550         struct inet6_ifaddr *ifa = NULL;
 551         struct rt6_info *rt;
 552         int hash;
 553         int err = 0;
 554
 555         rcu_read_lock_bh();
 556         if (idev->dead) {
 557                 err = -ENODEV;                  /*XXX*/
 558                 goto out2;
 559         }
 560
 561         write_lock(&addrconf_hash_lock);
 562
 563         /* Ignore adding duplicate addresses on an interface */
 564         if (ipv6_chk_same_addr(addr, idev->dev)) {
 565                 ADBG(("ipv6_add_addr: already assigned\n"));
 566                 err = -EEXIST;
 567                 goto out;
 568         }
 569
 570         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 571
 572         if (ifa == NULL) {
 573                 ADBG(("ipv6_add_addr: malloc failed\n"));
 574                 err = -ENOBUFS;
 575                 goto out;
 576         }
 577
 578         rt = addrconf_dst_alloc(idev, addr, 0);
 579         if (IS_ERR(rt)) {
 580                 err = PTR_ERR(rt);
 581                 goto out;
 582         }
 583
 584         ipv6_addr_copy(&ifa->addr, addr);
 585
 586         spin_lock_init(&ifa->lock);
 587         init_timer(&ifa->timer);
 588         ifa->timer.data = (unsigned long) ifa;
 589         ifa->scope = scope;
 590         ifa->prefix_len = pfxlen;
 591         ifa->flags = flags | IFA_F_TENTATIVE;
 592         ifa->cstamp = ifa->tstamp = jiffies;
 593
 594         ifa->rt = rt;
 595
 596         ifa->idev = idev;
 597         in6_dev_hold(idev);
 598         /* For caller */
 599         in6_ifa_hold(ifa);
 600
 601         /* Add to big hash table */
 602         hash = ipv6_addr_hash(addr);
 603
 604         ifa->lst_next = inet6_addr_lst[hash];
 605         inet6_addr_lst[hash] = ifa;
 606         in6_ifa_hold(ifa);
 607         write_unlock(&addrconf_hash_lock);
 608
 609         write_lock(&idev->lock);
 610         /* Add to inet6_dev unicast addr list. */
 611         ipv6_link_dev_addr(idev, ifa);
 612
 613 #ifdef CONFIG_IPV6_PRIVACY
 614         if (ifa->flags&IFA_F_TEMPORARY) {
 615                 ifa->tmp_next = idev->tempaddr_list;
 616                 idev->tempaddr_list = ifa;
 617                 in6_ifa_hold(ifa);
 618         }
 619 #endif
 620
 621         in6_ifa_hold(ifa);
 622         write_unlock(&idev->lock);
 623 out2:
 624         rcu_read_unlock_bh();
 625
 626         if (likely(err == 0))
 627                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 628         else {
 629                 kfree(ifa);
 630                 ifa = ERR_PTR(err);
 631         }
 632
 633         return ifa;
 634 out:
 635         write_unlock(&addrconf_hash_lock);
 636         goto out2;
 637 }
 638
 639 /* This function wants to get referenced ifp and releases it before return */
 640
 641 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 642 {
 643         struct inet6_ifaddr *ifa, **ifap;
 644         struct inet6_dev *idev = ifp->idev;
 645         int hash;
 646         int deleted = 0, onlink = 0;
 647         unsigned long expires = jiffies;
 648
 649         hash = ipv6_addr_hash(&ifp->addr);
 650
 651         ifp->dead = 1;
 652
 653         write_lock_bh(&addrconf_hash_lock);
 654         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 655              ifap = &ifa->lst_next) {
 656                 if (ifa == ifp) {
 657                         *ifap = ifa->lst_next;
 658                         __in6_ifa_put(ifp);
 659                         ifa->lst_next = NULL;
 660                         break;
 661                 }
 662         }
 663         write_unlock_bh(&addrconf_hash_lock);
 664
 665         write_lock_bh(&idev->lock);
 666 #ifdef CONFIG_IPV6_PRIVACY
 667         if (ifp->flags&IFA_F_TEMPORARY) {
 668                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 669                      ifap = &ifa->tmp_next) {
 670                         if (ifa == ifp) {
 671                                 *ifap = ifa->tmp_next;
 672                                 if (ifp->ifpub) {
 673                                         in6_ifa_put(ifp->ifpub);
 674                                         ifp->ifpub = NULL;
 675                                 }
 676                                 __in6_ifa_put(ifp);
 677                                 ifa->tmp_next = NULL;
 678                                 break;
 679                         }
 680                 }
 681         }
 682 #endif
 683
 684         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 685                 if (ifa == ifp) {
 686                         *ifap = ifa->if_next;
 687                         __in6_ifa_put(ifp);
 688                         ifa->if_next = NULL;
 689                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 690                                 break;
 691                         deleted = 1;
 692                         continue;
 693                 } else if (ifp->flags & IFA_F_PERMANENT) {
 694                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 695                                               ifp->prefix_len)) {
 696                                 if (ifa->flags & IFA_F_PERMANENT) {
 697                                         onlink = 1;
 698                                         if (deleted)
 699                                                 break;
 700                                 } else {
 701                                         unsigned long lifetime;
 702
 703                                         if (!onlink)
 704                                                 onlink = -1;
 705
 706                                         spin_lock(&ifa->lock);
 707                                         lifetime = min_t(unsigned long,
 708                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 709                                         if (time_before(expires,
 710                                                         ifa->tstamp + lifetime * HZ))
 711                                                 expires = ifa->tstamp + lifetime * HZ;
 712                                         spin_unlock(&ifa->lock);
 713                                 }
 714                         }
 715                 }
 716                 ifap = &ifa->if_next;
 717         }
 718         write_unlock_bh(&idev->lock);
 719
 720         ipv6_ifa_notify(RTM_DELADDR, ifp);
 721
 722         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 723
 724         addrconf_del_timer(ifp);
 725
 726         /*
 727          * Purge or update corresponding prefix
 728          *
 729          * 1) we don't purge prefix here if address was not permanent.
 730          *    prefix is managed by its own lifetime.
 731          * 2) if there're no addresses, delete prefix.
 732          * 3) if there're still other permanent address(es),
 733          *    corresponding prefix is still permanent.
 734          * 4) otherwise, update prefix lifetime to the
 735          *    longest valid lifetime among the corresponding
 736          *    addresses on the device.
 737          *    Note: subsequent RA will update lifetime.
 738          *
 739          * --yoshfuji
 740          */
 741         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 742                 struct in6_addr prefix;
 743                 struct rt6_info *rt;
 744
 745                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 746                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
 747
 748                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 749                         if (onlink == 0) {
 750                                 ip6_del_rt(rt);
 751                                 rt = NULL;
 752                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 753                                 rt->rt6i_expires = expires;
 754                                 rt->rt6i_flags |= RTF_EXPIRES;
 755                         }
 756                 }
 757                 dst_release(&rt->u.dst);
 758         }
 759
 760         in6_ifa_put(ifp);
 761 }
 762
 763 #ifdef CONFIG_IPV6_PRIVACY
 764 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 765 {
 766         struct inet6_dev *idev = ifp->idev;
 767         struct in6_addr addr, *tmpaddr;
 768         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 769         int tmp_plen;
 770         int ret = 0;
 771         int max_addresses;
 772
 773         write_lock(&idev->lock);
 774         if (ift) {
 775                 spin_lock_bh(&ift->lock);
 776                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 777                 spin_unlock_bh(&ift->lock);
 778                 tmpaddr = &addr;
 779         } else {
 780                 tmpaddr = NULL;
 781         }
 782 retry:
 783         in6_dev_hold(idev);
 784         if (idev->cnf.use_tempaddr <= 0) {
 785                 write_unlock(&idev->lock);
 786                 printk(KERN_INFO
 787                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 788                 in6_dev_put(idev);
 789                 ret = -1;
 790                 goto out;
 791         }
 792         spin_lock_bh(&ifp->lock);
 793         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 794                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 795                 spin_unlock_bh(&ifp->lock);
 796                 write_unlock(&idev->lock);
 797                 printk(KERN_WARNING
 798                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 799                 in6_dev_put(idev);
 800                 ret = -1;
 801                 goto out;
 802         }
 803         in6_ifa_hold(ifp);
 804         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 805         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 806                 spin_unlock_bh(&ifp->lock);
 807                 write_unlock(&idev->lock);
 808                 printk(KERN_WARNING
 809                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 810                 in6_ifa_put(ifp);
 811                 in6_dev_put(idev);
 812                 ret = -1;
 813                 goto out;
 814         }
 815         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 816         tmp_valid_lft = min_t(__u32,
 817                               ifp->valid_lft,
 818                               idev->cnf.temp_valid_lft);
 819         tmp_prefered_lft = min_t(__u32,
 820                                  ifp->prefered_lft,
 821                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 822         tmp_plen = ifp->prefix_len;
 823         max_addresses = idev->cnf.max_addresses;
 824         tmp_cstamp = ifp->cstamp;
 825         tmp_tstamp = ifp->tstamp;
 826         spin_unlock_bh(&ifp->lock);
 827
 828         write_unlock(&idev->lock);
 829         ift = !max_addresses ||
 830               ipv6_count_addresses(idev) < max_addresses ?
 831                 ipv6_add_addr(idev, &addr, tmp_plen,
 832                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
 833         if (!ift || IS_ERR(ift)) {
 834                 in6_ifa_put(ifp);
 835                 in6_dev_put(idev);
 836                 printk(KERN_INFO
 837                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 838                 tmpaddr = &addr;
 839                 write_lock(&idev->lock);
 840                 goto retry;
 841         }
 842
 843         spin_lock_bh(&ift->lock);
 844         ift->ifpub = ifp;
 845         ift->valid_lft = tmp_valid_lft;
 846         ift->prefered_lft = tmp_prefered_lft;
 847         ift->cstamp = tmp_cstamp;
 848         ift->tstamp = tmp_tstamp;
 849         spin_unlock_bh(&ift->lock);
 850
 851         addrconf_dad_start(ift, 0);
 852         in6_ifa_put(ift);
 853         in6_dev_put(idev);
 854 out:
 855         return ret;
 856 }
 857 #endif
 858
 859 /*
 860  *      Choose an appropriate source address (RFC3484)
 861  */
 862 struct ipv6_saddr_score {
 863         int             addr_type;
 864         unsigned int    attrs;
 865         int             matchlen;
 866         int             scope;
 867         unsigned int    rule;
 868 };
 869
 870 #define IPV6_SADDR_SCORE_LOCAL          0x0001
 871 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
 872 #define IPV6_SADDR_SCORE_HOA            0x0008
 873 #define IPV6_SADDR_SCORE_OIF            0x0010
 874 #define IPV6_SADDR_SCORE_LABEL          0x0020
 875 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
 876
 877 static int inline ipv6_saddr_preferred(int type)
 878 {
 879         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 880                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 881                 return 1;
 882         return 0;
 883 }
 884
 885 /* static matching label */
 886 static int inline ipv6_saddr_label(const struct in6_addr *addr, int type)
 887 {
 888  /*
 889   *     prefix (longest match)  label
 890   *     -----------------------------
 891   *     ::1/128                 0
 892   *     ::/0                    1
 893   *     2002::/16               2
 894   *     ::/96                   3
 895   *     ::ffff:0:0/96           4
 896   *     fc00::/7                5
 897   *     2001::/32               6
 898   */
 899         if (type & IPV6_ADDR_LOOPBACK)
 900                 return 0;
 901         else if (type & IPV6_ADDR_COMPATv4)
 902                 return 3;
 903         else if (type & IPV6_ADDR_MAPPED)
 904                 return 4;
 905         else if (addr->s6_addr32[0] == htonl(0x20010000))
 906                 return 6;
 907         else if (addr->s6_addr16[0] == htons(0x2002))
 908                 return 2;
 909         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
 910                 return 5;
 911         return 1;
 912 }
 913
 914 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
 915                        struct in6_addr *daddr, struct in6_addr *saddr)
 916 {
 917         struct ipv6_saddr_score hiscore;
 918         struct inet6_ifaddr *ifa_result = NULL;
 919         int daddr_type = __ipv6_addr_type(daddr);
 920         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
 921         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
 922         struct net_device *dev;
 923
 924         memset(&hiscore, 0, sizeof(hiscore));
 925
 926         read_lock(&dev_base_lock);
 927         rcu_read_lock();
 928
 929         for (dev = dev_base; dev; dev=dev->next) {
 930                 struct inet6_dev *idev;
 931                 struct inet6_ifaddr *ifa;
 932
 933                 /* Rule 0: Candidate Source Address (section 4)
 934                  *  - multicast and link-local destination address,
 935                  *    the set of candidate source address MUST only
 936                  *    include addresses assigned to interfaces
 937                  *    belonging to the same link as the outgoing
 938                  *    interface.
 939                  * (- For site-local destination addresses, the
 940                  *    set of candidate source addresses MUST only
 941                  *    include addresses assigned to interfaces
 942                  *    belonging to the same site as the outgoing
 943                  *    interface.)
 944                  */
 945                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
 946                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
 947                     daddr_dev && dev != daddr_dev)
 948                         continue;
 949
 950                 idev = __in6_dev_get(dev);
 951                 if (!idev)
 952                         continue;
 953
 954                 read_lock_bh(&idev->lock);
 955                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
 956                         struct ipv6_saddr_score score;
 957
 958                         score.addr_type = __ipv6_addr_type(&ifa->addr);
 959
 960                         /* Rule 0:
 961                          * - Tentative Address (RFC2462 section 5.4)
 962                          *  - A tentative address is not considered
 963                          *    "assigned to an interface" in the traditional
 964                          *    sense.
