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         u32 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             dev->type == ARPHRD_NONE ||
 400             dev->type == ARPHRD_SIT) {
 401                 printk(KERN_INFO
 402                        "%s: Disabled Privacy Extensions\n",
 403                        dev->name);
 404                 ndev->cnf.use_tempaddr = -1;
 405         } else {
 406                 in6_dev_hold(ndev);
 407                 ipv6_regen_rndid((unsigned long) ndev);
 408         }
 409 #endif
 410
 411         if (netif_carrier_ok(dev))
 412                 ndev->if_flags |= IF_READY;
 413
 414         /* protected by rtnl_lock */
 415         rcu_assign_pointer(dev->ip6_ptr, ndev);
 416
 417         ipv6_mc_init_dev(ndev);
 418         ndev->tstamp = jiffies;
 419 #ifdef CONFIG_SYSCTL
 420         neigh_sysctl_register(dev, ndev->nd_parms, NET_IPV6,
 421                               NET_IPV6_NEIGH, "ipv6",
 422                               &ndisc_ifinfo_sysctl_change,
 423                               NULL);
 424         addrconf_sysctl_register(ndev, &ndev->cnf);
 425 #endif
 426         return ndev;
 427 }
 428
 429 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 430 {
 431         struct inet6_dev *idev;
 432
 433         ASSERT_RTNL();
 434
 435         if ((idev = __in6_dev_get(dev)) == NULL) {
 436                 if ((idev = ipv6_add_dev(dev)) == NULL)
 437                         return NULL;
 438         }
 439
 440         if (dev->flags&IFF_UP)
 441                 ipv6_mc_up(idev);
 442         return idev;
 443 }
 444
 445 #ifdef CONFIG_SYSCTL
 446 static void dev_forward_change(struct inet6_dev *idev)
 447 {
 448         struct net_device *dev;
 449         struct inet6_ifaddr *ifa;
 450         struct in6_addr addr;
 451
 452         if (!idev)
 453                 return;
 454         dev = idev->dev;
 455         if (dev && (dev->flags & IFF_MULTICAST)) {
 456                 ipv6_addr_all_routers(&addr);
 457
 458                 if (idev->cnf.forwarding)
 459                         ipv6_dev_mc_inc(dev, &addr);
 460                 else
 461                         ipv6_dev_mc_dec(dev, &addr);
 462         }
 463         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 464                 if (idev->cnf.forwarding)
 465                         addrconf_join_anycast(ifa);
 466                 else
 467                         addrconf_leave_anycast(ifa);
 468         }
 469 }
 470
 471
 472 static void addrconf_forward_change(void)
 473 {
 474         struct net_device *dev;
 475         struct inet6_dev *idev;
 476
 477         read_lock(&dev_base_lock);
 478         for (dev=dev_base; dev; dev=dev->next) {
 479                 rcu_read_lock();
 480                 idev = __in6_dev_get(dev);
 481                 if (idev) {
 482                         int changed = (!idev->cnf.forwarding) ^ (!ipv6_devconf.forwarding);
 483                         idev->cnf.forwarding = ipv6_devconf.forwarding;
 484                         if (changed)
 485                                 dev_forward_change(idev);
 486                 }
 487                 rcu_read_unlock();
 488         }
 489         read_unlock(&dev_base_lock);
 490 }
 491 #endif
 492
 493 /* Nobody refers to this ifaddr, destroy it */
 494
 495 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 496 {
 497         BUG_TRAP(ifp->if_next==NULL);
 498         BUG_TRAP(ifp->lst_next==NULL);
 499 #ifdef NET_REFCNT_DEBUG
 500         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 501 #endif
 502
 503         in6_dev_put(ifp->idev);
 504
 505         if (del_timer(&ifp->timer))
 506                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 507
 508         if (!ifp->dead) {
 509                 printk("Freeing alive inet6 address %p\n", ifp);
 510                 return;
 511         }
 512         dst_release(&ifp->rt->u.dst);
 513
 514         kfree(ifp);
 515 }
 516
 517 static void
 518 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 519 {
 520         struct inet6_ifaddr *ifa, **ifap;
 521         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 522
 523         /*
 524          * Each device address list is sorted in order of scope -
 525          * global before linklocal.
 526          */
 527         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 528              ifap = &ifa->if_next) {
 529                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 530                         break;
 531         }
 532
 533         ifp->if_next = *ifap;
 534         *ifap = ifp;
 535 }
 536
 537 /* On success it returns ifp with increased reference count */
 538
 539 static struct inet6_ifaddr *
 540 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 541               int scope, u32 flags)
 542 {
 543         struct inet6_ifaddr *ifa = NULL;
 544         struct rt6_info *rt;
 545         int hash;
 546         int err = 0;
 547
 548         rcu_read_lock_bh();
 549         if (idev->dead) {
 550                 err = -ENODEV;                  /*XXX*/
 551                 goto out2;
 552         }
 553
 554         write_lock(&addrconf_hash_lock);
 555
 556         /* Ignore adding duplicate addresses on an interface */
 557         if (ipv6_chk_same_addr(addr, idev->dev)) {
 558                 ADBG(("ipv6_add_addr: already assigned\n"));
 559                 err = -EEXIST;
 560                 goto out;
 561         }
 562
 563         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 564
 565         if (ifa == NULL) {
 566                 ADBG(("ipv6_add_addr: malloc failed\n"));
 567                 err = -ENOBUFS;
 568                 goto out;
 569         }
 570
 571         rt = addrconf_dst_alloc(idev, addr, 0);
 572         if (IS_ERR(rt)) {
 573                 err = PTR_ERR(rt);
 574                 goto out;
 575         }
 576
 577         ipv6_addr_copy(&ifa->addr, addr);
 578
 579         spin_lock_init(&ifa->lock);
 580         init_timer(&ifa->timer);
 581         ifa->timer.data = (unsigned long) ifa;
 582         ifa->scope = scope;
 583         ifa->prefix_len = pfxlen;
 584         ifa->flags = flags | IFA_F_TENTATIVE;
 585         ifa->cstamp = ifa->tstamp = jiffies;
 586
 587         ifa->rt = rt;
 588
 589         ifa->idev = idev;
 590         in6_dev_hold(idev);
 591         /* For caller */
 592         in6_ifa_hold(ifa);
 593
 594         /* Add to big hash table */
 595         hash = ipv6_addr_hash(addr);
 596
 597         ifa->lst_next = inet6_addr_lst[hash];
 598         inet6_addr_lst[hash] = ifa;
 599         in6_ifa_hold(ifa);
 600         write_unlock(&addrconf_hash_lock);
 601
 602         write_lock(&idev->lock);
 603         /* Add to inet6_dev unicast addr list. */
 604         ipv6_link_dev_addr(idev, ifa);
 605
 606 #ifdef CONFIG_IPV6_PRIVACY
 607         if (ifa->flags&IFA_F_TEMPORARY) {
 608                 ifa->tmp_next = idev->tempaddr_list;
 609                 idev->tempaddr_list = ifa;
 610                 in6_ifa_hold(ifa);
 611         }
 612 #endif
 613
 614         in6_ifa_hold(ifa);
 615         write_unlock(&idev->lock);
 616 out2:
 617         rcu_read_unlock_bh();
 618
 619         if (likely(err == 0))
 620                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 621         else {
 622                 kfree(ifa);
 623                 ifa = ERR_PTR(err);
 624         }
 625
 626         return ifa;
 627 out:
 628         write_unlock(&addrconf_hash_lock);
 629         goto out2;
 630 }
 631
 632 /* This function wants to get referenced ifp and releases it before return */
 633
 634 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 635 {
 636         struct inet6_ifaddr *ifa, **ifap;
 637         struct inet6_dev *idev = ifp->idev;
 638         int hash;
 639         int deleted = 0, onlink = 0;
 640         unsigned long expires = jiffies;
 641
 642         hash = ipv6_addr_hash(&ifp->addr);
 643
 644         ifp->dead = 1;
 645
 646         write_lock_bh(&addrconf_hash_lock);
 647         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 648              ifap = &ifa->lst_next) {
 649                 if (ifa == ifp) {
 650                         *ifap = ifa->lst_next;
 651                         __in6_ifa_put(ifp);
 652                         ifa->lst_next = NULL;
 653                         break;
 654                 }
 655         }
 656         write_unlock_bh(&addrconf_hash_lock);
 657
 658         write_lock_bh(&idev->lock);
 659 #ifdef CONFIG_IPV6_PRIVACY
 660         if (ifp->flags&IFA_F_TEMPORARY) {
 661                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 662                      ifap = &ifa->tmp_next) {
 663                         if (ifa == ifp) {
 664                                 *ifap = ifa->tmp_next;
 665                                 if (ifp->ifpub) {
 666                                         in6_ifa_put(ifp->ifpub);
 667                                         ifp->ifpub = NULL;
 668                                 }
 669                                 __in6_ifa_put(ifp);
 670                                 ifa->tmp_next = NULL;
 671                                 break;
 672                         }
 673                 }
 674         }
 675 #endif
 676
 677         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 678                 if (ifa == ifp) {
 679                         *ifap = ifa->if_next;
 680                         __in6_ifa_put(ifp);
 681                         ifa->if_next = NULL;
 682                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 683                                 break;
 684                         deleted = 1;
 685                         continue;
 686                 } else if (ifp->flags & IFA_F_PERMANENT) {
 687                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 688                                               ifp->prefix_len)) {
 689                                 if (ifa->flags & IFA_F_PERMANENT) {
 690                                         onlink = 1;
 691                                         if (deleted)
 692                                                 break;
 693                                 } else {
 694                                         unsigned long lifetime;
 695
 696                                         if (!onlink)
 697                                                 onlink = -1;
 698
 699                                         spin_lock(&ifa->lock);
 700                                         lifetime = min_t(unsigned long,
 701                                                          ifa->valid_lft, 0x7fffffffUL/HZ);
 702                                         if (time_before(expires,
 703                                                         ifa->tstamp + lifetime * HZ))
 704                                                 expires = ifa->tstamp + lifetime * HZ;
 705                                         spin_unlock(&ifa->lock);
 706                                 }
 707                         }
 708                 }
 709                 ifap = &ifa->if_next;
 710         }
 711         write_unlock_bh(&idev->lock);
 712
 713         ipv6_ifa_notify(RTM_DELADDR, ifp);
 714
 715         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 716
 717         addrconf_del_timer(ifp);
 718
 719         /*
 720          * Purge or update corresponding prefix
 721          *
 722          * 1) we don't purge prefix here if address was not permanent.
 723          *    prefix is managed by its own lifetime.
 724          * 2) if there're no addresses, delete prefix.
 725          * 3) if there're still other permanent address(es),
 726          *    corresponding prefix is still permanent.
 727          * 4) otherwise, update prefix lifetime to the
 728          *    longest valid lifetime among the corresponding
 729          *    addresses on the device.
 730          *    Note: subsequent RA will update lifetime.
 731          *
 732          * --yoshfuji
 733          */
 734         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 735                 struct in6_addr prefix;
 736                 struct rt6_info *rt;
 737
 738                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 739                 rt = rt6_lookup(&prefix, NULL, ifp->idev->dev->ifindex, 1);
 740
 741                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
 742                         if (onlink == 0) {
 743                                 ip6_del_rt(rt);
 744                                 rt = NULL;
 745                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 746                                 rt->rt6i_expires = expires;
 747                                 rt->rt6i_flags |= RTF_EXPIRES;
 748                         }
 749                 }
 750                 dst_release(&rt->u.dst);
 751         }
 752
 753         in6_ifa_put(ifp);
 754 }
 755
 756 #ifdef CONFIG_IPV6_PRIVACY
 757 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 758 {
 759         struct inet6_dev *idev = ifp->idev;
 760         struct in6_addr addr, *tmpaddr;
 761         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 762         int tmp_plen;
 763         int ret = 0;
 764         int max_addresses;
 765
 766         write_lock(&idev->lock);
 767         if (ift) {
 768                 spin_lock_bh(&ift->lock);
 769                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 770                 spin_unlock_bh(&ift->lock);
 771                 tmpaddr = &addr;
 772         } else {
 773                 tmpaddr = NULL;
 774         }
 775 retry:
 776         in6_dev_hold(idev);
 777         if (idev->cnf.use_tempaddr <= 0) {
 778                 write_unlock(&idev->lock);
 779                 printk(KERN_INFO
 780                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 781                 in6_dev_put(idev);
 782                 ret = -1;
 783                 goto out;
 784         }
 785         spin_lock_bh(&ifp->lock);
 786         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 787                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 788                 spin_unlock_bh(&ifp->lock);
 789                 write_unlock(&idev->lock);
 790                 printk(KERN_WARNING
 791                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 792                 in6_dev_put(idev);
 793                 ret = -1;
 794                 goto out;
 795         }
 796         in6_ifa_hold(ifp);
 797         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 798         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 799                 spin_unlock_bh(&ifp->lock);
 800                 write_unlock(&idev->lock);
 801                 printk(KERN_WARNING
 802                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 803                 in6_ifa_put(ifp);
 804                 in6_dev_put(idev);
 805                 ret = -1;
 806                 goto out;
 807         }
 808         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 809         tmp_valid_lft = min_t(__u32,
 810                               ifp->valid_lft,
 811                               idev->cnf.temp_valid_lft);
 812         tmp_prefered_lft = min_t(__u32,
 813                                  ifp->prefered_lft,
 814                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 815         tmp_plen = ifp->prefix_len;
 816         max_addresses = idev->cnf.max_addresses;
 817         tmp_cstamp = ifp->cstamp;
 818         tmp_tstamp = ifp->tstamp;
 819         spin_unlock_bh(&ifp->lock);
 820
 821         write_unlock(&idev->lock);
 822         ift = !max_addresses ||
 823               ipv6_count_addresses(idev) < max_addresses ?
