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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.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 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188         .disable_ipv6           = 0,
 189         .accept_dad             = 1,
 190 };
 191
 192 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 193         .forwarding             = 0,
 194         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 195         .mtu6                   = IPV6_MIN_MTU,
 196         .accept_ra              = 1,
 197         .accept_redirects       = 1,
 198         .autoconf               = 1,
 199         .dad_transmits          = 1,
 200         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 201         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 202         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 203 #ifdef CONFIG_IPV6_PRIVACY
 204         .use_tempaddr           = 0,
 205         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 206         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 207         .regen_max_retry        = REGEN_MAX_RETRY,
 208         .max_desync_factor      = MAX_DESYNC_FACTOR,
 209 #endif
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 1,
 224 };
 225
 226 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 227 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 228 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 229 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 230 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 231
 232 /* Check if a valid qdisc is available */
 233 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 234 {
 235         return !qdisc_tx_is_noop(dev);
 236 }
 237
 238 /* Check if a route is valid prefix route */
 239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 240 {
 241         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 242 }
 243
 244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 245 {
 246         if (del_timer(&ifp->timer))
 247                 __in6_ifa_put(ifp);
 248 }
 249
 250 enum addrconf_timer_t
 251 {
 252         AC_NONE,
 253         AC_DAD,
 254         AC_RS,
 255 };
 256
 257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 258                                enum addrconf_timer_t what,
 259                                unsigned long when)
 260 {
 261         if (!del_timer(&ifp->timer))
 262                 in6_ifa_hold(ifp);
 263
 264         switch (what) {
 265         case AC_DAD:
 266                 ifp->timer.function = addrconf_dad_timer;
 267                 break;
 268         case AC_RS:
 269                 ifp->timer.function = addrconf_rs_timer;
 270                 break;
 271         default:;
 272         }
 273         ifp->timer.expires = jiffies + when;
 274         add_timer(&ifp->timer);
 275 }
 276
 277 static int snmp6_alloc_dev(struct inet6_dev *idev)
 278 {
 279         if (snmp_mib_init((void **)idev->stats.ipv6,
 280                           sizeof(struct ipstats_mib)) < 0)
 281                 goto err_ip;
 282         if (snmp_mib_init((void **)idev->stats.icmpv6,
 283                           sizeof(struct icmpv6_mib)) < 0)
 284                 goto err_icmp;
 285         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 286                           sizeof(struct icmpv6msg_mib)) < 0)
 287                 goto err_icmpmsg;
 288
 289         return 0;
 290
 291 err_icmpmsg:
 292         snmp_mib_free((void **)idev->stats.icmpv6);
 293 err_icmp:
 294         snmp_mib_free((void **)idev->stats.ipv6);
 295 err_ip:
 296         return -ENOMEM;
 297 }
 298
 299 static void snmp6_free_dev(struct inet6_dev *idev)
 300 {
 301         snmp_mib_free((void **)idev->stats.icmpv6msg);
 302         snmp_mib_free((void **)idev->stats.icmpv6);
 303         snmp_mib_free((void **)idev->stats.ipv6);
 304 }
 305
 306 /* Nobody refers to this device, we may destroy it. */
 307
 308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 309 {
 310         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 311         kfree(idev);
 312 }
 313
 314 void in6_dev_finish_destroy(struct inet6_dev *idev)
 315 {
 316         struct net_device *dev = idev->dev;
 317
 318         WARN_ON(idev->addr_list != NULL);
 319         WARN_ON(idev->mc_list != NULL);
 320
 321 #ifdef NET_REFCNT_DEBUG
 322         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 323 #endif
 324         dev_put(dev);
 325         if (!idev->dead) {
 326                 printk("Freeing alive inet6 device %p\n", idev);
 327                 return;
 328         }
 329         snmp6_free_dev(idev);
 330         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 331 }
 332
 333 EXPORT_SYMBOL(in6_dev_finish_destroy);
 334
 335 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338
 339         ASSERT_RTNL();
 340
 341         if (dev->mtu < IPV6_MIN_MTU)
 342                 return NULL;
 343
 344         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 345
 346         if (ndev == NULL)
 347                 return NULL;
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 352         ndev->cnf.mtu6 = dev->mtu;
 353         ndev->cnf.sysctl = NULL;
 354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 355         if (ndev->nd_parms == NULL) {
 356                 kfree(ndev);
 357                 return NULL;
 358         }
 359         if (ndev->cnf.forwarding)
 360                 dev_disable_lro(dev);
 361         /* We refer to the device */
 362         dev_hold(dev);
 363
 364         if (snmp6_alloc_dev(ndev) < 0) {
 365                 ADBG((KERN_WARNING
 366                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 367                         __func__, dev->name));
 368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 369                 ndev->dead = 1;
 370                 in6_dev_finish_destroy(ndev);
 371                 return NULL;
 372         }
 373
 374         if (snmp6_register_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         /* One reference from device.  We must do this before
 385          * we invoke __ipv6_regen_rndid().