 965                          * - Candidate Source Address (section 4)
 966                          *  - In any case, anycast addresses, multicast
 967                          *    addresses, and the unspecified address MUST
 968                          *    NOT be included in a candidate set.
 969                          */
 970                         if (ifa->flags & IFA_F_TENTATIVE)
 971                                 continue;
 972                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
 973                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
 974                                 LIMIT_NETDEBUG(KERN_DEBUG
 975                                                "ADDRCONF: unspecified / multicast address"
 976                                                "assigned as unicast address on %s",
 977                                                dev->name);
 978                                 continue;
 979                         }
 980
 981                         score.attrs = 0;
 982                         score.matchlen = 0;
 983                         score.scope = 0;
 984                         score.rule = 0;
 985
 986                         if (ifa_result == NULL) {
 987                                 /* record it if the first available entry */
 988                                 goto record_it;
 989                         }
 990
 991                         /* Rule 1: Prefer same address */
 992                         if (hiscore.rule < 1) {
 993                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
 994                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
 995                                 hiscore.rule++;
 996                         }
 997                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
 998                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
 999                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
1000                                         score.rule = 1;
1001                                         goto record_it;
1002                                 }
1003                         } else {
1004                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
1005                                         continue;
1006                         }
1007
1008                         /* Rule 2: Prefer appropriate scope */
1009                         if (hiscore.rule < 2) {
1010                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1011                                 hiscore.rule++;
1012                         }
1013                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1014                         if (hiscore.scope < score.scope) {
1015                                 if (hiscore.scope < daddr_scope) {
1016                                         score.rule = 2;
1017                                         goto record_it;
1018                                 } else
1019                                         continue;
1020                         } else if (score.scope < hiscore.scope) {
1021                                 if (score.scope < daddr_scope)
1022                                         break; /* addresses sorted by scope */
1023                                 else {
1024                                         score.rule = 2;
1025                                         goto record_it;
1026                                 }
1027                         }
1028
1029                         /* Rule 3: Avoid deprecated address */
1030                         if (hiscore.rule < 3) {
1031                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1032                                     !(ifa_result->flags & IFA_F_DEPRECATED))
1033                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1034                                 hiscore.rule++;
1035                         }
1036                         if (ipv6_saddr_preferred(score.addr_type) ||
1037                             !(ifa->flags & IFA_F_DEPRECATED)) {
1038                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1039                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1040                                         score.rule = 3;
1041                                         goto record_it;
1042                                 }
1043                         } else {
1044                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1045                                         continue;
1046                         }
1047
1048                         /* Rule 4: Prefer home address */
1049 #ifdef CONFIG_IPV6_MIP6
1050                         if (hiscore.rule < 4) {
1051                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1052                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1053                                 hiscore.rule++;
1054                         }
1055                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1056                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1057                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1058                                         score.rule = 4;
1059                                         goto record_it;
1060                                 }
1061                         } else {
1062                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1063                                         continue;
1064                         }
1065 #else
1066                         if (hiscore.rule < 4)
1067                                 hiscore.rule++;
1068 #endif
1069
1070                         /* Rule 5: Prefer outgoing interface */
1071                         if (hiscore.rule < 5) {
1072                                 if (daddr_dev == NULL ||
1073                                     daddr_dev == ifa_result->idev->dev)
1074                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1075                                 hiscore.rule++;
1076                         }
1077                         if (daddr_dev == NULL ||
1078                             daddr_dev == ifa->idev->dev) {
1079                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1080                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1081                                         score.rule = 5;
1082                                         goto record_it;
1083                                 }
1084                         } else {
1085                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1086                                         continue;
1087                         }
1088
1089                         /* Rule 6: Prefer matching label */
1090                         if (hiscore.rule < 6) {
1091                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1092                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1093                                 hiscore.rule++;
1094                         }
1095                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1096                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1097                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1098                                         score.rule = 6;
1099                                         goto record_it;
1100                                 }
1101                         } else {
1102                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1103                                         continue;
1104                         }
1105
1106 #ifdef CONFIG_IPV6_PRIVACY
1107                         /* Rule 7: Prefer public address
1108                          * Note: prefer temprary address if use_tempaddr >= 2
1109                          */
1110                         if (hiscore.rule < 7) {
1111                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1112                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1113                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1114                                 hiscore.rule++;
1115                         }
1116                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1117                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1118                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1119                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1120                                         score.rule = 7;
1121                                         goto record_it;
1122                                 }
1123                         } else {
1124                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1125                                         continue;
1126                         }
1127 #else
1128                         if (hiscore.rule < 7)
1129                                 hiscore.rule++;
1130 #endif
1131                         /* Rule 8: Use longest matching prefix */
1132                         if (hiscore.rule < 8) {
1133                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1134                                 hiscore.rule++;
1135                         }
1136                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1137                         if (score.matchlen > hiscore.matchlen) {
1138                                 score.rule = 8;
1139                                 goto record_it;
1140                         }
1141 #if 0
1142                         else if (score.matchlen < hiscore.matchlen)
1143                                 continue;
1144 #endif
1145
1146                         /* Final Rule: choose first available one */
1147                         continue;
1148 record_it:
1149                         if (ifa_result)
1150                                 in6_ifa_put(ifa_result);
1151                         in6_ifa_hold(ifa);
1152                         ifa_result = ifa;
1153                         hiscore = score;
1154                 }
1155                 read_unlock_bh(&idev->lock);
1156         }
1157         rcu_read_unlock();
1158         read_unlock(&dev_base_lock);
1159
1160         if (!ifa_result)
1161                 return -EADDRNOTAVAIL;
1162
1163         ipv6_addr_copy(saddr, &ifa_result->addr);
1164         in6_ifa_put(ifa_result);
1165         return 0;
1166 }
1167
1168
1169 int ipv6_get_saddr(struct dst_entry *dst,
1170                    struct in6_addr *daddr, struct in6_addr *saddr)
1171 {
1172         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1173 }
1174
1175
1176 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1177 {
1178         struct inet6_dev *idev;
1179         int err = -EADDRNOTAVAIL;
1180
1181         rcu_read_lock();
1182         if ((idev = __in6_dev_get(dev)) != NULL) {
1183                 struct inet6_ifaddr *ifp;
1184
1185                 read_lock_bh(&idev->lock);
1186                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1187                         if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1188                                 ipv6_addr_copy(addr, &ifp->addr);
1189                                 err = 0;
1190                                 break;
1191                         }
1192                 }
1193                 read_unlock_bh(&idev->lock);
1194         }
1195         rcu_read_unlock();
1196         return err;
1197 }
1198
1199 static int ipv6_count_addresses(struct inet6_dev *idev)
1200 {
1201         int cnt = 0;
1202         struct inet6_ifaddr *ifp;
1203
1204         read_lock_bh(&idev->lock);
1205         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1206                 cnt++;
1207         read_unlock_bh(&idev->lock);
1208         return cnt;
1209 }
1210
1211 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1212 {
1213         struct inet6_ifaddr * ifp;
1214         u8 hash = ipv6_addr_hash(addr);
1215
1216         read_lock_bh(&addrconf_hash_lock);
1217         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1218                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1219                     !(ifp->flags&IFA_F_TENTATIVE)) {
1220                         if (dev == NULL || ifp->idev->dev == dev ||
1221                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1222                                 break;
1223                 }
1224         }
1225         read_unlock_bh(&addrconf_hash_lock);
1226         return ifp != NULL;
1227 }
1228
1229 static
1230 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1231 {
1232         struct inet6_ifaddr * ifp;
1233         u8 hash = ipv6_addr_hash(addr);
1234
1235         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1236                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1237                         if (dev == NULL || ifp->idev->dev == dev)
1238                                 break;
1239                 }
1240         }
1241         return ifp != NULL;
1242 }
1243
1244 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1245 {
1246         struct inet6_ifaddr * ifp;
1247         u8 hash = ipv6_addr_hash(addr);
1248
1249         read_lock_bh(&addrconf_hash_lock);
1250         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1251                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1252                         if (dev == NULL || ifp->idev->dev == dev ||
1253                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1254                                 in6_ifa_hold(ifp);
1255                                 break;
1256                         }
1257                 }
1258         }
1259         read_unlock_bh(&addrconf_hash_lock);
1260
1261         return ifp;
1262 }
1263
1264 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1265 {
1266         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1267         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1268         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1269         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1270         int sk_ipv6only = ipv6_only_sock(sk);
1271         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1272         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1273         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1274
1275         if (!sk2_rcv_saddr && !sk_ipv6only)
1276                 return 1;
1277
1278         if (addr_type2 == IPV6_ADDR_ANY &&
1279             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1280                 return 1;
1281
1282         if (addr_type == IPV6_ADDR_ANY &&
1283             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1284                 return 1;
1285
1286         if (sk2_rcv_saddr6 &&
1287             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1288                 return 1;
1289
1290         if (addr_type == IPV6_ADDR_MAPPED &&
1291             !sk2_ipv6only &&
1292             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1293                 return 1;
1294
1295         return 0;
1296 }
1297
1298 /* Gets referenced address, destroys ifaddr */
1299
1300 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1301 {
1302         if (ifp->flags&IFA_F_PERMANENT) {
1303                 spin_lock_bh(&ifp->lock);
1304                 addrconf_del_timer(ifp);
1305                 ifp->flags |= IFA_F_TENTATIVE;
1306                 spin_unlock_bh(&ifp->lock);
1307                 in6_ifa_put(ifp);
1308 #ifdef CONFIG_IPV6_PRIVACY
1309         } else if (ifp->flags&IFA_F_TEMPORARY) {
1310                 struct inet6_ifaddr *ifpub;
1311                 spin_lock_bh(&ifp->lock);
1312                 ifpub = ifp->ifpub;
1313                 if (ifpub) {
1314                         in6_ifa_hold(ifpub);
1315                         spin_unlock_bh(&ifp->lock);
1316                         ipv6_create_tempaddr(ifpub, ifp);
1317                         in6_ifa_put(ifpub);
1318                 } else {
1319                         spin_unlock_bh(&ifp->lock);
1320                 }
1321                 ipv6_del_addr(ifp);
1322 #endif
1323         } else
1324                 ipv6_del_addr(ifp);
1325 }
1326
1327 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1328 {
1329         if (net_ratelimit())
1330                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1331         addrconf_dad_stop(ifp);
1332 }
1333
1334 /* Join to solicited addr multicast group. */
1335
1336 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1337 {
1338         struct in6_addr maddr;
1339
1340         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1341                 return;
1342
1343         addrconf_addr_solict_mult(addr, &maddr);
1344         ipv6_dev_mc_inc(dev, &maddr);
1345 }
1346
1347 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1348 {
1349         struct in6_addr maddr;
1350
1351         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1352                 return;
1353
1354         addrconf_addr_solict_mult(addr, &maddr);
1355         __ipv6_dev_mc_dec(idev, &maddr);
1356 }
1357
1358 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1359 {
1360         struct in6_addr addr;
1361         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1362         if (ipv6_addr_any(&addr))
1363                 return;
1364         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1365 }
1366
1367 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1368 {
1369         struct in6_addr addr;
1370         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1371         if (ipv6_addr_any(&addr))
1372                 return;
1373         __ipv6_dev_ac_dec(ifp->idev, &addr);
1374 }
1375
1376 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1377 {
1378         if (dev->addr_len != ETH_ALEN)
1379                 return -1;
1380         memcpy(eui, dev->dev_addr, 3);
1381         memcpy(eui + 5, dev->dev_addr + 3, 3);
1382
1383         /*
1384          * The zSeries OSA network cards can be shared among various
1385          * OS instances, but the OSA cards have only one MAC address.
1386          * This leads to duplicate address conflicts in conjunction
1387          * with IPv6 if more than one instance uses the same card.
1388          *
1389          * The driver for these cards can deliver a unique 16-bit
1390          * identifier for each instance sharing the same card.  It is
1391          * placed instead of 0xFFFE in the interface identifier.  The
1392          * "u" bit of the interface identifier is not inverted in this
1393          * case.  Hence the resulting interface identifier has local
1394          * scope according to RFC2373.