 824                 ipv6_add_addr(idev, &addr, tmp_plen,
 825                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK, IFA_F_TEMPORARY) : NULL;
 826         if (!ift || IS_ERR(ift)) {
 827                 in6_ifa_put(ifp);
 828                 in6_dev_put(idev);
 829                 printk(KERN_INFO
 830                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 831                 tmpaddr = &addr;
 832                 write_lock(&idev->lock);
 833                 goto retry;
 834         }
 835
 836         spin_lock_bh(&ift->lock);
 837         ift->ifpub = ifp;
 838         ift->valid_lft = tmp_valid_lft;
 839         ift->prefered_lft = tmp_prefered_lft;
 840         ift->cstamp = tmp_cstamp;
 841         ift->tstamp = tmp_tstamp;
 842         spin_unlock_bh(&ift->lock);
 843
 844         addrconf_dad_start(ift, 0);
 845         in6_ifa_put(ift);
 846         in6_dev_put(idev);
 847 out:
 848         return ret;
 849 }
 850 #endif
 851
 852 /*
 853  *      Choose an appropriate source address (RFC3484)
 854  */
 855 struct ipv6_saddr_score {
 856         int             addr_type;
 857         unsigned int    attrs;
 858         int             matchlen;
 859         int             scope;
 860         unsigned int    rule;
 861 };
 862
 863 #define IPV6_SADDR_SCORE_LOCAL          0x0001
 864 #define IPV6_SADDR_SCORE_PREFERRED      0x0004
 865 #define IPV6_SADDR_SCORE_HOA            0x0008
 866 #define IPV6_SADDR_SCORE_OIF            0x0010
 867 #define IPV6_SADDR_SCORE_LABEL          0x0020
 868 #define IPV6_SADDR_SCORE_PRIVACY        0x0040
 869
 870 static int inline ipv6_saddr_preferred(int type)
 871 {
 872         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 873                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 874                 return 1;
 875         return 0;
 876 }
 877
 878 /* static matching label */
 879 static int inline ipv6_saddr_label(const struct in6_addr *addr, int type)
 880 {
 881  /*
 882   *     prefix (longest match)  label
 883   *     -----------------------------
 884   *     ::1/128                 0
 885   *     ::/0                    1
 886   *     2002::/16               2
 887   *     ::/96                   3
 888   *     ::ffff:0:0/96           4
 889   *     fc00::/7                5
 890   *     2001::/32               6
 891   */
 892         if (type & IPV6_ADDR_LOOPBACK)
 893                 return 0;
 894         else if (type & IPV6_ADDR_COMPATv4)
 895                 return 3;
 896         else if (type & IPV6_ADDR_MAPPED)
 897                 return 4;
 898         else if (addr->s6_addr32[0] == htonl(0x20010000))
 899                 return 6;
 900         else if (addr->s6_addr16[0] == htons(0x2002))
 901                 return 2;
 902         else if ((addr->s6_addr[0] & 0xfe) == 0xfc)
 903                 return 5;
 904         return 1;
 905 }
 906
 907 int ipv6_dev_get_saddr(struct net_device *daddr_dev,
 908                        struct in6_addr *daddr, struct in6_addr *saddr)
 909 {
 910         struct ipv6_saddr_score hiscore;
 911         struct inet6_ifaddr *ifa_result = NULL;
 912         int daddr_type = __ipv6_addr_type(daddr);
 913         int daddr_scope = __ipv6_addr_src_scope(daddr_type);
 914         u32 daddr_label = ipv6_saddr_label(daddr, daddr_type);
 915         struct net_device *dev;
 916
 917         memset(&hiscore, 0, sizeof(hiscore));
 918
 919         read_lock(&dev_base_lock);
 920         rcu_read_lock();
 921
 922         for (dev = dev_base; dev; dev=dev->next) {
 923                 struct inet6_dev *idev;
 924                 struct inet6_ifaddr *ifa;
 925
 926                 /* Rule 0: Candidate Source Address (section 4)
 927                  *  - multicast and link-local destination address,
 928                  *    the set of candidate source address MUST only
 929                  *    include addresses assigned to interfaces
 930                  *    belonging to the same link as the outgoing
 931                  *    interface.
 932                  * (- For site-local destination addresses, the
 933                  *    set of candidate source addresses MUST only
 934                  *    include addresses assigned to interfaces
 935                  *    belonging to the same site as the outgoing
 936                  *    interface.)
 937                  */
 938                 if ((daddr_type & IPV6_ADDR_MULTICAST ||
 939                      daddr_scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
 940                     daddr_dev && dev != daddr_dev)
 941                         continue;
 942
 943                 idev = __in6_dev_get(dev);
 944                 if (!idev)
 945                         continue;
 946
 947                 read_lock_bh(&idev->lock);
 948                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
 949                         struct ipv6_saddr_score score;
 950
 951                         score.addr_type = __ipv6_addr_type(&ifa->addr);
 952
 953                         /* Rule 0:
 954                          * - Tentative Address (RFC2462 section 5.4)
 955                          *  - A tentative address is not considered
 956                          *    "assigned to an interface" in the traditional
 957                          *    sense.
 958                          * - Candidate Source Address (section 4)
 959                          *  - In any case, anycast addresses, multicast
 960                          *    addresses, and the unspecified address MUST
 961                          *    NOT be included in a candidate set.
 962                          */
 963                         if (ifa->flags & IFA_F_TENTATIVE)
 964                                 continue;
 965                         if (unlikely(score.addr_type == IPV6_ADDR_ANY ||
 966                                      score.addr_type & IPV6_ADDR_MULTICAST)) {
 967                                 LIMIT_NETDEBUG(KERN_DEBUG
 968                                                "ADDRCONF: unspecified / multicast address"
 969                                                "assigned as unicast address on %s",
 970                                                dev->name);
 971                                 continue;
 972                         }
 973
 974                         score.attrs = 0;
 975                         score.matchlen = 0;
 976                         score.scope = 0;
 977                         score.rule = 0;
 978
 979                         if (ifa_result == NULL) {
 980                                 /* record it if the first available entry */
 981                                 goto record_it;
 982                         }
 983
 984                         /* Rule 1: Prefer same address */
 985                         if (hiscore.rule < 1) {
 986                                 if (ipv6_addr_equal(&ifa_result->addr, daddr))
 987                                         hiscore.attrs |= IPV6_SADDR_SCORE_LOCAL;
 988                                 hiscore.rule++;
 989                         }
 990                         if (ipv6_addr_equal(&ifa->addr, daddr)) {
 991                                 score.attrs |= IPV6_SADDR_SCORE_LOCAL;
 992                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)) {
 993                                         score.rule = 1;
 994                                         goto record_it;
 995                                 }
 996                         } else {
 997                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LOCAL)
 998                                         continue;
 999                         }
1000
1001                         /* Rule 2: Prefer appropriate scope */
1002                         if (hiscore.rule < 2) {
1003                                 hiscore.scope = __ipv6_addr_src_scope(hiscore.addr_type);
1004                                 hiscore.rule++;
1005                         }
1006                         score.scope = __ipv6_addr_src_scope(score.addr_type);
1007                         if (hiscore.scope < score.scope) {
1008                                 if (hiscore.scope < daddr_scope) {
1009                                         score.rule = 2;
1010                                         goto record_it;
1011                                 } else
1012                                         continue;
1013                         } else if (score.scope < hiscore.scope) {
1014                                 if (score.scope < daddr_scope)
1015                                         break; /* addresses sorted by scope */
1016                                 else {
1017                                         score.rule = 2;
1018                                         goto record_it;
1019                                 }
1020                         }
1021
1022                         /* Rule 3: Avoid deprecated address */
1023                         if (hiscore.rule < 3) {
1024                                 if (ipv6_saddr_preferred(hiscore.addr_type) ||
1025                                     !(ifa_result->flags & IFA_F_DEPRECATED))
1026                                         hiscore.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1027                                 hiscore.rule++;
1028                         }
1029                         if (ipv6_saddr_preferred(score.addr_type) ||
1030                             !(ifa->flags & IFA_F_DEPRECATED)) {
1031                                 score.attrs |= IPV6_SADDR_SCORE_PREFERRED;
1032                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)) {
1033                                         score.rule = 3;
1034                                         goto record_it;
1035                                 }
1036                         } else {
1037                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PREFERRED)
1038                                         continue;
1039                         }
1040
1041                         /* Rule 4: Prefer home address -- not implemented yet */
1042                         if (hiscore.rule < 4)
1043                                 hiscore.rule++;
1044
1045                         /* Rule 5: Prefer outgoing interface */
1046                         if (hiscore.rule < 5) {
1047                                 if (daddr_dev == NULL ||
1048                                     daddr_dev == ifa_result->idev->dev)
1049                                         hiscore.attrs |= IPV6_SADDR_SCORE_OIF;
1050                                 hiscore.rule++;
1051                         }
1052                         if (daddr_dev == NULL ||
1053                             daddr_dev == ifa->idev->dev) {
1054                                 score.attrs |= IPV6_SADDR_SCORE_OIF;
1055                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_OIF)) {
1056                                         score.rule = 5;
1057                                         goto record_it;
1058                                 }
1059                         } else {
1060                                 if (hiscore.attrs & IPV6_SADDR_SCORE_OIF)
1061                                         continue;
1062                         }
1063
1064                         /* Rule 6: Prefer matching label */
1065                         if (hiscore.rule < 6) {
1066                                 if (ipv6_saddr_label(&ifa_result->addr, hiscore.addr_type) == daddr_label)
1067                                         hiscore.attrs |= IPV6_SADDR_SCORE_LABEL;
1068                                 hiscore.rule++;
1069                         }
1070                         if (ipv6_saddr_label(&ifa->addr, score.addr_type) == daddr_label) {
1071                                 score.attrs |= IPV6_SADDR_SCORE_LABEL;
1072                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_LABEL)) {
1073                                         score.rule = 6;
1074                                         goto record_it;
1075                                 }
1076                         } else {
1077                                 if (hiscore.attrs & IPV6_SADDR_SCORE_LABEL)
1078                                         continue;
1079                         }
1080
1081 #ifdef CONFIG_IPV6_PRIVACY
1082                         /* Rule 7: Prefer public address
1083                          * Note: prefer temprary address if use_tempaddr >= 2
1084                          */
1085                         if (hiscore.rule < 7) {
1086                                 if ((!(ifa_result->flags & IFA_F_TEMPORARY)) ^
1087                                     (ifa_result->idev->cnf.use_tempaddr >= 2))
1088                                         hiscore.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1089                                 hiscore.rule++;
1090                         }
1091                         if ((!(ifa->flags & IFA_F_TEMPORARY)) ^
1092                             (ifa->idev->cnf.use_tempaddr >= 2)) {
1093                                 score.attrs |= IPV6_SADDR_SCORE_PRIVACY;
1094                                 if (!(hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)) {
1095                                         score.rule = 7;
1096                                         goto record_it;
1097                                 }
1098                         } else {
1099                                 if (hiscore.attrs & IPV6_SADDR_SCORE_PRIVACY)
1100                                         continue;
1101                         }
1102 #else
1103                         if (hiscore.rule < 7)
1104                                 hiscore.rule++;
1105 #endif
1106                         /* Rule 8: Use longest matching prefix */
1107                         if (hiscore.rule < 8) {
1108                                 hiscore.matchlen = ipv6_addr_diff(&ifa_result->addr, daddr);
1109                                 hiscore.rule++;
1110                         }
1111                         score.matchlen = ipv6_addr_diff(&ifa->addr, daddr);
1112                         if (score.matchlen > hiscore.matchlen) {
1113                                 score.rule = 8;
1114                                 goto record_it;
1115                         }
1116 #if 0
1117                         else if (score.matchlen < hiscore.matchlen)
1118                                 continue;
1119 #endif
1120
1121                         /* Final Rule: choose first available one */
1122                         continue;
1123 record_it:
1124                         if (ifa_result)
1125                                 in6_ifa_put(ifa_result);
1126                         in6_ifa_hold(ifa);
1127                         ifa_result = ifa;
1128                         hiscore = score;
1129                 }
1130                 read_unlock_bh(&idev->lock);
1131         }
1132         rcu_read_unlock();
1133         read_unlock(&dev_base_lock);
1134
1135         if (!ifa_result)
1136                 return -EADDRNOTAVAIL;
1137
1138         ipv6_addr_copy(saddr, &ifa_result->addr);
1139         in6_ifa_put(ifa_result);
1140         return 0;
1141 }
1142
1143
1144 int ipv6_get_saddr(struct dst_entry *dst,
1145                    struct in6_addr *daddr, struct in6_addr *saddr)
1146 {
1147         return ipv6_dev_get_saddr(dst ? ((struct rt6_info *)dst)->rt6i_idev->dev : NULL, daddr, saddr);
1148 }
1149
1150
1151 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr)
1152 {
1153         struct inet6_dev *idev;
1154         int err = -EADDRNOTAVAIL;
1155
1156         rcu_read_lock();
1157         if ((idev = __in6_dev_get(dev)) != NULL) {
1158                 struct inet6_ifaddr *ifp;
1159
1160                 read_lock_bh(&idev->lock);
1161                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1162                         if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1163                                 ipv6_addr_copy(addr, &ifp->addr);
1164                                 err = 0;
1165                                 break;
1166                         }
1167                 }
1168                 read_unlock_bh(&idev->lock);
1169         }
1170         rcu_read_unlock();
1171         return err;
1172 }
1173
1174 static int ipv6_count_addresses(struct inet6_dev *idev)
1175 {
1176         int cnt = 0;
1177         struct inet6_ifaddr *ifp;
1178
1179         read_lock_bh(&idev->lock);
1180         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1181                 cnt++;
1182         read_unlock_bh(&idev->lock);
1183         return cnt;
1184 }
1185
1186 int ipv6_chk_addr(struct in6_addr *addr, struct net_device *dev, int strict)
1187 {
1188         struct inet6_ifaddr * ifp;
1189         u8 hash = ipv6_addr_hash(addr);
1190
1191         read_lock_bh(&addrconf_hash_lock);
1192         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1193                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1194                     !(ifp->flags&IFA_F_TENTATIVE)) {
1195                         if (dev == NULL || ifp->idev->dev == dev ||
1196                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1197                                 break;
1198                 }
1199         }
1200         read_unlock_bh(&addrconf_hash_lock);
1201         return ifp != NULL;
1202 }
1203
1204 static
1205 int ipv6_chk_same_addr(const struct in6_addr *addr, struct net_device *dev)
1206 {
1207         struct inet6_ifaddr * ifp;
1208         u8 hash = ipv6_addr_hash(addr);
1209
1210         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1211                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1212                         if (dev == NULL || ifp->idev->dev == dev)
1213                                 break;
1214                 }
1215         }
1216         return ifp != NULL;
1217 }
1218
1219 struct inet6_ifaddr * ipv6_get_ifaddr(struct in6_addr *addr, struct net_device *dev, int strict)
1220 {
1221         struct inet6_ifaddr * ifp;
1222         u8 hash = ipv6_addr_hash(addr);
1223
1224         read_lock_bh(&addrconf_hash_lock);
1225         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1226                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1227                         if (dev == NULL || ifp->idev->dev == dev ||
1228                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1229                                 in6_ifa_hold(ifp);
1230                                 break;
1231                         }
1232                 }
1233         }
1234         read_unlock_bh(&addrconf_hash_lock);
1235
1236         return ifp;
1237 }
1238
1239 int ipv6_rcv_saddr_equal(const struct sock *sk, const struct sock *sk2)
1240 {
1241         const struct in6_addr *sk_rcv_saddr6 = &inet6_sk(sk)->rcv_saddr;
1242         const struct in6_addr *sk2_rcv_saddr6 = inet6_rcv_saddr(sk2);
1243         u32 sk_rcv_saddr = inet_sk(sk)->rcv_saddr;
1244         u32 sk2_rcv_saddr = inet_rcv_saddr(sk2);
1245         int sk_ipv6only = ipv6_only_sock(sk);
1246         int sk2_ipv6only = inet_v6_ipv6only(sk2);
1247         int addr_type = ipv6_addr_type(sk_rcv_saddr6);
1248         int addr_type2 = sk2_rcv_saddr6 ? ipv6_addr_type(sk2_rcv_saddr6) : IPV6_ADDR_MAPPED;
1249
1250         if (!sk2_rcv_saddr && !sk_ipv6only)
1251                 return 1;
1252
1253         if (addr_type2 == IPV6_ADDR_ANY &&
1254             !(sk2_ipv6only && addr_type == IPV6_ADDR_MAPPED))
1255                 return 1;
1256
1257         if (addr_type == IPV6_ADDR_ANY &&
1258             !(sk_ipv6only && addr_type2 == IPV6_ADDR_MAPPED))
1259                 return 1;
1260
1261         if (sk2_rcv_saddr6 &&
1262             ipv6_addr_equal(sk_rcv_saddr6, sk2_rcv_saddr6))
1263                 return 1;
1264
1265         if (addr_type == IPV6_ADDR_MAPPED &&
1266             !sk2_ipv6only &&
1267             (!sk2_rcv_saddr || !sk_rcv_saddr || sk_rcv_saddr == sk2_rcv_saddr))
1268                 return 1;
1269
1270         return 0;
1271 }
1272
1273 /* Gets referenced address, destroys ifaddr */
1274
1275 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1276 {
1277         if (ifp->flags&IFA_F_PERMANENT) {
1278                 spin_lock_bh(&ifp->lock);
1279                 addrconf_del_timer(ifp);
1280                 ifp->flags |= IFA_F_TENTATIVE;
1281                 spin_unlock_bh(&ifp->lock);
1282                 in6_ifa_put(ifp);
1283 #ifdef CONFIG_IPV6_PRIVACY
1284         } else if (ifp->flags&IFA_F_TEMPORARY) {
1285                 struct inet6_ifaddr *ifpub;
1286                 spin_lock_bh(&ifp->lock);
1287                 ifpub = ifp->ifpub;
1288                 if (ifpub) {
1289                         in6_ifa_hold(ifpub);
1290                         spin_unlock_bh(&ifp->lock);
1291                         ipv6_create_tempaddr(ifpub, ifp);
1292                         in6_ifa_put(ifpub);
1293                 } else {
1294                         spin_unlock_bh(&ifp->lock);
1295                 }
1296                 ipv6_del_addr(ifp);
1297 #endif
1298         } else
1299                 ipv6_del_addr(ifp);
1300 }
1301
1302 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1303 {
1304         if (net_ratelimit())
1305                 printk(KERN_INFO "%s: duplicate address detected!\n", ifp->idev->dev->name);
1306         addrconf_dad_stop(ifp);
1307 }
1308
1309 /* Join to solicited addr multicast group. */
1310
1311 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1312 {
1313         struct in6_addr maddr;
1314
1315         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1316                 return;
1317
1318         addrconf_addr_solict_mult(addr, &maddr);
1319         ipv6_dev_mc_inc(dev, &maddr);
1320 }
1321
1322 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1323 {
1324         struct in6_addr maddr;
1325
1326         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1327                 return;
1328
1329         addrconf_addr_solict_mult(addr, &maddr);
1330         __ipv6_dev_mc_dec(idev, &maddr);
1331 }
1332
1333 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1334 {
1335         struct in6_addr addr;
1336         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1337         if (ipv6_addr_any(&addr))
1338                 return;
1339         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1340 }
1341
1342 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1343 {
1344         struct in6_addr addr;
1345         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1346         if (ipv6_addr_any(&addr))
1347                 return;
1348         __ipv6_dev_ac_dec(ifp->idev, &addr);
1349 }
1350
1351 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1352 {
1353         if (dev->addr_len != ETH_ALEN)
1354                 return -1;
1355         memcpy(eui, dev->dev_addr, 3);
1356         memcpy(eui + 5, dev->dev_addr + 3, 3);
1357
1358         /*
1359          * The zSeries OSA network cards can be shared among various
1360          * OS instances, but the OSA cards have only one MAC address.