 386          */
 387         in6_dev_hold(ndev);
 388
 389         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 390                 ndev->cnf.accept_dad = -1;
 391
 392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 393         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 394                 printk(KERN_INFO
 395                        "%s: Disabled Multicast RS\n",
 396                        dev->name);
 397                 ndev->cnf.rtr_solicits = 0;
 398         }
 399 #endif
 400
 401 #ifdef CONFIG_IPV6_PRIVACY
 402         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 403         if ((dev->flags&IFF_LOOPBACK) ||
 404             dev->type == ARPHRD_TUNNEL ||
 405             dev->type == ARPHRD_TUNNEL6 ||
 406             dev->type == ARPHRD_SIT ||
 407             dev->type == ARPHRD_NONE) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Privacy Extensions\n",
 410                        dev->name);
 411                 ndev->cnf.use_tempaddr = -1;
 412         } else {
 413                 in6_dev_hold(ndev);
 414                 ipv6_regen_rndid((unsigned long) ndev);
 415         }
 416 #endif
 417
 418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 419                 ndev->if_flags |= IF_READY;
 420
 421         ipv6_mc_init_dev(ndev);
 422         ndev->tstamp = jiffies;
 423         addrconf_sysctl_register(ndev);
 424         /* protected by rtnl_lock */
 425         rcu_assign_pointer(dev->ip6_ptr, ndev);
 426
 427         /* Join all-node multicast group */
 428         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 429
 430         return ndev;
 431 }
 432
 433 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 434 {
 435         struct inet6_dev *idev;
 436
 437         ASSERT_RTNL();
 438
 439         if ((idev = __in6_dev_get(dev)) == NULL) {
 440                 if ((idev = ipv6_add_dev(dev)) == NULL)
 441                         return NULL;
 442         }
 443
 444         if (dev->flags&IFF_UP)
 445                 ipv6_mc_up(idev);
 446         return idev;
 447 }
 448
 449 #ifdef CONFIG_SYSCTL
 450 static void dev_forward_change(struct inet6_dev *idev)
 451 {
 452         struct net_device *dev;
 453         struct inet6_ifaddr *ifa;
 454
 455         if (!idev)
 456                 return;
 457         dev = idev->dev;
 458         if (idev->cnf.forwarding)
 459                 dev_disable_lro(dev);
 460         if (dev && (dev->flags & IFF_MULTICAST)) {
 461                 if (idev->cnf.forwarding)
 462                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463                 else
 464                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 465         }
 466         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 467                 if (ifa->flags&IFA_F_TENTATIVE)
 468                         continue;
 469                 if (idev->cnf.forwarding)
 470                         addrconf_join_anycast(ifa);
 471                 else
 472                         addrconf_leave_anycast(ifa);
 473         }
 474 }
 475
 476
 477 static void addrconf_forward_change(struct net *net, __s32 newf)
 478 {
 479         struct net_device *dev;
 480         struct inet6_dev *idev;
 481
 482         read_lock(&dev_base_lock);
 483         for_each_netdev(net, dev) {
 484                 rcu_read_lock();
 485                 idev = __in6_dev_get(dev);
 486                 if (idev) {
 487                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 488                         idev->cnf.forwarding = newf;
 489                         if (changed)
 490                                 dev_forward_change(idev);
 491                 }
 492                 rcu_read_unlock();
 493         }
 494         read_unlock(&dev_base_lock);
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock())
 506                 return restart_syscall();
 507
 508         if (p == &net->ipv6.devconf_all->forwarding) {
 509                 __s32 newf = net->ipv6.devconf_all->forwarding;
 510                 net->ipv6.devconf_dflt->forwarding = newf;
 511                 addrconf_forward_change(net, newf);
 512         } else if ((!*p) ^ (!old))
 513                 dev_forward_change((struct inet6_dev *)table->extra1);
 514         rtnl_unlock();
 515
 516         if (*p)
 517                 rt6_purge_dflt_routers(net);
 518         return 1;
 519 }
 520 #endif
 521
 522 /* Nobody refers to this ifaddr, destroy it */
 523
 524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 525 {
 526         WARN_ON(ifp->if_next != NULL);
 527         WARN_ON(ifp->lst_next != NULL);
 528
 529 #ifdef NET_REFCNT_DEBUG
 530         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 531 #endif
 532
 533         in6_dev_put(ifp->idev);
 534
 535         if (del_timer(&ifp->timer))
 536                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 537
 538         if (!ifp->dead) {
 539                 printk("Freeing alive inet6 address %p\n", ifp);
 540                 return;
 541         }
 542         dst_release(&ifp->rt->u.dst);
 543
 544         kfree(ifp);
 545 }
 546
 547 static void
 548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 549 {
 550         struct inet6_ifaddr *ifa, **ifap;
 551         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 552
 553         /*
 554          * Each device address list is sorted in order of scope -
 555          * global before linklocal.
 556          */
 557         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 558              ifap = &ifa->if_next) {
 559                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 560                         break;
 561         }
 562
 563         ifp->if_next = *ifap;
 564         *ifap = ifp;
 565 }
 566
 567 /*
 568  *      Hash function taken from net_alias.c
 569  */
 570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 571 {
 572         __u32 word;
 573
 574         /*
 575          * We perform the hash function over the last 64 bits of the address
 576          * This will include the IEEE address token on links that support it.
