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/sched.h>
  48 #include <linux/net.h>
  49 #include <linux/in6.h>
  50 #include <linux/netdevice.h>
  51 #include <linux/if_addr.h>
  52 #include <linux/if_arp.h>
  53 #include <linux/if_arcnet.h>
  54 #include <linux/if_infiniband.h>
  55 #include <linux/route.h>
  56 #include <linux/inetdevice.h>
  57 #include <linux/init.h>
  58 #ifdef CONFIG_SYSCTL
  59 #include <linux/sysctl.h>
  60 #endif
  61 #include <linux/capability.h>
  62 #include <linux/delay.h>
  63 #include <linux/notifier.h>
  64 #include <linux/string.h>
  65
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85
  86 #include <linux/proc_fs.h>
  87 #include <linux/seq_file.h>
  88
  89 /* Set to 3 to get tracing... */
  90 #define ACONF_DEBUG 2
  91
  92 #if ACONF_DEBUG >= 3
  93 #define ADBG(x) printk x
  94 #else
  95 #define ADBG(x)
  96 #endif
  97
  98 #define INFINITY_LIFE_TIME      0xFFFFFFFF
  99 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 100
 101 #ifdef CONFIG_SYSCTL
 102 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p);
 103 static void addrconf_sysctl_unregister(struct ipv6_devconf *p);
 104 #endif
 105
 106 #ifdef CONFIG_IPV6_PRIVACY
 107 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 108 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 109 static void ipv6_regen_rndid(unsigned long data);
 110
 111 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 112 #endif
 113
 114 static int ipv6_count_addresses(struct inet6_dev *idev);
 115
 116 /*
 117  *      Configured unicast address hash table
 118  */
 119 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 120 static DEFINE_RWLOCK(addrconf_hash_lock);
 121
 122 static void addrconf_verify(unsigned long);
 123
 124 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 125 static DEFINE_SPINLOCK(addrconf_verify_lock);
 126
 127 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 128 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 129
 130 static int addrconf_ifdown(struct net_device *dev, int how);
 131
 132 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 133 static void addrconf_dad_timer(unsigned long data);
 134 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 135 static void addrconf_dad_run(struct inet6_dev *idev);
 136 static void addrconf_rs_timer(unsigned long data);
 137 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 138 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 139
 140 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 141                                 struct prefix_info *pinfo);
 142 static int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev);
 143
 144 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 145
 146 struct ipv6_devconf ipv6_devconf __read_mostly = {
 147         .forwarding             = 0,
 148         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 149         .mtu6                   = IPV6_MIN_MTU,
 150         .accept_ra              = 1,
 151         .accept_redirects       = 1,
 152         .autoconf               = 1,
 153         .force_mld_version      = 0,
 154         .dad_transmits          = 1,
 155         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 156         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 157         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 158 #ifdef CONFIG_IPV6_PRIVACY
 159         .use_tempaddr           = 0,
 160         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 161         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 162         .regen_max_retry        = REGEN_MAX_RETRY,
 163         .max_desync_factor      = MAX_DESYNC_FACTOR,
 164 #endif
 165         .max_addresses          = IPV6_MAX_ADDRESSES,
 166         .accept_ra_defrtr       = 1,
 167         .accept_ra_pinfo        = 1,
 168 #ifdef CONFIG_IPV6_ROUTER_PREF
 169         .accept_ra_rtr_pref     = 1,
 170         .rtr_probe_interval     = 60 * HZ,
 171 #ifdef CONFIG_IPV6_ROUTE_INFO
 172         .accept_ra_rt_info_max_plen = 0,
 173 #endif
 174 #endif
 175         .proxy_ndp              = 0,
 176 };
 177
 178 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 179         .forwarding             = 0,
 180         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 181         .mtu6                   = IPV6_MIN_MTU,
 182         .accept_ra              = 1,
 183         .accept_redirects       = 1,
 184         .autoconf               = 1,
 185         .dad_transmits          = 1,
 186         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 187         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 188         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 189 #ifdef CONFIG_IPV6_PRIVACY
 190         .use_tempaddr           = 0,
 191         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 192         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 193         .regen_max_retry        = REGEN_MAX_RETRY,
 194         .max_desync_factor      = MAX_DESYNC_FACTOR,
 195 #endif
 196         .max_addresses          = IPV6_MAX_ADDRESSES,
 197         .accept_ra_defrtr       = 1,
 198         .accept_ra_pinfo        = 1,
 199 #ifdef CONFIG_IPV6_ROUTER_PREF
 200         .accept_ra_rtr_pref     = 1,
 201         .rtr_probe_interval     = 60 * HZ,
 202 #ifdef CONFIG_IPV6_ROUTE_INFO
 203         .accept_ra_rt_info_max_plen = 0,
 204 #endif
 205 #endif
 206         .proxy_ndp              = 0,
 207 };
 208
 209 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 210 #if 0
 211 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 212 #endif
 213 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 214
 215 #define IPV6_ADDR_SCOPE_TYPE(scope)     ((scope) << 16)
 216
 217 static inline unsigned ipv6_addr_scope2type(unsigned scope)
 218 {
 219         switch(scope) {
 220         case IPV6_ADDR_SCOPE_NODELOCAL:
 221                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_NODELOCAL) |
 222                         IPV6_ADDR_LOOPBACK);
 223         case IPV6_ADDR_SCOPE_LINKLOCAL:
 224                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL) |
 225                         IPV6_ADDR_LINKLOCAL);
 226         case IPV6_ADDR_SCOPE_SITELOCAL:
 227                 return (IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL) |
 228                         IPV6_ADDR_SITELOCAL);
 229         }
 230         return IPV6_ADDR_SCOPE_TYPE(scope);
 231 }
 232
 233 int __ipv6_addr_type(const struct in6_addr *addr)
 234 {
 235         __be32 st;
 236
 237         st = addr->s6_addr32[0];
 238
 239         /* Consider all addresses with the first three bits different of
 240            000 and 111 as unicasts.
 241          */
 242         if ((st & htonl(0xE0000000)) != htonl(0x00000000) &&
 243             (st & htonl(0xE0000000)) != htonl(0xE0000000))
 244                 return (IPV6_ADDR_UNICAST |
 245                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));
 246
 247         if ((st & htonl(0xFF000000)) == htonl(0xFF000000)) {
 248                 /* multicast */
 249                 /* addr-select 3.1 */
 250                 return (IPV6_ADDR_MULTICAST |
 251                         ipv6_addr_scope2type(IPV6_ADDR_MC_SCOPE(addr)));
 252         }
 253
 254         if ((st & htonl(0xFFC00000)) == htonl(0xFE800000))
 255                 return (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST |
 256                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));               /* addr-select 3.1 */
 257         if ((st & htonl(0xFFC00000)) == htonl(0xFEC00000))
 258                 return (IPV6_ADDR_SITELOCAL | IPV6_ADDR_UNICAST |
 259                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_SITELOCAL));               /* addr-select 3.1 */
 260
 261         if ((addr->s6_addr32[0] | addr->s6_addr32[1]) == 0) {
 262                 if (addr->s6_addr32[2] == 0) {
 263                         if (addr->s6_addr32[3] == 0)
 264                                 return IPV6_ADDR_ANY;
 265
 266                         if (addr->s6_addr32[3] == htonl(0x00000001))
 267                                 return (IPV6_ADDR_LOOPBACK | IPV6_ADDR_UNICAST |
 268                                         IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_LINKLOCAL));       /* addr-select 3.4 */
 269
 270                         return (IPV6_ADDR_COMPATv4 | IPV6_ADDR_UNICAST |
 271                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
 272                 }
 273
 274                 if (addr->s6_addr32[2] == htonl(0x0000ffff))
 275                         return (IPV6_ADDR_MAPPED |
 276                                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.3 */
 277         }
 278
 279         return (IPV6_ADDR_RESERVED |
 280                 IPV6_ADDR_SCOPE_TYPE(IPV6_ADDR_SCOPE_GLOBAL));  /* addr-select 3.4 */
 281 }
 282
 283 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 284 {
 285         if (del_timer(&ifp->timer))
 286                 __in6_ifa_put(ifp);
 287 }
 288
 289 enum addrconf_timer_t
 290 {
 291         AC_NONE,
 292         AC_DAD,
 293         AC_RS,
 294 };
 295
 296 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 297                                enum addrconf_timer_t what,
 298                                unsigned long when)
 299 {
 300         if (!del_timer(&ifp->timer))
 301                 in6_ifa_hold(ifp);
 302
 303         switch (what) {
 304         case AC_DAD:
 305                 ifp->timer.function = addrconf_dad_timer;
 306                 break;
 307         case AC_RS:
 308                 ifp->timer.function = addrconf_rs_timer;
 309                 break;
 310         default:;
 311         }
 312         ifp->timer.expires = jiffies + when;
 313         add_timer(&ifp->timer);
 314 }
 315
 316 /* Nobody refers to this device, we may destroy it. */
 317
 318 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 319 {
 320         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 321         kfree(idev);
 322 }
 323
 324 void in6_dev_finish_destroy(struct inet6_dev *idev)
 325 {
 326         struct net_device *dev = idev->dev;
 327         BUG_TRAP(idev->addr_list==NULL);
 328         BUG_TRAP(idev->mc_list==NULL);
 329 #ifdef NET_REFCNT_DEBUG
 330         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 331 #endif
 332         dev_put(dev);
 333         if (!idev->dead) {
 334                 printk("Freeing alive inet6 device %p\n", idev);
 335                 return;
 336         }
 337         snmp6_free_dev(idev);
 338         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 339 }
 340
 341 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 342 {
 343         struct inet6_dev *ndev;
 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         /* protected by rtnl_lock */
 417         rcu_assign_pointer(dev->ip6_ptr, ndev);
 418
 419         ipv6_mc_init_dev(ndev);
 420         ndev->tstamp = jiffies;
 421 #ifdef CONFIG_SYSCTL
 422         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
 423                               NET_IPV6_NEIGH, "ipv6",
 424                               &ndisc_ifinfo_sysctl_change,
 425                               NULL);
 426         addrconf_sysctl_register(ndev, &ndev->cnf);
 427 #endif
 428         return ndev;
 429 }
 430
 431 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 432 {
 433         struct inet6_dev *idev;
 434
 435         ASSERT_RTNL();
 436
 437         if ((idev = __in6_dev_get(dev)) == NULL) {
 438                 if ((idev = ipv6_add_dev(dev)) == NULL)
 439                         return NULL;
 440         }
 441
 442         if (dev->flags&IFF_UP)
 443                 ipv6_mc_up(idev);
 444         return idev;
 445 }
 446
 447 #ifdef CONFIG_SYSCTL
 448 static void dev_forward_change(struct inet6_dev *idev)
 449 {
 450         struct net_device *dev;
 451         struct inet6_ifaddr *ifa;
 452         struct in6_addr addr;
 453
 454         if (!idev)
 455                 return;
 456         dev = idev->dev;
 457         if (dev && (dev->flags & IFF_MULTICAST)) {
 458                 ipv6_addr_all_routers(&addr);
 459
 460                 if (idev->cnf.forwarding)
 461                         ipv6_dev_mc_inc(dev, &addr);
 462                 else
 463                         ipv6_dev_mc_dec(dev, &addr);
 464         }
 465         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 466                 if (idev->cnf.forwarding)
 467                         addrconf_join_anycast(ifa);
 468                 else
 469                         addrconf_leave_anycast(ifa);
 470         }
 471 }
 472
 473
 474 static void addrconf_forward_change(void)
 475 {
 476         struct net_device *dev;
 477         struct inet6_dev *idev;
 478
 479         read_lock(&dev_base_lock);
 480         for (dev=dev_base; dev; dev=dev->next) {
 481                 rcu_read_lock();
 482                 idev = __in6_dev_get(dev);
 483                 if (idev) {
 484                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
 485                         idev->cnf.forwarding = ipv6_devconf.forwarding;
 486                         if (changed)
 487                                 dev_forward_change(idev);
 488                 }
 489                 rcu_read_unlock();
 490         }
 491         read_unlock(&dev_base_lock);
 492 }
 493 #endif
 494
 495 /* Nobody refers to this ifaddr, destroy it */
 496
 497 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 498 {
 499         BUG_TRAP(ifp->if_next==NULL);
 500         BUG_TRAP(ifp->lst_next==NULL);
 501 #ifdef NET_REFCNT_DEBUG
 502         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 503 #endif
 504
 505         in6_dev_put(ifp->idev);
 506
 507         if (del_timer(&ifp->timer))
 508                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 509
 510         if (!ifp->dead) {
 511                 printk("Freeing alive inet6 address %p\n", ifp);
 512                 return;
 513         }
 514         dst_release(&ifp->rt->u.dst);
 515
 516         kfree(ifp);
 517 }
 518
 519 static void
 520 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 521 {
 522         struct inet6_ifaddr *ifa, **ifap;
 523         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 524
 525         /*
 526          * Each device address list is sorted in order of scope -
 527          * global before linklocal.