1361          * This leads to duplicate address conflicts in conjunction
1362          * with IPv6 if more than one instance uses the same card.
1363          *
1364          * The driver for these cards can deliver a unique 16-bit
1365          * identifier for each instance sharing the same card.  It is
1366          * placed instead of 0xFFFE in the interface identifier.  The
1367          * "u" bit of the interface identifier is not inverted in this
1368          * case.  Hence the resulting interface identifier has local
1369          * scope according to RFC2373.
1370          */
1371         if (dev->dev_id) {
1372                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1373                 eui[4] = dev->dev_id & 0xFF;
1374         } else {
1375                 eui[3] = 0xFF;
1376                 eui[4] = 0xFE;
1377                 eui[0] ^= 2;
1378         }
1379         return 0;
1380 }
1381
1382 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1383 {
1384         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1385         if (dev->addr_len != ARCNET_ALEN)
1386                 return -1;
1387         memset(eui, 0, 7);
1388         eui[7] = *(u8*)dev->dev_addr;
1389         return 0;
1390 }
1391
1392 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1393 {
1394         if (dev->addr_len != INFINIBAND_ALEN)
1395                 return -1;
1396         memcpy(eui, dev->dev_addr + 12, 8);
1397         eui[0] |= 2;
1398         return 0;
1399 }
1400
1401 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1402 {
1403         switch (dev->type) {
1404         case ARPHRD_ETHER:
1405         case ARPHRD_FDDI:
1406         case ARPHRD_IEEE802_TR:
1407                 return addrconf_ifid_eui48(eui, dev);
1408         case ARPHRD_ARCNET:
1409                 return addrconf_ifid_arcnet(eui, dev);
1410         case ARPHRD_INFINIBAND:
1411                 return addrconf_ifid_infiniband(eui, dev);
1412         }
1413         return -1;
1414 }
1415
1416 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1417 {
1418         int err = -1;
1419         struct inet6_ifaddr *ifp;
1420
1421         read_lock_bh(&idev->lock);
1422         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1423                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1424                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1425                         err = 0;
1426                         break;
1427                 }
1428         }
1429         read_unlock_bh(&idev->lock);
1430         return err;
1431 }
1432
1433 #ifdef CONFIG_IPV6_PRIVACY
1434 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1435 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1436 {
1437 regen:
1438         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1439         idev->rndid[0] &= ~0x02;
1440
1441         /*
1442          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1443          * check if generated address is not inappropriate
1444          *
1445          *  - Reserved subnet anycast (RFC 2526)
1446          *      11111101 11....11 1xxxxxxx
1447          *  - ISATAP (draft-ietf-ngtrans-isatap-13.txt) 5.1
1448          *      00-00-5E-FE-xx-xx-xx-xx
1449          *  - value 0
1450          *  - XXX: already assigned to an address on the device
1451          */
1452         if (idev->rndid[0] == 0xfd &&
1453             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1454             (idev->rndid[7]&0x80))
1455                 goto regen;
1456         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1457                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1458                         goto regen;
1459                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1460                         goto regen;
1461         }
1462
1463         return 0;
1464 }
1465
1466 static void ipv6_regen_rndid(unsigned long data)
1467 {
1468         struct inet6_dev *idev = (struct inet6_dev *) data;
1469         unsigned long expires;
1470
1471         rcu_read_lock_bh();
1472         write_lock_bh(&idev->lock);
1473
1474         if (idev->dead)
1475                 goto out;
1476
1477         if (__ipv6_regen_rndid(idev) < 0)
1478                 goto out;
1479
1480         expires = jiffies +
1481                 idev->cnf.temp_prefered_lft * HZ -
1482                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1483         if (time_before(expires, jiffies)) {
1484                 printk(KERN_WARNING
1485                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1486                         idev->dev->name);
1487                 goto out;
1488         }
1489
1490         if (!mod_timer(&idev->regen_timer, expires))
1491                 in6_dev_hold(idev);
1492
1493 out:
1494         write_unlock_bh(&idev->lock);
1495         rcu_read_unlock_bh();
1496         in6_dev_put(idev);
1497 }
1498
1499 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1500         int ret = 0;
1501
1502         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1503                 ret = __ipv6_regen_rndid(idev);
1504         return ret;
1505 }
1506 #endif
1507
1508 /*
1509  *      Add prefix route.
1510  */
1511
1512 static void
1513 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1514                       unsigned long expires, u32 flags)
1515 {
1516         struct fib6_config cfg = {
1517                 .fc_table = RT6_TABLE_PREFIX,
1518                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1519                 .fc_ifindex = dev->ifindex,
1520                 .fc_expires = expires,
1521                 .fc_dst_len = plen,
1522                 .fc_flags = RTF_UP | flags,
1523         };
1524
1525         ipv6_addr_copy(&cfg.fc_dst, pfx);
1526
1527         /* Prevent useless cloning on PtP SIT.
1528            This thing is done here expecting that the whole
1529            class of non-broadcast devices need not cloning.
1530          */
1531         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1532                 cfg.fc_flags |= RTF_NONEXTHOP;
1533
1534         ip6_route_add(&cfg);
1535 }
1536
1537 /* Create "default" multicast route to the interface */
1538
1539 static void addrconf_add_mroute(struct net_device *dev)
1540 {
1541         struct fib6_config cfg = {
1542                 .fc_table = RT6_TABLE_LOCAL,
1543                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1544                 .fc_ifindex = dev->ifindex,
1545                 .fc_dst_len = 8,
1546                 .fc_flags = RTF_UP,
1547         };
1548
1549         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1550
1551         ip6_route_add(&cfg);
1552 }
1553
1554 static void sit_route_add(struct net_device *dev)
1555 {
1556         struct fib6_config cfg = {
1557                 .fc_table = RT6_TABLE_MAIN,
1558                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1559                 .fc_ifindex = dev->ifindex,
1560                 .fc_dst_len = 96,
1561                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1562         };
1563
1564         /* prefix length - 96 bits "::d.d.d.d" */
1565         ip6_route_add(&cfg);
1566 }
1567
1568 static void addrconf_add_lroute(struct net_device *dev)
1569 {
1570         struct in6_addr addr;
1571
1572         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1573         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1574 }
1575
1576 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1577 {
1578         struct inet6_dev *idev;
1579
1580         ASSERT_RTNL();
1581
1582         if ((idev = ipv6_find_idev(dev)) == NULL)
1583                 return NULL;
1584
1585         /* Add default multicast route */
1586         addrconf_add_mroute(dev);
1587
1588         /* Add link local route */
1589         addrconf_add_lroute(dev);
1590         return idev;
1591 }
1592
1593 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1594 {
1595         struct prefix_info *pinfo;
1596         __u32 valid_lft;
1597         __u32 prefered_lft;
1598         int addr_type;
1599         unsigned long rt_expires;
1600         struct inet6_dev *in6_dev;
1601
1602         pinfo = (struct prefix_info *) opt;
1603
1604         if (len < sizeof(struct prefix_info)) {
1605                 ADBG(("addrconf: prefix option too short\n"));
1606                 return;
1607         }
1608
1609         /*
1610          *      Validation checks ([ADDRCONF], page 19)
1611          */
1612
1613         addr_type = ipv6_addr_type(&pinfo->prefix);
1614
1615         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1616                 return;
1617
1618         valid_lft = ntohl(pinfo->valid);
1619         prefered_lft = ntohl(pinfo->prefered);
1620
1621         if (prefered_lft > valid_lft) {
1622                 if (net_ratelimit())
1623                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1624                 return;
1625         }
1626
1627         in6_dev = in6_dev_get(dev);
1628
1629         if (in6_dev == NULL) {
1630                 if (net_ratelimit())
1631                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1632                 return;
1633         }
1634
1635         /*
1636          *      Two things going on here:
1637          *      1) Add routes for on-link prefixes
1638          *      2) Configure prefixes with the auto flag set
1639          */
1640
1641         /* Avoid arithmetic overflow. Really, we could
1642            save rt_expires in seconds, likely valid_lft,
1643            but it would require division in fib gc, that it
1644            not good.
1645          */
1646         if (valid_lft >= 0x7FFFFFFF/HZ)
1647                 rt_expires = 0x7FFFFFFF - (0x7FFFFFFF % HZ);
1648         else
1649                 rt_expires = valid_lft * HZ;
1650
1651         /*
1652          * We convert this (in jiffies) to clock_t later.
1653          * Avoid arithmetic overflow there as well.
1654          * Overflow can happen only if HZ < USER_HZ.
1655          */
1656         if (HZ < USER_HZ && rt_expires > 0x7FFFFFFF / USER_HZ)
1657                 rt_expires = 0x7FFFFFFF / USER_HZ;
1658
1659         if (pinfo->onlink) {
1660                 struct rt6_info *rt;
1661                 rt = rt6_lookup(&pinfo->prefix, NULL, dev->ifindex, 1);
1662
1663                 if (rt && ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0)) {
1664                         if (rt->rt6i_flags&RTF_EXPIRES) {
1665                                 if (valid_lft == 0) {
1666                                         ip6_del_rt(rt);
1667                                         rt = NULL;
1668                                 } else {
1669                                         rt->rt6i_expires = jiffies + rt_expires;
1670                                 }
1671                         }
1672                 } else if (valid_lft) {
1673                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1674                                               dev, jiffies_to_clock_t(rt_expires), RTF_ADDRCONF|RTF_EXPIRES|RTF_PREFIX_RT);
1675                 }
1676                 if (rt)
1677                         dst_release(&rt->u.dst);
1678         }
1679
1680         /* Try to figure out our local address for this prefix */
1681
1682         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1683                 struct inet6_ifaddr * ifp;
1684                 struct in6_addr addr;
1685                 int create = 0, update_lft = 0;
1686
1687                 if (pinfo->prefix_len == 64) {
1688                         memcpy(&addr, &pinfo->prefix, 8);
1689                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1690                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1691                                 in6_dev_put(in6_dev);
1692                                 return;
1693                         }
1694                         goto ok;
1695                 }
1696                 if (net_ratelimit())
1697                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1698                                pinfo->prefix_len);
1699                 in6_dev_put(in6_dev);
1700                 return;
1701
1702 ok:
1703
1704                 ifp = ipv6_get_ifaddr(&addr, dev, 1);
1705
1706                 if (ifp == NULL && valid_lft) {
1707                         int max_addresses = in6_dev->cnf.max_addresses;
1708
1709                         /* Do not allow to create too much of autoconfigured
1710                          * addresses; this would be too easy way to crash kernel.
1711                          */
1712                         if (!max_addresses ||
1713                             ipv6_count_addresses(in6_dev) < max_addresses)
1714                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1715                                                     addr_type&IPV6_ADDR_SCOPE_MASK, 0);
1716
1717                         if (!ifp || IS_ERR(ifp)) {
1718                                 in6_dev_put(in6_dev);
1719                                 return;
1720                         }
1721
1722                         update_lft = create = 1;
1723                         ifp->cstamp = jiffies;
1724                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1725                 }
1726
1727                 if (ifp) {
1728                         int flags;
1729                         unsigned long now;
1730 #ifdef CONFIG_IPV6_PRIVACY
1731                         struct inet6_ifaddr *ift;
1732 #endif
1733                         u32 stored_lft;
1734
1735                         /* update lifetime (RFC2462 5.5.3 e) */
1736                         spin_lock(&ifp->lock);
1737                         now = jiffies;
1738                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1739                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1740                         else
1741                                 stored_lft = 0;
1742                         if (!update_lft && stored_lft) {
1743                                 if (valid_lft > MIN_VALID_LIFETIME ||
1744                                     valid_lft > stored_lft)
1745                                         update_lft = 1;
1746                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1747                                         /* valid_lft <= stored_lft is always true */
1748                                         /* XXX: IPsec */
1749                                         update_lft = 0;
1750                                 } else {
1751                                         valid_lft = MIN_VALID_LIFETIME;
1752                                         if (valid_lft < prefered_lft)
1753                                                 prefered_lft = valid_lft;
1754                                         update_lft = 1;
1755                                 }
1756                         }
1757
1758                         if (update_lft) {
1759                                 ifp->valid_lft = valid_lft;
1760                                 ifp->prefered_lft = prefered_lft;
1761                                 ifp->tstamp = now;
1762                                 flags = ifp->flags;
1763                                 ifp->flags &= ~IFA_F_DEPRECATED;
1764                                 spin_unlock(&ifp->lock);
1765
1766                                 if (!(flags&IFA_F_TENTATIVE))
1767                                         ipv6_ifa_notify(0, ifp);
1768                         } else
1769                                 spin_unlock(&ifp->lock);
1770
1771 #ifdef CONFIG_IPV6_PRIVACY
1772                         read_lock_bh(&in6_dev->lock);
1773                         /* update all temporary addresses in the list */
1774                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1775                                 /*
1776                                  * When adjusting the lifetimes of an existing
1777                                  * temporary address, only lower the lifetimes.