 577          */
 578
 579         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 580         word ^= (word >> 16);
 581         word ^= (word >> 8);
 582
 583         return ((word ^ (word >> 4)) & 0x0f);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         struct net *net = dev_net(idev->dev);
 595         int hash;
 596         int err = 0;
 597         int addr_type = ipv6_addr_type(addr);
 598
 599         if (addr_type == IPV6_ADDR_ANY ||
 600             addr_type & IPV6_ADDR_MULTICAST ||
 601             (!(idev->dev->flags & IFF_LOOPBACK) &&
 602              addr_type & IPV6_ADDR_LOOPBACK))
 603                 return ERR_PTR(-EADDRNOTAVAIL);
 604
 605         rcu_read_lock_bh();
 606         if (idev->dead) {
 607                 err = -ENODEV;                  /*XXX*/
 608                 goto out2;
 609         }
 610
 611         if (idev->cnf.disable_ipv6 || net->ipv6.devconf_all->disable_ipv6) {
 612                 err = -EACCES;
 613                 goto out2;
 614         }
 615
 616         write_lock(&addrconf_hash_lock);
 617
 618         /* Ignore adding duplicate addresses on an interface */
 619         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 620                 ADBG(("ipv6_add_addr: already assigned\n"));
 621                 err = -EEXIST;
 622                 goto out;
 623         }
 624
 625         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 626
 627         if (ifa == NULL) {
 628                 ADBG(("ipv6_add_addr: malloc failed\n"));
 629                 err = -ENOBUFS;
 630                 goto out;
 631         }
 632
 633         rt = addrconf_dst_alloc(idev, addr, 0);
 634         if (IS_ERR(rt)) {
 635                 err = PTR_ERR(rt);
 636                 goto out;
 637         }
 638
 639         ipv6_addr_copy(&ifa->addr, addr);
 640
 641         spin_lock_init(&ifa->lock);
 642         init_timer(&ifa->timer);
 643         ifa->timer.data = (unsigned long) ifa;
 644         ifa->scope = scope;
 645         ifa->prefix_len = pfxlen;
 646         ifa->flags = flags | IFA_F_TENTATIVE;
 647         ifa->cstamp = ifa->tstamp = jiffies;
 648
 649         ifa->rt = rt;
 650
 651         /*
 652          * part one of RFC 4429, section 3.3
 653          * We should not configure an address as
 654          * optimistic if we do not yet know the link
 655          * layer address of our nexhop router
 656          */
 657
 658         if (rt->rt6i_nexthop == NULL)
 659                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 660
 661         ifa->idev = idev;
 662         in6_dev_hold(idev);
 663         /* For caller */
 664         in6_ifa_hold(ifa);
 665
 666         /* Add to big hash table */
 667         hash = ipv6_addr_hash(addr);
 668
 669         ifa->lst_next = inet6_addr_lst[hash];
 670         inet6_addr_lst[hash] = ifa;
 671         in6_ifa_hold(ifa);
 672         write_unlock(&addrconf_hash_lock);
 673
 674         write_lock(&idev->lock);
 675         /* Add to inet6_dev unicast addr list. */
 676         ipv6_link_dev_addr(idev, ifa);
 677
 678 #ifdef CONFIG_IPV6_PRIVACY
 679         if (ifa->flags&IFA_F_TEMPORARY) {
 680                 ifa->tmp_next = idev->tempaddr_list;
 681                 idev->tempaddr_list = ifa;
 682                 in6_ifa_hold(ifa);
 683         }
 684 #endif
 685
 686         in6_ifa_hold(ifa);
 687         write_unlock(&idev->lock);
 688 out2:
 689         rcu_read_unlock_bh();
 690
 691         if (likely(err == 0))
 692                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 693         else {
 694                 kfree(ifa);
 695                 ifa = ERR_PTR(err);
 696         }
 697
 698         return ifa;
 699 out:
 700         write_unlock(&addrconf_hash_lock);
 701         goto out2;
 702 }
 703
 704 /* This function wants to get referenced ifp and releases it before return */
 705
 706 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 707 {
 708         struct inet6_ifaddr *ifa, **ifap;
 709         struct inet6_dev *idev = ifp->idev;
 710         int hash;
 711         int deleted = 0, onlink = 0;
 712         unsigned long expires = jiffies;
 713
 714         hash = ipv6_addr_hash(&ifp->addr);
 715
 716         ifp->dead = 1;
 717
 718         write_lock_bh(&addrconf_hash_lock);
 719         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 720              ifap = &ifa->lst_next) {
 721                 if (ifa == ifp) {
 722                         *ifap = ifa->lst_next;
 723                         __in6_ifa_put(ifp);
 724                         ifa->lst_next = NULL;
 725                         break;
 726                 }
 727         }
 728         write_unlock_bh(&addrconf_hash_lock);
 729
 730         write_lock_bh(&idev->lock);
 731 #ifdef CONFIG_IPV6_PRIVACY
 732         if (ifp->flags&IFA_F_TEMPORARY) {
 733                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 734                      ifap = &ifa->tmp_next) {
 735                         if (ifa == ifp) {
 736                                 *ifap = ifa->tmp_next;
 737                                 if (ifp->ifpub) {
 738                                         in6_ifa_put(ifp->ifpub);
 739                                         ifp->ifpub = NULL;
 740                                 }
 741                                 __in6_ifa_put(ifp);
 742                                 ifa->tmp_next = NULL;
 743                                 break;
 744                         }
 745                 }
 746         }
 747 #endif
 748
 749         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 750                 if (ifa == ifp) {
 751                         *ifap = ifa->if_next;
 752                         __in6_ifa_put(ifp);
 753                         ifa->if_next = NULL;
 754                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 755                                 break;
 756                         deleted = 1;
 757                         continue;
 758                 } else if (ifp->flags & IFA_F_PERMANENT) {
 759                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 760                                               ifp->prefix_len)) {
 761                                 if (ifa->flags & IFA_F_PERMANENT) {
 762                                         onlink = 1;
 763                                         if (deleted)
 764                                                 break;
 765                                 } else {
 766                                         unsigned long lifetime;
 767
 768                                         if (!onlink)
 769                                                 onlink = -1;
 770
 771                                         spin_lock(&ifa->lock);
 772
 773                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 774                                         /*
 775                                          * Note: Because this address is
 776                                          * not permanent, lifetime <
 777                                          * LONG_MAX / HZ here.