1395          */
1396         if (dev->dev_id) {
1397                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1398                 eui[4] = dev->dev_id & 0xFF;
1399         } else {
1400                 eui[3] = 0xFF;
1401                 eui[4] = 0xFE;
1402                 eui[0] ^= 2;
1403         }
1404         return 0;
1405 }
1406
1407 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1408 {
1409         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1410         if (dev->addr_len != ARCNET_ALEN)
1411                 return -1;
1412         memset(eui, 0, 7);
1413         eui[7] = *(u8*)dev->dev_addr;
1414         return 0;
1415 }
1416
1417 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1418 {
1419         if (dev->addr_len != INFINIBAND_ALEN)
1420                 return -1;
1421         memcpy(eui, dev->dev_addr + 12, 8);
1422         eui[0] |= 2;
1423         return 0;
1424 }
1425
1426 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1427 {
1428         switch (dev->type) {
1429         case ARPHRD_ETHER:
1430         case ARPHRD_FDDI:
1431         case ARPHRD_IEEE802_TR:
1432                 return addrconf_ifid_eui48(eui, dev);
1433         case ARPHRD_ARCNET:
1434                 return addrconf_ifid_arcnet(eui, dev);
1435         case ARPHRD_INFINIBAND:
1436                 return addrconf_ifid_infiniband(eui, dev);
1437         }
1438         return -1;
1439 }
1440
1441 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1442 {
1443         int err = -1;
1444         struct inet6_ifaddr *ifp;
1445
1446         read_lock_bh(&idev->lock);
1447         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1448                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1449                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1450                         err = 0;
1451                         break;
1452                 }
1453         }
1454         read_unlock_bh(&idev->lock);
1455         return err;
1456 }
1457
1458 #ifdef CONFIG_IPV6_PRIVACY
1459 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1460 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1461 {
1462 regen:
1463         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1464         idev->rndid[0] &= ~0x02;
1465
1466         /*
1467          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1468          * check if generated address is not inappropriate
1469          *
1470          *  - Reserved subnet anycast (RFC 2526)
1471          *      11111101 11....11 1xxxxxxx
1472          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1473          *      00-00-5E-FE-xx-xx-xx-xx
1474          *  - value 0
1475          *  - XXX: already assigned to an address on the device
1476          */
1477         if (idev->rndid[0] == 0xfd &&
1478             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1479             (idev->rndid[7]&0x80))
1480                 goto regen;
1481         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1482                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1483                         goto regen;
1484                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1485                         goto regen;
1486         }
1487
1488         return 0;
1489 }
1490
1491 static void ipv6_regen_rndid(unsigned long data)
1492 {
1493         struct inet6_dev *idev = (struct inet6_dev *) data;
1494         unsigned long expires;
1495
1496         rcu_read_lock_bh();
1497         write_lock_bh(&idev->lock);
1498
1499         if (idev->dead)
1500                 goto out;
1501
1502         if (__ipv6_regen_rndid(idev) < 0)
1503                 goto out;
1504
1505         expires = jiffies +
1506                 idev->cnf.temp_prefered_lft * HZ -
1507                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1508         if (time_before(expires, jiffies)) {
1509                 printk(KERN_WARNING
1510                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1511                         idev->dev->name);
1512                 goto out;
1513         }
1514
1515         if (!mod_timer(&idev->regen_timer, expires))
1516                 in6_dev_hold(idev);
1517
1518 out:
1519         write_unlock_bh(&idev->lock);
1520         rcu_read_unlock_bh();
1521         in6_dev_put(idev);
1522 }
1523
1524 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1525         int ret = 0;
1526
1527         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1528                 ret = __ipv6_regen_rndid(idev);
1529         return ret;
1530 }
1531 #endif
1532
1533 /*
1534  *      Add prefix route.
1535  */
1536
1537 static void
1538 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1539                       unsigned long expires, u32 flags)
1540 {
1541         struct fib6_config cfg = {
1542                 .fc_table = RT6_TABLE_PREFIX,
1543                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1544                 .fc_ifindex = dev->ifindex,
1545                 .fc_expires = expires,
1546                 .fc_dst_len = plen,
1547                 .fc_flags = RTF_UP | flags,
1548         };
1549
1550         ipv6_addr_copy(&cfg.fc_dst, pfx);
1551
1552         /* Prevent useless cloning on PtP SIT.
1553            This thing is done here expecting that the whole
1554            class of non-broadcast devices need not cloning.
1555          */
1556 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1557         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1558                 cfg.fc_flags |= RTF_NONEXTHOP;
1559 #endif
1560
1561         ip6_route_add(&cfg);
1562 }
1563
1564 /* Create "default" multicast route to the interface */
1565
1566 static void addrconf_add_mroute(struct net_device *dev)
1567 {
1568         struct fib6_config cfg = {
1569                 .fc_table = RT6_TABLE_LOCAL,
1570                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1571                 .fc_ifindex = dev->ifindex,
1572                 .fc_dst_len = 8,
1573                 .fc_flags = RTF_UP,
1574         };
1575
1576         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1577
1578         ip6_route_add(&cfg);
1579 }
1580
1581 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1582 static void sit_route_add(struct net_device *dev)
1583 {
1584         struct fib6_config cfg = {
1585                 .fc_table = RT6_TABLE_MAIN,
1586                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1587                 .fc_ifindex = dev->ifindex,
1588                 .fc_dst_len = 96,
1589                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1590         };
1591
1592         /* prefix length - 96 bits "::d.d.d.d" */
1593         ip6_route_add(&cfg);
1594 }
1595 #endif
1596
1597 static void addrconf_add_lroute(struct net_device *dev)
1598 {
1599         struct in6_addr addr;
1600
1601         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1602         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1603 }
1604
1605 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1606 {
1607         struct inet6_dev *idev;
1608
1609         ASSERT_RTNL();
1610
1611         if ((idev = ipv6_find_idev(dev)) == NULL)
1612                 return NULL;
1613
1614         /* Add default multicast route */
1615         addrconf_add_mroute(dev);
1616
1617         /* Add link local route */
1618         addrconf_add_lroute(dev);
1619         return idev;
1620 }
1621
1622 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1623 {
1624         struct prefix_info *pinfo;
1625         __u32 valid_lft;
1626         __u32 prefered_lft;
1627         int addr_type;
1628         unsigned long rt_expires;
1629         struct inet6_dev *in6_dev;
1630
1631         pinfo = (struct prefix_info *) opt;
1632
1633         if (len < sizeof(struct prefix_info)) {
1634                 ADBG(("addrconf: prefix option too short\n"));
1635                 return;
1636         }
1637
1638         /*
1639          *      Validation checks ([ADDRCONF], page 19)
1640          */
1641
1642         addr_type = ipv6_addr_type(&pinfo->prefix);
1643
1644         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1645                 return;
1646
1647         valid_lft = ntohl(pinfo->valid);
1648         prefered_lft = ntohl(pinfo->prefered);
1649
1650         if (prefered_lft > valid_lft) {
1651                 if (net_ratelimit())
1652                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1653                 return;
1654         }
1655
1656         in6_dev = in6_dev_get(dev);
1657
1658         if (in6_dev == NULL) {
1659                 if (net_ratelimit())
1660                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1661                 return;
1662         }
1663
1664         /*
1665          *      Two things going on here:
1666          *      1) Add routes for on-link prefixes
1667          *      2) Configure prefixes with the auto flag set
1668          */
1669
1670         /* Avoid arithmetic overflow. Really, we could
1671            save rt_expires in seconds, likely valid_lft,
1672            but it would require division in fib gc, that it
1673            not good.
1674          */
1675         if (valid_lft >= 0x7FFFFFFF/HZ)
1676                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1677         else
1678                 rt_expires = valid_lft * HZ;
1679
1680         /*
1681          * We convert this (in jiffies) to clock_t later.
1682          * Avoid arithmetic overflow there as well.
1683          * Overflow can happen only if HZ < USER_HZ.
1684          */
1685         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1686                 rt_expires = 0x7FFFFFFF / USER_HZ;
1687
1688         if (pinfo->onlink) {
1689                 struct rt6_info *rt;
1690                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1691
1692                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1693                         if (rt->rt6i_flags&RTF_EXPIRES) {
1694                                 if (valid_lft == 0) {
1695                                         ip6_del_rt(rt);
1696                                         rt = NULL;
1697                                 } else {
1698                                         rt->rt6i_expires = jiffies + rt_expires;
1699                                 }
1700                         }
1701                 } else if (valid_lft) {
1702                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1703                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1704                 }
1705                 if (rt)
1706                         dst_release(&rt->u.dst);
1707         }
1708
1709         /* Try to figure out our local address for this prefix */
1710
1711         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1712                 struct inet6_ifaddr * ifp;
1713                 struct in6_addr addr;
1714                 int create = 0, update_lft = 0;
1715
1716                 if (pinfo->prefix_len == 64) {
1717                         memcpy(&addr, &pinfo->prefix, 8);
1718                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1719                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1720                                 in6_dev_put(in6_dev);
1721                                 return;
1722                         }
1723                         goto ok;
1724                 }
1725                 if (net_ratelimit())
1726                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1727                                pinfo->prefix_len);
1728                 in6_dev_put(in6_dev);
1729                 return;
1730
1731 ok:
1732
1733                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1734
1735                 if (ifp == NULL && valid_lft) {
1736                         int max_addresses = in6_dev->cnf.max_addresses;
1737
1738                         /* Do not allow to create too much of autoconfigured
1739                          * addresses; this would be too easy way to crash kernel.
1740                          */
1741                         if (!max_addresses ||
1742                             ipv6_count_addresses(in6_dev) < max_addresses)
1743                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1744                                                     addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1745
1746                         if (!ifp || IS_ERR(ifp)) {
1747                                 in6_dev_put(in6_dev);
1748                                 return;
1749                         }
1750
1751                         update_lft = create = 1;
1752                         ifp->cstamp = jiffies;
1753                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1754                 }
1755
1756                 if (ifp) {
1757                         int flags;
1758                         unsigned long now;
1759 #ifdef CONFIG_IPV6_PRIVACY
1760                         struct inet6_ifaddr *ift;
1761 #endif
1762                         u32 stored_lft;
1763
1764                         /* update lifetime (RFC2462 5.5.3 e) */
1765                         spin_lock(&ifp->lock);
1766                         now = jiffies;
1767                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1768                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1769                         else
1770                                 stored_lft = 0;
1771                         if (!update_lft && stored_lft) {
1772                                 if (valid_lft > MIN_VALID_LIFETIME ||
1773                                     valid_lft > stored_lft)
1774                                         update_lft = 1;
1775                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1776                                         /* valid_lft <= stored_lft is always true */
1777                                         /* XXX: IPsec */
1778                                         update_lft = 0;
1779                                 } else {
1780                                         valid_lft = MIN_VALID_LIFETIME;
1781                                         if (valid_lft < prefered_lft)
1782                                                 prefered_lft = valid_lft;
1783                                         update_lft = 1;
1784                                 }
1785                         }
1786
1787                         if (update_lft) {
1788                                 ifp->valid_lft = valid_lft;
1789                                 ifp->prefered_lft = prefered_lft;
1790                                 ifp->tstamp = now;
1791                                 flags = ifp->flags;
1792                                 ifp->flags &= ~IFA_F_DEPRECATED;
1793                                 spin_unlock(&ifp->lock);
1794
1795                                 if (!(flags&IFA_F_TENTATIVE))
1796                                         ipv6_ifa_notify(0, ifp);
1797                         } else
1798                                 spin_unlock(&ifp->lock);
1799
1800 #ifdef CONFIG_IPV6_PRIVACY
1801                         read_lock_bh(&in6_dev->lock);
1802                         /* update all temporary addresses in the list */
1803                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1804                                 /*
1805                                  * When adjusting the lifetimes of an existing
1806                                  * temporary address, only lower the lifetimes.
1807                                  * Implementations must not increase the
1808                                  * lifetimes of an existing temporary address
1809                                  * when processing a Prefix Information Option.
1810                                  */
1811                                 spin_lock(&ift->lock);
1812                                 flags = ift->flags;
1813                                 if (ift->valid_lft > valid_lft &&
1814                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1815                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1816                                 if (ift->prefered_lft > prefered_lft &&
1817                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1818                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1819                                 spin_unlock(&ift->lock);
1820                                 if (!(flags&IFA_F_TENTATIVE))
1821                                         ipv6_ifa_notify(0, ift);
1822                         }
1823
1824                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1825                                 /*
1826                                  * When a new public address is created as described in [ADDRCONF],
1827                                  * also create a new temporary address.
1828                                  */
1829                                 read_unlock_bh(&in6_dev->lock);
1830                                 ipv6_create_tempaddr(ifp, NULL);
1831                         } else {
1832                                 read_unlock_bh(&in6_dev->lock);
1833                         }
1834 #endif
1835                         in6_ifa_put(ifp);
1836                         addrconf_verify(0);
1837                 }
1838         }
1839         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1840         in6_dev_put(in6_dev);
1841 }
1842
1843 /*
1844  *      Set destination address.
1845  *      Special case for SIT interfaces where we create a new "virtual"
1846  *      device.
1847  */
1848 int addrconf_set_dstaddr(void __user *arg)
1849 {
1850         struct in6_ifreq ireq;
1851         struct net_device *dev;
1852         int err = -EINVAL;
1853
1854         rtnl_lock();
1855
1856         err = -EFAULT;
1857         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1858                 goto err_exit;
1859
1860         dev = __dev_get_by_index(ireq.ifr6_ifindex);
1861
1862         err = -ENODEV;
1863         if (dev == NULL)
1864                 goto err_exit;
1865
1866 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1867         if (dev->type == ARPHRD_SIT) {
1868                 struct ifreq ifr;
1869                 mm_segment_t    oldfs;
1870                 struct ip_tunnel_parm p;
1871
1872                 err = -EADDRNOTAVAIL;
1873                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1874                         goto err_exit;
1875
1876                 memset(&p, 0, sizeof(p));
1877                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1878                 p.iph.saddr = 0;
1879                 p.iph.version = 4;
1880                 p.iph.ihl = 5;
1881                 p.iph.protocol = IPPROTO_IPV6;
1882                 p.iph.ttl = 64;
1883                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1884
1885                 oldfs = get_fs(); set_fs(KERNEL_DS);
1886                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1887                 set_fs(oldfs);
1888
1889                 if (err == 0) {
1890                         err = -ENOBUFS;
1891                         if ((dev = __dev_get_by_name(p.name)) == NULL)
1892                                 goto err_exit;
1893                         err = dev_open(dev);
1894                 }
1895         }
1896 #endif
1897
1898 err_exit:
1899         rtnl_unlock();
1900         return err;
1901 }
1902
1903 /*
1904  *      Manual configuration of address on an interface
1905  */
1906 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1907                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1908 {
1909         struct inet6_ifaddr *ifp;
1910         struct inet6_dev *idev;
1911         struct net_device *dev;
1912         int scope;
1913
1914         ASSERT_RTNL();
1915
1916         /* check the lifetime */
1917         if (!valid_lft || prefered_lft > valid_lft)
1918                 return -EINVAL;
1919
1920         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1921                 return -ENODEV;
1922
1923         if ((idev = addrconf_add_dev(dev)) == NULL)
1924                 return -ENOBUFS;
1925
1926         scope = ipv6_addr_scope(pfx);
1927
1928         if (valid_lft == INFINITY_LIFE_TIME)
1929                 ifa_flags |= IFA_F_PERMANENT;
1930         else if (valid_lft >= 0x7FFFFFFF/HZ)
1931                 valid_lft = 0x7FFFFFFF/HZ;
1932
1933         if (prefered_lft == 0)
1934                 ifa_flags |= IFA_F_DEPRECATED;
1935         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1936                  (prefered_lft != INFINITY_LIFE_TIME))
1937                 prefered_lft = 0x7FFFFFFF/HZ;
1938
1939         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1940
1941         if (!IS_ERR(ifp)) {
1942                 spin_lock_bh(&ifp->lock);
1943                 ifp->valid_lft = valid_lft;
1944                 ifp->prefered_lft = prefered_lft;
1945                 ifp->tstamp = jiffies;
1946                 spin_unlock_bh(&ifp->lock);
1947
1948                 addrconf_dad_start(ifp, 0);
1949                 in6_ifa_put(ifp);
1950                 addrconf_verify(0);
1951                 return 0;
1952         }
1953
1954         return PTR_ERR(ifp);
1955 }
1956
1957 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1958 {
1959         struct inet6_ifaddr *ifp;
1960         struct inet6_dev *idev;
1961         struct net_device *dev;
1962
1963         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1964                 return -ENODEV;
1965
1966         if ((idev = __in6_dev_get(dev)) == NULL)
1967                 return -ENXIO;
1968
1969         read_lock_bh(&idev->lock);
1970         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1971                 if (ifp->prefix_len == plen &&
1972                     ipv6_addr_equal(pfx, &ifp->addr)) {
1973                         in6_ifa_hold(ifp);
1974                         read_unlock_bh(&idev->lock);
1975
1976                         ipv6_del_addr(ifp);
1977
1978                         /* If the last address is deleted administratively,
1979                            disable IPv6 on this interface.