 528          */
 529         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 530              ifap = &ifa->if_next) {
 531                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 532                         break;
 533         }
 534
 535         ifp->if_next = *ifap;
 536         *ifap = ifp;
 537 }
 538
 539 /* On success it returns ifp with increased reference count */
 540
 541 static struct inet6_ifaddr *
 542 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 543               int scope, u32 flags)
 544 {
 545         struct inet6_ifaddr *ifa = NULL;
 546         struct rt6_info *rt;
 547         int hash;
 548         int err = 0;
 549
 550         rcu_read_lock_bh();
 551         if (idev->dead) {
 552                 err = -ENODEV;                  /*XXX*/
 553                 goto out2;
 554         }
 555
 556         write_lock(&addrconf_hash_lock);
 557
 558         /* Ignore adding duplicate addresses on an interface */
 559         if (ipv6_chk_same_addr(addr, idev->dev)) {
 560                 ADBG(("ipv6_add_addr: already assigned\n"));
 561                 err = -EEXIST;
 562                 goto out;
 563         }
 564
 565         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 566
 567         if (ifa == NULL) {
 568                 ADBG(("ipv6_add_addr: malloc failed\n"));
 569                 err = -ENOBUFS;
 570                 goto out;
 571         }
 572
 573         rt = addrconf_dst_alloc(idev, addr, 0);
 574         if (IS_ERR(rt)) {
 575                 err = PTR_ERR(rt);
 576                 goto out;
 577         }
 578
 579         ipv6_addr_copy(&ifa->addr, addr);
 580
 581         spin_lock_init(&ifa->lock);
 582         init_timer(&ifa->timer);
 583         ifa->timer.data = (unsigned long) ifa;
 584         ifa->scope = scope;
 585         ifa->prefix_len = pfxlen;
 586         ifa->flags = flags | IFA_F_TENTATIVE;
 587         ifa->cstamp = ifa->tstamp = jiffies;
 588
 589         ifa->rt = rt;
 590
 591         ifa->idev = idev;
 592         in6_dev_hold(idev);
 593         /* For caller */
 594         in6_ifa_hold(ifa);
 595
 596         /* Add to big hash table */
 597         hash = ipv6_addr_hash(addr);
 598
 599         ifa->lst_next = inet6_addr_lst[hash];
 600         inet6_addr_lst[hash] = ifa;
 601         in6_ifa_hold(ifa);
 602         write_unlock(&addrconf_hash_lock);
 603
 604         write_lock(&idev->lock);
 605         /* Add to inet6_dev unicast addr list. */
 606         ipv6_link_dev_addr(idev, ifa);
 607
 608 #ifdef CONFIG_IPV6_PRIVACY
 609         if (ifa->flags&IFA_F_TEMPORARY) {
 610                 ifa->tmp_next = idev->tempaddr_list;
 611                 idev->tempaddr_list = ifa;
 612                 in6_ifa_hold(ifa);
 613         }
 614 #endif
 615
 616         in6_ifa_hold(ifa);
 617         write_unlock(&idev->lock);
 618 out2:
 619         rcu_read_unlock_bh();
 620
 621         if (likely(err == 0))
 622                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 623         else {
 624                 kfree(ifa);
 625                 ifa = ERR_PTR(err);
 626         }
 627
 628         return ifa;
 629 out:
 630         write_unlock(&addrconf_hash_lock);
 631         goto out2;
 632 }
 633
 634 /* This function wants to get referenced ifp and releases it before return */
 635
 636 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 637 {
 638         struct inet6_ifaddr *ifa, **ifap;
 639         struct inet6_dev *idev = ifp->idev;
 640         int hash;
 641         int deleted = 0, onlink = 0;
 642         unsigned long expires = jiffies;
 643
 644         hash = ipv6_addr_hash(&ifp->addr);
 645
 646         ifp->dead = 1;
 647
 648         write_lock_bh(&addrconf_hash_lock);
 649         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 650              ifap = &ifa->lst_next) {
 651                 if (ifa == ifp) {
 652                         *ifap = ifa->lst_next;
 653                         __in6_ifa_put(ifp);
 654                         ifa->lst_next = NULL;
 655                         break;
 656                 }
 657         }
 658         write_unlock_bh(&addrconf_hash_lock);
 659
 660         write_lock_bh(&idev->lock);
 661 #ifdef CONFIG_IPV6_PRIVACY
 662         if (ifp->flags&IFA_F_TEMPORARY) {
 663                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 664                      ifap = &ifa->tmp_next) {
 665                         if (ifa == ifp) {
 666                                 *ifap = ifa->tmp_next;
 667                                 if (ifp->ifpub) {
 668                                         in6_ifa_put(ifp->ifpub);
 669                                         ifp->ifpub = NULL;
 670                                 }
 671                                 __in6_ifa_put(ifp);
 672                                 ifa->tmp_next = NULL;
 673                                 break;
 674                         }
 675                 }
 676         }
 677 #endif
 678
 679         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 680                 if (ifa == ifp) {
 681                         *ifap = ifa->if_next;
 682                         __in6_ifa_put(ifp);
 683                         ifa->if_next = NULL;
 684                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 685                                 break;
 686                         deleted = 1;
 687                         continue;
 688                 } else if (ifp->flags & IFA_F_PERMANENT) {
 689                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 690                                               ifp->prefix_len)) {
 691                                 if (ifa->flags & IFA_F_PERMANENT) {
 692                                         onlink = 1;
 693                                         if (deleted)
 694                                                 break;
 695                                 } else {
 696                                         unsigned long lifetime;
 697
 698                                         if (!onlink)
 699                                                 onlink = -1;
 700
 701                                         spin_lock(&ifa->lock);
 702                                         lifetime = min_t(unsigned long,
 703                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 704                                         if (time_before(expires,
 705                                                         ifa->tstamp + lifetime * HZ))
 706                                                 expires = ifa->tstamp + lifetime * HZ;
 707                                         spin_unlock(&ifa->lock);
 708                                 }
 709                         }
 710                 }
 711                 ifap = &ifa->if_next;
 712         }
 713         write_unlock_bh(&idev->lock);
 714
 715         ipv6_ifa_notify(RTM_DELADDR, ifp);
 716
 717         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 718
 719         addrconf_del_timer(ifp);
 720
 721         /*
 722          * Purge or update corresponding prefix
 723          *
 724          * 1) we don't purge prefix here if address was not permanent.
 725          *    prefix is managed by its own lifetime.
 726          * 2) if there're no addresses, delete prefix.
 727          * 3) if there're still other permanent address(es),
 728          *    corresponding prefix is still permanent.
 729          * 4) otherwise, update prefix lifetime to the
 730          *    longest valid lifetime among the corresponding
 731          *    addresses on the device.
 732          *    Note: subsequent RA will update lifetime.
 733          *
 734          * --yoshfuji
 735          */
 736         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 737                 struct in6_addr prefix;
 738                 struct rt6_info *rt;
 739
 740                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 741                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
 742
 743                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 744                         if (onlink == 0) {
 745                                 ip6_del_rt(rt);
 746                                 rt = NULL;
 747                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 748                                 rt->rt6i_expires = expires;
 749                                 rt->rt6i_flags |= RTF_EXPIRES;
 750                         }
 751                 }
 752                 dst_release(&rt->u.dst);
 753         }
 754
 755         in6_ifa_put(ifp);
 756 }
 757
 758 #ifdef CONFIG_IPV6_PRIVACY
 759 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 760 {
 761         struct inet6_dev *idev = ifp->idev;
 762         struct in6_addr addr, *tmpaddr;
 763         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 764         int tmp_plen;
 765         int ret = 0;
 766         int max_addresses;
 767
 768         write_lock(&idev->lock);
 769         if (ift) {
 770                 spin_lock_bh(&ift->lock);
 771                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 772                 spin_unlock_bh(&ift->lock);
 773                 tmpaddr = &addr;
 774         } else {
 775                 tmpaddr = NULL;
 776         }
 777 retry:
 778         in6_dev_hold(idev);
 779         if (idev->cnf.use_tempaddr <= 0) {
 780                 write_unlock(&idev->lock);
 781                 printk(KERN_INFO
 782                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 783                 in6_dev_put(idev);
 784                 ret = -1;
 785                 goto out;
 786         }
 787         spin_lock_bh(&ifp->lock);
 788         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 789                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 790                 spin_unlock_bh(&ifp->lock);
 791                 write_unlock(&idev->lock);
 792                 printk(KERN_WARNING
 793                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 794                 in6_dev_put(idev);
 795                 ret = -1;
 796                 goto out;
 797         }
 798         in6_ifa_hold(ifp);
 799         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 800         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 801                 spin_unlock_bh(&ifp->lock);
 802                 write_unlock(&idev->lock);
 803                 printk(KERN_WARNING
 804                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 805                 in6_ifa_put(ifp);
 806                 in6_dev_put(idev);
 807                 ret = -1;
 808                 goto out;
 809         }
 810         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 811         tmp_valid_lft = min_t(__u32,
 812                               ifp->valid_lft,
 813                               idev->cnf.temp_valid_lft);
 814         tmp_prefered_lft = min_t(__u32,
 815                                  ifp->prefered_lft,
 816                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 817         tmp_plen = ifp->prefix_len;
 818         max_addresses = idev->cnf.max_addresses;
 819         tmp_cstamp = ifp->cstamp;
 820         tmp_tstamp = ifp->tstamp;
 821         spin_unlock_bh(&ifp->lock);
 822
 823         write_unlock(&idev->lock);
 824         ift = !max_addresses ||
 825               ipv6_count_addresses(idev) < max_addresses ?
 826                 ipv6_add_addr(idev, &addr, tmp_plen,
 827                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
 828         if (!ift || IS_ERR(ift)) {
 829                 in6_ifa_put(ifp);
 830                 in6_dev_put(idev);
 831                 printk(KERN_INFO
 832                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 833                 tmpaddr = &addr;
 834                 write_lock(&idev->lock);
 835                 goto retry;
 836         }
 837
 838         spin_lock_bh(&ift->lock);
 839         ift->ifpub = ifp;
 840         ift->valid_lft = tmp_valid_lft;
 841         ift->prefered_lft = tmp_prefered_lft;
 842         ift->cstamp = tmp_cstamp;
 843         ift->tstamp = tmp_tstamp;
 844         spin_unlock_bh(&ift->lock);
 845
 846         addrconf_dad_start(ift, 0);
 847         in6_ifa_put(ift);
 848         in6_dev_put(idev);
 849 out:
 850         return ret;
 851 }
 852 #endif
 853
 854 /*
 855  *      Choose an appropriate source address (RFC3484)
 856  */
 857 struct ipv6_saddr_score {
 858         int             addr_type;
 859         unsigned int    attrs;
 860         int             matchlen;
 861         int             scope;
 862         unsigned int    rule;
 863 };
 864
 865 #define IPV6_SADDR_SCORE_LOCAL          0x0001
 866 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
 867 #define IPV6_SADDR_SCORE_HOA            0x0008
 868 #define IPV6_SADDR_SCORE_OIF            0x0010
 869 #define IPV6_SADDR_SCORE_LABEL          0x0020
 870 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
 871
 872 static int inline ipv6_saddr_preferred(int type)
 873 {
 874         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 875                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 876                 return 1;
 877         return 0;
 878 }
 879
 880 /* static matching label */
 881 static int inline ipv6_saddr_label(const struct in6_addr *addr, int type)
 882 {
 883  /*
 884   *     prefix (longest match)  label
 885   *     -----------------------------
 886   *     ::1/128                 0
 887   *     ::/0                    1
 888   *     2002::/16               2
 889   *     ::/96                   3
 890   *     ::ffff:0:0/96           4
 891   *     fc00::/7                5
 892   *     2001::/32               6
 893   */
 894         if (type & IPV6_ADDR_LOOPBACK)
 895                 return 0;
 896         else if (type & IPV6_ADDR_COMPATv4)
 897                 return 3;
 898         else if (type & IPV6_ADDR_MAPPED)
 899                 return 4;
 900         else if (addr->s6_addr32[0] == htonl(0x20010000))
 901                 return 6;
 902         else if (addr->s6_addr16[0] == htons(0x2002))
 903                 return 2;
 904         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
 905                 return 5;
 906         return 1;
 907 }
 908
 909 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
 910                        struct in6_addr *daddr, struct in6_addr *saddr)
 911 {
 912         struct ipv6_saddr_score hiscore;
 913         struct inet6_ifaddr *ifa_result = NULL;
 914         int daddr_type = __ipv6_addr_type(daddr);
 915         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
 916         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
 917         struct net_device *dev;
 918
 919         memset(&hiscore, 0, sizeof(hiscore));
 920
 921         read_lock(&dev_base_lock);
 922         rcu_read_lock();
 923
 924         for (dev = dev_base; dev; dev=dev->next) {
 925                 struct inet6_dev *idev;
 926                 struct inet6_ifaddr *ifa;
 927
 928                 /* Rule 0: Candidate Source Address (section 4)
 929                  *  - multicast and link-local destination address,
 930                  *    the set of candidate source address MUST only
 931                  *    include addresses assigned to interfaces
 932                  *    belonging to the same link as the outgoing
 933                  *    interface.
 934                  * (- For site-local destination addresses, the
 935                  *    set of candidate source addresses MUST only
 936                  *    include addresses assigned to interfaces
 937                  *    belonging to the same site as the outgoing
 938                  *    interface.)
 939                  */
 940                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
 941                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
 942                     daddr_dev && dev != daddr_dev)
 943                         continue;
 944
 945                 idev = __in6_dev_get(dev);
 946                 if (!idev)
 947                         continue;
 948
 949                 read_lock_bh(&idev->lock);
 950                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
 951                         struct ipv6_saddr_score score;
 952
 953                         score.addr_type = __ipv6_addr_type(&ifa->addr);
 954
 955                         /* Rule 0:
 956                          * - Tentative Address (RFC2462 section 5.4)
 957                          *  - A tentative address is not considered
 958                          *    "assigned to an interface" in the traditional
 959                          *    sense.
 960                          * - Candidate Source Address (section 4)
 961                          *  - In any case, anycast addresses, multicast
 962                          *    addresses, and the unspecified address MUST
 963                          *    NOT be included in a candidate set.