1778                                  * Implementations must not increase the
1779                                  * lifetimes of an existing temporary address
1780                                  * when processing a Prefix Information Option.
1781                                  */
1782                                 spin_lock(&ift->lock);
1783                                 flags = ift->flags;
1784                                 if (ift->valid_lft > valid_lft &&
1785                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1786                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1787                                 if (ift->prefered_lft > prefered_lft &&
1788                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1789                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1790                                 spin_unlock(&ift->lock);
1791                                 if (!(flags&IFA_F_TENTATIVE))
1792                                         ipv6_ifa_notify(0, ift);
1793                         }
1794
1795                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1796                                 /*
1797                                  * When a new public address is created as described in [ADDRCONF],
1798                                  * also create a new temporary address.
1799                                  */
1800                                 read_unlock_bh(&in6_dev->lock);
1801                                 ipv6_create_tempaddr(ifp, NULL);
1802                         } else {
1803                                 read_unlock_bh(&in6_dev->lock);
1804                         }
1805 #endif
1806                         in6_ifa_put(ifp);
1807                         addrconf_verify(0);
1808                 }
1809         }
1810         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1811         in6_dev_put(in6_dev);
1812 }
1813
1814 /*
1815  *      Set destination address.
1816  *      Special case for SIT interfaces where we create a new "virtual"
1817  *      device.
1818  */
1819 int addrconf_set_dstaddr(void __user *arg)
1820 {
1821         struct in6_ifreq ireq;
1822         struct net_device *dev;
1823         int err = -EINVAL;
1824
1825         rtnl_lock();
1826
1827         err = -EFAULT;
1828         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1829                 goto err_exit;
1830
1831         dev = __dev_get_by_index(ireq.ifr6_ifindex);
1832
1833         err = -ENODEV;
1834         if (dev == NULL)
1835                 goto err_exit;
1836
1837         if (dev->type == ARPHRD_SIT) {
1838                 struct ifreq ifr;
1839                 mm_segment_t    oldfs;
1840                 struct ip_tunnel_parm p;
1841
1842                 err = -EADDRNOTAVAIL;
1843                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
1844                         goto err_exit;
1845
1846                 memset(&p, 0, sizeof(p));
1847                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
1848                 p.iph.saddr = 0;
1849                 p.iph.version = 4;
1850                 p.iph.ihl = 5;
1851                 p.iph.protocol = IPPROTO_IPV6;
1852                 p.iph.ttl = 64;
1853                 ifr.ifr_ifru.ifru_data = (void __user *)&p;
1854
1855                 oldfs = get_fs(); set_fs(KERNEL_DS);
1856                 err = dev->do_ioctl(dev, &ifr, SIOCADDTUNNEL);
1857                 set_fs(oldfs);
1858
1859                 if (err == 0) {
1860                         err = -ENOBUFS;
1861                         if ((dev = __dev_get_by_name(p.name)) == NULL)
1862                                 goto err_exit;
1863                         err = dev_open(dev);
1864                 }
1865         }
1866
1867 err_exit:
1868         rtnl_unlock();
1869         return err;
1870 }
1871
1872 /*
1873  *      Manual configuration of address on an interface
1874  */
1875 static int inet6_addr_add(int ifindex, struct in6_addr *pfx, int plen,
1876                           __u32 prefered_lft, __u32 valid_lft)
1877 {
1878         struct inet6_ifaddr *ifp;
1879         struct inet6_dev *idev;
1880         struct net_device *dev;
1881         __u8 ifa_flags = 0;
1882         int scope;
1883
1884         ASSERT_RTNL();
1885
1886         /* check the lifetime */
1887         if (!valid_lft || prefered_lft > valid_lft)
1888                 return -EINVAL;
1889
1890         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1891                 return -ENODEV;
1892
1893         if ((idev = addrconf_add_dev(dev)) == NULL)
1894                 return -ENOBUFS;
1895
1896         scope = ipv6_addr_scope(pfx);
1897
1898         if (valid_lft == INFINITY_LIFE_TIME)
1899                 ifa_flags |= IFA_F_PERMANENT;
1900         else if (valid_lft >= 0x7FFFFFFF/HZ)
1901                 valid_lft = 0x7FFFFFFF/HZ;
1902
1903         if (prefered_lft == 0)
1904                 ifa_flags |= IFA_F_DEPRECATED;
1905         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
1906                  (prefered_lft != INFINITY_LIFE_TIME))
1907                 prefered_lft = 0x7FFFFFFF/HZ;
1908
1909         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
1910
1911         if (!IS_ERR(ifp)) {
1912                 spin_lock_bh(&ifp->lock);
1913                 ifp->valid_lft = valid_lft;
1914                 ifp->prefered_lft = prefered_lft;
1915                 ifp->tstamp = jiffies;
1916                 spin_unlock_bh(&ifp->lock);
1917
1918                 addrconf_dad_start(ifp, 0);
1919                 in6_ifa_put(ifp);
1920                 addrconf_verify(0);
1921                 return 0;
1922         }
1923
1924         return PTR_ERR(ifp);
1925 }
1926
1927 static int inet6_addr_del(int ifindex, struct in6_addr *pfx, int plen)
1928 {
1929         struct inet6_ifaddr *ifp;
1930         struct inet6_dev *idev;
1931         struct net_device *dev;
1932
1933         if ((dev = __dev_get_by_index(ifindex)) == NULL)
1934                 return -ENODEV;
1935
1936         if ((idev = __in6_dev_get(dev)) == NULL)
1937                 return -ENXIO;
1938
1939         read_lock_bh(&idev->lock);
1940         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
1941                 if (ifp->prefix_len == plen &&
1942                     ipv6_addr_equal(pfx, &ifp->addr)) {
1943                         in6_ifa_hold(ifp);
1944                         read_unlock_bh(&idev->lock);
1945
1946                         ipv6_del_addr(ifp);
1947
1948                         /* If the last address is deleted administratively,
1949                            disable IPv6 on this interface.
1950                          */
1951                         if (idev->addr_list == NULL)
1952                                 addrconf_ifdown(idev->dev, 1);
1953                         return 0;
1954                 }
1955         }
1956         read_unlock_bh(&idev->lock);
1957         return -EADDRNOTAVAIL;
1958 }
1959
1960
1961 int addrconf_add_ifaddr(void __user *arg)
1962 {
1963         struct in6_ifreq ireq;
1964         int err;
1965
1966         if (!capable(CAP_NET_ADMIN))
1967                 return -EPERM;
1968
1969         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1970                 return -EFAULT;
1971
1972         rtnl_lock();
1973         err = inet6_addr_add(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen,
1974                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
1975         rtnl_unlock();
1976         return err;
1977 }
1978
1979 int addrconf_del_ifaddr(void __user *arg)
1980 {
1981         struct in6_ifreq ireq;
1982         int err;
1983
1984         if (!capable(CAP_NET_ADMIN))
1985                 return -EPERM;
1986
1987         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
1988                 return -EFAULT;
1989
1990         rtnl_lock();
1991         err = inet6_addr_del(ireq.ifr6_ifindex, &ireq.ifr6_addr, ireq.ifr6_prefixlen);
1992         rtnl_unlock();
1993         return err;
1994 }
1995
1996 static void sit_add_v4_addrs(struct inet6_dev *idev)
1997 {
1998         struct inet6_ifaddr * ifp;
1999         struct in6_addr addr;
2000         struct net_device *dev;
2001         int scope;
2002
2003         ASSERT_RTNL();
2004
2005         memset(&addr, 0, sizeof(struct in6_addr));
2006         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2007
2008         if (idev->dev->flags&IFF_POINTOPOINT) {
2009                 addr.s6_addr32[0] = htonl(0xfe800000);
2010                 scope = IFA_LINK;
2011         } else {
2012                 scope = IPV6_ADDR_COMPATv4;
2013         }
2014
2015         if (addr.s6_addr32[3]) {
2016                 ifp = ipv6_add_addr(idev, &addr, 128, scope, IFA_F_PERMANENT);
2017                 if (!IS_ERR(ifp)) {
2018                         spin_lock_bh(&ifp->lock);
2019                         ifp->flags &= ~IFA_F_TENTATIVE;
2020                         spin_unlock_bh(&ifp->lock);
2021                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2022                         in6_ifa_put(ifp);
2023                 }
2024                 return;
2025         }
2026
2027         for (dev = dev_base; dev != NULL; dev = dev->next) {
2028                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2029                 if (in_dev && (dev->flags & IFF_UP)) {
2030                         struct in_ifaddr * ifa;
2031
2032                         int flag = scope;
2033
2034                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2035                                 int plen;
2036
2037                                 addr.s6_addr32[3] = ifa->ifa_local;
2038
2039                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2040                                         continue;
2041                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2042                                         if (idev->dev->flags&IFF_POINTOPOINT)
2043                                                 continue;
2044                                         flag |= IFA_HOST;
2045                                 }
2046                                 if (idev->dev->flags&IFF_POINTOPOINT)
2047                                         plen = 64;
2048                                 else
2049                                         plen = 96;
2050
2051                                 ifp = ipv6_add_addr(idev, &addr, plen, flag,
2052                                                     IFA_F_PERMANENT);
2053                                 if (!IS_ERR(ifp)) {
2054                                         spin_lock_bh(&ifp->lock);
2055                                         ifp->flags &= ~IFA_F_TENTATIVE;
2056                                         spin_unlock_bh(&ifp->lock);
2057                                         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2058                                         in6_ifa_put(ifp);
2059                                 }
2060                         }
2061                 }
2062         }
2063 }
2064
2065 static void init_loopback(struct net_device *dev)
2066 {
2067         struct inet6_dev  *idev;
2068         struct inet6_ifaddr * ifp;
2069
2070         /* ::1 */
2071
2072         ASSERT_RTNL();
2073
2074         if ((idev = ipv6_find_idev(dev)) == NULL) {
2075                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2076                 return;
2077         }
2078
2079         ifp = ipv6_add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFA_F_PERMANENT);
2080         if (!IS_ERR(ifp)) {
2081                 spin_lock_bh(&ifp->lock);
2082                 ifp->flags &= ~IFA_F_TENTATIVE;
2083                 spin_unlock_bh(&ifp->lock);
2084                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2085                 in6_ifa_put(ifp);
2086         }
2087 }
2088
2089 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2090 {
2091         struct inet6_ifaddr * ifp;
2092
2093         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, IFA_F_PERMANENT);
2094         if (!IS_ERR(ifp)) {
2095                 addrconf_dad_start(ifp, 0);
2096                 in6_ifa_put(ifp);
2097         }
2098 }
2099
2100 static void addrconf_dev_config(struct net_device *dev)
2101 {
2102         struct in6_addr addr;
2103         struct inet6_dev    * idev;
2104
2105         ASSERT_RTNL();
2106
2107         if ((dev->type != ARPHRD_ETHER) &&
2108             (dev->type != ARPHRD_FDDI) &&
2109             (dev->type != ARPHRD_IEEE802_TR) &&
2110             (dev->type != ARPHRD_ARCNET) &&
2111             (dev->type != ARPHRD_INFINIBAND)) {
2112                 /* Alas, we support only Ethernet autoconfiguration. */
2113                 return;
2114         }
2115
2116         idev = addrconf_add_dev(dev);
2117         if (idev == NULL)
2118                 return;
2119
2120         memset(&addr, 0, sizeof(struct in6_addr));
2121         addr.s6_addr32[0] = htonl(0xFE800000);
2122
2123         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2124                 addrconf_add_linklocal(idev, &addr);
2125 }
2126
2127 static void addrconf_sit_config(struct net_device *dev)
2128 {
2129         struct inet6_dev *idev;
2130
2131         ASSERT_RTNL();
2132
2133         /*
2134          * Configure the tunnel with one of our IPv4
2135          * addresses... we should configure all of
2136          * our v4 addrs in the tunnel
2137          */
2138
2139         if ((idev = ipv6_find_idev(dev)) == NULL) {
2140                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2141                 return;
2142         }
2143
2144         sit_add_v4_addrs(idev);
2145
2146         if (dev->flags&IFF_POINTOPOINT) {
2147                 addrconf_add_mroute(dev);
2148                 addrconf_add_lroute(dev);
2149         } else
2150                 sit_route_add(dev);
2151 }
2152
2153 static inline int
2154 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2155 {
2156         struct in6_addr lladdr;
2157
2158         if (!ipv6_get_lladdr(link_dev, &lladdr)) {
2159                 addrconf_add_linklocal(idev, &lladdr);
2160                 return 0;
2161         }
2162         return -1;
2163 }
2164
2165 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2166 {
2167         struct net_device *link_dev;
2168
2169         /* first try to inherit the link-local address from the link device */
2170         if (idev->dev->iflink &&
2171             (link_dev = __dev_get_by_index(idev->dev->iflink))) {
2172                 if (!ipv6_inherit_linklocal(idev, link_dev))
2173                         return;
2174         }
2175         /* then try to inherit it from any device */
2176         for (link_dev = dev_base; link_dev; link_dev = link_dev->next) {
2177                 if (!ipv6_inherit_linklocal(idev, link_dev))
2178                         return;
2179         }
2180         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2181 }
2182
2183 /*
2184  * Autoconfigure tunnel with a link-local address so routing protocols,
2185  * DHCPv6, MLD etc. can be run over the virtual link
2186  */
2187
2188 static void addrconf_ip6_tnl_config(struct net_device *dev)
2189 {
2190         struct inet6_dev *idev;
2191
2192         ASSERT_RTNL();
2193
2194         if ((idev = addrconf_add_dev(dev)) == NULL) {
2195                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2196                 return;
2197         }
2198         ip6_tnl_add_linklocal(idev);
2199 }
2200
2201 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2202                            void * data)
2203 {
2204         struct net_device *dev = (struct net_device *) data;
2205         struct inet6_dev *idev = __in6_dev_get(dev);
2206         int run_pending = 0;
2207
2208         switch(event) {
2209         case NETDEV_UP:
2210         case NETDEV_CHANGE:
2211                 if (event == NETDEV_UP) {
2212                         if (!netif_carrier_ok(dev)) {
2213                                 /* device is not ready yet. */
2214                                 printk(KERN_INFO
2215                                         "ADDRCONF(NETDEV_UP): %s: "
2216                                         "link is not ready\n",
2217                                         dev->name);
2218                                 break;
2219                         }
2220
2221                         if (idev)
2222                                 idev->if_flags |= IF_READY;
2223                 } else {
2224                         if (!netif_carrier_ok(dev)) {
2225                                 /* device is still not ready. */
2226                                 break;
2227                         }
2228
2229                         if (idev) {
2230                                 if (idev->if_flags & IF_READY) {
2231                                         /* device is already configured. */
2232                                         break;
2233                                 }
2234                                 idev->if_flags |= IF_READY;
2235                         }
2236
2237                         printk(KERN_INFO
2238                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2239                                         "link becomes ready\n",
2240                                         dev->name);
2241
2242                         run_pending = 1;
2243                 }
2244
2245                 switch(dev->type) {
2246                 case ARPHRD_SIT:
2247                         addrconf_sit_config(dev);
2248                         break;
2249                 case ARPHRD_TUNNEL6:
2250                         addrconf_ip6_tnl_config(dev);
2251                         break;
2252                 case ARPHRD_LOOPBACK:
2253                         init_loopback(dev);
2254                         break;
2255
2256                 default:
2257                         addrconf_dev_config(dev);
2258                         break;
2259                 };
2260                 if (idev) {
2261                         if (run_pending)
2262                                 addrconf_dad_run(idev);
2263
2264                         /* If the MTU changed during the interface down, when the
2265                            interface up, the changed MTU must be reflected in the
2266                            idev as well as routers.