1980                          */
1981                         if (idev->addr_list == NULL)
1982                                 addrconf_ifdown(idev->dev, 1);
1983                         return 0;
1984                 }
1985         }
1986         read_unlock_bh(&idev->lock);
1987         return -EADDRNOTAVAIL;
1988 }
1989
1990
1991 int addrconf_add_ifaddr(void __user *arg)
1992 {
1993         struct in6_ifreq ireq;
1994         int err;
1995
1996         if (!capable(CAP_NET_ADMIN))
1997                 return -EPERM;
1998
1999         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2000                 return -EFAULT;
2001
2002         rtnl_lock();
2003         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
2004                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2005         rtnl_unlock();
2006         return err;
2007 }
2008
2009 int addrconf_del_ifaddr(void __user *arg)
2010 {
2011         struct in6_ifreq ireq;
2012         int err;
2013
2014         if (!capable(CAP_NET_ADMIN))
2015                 return -EPERM;
2016
2017         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2018                 return -EFAULT;
2019
2020         rtnl_lock();
2021         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2022         rtnl_unlock();
2023         return err;
2024 }
2025
2026 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2027 static void sit_add_v4_addrs(struct inet6_dev *idev)
2028 {
2029         struct inet6_ifaddr * ifp;
2030         struct in6_addr addr;
2031         struct net_device *dev;
2032         int scope;
2033
2034         ASSERT_RTNL();
2035
2036         memset(&addr, 0, sizeof(struct in6_addr));
2037         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2038
2039         if (idev->dev->flags&IFF_POINTOPOINT) {
2040                 addr.s6_addr32[0] = htonl(0xfe800000);
2041                 scope = IFA_LINK;
2042         } else {
2043                 scope = IPV6_ADDR_COMPATv4;
2044         }
2045
2046         if (addr.s6_addr32[3]) {
2047                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2048                 if (!IS_ERR(ifp)) {
2049                         spin_lock_bh(&ifp->lock);
2050                         ifp->flags &= ~IFA_F_TENTATIVE;
2051                         spin_unlock_bh(&ifp->lock);
2052                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2053                         in6_ifa_put(ifp);
2054                 }
2055                 return;
2056         }
2057
2058         for (dev = dev_base; dev != NULL; dev = dev->next) {
2059                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2060                 if (in_dev && (dev->flags & IFF_UP)) {
2061                         struct in_ifaddr * ifa;
2062
2063                         int flag = scope;
2064
2065                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2066                                 int plen;
2067
2068                                 addr.s6_addr32[3] = ifa->ifa_local;
2069
2070                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2071                                         continue;
2072                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2073                                         if (idev->dev->flags&IFF_POINTOPOINT)
2074                                                 continue;
2075                                         flag |= IFA_HOST;
2076                                 }
2077                                 if (idev->dev->flags&IFF_POINTOPOINT)
2078                                         plen = 64;
2079                                 else
2080                                         plen = 96;
2081
2082                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2083                                                     IFA_F_PERMANENT);
2084                                 if (!IS_ERR(ifp)) {
2085                                         spin_lock_bh(&ifp->lock);
2086                                         ifp->flags &= ~IFA_F_TENTATIVE;
2087                                         spin_unlock_bh(&ifp->lock);
2088                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2089                                         in6_ifa_put(ifp);
2090                                 }
2091                         }
2092                 }
2093         }
2094 }
2095 #endif
2096
2097 static void init_loopback(struct net_device *dev)
2098 {
2099         struct inet6_dev  *idev;
2100         struct inet6_ifaddr * ifp;
2101
2102         /* ::1 */
2103
2104         ASSERT_RTNL();
2105
2106         if ((idev = ipv6_find_idev(dev)) == NULL) {
2107                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2108                 return;
2109         }
2110
2111         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2112         if (!IS_ERR(ifp)) {
2113                 spin_lock_bh(&ifp->lock);
2114                 ifp->flags &= ~IFA_F_TENTATIVE;
2115                 spin_unlock_bh(&ifp->lock);
2116                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2117                 in6_ifa_put(ifp);
2118         }
2119 }
2120
2121 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2122 {
2123         struct inet6_ifaddr * ifp;
2124
2125         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2126         if (!IS_ERR(ifp)) {
2127                 addrconf_dad_start(ifp, 0);
2128                 in6_ifa_put(ifp);
2129         }
2130 }
2131
2132 static void addrconf_dev_config(struct net_device *dev)
2133 {
2134         struct in6_addr addr;
2135         struct inet6_dev    * idev;
2136
2137         ASSERT_RTNL();
2138
2139         if ((dev->type != ARPHRD_ETHER) &&
2140             (dev->type != ARPHRD_FDDI) &&
2141             (dev->type != ARPHRD_IEEE802_TR) &&
2142             (dev->type != ARPHRD_ARCNET) &&
2143             (dev->type != ARPHRD_INFINIBAND)) {
2144                 /* Alas, we support only Ethernet autoconfiguration. */
2145                 return;
2146         }
2147
2148         idev = addrconf_add_dev(dev);
2149         if (idev == NULL)
2150                 return;
2151
2152         memset(&addr, 0, sizeof(struct in6_addr));
2153         addr.s6_addr32[0] = htonl(0xFE800000);
2154
2155         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2156                 addrconf_add_linklocal(idev, &addr);
2157 }
2158
2159 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2160 static void addrconf_sit_config(struct net_device *dev)
2161 {
2162         struct inet6_dev *idev;
2163
2164         ASSERT_RTNL();
2165
2166         /*
2167          * Configure the tunnel with one of our IPv4
2168          * addresses... we should configure all of
2169          * our v4 addrs in the tunnel
2170          */
2171
2172         if ((idev = ipv6_find_idev(dev)) == NULL) {
2173                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2174                 return;
2175         }
2176
2177         sit_add_v4_addrs(idev);
2178
2179         if (dev->flags&IFF_POINTOPOINT) {
2180                 addrconf_add_mroute(dev);
2181                 addrconf_add_lroute(dev);
2182         } else
2183                 sit_route_add(dev);
2184 }
2185 #endif
2186
2187 static inline int
2188 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2189 {
2190         struct in6_addr lladdr;
2191
2192         if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2193                 addrconf_add_linklocal(idev, &lladdr);
2194                 return 0;
2195         }
2196         return -1;
2197 }
2198
2199 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2200 {
2201         struct net_device *link_dev;
2202
2203         /* first try to inherit the link-local address from the link device */
2204         if (idev->dev->iflink &&
2205             (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2206                 if (!ipv6_inherit_linklocal(idev, link_dev))
2207                         return;
2208         }
2209         /* then try to inherit it from any device */
2210         for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2211                 if (!ipv6_inherit_linklocal(idev, link_dev))
2212                         return;
2213         }
2214         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2215 }
2216
2217 /*
2218  * Autoconfigure tunnel with a link-local address so routing protocols,
2219  * DHCPv6, MLD etc. can be run over the virtual link
2220  */
2221
2222 static void addrconf_ip6_tnl_config(struct net_device *dev)
2223 {
2224         struct inet6_dev *idev;
2225
2226         ASSERT_RTNL();
2227
2228         if ((idev = addrconf_add_dev(dev)) == NULL) {
2229                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2230                 return;
2231         }
2232         ip6_tnl_add_linklocal(idev);
2233 }
2234
2235 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2236                            void * data)
2237 {
2238         struct net_device *dev = (struct net_device *) data;
2239         struct inet6_dev *idev = __in6_dev_get(dev);
2240         int run_pending = 0;
2241
2242         switch(event) {
2243         case NETDEV_UP:
2244         case NETDEV_CHANGE:
2245                 if (event == NETDEV_UP) {
2246                         if (!netif_carrier_ok(dev)) {
2247                                 /* device is not ready yet. */
2248                                 printk(KERN_INFO
2249                                         "ADDRCONF(NETDEV_UP): %s: "
2250                                         "link is not ready\n",
2251                                         dev->name);
2252                                 break;
2253                         }
2254
2255                         if (idev)
2256                                 idev->if_flags |= IF_READY;
2257                 } else {
2258                         if (!netif_carrier_ok(dev)) {
2259                                 /* device is still not ready. */
2260                                 break;
2261                         }
2262
2263                         if (idev) {
2264                                 if (idev->if_flags & IF_READY) {
2265                                         /* device is already configured. */
2266                                         break;
2267                                 }
2268                                 idev->if_flags |= IF_READY;
2269                         }
2270
2271                         printk(KERN_INFO
2272                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2273                                         "link becomes ready\n",
2274                                         dev->name);
2275
2276                         run_pending = 1;
2277                 }
2278
2279                 switch(dev->type) {
2280 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2281                 case ARPHRD_SIT:
2282                         addrconf_sit_config(dev);
2283                         break;
2284 #endif
2285                 case ARPHRD_TUNNEL6:
2286                         addrconf_ip6_tnl_config(dev);
2287                         break;
2288                 case ARPHRD_LOOPBACK:
2289                         init_loopback(dev);
2290                         break;
2291
2292                 default:
2293                         addrconf_dev_config(dev);
2294                         break;
2295                 };
2296                 if (idev) {
2297                         if (run_pending)
2298                                 addrconf_dad_run(idev);
2299
2300                         /* If the MTU changed during the interface down, when the
2301                            interface up, the changed MTU must be reflected in the
2302                            idev as well as routers.
2303                          */
2304                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2305                                 rt6_mtu_change(dev, dev->mtu);
2306                                 idev->cnf.mtu6 = dev->mtu;
2307                         }
2308                         idev->tstamp = jiffies;
2309                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2310                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2311                            stop IPv6 on this interface.
2312                          */
2313                         if (dev->mtu < IPV6_MIN_MTU)
2314                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2315                 }
2316                 break;
2317
2318         case NETDEV_CHANGEMTU:
2319                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2320                         rt6_mtu_change(dev, dev->mtu);
2321                         idev->cnf.mtu6 = dev->mtu;
2322                         break;
2323                 }
2324
2325                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2326
2327         case NETDEV_DOWN:
2328         case NETDEV_UNREGISTER:
2329                 /*
2330                  *      Remove all addresses from this interface.
2331                  */
2332                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2333                 break;
2334
2335         case NETDEV_CHANGENAME:
2336 #ifdef CONFIG_SYSCTL
2337                 if (idev) {
2338                         addrconf_sysctl_unregister(&idev->cnf);
2339                         neigh_sysctl_unregister(idev->nd_parms);
2340                         neigh_sysctl_register(dev, idev->nd_parms,
2341                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2342                                               &ndisc_ifinfo_sysctl_change,
2343                                               NULL);
2344                         addrconf_sysctl_register(idev, &idev->cnf);
2345                 }
2346 #endif
2347                 break;
2348         };
2349
2350         return NOTIFY_OK;
2351 }
2352
2353 /*
2354  *      addrconf module should be notified of a device going up
2355  */
2356 static struct notifier_block ipv6_dev_notf = {
2357         .notifier_call = addrconf_notify,
2358         .priority = 0
2359 };
2360
2361 static int addrconf_ifdown(struct net_device *dev, int how)
2362 {
2363         struct inet6_dev *idev;
2364         struct inet6_ifaddr *ifa, **bifa;
2365         int i;
2366
2367         ASSERT_RTNL();
2368
2369         if (dev == &loopback_dev && how == 1)
2370                 how = 0;
2371
2372         rt6_ifdown(dev);
2373         neigh_ifdown(&nd_tbl, dev);
2374
2375         idev = __in6_dev_get(dev);
2376         if (idev == NULL)
2377                 return -ENODEV;
2378
2379         /* Step 1: remove reference to ipv6 device from parent device.