 778                                          */
 779                                         if (time_before(expires,
 780                                                         ifa->tstamp + lifetime * HZ))
 781                                                 expires = ifa->tstamp + lifetime * HZ;
 782                                         spin_unlock(&ifa->lock);
 783                                 }
 784                         }
 785                 }
 786                 ifap = &ifa->if_next;
 787         }
 788         write_unlock_bh(&idev->lock);
 789
 790         addrconf_del_timer(ifp);
 791
 792         ipv6_ifa_notify(RTM_DELADDR, ifp);
 793
 794         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 795
 796         /*
 797          * Purge or update corresponding prefix
 798          *
 799          * 1) we don't purge prefix here if address was not permanent.
 800          *    prefix is managed by its own lifetime.
 801          * 2) if there're no addresses, delete prefix.
 802          * 3) if there're still other permanent address(es),
 803          *    corresponding prefix is still permanent.
 804          * 4) otherwise, update prefix lifetime to the
 805          *    longest valid lifetime among the corresponding
 806          *    addresses on the device.
 807          *    Note: subsequent RA will update lifetime.
 808          *
 809          * --yoshfuji
 810          */
 811         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 812                 struct in6_addr prefix;
 813                 struct rt6_info *rt;
 814                 struct net *net = dev_net(ifp->idev->dev);
 815                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 816                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 817
 818                 if (rt && addrconf_is_prefix_route(rt)) {
 819                         if (onlink == 0) {
 820                                 ip6_del_rt(rt);
 821                                 rt = NULL;
 822                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 823                                 rt->rt6i_expires = expires;
 824                                 rt->rt6i_flags |= RTF_EXPIRES;
 825                         }
 826                 }
 827                 dst_release(&rt->u.dst);
 828         }
 829
 830         in6_ifa_put(ifp);
 831 }
 832
 833 #ifdef CONFIG_IPV6_PRIVACY
 834 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 835 {
 836         struct inet6_dev *idev = ifp->idev;
 837         struct in6_addr addr, *tmpaddr;
 838         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 839         unsigned long regen_advance;
 840         int tmp_plen;
 841         int ret = 0;
 842         int max_addresses;
 843         u32 addr_flags;
 844
 845         write_lock(&idev->lock);
 846         if (ift) {
 847                 spin_lock_bh(&ift->lock);
 848                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 849                 spin_unlock_bh(&ift->lock);
 850                 tmpaddr = &addr;
 851         } else {
 852                 tmpaddr = NULL;
 853         }
 854 retry:
 855         in6_dev_hold(idev);
 856         if (idev->cnf.use_tempaddr <= 0) {
 857                 write_unlock(&idev->lock);
 858                 printk(KERN_INFO
 859                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 860                 in6_dev_put(idev);
 861                 ret = -1;
 862                 goto out;
 863         }
 864         spin_lock_bh(&ifp->lock);
 865         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 866                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 867                 spin_unlock_bh(&ifp->lock);
 868                 write_unlock(&idev->lock);
 869                 printk(KERN_WARNING
 870                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 871                 in6_dev_put(idev);
 872                 ret = -1;
 873                 goto out;
 874         }
 875         in6_ifa_hold(ifp);
 876         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 877         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 878                 spin_unlock_bh(&ifp->lock);
 879                 write_unlock(&idev->lock);
 880                 printk(KERN_WARNING
 881                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 882                 in6_ifa_put(ifp);
 883                 in6_dev_put(idev);
 884                 ret = -1;
 885                 goto out;
 886         }
 887         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 888         tmp_valid_lft = min_t(__u32,
 889                               ifp->valid_lft,
 890                               idev->cnf.temp_valid_lft);
 891         tmp_prefered_lft = min_t(__u32,
 892                                  ifp->prefered_lft,
 893                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 894         tmp_plen = ifp->prefix_len;
 895         max_addresses = idev->cnf.max_addresses;
 896         tmp_cstamp = ifp->cstamp;
 897         tmp_tstamp = ifp->tstamp;
 898         spin_unlock_bh(&ifp->lock);
 899
 900         regen_advance = idev->cnf.regen_max_retry *
 901                         idev->cnf.dad_transmits *
 902                         idev->nd_parms->retrans_time / HZ;
 903         write_unlock(&idev->lock);
 904
 905         /* A temporary address is created only if this calculated Preferred
 906          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 907          * an implementation must not create a temporary address with a zero
 908          * Preferred Lifetime.
 909          */
 910         if (tmp_prefered_lft <= regen_advance) {
 911                 in6_ifa_put(ifp);
 912                 in6_dev_put(idev);
 913                 ret = -1;
 914                 goto out;
 915         }
 916
 917         addr_flags = IFA_F_TEMPORARY;
 918         /* set in addrconf_prefix_rcv() */
 919         if (ifp->flags & IFA_F_OPTIMISTIC)
 920                 addr_flags |= IFA_F_OPTIMISTIC;
 921
 922         ift = !max_addresses ||
 923               ipv6_count_addresses(idev) < max_addresses ?