2267                          */
2268                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2269                                 rt6_mtu_change(dev, dev->mtu);
2270                                 idev->cnf.mtu6 = dev->mtu;
2271                         }
2272                         idev->tstamp = jiffies;
2273                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2274                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2275                            stop IPv6 on this interface.
2276                          */
2277                         if (dev->mtu < IPV6_MIN_MTU)
2278                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2279                 }
2280                 break;
2281
2282         case NETDEV_CHANGEMTU:
2283                 if ( idev && dev->mtu >= IPV6_MIN_MTU) {
2284                         rt6_mtu_change(dev, dev->mtu);
2285                         idev->cnf.mtu6 = dev->mtu;
2286                         break;
2287                 }
2288
2289                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2290
2291         case NETDEV_DOWN:
2292         case NETDEV_UNREGISTER:
2293                 /*
2294                  *      Remove all addresses from this interface.
2295                  */
2296                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2297                 break;
2298
2299         case NETDEV_CHANGENAME:
2300 #ifdef CONFIG_SYSCTL
2301                 if (idev) {
2302                         addrconf_sysctl_unregister(&idev->cnf);
2303                         neigh_sysctl_unregister(idev->nd_parms);
2304                         neigh_sysctl_register(dev, idev->nd_parms,
2305                                               NET_IPV6, NET_IPV6_NEIGH, "ipv6",
2306                                               &ndisc_ifinfo_sysctl_change,
2307                                               NULL);
2308                         addrconf_sysctl_register(idev, &idev->cnf);
2309                 }
2310 #endif
2311                 break;
2312         };
2313
2314         return NOTIFY_OK;
2315 }
2316
2317 /*
2318  *      addrconf module should be notified of a device going up
2319  */
2320 static struct notifier_block ipv6_dev_notf = {
2321         .notifier_call = addrconf_notify,
2322         .priority = 0
2323 };
2324
2325 static int addrconf_ifdown(struct net_device *dev, int how)
2326 {
2327         struct inet6_dev *idev;
2328         struct inet6_ifaddr *ifa, **bifa;
2329         int i;
2330
2331         ASSERT_RTNL();
2332
2333         if (dev == &loopback_dev && how == 1)
2334                 how = 0;
2335
2336         rt6_ifdown(dev);
2337         neigh_ifdown(&nd_tbl, dev);
2338
2339         idev = __in6_dev_get(dev);
2340         if (idev == NULL)
2341                 return -ENODEV;
2342
2343         /* Step 1: remove reference to ipv6 device from parent device.
2344                    Do not dev_put!
2345          */
2346         if (how == 1) {
2347                 idev->dead = 1;
2348
2349                 /* protected by rtnl_lock */
2350                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2351
2352                 /* Step 1.5: remove snmp6 entry */
2353                 snmp6_unregister_dev(idev);
2354
2355         }
2356
2357         /* Step 2: clear hash table */
2358         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2359                 bifa = &inet6_addr_lst[i];
2360
2361                 write_lock_bh(&addrconf_hash_lock);
2362                 while ((ifa = *bifa) != NULL) {
2363                         if (ifa->idev == idev) {
2364                                 *bifa = ifa->lst_next;
2365                                 ifa->lst_next = NULL;
2366                                 addrconf_del_timer(ifa);
2367                                 in6_ifa_put(ifa);
2368                                 continue;
2369                         }
2370                         bifa = &ifa->lst_next;
2371                 }
2372                 write_unlock_bh(&addrconf_hash_lock);
2373         }
2374
2375         write_lock_bh(&idev->lock);
2376
2377         /* Step 3: clear flags for stateless addrconf */
2378         if (how != 1)
2379                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2380
2381         /* Step 4: clear address list */
2382 #ifdef CONFIG_IPV6_PRIVACY
2383         if (how == 1 && del_timer(&idev->regen_timer))
2384                 in6_dev_put(idev);
2385
2386         /* clear tempaddr list */
2387         while ((ifa = idev->tempaddr_list) != NULL) {
2388                 idev->tempaddr_list = ifa->tmp_next;
2389                 ifa->tmp_next = NULL;
2390                 ifa->dead = 1;
2391                 write_unlock_bh(&idev->lock);
2392                 spin_lock_bh(&ifa->lock);
2393
2394                 if (ifa->ifpub) {
2395                         in6_ifa_put(ifa->ifpub);
2396                         ifa->ifpub = NULL;
2397                 }
2398                 spin_unlock_bh(&ifa->lock);
2399                 in6_ifa_put(ifa);
2400                 write_lock_bh(&idev->lock);
2401         }
2402 #endif
2403         while ((ifa = idev->addr_list) != NULL) {
2404                 idev->addr_list = ifa->if_next;
2405                 ifa->if_next = NULL;
2406                 ifa->dead = 1;
2407                 addrconf_del_timer(ifa);
2408                 write_unlock_bh(&idev->lock);
2409
2410                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2411                 in6_ifa_put(ifa);
2412
2413                 write_lock_bh(&idev->lock);
2414         }
2415         write_unlock_bh(&idev->lock);
2416
2417         /* Step 5: Discard multicast list */
2418
2419         if (how == 1)
2420                 ipv6_mc_destroy_dev(idev);
2421         else
2422                 ipv6_mc_down(idev);
2423
2424         /* Step 5: netlink notification of this interface */
2425         idev->tstamp = jiffies;
2426         inet6_ifinfo_notify(RTM_DELLINK, idev);
2427
2428         /* Shot the device (if unregistered) */
2429
2430         if (how == 1) {
2431 #ifdef CONFIG_SYSCTL
2432                 addrconf_sysctl_unregister(&idev->cnf);
2433                 neigh_sysctl_unregister(idev->nd_parms);
2434 #endif
2435                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2436                 neigh_ifdown(&nd_tbl, dev);
2437                 in6_dev_put(idev);
2438         }
2439         return 0;
2440 }
2441
2442 static void addrconf_rs_timer(unsigned long data)
2443 {
2444         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2445
2446         if (ifp->idev->cnf.forwarding)
2447                 goto out;
2448
2449         if (ifp->idev->if_flags & IF_RA_RCVD) {
2450                 /*
2451                  *      Announcement received after solicitation
2452                  *      was sent
2453                  */
2454                 goto out;
2455         }
2456
2457         spin_lock(&ifp->lock);
2458         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2459                 struct in6_addr all_routers;
2460
2461                 /* The wait after the last probe can be shorter */
2462                 addrconf_mod_timer(ifp, AC_RS,
2463                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2464                                    ifp->idev->cnf.rtr_solicit_delay :
2465                                    ifp->idev->cnf.rtr_solicit_interval);
2466                 spin_unlock(&ifp->lock);
2467
2468                 ipv6_addr_all_routers(&all_routers);
2469
2470                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2471         } else {
2472                 spin_unlock(&ifp->lock);
2473                 /*
2474                  * Note: we do not support deprecated "all on-link"
2475                  * assumption any longer.
2476                  */
2477                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2478                        ifp->idev->dev->name);
2479         }
2480
2481 out:
2482         in6_ifa_put(ifp);
2483 }
2484
2485 /*
2486  *      Duplicate Address Detection
2487  */
2488 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2489 {
2490         unsigned long rand_num;
2491         struct inet6_dev *idev = ifp->idev;
2492
2493         rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2494         ifp->probes = idev->cnf.dad_transmits;
2495         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2496 }
2497
2498 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2499 {
2500         struct inet6_dev *idev = ifp->idev;
2501         struct net_device *dev = idev->dev;
2502
2503         addrconf_join_solict(dev, &ifp->addr);
2504
2505         if (ifp->prefix_len != 128 && (ifp->flags&IFA_F_PERMANENT))
2506                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 0,
2507                                         flags);
2508
2509         net_srandom(ifp->addr.s6_addr32[3]);
2510
2511         read_lock_bh(&idev->lock);
2512         if (ifp->dead)
2513                 goto out;
2514         spin_lock_bh(&ifp->lock);
2515
2516         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2517             !(ifp->flags&IFA_F_TENTATIVE)) {
2518                 ifp->flags &= ~IFA_F_TENTATIVE;
2519                 spin_unlock_bh(&ifp->lock);
2520                 read_unlock_bh(&idev->lock);
2521
2522                 addrconf_dad_completed(ifp);
2523                 return;
2524         }
2525
2526         if (!(idev->if_flags & IF_READY)) {
2527                 spin_unlock_bh(&ifp->lock);
2528                 read_unlock_bh(&idev->lock);
2529                 /*
2530                  * If the defice is not ready:
2531                  * - keep it tentative if it is a permanent address.
2532                  * - otherwise, kill it.
2533                  */
2534                 in6_ifa_hold(ifp);
2535                 addrconf_dad_stop(ifp);
2536                 return;
2537         }
2538         addrconf_dad_kick(ifp);
2539         spin_unlock_bh(&ifp->lock);
2540 out:
2541         read_unlock_bh(&idev->lock);
2542 }
2543
2544 static void addrconf_dad_timer(unsigned long data)
2545 {
2546         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2547         struct inet6_dev *idev = ifp->idev;
2548         struct in6_addr unspec;
2549         struct in6_addr mcaddr;
2550
2551         read_lock_bh(&idev->lock);
2552         if (idev->dead) {
2553                 read_unlock_bh(&idev->lock);
2554                 goto out;
2555         }
2556         spin_lock_bh(&ifp->lock);
2557         if (ifp->probes == 0) {
2558                 /*
2559                  * DAD was successful
2560                  */
2561
2562                 ifp->flags &= ~IFA_F_TENTATIVE;
2563                 spin_unlock_bh(&ifp->lock);
2564                 read_unlock_bh(&idev->lock);
2565
2566                 addrconf_dad_completed(ifp);
2567
2568                 goto out;
2569         }
2570
2571         ifp->probes--;
2572         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2573         spin_unlock_bh(&ifp->lock);
2574         read_unlock_bh(&idev->lock);
2575
2576         /* send a neighbour solicitation for our addr */
2577         memset(&unspec, 0, sizeof(unspec));
2578         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2579         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &unspec);
2580 out:
2581         in6_ifa_put(ifp);
2582 }
2583
2584 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2585 {
2586         struct net_device *     dev = ifp->idev->dev;
2587
2588         /*
2589          *      Configure the address for reception. Now it is valid.
2590          */
2591
2592         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2593
2594         /* If added prefix is link local and forwarding is off,
2595            start sending router solicitations.
2596          */
2597
2598         if (ifp->idev->cnf.forwarding == 0 &&
2599             ifp->idev->cnf.rtr_solicits > 0 &&
2600             (dev->flags&IFF_LOOPBACK) == 0 &&
2601             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2602                 struct in6_addr all_routers;
2603
2604                 ipv6_addr_all_routers(&all_routers);
2605
2606                 /*
2607                  *      If a host as already performed a random delay
2608                  *      [...] as part of DAD [...] there is no need
2609                  *      to delay again before sending the first RS
2610                  */
2611                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &all_routers);
2612
2613                 spin_lock_bh(&ifp->lock);
2614                 ifp->probes = 1;
2615                 ifp->idev->if_flags |= IF_RS_SENT;
2616                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2617                 spin_unlock_bh(&ifp->lock);
2618         }
2619 }
2620
2621 static void addrconf_dad_run(struct inet6_dev *idev) {
2622         struct inet6_ifaddr *ifp;
2623
2624         read_lock_bh(&idev->lock);
2625         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2626                 spin_lock_bh(&ifp->lock);
2627                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2628                         spin_unlock_bh(&ifp->lock);
2629                         continue;
2630                 }
2631                 spin_unlock_bh(&ifp->lock);
2632                 addrconf_dad_kick(ifp);
2633         }
2634         read_unlock_bh(&idev->lock);
2635 }
2636
2637 #ifdef CONFIG_PROC_FS
2638 struct if6_iter_state {
2639         int bucket;
2640 };
2641
2642 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2643 {
2644         struct inet6_ifaddr *ifa = NULL;
2645         struct if6_iter_state *state = seq->private;
2646
2647         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2648                 ifa = inet6_addr_lst[state->bucket];
2649                 if (ifa)
2650                         break;
2651         }
2652         return ifa;
2653 }
2654
2655 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2656 {
2657         struct if6_iter_state *state = seq->private;
2658
2659         ifa = ifa->lst_next;
2660 try_again:
2661         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2662                 ifa = inet6_addr_lst[state->bucket];
2663                 goto try_again;
2664         }
2665         return ifa;
2666 }
2667
2668 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2669 {
2670         struct inet6_ifaddr *ifa = if6_get_first(seq);
2671
2672         if (ifa)
2673                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2674                         --pos;
2675         return pos ? NULL : ifa;
2676 }
2677
2678 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2679 {
2680         read_lock_bh(&addrconf_hash_lock);
2681         return if6_get_idx(seq, *pos);
2682 }
2683
2684 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2685 {
2686         struct inet6_ifaddr *ifa;
2687
2688         ifa = if6_get_next(seq, v);
2689         ++*pos;
2690         return ifa;
2691 }
2692
2693 static void if6_seq_stop(struct seq_file *seq, void *v)
2694 {
2695         read_unlock_bh(&addrconf_hash_lock);
2696 }
2697
2698 static int if6_seq_show(struct seq_file *seq, void *v)
2699 {
2700         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2701         seq_printf(seq,
2702                    NIP6_SEQFMT " %02x %02x %02x %02x %8s\n",
2703                    NIP6(ifp->addr),
2704                    ifp->idev->dev->ifindex,
2705                    ifp->prefix_len,
2706                    ifp->scope,
2707                    ifp->flags,
2708                    ifp->idev->dev->name);
2709         return 0;
2710 }
2711
2712 static struct seq_operations if6_seq_ops = {
2713         .start  = if6_seq_start,
2714         .next   = if6_seq_next,
2715         .show   = if6_seq_show,
2716         .stop   = if6_seq_stop,
2717 };
2718
2719 static int if6_seq_open(struct inode *inode, struct file *file)
2720 {
2721         struct seq_file *seq;
2722         int rc = -ENOMEM;
2723         struct if6_iter_state *s = kzalloc(sizeof(*s), GFP_KERNEL);
2724
2725         if (!s)
2726                 goto out;
2727
2728         rc = seq_open(file, &if6_seq_ops);
2729         if (rc)
2730                 goto out_kfree;
2731
2732         seq = file->private_data;
2733         seq->private = s;
2734 out:
2735         return rc;
2736 out_kfree:
2737         kfree(s);
2738         goto out;
2739 }
2740
2741 static struct file_operations if6_fops = {
2742         .owner          = THIS_MODULE,
2743         .open           = if6_seq_open,
2744         .read           = seq_read,
2745         .llseek         = seq_lseek,
2746         .release        = seq_release_private,
2747 };
2748
2749 int __init if6_proc_init(void)
2750 {
2751         if (!proc_net_fops_create("if_inet6", S_IRUGO, &if6_fops))
2752                 return -ENOMEM;
2753         return 0;
2754 }
2755
2756 void if6_proc_exit(void)
2757 {
2758         proc_net_remove("if_inet6");
2759 }
2760 #endif  /* CONFIG_PROC_FS */
2761
2762 /*
2763  *      Periodic address status verification
2764  */
2765
2766 static void addrconf_verify(unsigned long foo)
2767 {
2768         struct inet6_ifaddr *ifp;
2769         unsigned long now, next;
2770         int i;
2771
2772         spin_lock_bh(&addrconf_verify_lock);
2773         now = jiffies;
2774         next = now + ADDR_CHECK_FREQUENCY;
2775
2776         del_timer(&addr_chk_timer);
2777
2778         for (i=0; i < IN6_ADDR_HSIZE; i++) {
2779
2780 restart:
2781                 read_lock(&addrconf_hash_lock);
2782                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
2783                         unsigned long age;
2784 #ifdef CONFIG_IPV6_PRIVACY
2785                         unsigned long regen_advance;
2786 #endif
2787
2788                         if (ifp->flags & IFA_F_PERMANENT)
2789                                 continue;
2790
2791                         spin_lock(&ifp->lock);
2792                         age = (now - ifp->tstamp) / HZ;
2793
2794 #ifdef CONFIG_IPV6_PRIVACY
2795                         regen_advance = ifp->idev->cnf.regen_max_retry *
2796                                         ifp->idev->cnf.dad_transmits *
2797                                         ifp->idev->nd_parms->retrans_time / HZ;
2798 #endif
2799
2800                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