 964                          */
 965                         if (ifa->flags & IFA_F_TENTATIVE)
 966                                 continue;
 967                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
 968                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
 969                                 LIMIT_NETDEBUG(KERN_DEBUG
 970                                                "ADDRCONF: unspecified / multicast address"
 971                                                "assigned as unicast address on %s",
 972                                                dev->name);
 973                                 continue;
 974                         }
 975
 976                         score.attrs = 0;
 977                         score.matchlen = 0;
 978                         score.scope = 0;
 979                         score.rule = 0;
 980
 981                         if (ifa_result == NULL) {
 982                                 /* record it if the first available entry */
 983                                 goto record_it;
 984                         }
 985
 986                         /* Rule 1: Prefer same address */
 987                         if (hiscore.rule < 1) {
 988                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
 989                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
 990                                 hiscore.rule++;
 991                         }
 992                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
 993                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
 994                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
 995                                         score.rule = 1;
 996                                         goto record_it;
 997                                 }
 998                         } else {
 999                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
1000                                         continue;
1001                         }
1002
1003                         /* Rule 2: Prefer appropriate scope */
1004                         if (hiscore.rule < 2) {
1005                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1006                                 hiscore.rule++;
1007                         }
1008                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1009                         if (hiscore.scope < score.scope) {
1010                                 if (hiscore.scope < daddr_scope) {
1011                                         score.rule = 2;
1012                                         goto record_it;
1013                                 } else
1014                                         continue;
1015                         } else if (score.scope < hiscore.scope) {
1016                                 if (score.scope < daddr_scope)
1017                                         break; /* addresses sorted by scope */
1018                                 else {
1019                                         score.rule = 2;
1020                                         goto record_it;
1021                                 }
1022                         }
1023
1024                         /* Rule 3: Avoid deprecated address */
1025                         if (hiscore.rule < 3) {
1026                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1027                                     !(ifa_result->flags & IFA_F_DEPRECATED))
1028                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1029                                 hiscore.rule++;
1030                         }
1031                         if (ipv6_saddr_preferred(score.addr_type) ||
1032                             !(ifa->flags & IFA_F_DEPRECATED)) {
1033                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1034                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1035                                         score.rule = 3;
1036                                         goto record_it;
1037                                 }
1038                         } else {
1039                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1040                                         continue;
1041                         }
1042
1043                         /* Rule 4: Prefer home address */
1044 #ifdef CONFIG_IPV6_MIP6
1045                         if (hiscore.rule < 4) {
1046                                 if (ifa_result->flags & IFA_F_HOMEADDRESS)
1047                                         hiscore.attrs |= IPV6_SADDR_SCORE_HOA;
1048                                 hiscore.rule++;
1049                         }
1050                         if (ifa->flags & IFA_F_HOMEADDRESS) {
1051                                 score.attrs |= IPV6_SADDR_SCORE_HOA;
1052                                 if (!(ifa_result->flags & IFA_F_HOMEADDRESS)) {
1053                                         score.rule = 4;
1054                                         goto record_it;
1055                                 }
1056                         } else {
1057                                 if (hiscore.attrs & IPV6_SADDR_SCORE_HOA)
1058                                         continue;
1059                         }
1060 #else
1061                         if (hiscore.rule < 4)
1062                                 hiscore.rule++;
1063 #endif
1064
1065                         /* Rule 5: Prefer outgoing interface */
1066                         if (hiscore.rule < 5) {
1067                                 if (daddr_dev == NULL ||
1068                                     daddr_dev == ifa_result->idev->dev)
1069                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1070                                 hiscore.rule++;
1071                         }
1072                         if (daddr_dev == NULL ||
1073                             daddr_dev == ifa->idev->dev) {
1074                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1075                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1076                                         score.rule = 5;
1077                                         goto record_it;
1078                                 }
1079                         } else {
1080                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1081                                         continue;
1082                         }
1083
1084                         /* Rule 6: Prefer matching label */
1085                         if (hiscore.rule < 6) {
1086                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1087                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1088                                 hiscore.rule++;
1089                         }
1090                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1091                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1092                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1093                                         score.rule = 6;
1094                                         goto record_it;
1095                                 }
1096                         } else {
1097                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1098                                         continue;
1099                         }
1100
1101 #ifdef CONFIG_IPV6_PRIVACY
1102                         /* Rule 7: Prefer public address
1103                          * Note: prefer temprary address if use_tempaddr >= 2
1104                          */
1105                         if (hiscore.rule < 7) {
1106                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1107                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1108                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1109                                 hiscore.rule++;
1110                         }
1111                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1112                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1113                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1114                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1115                                         score.rule = 7;
1116                                         goto record_it;
1117                                 }
1118                         } else {
1119                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1120                                         continue;
1121                         }
1122 #else
1123                         if (hiscore.rule < 7)
1124                                 hiscore.rule++;
1125 #endif
1126                         /* Rule 8: Use longest matching prefix */
1127                         if (hiscore.rule < 8) {
1128                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1129                                 hiscore.rule++;
1130                         }
1131                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1132                         if (score.matchlen > hiscore.matchlen) {
1133                                 score.rule = 8;
1134                                 goto record_it;
1135                         }
1136 #if 0
1137                         else if (score.matchlen < hiscore.matchlen)
1138                                 continue;
1139 #endif
1140
1141                         /* Final Rule: choose first available one */
1142                         continue;
1143 record_it:
1144                         if (ifa_result)
1145                                 in6_ifa_put(ifa_result);
1146                         in6_ifa_hold(ifa);
1147                         ifa_result = ifa;
1148                         hiscore = score;
1149                 }
1150                 read_unlock_bh(&idev->lock);
1151         }
1152         rcu_read_unlock();
1153         read_unlock(&dev_base_lock);
1154
1155         if (!ifa_result)
1156                 return -EADDRNOTAVAIL;
1157
1158         ipv6_addr_copy(saddr, &ifa_result->addr);
1159         in6_ifa_put(ifa_result);
1160         return 0;
1161 }
1162
1163
1164 int ipv6_get_saddr(struct dst_entry *dst,
1165                    struct in6_addr *daddr, struct in6_addr *saddr)
1166 {
1167         return ipv6_dev_get_saddr(dst ? ip6_dst_idev(dst)->dev : NULL, daddr, saddr);
1168 }
1169
1170
1171 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1172 {
1173         struct inet6_dev *idev;
1174         int err = -EADDRNOTAVAIL;
1175
1176         rcu_read_lock();
1177         if ((idev = __in6_dev_get(dev)) != NULL) {
1178                 struct inet6_ifaddr *ifp;
1179
1180                 read_lock_bh(&idev->lock);
1181                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1182                         if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1183                                 ipv6_addr_copy(addr, &ifp->addr);
1184                                 err = 0;
1185                                 break;
1186                         }
1187                 }
1188                 read_unlock_bh(&idev->lock);
1189         }
1190         rcu_read_unlock();
1191         return err;
1192 }
1193
1194 static int ipv6_count_addresses(struct inet6_dev *idev)
1195 {
1196         int cnt = 0;
1197         struct inet6_ifaddr *ifp;
1198
1199         read_lock_bh(&idev->lock);
1200         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1201                 cnt++;
1202         read_unlock_bh(&idev->lock);
1203         return cnt;
1204 }
1205
1206 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1207 {
1208         struct inet6_ifaddr * ifp;
1209         u8 hash = ipv6_addr_hash(addr);
1210
1211         read_lock_bh(&addrconf_hash_lock);
1212         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1213                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1214                     !(ifp->flags&IFA_F_TENTATIVE)) {
1215                         if (dev == NULL || ifp->idev->dev == dev ||
1216                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1217                                 break;
1218                 }
1219         }
1220         read_unlock_bh(&addrconf_hash_lock);
1221         return ifp != NULL;
1222 }
1223
1224 static
1225 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1226 {
1227         struct inet6_ifaddr * ifp;
1228         u8 hash = ipv6_addr_hash(addr);
1229
1230         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1231                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1232                         if (dev == NULL || ifp->idev->dev == dev)
1233                                 break;
1234                 }
1235         }
1236         return ifp != NULL;
1237 }
1238
1239 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1240 {
1241         struct inet6_ifaddr * ifp;
1242         u8 hash = ipv6_addr_hash(addr);
1243
1244         read_lock_bh(&addrconf_hash_lock);
1245         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1246                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1247                         if (dev == NULL || ifp->idev->dev == dev ||
1248                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1249                                 in6_ifa_hold(ifp);
1250                                 break;
1251                         }
1252                 }
1253         }
1254         read_unlock_bh(&addrconf_hash_lock);
1255
1256         return ifp;
1257 }
1258
1259 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1260 {
1261         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1262         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1263         __be32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1264         __be32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1265         int sk_ipv6only = ipv6_only_sock(sk);
1266         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1267         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1268         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1269
1270         if (!sk2_rcv_saddr && !sk_ipv6only)
1271                 return 1;
1272
1273         if (addr_type2 == IPV6_ADDR_ANY &&
1274             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1275                 return 1;
1276
1277         if (addr_type == IPV6_ADDR_ANY &&
1278             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1279                 return 1;
1280
1281         if (sk2_rcv_saddr6 &&
1282             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1283                 return 1;
1284
1285         if (addr_type == IPV6_ADDR_MAPPED &&
1286             !sk2_ipv6only &&
1287             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1288                 return 1;
1289
1290         return 0;
1291 }
1292
1293 /* Gets referenced address, destroys ifaddr */
1294
1295 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1296 {
1297         if (ifp->flags&IFA_F_PERMANENT) {
1298                 spin_lock_bh(&ifp->lock);
1299                 addrconf_del_timer(ifp);
1300                 ifp->flags |= IFA_F_TENTATIVE;
1301                 spin_unlock_bh(&ifp->lock);
1302                 in6_ifa_put(ifp);
1303 #ifdef CONFIG_IPV6_PRIVACY
1304         } else if (ifp->flags&IFA_F_TEMPORARY) {
1305                 struct inet6_ifaddr *ifpub;
1306                 spin_lock_bh(&ifp->lock);
1307                 ifpub = ifp->ifpub;
1308                 if (ifpub) {
1309                         in6_ifa_hold(ifpub);
1310                         spin_unlock_bh(&ifp->lock);
1311                         ipv6_create_tempaddr(ifpub, ifp);
1312                         in6_ifa_put(ifpub);
1313                 } else {
1314                         spin_unlock_bh(&ifp->lock);
1315                 }
1316                 ipv6_del_addr(ifp);
1317 #endif
1318         } else
1319                 ipv6_del_addr(ifp);
1320 }
1321
1322 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1323 {
1324         if (net_ratelimit())
1325                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1326         addrconf_dad_stop(ifp);
1327 }
1328
1329 /* Join to solicited addr multicast group. */
1330
1331 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1332 {
1333         struct in6_addr maddr;
1334
1335         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1336                 return;
1337
1338         addrconf_addr_solict_mult(addr, &maddr);
1339         ipv6_dev_mc_inc(dev, &maddr);
1340 }
1341
1342 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1343 {
1344         struct in6_addr maddr;
1345
1346         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1347                 return;
1348
1349         addrconf_addr_solict_mult(addr, &maddr);
1350         __ipv6_dev_mc_dec(idev, &maddr);
1351 }
1352
1353 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1354 {
1355         struct in6_addr addr;
1356         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1357         if (ipv6_addr_any(&addr))
1358                 return;
1359         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1360 }
1361
1362 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1363 {
1364         struct in6_addr addr;
1365         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1366         if (ipv6_addr_any(&addr))
1367                 return;
1368         __ipv6_dev_ac_dec(ifp->idev, &addr);
1369 }
1370
1371 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1372 {
1373         if (dev->addr_len != ETH_ALEN)
1374                 return -1;
1375         memcpy(eui, dev->dev_addr, 3);
1376         memcpy(eui + 5, dev->dev_addr + 3, 3);
1377
1378         /*
1379          * The zSeries OSA network cards can be shared among various
1380          * OS instances, but the OSA cards have only one MAC address.
1381          * This leads to duplicate address conflicts in conjunction
1382          * with IPv6 if more than one instance uses the same card.
1383          *
1384          * The driver for these cards can deliver a unique 16-bit
1385          * identifier for each instance sharing the same card.  It is
1386          * placed instead of 0xFFFE in the interface identifier.  The
1387          * "u" bit of the interface identifier is not inverted in this
1388          * case.  Hence the resulting interface identifier has local
1389          * scope according to RFC2373.
1390          */
1391         if (dev->dev_id) {
1392                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1393                 eui[4] = dev->dev_id & 0xFF;
1394         } else {
1395                 eui[3] = 0xFF;
1396                 eui[4] = 0xFE;
1397                 eui[0] ^= 2;
1398         }
1399         return 0;
1400 }
1401
1402 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1403 {
1404         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1405         if (dev->addr_len != ARCNET_ALEN)
1406                 return -1;
1407         memset(eui, 0, 7);
1408         eui[7] = *(u8*)dev->dev_addr;
1409         return 0;
1410 }
1411
1412 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1413 {
1414         if (dev->addr_len != INFINIBAND_ALEN)
1415                 return -1;
1416         memcpy(eui, dev->dev_addr + 12, 8);
1417         eui[0] |= 2;
1418         return 0;
1419 }
1420
1421 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1422 {
1423         switch (dev->type) {
1424         case ARPHRD_ETHER:
1425         case ARPHRD_FDDI:
1426         case ARPHRD_IEEE802_TR:
1427                 return addrconf_ifid_eui48(eui, dev);
1428         case ARPHRD_ARCNET:
1429                 return addrconf_ifid_arcnet(eui, dev);
1430         case ARPHRD_INFINIBAND:
1431                 return addrconf_ifid_infiniband(eui, dev);
1432         }
1433         return -1;
1434 }
1435
1436 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1437 {
1438         int err = -1;
1439         struct inet6_ifaddr *ifp;
1440
1441         read_lock_bh(&idev->lock);
1442         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1443                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1444                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1445                         err = 0;
1446                         break;
1447                 }
1448         }
1449         read_unlock_bh(&idev->lock);
1450         return err;
1451 }
1452
1453 #ifdef CONFIG_IPV6_PRIVACY
1454 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1455 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1456 {
1457 regen:
1458         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1459         idev->rndid[0] &= ~0x02;
1460
1461         /*
1462          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1463          * check if generated address is not inappropriate
1464          *
1465          *  - Reserved subnet anycast (RFC 2526)
1466          *      11111101 11....11 1xxxxxxx
1467          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1468          *      00-00-5E-FE-xx-xx-xx-xx
1469          *  - value 0
1470          *  - XXX: already assigned to an address on the device
1471          */
1472         if (idev->rndid[0] == 0xfd &&
1473             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1474             (idev->rndid[7]&0x80))
1475                 goto regen;
1476         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1477                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1478                         goto regen;
1479                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1480                         goto regen;
1481         }
1482
1483         return 0;
1484 }
1485
1486 static void ipv6_regen_rndid(unsigned long data)
1487 {
1488         struct inet6_dev *idev = (struct inet6_dev *) data;
1489         unsigned long expires;
1490
1491         rcu_read_lock_bh();
1492         write_lock_bh(&idev->lock);
1493
1494         if (idev->dead)
1495                 goto out;
1496
1497         if (__ipv6_regen_rndid(idev) < 0)
1498                 goto out;
1499
1500         expires = jiffies +
1501                 idev->cnf.temp_prefered_lft * HZ -
1502                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1503         if (time_before(expires, jiffies)) {
1504                 printk(KERN_WARNING
1505                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1506                         idev->dev->name);
1507                 goto out;
1508         }
1509
1510         if (!mod_timer(&idev->regen_timer, expires))
1511                 in6_dev_hold(idev);
1512
1513 out:
1514         write_unlock_bh(&idev->lock);
1515         rcu_read_unlock_bh();
1516         in6_dev_put(idev);
1517 }
1518
1519 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1520         int ret = 0;
1521
1522         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1523                 ret = __ipv6_regen_rndid(idev);
1524         return ret;
1525 }
1526 #endif
1527
1528 /*
1529  *      Add prefix route.
1530  */
1531
1532 static void
1533 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1534                       unsigned long expires, u32 flags)
1535 {
1536         struct fib6_config cfg = {
1537                 .fc_table = RT6_TABLE_PREFIX,
1538                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1539                 .fc_ifindex = dev->ifindex,
1540                 .fc_expires = expires,
1541                 .fc_dst_len = plen,
1542                 .fc_flags = RTF_UP | flags,
1543         };
1544
1545         ipv6_addr_copy(&cfg.fc_dst, pfx);
1546
1547         /* Prevent useless cloning on PtP SIT.
1548            This thing is done here expecting that the whole
1549            class of non-broadcast devices need not cloning.
1550          */
1551 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1552         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1553                 cfg.fc_flags |= RTF_NONEXTHOP;
1554 #endif
1555
1556         ip6_route_add(&cfg);
1557 }
1558
1559 /* Create "default" multicast route to the interface */
1560
1561 static void addrconf_add_mroute(struct net_device *dev)
1562 {
1563         struct fib6_config cfg = {
1564                 .fc_table = RT6_TABLE_LOCAL,
1565                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1566                 .fc_ifindex = dev->ifindex,
1567                 .fc_dst_len = 8,
1568                 .fc_flags = RTF_UP,
1569         };
1570
1571         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1572
1573         ip6_route_add(&cfg);
1574 }
1575
1576 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1577 static void sit_route_add(struct net_device *dev)
1578 {
1579         struct fib6_config cfg = {
1580                 .fc_table = RT6_TABLE_MAIN,
1581                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1582                 .fc_ifindex = dev->ifindex,
1583                 .fc_dst_len = 96,
1584                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1585         };
1586
1587         /* prefix length - 96 bits "::d.d.d.d" */
1588         ip6_route_add(&cfg);
1589 }
1590 #endif
1591
1592 static void addrconf_add_lroute(struct net_device *dev)
1593 {
1594         struct in6_addr addr;
1595
1596         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1597         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1598 }
1599
1600 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1601 {
1602         struct inet6_dev *idev;
1603
1604         ASSERT_RTNL();
1605
1606         if ((idev = ipv6_find_idev(dev)) == NULL)
1607                 return NULL;
1608
1609         /* Add default multicast route */
1610         addrconf_add_mroute(dev);
1611
1612         /* Add link local route */
1613         addrconf_add_lroute(dev);
1614         return idev;
1615 }
1616
1617 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1618 {
1619         struct prefix_info *pinfo;
1620         __u32 valid_lft;
1621         __u32 prefered_lft;
1622         int addr_type;
1623         unsigned long rt_expires;
1624         struct inet6_dev *in6_dev;
1625
1626         pinfo = (struct prefix_info *) opt;
1627
1628         if (len < sizeof(struct prefix_info)) {
1629                 ADBG(("addrconf: prefix option too short\n"));
1630                 return;
1631         }
1632
1633         /*
1634          *      Validation checks ([ADDRCONF], page 19)
1635          */
1636
1637         addr_type = ipv6_addr_type(&pinfo->prefix);
1638
1639         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1640                 return;
1641
1642         valid_lft = ntohl(pinfo->valid);
1643         prefered_lft = ntohl(pinfo->prefered);
1644
1645         if (prefered_lft > valid_lft) {
1646                 if (net_ratelimit())
1647                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1648                 return;
1649         }
1650
1651         in6_dev = in6_dev_get(dev);
1652
1653         if (in6_dev == NULL) {
1654                 if (net_ratelimit())
1655                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1656                 return;
1657         }
1658
1659         /*
1660          *      Two things going on here:
1661          *      1) Add routes for on-link prefixes
1662          *      2) Configure prefixes with the auto flag set
1663          */
1664
1665         /* Avoid arithmetic overflow. Really, we could
1666            save rt_expires in seconds, likely valid_lft,
1667            but it would require division in fib gc, that it
1668            not good.