 924                 ipv6_add_addr(idev, &addr, tmp_plen,
 925                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 926                               addr_flags) : NULL;
 927         if (!ift || IS_ERR(ift)) {
 928                 in6_ifa_put(ifp);
 929                 in6_dev_put(idev);
 930                 printk(KERN_INFO
 931                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 932                 tmpaddr = &addr;
 933                 write_lock(&idev->lock);
 934                 goto retry;
 935         }
 936
 937         spin_lock_bh(&ift->lock);
 938         ift->ifpub = ifp;
 939         ift->valid_lft = tmp_valid_lft;
 940         ift->prefered_lft = tmp_prefered_lft;
 941         ift->cstamp = tmp_cstamp;
 942         ift->tstamp = tmp_tstamp;
 943         spin_unlock_bh(&ift->lock);
 944
 945         addrconf_dad_start(ift, 0);
 946         in6_ifa_put(ift);
 947         in6_dev_put(idev);
 948 out:
 949         return ret;
 950 }
 951 #endif
 952
 953 /*
 954  *      Choose an appropriate source address (RFC3484)
 955  */
 956 enum {
 957         IPV6_SADDR_RULE_INIT = 0,
 958         IPV6_SADDR_RULE_LOCAL,
 959         IPV6_SADDR_RULE_SCOPE,
 960         IPV6_SADDR_RULE_PREFERRED,
 961 #ifdef CONFIG_IPV6_MIP6
 962         IPV6_SADDR_RULE_HOA,
 963 #endif
 964         IPV6_SADDR_RULE_OIF,
 965         IPV6_SADDR_RULE_LABEL,
 966 #ifdef CONFIG_IPV6_PRIVACY
 967         IPV6_SADDR_RULE_PRIVACY,
 968 #endif
 969         IPV6_SADDR_RULE_ORCHID,
 970         IPV6_SADDR_RULE_PREFIX,
 971         IPV6_SADDR_RULE_MAX
 972 };
 973
 974 struct ipv6_saddr_score {
 975         int                     rule;
 976         int                     addr_type;
 977         struct inet6_ifaddr     *ifa;
 978         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 979         int                     scopedist;
 980         int                     matchlen;
 981 };
 982
 983 struct ipv6_saddr_dst {
 984         const struct in6_addr *addr;
 985         int ifindex;
 986         int scope;
 987         int label;
 988         unsigned int prefs;
 989 };
 990
 991 static inline int ipv6_saddr_preferred(int type)
 992 {
 993         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 994                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 995                 return 1;
 996         return 0;
 997 }
 998
 999 static int ipv6_get_saddr_eval(struct net *net,
1000                                struct ipv6_saddr_score *score,
1001                                struct ipv6_saddr_dst *dst,
1002                                int i)
1003 {
1004         int ret;
1005
1006         if (i <= score->rule) {
1007                 switch (i) {
1008                 case IPV6_SADDR_RULE_SCOPE:
1009                         ret = score->scopedist;
1010                         break;
1011                 case IPV6_SADDR_RULE_PREFIX:
1012                         ret = score->matchlen;
1013                         break;
1014                 default:
1015                         ret = !!test_bit(i, score->scorebits);
1016                 }
1017                 goto out;
1018         }
1019
1020         switch (i) {
1021         case IPV6_SADDR_RULE_INIT:
1022                 /* Rule 0: remember if hiscore is not ready yet */
1023                 ret = !!score->ifa;
1024                 break;
1025         case IPV6_SADDR_RULE_LOCAL:
1026                 /* Rule 1: Prefer same address */
1027                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1028                 break;
1029         case IPV6_SADDR_RULE_SCOPE:
1030                 /* Rule 2: Prefer appropriate scope
1031                  *
1032                  *      ret
1033                  *       ^
1034                  *    -1 |  d 15
1035                  *    ---+--+-+---> scope
1036                  *       |
1037                  *       |             d is scope of the destination.
1038                  *  B-d  |  \
1039                  *       |   \      <- smaller scope is better if
1040                  *  B-15 |    \        if scope is enough for destinaion.
1041                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1042                  * d-C-1 | /
1043                  *       |/         <- greater is better
1044                  *   -C  /             if scope is not enough for destination.
1045                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1046                  *
1047                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1048                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1049                  * Assume B = 0 and we get C > 29.
1050                  */
1051                 ret = __ipv6_addr_src_scope(score->addr_type);
1052                 if (ret >= dst->scope)
1053                         ret = -ret;
1054                 else
1055                         ret -= 128;     /* 30 is enough */
1056                 score->scopedist = ret;
1057                 break;
1058         case IPV6_SADDR_RULE_PREFERRED:
1059                 /* Rule 3: Avoid deprecated and optimistic addresses */
1060                 ret = ipv6_saddr_preferred(score->addr_type) ||
1061                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1062                 break;
1063 #ifdef CONFIG_IPV6_MIP6
1064         case IPV6_SADDR_RULE_HOA:
1065             {
1066                 /* Rule 4: Prefer home address */
1067                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1068                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1069                 break;
1070             }
1071 #endif
1072         case IPV6_SADDR_RULE_OIF:
1073                 /* Rule 5: Prefer outgoing interface */
1074                 ret = (!dst->ifindex ||
1075                        dst->ifindex == score->ifa->idev->dev->ifindex);
1076                 break;
1077         case IPV6_SADDR_RULE_LABEL:
1078                 /* Rule 6: Prefer matching label */
1079                 ret = ipv6_addr_label(net,
1080                                       &score->ifa->addr, score->addr_type,
1081                                       score->ifa->idev->dev->ifindex) == dst->label;
1082                 break;
1083 #ifdef CONFIG_IPV6_PRIVACY
1084         case IPV6_SADDR_RULE_PRIVACY:
1085             {
1086                 /* Rule 7: Prefer public address
1087                  * Note: prefer temprary address if use_tempaddr >= 2
1088                  */
1089                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1090                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1091                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1092                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1093                 break;
1094             }
1095 #endif
1096         case IPV6_SADDR_RULE_ORCHID:
1097                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1098                  *          non-ORCHID vs non-ORCHID
1099                  */
1100                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1101                         ipv6_addr_orchid(dst->addr));
1102                 break;
1103         case IPV6_SADDR_RULE_PREFIX:
1104                 /* Rule 8: Use longest matching prefix */
1105                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1106                                                        dst->addr);
1107                 break;
1108         default:
1109                 ret = 0;
1110         }
1111
1112         if (ret)
1113                 __set_bit(i, score->scorebits);
1114         score->rule = i;
1115 out:
1116         return ret;
1117 }
1118
1119 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1120                        const struct in6_addr *daddr, unsigned int prefs,
1121                        struct in6_addr *saddr)
1122 {
1123         struct ipv6_saddr_score scores[2],
1124                                 *score = &scores[0], *hiscore = &scores[1];
1125         struct ipv6_saddr_dst dst;
1126         struct net_device *dev;
1127         int dst_type;
1128
1129         dst_type = __ipv6_addr_type(daddr);
1130         dst.addr = daddr;
1131         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1132         dst.scope = __ipv6_addr_src_scope(dst_type);
1133         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1134         dst.prefs = prefs;
1135
1136         hiscore->rule = -1;
1137         hiscore->ifa = NULL;
1138
1139         read_lock(&dev_base_lock);
1140         rcu_read_lock();
1141
1142         for_each_netdev(net, dev) {
1143                 struct inet6_dev *idev;
1144
1145                 /* Candidate Source Address (section 4)
1146                  *  - multicast and link-local destination address,
1147                  *    the set of candidate source address MUST only
1148                  *    include addresses assigned to interfaces
1149                  *    belonging to the same link as the outgoing
1150                  *    interface.