2380                    Do not dev_put!
2381          */
2382         if (how == 1) {
2383                 idev->dead = 1;
2384
2385                 /* protected by rtnl_lock */
2386                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2387
2388                 /* Step 1.5: remove snmp6 entry */
2389                 snmp6_unregister_dev(idev);
2390
2391         }
2392
2393         /* Step 2: clear hash table */
2394         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2395                 bifa = &inet6_addr_lst[i];
2396
2397                 write_lock_bh(&addrconf_hash_lock);
2398                 while ((ifa = *bifa) != NULL) {
2399                         if (ifa->idev == idev) {
2400                                 *bifa = ifa->lst_next;
2401                                 ifa->lst_next = NULL;
2402                                 addrconf_del_timer(ifa);
2403                                 in6_ifa_put(ifa);
2404                                 continue;
2405                         }
2406                         bifa = &ifa->lst_next;
2407                 }
2408                 write_unlock_bh(&addrconf_hash_lock);
2409         }
2410
2411         write_lock_bh(&idev->lock);
2412
2413         /* Step 3: clear flags for stateless addrconf */
2414         if (how != 1)
2415                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2416
2417         /* Step 4: clear address list */
2418 #ifdef CONFIG_IPV6_PRIVACY
2419         if (how == 1 && del_timer(&idev->regen_timer))
2420                 in6_dev_put(idev);
2421
2422         /* clear tempaddr list */
2423         while ((ifa = idev->tempaddr_list) != NULL) {
2424                 idev->tempaddr_list = ifa->tmp_next;
2425                 ifa->tmp_next = NULL;
2426                 ifa->dead = 1;
2427                 write_unlock_bh(&idev->lock);
2428                 spin_lock_bh(&ifa->lock);
2429
2430                 if (ifa->ifpub) {
2431                         in6_ifa_put(ifa->ifpub);
2432                         ifa->ifpub = NULL;
2433                 }
2434                 spin_unlock_bh(&ifa->lock);
2435                 in6_ifa_put(ifa);
2436                 write_lock_bh(&idev->lock);
2437         }
2438 #endif
2439         while ((ifa = idev->addr_list) != NULL) {
2440                 idev->addr_list = ifa->if_next;
2441                 ifa->if_next = NULL;
2442                 ifa->dead = 1;
2443                 addrconf_del_timer(ifa);
2444                 write_unlock_bh(&idev->lock);
2445
2446                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2447                 in6_ifa_put(ifa);
2448
2449                 write_lock_bh(&idev->lock);
2450         }
2451         write_unlock_bh(&idev->lock);
2452
2453         /* Step 5: Discard multicast list */
2454
2455         if (how == 1)
2456                 ipv6_mc_destroy_dev(idev);
2457         else
2458                 ipv6_mc_down(idev);
2459
2460         /* Step 5: netlink notification of this interface */
2461         idev->tstamp = jiffies;
2462         inet6_ifinfo_notify(RTM_DELLINK, idev);
2463
2464         /* Shot the device (if unregistered) */
2465
2466         if (how == 1) {
2467 #ifdef CONFIG_SYSCTL
2468                 addrconf_sysctl_unregister(&idev->cnf);
2469                 neigh_sysctl_unregister(idev->nd_parms);
2470 #endif
2471                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2472                 neigh_ifdown(&nd_tbl, dev);
2473                 in6_dev_put(idev);
2474         }
2475         return 0;
2476 }
2477
2478 static void addrconf_rs_timer(unsigned long data)
2479 {
2480         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2481
2482         if (ifp->idev->cnf.forwarding)
2483                 goto out;
2484
2485         if (ifp->idev->if_flags & IF_RA_RCVD) {
2486                 /*
2487                  *      Announcement received after solicitation
2488                  *      was sent
2489                  */
2490                 goto out;
2491         }
2492
2493         spin_lock(&ifp->lock);
2494         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2495                 struct in6_addr all_routers;
2496
2497                 /* The wait after the last probe can be shorter */
2498                 addrconf_mod_timer(ifp, AC_RS,
2499                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2500                                    ifp->idev->cnf.rtr_solicit_delay :
2501                                    ifp->idev->cnf.rtr_solicit_interval);
2502                 spin_unlock(&ifp->lock);
2503
2504                 ipv6_addr_all_routers(&all_routers);
2505
2506                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2507         } else {
2508                 spin_unlock(&ifp->lock);
2509                 /*
2510                  * Note: we do not support deprecated "all on-link"
2511                  * assumption any longer.
2512                  */
2513                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2514                        ifp->idev->dev->name);
2515         }
2516
2517 out:
2518         in6_ifa_put(ifp);
2519 }
2520
2521 /*
2522  *      Duplicate Address Detection
2523  */
2524 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2525 {
2526         unsigned long rand_num;
2527         struct inet6_dev *idev = ifp->idev;
2528
2529         rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2530         ifp->probes = idev->cnf.dad_transmits;
2531         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2532 }
2533
2534 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2535 {
2536         struct inet6_dev *idev = ifp->idev;
2537         struct net_device *dev = idev->dev;
2538
2539         addrconf_join_solict(dev, &ifp->addr);
2540
2541         if (ifp->prefix_len != 128 && (ifp->flags&IFA_F_PERMANENT))
2542                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 0,
2543                                         flags);
2544
2545         net_srandom(ifp->addr.s6_addr32[3]);
2546
2547         read_lock_bh(&idev->lock);
2548         if (ifp->dead)
2549                 goto out;
2550         spin_lock_bh(&ifp->lock);
2551
2552         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2553             !(ifp->flags&IFA_F_TENTATIVE) ||
2554             ifp->flags & IFA_F_NODAD) {
2555                 ifp->flags &= ~IFA_F_TENTATIVE;
2556                 spin_unlock_bh(&ifp->lock);
2557                 read_unlock_bh(&idev->lock);
2558
2559                 addrconf_dad_completed(ifp);
2560                 return;
2561         }
2562
2563         if (!(idev->if_flags & IF_READY)) {
2564                 spin_unlock_bh(&ifp->lock);
2565                 read_unlock_bh(&idev->lock);
2566                 /*
2567                  * If the defice is not ready:
2568                  * - keep it tentative if it is a permanent address.
2569                  * - otherwise, kill it.
2570                  */
2571                 in6_ifa_hold(ifp);
2572                 addrconf_dad_stop(ifp);
2573                 return;
2574         }
2575         addrconf_dad_kick(ifp);
2576         spin_unlock_bh(&ifp->lock);
2577 out:
2578         read_unlock_bh(&idev->lock);
2579 }
2580
2581 static void addrconf_dad_timer(unsigned long data)
2582 {
2583         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2584         struct inet6_dev *idev = ifp->idev;
2585         struct in6_addr unspec;
2586         struct in6_addr mcaddr;
2587
2588         read_lock_bh(&idev->lock);
2589         if (idev->dead) {
2590                 read_unlock_bh(&idev->lock);
2591                 goto out;
2592         }
2593         spin_lock_bh(&ifp->lock);
2594         if (ifp->probes == 0) {
2595                 /*
2596                  * DAD was successful
2597                  */
2598
2599                 ifp->flags &= ~IFA_F_TENTATIVE;
2600                 spin_unlock_bh(&ifp->lock);
2601                 read_unlock_bh(&idev->lock);
2602
2603                 addrconf_dad_completed(ifp);
2604
2605                 goto out;
2606         }
2607
2608         ifp->probes--;
2609         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2610         spin_unlock_bh(&ifp->lock);
2611         read_unlock_bh(&idev->lock);
2612
2613         /* send a neighbour solicitation for our addr */
2614         memset(&unspec, 0, sizeof(unspec));
2615         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2616         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2617 out:
2618         in6_ifa_put(ifp);
2619 }
2620
2621 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2622 {
2623         struct net_device *     dev = ifp->idev->dev;
2624
2625         /*
2626          *      Configure the address for reception. Now it is valid.
2627          */
2628
2629         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2630
2631         /* If added prefix is link local and forwarding is off,
2632            start sending router solicitations.
2633          */
2634
2635         if (ifp->idev->cnf.forwarding == 0 &&
2636             ifp->idev->cnf.rtr_solicits > 0 &&
2637             (dev->flags&IFF_LOOPBACK) == 0 &&
2638             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2639                 struct in6_addr all_routers;
2640
2641                 ipv6_addr_all_routers(&all_routers);
2642
2643                 /*
2644                  *      If a host as already performed a random delay
2645                  *      [...] as part of DAD [...] there is no need
2646                  *      to delay again before sending the first RS
2647                  */
2648                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2649
2650                 spin_lock_bh(&ifp->lock);
2651                 ifp->probes = 1;
2652                 ifp->idev->if_flags |= IF_RS_SENT;
2653                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2654                 spin_unlock_bh(&ifp->lock);
2655         }
2656 }
2657
2658 static void addrconf_dad_run(struct inet6_dev *idev) {
2659         struct inet6_ifaddr *ifp;
2660
2661         read_lock_bh(&idev->lock);
2662         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2663                 spin_lock_bh(&ifp->lock);
2664                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2665                         spin_unlock_bh(&ifp->lock);
2666                         continue;
2667                 }
2668                 spin_unlock_bh(&ifp->lock);
2669                 addrconf_dad_kick(ifp);
2670         }
2671         read_unlock_bh(&idev->lock);
2672 }
2673
2674 #ifdef CONFIG_PROC_FS
2675 struct if6_iter_state {
2676         int bucket;
2677 };
2678
2679 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2680 {
2681         struct inet6_ifaddr *ifa = NULL;
2682         struct if6_iter_state *state = seq->private;
2683
2684         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2685                 ifa = inet6_addr_lst[state->bucket];
2686                 if (ifa)
2687                         break;
2688         }
2689         return ifa;
2690 }
2691
2692 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2693 {
2694         struct if6_iter_state *state = seq->private;
2695
2696         ifa = ifa->lst_next;
2697 try_again:
2698         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2699                 ifa = inet6_addr_lst[state->bucket];
2700                 goto try_again;
2701         }
2702         return ifa;
2703 }
2704
2705 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2706 {
2707         struct inet6_ifaddr *ifa = if6_get_first(seq);
2708
2709         if (ifa)
2710                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2711                         --pos;
2712         return pos ? NULL : ifa;
2713 }
2714
2715 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2716 {
2717         read_lock_bh(&addrconf_hash_lock);
2718         return if6_get_idx(seq, *pos);
2719 }
2720
2721 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2722 {
2723         struct inet6_ifaddr *ifa;
2724
2725         ifa = if6_get_next(seq, v);
2726         ++*pos;
2727         return ifa;
2728 }
2729
2730 static void if6_seq_stop(struct seq_file *seq, void *v)
2731 {
2732         read_unlock_bh(&addrconf_hash_lock);
2733 }
2734
2735 static int if6_seq_show(struct seq_file *seq, void *v)
2736 {
2737         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2738         seq_printf(seq,
2739                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2740                    NIP6(ifp->addr),
2741                    ifp->idev->dev->ifindex,
2742                    ifp->prefix_len,
2743                    ifp->scope,
2744                    ifp->flags,
2745                    ifp->idev->dev->name);
2746         return 0;
2747 }
2748
2749 static struct seq_operations if6_seq_ops = {
2750         .start  = if6_seq_start,
2751         .next   = if6_seq_next,
2752         .show   = if6_seq_show,
2753         .stop   = if6_seq_stop,
2754 };
2755
2756 static int if6_seq_open(struct inode *inode, struct file *file)
2757 {
2758         struct seq_file *seq;
2759         int rc = -ENOMEM;
2760         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2761
2762         if (!s)
2763                 goto out;
2764
2765         rc = seq_open(file, &if6_seq_ops);
2766         if (rc)
2767                 goto out_kfree;
2768
2769         seq = file->private_data;
2770         seq->private = s;
2771 out:
2772         return rc;
2773 out_kfree:
2774         kfree(s);
2775         goto out;
2776 }
2777
2778 static const struct file_operations if6_fops = {
2779         .owner          = THIS_MODULE,
2780         .open           = if6_seq_open,
2781         .read           = seq_read,
2782         .llseek         = seq_lseek,
2783         .release        = seq_release_private,
2784 };
2785
2786 int __init if6_proc_init(void)
2787 {
2788         if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2789                 return -ENOMEM;
2790         return 0;
2791 }
2792
2793 void if6_proc_exit(void)
2794 {
2795         proc_net_remove("if_inet6");
2796 }
2797 #endif  /* CONFIG_PROC_FS */
2798
2799 #ifdef CONFIG_IPV6_MIP6
2800 /* Check if address is a home address configured on any interface. */
2801 int ipv6_chk_home_addr(struct in6_addr *addr)
2802 {
2803         int ret = 0;
2804         struct inet6_ifaddr * ifp;
2805         u8 hash = ipv6_addr_hash(addr);
2806         read_lock_bh(&addrconf_hash_lock);
2807         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2808                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2809                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2810                         ret = 1;
2811                         break;
2812                 }
2813         }
2814         read_unlock_bh(&addrconf_hash_lock);
2815         return ret;
2816 }
2817 #endif
2818
2819 /*
2820  *      Periodic address status verification
2821  */
2822
2823 static void addrconf_verify(unsigned long foo)
2824 {
2825         struct inet6_ifaddr *ifp;
2826         unsigned long now, next;
2827         int i;
2828
2829         spin_lock_bh(&addrconf_verify_lock);
2830         now = jiffies;
2831         next = now + ADDR_CHECK_FREQUENCY;
2832
2833         del_timer(&addr_chk_timer);
2834
2835         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2836
2837 restart:
2838                 read_lock(&addrconf_hash_lock);
2839                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2840                         unsigned long age;
2841 #ifdef CONFIG_IPV6_PRIVACY
2842                         unsigned long regen_advance;
2843 #endif
2844
2845                         if (ifp->flags & IFA_F_PERMANENT)
2846                                 continue;
2847
2848                         spin_lock(&ifp->lock);
2849                         age = (now - ifp->tstamp) / HZ;
2850
2851 #ifdef CONFIG_IPV6_PRIVACY
2852                         regen_advance = ifp->idev->cnf.regen_max_retry *
2853                                         ifp->idev->cnf.dad_transmits *
2854                                         ifp->idev->nd_parms->retrans_time / HZ;
2855 #endif
2856
2857                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2858                             age >= ifp->valid_lft) {
2859                                 spin_unlock(&ifp->lock);
2860                                 in6_ifa_hold(ifp);
2861                                 read_unlock(&addrconf_hash_lock);
2862                                 ipv6_del_addr(ifp);
2863                                 goto restart;
2864                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2865                                 spin_unlock(&ifp->lock);
2866                                 continue;
2867                         } else if (age >= ifp->prefered_lft) {
2868                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2869                                 int deprecate = 0;
2870
2871                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2872                                         deprecate = 1;
2873                                         ifp->flags |= IFA_F_DEPRECATED;
2874                                 }
2875
2876                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2877                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2878
2879                                 spin_unlock(&ifp->lock);
2880
2881                                 if (deprecate) {
2882                                         in6_ifa_hold(ifp);
2883                                         read_unlock(&addrconf_hash_lock);
2884
2885                                         ipv6_ifa_notify(0, ifp);
2886                                         in6_ifa_put(ifp);