1669          */
1670         if (valid_lft >= 0x7FFFFFFF/HZ)
1671                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1672         else
1673                 rt_expires = valid_lft * HZ;
1674
1675         /*
1676          * We convert this (in jiffies) to clock_t later.
1677          * Avoid arithmetic overflow there as well.
1678          * Overflow can happen only if HZ < USER_HZ.
1679          */
1680         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1681                 rt_expires = 0x7FFFFFFF / USER_HZ;
1682
1683         if (pinfo->onlink) {
1684                 struct rt6_info *rt;
1685                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1686
1687                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1688                         if (rt->rt6i_flags&RTF_EXPIRES) {
1689                                 if (valid_lft == 0) {
1690                                         ip6_del_rt(rt);
1691                                         rt = NULL;
1692                                 } else {
1693                                         rt->rt6i_expires = jiffies + rt_expires;
1694                                 }
1695                         }
1696                 } else if (valid_lft) {
1697                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1698                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1699                 }
1700                 if (rt)
1701                         dst_release(&rt->u.dst);
1702         }
1703
1704         /* Try to figure out our local address for this prefix */
1705
1706         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1707                 struct inet6_ifaddr * ifp;
1708                 struct in6_addr addr;
1709                 int create = 0, update_lft = 0;
1710
1711                 if (pinfo->prefix_len == 64) {
1712                         memcpy(&addr, &pinfo->prefix, 8);
1713                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1714                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1715                                 in6_dev_put(in6_dev);
1716                                 return;
1717                         }
1718                         goto ok;
1719                 }
1720                 if (net_ratelimit())
1721                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1722                                pinfo->prefix_len);
1723                 in6_dev_put(in6_dev);
1724                 return;
1725
1726 ok:
1727
1728                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1729
1730                 if (ifp == NULL && valid_lft) {
1731                         int max_addresses = in6_dev->cnf.max_addresses;
1732
1733                         /* Do not allow to create too much of autoconfigured
1734                          * addresses; this would be too easy way to crash kernel.
1735                          */
1736                         if (!max_addresses ||
1737                             ipv6_count_addresses(in6_dev) < max_addresses)
1738                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1739                                                     addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1740
1741                         if (!ifp || IS_ERR(ifp)) {
1742                                 in6_dev_put(in6_dev);
1743                                 return;
1744                         }
1745
1746                         update_lft = create = 1;
1747                         ifp->cstamp = jiffies;
1748                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1749                 }
1750
1751                 if (ifp) {
1752                         int flags;
1753                         unsigned long now;
1754 #ifdef CONFIG_IPV6_PRIVACY
1755                         struct inet6_ifaddr *ift;
1756 #endif
1757                         u32 stored_lft;
1758
1759                         /* update lifetime (RFC2462 5.5.3 e) */
1760                         spin_lock(&ifp->lock);
1761                         now = jiffies;
1762                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1763                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1764                         else
1765                                 stored_lft = 0;
1766                         if (!update_lft && stored_lft) {
1767                                 if (valid_lft > MIN_VALID_LIFETIME ||
1768                                     valid_lft > stored_lft)
1769                                         update_lft = 1;
1770                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1771                                         /* valid_lft <= stored_lft is always true */
1772                                         /* XXX: IPsec */
1773                                         update_lft = 0;
1774                                 } else {
1775                                         valid_lft = MIN_VALID_LIFETIME;
1776                                         if (valid_lft < prefered_lft)
1777                                                 prefered_lft = valid_lft;
1778                                         update_lft = 1;
1779                                 }
1780                         }
1781
1782                         if (update_lft) {
1783                                 ifp->valid_lft = valid_lft;
1784                                 ifp->prefered_lft = prefered_lft;
1785                                 ifp->tstamp = now;
1786                                 flags = ifp->flags;
1787                                 ifp->flags &= ~IFA_F_DEPRECATED;
1788                                 spin_unlock(&ifp->lock);
1789
1790                                 if (!(flags&IFA_F_TENTATIVE))
1791                                         ipv6_ifa_notify(0, ifp);
1792                         } else
1793                                 spin_unlock(&ifp->lock);
1794
1795 #ifdef CONFIG_IPV6_PRIVACY
1796                         read_lock_bh(&in6_dev->lock);
1797                         /* update all temporary addresses in the list */
1798                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1799                                 /*
1800                                  * When adjusting the lifetimes of an existing
1801                                  * temporary address, only lower the lifetimes.
1802                                  * Implementations must not increase the
1803                                  * lifetimes of an existing temporary address
1804                                  * when processing a Prefix Information Option.
1805                                  */
1806                                 spin_lock(&ift->lock);
1807                                 flags = ift->flags;
1808                                 if (ift->valid_lft > valid_lft &&
1809                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1810                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1811                                 if (ift->prefered_lft > prefered_lft &&
1812                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1813                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1814                                 spin_unlock(&ift->lock);
1815                                 if (!(flags&IFA_F_TENTATIVE))
1816                                         ipv6_ifa_notify(0, ift);
1817                         }
1818
1819                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1820                                 /*
1821                                  * When a new public address is created as described in [ADDRCONF],
1822                                  * also create a new temporary address.
1823                                  */
1824                                 read_unlock_bh(&in6_dev->lock);
1825                                 ipv6_create_tempaddr(ifp, NULL);
1826                         } else {
1827                                 read_unlock_bh(&in6_dev->lock);
1828                         }
1829 #endif
1830                         in6_ifa_put(ifp);
1831                         addrconf_verify(0);
1832                 }
1833         }
1834         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1835         in6_dev_put(in6_dev);
1836 }
1837
1838 /*
1839  *      Set destination address.
1840  *      Special case for SIT interfaces where we create a new "virtual"
1841  *      device.
1842  */
1843 int addrconf_set_dstaddr(void __user *arg)
1844 {
1845         struct in6_ifreq ireq;
1846         struct net_device *dev;
1847         int err = -EINVAL;
1848
1849         rtnl_lock();
1850
1851         err = -EFAULT;
1852         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1853                 goto err_exit;
1854
1855         dev = __dev_get_by_index(ireq.ifr6_ifindex);
1856
1857         err = -ENODEV;
1858         if (dev == NULL)
1859                 goto err_exit;
1860
1861 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1862         if (dev->type == ARPHRD_SIT) {
1863                 struct ifreq ifr;
1864                 mm_segment_t    oldfs;
1865                 struct ip_tunnel_parm p;
1866
1867                 err = -EADDRNOTAVAIL;
1868                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1869                         goto err_exit;
1870
1871                 memset(&p, 0, sizeof(p));
1872                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1873                 p.iph.saddr = 0;
1874                 p.iph.version = 4;
1875                 p.iph.ihl = 5;
1876                 p.iph.protocol = IPPROTO_IPV6;
1877                 p.iph.ttl = 64;
1878                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1879
1880                 oldfs = get_fs(); set_fs(KERNEL_DS);
1881                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1882                 set_fs(oldfs);
1883
1884                 if (err == 0) {
1885                         err = -ENOBUFS;
1886                         if ((dev = __dev_get_by_name(p.name)) == NULL)
1887                                 goto err_exit;
1888                         err = dev_open(dev);
1889                 }
1890         }
1891 #endif
1892
1893 err_exit:
1894         rtnl_unlock();
1895         return err;
1896 }
1897
1898 /*
1899  *      Manual configuration of address on an interface
1900  */
1901 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1902                           __u8 ifa_flags, __u32 prefered_lft, __u32 valid_lft)
1903 {
1904         struct inet6_ifaddr *ifp;
1905         struct inet6_dev *idev;
1906         struct net_device *dev;
1907         int scope;
1908
1909         ASSERT_RTNL();
1910
1911         /* check the lifetime */
1912         if (!valid_lft || prefered_lft > valid_lft)
1913                 return -EINVAL;
1914
1915         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1916                 return -ENODEV;
1917
1918         if ((idev = addrconf_add_dev(dev)) == NULL)
1919                 return -ENOBUFS;
1920
1921         scope = ipv6_addr_scope(pfx);
1922
1923         if (valid_lft == INFINITY_LIFE_TIME)
1924                 ifa_flags |= IFA_F_PERMANENT;
1925         else if (valid_lft >= 0x7FFFFFFF/HZ)
1926                 valid_lft = 0x7FFFFFFF/HZ;
1927
1928         if (prefered_lft == 0)
1929                 ifa_flags |= IFA_F_DEPRECATED;
1930         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1931                  (prefered_lft != INFINITY_LIFE_TIME))
1932                 prefered_lft = 0x7FFFFFFF/HZ;
1933
1934         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1935
1936         if (!IS_ERR(ifp)) {
1937                 spin_lock_bh(&ifp->lock);
1938                 ifp->valid_lft = valid_lft;
1939                 ifp->prefered_lft = prefered_lft;
1940                 ifp->tstamp = jiffies;
1941                 spin_unlock_bh(&ifp->lock);
1942
1943                 addrconf_dad_start(ifp, 0);
1944                 in6_ifa_put(ifp);
1945                 addrconf_verify(0);
1946                 return 0;
1947         }
1948
1949         return PTR_ERR(ifp);
1950 }
1951
1952 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1953 {
1954         struct inet6_ifaddr *ifp;
1955         struct inet6_dev *idev;
1956         struct net_device *dev;
1957
1958         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1959                 return -ENODEV;
1960
1961         if ((idev = __in6_dev_get(dev)) == NULL)
1962                 return -ENXIO;
1963
1964         read_lock_bh(&idev->lock);
1965         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1966                 if (ifp->prefix_len == plen &&
1967                     ipv6_addr_equal(pfx, &ifp->addr)) {
1968                         in6_ifa_hold(ifp);
1969                         read_unlock_bh(&idev->lock);
1970
1971                         ipv6_del_addr(ifp);
1972
1973                         /* If the last address is deleted administratively,
1974                            disable IPv6 on this interface.