1151                  * (- For site-local destination addresses, the
1152                  *    set of candidate source addresses MUST only
1153                  *    include addresses assigned to interfaces
1154                  *    belonging to the same site as the outgoing
1155                  *    interface.)
1156                  */
1157                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1158                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1159                     dst.ifindex && dev->ifindex != dst.ifindex)
1160                         continue;
1161
1162                 idev = __in6_dev_get(dev);
1163                 if (!idev)
1164                         continue;
1165
1166                 read_lock_bh(&idev->lock);
1167                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1168                         int i;
1169
1170                         /*
1171                          * - Tentative Address (RFC2462 section 5.4)
1172                          *  - A tentative address is not considered
1173                          *    "assigned to an interface" in the traditional
1174                          *    sense, unless it is also flagged as optimistic.
1175                          * - Candidate Source Address (section 4)
1176                          *  - In any case, anycast addresses, multicast
1177                          *    addresses, and the unspecified address MUST
1178                          *    NOT be included in a candidate set.
1179                          */
1180                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1181                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1182                                 continue;
1183
1184                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1185
1186                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1187                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1188                                 LIMIT_NETDEBUG(KERN_DEBUG
1189                                                "ADDRCONF: unspecified / multicast address "
1190                                                "assigned as unicast address on %s",
1191                                                dev->name);
1192                                 continue;
1193                         }
1194
1195                         score->rule = -1;
1196                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1197
1198                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1199                                 int minihiscore, miniscore;
1200
1201                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1202                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1203
1204                                 if (minihiscore > miniscore) {
1205                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1206                                             score->scopedist > 0) {
1207                                                 /*
1208                                                  * special case:
1209                                                  * each remaining entry
1210                                                  * has too small (not enough)
1211                                                  * scope, because ifa entries
1212                                                  * are sorted by their scope
1213                                                  * values.
1214                                                  */
1215                                                 goto try_nextdev;
1216                                         }
1217                                         break;
1218                                 } else if (minihiscore < miniscore) {
1219                                         if (hiscore->ifa)
1220                                                 in6_ifa_put(hiscore->ifa);
1221
1222                                         in6_ifa_hold(score->ifa);
1223
1224                                         swap(hiscore, score);
1225
1226                                         /* restore our iterator */
1227                                         score->ifa = hiscore->ifa;
1228
1229                                         break;
1230                                 }
1231                         }
1232                 }
1233 try_nextdev:
1234                 read_unlock_bh(&idev->lock);
1235         }
1236         rcu_read_unlock();
1237         read_unlock(&dev_base_lock);
1238
1239         if (!hiscore->ifa)
1240                 return -EADDRNOTAVAIL;
1241
1242         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1243         in6_ifa_put(hiscore->ifa);
1244         return 0;
1245 }
1246
1247 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1248
1249 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1250                     unsigned char banned_flags)
1251 {
1252         struct inet6_dev *idev;
1253         int err = -EADDRNOTAVAIL;
1254
1255         rcu_read_lock();
1256         if ((idev = __in6_dev_get(dev)) != NULL) {
1257                 struct inet6_ifaddr *ifp;
1258
1259                 read_lock_bh(&idev->lock);
1260                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1261                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1262                                 ipv6_addr_copy(addr, &ifp->addr);
1263                                 err = 0;
1264                                 break;
1265                         }
1266                 }
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr * ifp;
1289         u8 hash = ipv6_addr_hash(addr);
1290
1291         read_lock_bh(&addrconf_hash_lock);
1292         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1293                 if (!net_eq(dev_net(ifp->idev->dev), net))
1294                         continue;
1295                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1296                     !(ifp->flags&IFA_F_TENTATIVE)) {
1297                         if (dev == NULL || ifp->idev->dev == dev ||
1298                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1299                                 break;
1300                 }
1301         }
1302         read_unlock_bh(&addrconf_hash_lock);
1303         return ifp != NULL;
1304 }
1305 EXPORT_SYMBOL(ipv6_chk_addr);
1306
1307 static
1308 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1309                        struct net_device *dev)
1310 {
1311         struct inet6_ifaddr * ifp;
1312         u8 hash = ipv6_addr_hash(addr);