2801                             age >= ifp->valid_lft) {
2802                                 spin_unlock(&ifp->lock);
2803                                 in6_ifa_hold(ifp);
2804                                 read_unlock(&addrconf_hash_lock);
2805                                 ipv6_del_addr(ifp);
2806                                 goto restart;
2807                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
2808                                 spin_unlock(&ifp->lock);
2809                                 continue;
2810                         } else if (age >= ifp->prefered_lft) {
2811                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
2812                                 int deprecate = 0;
2813
2814                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
2815                                         deprecate = 1;
2816                                         ifp->flags |= IFA_F_DEPRECATED;
2817                                 }
2818
2819                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
2820                                         next = ifp->tstamp + ifp->valid_lft * HZ;
2821
2822                                 spin_unlock(&ifp->lock);
2823
2824                                 if (deprecate) {
2825                                         in6_ifa_hold(ifp);
2826                                         read_unlock(&addrconf_hash_lock);
2827
2828                                         ipv6_ifa_notify(0, ifp);
2829                                         in6_ifa_put(ifp);
2830                                         goto restart;
2831                                 }
2832 #ifdef CONFIG_IPV6_PRIVACY
2833                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
2834                                    !(ifp->flags&IFA_F_TENTATIVE)) {
2835                                 if (age >= ifp->prefered_lft - regen_advance) {
2836                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
2837                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2838                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
2839                                         if (!ifp->regen_count && ifpub) {
2840                                                 ifp->regen_count++;
2841                                                 in6_ifa_hold(ifp);
2842                                                 in6_ifa_hold(ifpub);
2843                                                 spin_unlock(&ifp->lock);
2844                                                 read_unlock(&addrconf_hash_lock);
2845                                                 spin_lock(&ifpub->lock);
2846                                                 ifpub->regen_count = 0;
2847                                                 spin_unlock(&ifpub->lock);
2848                                                 ipv6_create_tempaddr(ifpub, ifp);
2849                                                 in6_ifa_put(ifpub);
2850                                                 in6_ifa_put(ifp);
2851                                                 goto restart;
2852                                         }
2853                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
2854                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
2855                                 spin_unlock(&ifp->lock);
2856 #endif
2857                         } else {
2858                                 /* ifp->prefered_lft <= ifp->valid_lft */
2859                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
2860                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
2861                                 spin_unlock(&ifp->lock);
2862                         }
2863                 }
2864                 read_unlock(&addrconf_hash_lock);
2865         }
2866
2867         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
2868         add_timer(&addr_chk_timer);
2869         spin_unlock_bh(&addrconf_verify_lock);
2870 }
2871
2872 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
2873 {
2874         struct in6_addr *pfx = NULL;
2875
2876         if (addr)
2877                 pfx = nla_data(addr);
2878
2879         if (local) {
2880                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
2881                         pfx = NULL;
2882                 else
2883                         pfx = nla_data(local);
2884         }
2885
2886         return pfx;
2887 }
2888
2889 static struct nla_policy ifa_ipv6_policy[IFA_MAX+1] __read_mostly = {
2890         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
2891         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
2892         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
2893 };
2894
2895 static int
2896 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2897 {
2898         struct ifaddrmsg *ifm;
2899         struct nlattr *tb[IFA_MAX+1];
2900         struct in6_addr *pfx;
2901         int err;
2902
2903         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2904         if (err < 0)
2905                 return err;
2906
2907         ifm = nlmsg_data(nlh);
2908         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2909         if (pfx == NULL)
2910                 return -EINVAL;
2911
2912         return inet6_addr_del(ifm->ifa_index, pfx, ifm->ifa_prefixlen);
2913 }
2914
2915 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 prefered_lft,
2916                              u32 valid_lft)
2917 {
2918         int ifa_flags = 0;
2919
2920         if (!valid_lft || (prefered_lft > valid_lft))
2921                 return -EINVAL;
2922
2923         if (valid_lft == INFINITY_LIFE_TIME)
2924                 ifa_flags = IFA_F_PERMANENT;
2925         else if (valid_lft >= 0x7FFFFFFF/HZ)
2926                 valid_lft = 0x7FFFFFFF/HZ;
2927
2928         if (prefered_lft == 0)
2929                 ifa_flags = IFA_F_DEPRECATED;
2930         else if ((prefered_lft >= 0x7FFFFFFF/HZ) &&
2931                  (prefered_lft != INFINITY_LIFE_TIME))
2932                 prefered_lft = 0x7FFFFFFF/HZ;
2933
2934         spin_lock_bh(&ifp->lock);
2935         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED|IFA_F_PERMANENT)) | ifa_flags;
2936
2937         ifp->tstamp = jiffies;
2938         ifp->valid_lft = valid_lft;
2939         ifp->prefered_lft = prefered_lft;
2940
2941         spin_unlock_bh(&ifp->lock);
2942         if (!(ifp->flags&IFA_F_TENTATIVE))
2943                 ipv6_ifa_notify(0, ifp);
2944
2945         addrconf_verify(0);
2946
2947         return 0;
2948 }
2949
2950 static int
2951 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
2952 {
2953         struct ifaddrmsg *ifm;
2954         struct nlattr *tb[IFA_MAX+1];
2955         struct in6_addr *pfx;
2956         struct inet6_ifaddr *ifa;
2957         struct net_device *dev;
2958         u32 valid_lft, preferred_lft;
2959         int err;
2960
2961         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
2962         if (err < 0)
2963                 return err;
2964
2965         ifm = nlmsg_data(nlh);
2966         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
2967         if (pfx == NULL)
2968                 return -EINVAL;
2969
2970         if (tb[IFA_CACHEINFO]) {
2971                 struct ifa_cacheinfo *ci;
2972
2973                 ci = nla_data(tb[IFA_CACHEINFO]);
2974                 valid_lft = ci->ifa_valid;
2975                 preferred_lft = ci->ifa_prefered;
2976         } else {
2977                 preferred_lft = INFINITY_LIFE_TIME;
2978                 valid_lft = INFINITY_LIFE_TIME;
2979         }
2980
2981         dev =  __dev_get_by_index(ifm->ifa_index);
2982         if (dev == NULL)
2983                 return -ENODEV;
2984
2985         ifa = ipv6_get_ifaddr(pfx, dev, 1);
2986         if (ifa == NULL) {
2987                 /*
2988                  * It would be best to check for !NLM_F_CREATE here but
2989                  * userspace alreay relies on not having to provide this.
2990                  */
2991                 return inet6_addr_add(ifm->ifa_index, pfx, ifm->ifa_prefixlen,
2992                                       preferred_lft, valid_lft);
2993         }
2994
2995         if (nlh->nlmsg_flags & NLM_F_EXCL ||
2996             !(nlh->nlmsg_flags & NLM_F_REPLACE))
2997                 err = -EEXIST;
2998         else
2999                 err = inet6_addr_modify(ifa, preferred_lft, valid_lft);
3000
3001         in6_ifa_put(ifa);
3002
3003         return err;
3004 }
3005
3006 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3007                           u8 scope, int ifindex)
3008 {
3009         struct ifaddrmsg *ifm;
3010
3011         ifm = nlmsg_data(nlh);
3012         ifm->ifa_family = AF_INET6;
3013         ifm->ifa_prefixlen = prefixlen;
3014         ifm->ifa_flags = flags;
3015         ifm->ifa_scope = scope;
3016         ifm->ifa_index = ifindex;
3017 }
3018
3019 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3020                          unsigned long tstamp, u32 preferred, u32 valid)
3021 {
3022         struct ifa_cacheinfo ci;
3023
3024         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3025                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3026         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3027                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3028         ci.ifa_prefered = preferred;
3029         ci.ifa_valid = valid;
3030
3031         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3032 }
3033
3034 static inline int rt_scope(int ifa_scope)
3035 {
3036         if (ifa_scope & IFA_HOST)
3037                 return RT_SCOPE_HOST;
3038         else if (ifa_scope & IFA_LINK)
3039                 return RT_SCOPE_LINK;
3040         else if (ifa_scope & IFA_SITE)
3041                 return RT_SCOPE_SITE;
3042         else
3043                 return RT_SCOPE_UNIVERSE;
3044 }
3045
3046 static inline int inet6_ifaddr_msgsize(void)
3047 {
3048         return nlmsg_total_size(sizeof(struct ifaddrmsg) +
3049                                 nla_total_size(16) +
3050                                 nla_total_size(sizeof(struct ifa_cacheinfo)) +
3051                                 128);
3052 }
3053
3054 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3055                              u32 pid, u32 seq, int event, unsigned int flags)
3056 {
3057         struct nlmsghdr  *nlh;
3058         u32 preferred, valid;
3059
3060         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3061         if (nlh == NULL)
3062                 return -ENOBUFS;
3063
3064         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3065                       ifa->idev->dev->ifindex);
3066
3067         if (!(ifa->flags&IFA_F_PERMANENT)) {
3068                 preferred = ifa->prefered_lft;
3069                 valid = ifa->valid_lft;
3070                 if (preferred != INFINITY_LIFE_TIME) {
3071                         long tval = (jiffies - ifa->tstamp)/HZ;
3072                         preferred -= tval;
3073                         if (valid != INFINITY_LIFE_TIME)
3074                                 valid -= tval;
3075                 }
3076         } else {
3077                 preferred = INFINITY_LIFE_TIME;
3078                 valid = INFINITY_LIFE_TIME;
3079         }
3080
3081         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3082             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0)
3083                 return nlmsg_cancel(skb, nlh);
3084
3085         return nlmsg_end(skb, nlh);
3086 }
3087
3088 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3089                                 u32 pid, u32 seq, int event, u16 flags)
3090 {
3091         struct nlmsghdr  *nlh;
3092         u8 scope = RT_SCOPE_UNIVERSE;
3093         int ifindex = ifmca->idev->dev->ifindex;
3094
3095         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3096                 scope = RT_SCOPE_SITE;
3097
3098         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3099         if (nlh == NULL)
3100                 return -ENOBUFS;
3101
3102         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3103         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3104             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3105                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3106                 return nlmsg_cancel(skb, nlh);
3107
3108         return nlmsg_end(skb, nlh);
3109 }
3110
3111 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3112                                 u32 pid, u32 seq, int event, unsigned int flags)
3113 {
3114         struct nlmsghdr  *nlh;
3115         u8 scope = RT_SCOPE_UNIVERSE;
3116         int ifindex = ifaca->aca_idev->dev->ifindex;
3117
3118         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3119                 scope = RT_SCOPE_SITE;
3120
3121         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3122         if (nlh == NULL)
3123                 return -ENOBUFS;
3124
3125         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3126         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3127             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3128                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0)
3129                 return nlmsg_cancel(skb, nlh);
3130
3131         return nlmsg_end(skb, nlh);
3132 }
3133
3134 enum addr_type_t
3135 {
3136         UNICAST_ADDR,
3137         MULTICAST_ADDR,
3138         ANYCAST_ADDR,
3139 };
3140
3141 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3142                            enum addr_type_t type)
3143 {
3144         int idx, ip_idx;
3145         int s_idx, s_ip_idx;
3146         int err = 1;
3147         struct net_device *dev;
3148         struct inet6_dev *idev = NULL;
3149         struct inet6_ifaddr *ifa;
3150         struct ifmcaddr6 *ifmca;
3151         struct ifacaddr6 *ifaca;
3152
3153         s_idx = cb->args[0];
3154         s_ip_idx = ip_idx = cb->args[1];
3155         read_lock(&dev_base_lock);
3156
3157         for (dev = dev_base, idx = 0; dev; dev = dev->next, idx++) {
3158                 if (idx < s_idx)
3159                         continue;
3160                 if (idx > s_idx)
3161                         s_ip_idx = 0;
3162                 ip_idx = 0;
3163                 if ((idev = in6_dev_get(dev)) == NULL)
3164                         continue;
3165                 read_lock_bh(&idev->lock);
3166                 switch (type) {
3167                 case UNICAST_ADDR:
3168                         /* unicast address incl. temp addr */
3169                         for (ifa = idev->addr_list; ifa;
3170                              ifa = ifa->if_next, ip_idx++) {
3171                                 if (ip_idx < s_ip_idx)
3172                                         continue;
3173                                 if ((err = inet6_fill_ifaddr(skb, ifa,
3174                                     NETLINK_CB(cb->skb).pid,
3175                                     cb->nlh->nlmsg_seq, RTM_NEWADDR,
3176                                     NLM_F_MULTI)) <= 0)
3177                                         goto done;
3178                         }
3179                         break;
3180                 case MULTICAST_ADDR:
3181                         /* multicast address */
3182                         for (ifmca = idev->mc_list; ifmca;
3183                              ifmca = ifmca->next, ip_idx++) {
3184                                 if (ip_idx < s_ip_idx)
3185                                         continue;
3186                                 if ((err = inet6_fill_ifmcaddr(skb, ifmca,
3187                                     NETLINK_CB(cb->skb).pid,
3188                                     cb->nlh->nlmsg_seq, RTM_GETMULTICAST,
3189                                     NLM_F_MULTI)) <= 0)
3190                                         goto done;
3191                         }
3192                         break;
3193                 case ANYCAST_ADDR:
3194                         /* anycast address */
3195                         for (ifaca = idev->ac_list; ifaca;
3196                              ifaca = ifaca->aca_next, ip_idx++) {
3197                                 if (ip_idx < s_ip_idx)
3198                                         continue;
3199                                 if ((err = inet6_fill_ifacaddr(skb, ifaca,
3200                                     NETLINK_CB(cb->skb).pid,
3201                                     cb->nlh->nlmsg_seq, RTM_GETANYCAST,
3202                                     NLM_F_MULTI)) <= 0)
3203                                         goto done;
3204                         }
3205                         break;
3206                 default:
3207                         break;
3208                 }
3209                 read_unlock_bh(&idev->lock);
3210                 in6_dev_put(idev);
3211         }
3212 done:
3213         if (err <= 0) {
3214                 read_unlock_bh(&idev->lock);
3215                 in6_dev_put(idev);
3216         }
3217         read_unlock(&dev_base_lock);
3218         cb->args[0] = idx;
3219         cb->args[1] = ip_idx;
3220         return skb->len;
3221 }
3222
3223 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3224 {
3225         enum addr_type_t type = UNICAST_ADDR;
3226         return inet6_dump_addr(skb, cb, type);
3227 }
3228
3229 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3230 {
3231         enum addr_type_t type = MULTICAST_ADDR;
3232         return inet6_dump_addr(skb, cb, type);
3233 }
3234
3235
3236 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3237 {
3238         enum addr_type_t type = ANYCAST_ADDR;
3239         return inet6_dump_addr(skb, cb, type);
3240 }
3241
3242 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3243                              void *arg)
3244 {
3245         struct ifaddrmsg *ifm;
3246         struct nlattr *tb[IFA_MAX+1];
3247         struct in6_addr *addr = NULL;
3248         struct net_device *dev = NULL;
3249         struct inet6_ifaddr *ifa;
3250         struct sk_buff *skb;
3251         int err;
3252
3253         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3254         if (err < 0)
3255                 goto errout;
3256
3257         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3258         if (addr == NULL) {