2887                                         goto restart;
2888                                 }
2889 #ifdef CONFIG_IPV6_PRIVACY
2890                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2891                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2892                                 if (age >= ifp->prefered_lft - regen_advance) {
2893                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2894                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2895                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2896                                         if (!ifp->regen_count && ifpub) {
2897                                                 ifp->regen_count++;
2898                                                 in6_ifa_hold(ifp);
2899                                                 in6_ifa_hold(ifpub);
2900                                                 spin_unlock(&ifp->lock);
2901                                                 read_unlock(&addrconf_hash_lock);
2902                                                 spin_lock(&ifpub->lock);
2903                                                 ifpub->regen_count = 0;
2904                                                 spin_unlock(&ifpub->lock);
2905                                                 ipv6_create_tempaddr(ifpub, ifp);
2906                                                 in6_ifa_put(ifpub);
2907                                                 in6_ifa_put(ifp);
2908                                                 goto restart;
2909                                         }
2910                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2911                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2912                                 spin_unlock(&ifp->lock);
2913 #endif
2914                         } else {
2915                                 /* ifp->prefered_lft <= ifp->valid_lft */
2916                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2917                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2918                                 spin_unlock(&ifp->lock);
2919                         }
2920                 }
2921                 read_unlock(&addrconf_hash_lock);
2922         }
2923
2924         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2925         add_timer(&addr_chk_timer);
2926         spin_unlock_bh(&addrconf_verify_lock);
2927 }
2928
2929 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2930 {
2931         struct in6_addr *pfx = NULL;
2932
2933         if (addr)
2934                 pfx = nla_data(addr);
2935
2936         if (local) {
2937                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2938                         pfx = NULL;
2939                 else
2940                         pfx = nla_data(local);
2941         }
2942
2943         return pfx;
2944 }
2945
2946 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2947         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2948         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2949         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2950 };
2951
2952 static int
2953 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2954 {
2955         struct ifaddrmsg *ifm;
2956         struct nlattr *tb[IFA_MAX+1];
2957         struct in6_addr *pfx;
2958         int err;
2959
2960         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2961         if (err < 0)
2962                 return err;
2963
2964         ifm = nlmsg_data(nlh);
2965         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2966         if (pfx == NULL)
2967                 return -EINVAL;
2968
2969         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2970 }
2971
2972 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
2973                              u32 prefered_lft, u32 valid_lft)
2974 {
2975         if (!valid_lft || (prefered_lft > valid_lft))
2976                 return -EINVAL;
2977
2978         if (valid_lft == INFINITY_LIFE_TIME)
2979                 ifa_flags |= IFA_F_PERMANENT;
2980         else if (valid_lft >= 0x7FFFFFFF/HZ)
2981                 valid_lft = 0x7FFFFFFF/HZ;
2982
2983         if (prefered_lft == 0)
2984                 ifa_flags |= IFA_F_DEPRECATED;
2985         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2986                  (prefered_lft != INFINITY_LIFE_TIME))
2987                 prefered_lft = 0x7FFFFFFF/HZ;
2988
2989         spin_lock_bh(&ifp->lock);
2990         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
2991         ifp->tstamp = jiffies;
2992         ifp->valid_lft = valid_lft;
2993         ifp->prefered_lft = prefered_lft;
2994
2995         spin_unlock_bh(&ifp->lock);
2996         if (!(ifp->flags&IFA_F_TENTATIVE))
2997                 ipv6_ifa_notify(0, ifp);
2998
2999         addrconf_verify(0);
3000
3001         return 0;
3002 }
3003
3004 static int
3005 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3006 {
3007         struct ifaddrmsg *ifm;
3008         struct nlattr *tb[IFA_MAX+1];
3009         struct in6_addr *pfx;
3010         struct inet6_ifaddr *ifa;
3011         struct net_device *dev;
3012         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3013         u8 ifa_flags;
3014         int err;
3015
3016         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3017         if (err < 0)
3018                 return err;
3019
3020         ifm = nlmsg_data(nlh);
3021         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3022         if (pfx == NULL)
3023                 return -EINVAL;
3024
3025         if (tb[IFA_CACHEINFO]) {
3026                 struct ifa_cacheinfo *ci;
3027
3028                 ci = nla_data(tb[IFA_CACHEINFO]);
3029                 valid_lft = ci->ifa_valid;
3030                 preferred_lft = ci->ifa_prefered;
3031         } else {
3032                 preferred_lft = INFINITY_LIFE_TIME;
3033                 valid_lft = INFINITY_LIFE_TIME;
3034         }
3035
3036         dev =  __dev_get_by_index(ifm->ifa_index);
3037         if (dev == NULL)
3038                 return -ENODEV;
3039
3040         /* We ignore other flags so far. */
3041         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3042
3043         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3044         if (ifa == NULL) {
3045                 /*
3046                  * It would be best to check for !NLM_F_CREATE here but
3047                  * userspace alreay relies on not having to provide this.
3048                  */
3049                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3050                                       ifa_flags, preferred_lft, valid_lft);
3051         }
3052
3053         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3054             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3055                 err = -EEXIST;
3056         else
3057                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3058
3059         in6_ifa_put(ifa);
3060
3061         return err;
3062 }
3063
3064 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3065                           u8 scope, int ifindex)
3066 {
3067         struct ifaddrmsg *ifm;
3068
3069         ifm = nlmsg_data(nlh);
3070         ifm->ifa_family = AF_INET6;
3071         ifm->ifa_prefixlen = prefixlen;
3072         ifm->ifa_flags = flags;
3073         ifm->ifa_scope = scope;
3074         ifm->ifa_index = ifindex;
3075 }
3076
3077 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3078                          unsigned long tstamp, u32 preferred, u32 valid)
3079 {
3080         struct ifa_cacheinfo ci;
3081
3082         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3083                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3084         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3085                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3086         ci.ifa_prefered = preferred;
3087         ci.ifa_valid = valid;
3088
3089         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3090 }
3091
3092 static inline int rt_scope(int ifa_scope)
3093 {
3094         if (ifa_scope & IFA_HOST)
3095                 return RT_SCOPE_HOST;
3096         else if (ifa_scope & IFA_LINK)
3097                 return RT_SCOPE_LINK;
3098         else if (ifa_scope & IFA_SITE)
3099                 return RT_SCOPE_SITE;
3100         else
3101                 return RT_SCOPE_UNIVERSE;
3102 }
3103
3104 static inline int inet6_ifaddr_msgsize(void)
3105 {
3106         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3107                + nla_total_size(16) /* IFA_ADDRESS */
3108                + nla_total_size(sizeof(struct ifa_cacheinfo));
3109 }
3110
3111 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3112                              u32 pid, u32 seq, int event, unsigned int flags)
3113 {
3114         struct nlmsghdr  *nlh;
3115         u32 preferred, valid;
3116
3117         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3118         if (nlh == NULL)
3119                 return -EMSGSIZE;
3120
3121         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3122                       ifa->idev->dev->ifindex);
3123
3124         if (!(ifa->flags&IFA_F_PERMANENT)) {
3125                 preferred = ifa->prefered_lft;
3126                 valid = ifa->valid_lft;
3127                 if (preferred != INFINITY_LIFE_TIME) {
3128                         long tval = (jiffies - ifa->tstamp)/HZ;
3129                         preferred -= tval;
3130                         if (valid != INFINITY_LIFE_TIME)
3131                                 valid -= tval;
3132                 }
3133         } else {
3134                 preferred = INFINITY_LIFE_TIME;
3135                 valid = INFINITY_LIFE_TIME;
3136         }
3137
3138         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3139             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3140                 nlmsg_cancel(skb, nlh);
3141                 return -EMSGSIZE;
3142         }
3143
3144         return nlmsg_end(skb, nlh);
3145 }
3146
3147 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3148                                 u32 pid, u32 seq, int event, u16 flags)
3149 {
3150         struct nlmsghdr  *nlh;
3151         u8 scope = RT_SCOPE_UNIVERSE;
3152         int ifindex = ifmca->idev->dev->ifindex;
3153
3154         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3155                 scope = RT_SCOPE_SITE;
3156
3157         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3158         if (nlh == NULL)
3159                 return -EMSGSIZE;
3160
3161         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3162         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3163             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3164                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3165                 nlmsg_cancel(skb, nlh);
3166                 return -EMSGSIZE;
3167         }
3168
3169         return nlmsg_end(skb, nlh);
3170 }
3171
3172 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3173                                 u32 pid, u32 seq, int event, unsigned int flags)
3174 {
3175         struct nlmsghdr  *nlh;
3176         u8 scope = RT_SCOPE_UNIVERSE;
3177         int ifindex = ifaca->aca_idev->dev->ifindex;
3178
3179         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3180                 scope = RT_SCOPE_SITE;
3181
3182         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3183         if (nlh == NULL)
3184                 return -EMSGSIZE;
3185
3186         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3187         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3188             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3189                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3190                 nlmsg_cancel(skb, nlh);
3191                 return -EMSGSIZE;
3192         }
3193
3194         return nlmsg_end(skb, nlh);
3195 }
3196
3197 enum addr_type_t
3198 {
3199         UNICAST_ADDR,
3200         MULTICAST_ADDR,
3201         ANYCAST_ADDR,
3202 };
3203
3204 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3205                            enum addr_type_t type)
3206 {
3207         int idx, ip_idx;
3208         int s_idx, s_ip_idx;
3209         int err = 1;
3210         struct net_device *dev;
3211         struct inet6_dev *idev = NULL;
3212         struct inet6_ifaddr *ifa;
3213         struct ifmcaddr6 *ifmca;
3214         struct ifacaddr6 *ifaca;
3215
3216         s_idx = cb->args[0];
3217         s_ip_idx = ip_idx = cb->args[1];
3218         read_lock(&dev_base_lock);
3219
3220         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3221                 if (idx < s_idx)
3222                         continue;
3223                 if (idx > s_idx)
3224                         s_ip_idx = 0;
3225                 ip_idx = 0;
3226                 if ((idev = in6_dev_get(dev)) == NULL)
3227                         continue;
3228                 read_lock_bh(&idev->lock);
3229                 switch (type) {
3230                 case UNICAST_ADDR:
3231                         /* unicast address incl. temp addr */
3232                         for (ifa = idev->addr_list; ifa;
3233                              ifa = ifa->if_next, ip_idx++) {
3234                                 if (ip_idx < s_ip_idx)
3235                                         continue;
3236                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3237                                     NETLINK_CB(cb->skb).pid,
3238                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3239                                     NLM_F_MULTI)) <= 0)
3240                                         goto done;
3241                         }
3242                         break;
3243                 case MULTICAST_ADDR:
3244                         /* multicast address */
3245                         for (ifmca = idev->mc_list; ifmca;
3246                              ifmca = ifmca->next, ip_idx++) {
3247                                 if (ip_idx < s_ip_idx)
3248                                         continue;
3249                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3250                                     NETLINK_CB(cb->skb).pid,
3251                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3252                                     NLM_F_MULTI)) <= 0)
3253                                         goto done;
3254                         }
3255                         break;
3256                 case ANYCAST_ADDR:
3257                         /* anycast address */
3258                         for (ifaca = idev->ac_list; ifaca;
3259                              ifaca = ifaca->aca_next, ip_idx++) {
3260                                 if (ip_idx < s_ip_idx)
3261                                         continue;
3262                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3263                                     NETLINK_CB(cb->skb).pid,
3264                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3265                                     NLM_F_MULTI)) <= 0)
3266                                         goto done;
3267                         }
3268                         break;
3269                 default:
3270                         break;
3271                 }
3272                 read_unlock_bh(&idev->lock);
3273                 in6_dev_put(idev);
3274         }
3275 done:
3276         if (err <= 0) {
3277                 read_unlock_bh(&idev->lock);
3278                 in6_dev_put(idev);
3279         }
3280         read_unlock(&dev_base_lock);
3281         cb->args[0] = idx;
3282         cb->args[1] = ip_idx;
3283         return skb->len;
3284 }
3285
3286 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3287 {
3288         enum addr_type_t type = UNICAST_ADDR;
3289         return inet6_dump_addr(skb, cb, type);
3290 }
3291
3292 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3293 {
3294         enum addr_type_t type = MULTICAST_ADDR;
3295         return inet6_dump_addr(skb, cb, type);
3296 }
3297
3298
3299 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3300 {
3301         enum addr_type_t type = ANYCAST_ADDR;
3302         return inet6_dump_addr(skb, cb, type);
3303 }
3304
3305 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3306                              void *arg)
3307 {
3308         struct ifaddrmsg *ifm;
3309         struct nlattr *tb[IFA_MAX+1];
3310         struct in6_addr *addr = NULL;
3311         struct net_device *dev = NULL;
3312         struct inet6_ifaddr *ifa;
3313         struct sk_buff *skb;
3314         int err;
3315
3316         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3317         if (err < 0)
3318                 goto errout;
3319
3320         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3321         if (addr == NULL) {
3322                 err = -EINVAL;
3323                 goto errout;
3324         }
3325
3326         ifm = nlmsg_data(nlh);
3327         if (ifm->ifa_index)
3328                 dev = __dev_get_by_index(ifm->ifa_index);
3329
3330         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3331                 err = -EADDRNOTAVAIL;
3332                 goto errout;
3333         }
3334
3335         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3336                 err = -ENOBUFS;
3337                 goto errout_ifa;
3338         }
3339
3340         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3341                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3342         if (err < 0) {