1313
1314         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1315                 if (!net_eq(dev_net(ifp->idev->dev), net))
1316                         continue;
1317                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1318                         if (dev == NULL || ifp->idev->dev == dev)
1319                                 break;
1320                 }
1321         }
1322         return ifp != NULL;
1323 }
1324
1325 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1326 {
1327         struct inet6_dev *idev;
1328         struct inet6_ifaddr *ifa;
1329         int     onlink;
1330
1331         onlink = 0;
1332         rcu_read_lock();
1333         idev = __in6_dev_get(dev);
1334         if (idev) {
1335                 read_lock_bh(&idev->lock);
1336                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1337                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1338                                                    ifa->prefix_len);
1339                         if (onlink)
1340                                 break;
1341                 }
1342                 read_unlock_bh(&idev->lock);
1343         }
1344         rcu_read_unlock();
1345         return onlink;
1346 }
1347
1348 EXPORT_SYMBOL(ipv6_chk_prefix);
1349
1350 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1351                                      struct net_device *dev, int strict)
1352 {
1353         struct inet6_ifaddr * ifp;
1354         u8 hash = ipv6_addr_hash(addr);
1355
1356         read_lock_bh(&addrconf_hash_lock);
1357         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1358                 if (!net_eq(dev_net(ifp->idev->dev), net))
1359                         continue;
1360                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1361                         if (dev == NULL || ifp->idev->dev == dev ||
1362                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1363                                 in6_ifa_hold(ifp);
1364                                 break;
1365                         }
1366                 }
1367         }
1368         read_unlock_bh(&addrconf_hash_lock);
1369
1370         return ifp;
1371 }
1372
1373 /* Gets referenced address, destroys ifaddr */
1374
1375 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1376 {
1377         if (ifp->flags&IFA_F_PERMANENT) {
1378                 spin_lock_bh(&ifp->lock);
1379                 addrconf_del_timer(ifp);
1380                 ifp->flags |= IFA_F_TENTATIVE;
1381                 spin_unlock_bh(&ifp->lock);
1382                 in6_ifa_put(ifp);
1383 #ifdef CONFIG_IPV6_PRIVACY
1384         } else if (ifp->flags&IFA_F_TEMPORARY) {
1385                 struct inet6_ifaddr *ifpub;
1386                 spin_lock_bh(&ifp->lock);
1387                 ifpub = ifp->ifpub;
1388                 if (ifpub) {
1389                         in6_ifa_hold(ifpub);
1390                         spin_unlock_bh(&ifp->lock);
1391                         ipv6_create_tempaddr(ifpub, ifp);
1392                         in6_ifa_put(ifpub);
1393                 } else {
1394                         spin_unlock_bh(&ifp->lock);
1395                 }
1396                 ipv6_del_addr(ifp);
1397 #endif
1398         } else
1399                 ipv6_del_addr(ifp);
1400 }
1401
1402 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1403 {
1404         struct inet6_dev *idev = ifp->idev;
1405
1406         if (net_ratelimit())
1407                 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1408                         ifp->idev->dev->name);
1409
1410         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1411                 struct in6_addr addr;
1412
1413                 addr.s6_addr32[0] = htonl(0xfe800000);
1414                 addr.s6_addr32[1] = 0;
1415
1416                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1417                     ipv6_addr_equal(&ifp->addr, &addr)) {
1418                         /* DAD failed for link-local based on MAC address */
1419                         idev->cnf.disable_ipv6 = 1;
1420
1421                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1422                                 ifp->idev->dev->name);
1423                 }
1424         }
1425
1426         addrconf_dad_stop(ifp);
1427 }
1428
1429 /* Join to solicited addr multicast group. */
1430
1431 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1432 {
1433         struct in6_addr maddr;
1434
1435         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1436                 return;
1437
1438         addrconf_addr_solict_mult(addr, &maddr);
1439         ipv6_dev_mc_inc(dev, &maddr);
1440 }
1441
1442 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1443 {
1444         struct in6_addr maddr;
1445
1446         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1447                 return;
1448
1449         addrconf_addr_solict_mult(addr, &maddr);
1450         __ipv6_dev_mc_dec(idev, &maddr);
1451 }
1452
1453 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1454 {
1455         struct in6_addr addr;
1456         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1457         if (ipv6_addr_any(&addr))
1458                 return;
1459         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1460 }
1461
1462 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1463 {
1464         struct in6_addr addr;
1465         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1466         if (ipv6_addr_any(&addr))
1467                 return;
1468         __ipv6_dev_ac_dec(ifp->idev, &addr);
1469 }
1470
1471 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1472 {
1473         if (dev->addr_len != ETH_ALEN)
1474                 return -1;
1475         memcpy(eui, dev->dev_addr, 3);
1476         memcpy(eui + 5, dev->dev_addr + 3, 3);
1477
1478         /*
1479          * The zSeries OSA network cards can be shared among various
1480          * OS instances, but the OSA cards have only one MAC address.
1481          * This leads to duplicate address conflicts in conjunction
1482          * with IPv6 if more than one instance uses the same card.
1483          *
1484          * The driver for these cards can deliver a unique 16-bit
1485          * identifier for each instance sharing the same card.  It is
1486          * placed instead of 0xFFFE in the interface identifier.  The
1487          * "u" bit of the interface identifier is not inverted in this
1488          * case.  Hence the resulting interface identifier has local
1489          * scope according to RFC2373.