1975                          */
1976                         if (idev->addr_list == NULL)
1977                                 addrconf_ifdown(idev->dev, 1);
1978                         return 0;
1979                 }
1980         }
1981         read_unlock_bh(&idev->lock);
1982         return -EADDRNOTAVAIL;
1983 }
1984
1985
1986 int addrconf_add_ifaddr(void __user *arg)
1987 {
1988         struct in6_ifreq ireq;
1989         int err;
1990
1991         if (!capable(CAP_NET_ADMIN))
1992                 return -EPERM;
1993
1994         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1995                 return -EFAULT;
1996
1997         rtnl_lock();
1998         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
1999                              IFA_F_PERMANENT, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2000         rtnl_unlock();
2001         return err;
2002 }
2003
2004 int addrconf_del_ifaddr(void __user *arg)
2005 {
2006         struct in6_ifreq ireq;
2007         int err;
2008
2009         if (!capable(CAP_NET_ADMIN))
2010                 return -EPERM;
2011
2012         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2013                 return -EFAULT;
2014
2015         rtnl_lock();
2016         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
2017         rtnl_unlock();
2018         return err;
2019 }
2020
2021 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2022 static void sit_add_v4_addrs(struct inet6_dev *idev)
2023 {
2024         struct inet6_ifaddr * ifp;
2025         struct in6_addr addr;
2026         struct net_device *dev;
2027         int scope;
2028
2029         ASSERT_RTNL();
2030
2031         memset(&addr, 0, sizeof(struct in6_addr));
2032         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2033
2034         if (idev->dev->flags&IFF_POINTOPOINT) {
2035                 addr.s6_addr32[0] = htonl(0xfe800000);
2036                 scope = IFA_LINK;
2037         } else {
2038                 scope = IPV6_ADDR_COMPATv4;
2039         }
2040
2041         if (addr.s6_addr32[3]) {
2042                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2043                 if (!IS_ERR(ifp)) {
2044                         spin_lock_bh(&ifp->lock);
2045                         ifp->flags &= ~IFA_F_TENTATIVE;
2046                         spin_unlock_bh(&ifp->lock);
2047                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2048                         in6_ifa_put(ifp);
2049                 }
2050                 return;
2051         }
2052
2053         for (dev = dev_base; dev != NULL; dev = dev->next) {
2054                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2055                 if (in_dev && (dev->flags & IFF_UP)) {
2056                         struct in_ifaddr * ifa;
2057
2058                         int flag = scope;
2059
2060                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2061                                 int plen;
2062
2063                                 addr.s6_addr32[3] = ifa->ifa_local;
2064
2065                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2066                                         continue;
2067                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2068                                         if (idev->dev->flags&IFF_POINTOPOINT)
2069                                                 continue;
2070                                         flag |= IFA_HOST;
2071                                 }
2072                                 if (idev->dev->flags&IFF_POINTOPOINT)
2073                                         plen = 64;
2074                                 else
2075                                         plen = 96;
2076
2077                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2078                                                     IFA_F_PERMANENT);
2079                                 if (!IS_ERR(ifp)) {
2080                                         spin_lock_bh(&ifp->lock);
2081                                         ifp->flags &= ~IFA_F_TENTATIVE;
2082                                         spin_unlock_bh(&ifp->lock);
2083                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2084                                         in6_ifa_put(ifp);
2085                                 }
2086                         }
2087                 }
2088         }
2089 }
2090 #endif
2091
2092 static void init_loopback(struct net_device *dev)
2093 {
2094         struct inet6_dev  *idev;
2095         struct inet6_ifaddr * ifp;
2096
2097         /* ::1 */
2098
2099         ASSERT_RTNL();
2100
2101         if ((idev = ipv6_find_idev(dev)) == NULL) {
2102                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2103                 return;
2104         }
2105
2106         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2107         if (!IS_ERR(ifp)) {
2108                 spin_lock_bh(&ifp->lock);
2109                 ifp->flags &= ~IFA_F_TENTATIVE;
2110                 spin_unlock_bh(&ifp->lock);
2111                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2112                 in6_ifa_put(ifp);
2113         }
2114 }
2115
2116 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2117 {
2118         struct inet6_ifaddr * ifp;
2119
2120         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2121         if (!IS_ERR(ifp)) {
2122                 addrconf_dad_start(ifp, 0);
2123                 in6_ifa_put(ifp);
2124         }
2125 }
2126
2127 static void addrconf_dev_config(struct net_device *dev)
2128 {
2129         struct in6_addr addr;
2130         struct inet6_dev    * idev;
2131
2132         ASSERT_RTNL();
2133
2134         if ((dev->type != ARPHRD_ETHER) &&
2135             (dev->type != ARPHRD_FDDI) &&
2136             (dev->type != ARPHRD_IEEE802_TR) &&
2137             (dev->type != ARPHRD_ARCNET) &&
2138             (dev->type != ARPHRD_INFINIBAND)) {
2139                 /* Alas, we support only Ethernet autoconfiguration. */
2140                 return;
2141         }
2142
2143         idev = addrconf_add_dev(dev);
2144         if (idev == NULL)
2145                 return;
2146
2147         memset(&addr, 0, sizeof(struct in6_addr));
2148         addr.s6_addr32[0] = htonl(0xFE800000);
2149
2150         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2151                 addrconf_add_linklocal(idev, &addr);
2152 }
2153
2154 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2155 static void addrconf_sit_config(struct net_device *dev)
2156 {
2157         struct inet6_dev *idev;
2158
2159         ASSERT_RTNL();
2160
2161         /*
2162          * Configure the tunnel with one of our IPv4
2163          * addresses... we should configure all of
2164          * our v4 addrs in the tunnel
2165          */
2166
2167         if ((idev = ipv6_find_idev(dev)) == NULL) {
2168                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2169                 return;
2170         }
2171
2172         sit_add_v4_addrs(idev);
2173
2174         if (dev->flags&IFF_POINTOPOINT) {
2175                 addrconf_add_mroute(dev);
2176                 addrconf_add_lroute(dev);
2177         } else
2178                 sit_route_add(dev);
2179 }
2180 #endif
2181
2182 static inline int
2183 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2184 {
2185         struct in6_addr lladdr;
2186
2187         if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2188                 addrconf_add_linklocal(idev, &lladdr);
2189                 return 0;
2190         }
2191         return -1;
2192 }
2193
2194 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2195 {
2196         struct net_device *link_dev;
2197
2198         /* first try to inherit the link-local address from the link device */
2199         if (idev->dev->iflink &&
2200             (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2201                 if (!ipv6_inherit_linklocal(idev, link_dev))
2202                         return;
2203         }
2204         /* then try to inherit it from any device */
2205         for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2206                 if (!ipv6_inherit_linklocal(idev, link_dev))
2207                         return;
2208         }
2209         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2210 }
2211
2212 /*
2213  * Autoconfigure tunnel with a link-local address so routing protocols,
2214  * DHCPv6, MLD etc. can be run over the virtual link
2215  */
2216
2217 static void addrconf_ip6_tnl_config(struct net_device *dev)
2218 {
2219         struct inet6_dev *idev;
2220
2221         ASSERT_RTNL();
2222
2223         if ((idev = addrconf_add_dev(dev)) == NULL) {
2224                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2225                 return;
2226         }
2227         ip6_tnl_add_linklocal(idev);
2228 }
2229
2230 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2231                            void * data)
2232 {
2233         struct net_device *dev = (struct net_device *) data;
2234         struct inet6_dev *idev = __in6_dev_get(dev);
2235         int run_pending = 0;
2236
2237         switch(event) {
2238         case NETDEV_UP:
2239         case NETDEV_CHANGE:
2240                 if (event == NETDEV_UP) {
2241                         if (!netif_carrier_ok(dev)) {
2242                                 /* device is not ready yet. */
2243                                 printk(KERN_INFO
2244                                         "ADDRCONF(NETDEV_UP): %s: "
2245                                         "link is not ready\n",
2246                                         dev->name);
2247                                 break;
2248                         }
2249
2250                         if (idev)
2251                                 idev->if_flags |= IF_READY;
2252                 } else {
2253                         if (!netif_carrier_ok(dev)) {
2254                                 /* device is still not ready. */
2255                                 break;
2256                         }
2257
2258                         if (idev) {
2259                                 if (idev->if_flags & IF_READY) {
2260                                         /* device is already configured. */
2261                                         break;
2262                                 }
2263                                 idev->if_flags |= IF_READY;
2264                         }
2265
2266                         printk(KERN_INFO
2267                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2268                                         "link becomes ready\n",
2269                                         dev->name);
2270
2271                         run_pending = 1;
2272                 }
2273
2274                 switch(dev->type) {
2275 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2276                 case ARPHRD_SIT:
2277                         addrconf_sit_config(dev);
2278                         break;
2279 #endif
2280                 case ARPHRD_TUNNEL6:
2281                         addrconf_ip6_tnl_config(dev);
2282                         break;
2283                 case ARPHRD_LOOPBACK:
2284                         init_loopback(dev);
2285                         break;
2286
2287                 default:
2288                         addrconf_dev_config(dev);
2289                         break;
2290                 };
2291                 if (idev) {
2292                         if (run_pending)
2293                                 addrconf_dad_run(idev);
2294
2295                         /* If the MTU changed during the interface down, when the
2296                            interface up, the changed MTU must be reflected in the
2297                            idev as well as routers.
2298                          */
2299                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2300                                 rt6_mtu_change(dev, dev->mtu);
2301                                 idev->cnf.mtu6 = dev->mtu;
2302                         }
2303                         idev->tstamp = jiffies;
2304                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2305                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2306                            stop IPv6 on this interface.
2307                          */
2308                         if (dev->mtu < IPV6_MIN_MTU)
2309                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2310                 }
2311                 break;
2312
2313         case NETDEV_CHANGEMTU:
2314                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2315                         rt6_mtu_change(dev, dev->mtu);
2316                         idev->cnf.mtu6 = dev->mtu;
2317                         break;
2318                 }
2319
2320                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2321
2322         case NETDEV_DOWN:
2323         case NETDEV_UNREGISTER:
2324                 /*
2325                  *      Remove all addresses from this interface.
2326                  */
2327                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2328                 break;
2329
2330         case NETDEV_CHANGENAME:
2331 #ifdef CONFIG_SYSCTL
2332                 if (idev) {
2333                         addrconf_sysctl_unregister(&idev->cnf);
2334                         neigh_sysctl_unregister(idev->nd_parms);
2335                         neigh_sysctl_register(dev, idev->nd_parms,
2336                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2337                                               &ndisc_ifinfo_sysctl_change,
2338                                               NULL);
2339                         addrconf_sysctl_register(idev, &idev->cnf);
2340                 }
2341 #endif
2342                 break;
2343         };
2344
2345         return NOTIFY_OK;
2346 }
2347
2348 /*
2349  *      addrconf module should be notified of a device going up
2350  */
2351 static struct notifier_block ipv6_dev_notf = {
2352         .notifier_call = addrconf_notify,
2353         .priority = 0
2354 };
2355
2356 static int addrconf_ifdown(struct net_device *dev, int how)
2357 {
2358         struct inet6_dev *idev;
2359         struct inet6_ifaddr *ifa, **bifa;
2360         int i;
2361
2362         ASSERT_RTNL();
2363
2364         if (dev == &loopback_dev && how == 1)
2365                 how = 0;
2366
2367         rt6_ifdown(dev);
2368         neigh_ifdown(&nd_tbl, dev);
2369
2370         idev = __in6_dev_get(dev);
2371         if (idev == NULL)
2372                 return -ENODEV;
2373
2374         /* Step 1: remove reference to ipv6 device from parent device.
2375                    Do not dev_put!
2376          */
2377         if (how == 1) {
2378                 idev->dead = 1;
2379
2380                 /* protected by rtnl_lock */
2381                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2382
2383                 /* Step 1.5: remove snmp6 entry */
2384                 snmp6_unregister_dev(idev);
2385
2386         }
2387
2388         /* Step 2: clear hash table */
2389         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2390                 bifa = &inet6_addr_lst[i];
2391
2392                 write_lock_bh(&addrconf_hash_lock);
2393                 while ((ifa = *bifa) != NULL) {
2394                         if (ifa->idev == idev) {
2395                                 *bifa = ifa->lst_next;
2396                                 ifa->lst_next = NULL;
2397                                 addrconf_del_timer(ifa);
2398                                 in6_ifa_put(ifa);
2399                                 continue;
2400                         }
2401                         bifa = &ifa->lst_next;
2402                 }
2403                 write_unlock_bh(&addrconf_hash_lock);
2404         }
2405
2406         write_lock_bh(&idev->lock);
2407
2408         /* Step 3: clear flags for stateless addrconf */
2409         if (how != 1)
2410                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2411
2412         /* Step 4: clear address list */
2413 #ifdef CONFIG_IPV6_PRIVACY
2414         if (how == 1 && del_timer(&idev->regen_timer))
2415                 in6_dev_put(idev);
2416
2417         /* clear tempaddr list */
2418         while ((ifa = idev->tempaddr_list) != NULL) {
2419                 idev->tempaddr_list = ifa->tmp_next;
2420                 ifa->tmp_next = NULL;
2421                 ifa->dead = 1;
2422                 write_unlock_bh(&idev->lock);
2423                 spin_lock_bh(&ifa->lock);
2424
2425                 if (ifa->ifpub) {
2426                         in6_ifa_put(ifa->ifpub);
2427                         ifa->ifpub = NULL;
2428                 }
2429                 spin_unlock_bh(&ifa->lock);
2430                 in6_ifa_put(ifa);
2431                 write_lock_bh(&idev->lock);
2432         }
2433 #endif
2434         while ((ifa = idev->addr_list) != NULL) {
2435                 idev->addr_list = ifa->if_next;
2436                 ifa->if_next = NULL;
2437                 ifa->dead = 1;
2438                 addrconf_del_timer(ifa);
2439                 write_unlock_bh(&idev->lock);
2440
2441                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2442                 in6_ifa_put(ifa);
2443
2444                 write_lock_bh(&idev->lock);
2445         }
2446         write_unlock_bh(&idev->lock);
2447
2448         /* Step 5: Discard multicast list */
2449
2450         if (how == 1)
2451                 ipv6_mc_destroy_dev(idev);
2452         else
2453                 ipv6_mc_down(idev);
2454
2455         /* Step 5: netlink notification of this interface */
2456         idev->tstamp = jiffies;
2457         inet6_ifinfo_notify(RTM_DELLINK, idev);
2458
2459         /* Shot the device (if unregistered) */
2460
2461         if (how == 1) {
2462 #ifdef CONFIG_SYSCTL
2463                 addrconf_sysctl_unregister(&idev->cnf);
2464                 neigh_sysctl_unregister(idev->nd_parms);
2465 #endif
2466                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2467                 neigh_ifdown(&nd_tbl, dev);
2468                 in6_dev_put(idev);
2469         }
2470         return 0;
2471 }
2472
2473 static void addrconf_rs_timer(unsigned long data)
2474 {
2475         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2476
2477         if (ifp->idev->cnf.forwarding)
2478                 goto out;
2479
2480         if (ifp->idev->if_flags & IF_RA_RCVD) {
2481                 /*
2482                  *      Announcement received after solicitation
2483                  *      was sent
2484                  */
2485                 goto out;
2486         }
2487
2488         spin_lock(&ifp->lock);
2489         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2490                 struct in6_addr all_routers;
2491
2492                 /* The wait after the last probe can be shorter */
2493                 addrconf_mod_timer(ifp, AC_RS,
2494                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2495                                    ifp->idev->cnf.rtr_solicit_delay :
2496                                    ifp->idev->cnf.rtr_solicit_interval);
2497                 spin_unlock(&ifp->lock);
2498
2499                 ipv6_addr_all_routers(&all_routers);
2500
2501                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2502         } else {
2503                 spin_unlock(&ifp->lock);
2504                 /*
2505                  * Note: we do not support deprecated "all on-link"
2506                  * assumption any longer.
2507                  */
2508                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2509                        ifp->idev->dev->name);
2510         }
2511
2512 out:
2513         in6_ifa_put(ifp);
2514 }
2515
2516 /*
2517  *      Duplicate Address Detection
2518  */
2519 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2520 {
2521         unsigned long rand_num;
2522         struct inet6_dev *idev = ifp->idev;
2523
2524         rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2525         ifp->probes = idev->cnf.dad_transmits;
2526         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2527 }
2528
2529 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2530 {
2531         struct inet6_dev *idev = ifp->idev;
2532         struct net_device *dev = idev->dev;
2533
2534         addrconf_join_solict(dev, &ifp->addr);
2535
2536         if (ifp->prefix_len != 128 && (ifp->flags&IFA_F_PERMANENT))
2537                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 0,
2538                                         flags);
2539
2540         net_srandom(ifp->addr.s6_addr32[3]);
2541
2542         read_lock_bh(&idev->lock);
2543         if (ifp->dead)
2544                 goto out;
2545         spin_lock_bh(&ifp->lock);
2546
2547         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2548             !(ifp->flags&IFA_F_TENTATIVE) ||
2549             ifp->flags & IFA_F_NODAD) {
2550                 ifp->flags &= ~IFA_F_TENTATIVE;
2551                 spin_unlock_bh(&ifp->lock);
2552                 read_unlock_bh(&idev->lock);
2553
2554                 addrconf_dad_completed(ifp);
2555                 return;
2556         }
2557
2558         if (!(idev->if_flags & IF_READY)) {
2559                 spin_unlock_bh(&ifp->lock);
2560                 read_unlock_bh(&idev->lock);
2561                 /*
2562                  * If the defice is not ready:
2563                  * - keep it tentative if it is a permanent address.
2564                  * - otherwise, kill it.
2565                  */
2566                 in6_ifa_hold(ifp);
2567                 addrconf_dad_stop(ifp);
2568                 return;
2569         }
2570         addrconf_dad_kick(ifp);
2571         spin_unlock_bh(&ifp->lock);
2572 out:
2573         read_unlock_bh(&idev->lock);
2574 }
2575
2576 static void addrconf_dad_timer(unsigned long data)
2577 {
2578         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2579         struct inet6_dev *idev = ifp->idev;
2580         struct in6_addr unspec;
2581         struct in6_addr mcaddr;
2582
2583         read_lock_bh(&idev->lock);
2584         if (idev->dead) {
2585                 read_unlock_bh(&idev->lock);
2586                 goto out;
2587         }
2588         spin_lock_bh(&ifp->lock);
2589         if (ifp->probes == 0) {
2590                 /*
2591                  * DAD was successful
2592                  */
2593
2594                 ifp->flags &= ~IFA_F_TENTATIVE;
2595                 spin_unlock_bh(&ifp->lock);
2596                 read_unlock_bh(&idev->lock);
2597
2598                 addrconf_dad_completed(ifp);
2599
2600                 goto out;
2601         }
2602
2603         ifp->probes--;
2604         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2605         spin_unlock_bh(&ifp->lock);
2606         read_unlock_bh(&idev->lock);
2607
2608         /* send a neighbour solicitation for our addr */
2609         memset(&unspec, 0, sizeof(unspec));
2610         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2611         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2612 out:
2613         in6_ifa_put(ifp);
2614 }
2615
2616 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2617 {
2618         struct net_device *     dev = ifp->idev->dev;
2619
2620         /*
2621          *      Configure the address for reception. Now it is valid.
2622          */
2623
2624         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2625
2626         /* If added prefix is link local and forwarding is off,
2627            start sending router solicitations.