3343                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3344                 WARN_ON(err == -EMSGSIZE);
3345                 kfree_skb(skb);
3346                 goto errout_ifa;
3347         }
3348         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3349 errout_ifa:
3350         in6_ifa_put(ifa);
3351 errout:
3352         return err;
3353 }
3354
3355 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3356 {
3357         struct sk_buff *skb;
3358         int err = -ENOBUFS;
3359
3360         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3361         if (skb == NULL)
3362                 goto errout;
3363
3364         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3365         if (err < 0) {
3366                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3367                 WARN_ON(err == -EMSGSIZE);
3368                 kfree_skb(skb);
3369                 goto errout;
3370         }
3371         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3372 errout:
3373         if (err < 0)
3374                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3375 }
3376
3377 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
3378                                 __s32 *array, int bytes)
3379 {
3380         BUG_ON(bytes < (DEVCONF_MAX * 4));
3381
3382         memset(array, 0, bytes);
3383         array[DEVCONF_FORWARDING] = cnf->forwarding;
3384         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3385         array[DEVCONF_MTU6] = cnf->mtu6;
3386         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3387         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3388         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3389         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3390         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3391         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3392         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3393         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3394 #ifdef CONFIG_IPV6_PRIVACY
3395         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3396         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3397         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3398         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3399         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3400 #endif
3401         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3402         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3403         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3404 #ifdef CONFIG_IPV6_ROUTER_PREF
3405         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3406         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3407 #ifdef CONFIG_IPV6_ROUTE_INFO
3408         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3409 #endif
3410 #endif
3411         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3412 }
3413
3414 static inline size_t inet6_if_nlmsg_size(void)
3415 {
3416         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3417                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3418                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3419                + nla_total_size(4) /* IFLA_MTU */
3420                + nla_total_size(4) /* IFLA_LINK */
3421                + nla_total_size( /* IFLA_PROTINFO */
3422                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3423                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3424                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3425                  );
3426 }
3427
3428 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3429                              u32 pid, u32 seq, int event, unsigned int flags)
3430 {
3431         struct net_device *dev = idev->dev;
3432         struct nlattr *conf;
3433         struct ifinfomsg *hdr;
3434         struct nlmsghdr *nlh;
3435         void *protoinfo;
3436         struct ifla_cacheinfo ci;
3437
3438         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3439         if (nlh == NULL)
3440                 return -EMSGSIZE;
3441
3442         hdr = nlmsg_data(nlh);
3443         hdr->ifi_family = AF_INET6;
3444         hdr->__ifi_pad = 0;
3445         hdr->ifi_type = dev->type;
3446         hdr->ifi_index = dev->ifindex;
3447         hdr->ifi_flags = dev_get_flags(dev);
3448         hdr->ifi_change = 0;
3449
3450         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3451
3452         if (dev->addr_len)
3453                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3454
3455         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3456         if (dev->ifindex != dev->iflink)
3457                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3458
3459         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3460         if (protoinfo == NULL)
3461                 goto nla_put_failure;
3462
3463         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3464
3465         ci.max_reasm_len = IPV6_MAXPLEN;
3466         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3467                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3468         ci.reachable_time = idev->nd_parms->reachable_time;
3469         ci.retrans_time = idev->nd_parms->retrans_time;
3470         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3471
3472         conf = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3473         if (conf == NULL)
3474                 goto nla_put_failure;
3475         ipv6_store_devconf(&idev->cnf, nla_data(conf), nla_len(conf));
3476
3477         /* XXX - Statistics/MC not implemented */
3478
3479         nla_nest_end(skb, protoinfo);
3480         return nlmsg_end(skb, nlh);
3481
3482 nla_put_failure:
3483         nlmsg_cancel(skb, nlh);
3484         return -EMSGSIZE;
3485 }
3486
3487 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3488 {
3489         int idx, err;
3490         int s_idx = cb->args[0];
3491         struct net_device *dev;
3492         struct inet6_dev *idev;
3493
3494         read_lock(&dev_base_lock);
3495         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3496                 if (idx < s_idx)
3497                         continue;
3498                 if ((idev = in6_dev_get(dev)) == NULL)
3499                         continue;
3500                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3501                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3502                 in6_dev_put(idev);
3503                 if (err <= 0)
3504                         break;
3505         }
3506         read_unlock(&dev_base_lock);
3507         cb->args[0] = idx;
3508
3509         return skb->len;
3510 }
3511
3512 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3513 {
3514         struct sk_buff *skb;
3515         int err = -ENOBUFS;
3516
3517         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3518         if (skb == NULL)
3519                 goto errout;
3520
3521         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3522         if (err < 0) {
3523                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3524                 WARN_ON(err == -EMSGSIZE);
3525                 kfree_skb(skb);
3526                 goto errout;
3527         }
3528         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3529 errout:
3530         if (err < 0)
3531                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3532 }
3533
3534 static inline size_t inet6_prefix_nlmsg_size(void)
3535 {
3536         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3537                + nla_total_size(sizeof(struct in6_addr))
3538                + nla_total_size(sizeof(struct prefix_cacheinfo));
3539 }
3540
3541 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3542                              struct prefix_info *pinfo, u32 pid, u32 seq,
3543                              int event, unsigned int flags)
3544 {
3545         struct prefixmsg *pmsg;
3546         struct nlmsghdr *nlh;
3547         struct prefix_cacheinfo ci;
3548
3549         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3550         if (nlh == NULL)
3551                 return -EMSGSIZE;
3552
3553         pmsg = nlmsg_data(nlh);
3554         pmsg->prefix_family = AF_INET6;
3555         pmsg->prefix_pad1 = 0;
3556         pmsg->prefix_pad2 = 0;
3557         pmsg->prefix_ifindex = idev->dev->ifindex;
3558         pmsg->prefix_len = pinfo->prefix_len;
3559         pmsg->prefix_type = pinfo->type;
3560         pmsg->prefix_pad3 = 0;
3561         pmsg->prefix_flags = 0;
3562         if (pinfo->onlink)
3563                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3564         if (pinfo->autoconf)
3565                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3566
3567         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3568
3569         ci.preferred_time = ntohl(pinfo->prefered);
3570         ci.valid_time = ntohl(pinfo->valid);
3571         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3572
3573         return nlmsg_end(skb, nlh);
3574
3575 nla_put_failure:
3576         nlmsg_cancel(skb, nlh);
3577         return -EMSGSIZE;
3578 }
3579
3580 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3581                          struct prefix_info *pinfo)
3582 {
3583         struct sk_buff *skb;
3584         int err = -ENOBUFS;
3585
3586         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3587         if (skb == NULL)
3588                 goto errout;
3589
3590         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3591         if (err < 0) {
3592                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3593                 WARN_ON(err == -EMSGSIZE);
3594                 kfree_skb(skb);
3595                 goto errout;
3596         }
3597         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3598 errout:
3599         if (err < 0)
3600                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3601 }
3602
3603 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3604         [RTM_GETLINK - RTM_BASE] = { .dumpit    = inet6_dump_ifinfo, },
3605         [RTM_NEWADDR - RTM_BASE] = { .doit      = inet6_rtm_newaddr, },
3606         [RTM_DELADDR - RTM_BASE] = { .doit      = inet6_rtm_deladdr, },
3607         [RTM_GETADDR - RTM_BASE] = { .doit      = inet6_rtm_getaddr,
3608                                      .dumpit    = inet6_dump_ifaddr, },
3609         [RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3610         [RTM_GETANYCAST - RTM_BASE] = { .dumpit = inet6_dump_ifacaddr, },
3611         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet6_rtm_newroute, },
3612         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
3613         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
3614                                       .dumpit   = inet6_dump_fib, },
3615 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3616         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3617 #endif
3618 };
3619
3620 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3621 {
3622         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3623
3624         switch (event) {
3625         case RTM_NEWADDR:
3626                 ip6_ins_rt(ifp->rt);
3627                 if (ifp->idev->cnf.forwarding)
3628                         addrconf_join_anycast(ifp);
3629                 break;
3630         case RTM_DELADDR:
3631                 if (ifp->idev->cnf.forwarding)
3632                         addrconf_leave_anycast(ifp);
3633                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3634                 dst_hold(&ifp->rt->u.dst);
3635                 if (ip6_del_rt(ifp->rt))
3636                         dst_free(&ifp->rt->u.dst);
3637                 break;
3638         }
3639 }
3640
3641 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3642 {
3643         rcu_read_lock_bh();
3644         if (likely(ifp->idev->dead == 0))
3645                 __ipv6_ifa_notify(event, ifp);
3646         rcu_read_unlock_bh();
3647 }
3648
3649 #ifdef CONFIG_SYSCTL
3650
3651 static
3652 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3653                            void __user *buffer, size_t *lenp, loff_t *ppos)
3654 {
3655         int *valp = ctl->data;
3656         int val = *valp;
3657         int ret;
3658
3659         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3660
3661         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3662                 if (valp != &ipv6_devconf.forwarding) {
3663                         if ((!*valp) ^ (!val)) {
3664                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3665                                 if (idev == NULL)
3666                                         return ret;
3667                                 dev_forward_change(idev);
3668                         }
3669                 } else {
3670                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3671                         addrconf_forward_change();
3672                 }
3673                 if (*valp)
3674                         rt6_purge_dflt_routers();
3675         }
3676
3677         return ret;
3678 }
3679
3680 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3681                                             int __user *name, int nlen,
3682                                             void __user *oldval,
3683                                             size_t __user *oldlenp,
3684                                             void __user *newval, size_t newlen)
3685 {
3686         int *valp = table->data;
3687         int new;
3688
3689         if (!newval || !newlen)
3690                 return 0;
3691         if (newlen != sizeof(int))
3692                 return -EINVAL;
3693         if (get_user(new, (int __user *)newval))
3694                 return -EFAULT;
3695         if (new == *valp)
3696                 return 0;
3697         if (oldval && oldlenp) {
3698                 size_t len;
3699                 if (get_user(len, oldlenp))
3700                         return -EFAULT;
3701                 if (len) {
3702                         if (len > table->maxlen)
3703                                 len = table->maxlen;
3704                         if (copy_to_user(oldval, valp, len))
3705                                 return -EFAULT;
3706                         if (put_user(len, oldlenp))
3707                                 return -EFAULT;
3708                 }
3709         }
3710
3711         if (valp != &ipv6_devconf_dflt.forwarding) {
3712                 if (valp != &ipv6_devconf.forwarding) {
3713                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3714                         int changed;
3715                         if (unlikely(idev == NULL))
3716                                 return -ENODEV;
3717                         changed = (!*valp) ^ (!new);
3718                         *valp = new;
3719                         if (changed)
3720                                 dev_forward_change(idev);
3721                 } else {
3722                         *valp = new;
3723                         addrconf_forward_change();
3724                 }
3725
3726                 if (*valp)
3727                         rt6_purge_dflt_routers();
3728         } else
3729                 *valp = new;
3730
3731         return 1;
3732 }
3733
3734 static struct addrconf_sysctl_table
3735 {
3736         struct ctl_table_header *sysctl_header;
3737         ctl_table addrconf_vars[__NET_IPV6_MAX];
3738         ctl_table addrconf_dev[2];
3739         ctl_table addrconf_conf_dir[2];
3740         ctl_table addrconf_proto_dir[2];
3741         ctl_table addrconf_root_dir[2];
3742 } addrconf_sysctl __read_mostly = {
3743         .sysctl_header = NULL,
3744         .addrconf_vars = {
3745                 {
3746                         .ctl_name       =       NET_IPV6_FORWARDING,
3747                         .procname       =       "forwarding",
3748                         .data           =       &ipv6_devconf.forwarding,
3749                         .maxlen         =       sizeof(int),
3750                         .mode           =       0644,
3751                         .proc_handler   =       &addrconf_sysctl_forward,
3752                         .strategy       =       &addrconf_sysctl_forward_strategy,
3753                 },
3754                 {
3755                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3756                         .procname       =       "hop_limit",
3757                         .data           =       &ipv6_devconf.hop_limit,
3758                         .maxlen         =       sizeof(int),
3759                         .mode           =       0644,
3760                         .proc_handler   =       proc_dointvec,
3761                 },
3762                 {
3763                         .ctl_name       =       NET_IPV6_MTU,
3764                         .procname       =       "mtu",
3765                         .data           =       &ipv6_devconf.mtu6,
3766                         .maxlen         =       sizeof(int),
3767                         .mode           =       0644,
3768                         .proc_handler   =       &proc_dointvec,
3769                 },
3770                 {
3771                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3772                         .procname       =       "accept_ra",
3773                         .data           =       &ipv6_devconf.accept_ra,
3774                         .maxlen         =       sizeof(int),
3775                         .mode           =       0644,
3776                         .proc_handler   =       &proc_dointvec,
3777                 },
3778                 {