3259                 err = -EINVAL;
3260                 goto errout;
3261         }
3262
3263         ifm = nlmsg_data(nlh);
3264         if (ifm->ifa_index)
3265                 dev = __dev_get_by_index(ifm->ifa_index);
3266
3267         if ((ifa = ipv6_get_ifaddr(addr, dev, 1)) == NULL) {
3268                 err = -EADDRNOTAVAIL;
3269                 goto errout;
3270         }
3271
3272         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3273                 err = -ENOBUFS;
3274                 goto errout_ifa;
3275         }
3276
3277         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3278                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3279         if (err < 0) {
3280                 kfree_skb(skb);
3281                 goto errout_ifa;
3282         }
3283
3284         err = rtnl_unicast(skb, NETLINK_CB(in_skb).pid);
3285 errout_ifa:
3286         in6_ifa_put(ifa);
3287 errout:
3288         return err;
3289 }
3290
3291 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3292 {
3293         struct sk_buff *skb;
3294         int err = -ENOBUFS;
3295
3296         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3297         if (skb == NULL)
3298                 goto errout;
3299
3300         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3301         if (err < 0) {
3302                 kfree_skb(skb);
3303                 goto errout;
3304         }
3305
3306         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3307 errout:
3308         if (err < 0)
3309                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3310 }
3311
3312 static void inline ipv6_store_devconf(struct ipv6_devconf *cnf,
3313                                 __s32 *array, int bytes)
3314 {
3315         memset(array, 0, bytes);
3316         array[DEVCONF_FORWARDING] = cnf->forwarding;
3317         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3318         array[DEVCONF_MTU6] = cnf->mtu6;
3319         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3320         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3321         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3322         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3323         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3324         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3325         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3326         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3327 #ifdef CONFIG_IPV6_PRIVACY
3328         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3329         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3330         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3331         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3332         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3333 #endif
3334         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3335         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3336         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3337 #ifdef CONFIG_IPV6_ROUTER_PREF
3338         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3339         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3340 #ifdef CONFIV_IPV6_ROUTE_INFO
3341         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3342 #endif
3343 #endif
3344         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3345 }
3346
3347 /* Maximum length of ifinfomsg attributes */
3348 #define INET6_IFINFO_RTA_SPACE \
3349                 RTA_SPACE(IFNAMSIZ) /* IFNAME */ + \
3350                 RTA_SPACE(MAX_ADDR_LEN) /* ADDRESS */ + \
3351                 RTA_SPACE(sizeof(u32)) /* MTU */ + \
3352                 RTA_SPACE(sizeof(int)) /* LINK */ + \
3353                 RTA_SPACE(0) /* PROTINFO */ + \
3354                 RTA_SPACE(sizeof(u32)) /* FLAGS */ + \
3355                 RTA_SPACE(sizeof(struct ifla_cacheinfo)) /* CACHEINFO */ + \
3356                 RTA_SPACE(sizeof(__s32[DEVCONF_MAX])) /* CONF */
3357
3358 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3359                              u32 pid, u32 seq, int event, unsigned int flags)
3360 {
3361         struct net_device       *dev = idev->dev;
3362         __s32                   *array = NULL;
3363         struct ifinfomsg        *r;
3364         struct nlmsghdr         *nlh;
3365         unsigned char           *b = skb->tail;
3366         struct rtattr           *subattr;
3367         __u32                   mtu = dev->mtu;
3368         struct ifla_cacheinfo   ci;
3369
3370         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*r), flags);
3371         r = NLMSG_DATA(nlh);
3372         r->ifi_family = AF_INET6;
3373         r->__ifi_pad = 0;
3374         r->ifi_type = dev->type;
3375         r->ifi_index = dev->ifindex;
3376         r->ifi_flags = dev_get_flags(dev);
3377         r->ifi_change = 0;
3378
3379         RTA_PUT(skb, IFLA_IFNAME, strlen(dev->name)+1, dev->name);
3380
3381         if (dev->addr_len)
3382                 RTA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3383
3384         RTA_PUT(skb, IFLA_MTU, sizeof(mtu), &mtu);
3385         if (dev->ifindex != dev->iflink)
3386                 RTA_PUT(skb, IFLA_LINK, sizeof(int), &dev->iflink);
3387
3388         subattr = (struct rtattr*)skb->tail;
3389
3390         RTA_PUT(skb, IFLA_PROTINFO, 0, NULL);
3391
3392         /* return the device flags */
3393         RTA_PUT(skb, IFLA_INET6_FLAGS, sizeof(__u32), &idev->if_flags);
3394
3395         /* return interface cacheinfo */
3396         ci.max_reasm_len = IPV6_MAXPLEN;
3397         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3398                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3399         ci.reachable_time = idev->nd_parms->reachable_time;
3400         ci.retrans_time = idev->nd_parms->retrans_time;
3401         RTA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3402
3403         /* return the device sysctl params */
3404         if ((array = kmalloc(DEVCONF_MAX * sizeof(*array), GFP_ATOMIC)) == NULL)
3405                 goto rtattr_failure;
3406         ipv6_store_devconf(&idev->cnf, array, DEVCONF_MAX * sizeof(*array));
3407         RTA_PUT(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(*array), array);
3408
3409         /* XXX - Statistics/MC not implemented */
3410         subattr->rta_len = skb->tail - (u8*)subattr;
3411
3412         nlh->nlmsg_len = skb->tail - b;
3413         kfree(array);
3414         return skb->len;
3415
3416 nlmsg_failure:
3417 rtattr_failure:
3418         kfree(array);
3419         skb_trim(skb, b - skb->data);
3420         return -1;
3421 }
3422
3423 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3424 {
3425         int idx, err;
3426         int s_idx = cb->args[0];
3427         struct net_device *dev;
3428         struct inet6_dev *idev;
3429
3430         read_lock(&dev_base_lock);
3431         for (dev=dev_base, idx=0; dev; dev = dev->next, idx++) {
3432                 if (idx < s_idx)
3433                         continue;
3434                 if ((idev = in6_dev_get(dev)) == NULL)
3435                         continue;
3436                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3437                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3438                 in6_dev_put(idev);
3439                 if (err <= 0)
3440                         break;
3441         }
3442         read_unlock(&dev_base_lock);
3443         cb->args[0] = idx;
3444
3445         return skb->len;
3446 }
3447
3448 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3449 {
3450         struct sk_buff *skb;
3451         int payload = sizeof(struct ifinfomsg) + INET6_IFINFO_RTA_SPACE;
3452         int err = -ENOBUFS;
3453
3454         skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
3455         if (skb == NULL)
3456                 goto errout;
3457
3458         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3459         if (err < 0) {
3460                 kfree_skb(skb);
3461                 goto errout;
3462         }
3463
3464         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3465 errout:
3466         if (err < 0)
3467                 rtnl_set_sk_err(RTNLGRP_IPV6_IFADDR, err);
3468 }
3469
3470 /* Maximum length of prefix_cacheinfo attributes */
3471 #define INET6_PREFIX_RTA_SPACE \
3472                 RTA_SPACE(sizeof(((struct prefix_info *)NULL)->prefix)) /* ADDRESS */ + \
3473                 RTA_SPACE(sizeof(struct prefix_cacheinfo)) /* CACHEINFO */
3474
3475 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3476                         struct prefix_info *pinfo, u32 pid, u32 seq,
3477                         int event, unsigned int flags)
3478 {
3479         struct prefixmsg        *pmsg;
3480         struct nlmsghdr         *nlh;
3481         unsigned char           *b = skb->tail;
3482         struct prefix_cacheinfo ci;
3483
3484         nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*pmsg), flags);
3485         pmsg = NLMSG_DATA(nlh);
3486         pmsg->prefix_family = AF_INET6;
3487         pmsg->prefix_pad1 = 0;
3488         pmsg->prefix_pad2 = 0;
3489         pmsg->prefix_ifindex = idev->dev->ifindex;
3490         pmsg->prefix_len = pinfo->prefix_len;
3491         pmsg->prefix_type = pinfo->type;
3492         pmsg->prefix_pad3 = 0;
3493
3494         pmsg->prefix_flags = 0;
3495         if (pinfo->onlink)
3496                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3497         if (pinfo->autoconf)
3498                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3499
3500         RTA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3501
3502         ci.preferred_time = ntohl(pinfo->prefered);
3503         ci.valid_time = ntohl(pinfo->valid);
3504         RTA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3505
3506         nlh->nlmsg_len = skb->tail - b;
3507         return skb->len;
3508
3509 nlmsg_failure:
3510 rtattr_failure:
3511         skb_trim(skb, b - skb->data);
3512         return -1;
3513 }
3514
3515 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3516                          struct prefix_info *pinfo)
3517 {
3518         struct sk_buff *skb;
3519         int payload = sizeof(struct prefixmsg) + INET6_PREFIX_RTA_SPACE;
3520         int err = -ENOBUFS;
3521
3522         skb = nlmsg_new(nlmsg_total_size(payload), GFP_ATOMIC);
3523         if (skb == NULL)
3524                 goto errout;
3525
3526         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3527         if (err < 0) {
3528                 kfree_skb(skb);
3529                 goto errout;
3530         }
3531
3532         err = rtnl_notify(skb, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3533 errout:
3534         if (err < 0)
3535                 rtnl_set_sk_err(RTNLGRP_IPV6_PREFIX, err);
3536 }
3537
3538 static struct rtnetlink_link inet6_rtnetlink_table[RTM_NR_MSGTYPES] = {
3539         [RTM_GETLINK - RTM_BASE] = { .dumpit    = inet6_dump_ifinfo, },
3540         [RTM_NEWADDR - RTM_BASE] = { .doit      = inet6_rtm_newaddr, },
3541         [RTM_DELADDR - RTM_BASE] = { .doit      = inet6_rtm_deladdr, },
3542         [RTM_GETADDR - RTM_BASE] = { .doit      = inet6_rtm_getaddr,
3543                                      .dumpit    = inet6_dump_ifaddr, },
3544         [RTM_GETMULTICAST - RTM_BASE] = { .dumpit = inet6_dump_ifmcaddr, },
3545         [RTM_GETANYCAST - RTM_BASE] = { .dumpit = inet6_dump_ifacaddr, },
3546         [RTM_NEWROUTE - RTM_BASE] = { .doit     = inet6_rtm_newroute, },
3547         [RTM_DELROUTE - RTM_BASE] = { .doit     = inet6_rtm_delroute, },
3548         [RTM_GETROUTE - RTM_BASE] = { .doit     = inet6_rtm_getroute,
3549                                       .dumpit   = inet6_dump_fib, },
3550 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3551         [RTM_GETRULE  - RTM_BASE] = { .dumpit   = fib6_rules_dump,   },
3552 #endif
3553 };
3554
3555 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3556 {
3557         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3558
3559         switch (event) {
3560         case RTM_NEWADDR:
3561                 ip6_ins_rt(ifp->rt);
3562                 if (ifp->idev->cnf.forwarding)
3563                         addrconf_join_anycast(ifp);
3564                 break;
3565         case RTM_DELADDR:
3566                 if (ifp->idev->cnf.forwarding)
3567                         addrconf_leave_anycast(ifp);
3568                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3569                 dst_hold(&ifp->rt->u.dst);
3570                 if (ip6_del_rt(ifp->rt))
3571                         dst_free(&ifp->rt->u.dst);
3572                 break;
3573         }
3574 }
3575
3576 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3577 {
3578         rcu_read_lock_bh();
3579         if (likely(ifp->idev->dead == 0))
3580                 __ipv6_ifa_notify(event, ifp);
3581         rcu_read_unlock_bh();
3582 }
3583
3584 #ifdef CONFIG_SYSCTL
3585
3586 static
3587 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3588                            void __user *buffer, size_t *lenp, loff_t *ppos)
3589 {
3590         int *valp = ctl->data;
3591         int val = *valp;
3592         int ret;
3593
3594         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3595
3596         if (write && valp != &ipv6_devconf_dflt.forwarding) {
3597                 if (valp != &ipv6_devconf.forwarding) {
3598                         if ((!*valp) ^ (!val)) {
3599                                 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1;
3600                                 if (idev == NULL)
3601                                         return ret;
3602                                 dev_forward_change(idev);
3603                         }
3604                 } else {
3605                         ipv6_devconf_dflt.forwarding = ipv6_devconf.forwarding;
3606                         addrconf_forward_change();
3607                 }
3608                 if (*valp)
3609                         rt6_purge_dflt_routers();
3610         }
3611
3612         return ret;
3613 }
3614
3615 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3616                                             int __user *name, int nlen,
3617                                             void __user *oldval,
3618                                             size_t __user *oldlenp,
3619                                             void __user *newval, size_t newlen,
3620                                             void **context)
3621 {
3622         int *valp = table->data;
3623         int new;
3624
3625         if (!newval || !newlen)
3626                 return 0;
3627         if (newlen != sizeof(int))
3628                 return -EINVAL;
3629         if (get_user(new, (int __user *)newval))
3630                 return -EFAULT;
3631         if (new == *valp)
3632                 return 0;
3633         if (oldval && oldlenp) {
3634                 size_t len;
3635                 if (get_user(len, oldlenp))
3636                         return -EFAULT;
3637                 if (len) {
3638                         if (len > table->maxlen)
3639                                 len = table->maxlen;
3640                         if (copy_to_user(oldval, valp, len))
3641                                 return -EFAULT;
3642                         if (put_user(len, oldlenp))
3643                                 return -EFAULT;
3644                 }
3645         }
3646
3647         if (valp != &ipv6_devconf_dflt.forwarding) {
3648                 if (valp != &ipv6_devconf.forwarding) {
3649                         struct inet6_dev *idev = (struct inet6_dev *)table->extra1;
3650                         int changed;
3651                         if (unlikely(idev == NULL))
3652                                 return -ENODEV;
3653                         changed = (!*valp) ^ (!new);
3654                         *valp = new;
3655                         if (changed)
3656                                 dev_forward_change(idev);
3657                 } else {
3658                         *valp = new;
3659                         addrconf_forward_change();
3660                 }
3661
3662                 if (*valp)
3663                         rt6_purge_dflt_routers();
3664         } else
3665                 *valp = new;
3666
3667         return 1;
3668 }
3669
3670 static struct addrconf_sysctl_table
3671 {
3672         struct ctl_table_header *sysctl_header;
3673         ctl_table addrconf_vars[__NET_IPV6_MAX];
3674         ctl_table addrconf_dev[2];
3675         ctl_table addrconf_conf_dir[2];
3676         ctl_table addrconf_proto_dir[2];
3677         ctl_table addrconf_root_dir[2];
3678 } addrconf_sysctl __read_mostly = {
3679         .sysctl_header = NULL,
3680         .addrconf_vars = {
3681                 {
3682                         .ctl_name       =       NET_IPV6_FORWARDING,
3683                         .procname       =       "forwarding",
3684                         .data           =       &ipv6_devconf.forwarding,
3685                         .maxlen         =       sizeof(int),
3686                         .mode           =       0644,
3687                         .proc_handler   =       &addrconf_sysctl_forward,
3688                         .strategy       =       &addrconf_sysctl_forward_strategy,
3689                 },
3690                 {
3691                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
3692                         .procname       =       "hop_limit",
3693                         .data           =       &ipv6_devconf.hop_limit,
3694                         .maxlen         =       sizeof(int),
3695                         .mode           =       0644,
3696                         .proc_handler   =       proc_dointvec,
3697                 },
3698                 {
3699                         .ctl_name       =       NET_IPV6_MTU,
3700                         .procname       =       "mtu",
3701                         .data           =       &ipv6_devconf.mtu6,
3702                         .maxlen         =       sizeof(int),
3703                         .mode           =       0644,
3704                         .proc_handler   =       &proc_dointvec,
3705                 },
3706                 {
3707                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