2628          */
2629
2630         if (ifp->idev->cnf.forwarding == 0 &&
2631             ifp->idev->cnf.rtr_solicits > 0 &&
2632             (dev->flags&IFF_LOOPBACK) == 0 &&
2633             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2634                 struct in6_addr all_routers;
2635
2636                 ipv6_addr_all_routers(&all_routers);
2637
2638                 /*
2639                  *      If a host as already performed a random delay
2640                  *      [...] as part of DAD [...] there is no need
2641                  *      to delay again before sending the first RS
2642                  */
2643                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2644
2645                 spin_lock_bh(&ifp->lock);
2646                 ifp->probes = 1;
2647                 ifp->idev->if_flags |= IF_RS_SENT;
2648                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2649                 spin_unlock_bh(&ifp->lock);
2650         }
2651 }
2652
2653 static void addrconf_dad_run(struct inet6_dev *idev) {
2654         struct inet6_ifaddr *ifp;
2655
2656         read_lock_bh(&idev->lock);
2657         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2658                 spin_lock_bh(&ifp->lock);
2659                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2660                         spin_unlock_bh(&ifp->lock);
2661                         continue;
2662                 }
2663                 spin_unlock_bh(&ifp->lock);
2664                 addrconf_dad_kick(ifp);
2665         }
2666         read_unlock_bh(&idev->lock);
2667 }
2668
2669 #ifdef CONFIG_PROC_FS
2670 struct if6_iter_state {
2671         int bucket;
2672 };
2673
2674 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2675 {
2676         struct inet6_ifaddr *ifa = NULL;
2677         struct if6_iter_state *state = seq->private;
2678
2679         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2680                 ifa = inet6_addr_lst[state->bucket];
2681                 if (ifa)
2682                         break;
2683         }
2684         return ifa;
2685 }
2686
2687 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2688 {
2689         struct if6_iter_state *state = seq->private;
2690
2691         ifa = ifa->lst_next;
2692 try_again:
2693         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2694                 ifa = inet6_addr_lst[state->bucket];
2695                 goto try_again;
2696         }
2697         return ifa;
2698 }
2699
2700 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2701 {
2702         struct inet6_ifaddr *ifa = if6_get_first(seq);
2703
2704         if (ifa)
2705                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2706                         --pos;
2707         return pos ? NULL : ifa;
2708 }
2709
2710 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2711 {
2712         read_lock_bh(&addrconf_hash_lock);
2713         return if6_get_idx(seq, *pos);
2714 }
2715
2716 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2717 {
2718         struct inet6_ifaddr *ifa;
2719
2720         ifa = if6_get_next(seq, v);
2721         ++*pos;
2722         return ifa;
2723 }
2724
2725 static void if6_seq_stop(struct seq_file *seq, void *v)
2726 {
2727         read_unlock_bh(&addrconf_hash_lock);
2728 }
2729
2730 static int if6_seq_show(struct seq_file *seq, void *v)
2731 {
2732         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2733         seq_printf(seq,
2734                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2735                    NIP6(ifp->addr),
2736                    ifp->idev->dev->ifindex,
2737                    ifp->prefix_len,
2738                    ifp->scope,
2739                    ifp->flags,
2740                    ifp->idev->dev->name);
2741         return 0;
2742 }
2743
2744 static struct seq_operations if6_seq_ops = {
2745         .start  = if6_seq_start,
2746         .next   = if6_seq_next,
2747         .show   = if6_seq_show,
2748         .stop   = if6_seq_stop,
2749 };
2750
2751 static int if6_seq_open(struct inode *inode, struct file *file)
2752 {
2753         struct seq_file *seq;
2754         int rc = -ENOMEM;
2755         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2756
2757         if (!s)
2758                 goto out;
2759
2760         rc = seq_open(file, &if6_seq_ops);
2761         if (rc)
2762                 goto out_kfree;
2763
2764         seq = file->private_data;
2765         seq->private = s;
2766 out:
2767         return rc;
2768 out_kfree:
2769         kfree(s);
2770         goto out;
2771 }
2772
2773 static struct file_operations if6_fops = {
2774         .owner          = THIS_MODULE,
2775         .open           = if6_seq_open,
2776         .read           = seq_read,
2777         .llseek         = seq_lseek,
2778         .release        = seq_release_private,
2779 };
2780
2781 int __init if6_proc_init(void)
2782 {
2783         if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2784                 return -ENOMEM;
2785         return 0;
2786 }
2787
2788 void if6_proc_exit(void)
2789 {
2790         proc_net_remove("if_inet6");
2791 }
2792 #endif  /* CONFIG_PROC_FS */
2793
2794 #ifdef CONFIG_IPV6_MIP6
2795 /* Check if address is a home address configured on any interface. */
2796 int ipv6_chk_home_addr(struct in6_addr *addr)
2797 {
2798         int ret = 0;
2799         struct inet6_ifaddr * ifp;
2800         u8 hash = ipv6_addr_hash(addr);
2801         read_lock_bh(&addrconf_hash_lock);
2802         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
2803                 if (ipv6_addr_cmp(&ifp->addr, addr) == 0 &&
2804                     (ifp->flags & IFA_F_HOMEADDRESS)) {
2805                         ret = 1;
2806                         break;
2807                 }
2808         }
2809         read_unlock_bh(&addrconf_hash_lock);
2810         return ret;
2811 }
2812 #endif
2813
2814 /*
2815  *      Periodic address status verification
2816  */
2817
2818 static void addrconf_verify(unsigned long foo)
2819 {
2820         struct inet6_ifaddr *ifp;
2821         unsigned long now, next;
2822         int i;
2823
2824         spin_lock_bh(&addrconf_verify_lock);
2825         now = jiffies;
2826         next = now + ADDR_CHECK_FREQUENCY;
2827
2828         del_timer(&addr_chk_timer);
2829
2830         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2831
2832 restart:
2833                 read_lock(&addrconf_hash_lock);
2834                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2835                         unsigned long age;
2836 #ifdef CONFIG_IPV6_PRIVACY
2837                         unsigned long regen_advance;
2838 #endif
2839
2840                         if (ifp->flags & IFA_F_PERMANENT)
2841                                 continue;
2842
2843                         spin_lock(&ifp->lock);
2844                         age = (now - ifp->tstamp) / HZ;
2845
2846 #ifdef CONFIG_IPV6_PRIVACY
2847                         regen_advance = ifp->idev->cnf.regen_max_retry *
2848                                         ifp->idev->cnf.dad_transmits *
2849                                         ifp->idev->nd_parms->retrans_time / HZ;
2850 #endif
2851
2852                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2853                             age >= ifp->valid_lft) {
2854                                 spin_unlock(&ifp->lock);
2855                                 in6_ifa_hold(ifp);
2856                                 read_unlock(&addrconf_hash_lock);
2857                                 ipv6_del_addr(ifp);
2858                                 goto restart;
2859                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2860                                 spin_unlock(&ifp->lock);
2861                                 continue;
2862                         } else if (age >= ifp->prefered_lft) {
2863                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2864                                 int deprecate = 0;
2865
2866                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2867                                         deprecate = 1;
2868                                         ifp->flags |= IFA_F_DEPRECATED;
2869                                 }
2870
2871                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2872                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2873
2874                                 spin_unlock(&ifp->lock);
2875
2876                                 if (deprecate) {
2877                                         in6_ifa_hold(ifp);
2878                                         read_unlock(&addrconf_hash_lock);
2879
2880                                         ipv6_ifa_notify(0, ifp);
2881                                         in6_ifa_put(ifp);
2882                                         goto restart;
2883                                 }
2884 #ifdef CONFIG_IPV6_PRIVACY
2885                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2886                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2887                                 if (age >= ifp->prefered_lft - regen_advance) {
2888                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2889                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2890                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2891                                         if (!ifp->regen_count && ifpub) {
2892                                                 ifp->regen_count++;
2893                                                 in6_ifa_hold(ifp);
2894                                                 in6_ifa_hold(ifpub);
2895                                                 spin_unlock(&ifp->lock);
2896                                                 read_unlock(&addrconf_hash_lock);
2897                                                 spin_lock(&ifpub->lock);
2898                                                 ifpub->regen_count = 0;
2899                                                 spin_unlock(&ifpub->lock);
2900                                                 ipv6_create_tempaddr(ifpub, ifp);
2901                                                 in6_ifa_put(ifpub);
2902                                                 in6_ifa_put(ifp);
2903                                                 goto restart;
2904                                         }
2905                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2906                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2907                                 spin_unlock(&ifp->lock);
2908 #endif
2909                         } else {
2910                                 /* ifp->prefered_lft <= ifp->valid_lft */
2911                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2912                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2913                                 spin_unlock(&ifp->lock);
2914                         }
2915                 }
2916                 read_unlock(&addrconf_hash_lock);
2917         }
2918
2919         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2920         add_timer(&addr_chk_timer);
2921         spin_unlock_bh(&addrconf_verify_lock);
2922 }
2923
2924 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2925 {
2926         struct in6_addr *pfx = NULL;
2927
2928         if (addr)
2929                 pfx = nla_data(addr);
2930
2931         if (local) {
2932                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2933                         pfx = NULL;
2934                 else
2935                         pfx = nla_data(local);
2936         }
2937
2938         return pfx;
2939 }
2940
2941 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2942         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2943         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2944         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2945 };
2946
2947 static int
2948 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2949 {
2950         struct ifaddrmsg *ifm;
2951         struct nlattr *tb[IFA_MAX+1];
2952         struct in6_addr *pfx;
2953         int err;
2954
2955         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2956         if (err < 0)
2957                 return err;
2958
2959         ifm = nlmsg_data(nlh);
2960         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2961         if (pfx == NULL)
2962                 return -EINVAL;
2963
2964         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2965 }
2966
2967 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
2968                              u32 prefered_lft, u32 valid_lft)
2969 {
2970         if (!valid_lft || (prefered_lft > valid_lft))
2971                 return -EINVAL;
2972
2973         if (valid_lft == INFINITY_LIFE_TIME)
2974                 ifa_flags |= IFA_F_PERMANENT;
2975         else if (valid_lft >= 0x7FFFFFFF/HZ)
2976                 valid_lft = 0x7FFFFFFF/HZ;
2977
2978         if (prefered_lft == 0)
2979                 ifa_flags |= IFA_F_DEPRECATED;
2980         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2981                  (prefered_lft != INFINITY_LIFE_TIME))
2982                 prefered_lft = 0x7FFFFFFF/HZ;
2983
2984         spin_lock_bh(&ifp->lock);
2985         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
2986         ifp->tstamp = jiffies;
2987         ifp->valid_lft = valid_lft;
2988         ifp->prefered_lft = prefered_lft;
2989
2990         spin_unlock_bh(&ifp->lock);
2991         if (!(ifp->flags&IFA_F_TENTATIVE))
2992                 ipv6_ifa_notify(0, ifp);
2993
2994         addrconf_verify(0);
2995
2996         return 0;
2997 }
2998
2999 static int
3000 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3001 {
3002         struct ifaddrmsg *ifm;
3003         struct nlattr *tb[IFA_MAX+1];
3004         struct in6_addr *pfx;
3005         struct inet6_ifaddr *ifa;
3006         struct net_device *dev;
3007         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3008         u8 ifa_flags;
3009         int err;
3010
3011         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3012         if (err < 0)
3013                 return err;
3014
3015         ifm = nlmsg_data(nlh);
3016         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3017         if (pfx == NULL)
3018                 return -EINVAL;
3019
3020         if (tb[IFA_CACHEINFO]) {
3021                 struct ifa_cacheinfo *ci;
3022
3023                 ci = nla_data(tb[IFA_CACHEINFO]);
3024                 valid_lft = ci->ifa_valid;
3025                 preferred_lft = ci->ifa_prefered;
3026         } else {
3027                 preferred_lft = INFINITY_LIFE_TIME;
3028                 valid_lft = INFINITY_LIFE_TIME;
3029         }
3030
3031         dev =  __dev_get_by_index(ifm->ifa_index);
3032         if (dev == NULL)
3033                 return -ENODEV;
3034
3035         /* We ignore other flags so far. */
3036         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3037
3038         ifa = ipv6_get_ifaddr(pfx, dev, 1);
3039         if (ifa == NULL) {
3040                 /*
3041                  * It would be best to check for !NLM_F_CREATE here but
3042                  * userspace alreay relies on not having to provide this.