3779                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3780                         .procname       =       "accept_redirects",
3781                         .data           =       &ipv6_devconf.accept_redirects,
3782                         .maxlen         =       sizeof(int),
3783                         .mode           =       0644,
3784                         .proc_handler   =       &proc_dointvec,
3785                 },
3786                 {
3787                         .ctl_name       =       NET_IPV6_AUTOCONF,
3788                         .procname       =       "autoconf",
3789                         .data           =       &ipv6_devconf.autoconf,
3790                         .maxlen         =       sizeof(int),
3791                         .mode           =       0644,
3792                         .proc_handler   =       &proc_dointvec,
3793                 },
3794                 {
3795                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3796                         .procname       =       "dad_transmits",
3797                         .data           =       &ipv6_devconf.dad_transmits,
3798                         .maxlen         =       sizeof(int),
3799                         .mode           =       0644,
3800                         .proc_handler   =       &proc_dointvec,
3801                 },
3802                 {
3803                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3804                         .procname       =       "router_solicitations",
3805                         .data           =       &ipv6_devconf.rtr_solicits,
3806                         .maxlen         =       sizeof(int),
3807                         .mode           =       0644,
3808                         .proc_handler   =       &proc_dointvec,
3809                 },
3810                 {
3811                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3812                         .procname       =       "router_solicitation_interval",
3813                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3814                         .maxlen         =       sizeof(int),
3815                         .mode           =       0644,
3816                         .proc_handler   =       &proc_dointvec_jiffies,
3817                         .strategy       =       &sysctl_jiffies,
3818                 },
3819                 {
3820                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3821                         .procname       =       "router_solicitation_delay",
3822                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3823                         .maxlen         =       sizeof(int),
3824                         .mode           =       0644,
3825                         .proc_handler   =       &proc_dointvec_jiffies,
3826                         .strategy       =       &sysctl_jiffies,
3827                 },
3828                 {
3829                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3830                         .procname       =       "force_mld_version",
3831                         .data           =       &ipv6_devconf.force_mld_version,
3832                         .maxlen         =       sizeof(int),
3833                         .mode           =       0644,
3834                         .proc_handler   =       &proc_dointvec,
3835                 },
3836 #ifdef CONFIG_IPV6_PRIVACY
3837                 {
3838                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3839                         .procname       =       "use_tempaddr",
3840                         .data           =       &ipv6_devconf.use_tempaddr,
3841                         .maxlen         =       sizeof(int),
3842                         .mode           =       0644,
3843                         .proc_handler   =       &proc_dointvec,
3844                 },
3845                 {
3846                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3847                         .procname       =       "temp_valid_lft",
3848                         .data           =       &ipv6_devconf.temp_valid_lft,
3849                         .maxlen         =       sizeof(int),
3850                         .mode           =       0644,
3851                         .proc_handler   =       &proc_dointvec,
3852                 },
3853                 {
3854                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3855                         .procname       =       "temp_prefered_lft",
3856                         .data           =       &ipv6_devconf.temp_prefered_lft,
3857                         .maxlen         =       sizeof(int),
3858                         .mode           =       0644,
3859                         .proc_handler   =       &proc_dointvec,
3860                 },
3861                 {
3862                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3863                         .procname       =       "regen_max_retry",
3864                         .data           =       &ipv6_devconf.regen_max_retry,
3865                         .maxlen         =       sizeof(int),
3866                         .mode           =       0644,
3867                         .proc_handler   =       &proc_dointvec,
3868                 },
3869                 {
3870                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3871                         .procname       =       "max_desync_factor",
3872                         .data           =       &ipv6_devconf.max_desync_factor,
3873                         .maxlen         =       sizeof(int),
3874                         .mode           =       0644,
3875                         .proc_handler   =       &proc_dointvec,
3876                 },
3877 #endif
3878                 {
3879                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3880                         .procname       =       "max_addresses",
3881                         .data           =       &ipv6_devconf.max_addresses,
3882                         .maxlen         =       sizeof(int),
3883                         .mode           =       0644,
3884                         .proc_handler   =       &proc_dointvec,
3885                 },
3886                 {
3887                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3888                         .procname       =       "accept_ra_defrtr",
3889                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3890                         .maxlen         =       sizeof(int),
3891                         .mode           =       0644,
3892                         .proc_handler   =       &proc_dointvec,
3893                 },
3894                 {
3895                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3896                         .procname       =       "accept_ra_pinfo",
3897                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3898                         .maxlen         =       sizeof(int),
3899                         .mode           =       0644,
3900                         .proc_handler   =       &proc_dointvec,
3901                 },
3902 #ifdef CONFIG_IPV6_ROUTER_PREF
3903                 {
3904                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3905                         .procname       =       "accept_ra_rtr_pref",
3906                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3907                         .maxlen         =       sizeof(int),
3908                         .mode           =       0644,
3909                         .proc_handler   =       &proc_dointvec,
3910                 },
3911                 {
3912                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
3913                         .procname       =       "router_probe_interval",
3914                         .data           =       &ipv6_devconf.rtr_probe_interval,
3915                         .maxlen         =       sizeof(int),
3916                         .mode           =       0644,
3917                         .proc_handler   =       &proc_dointvec_jiffies,
3918                         .strategy       =       &sysctl_jiffies,
3919                 },
3920 #ifdef CONFIG_IPV6_ROUTE_INFO
3921                 {
3922                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3923                         .procname       =       "accept_ra_rt_info_max_plen",
3924                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
3925                         .maxlen         =       sizeof(int),
3926                         .mode           =       0644,
3927                         .proc_handler   =       &proc_dointvec,
3928                 },
3929 #endif
3930 #endif
3931                 {
3932                         .ctl_name       =       NET_IPV6_PROXY_NDP,
3933                         .procname       =       "proxy_ndp",
3934                         .data           =       &ipv6_devconf.proxy_ndp,
3935                         .maxlen         =       sizeof(int),
3936                         .mode           =       0644,
3937                         .proc_handler   =       &proc_dointvec,
3938                 },
3939                 {
3940                         .ctl_name       =       0,      /* sentinel */
3941                 }
3942         },
3943         .addrconf_dev = {
3944                 {
3945                         .ctl_name       =       NET_PROTO_CONF_ALL,
3946                         .procname       =       "all",
3947                         .mode           =       0555,
3948                         .child          =       addrconf_sysctl.addrconf_vars,
3949                 },
3950                 {
3951                         .ctl_name       =       0,      /* sentinel */
3952                 }
3953         },
3954         .addrconf_conf_dir = {
3955                 {
3956                         .ctl_name       =       NET_IPV6_CONF,
3957                         .procname       =       "conf",
3958                         .mode           =       0555,
3959                         .child          =       addrconf_sysctl.addrconf_dev,
3960                 },
3961                 {
3962                         .ctl_name       =       0,      /* sentinel */
3963                 }
3964         },
3965         .addrconf_proto_dir = {
3966                 {
3967                         .ctl_name       =       NET_IPV6,
3968                         .procname       =       "ipv6",
3969                         .mode           =       0555,
3970                         .child          =       addrconf_sysctl.addrconf_conf_dir,
3971                 },
3972                 {
3973                         .ctl_name       =       0,      /* sentinel */
3974                 }
3975         },
3976         .addrconf_root_dir = {
3977                 {
3978                         .ctl_name       =       CTL_NET,
3979                         .procname       =       "net",
3980                         .mode           =       0555,
3981                         .child          =       addrconf_sysctl.addrconf_proto_dir,
3982                 },
3983                 {
3984                         .ctl_name       =       0,      /* sentinel */
3985                 }
3986         },
3987 };
3988
3989 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3990 {
3991         int i;
3992         struct net_device *dev = idev ? idev->dev : NULL;
3993         struct addrconf_sysctl_table *t;
3994         char *dev_name = NULL;
3995
3996         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
3997         if (t == NULL)
3998                 return;
3999         for (i=0; t->addrconf_vars[i].data; i++) {
4000                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4001                 t->addrconf_vars[i].de = NULL;
4002                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4003         }
4004         if (dev) {
4005                 dev_name = dev->name;
4006                 t->addrconf_dev[0].ctl_name = dev->ifindex;
4007         } else {
4008                 dev_name = "default";
4009                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
4010         }
4011
4012         /*
4013          * Make a copy of dev_name, because '.procname' is regarded as const
4014          * by sysctl and we wouldn't want anyone to change it under our feet
4015          * (see SIOCSIFNAME).
4016          */
4017         dev_name = kstrdup(dev_name, GFP_KERNEL);
4018         if (!dev_name)
4019             goto free;
4020
4021         t->addrconf_dev[0].procname = dev_name;
4022
4023         t->addrconf_dev[0].child = t->addrconf_vars;
4024         t->addrconf_dev[0].de = NULL;
4025         t->addrconf_conf_dir[0].child = t->addrconf_dev;
4026         t->addrconf_conf_dir[0].de = NULL;
4027         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4028         t->addrconf_proto_dir[0].de = NULL;
4029         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4030         t->addrconf_root_dir[0].de = NULL;
4031
4032         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir, 0);
4033         if (t->sysctl_header == NULL)
4034                 goto free_procname;
4035         else
4036                 p->sysctl = t;
4037         return;
4038
4039         /* error path */
4040  free_procname:
4041         kfree(dev_name);
4042  free:
4043         kfree(t);
4044
4045         return;
4046 }
4047
4048 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4049 {
4050         if (p->sysctl) {
4051                 struct addrconf_sysctl_table *t = p->sysctl;
4052                 p->sysctl = NULL;
4053                 unregister_sysctl_table(t->sysctl_header);
4054                 kfree(t->addrconf_dev[0].procname);
4055                 kfree(t);
4056         }
4057 }
4058
4059
4060 #endif
4061
4062 /*
4063  *      Device notifier
4064  */
4065
4066 int register_inet6addr_notifier(struct notifier_block *nb)
4067 {
4068         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4069 }
4070
4071 int unregister_inet6addr_notifier(struct notifier_block *nb)
4072 {
4073         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4074 }
4075
4076 /*
4077  *      Init / cleanup code
4078  */
4079
4080 int __init addrconf_init(void)
4081 {
4082         int err = 0;
4083
4084         /* The addrconf netdev notifier requires that loopback_dev
4085          * has it's ipv6 private information allocated and setup
4086          * before it can bring up and give link-local addresses
4087          * to other devices which are up.
4088          *
4089          * Unfortunately, loopback_dev is not necessarily the first
4090          * entry in the global dev_base list of net devices.  In fact,
4091          * it is likely to be the very last entry on that list.
4092          * So this causes the notifier registry below to try and
4093          * give link-local addresses to all devices besides loopback_dev
4094          * first, then loopback_dev, which cases all the non-loopback_dev
4095          * devices to fail to get a link-local address.
4096          *
4097          * So, as a temporary fix, allocate the ipv6 structure for
4098          * loopback_dev first by hand.
4099          * Longer term, all of the dependencies ipv6 has upon the loopback
4100          * device and it being up should be removed.
4101          */
4102         rtnl_lock();
4103         if (!ipv6_add_dev(&loopback_dev))
4104                 err = -ENOMEM;
4105         rtnl_unlock();
4106         if (err)
4107                 return err;
4108
4109         ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4110
4111         register_netdevice_notifier(&ipv6_dev_notf);
4112
4113         addrconf_verify(0);
4114         rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4115 #ifdef CONFIG_SYSCTL
4116         addrconf_sysctl.sysctl_header =
4117                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir, 0);
4118         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4119 #endif
4120
4121         return 0;
4122 }
4123
4124 void __exit addrconf_cleanup(void)
4125 {
4126         struct net_device *dev;
4127         struct inet6_dev *idev;
4128         struct inet6_ifaddr *ifa;
4129         int i;
4130
4131         unregister_netdevice_notifier(&ipv6_dev_notf);
4132
4133         rtnetlink_links[PF_INET6] = NULL;
4134 #ifdef CONFIG_SYSCTL
4135         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4136         addrconf_sysctl_unregister(&ipv6_devconf);
4137 #endif
4138
4139         rtnl_lock();
4140
4141         /*
4142          *      clean dev list.
4143          */
4144
4145         for (dev=dev_base; dev; dev=dev->next) {
4146                 if ((idev = __in6_dev_get(dev)) == NULL)
4147                         continue;
4148                 addrconf_ifdown(dev, 1);
4149         }
4150         addrconf_ifdown(&loopback_dev, 2);
4151
4152         /*
4153          *      Check hash table.
4154          */
4155
4156         write_lock_bh(&addrconf_hash_lock);
4157         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4158                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4159                         struct inet6_ifaddr *bifa;
4160
4161                         bifa = ifa;
4162                         ifa = ifa->lst_next;
4163                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4164                         /* Do not free it; something is wrong.
4165                            Now we can investigate it with debugger.
4166                          */
4167                 }
4168         }
4169         write_unlock_bh(&addrconf_hash_lock);
4170
4171         del_timer(&addr_chk_timer);
4172
4173         rtnl_unlock();
4174
4175 #ifdef CONFIG_PROC_FS
4176         proc_net_remove("if_inet6");
4177 #endif
4178 }