3708                         .procname       =       "accept_ra",
3709                         .data           =       &ipv6_devconf.accept_ra,
3710                         .maxlen         =       sizeof(int),
3711                         .mode           =       0644,
3712                         .proc_handler   =       &proc_dointvec,
3713                 },
3714                 {
3715                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
3716                         .procname       =       "accept_redirects",
3717                         .data           =       &ipv6_devconf.accept_redirects,
3718                         .maxlen         =       sizeof(int),
3719                         .mode           =       0644,
3720                         .proc_handler   =       &proc_dointvec,
3721                 },
3722                 {
3723                         .ctl_name       =       NET_IPV6_AUTOCONF,
3724                         .procname       =       "autoconf",
3725                         .data           =       &ipv6_devconf.autoconf,
3726                         .maxlen         =       sizeof(int),
3727                         .mode           =       0644,
3728                         .proc_handler   =       &proc_dointvec,
3729                 },
3730                 {
3731                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
3732                         .procname       =       "dad_transmits",
3733                         .data           =       &ipv6_devconf.dad_transmits,
3734                         .maxlen         =       sizeof(int),
3735                         .mode           =       0644,
3736                         .proc_handler   =       &proc_dointvec,
3737                 },
3738                 {
3739                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
3740                         .procname       =       "router_solicitations",
3741                         .data           =       &ipv6_devconf.rtr_solicits,
3742                         .maxlen         =       sizeof(int),
3743                         .mode           =       0644,
3744                         .proc_handler   =       &proc_dointvec,
3745                 },
3746                 {
3747                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
3748                         .procname       =       "router_solicitation_interval",
3749                         .data           =       &ipv6_devconf.rtr_solicit_interval,
3750                         .maxlen         =       sizeof(int),
3751                         .mode           =       0644,
3752                         .proc_handler   =       &proc_dointvec_jiffies,
3753                         .strategy       =       &sysctl_jiffies,
3754                 },
3755                 {
3756                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
3757                         .procname       =       "router_solicitation_delay",
3758                         .data           =       &ipv6_devconf.rtr_solicit_delay,
3759                         .maxlen         =       sizeof(int),
3760                         .mode           =       0644,
3761                         .proc_handler   =       &proc_dointvec_jiffies,
3762                         .strategy       =       &sysctl_jiffies,
3763                 },
3764                 {
3765                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
3766                         .procname       =       "force_mld_version",
3767                         .data           =       &ipv6_devconf.force_mld_version,
3768                         .maxlen         =       sizeof(int),
3769                         .mode           =       0644,
3770                         .proc_handler   =       &proc_dointvec,
3771                 },
3772 #ifdef CONFIG_IPV6_PRIVACY
3773                 {
3774                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
3775                         .procname       =       "use_tempaddr",
3776                         .data           =       &ipv6_devconf.use_tempaddr,
3777                         .maxlen         =       sizeof(int),
3778                         .mode           =       0644,
3779                         .proc_handler   =       &proc_dointvec,
3780                 },
3781                 {
3782                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
3783                         .procname       =       "temp_valid_lft",
3784                         .data           =       &ipv6_devconf.temp_valid_lft,
3785                         .maxlen         =       sizeof(int),
3786                         .mode           =       0644,
3787                         .proc_handler   =       &proc_dointvec,
3788                 },
3789                 {
3790                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
3791                         .procname       =       "temp_prefered_lft",
3792                         .data           =       &ipv6_devconf.temp_prefered_lft,
3793                         .maxlen         =       sizeof(int),
3794                         .mode           =       0644,
3795                         .proc_handler   =       &proc_dointvec,
3796                 },
3797                 {
3798                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
3799                         .procname       =       "regen_max_retry",
3800                         .data           =       &ipv6_devconf.regen_max_retry,
3801                         .maxlen         =       sizeof(int),
3802                         .mode           =       0644,
3803                         .proc_handler   =       &proc_dointvec,
3804                 },
3805                 {
3806                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
3807                         .procname       =       "max_desync_factor",
3808                         .data           =       &ipv6_devconf.max_desync_factor,
3809                         .maxlen         =       sizeof(int),
3810                         .mode           =       0644,
3811                         .proc_handler   =       &proc_dointvec,
3812                 },
3813 #endif
3814                 {
3815                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
3816                         .procname       =       "max_addresses",
3817                         .data           =       &ipv6_devconf.max_addresses,
3818                         .maxlen         =       sizeof(int),
3819                         .mode           =       0644,
3820                         .proc_handler   =       &proc_dointvec,
3821                 },
3822                 {
3823                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
3824                         .procname       =       "accept_ra_defrtr",
3825                         .data           =       &ipv6_devconf.accept_ra_defrtr,
3826                         .maxlen         =       sizeof(int),
3827                         .mode           =       0644,
3828                         .proc_handler   =       &proc_dointvec,
3829                 },
3830                 {
3831                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
3832                         .procname       =       "accept_ra_pinfo",
3833                         .data           =       &ipv6_devconf.accept_ra_pinfo,
3834                         .maxlen         =       sizeof(int),
3835                         .mode           =       0644,
3836                         .proc_handler   =       &proc_dointvec,
3837                 },
3838 #ifdef CONFIG_IPV6_ROUTER_PREF
3839                 {
3840                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
3841                         .procname       =       "accept_ra_rtr_pref",
3842                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
3843                         .maxlen         =       sizeof(int),
3844                         .mode           =       0644,
3845                         .proc_handler   =       &proc_dointvec,
3846                 },
3847                 {
3848                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
3849                         .procname       =       "router_probe_interval",
3850                         .data           =       &ipv6_devconf.rtr_probe_interval,
3851                         .maxlen         =       sizeof(int),
3852                         .mode           =       0644,
3853                         .proc_handler   =       &proc_dointvec_jiffies,
3854                         .strategy       =       &sysctl_jiffies,
3855                 },
3856 #ifdef CONFIV_IPV6_ROUTE_INFO
3857                 {
3858                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
3859                         .procname       =       "accept_ra_rt_info_max_plen",
3860                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
3861                         .maxlen         =       sizeof(int),
3862                         .mode           =       0644,
3863                         .proc_handler   =       &proc_dointvec,
3864                 },
3865 #endif
3866 #endif
3867                 {
3868                         .ctl_name       =       NET_IPV6_PROXY_NDP,
3869                         .procname       =       "proxy_ndp",
3870                         .data           =       &ipv6_devconf.proxy_ndp,
3871                         .maxlen         =       sizeof(int),
3872                         .mode           =       0644,
3873                         .proc_handler   =       &proc_dointvec,
3874                 },
3875                 {
3876                         .ctl_name       =       0,      /* sentinel */
3877                 }
3878         },
3879         .addrconf_dev = {
3880                 {
3881                         .ctl_name       =       NET_PROTO_CONF_ALL,
3882                         .procname       =       "all",
3883                         .mode           =       0555,
3884                         .child          =       addrconf_sysctl.addrconf_vars,
3885                 },
3886                 {
3887                         .ctl_name       =       0,      /* sentinel */
3888                 }
3889         },
3890         .addrconf_conf_dir = {
3891                 {
3892                         .ctl_name       =       NET_IPV6_CONF,
3893                         .procname       =       "conf",
3894                         .mode           =       0555,
3895                         .child          =       addrconf_sysctl.addrconf_dev,
3896                 },
3897                 {
3898                         .ctl_name       =       0,      /* sentinel */
3899                 }
3900         },
3901         .addrconf_proto_dir = {
3902                 {
3903                         .ctl_name       =       NET_IPV6,
3904                         .procname       =       "ipv6",
3905                         .mode           =       0555,
3906                         .child          =       addrconf_sysctl.addrconf_conf_dir,
3907                 },
3908                 {
3909                         .ctl_name       =       0,      /* sentinel */
3910                 }
3911         },
3912         .addrconf_root_dir = {
3913                 {
3914                         .ctl_name       =       CTL_NET,
3915                         .procname       =       "net",
3916                         .mode           =       0555,
3917                         .child          =       addrconf_sysctl.addrconf_proto_dir,
3918                 },
3919                 {
3920                         .ctl_name       =       0,      /* sentinel */
3921                 }
3922         },
3923 };
3924
3925 static void addrconf_sysctl_register(struct inet6_dev *idev, struct ipv6_devconf *p)
3926 {
3927         int i;
3928         struct net_device *dev = idev ? idev->dev : NULL;
3929         struct addrconf_sysctl_table *t;
3930         char *dev_name = NULL;
3931
3932         t = kmalloc(sizeof(*t), GFP_KERNEL);
3933         if (t == NULL)
3934                 return;
3935         memcpy(t, &addrconf_sysctl, sizeof(*t));
3936         for (i=0; t->addrconf_vars[i].data; i++) {
3937                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
3938                 t->addrconf_vars[i].de = NULL;
3939                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
3940         }
3941         if (dev) {
3942                 dev_name = dev->name;
3943                 t->addrconf_dev[0].ctl_name = dev->ifindex;
3944         } else {
3945                 dev_name = "default";
3946                 t->addrconf_dev[0].ctl_name = NET_PROTO_CONF_DEFAULT;
3947         }
3948
3949         /*
3950          * Make a copy of dev_name, because '.procname' is regarded as const
3951          * by sysctl and we wouldn't want anyone to change it under our feet
3952          * (see SIOCSIFNAME).
3953          */
3954         dev_name = kstrdup(dev_name, GFP_KERNEL);
3955         if (!dev_name)
3956             goto free;
3957
3958         t->addrconf_dev[0].procname = dev_name;
3959
3960         t->addrconf_dev[0].child = t->addrconf_vars;
3961         t->addrconf_dev[0].de = NULL;
3962         t->addrconf_conf_dir[0].child = t->addrconf_dev;
3963         t->addrconf_conf_dir[0].de = NULL;
3964         t->addrconf_proto_dir[0].child = t->addrconf_conf_dir;
3965         t->addrconf_proto_dir[0].de = NULL;
3966         t->addrconf_root_dir[0].child = t->addrconf_proto_dir;
3967         t->addrconf_root_dir[0].de = NULL;
3968
3969         t->sysctl_header = register_sysctl_table(t->addrconf_root_dir, 0);
3970         if (t->sysctl_header == NULL)
3971                 goto free_procname;
3972         else
3973                 p->sysctl = t;
3974         return;
3975
3976         /* error path */
3977  free_procname:
3978         kfree(dev_name);
3979  free:
3980         kfree(t);
3981
3982         return;
3983 }
3984
3985 static void addrconf_sysctl_unregister(struct ipv6_devconf *p)
3986 {
3987         if (p->sysctl) {
3988                 struct addrconf_sysctl_table *t = p->sysctl;
3989                 p->sysctl = NULL;
3990                 unregister_sysctl_table(t->sysctl_header);
3991                 kfree(t->addrconf_dev[0].procname);
3992                 kfree(t);
3993         }
3994 }
3995
3996
3997 #endif
3998
3999 /*
4000  *      Device notifier
4001  */
4002
4003 int register_inet6addr_notifier(struct notifier_block *nb)
4004 {
4005         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4006 }
4007
4008 int unregister_inet6addr_notifier(struct notifier_block *nb)
4009 {
4010         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4011 }
4012
4013 /*
4014  *      Init / cleanup code
4015  */
4016
4017 int __init addrconf_init(void)
4018 {
4019         int err = 0;
4020
4021         /* The addrconf netdev notifier requires that loopback_dev
4022          * has it's ipv6 private information allocated and setup
4023          * before it can bring up and give link-local addresses
4024          * to other devices which are up.
4025          *
4026          * Unfortunately, loopback_dev is not necessarily the first
4027          * entry in the global dev_base list of net devices.  In fact,
4028          * it is likely to be the very last entry on that list.
4029          * So this causes the notifier registry below to try and
4030          * give link-local addresses to all devices besides loopback_dev
4031          * first, then loopback_dev, which cases all the non-loopback_dev
4032          * devices to fail to get a link-local address.
4033          *
4034          * So, as a temporary fix, allocate the ipv6 structure for
4035          * loopback_dev first by hand.
4036          * Longer term, all of the dependencies ipv6 has upon the loopback
4037          * device and it being up should be removed.
4038          */
4039         rtnl_lock();
4040         if (!ipv6_add_dev(&loopback_dev))
4041                 err = -ENOMEM;
4042         rtnl_unlock();
4043         if (err)
4044                 return err;
4045
4046         ip6_null_entry.rt6i_idev = in6_dev_get(&loopback_dev);
4047
4048         register_netdevice_notifier(&ipv6_dev_notf);
4049
4050         addrconf_verify(0);
4051         rtnetlink_links[PF_INET6] = inet6_rtnetlink_table;
4052 #ifdef CONFIG_SYSCTL
4053         addrconf_sysctl.sysctl_header =
4054                 register_sysctl_table(addrconf_sysctl.addrconf_root_dir, 0);
4055         addrconf_sysctl_register(NULL, &ipv6_devconf_dflt);
4056 #endif
4057
4058         return 0;
4059 }
4060
4061 void __exit addrconf_cleanup(void)
4062 {
4063         struct net_device *dev;
4064         struct inet6_dev *idev;
4065         struct inet6_ifaddr *ifa;
4066         int i;
4067
4068         unregister_netdevice_notifier(&ipv6_dev_notf);
4069
4070         rtnetlink_links[PF_INET6] = NULL;
4071 #ifdef CONFIG_SYSCTL
4072         addrconf_sysctl_unregister(&ipv6_devconf_dflt);
4073         addrconf_sysctl_unregister(&ipv6_devconf);
4074 #endif
4075
4076         rtnl_lock();
4077
4078         /*
4079          *      clean dev list.
4080          */
4081
4082         for (dev=dev_base; dev; dev=dev->next) {
4083                 if ((idev = __in6_dev_get(dev)) == NULL)
4084                         continue;
4085                 addrconf_ifdown(dev, 1);
4086         }
4087         addrconf_ifdown(&loopback_dev, 2);
4088
4089         /*
4090          *      Check hash table.
4091          */
4092
4093         write_lock_bh(&addrconf_hash_lock);
4094         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4095                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4096                         struct inet6_ifaddr *bifa;
4097
4098                         bifa = ifa;
4099                         ifa = ifa->lst_next;
4100                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4101                         /* Do not free it; something is wrong.
4102                            Now we can investigate it with debugger.
4103                          */
4104                 }
4105         }
4106         write_unlock_bh(&addrconf_hash_lock);
4107
4108         del_timer(&addr_chk_timer);
4109
4110         rtnl_unlock();
4111
4112 #ifdef CONFIG_PROC_FS
4113         proc_net_remove("if_inet6");
4114 #endif
4115 }