3043                  */
3044                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
3045                                       ifa_flags, preferred_lft, valid_lft);
3046         }
3047
3048         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3049             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3050                 err = -EEXIST;
3051         else
3052                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3053
3054         in6_ifa_put(ifa);
3055
3056         return err;
3057 }
3058
3059 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3060                           u8 scope, int ifindex)
3061 {
3062         struct ifaddrmsg *ifm;
3063
3064         ifm = nlmsg_data(nlh);
3065         ifm->ifa_family = AF_INET6;
3066         ifm->ifa_prefixlen = prefixlen;
3067         ifm->ifa_flags = flags;
3068         ifm->ifa_scope = scope;
3069         ifm->ifa_index = ifindex;
3070 }
3071
3072 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3073                          unsigned long tstamp, u32 preferred, u32 valid)
3074 {
3075         struct ifa_cacheinfo ci;
3076
3077         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3078                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3079         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3080                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3081         ci.ifa_prefered = preferred;
3082         ci.ifa_valid = valid;
3083
3084         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3085 }
3086
3087 static inline int rt_scope(int ifa_scope)
3088 {
3089         if (ifa_scope & IFA_HOST)
3090                 return RT_SCOPE_HOST;
3091         else if (ifa_scope & IFA_LINK)
3092                 return RT_SCOPE_LINK;
3093         else if (ifa_scope & IFA_SITE)
3094                 return RT_SCOPE_SITE;
3095         else
3096                 return RT_SCOPE_UNIVERSE;
3097 }
3098
3099 static inline int inet6_ifaddr_msgsize(void)
3100 {
3101         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3102                + nla_total_size(16) /* IFA_ADDRESS */
3103                + nla_total_size(sizeof(struct ifa_cacheinfo));
3104 }
3105
3106 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3107                              u32 pid, u32 seq, int event, unsigned int flags)
3108 {
3109         struct nlmsghdr  *nlh;
3110         u32 preferred, valid;
3111
3112         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3113         if (nlh == NULL)
3114                 return -ENOBUFS;
3115
3116         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3117                       ifa->idev->dev->ifindex);
3118
3119         if (!(ifa->flags&IFA_F_PERMANENT)) {
3120                 preferred = ifa->prefered_lft;
3121                 valid = ifa->valid_lft;
3122                 if (preferred != INFINITY_LIFE_TIME) {
3123                         long tval = (jiffies - ifa->tstamp)/HZ;
3124                         preferred -= tval;
3125                         if (valid != INFINITY_LIFE_TIME)
3126                                 valid -= tval;
3127                 }
3128         } else {
3129                 preferred = INFINITY_LIFE_TIME;
3130                 valid = INFINITY_LIFE_TIME;
3131         }
3132
3133         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3134             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3135                 return nlmsg_cancel(skb, nlh);
3136
3137         return nlmsg_end(skb, nlh);
3138 }
3139
3140 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3141                                 u32 pid, u32 seq, int event, u16 flags)
3142 {
3143         struct nlmsghdr  *nlh;
3144         u8 scope = RT_SCOPE_UNIVERSE;
3145         int ifindex = ifmca->idev->dev->ifindex;
3146
3147         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3148                 scope = RT_SCOPE_SITE;
3149
3150         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3151         if (nlh == NULL)
3152                 return -ENOBUFS;
3153
3154         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3155         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3156             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3157                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3158                 return nlmsg_cancel(skb, nlh);
3159
3160         return nlmsg_end(skb, nlh);
3161 }
3162
3163 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3164                                 u32 pid, u32 seq, int event, unsigned int flags)
3165 {
3166         struct nlmsghdr  *nlh;
3167         u8 scope = RT_SCOPE_UNIVERSE;
3168         int ifindex = ifaca->aca_idev->dev->ifindex;
3169
3170         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3171                 scope = RT_SCOPE_SITE;
3172
3173         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3174         if (nlh == NULL)
3175                 return -ENOBUFS;
3176
3177         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3178         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3179             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3180                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3181                 return nlmsg_cancel(skb, nlh);
3182
3183         return nlmsg_end(skb, nlh);
3184 }
3185
3186 enum addr_type_t
3187 {
3188         UNICAST_ADDR,
3189         MULTICAST_ADDR,
3190         ANYCAST_ADDR,
3191 };
3192
3193 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3194                            enum addr_type_t type)
3195 {
3196         int idx, ip_idx;
3197         int s_idx, s_ip_idx;
3198         int err = 1;
3199         struct net_device *dev;
3200         struct inet6_dev *idev = NULL;
3201         struct inet6_ifaddr *ifa;
3202         struct ifmcaddr6 *ifmca;
3203         struct ifacaddr6 *ifaca;
3204
3205         s_idx = cb->args[0];
3206         s_ip_idx = ip_idx = cb->args[1];
3207         read_lock(&dev_base_lock);
3208
3209         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3210                 if (idx < s_idx)
3211                         continue;
3212                 if (idx > s_idx)
3213                         s_ip_idx = 0;
3214                 ip_idx = 0;
3215                 if ((idev = in6_dev_get(dev)) == NULL)
3216                         continue;
3217                 read_lock_bh(&idev->lock);
3218                 switch (type) {
3219                 case UNICAST_ADDR:
3220                         /* unicast address incl. temp addr */
3221                         for (ifa = idev->addr_list; ifa;
3222                              ifa = ifa->if_next, ip_idx++) {
3223                                 if (ip_idx < s_ip_idx)
3224                                         continue;
3225                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3226                                     NETLINK_CB(cb->skb).pid,
3227                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3228                                     NLM_F_MULTI)) <= 0)
3229                                         goto done;
3230                         }
3231                         break;
3232                 case MULTICAST_ADDR:
3233                         /* multicast address */
3234                         for (ifmca = idev->mc_list; ifmca;
3235                              ifmca = ifmca->next, ip_idx++) {
3236                                 if (ip_idx < s_ip_idx)
3237                                         continue;
3238                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3239                                     NETLINK_CB(cb->skb).pid,
3240                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3241                                     NLM_F_MULTI)) <= 0)
3242                                         goto done;
3243                         }
3244                         break;
3245                 case ANYCAST_ADDR:
3246                         /* anycast address */
3247                         for (ifaca = idev->ac_list; ifaca;
3248                              ifaca = ifaca->aca_next, ip_idx++) {
3249                                 if (ip_idx < s_ip_idx)
3250                                         continue;
3251                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3252                                     NETLINK_CB(cb->skb).pid,
3253                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3254                                     NLM_F_MULTI)) <= 0)
3255                                         goto done;
3256                         }
3257                         break;
3258                 default:
3259                         break;
3260                 }
3261                 read_unlock_bh(&idev->lock);
3262                 in6_dev_put(idev);
3263         }
3264 done:
3265         if (err <= 0) {
3266                 read_unlock_bh(&idev->lock);
3267                 in6_dev_put(idev);
3268         }
3269         read_unlock(&dev_base_lock);
3270         cb->args[0] = idx;
3271         cb->args[1] = ip_idx;
3272         return skb->len;
3273 }
3274
3275 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3276 {
3277         enum addr_type_t type = UNICAST_ADDR;
3278         return inet6_dump_addr(skb, cb, type);
3279 }
3280
3281 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3282 {
3283         enum addr_type_t type = MULTICAST_ADDR;
3284         return inet6_dump_addr(skb, cb, type);
3285 }
3286
3287
3288 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3289 {
3290         enum addr_type_t type = ANYCAST_ADDR;
3291         return inet6_dump_addr(skb, cb, type);
3292 }
3293
3294 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3295                              void *arg)
3296 {
3297         struct ifaddrmsg *ifm;
3298         struct nlattr *tb[IFA_MAX+1];
3299         struct in6_addr *addr = NULL;
3300         struct net_device *dev = NULL;
3301         struct inet6_ifaddr *ifa;
3302         struct sk_buff *skb;
3303         int err;
3304
3305         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3306         if (err < 0)
3307                 goto errout;
3308
3309         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3310         if (addr == NULL) {
3311                 err = -EINVAL;
3312                 goto errout;
3313         }
3314
3315         ifm = nlmsg_data(nlh);
3316         if (ifm->ifa_index)
3317                 dev = __dev_get_by_index(ifm->ifa_index);
3318
3319         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3320                 err = -EADDRNOTAVAIL;
3321                 goto errout;
3322         }
3323
3324         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3325                 err = -ENOBUFS;
3326                 goto errout_ifa;
3327         }
3328
3329         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3330                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3331         /* failure implies BUG in inet6_ifaddr_msgsize() */
3332         BUG_ON(err < 0);
3333
3334         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3335 errout_ifa:
3336         in6_ifa_put(ifa);
3337 errout:
3338         return err;
3339 }
3340
3341 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3342 {
3343         struct sk_buff *skb;
3344         int err = -ENOBUFS;
3345
3346         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3347         if (skb == NULL)
3348                 goto errout;
3349
3350         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3351         /* failure implies BUG in inet6_ifaddr_msgsize() */
3352         BUG_ON(err < 0);
3353
3354         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3355 errout:
3356         if (err < 0)
3357                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3358 }
3359
3360 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
3361                                 __s32 *array, int bytes)
3362 {
3363         BUG_ON(bytes < (DEVCONF_MAX * 4));
3364
3365         memset(array, 0, bytes);
3366         array[DEVCONF_FORWARDING] = cnf->forwarding;
3367         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3368         array[DEVCONF_MTU6] = cnf->mtu6;
3369         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3370         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3371         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3372         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3373         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3374         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3375         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3376         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3377 #ifdef CONFIG_IPV6_PRIVACY
3378         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3379         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3380         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3381         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3382         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3383 #endif
3384         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3385         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3386         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3387 #ifdef CONFIG_IPV6_ROUTER_PREF
3388         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3389         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3390 #ifdef CONFIV_IPV6_ROUTE_INFO
3391         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3392 #endif
3393 #endif
3394         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3395 }
3396
3397 static inline size_t inet6_if_nlmsg_size(void)
3398 {
3399         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3400                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3401                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3402                + nla_total_size(4) /* IFLA_MTU */
3403                + nla_total_size(4) /* IFLA_LINK */
3404                + nla_total_size( /* IFLA_PROTINFO */
3405                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3406                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3407                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3408                  );
3409 }
3410
3411 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3412                              u32 pid, u32 seq, int event, unsigned int flags)
3413 {
3414         struct net_device *dev = idev->dev;
3415         struct nlattr *conf;
3416         struct ifinfomsg *hdr;
3417         struct nlmsghdr *nlh;
3418         void *protoinfo;
3419         struct ifla_cacheinfo ci;
3420
3421         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3422         if (nlh == NULL)
3423                 return -ENOBUFS;
3424
3425         hdr = nlmsg_data(nlh);
3426         hdr->ifi_family = AF_INET6;
3427         hdr->__ifi_pad = 0;
3428         hdr->ifi_type = dev->type;
3429         hdr->ifi_index = dev->ifindex;
3430         hdr->ifi_flags = dev_get_flags(dev);
3431         hdr->ifi_change = 0;
3432
3433         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3434
3435         if (dev->addr_len)
3436                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3437
3438         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3439         if (dev->ifindex != dev->iflink)
3440                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3441
3442         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3443         if (protoinfo == NULL)
3444                 goto nla_put_failure;
3445
3446         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3447
3448         ci.max_reasm_len = IPV6_MAXPLEN;
3449         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3450                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3451         ci.reachable_time = idev->nd_parms->reachable_time;
3452         ci.retrans_time = idev->nd_parms->retrans_time;
3453         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3454
3455         conf = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3456         if (conf == NULL)
3457                 goto nla_put_failure;
3458         ipv6_store_devconf(&idev->cnf, nla_data(conf), nla_len(conf));
3459
3460         /* XXX - Statistics/MC not implemented */
3461
3462         nla_nest_end(skb, protoinfo);
3463         return nlmsg_end(skb, nlh);
3464
3465 nla_put_failure:
3466         return nlmsg_cancel(skb, nlh);
3467 }
3468
3469 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3470 {
3471         int idx, err;
3472         int s_idx = cb->args[0];
3473         struct net_device *dev;
3474         struct inet6_dev *idev;
3475
3476         read_lock(&dev_base_lock);
3477         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3478                 if (idx < s_idx)
3479                         continue;
3480                 if ((idev = in6_dev_get(dev)) == NULL)
3481                         continue;
3482                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3483                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3484                 in6_dev_put(idev);
3485                 if (err <= 0)
3486                         break;
3487         }
3488         read_unlock(&dev_base_lock);
3489         cb->args[0] = idx;
3490
3491         return skb->len;
3492 }
3493
3494 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3495 {
3496         struct sk_buff *skb;
3497         int err = -ENOBUFS;
3498
3499         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3500         if (skb == NULL)
3501                 goto errout;
3502
3503         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3504         /* failure implies BUG in inet6_if_nlmsg_size() */
3505         BUG_ON(err < 0);
3506
3507         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3508 errout:
3509         if (err < 0)
3510                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3511 }
3512
3513 static inline size_t inet6_prefix_nlmsg_size(void)
3514 {
3515         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3516                + nla_total_size(sizeof(struct in6_addr))
3517                + nla_total_size(sizeof(struct prefix_cacheinfo));
3518 }
3519
3520 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3521                              struct prefix_info *pinfo, u32 pid, u32 seq,
3522                              int event, unsigned int flags)
3523 {
3524         struct prefixmsg *pmsg;
3525         struct nlmsghdr *nlh;
3526         struct prefix_cacheinfo ci;
3527
3528         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3529         if (nlh == NULL)
3530                 return -ENOBUFS;
3531
3532         pmsg = nlmsg_data(nlh);
3533         pmsg->prefix_family = AF_INET6;
3534         pmsg->prefix_pad1 = 0;
3535         pmsg->prefix_pad2 = 0;
3536         pmsg->prefix_ifindex = idev->dev->ifindex;
3537         pmsg->prefix_len = pinfo->prefix_len;
3538         pmsg->prefix_type = pinfo->type;
3539         pmsg->prefix_pad3 = 0;
3540         pmsg->prefix_flags = 0;
3541         if (pinfo->onlink)
3542                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3543         if (pinfo->autoconf)
3544                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3545
3546         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3547
3548         ci.preferred_time = ntohl(pinfo->prefered);
3549         ci.valid_time = ntohl(pinfo->valid);
3550         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3551
3552         return nlmsg_end(skb, nlh);
3553
3554 nla_put_failure:
3555         return nlmsg_cancel(skb, nlh);
3556 }
3557
3558 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3559                          struct prefix_info *pinfo)
3560 {
3561         struct sk_buff *skb;
3562         int err = -ENOBUFS;
3563
3564         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3565         if (skb == NULL)
3566                 goto errout;
3567
3568         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3569         /* failure implies BUG in inet6_prefix_nlmsg_size() */
3570         BUG_ON(err < 0);
3571
3572         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3573 errout:
3574         if (err < 0)
3575                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3576 }
3577
3578 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3579         [RTM_GETLINK - RTM_BASE] = { .dumpit    = inet6_dump_ifinfo, },
3580         [RTM_NEWADDR - RTM_BASE] = { .doit      = inet6_rtm_newaddr, },
3581         [RTM_DELADDR - RTM_BASE] = { .doit      = inet6_rtm_deladdr, },
3582         [RTM_GETADDR - RTM_BASE] = { .doit      = inet6_rtm_getaddr,
3583                                      .dumpit    = inet6_dump_ifaddr, },
3584         [RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3585         [RTM_GETANYCAST - RTM_BASE] = { .dumpit = inet6_dump_ifacaddr, },
3586         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet6_rtm_newroute, },
3587         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
3588         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
3589                                       .dumpit   = inet6_dump_fib, },
3590 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3591         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3592 #endif
3593 };
3594
3595 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3596 {
3597         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3598
3599         switch (event) {
3600         case RTM_NEWADDR:
3601                 ip6_ins_rt(ifp->rt);
3602                 if (ifp->idev->cnf.forwarding)
3603                         addrconf_join_anycast(ifp);
3604                 break;
3605         case RTM_DELADDR:
3606                 if (ifp->idev->cnf.forwarding)
3607                         addrconf_leave_anycast(ifp);
3608                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3609                 dst_hold(&ifp->rt->u.dst);
3610                 if (ip6_del_rt(ifp->rt))
3611                         dst_free(&ifp->rt->u.dst);
3612                 break;
3613         }
3614 }
3615
3616 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3617 {
3618         rcu_read_lock_bh();
3619         if (likely(ifp->idev->dead == 0))
3620                 __ipv6_ifa_notify(event, ifp);
3621         rcu_read_unlock_bh();
3622 }
3623
3624 #ifdef CONFIG_SYSCTL
3625
3626 static
3627 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3628                            void __user *buffer, size_t *lenp, loff_t *ppos)
3629 {
3630         int *valp = ctl->data;
3631         int val = *valp;
3632         int ret;
3633
3634         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3635
3636         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3637                 if (valp != &ipv6_devconf.forwarding) {
3638                         if ((!*valp) ^ (!val)) {
3639                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3640                                 if (idev == NULL)
3641                                         return ret;
3642                                 dev_forward_change(idev);
3643                         }
3644                 } else {
3645                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3646                         addrconf_forward_change();
3647                 }
3648                 if (*valp)
3649                         rt6_purge_dflt_routers();
3650         }
3651
3652         return ret;
3653 }
3654
3655 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3656                                             int __user *name, int nlen,
3657                                             void __user *oldval,
3658                                             size_t __user *oldlenp,
3659                                             void __user *newval, size_t newlen)
3660 {
3661         int *valp = table->data;
3662         int new;
3663
3664         if (!newval || !newlen)
3665                 return 0;
3666         if (newlen != sizeof(int))
3667                 return -EINVAL;
3668         if (get_user(new, (int __user *)newval))
3669                 return -EFAULT;
3670         if (new == *valp)
3671                 return 0;
3672         if (oldval && oldlenp) {
3673                 size_t len;
3674                 if (get_user(len, oldlenp))
3675                         return -EFAULT;
3676                 if (len) {
3677                         if (len > table->maxlen)
3678                                 len = table->maxlen;
3679                         if (copy_to_user(oldval, valp, len))
3680                                 return -EFAULT;
3681                         if (put_user(len, oldlenp))
3682                                 return -EFAULT;
3683                 }
3684         }
3685
3686         if (valp != &ipv6_devconf_dflt.forwarding) {
3687                 if (valp != &ipv6_devconf.forwarding) {
3688                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3689                         int changed;
3690                         if (unlikely(idev == NULL))
3691                                 return -ENODEV;
3692                         changed = (!*valp) ^ (!new);
3693                         *valp = new;
3694                         if (changed)
3695                                 dev_forward_change(idev);
3696                 } else {
3697                         *valp = new;
3698                         addrconf_forward_change();
3699                 }
3700
3701                 if (*valp)
3702                         rt6_purge_dflt_routers();
3703         } else
3704                 *valp = new;
3705
3706         return 1;
3707 }
3708
3709 static struct addrconf_sysctl_table
3710 {
3711         struct ctl_table_header *sysctl_header;
3712         ctl_table addrconf_vars[__NET_IPV6_MAX];
3713         ctl_table addrconf_dev[2];
3714         ctl_table addrconf_conf_dir[2];
3715         ctl_table addrconf_proto_dir[2];
3716         ctl_table addrconf_root_dir[2];
3717 } addrconf_sysctl __read_mostly = {
3718         .sysctl_header = NULL,
3719         .addrconf_vars = {
3720                 {
3721                         .ctl_name       =       NET_IPV6_FORWARDING,
3722                         .procname       =       "forwarding",
3723                         .data           =       &ipv6_devconf.forwarding,
3724                         .maxlen         =       sizeof(int),
3725                         .mode           =       0644,
3726                         .proc_handler   =       &addrconf_sysctl_forward,
3727                         .strategy       =       &addrconf_sysctl_forward_strategy,
3728                 },
3729                 {
3730                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3731                         .procname       =       "hop_limit",
3732                         .data           =       &ipv6_devconf.hop_limit,
3733                         .maxlen         =       sizeof(int),
3734                         .mode           =       0644,
3735                         .proc_handler   =       proc_dointvec,
3736                 },
3737                 {
3738                         .ctl_name       =       NET_IPV6_MTU,
3739                         .procname       =       "mtu",
3740                         .data           =       &ipv6_devconf.mtu6,
3741                         .maxlen         =       sizeof(int),
3742                         .mode           =       0644,
3743                         .proc_handler   =       &proc_dointvec,
3744                 },
3745                 {
3746                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3747                         .procname       =       "accept_ra",
3748                         .data           =       &ipv6_devconf.accept_ra,
3749                         .maxlen         =       sizeof(int),
3750                         .mode           =       0644,
3751                         .proc_handler   =       &proc_dointvec,
3752                 },
3753                 {
3754                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3755                         .procname       =       "accept_redirects",
3756                         .data           =       &ipv6_devconf.accept_redirects,
3757                         .maxlen         =       sizeof(int),
3758                         .mode           =       0644,
3759                         .proc_handler   =       &proc_dointvec,
3760                 },
3761                 {
3762                         .ctl_name       =       NET_IPV6_AUTOCONF,
3763                         .procname       =       "autoconf",
3764                         .data           =       &ipv6_devconf.autoconf,
3765                         .maxlen         =       sizeof(int),
3766                         .mode           =       0644,
3767                         .proc_handler   =       &proc_dointvec,
3768                 },
3769                 {
3770                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3771                         .procname       =       "dad_transmits",
3772                         .data           =       &ipv6_devconf.dad_transmits,
3773                         .maxlen         =       sizeof(int),
3774                         .mode           =       0644,
3775                         .proc_handler   =       &proc_dointvec,
3776                 },
3777                 {
3778                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3779                         .procname       =       "router_solicitations",
3780                         .data           =       &ipv6_devconf.rtr_solicits,
3781                         .maxlen         =       sizeof(int),
3782                         .mode           =       0644,
3783                         .proc_handler   =       &proc_dointvec,
3784                 },
3785                 {
3786                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3787                         .procname       =       "router_solicitation_interval",
3788                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3789                         .maxlen         =       sizeof(int),
3790                         .mode           =       0644,
3791                         .proc_handler   =       &proc_dointvec_jiffies,
3792                         .strategy       =       &sysctl_jiffies,
3793                 },
3794                 {
3795                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3796                         .procname       =       "router_solicitation_delay",
3797                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3798                         .maxlen         =       sizeof(int),
3799                         .mode           =       0644,
3800                         .proc_handler   =       &proc_dointvec_jiffies,
3801                         .strategy       =       &sysctl_jiffies,
3802                 },
3803                 {
3804                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3805                         .procname       =       "force_mld_version",
3806                         .data           =       &ipv6_devconf.force_mld_version,
3807                         .maxlen         =       sizeof(int),
3808                         .mode           =       0644,
3809                         .proc_handler   =       &proc_dointvec,
3810                 },
3811 #ifdef CONFIG_IPV6_PRIVACY
3812                 {
3813                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3814                         .procname       =       "use_tempaddr",
3815                         .data           =       &ipv6_devconf.use_tempaddr,
3816                         .maxlen         =       sizeof(int),
3817                         .mode           =       0644,
3818                         .proc_handler   =       &proc_dointvec,
3819                 },
3820                 {
3821                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3822                         .procname       =       "temp_valid_lft",
3823                         .data           =       &ipv6_devconf.temp_valid_lft,
3824                         .maxlen         =       sizeof(int),
3825                         .mode           =       0644,
3826                         .proc_handler   =       &proc_dointvec,
3827                 },
3828                 {
3829                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3830                         .procname       =       "temp_prefered_lft",
3831                         .data           =       &ipv6_devconf.temp_prefered_lft,
3832                         .maxlen         =       sizeof(int),
3833                         .mode           =       0644,
3834                         .proc_handler   =       &proc_dointvec,
3835                 },
3836                 {
3837                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3838                         .procname       =       "regen_max_retry",
3839                         .data           =       &ipv6_devconf.regen_max_retry,
3840                         .maxlen         =       sizeof(int),
3841                         .mode           =       0644,
3842                         .proc_handler   =       &proc_dointvec,
3843                 },
3844                 {
3845                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3846                         .procname       =       "max_desync_factor",
3847                         .data           =       &ipv6_devconf.max_desync_factor,
3848                         .maxlen         =       sizeof(int),
3849                         .mode           =       0644,
3850                         .proc_handler   =       &proc_dointvec,
3851                 },
3852 #endif
3853                 {
3854                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3855                         .procname       =       "max_addresses",
3856                         .data           =       &ipv6_devconf.max_addresses,
3857                         .maxlen         =       sizeof(int),
3858                         .mode           =       0644,
3859                         .proc_handler   =       &proc_dointvec,
3860                 },
3861                 {
3862                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3863                         .procname       =       "accept_ra_defrtr",
3864                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3865                         .maxlen         =       sizeof(int),
3866                         .mode           =       0644,
3867                         .proc_handler   =       &proc_dointvec,
3868                 },
3869                 {
3870                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3871                         .procname       =       "accept_ra_pinfo",
3872                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3873                         .maxlen         =       sizeof(int),
3874                         .mode           =       0644,
3875                         .proc_handler   =       &proc_dointvec,
3876                 },
3877 #ifdef CONFIG_IPV6_ROUTER_PREF
3878                 {
3879                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3880                         .procname       =       "accept_ra_rtr_pref",
3881                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3882                         .maxlen         =       sizeof(int),
3883                         .mode           =       0644,
3884                         .proc_handler   =       &proc_dointvec,
3885                 },
3886                 {
3887                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
3888                         .procname       =       "router_probe_interval",
3889                         .data           =       &ipv6_devconf.rtr_probe_interval,
3890                         .maxlen         =       sizeof(int),
3891                         .mode           =       0644,
3892                         .proc_handler   =       &proc_dointvec_jiffies,
3893                         .strategy       =       &sysctl_jiffies,
3894                 },
3895 #ifdef CONFIV_IPV6_ROUTE_INFO
3896                 {
3897                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3898                         .procname       =       "accept_ra_rt_info_max_plen",
3899                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
3900                         .maxlen         =       sizeof(int),
3901                         .mode           =       0644,
3902                         .proc_handler   =       &proc_dointvec,
3903                 },
3904 #endif
3905 #endif
3906                 {
3907                         .ctl_name       =       NET_IPV6_PROXY_NDP,
3908                         .procname       =       "proxy_ndp",
3909                         .data           =       &ipv6_devconf.proxy_ndp,
3910                         .maxlen         =       sizeof(int),
3911                         .mode           =       0644,
3912                         .proc_handler   =       &proc_dointvec,
3913                 },
3914                 {
3915                         .ctl_name       =       0,      /* sentinel */
3916                 }
3917         },
3918         .addrconf_dev = {
3919                 {
3920                         .ctl_name       =       NET_PROTO_CONF_ALL,
3921                         .procname       =       "all",
3922                         .mode           =       0555,
3923                         .child          =       addrconf_sysctl.addrconf_vars,
3924                 },
3925                 {
3926                         .ctl_name       =       0,      /* sentinel */
3927                 }
3928         },
3929         .addrconf_conf_dir = {
3930                 {
3931                         .ctl_name       =       NET_IPV6_CONF,
3932                         .procname       =       "conf",
3933                         .mode           =       0555,
3934                         .child          =       addrconf_sysctl.addrconf_dev,
3935                 },
3936                 {
3937                         .ctl_name       =       0,      /* sentinel */
3938                 }
3939         },
3940         .addrconf_proto_dir = {
3941                 {
3942                         .ctl_name       =       NET_IPV6,
3943                         .procname       =       "ipv6",
3944                         .mode           =       0555,
3945                         .child          =       addrconf_sysctl.addrconf_conf_dir,
3946                 },
3947                 {
3948                         .ctl_name       =       0,      /* sentinel */
3949                 }
3950         },
3951         .addrconf_root_dir = {
3952                 {
3953                         .ctl_name       =       CTL_NET,
3954                         .procname       =       "net",
3955                         .mode           =       0555,
3956                         .child          =       addrconf_sysctl.addrconf_proto_dir,
3957                 },
3958                 {
3959                         .ctl_name       =       0,      /* sentinel */
3960                 }
3961         },
3962 };
3963
3964 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3965 {
3966         int i;
3967         struct net_device *dev = idev ? idev->dev : NULL;
3968         struct addrconf_sysctl_table *t;
3969         char *dev_name = NULL;
3970
3971         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
3972         if (t == NULL)
3973                 return;
3974         for (i=0; t->addrconf_vars[i].data; i++) {
3975                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
3976                 t->addrconf_vars[i].de = NULL;
3977                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
3978         }
3979         if (dev) {
3980                 dev_name = dev->name;
3981                 t->addrconf_dev[0].ctl_name = dev->ifindex;
3982         } else {
3983                 dev_name = "default";
3984                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
3985         }
3986
3987         /*
3988          * Make a copy of dev_name, because '.procname' is regarded as const
3989          * by sysctl and we wouldn't want anyone to change it under our feet
3990          * (see SIOCSIFNAME).
3991          */
3992         dev_name = kstrdup(dev_name, GFP_KERNEL);
3993         if (!dev_name)
3994             goto free;
3995
3996         t->addrconf_dev[0].procname = dev_name;
3997
3998         t->addrconf_dev[0].child = t->addrconf_vars;
3999         t->addrconf_dev[0].de = NULL;
4000         t->addrconf_conf_dir[0].child = t->addrconf_dev;
4001         t->addrconf_conf_dir[0].de = NULL;
4002         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
4003         t->addrconf_proto_dir[0].de = NULL;
4004         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
4005         t->addrconf_root_dir[0].de = NULL;
4006
4007         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir, 0);
4008         if (t->sysctl_header == NULL)
4009                 goto free_procname;
4010         else
4011                 p->sysctl = t;
4012         return;
4013
4014         /* error path */
4015  free_procname:
4016         kfree(dev_name);
4017  free:
4018         kfree(t);
4019
4020         return;
4021 }
4022
4023 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
4024 {
4025         if (p->sysctl) {
4026                 struct addrconf_sysctl_table *t = p->sysctl;
4027                 p->sysctl = NULL;
4028                 unregister_sysctl_table(t->sysctl_header);
4029                 kfree(t->addrconf_dev[0].procname);
4030                 kfree(t);
4031         }
4032 }
4033
4034
4035 #endif
4036
4037 /*
4038  *      Device notifier
4039  */
4040
4041 int register_inet6addr_notifier(struct notifier_block *nb)
4042 {
4043         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4044 }
4045
4046 int unregister_inet6addr_notifier(struct notifier_block *nb)
4047 {
4048         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4049 }
4050
4051 /*
4052  *      Init / cleanup code
4053  */
4054
4055 int __init addrconf_init(void)
4056 {
4057         int err = 0;
4058
4059         /* The addrconf netdev notifier requires that loopback_dev
4060          * has it's ipv6 private information allocated and setup
4061          * before it can bring up and give link-local addresses
4062          * to other devices which are up.
4063          *
4064          * Unfortunately, loopback_dev is not necessarily the first
4065          * entry in the global dev_base list of net devices.  In fact,
4066          * it is likely to be the very last entry on that list.
4067          * So this causes the notifier registry below to try and
4068          * give link-local addresses to all devices besides loopback_dev
4069          * first, then loopback_dev, which cases all the non-loopback_dev
4070          * devices to fail to get a link-local address.
4071          *
4072          * So, as a temporary fix, allocate the ipv6 structure for
4073          * loopback_dev first by hand.
4074          * Longer term, all of the dependencies ipv6 has upon the loopback
4075          * device and it being up should be removed.
4076          */
4077         rtnl_lock();
4078         if (!ipv6_add_dev(&loopback_dev))
4079                 err = -ENOMEM;
4080         rtnl_unlock();
4081         if (err)
4082                 return err;
4083
4084         ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4085
4086         register_netdevice_notifier(&ipv6_dev_notf);
4087
4088         addrconf_verify(0);
4089         rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4090 #ifdef CONFIG_SYSCTL
4091         addrconf_sysctl.sysctl_header =
4092                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir, 0);
4093         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4094 #endif
4095
4096         return 0;
4097 }
4098
4099 void __exit addrconf_cleanup(void)
4100 {
4101         struct net_device *dev;
4102         struct inet6_dev *idev;
4103         struct inet6_ifaddr *ifa;
4104         int i;
4105
4106         unregister_netdevice_notifier(&ipv6_dev_notf);
4107
4108         rtnetlink_links[PF_INET6] = NULL;
4109 #ifdef CONFIG_SYSCTL
4110         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4111         addrconf_sysctl_unregister(&ipv6_devconf);
4112 #endif
4113
4114         rtnl_lock();
4115
4116         /*
4117          *      clean dev list.
4118          */
4119
4120         for (dev=dev_base; dev; dev=dev->next) {
4121                 if ((idev = __in6_dev_get(dev)) == NULL)
4122                         continue;
4123                 addrconf_ifdown(dev, 1);
4124         }
4125         addrconf_ifdown(&loopback_dev, 2);
4126
4127         /*
4128          *      Check hash table.
4129          */
4130
4131         write_lock_bh(&addrconf_hash_lock);
4132         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4133                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4134                         struct inet6_ifaddr *bifa;
4135
4136                         bifa = ifa;
4137                         ifa = ifa->lst_next;
4138                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4139                         /* Do not free it; something is wrong.
4140                            Now we can investigate it with debugger.
4141                          */
4142                 }
4143         }
4144         write_unlock_bh(&addrconf_hash_lock);
4145
4146         del_timer(&addr_chk_timer);
4147
4148         rtnl_unlock();
4149
4150 #ifdef CONFIG_PROC_FS
4151         proc_net_remove("if_inet6");
4152 #endif
4153 }