1490          */
1491         if (dev->dev_id) {
1492                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1493                 eui[4] = dev->dev_id & 0xFF;
1494         } else {
1495                 eui[3] = 0xFF;
1496                 eui[4] = 0xFE;
1497                 eui[0] ^= 2;
1498         }
1499         return 0;
1500 }
1501
1502 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1503 {
1504         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1505         if (dev->addr_len != ARCNET_ALEN)
1506                 return -1;
1507         memset(eui, 0, 7);
1508         eui[7] = *(u8*)dev->dev_addr;
1509         return 0;
1510 }
1511
1512 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1513 {
1514         if (dev->addr_len != INFINIBAND_ALEN)
1515                 return -1;
1516         memcpy(eui, dev->dev_addr + 12, 8);
1517         eui[0] |= 2;
1518         return 0;
1519 }
1520
1521 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1522 {
1523         if (addr == 0)
1524                 return -1;
1525         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1526                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1527                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1528                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1529                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1530                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1531         eui[1] = 0;
1532         eui[2] = 0x5E;
1533         eui[3] = 0xFE;
1534         memcpy(eui + 4, &addr, 4);
1535         return 0;
1536 }
1537 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1538
1539 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1540 {
1541         if (dev->priv_flags & IFF_ISATAP)
1542                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1543         return -1;
1544 }
1545
1546 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1547 {
1548         switch (dev->type) {
1549         case ARPHRD_ETHER:
1550         case ARPHRD_FDDI:
1551         case ARPHRD_IEEE802_TR:
1552                 return addrconf_ifid_eui48(eui, dev);
1553         case ARPHRD_ARCNET:
1554                 return addrconf_ifid_arcnet(eui, dev);
1555         case ARPHRD_INFINIBAND:
1556                 return addrconf_ifid_infiniband(eui, dev);
1557         case ARPHRD_SIT:
1558                 return addrconf_ifid_sit(eui, dev);
1559         }
1560         return -1;
1561 }
1562
1563 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1564 {
1565         int err = -1;
1566         struct inet6_ifaddr *ifp;
1567
1568         read_lock_bh(&idev->lock);
1569         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1570                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1571                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1572                         err = 0;
1573                         break;
1574                 }
1575         }
1576         read_unlock_bh(&idev->lock);
1577         return err;
1578 }
1579
1580 #ifdef CONFIG_IPV6_PRIVACY
1581 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1582 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1583 {
1584 regen:
1585         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1586         idev->rndid[0] &= ~0x02;
1587
1588         /*
1589          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1590          * check if generated address is not inappropriate
1591          *
1592          *  - Reserved subnet anycast (RFC 2526)
1593          *      11111101 11....11 1xxxxxxx
1594          *  - ISATAP (RFC4214) 6.1
1595          *      00-00-5E-FE-xx-xx-xx-xx
1596          *  - value 0
1597          *  - XXX: already assigned to an address on the device
1598          */
1599         if (idev->rndid[0] == 0xfd &&
1600             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1601             (idev->rndid[7]&0x80))
1602                 goto regen;
1603         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1604                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1605                         goto regen;
1606                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1607                         goto regen;
1608         }
1609
1610         return 0;
1611 }
1612
1613 static void ipv6_regen_rndid(unsigned long data)
1614 {
1615         struct inet6_dev *idev = (struct inet6_dev *) data;
1616         unsigned long expires;
1617
1618         rcu_read_lock_bh();
1619         write_lock_bh(&idev->lock);
1620
1621         if (idev->dead)
1622                 goto out;
1623
1624         if (__ipv6_regen_rndid(idev) < 0)
1625                 goto out;
1626
1627         expires = jiffies +
1628                 idev->cnf.temp_prefered_lft * HZ -
1629                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1630         if (time_before(expires, jiffies)) {
1631                 printk(KERN_WARNING
1632                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1633                         idev->dev->name);
1634                 goto out;
1635         }
1636
1637         if (!mod_timer(&idev->regen_timer, expires))
1638                 in6_dev_hold(idev);
1639
1640 out:
1641         write_unlock_bh(&idev->lock);
1642         rcu_read_unlock_bh();
1643         in6_dev_put(idev);
1644 }
1645
1646 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1647         int ret = 0;
1648
1649         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1650                 ret = __ipv6_regen_rndid(idev);
1651         return ret;
1652 }
1653 #endif
1654
1655 /*
1656  *      Add prefix route.
1657  */
1658
1659 static void
1660 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1661                       unsigned long expires, u32 flags)
1662 {
1663         struct fib6_config cfg = {
1664                 .fc_table = RT6_TABLE_PREFIX,
1665                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1666                 .fc_ifindex = dev->ifindex,
1667                 .fc_expires = expires,
1668                 .fc_dst_len = plen,
1669                 .fc_flags = RTF_UP | flags,
1670                 .fc_nlinfo.nl_net = dev_net(dev),
1671                 .fc_protocol = RTPROT_KERNEL,
1672         };
1673
1674         ipv6_addr_copy(&cfg.fc_dst, pfx);
1675
1676         /* Prevent useless cloning on PtP SIT.
1677            This thing is done here expecting that the whole
1678            class of non-broadcast devices need not cloning.
1679          */
1680 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1681         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1682                 cfg.fc_flags |= RTF_NONEXTHOP;
1683 #endif
1684
1685         ip6_route_add(&cfg);
1686 }
1687
1688 /* Create "default" multicast route to the interface */
1689
1690 static void addrconf_add_mroute(struct net_device *dev)
1691 {
1692         struct fib6_config cfg = {
1693                 .fc_table = RT6_TABLE_LOCAL,
1694                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1695                 .fc_ifindex = dev->ifindex,
1696                 .fc_dst_len = 8,
1697                 .fc_flags = RTF_UP,
1698                 .fc_nlinfo.nl_net = dev_net(dev),
1699         };
1700
1701         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1702
1703         ip6_route_add(&cfg);
1704 }
1705
1706 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1707 static void sit_route_add(struct net_device *dev)
1708 {
1709         struct fib6_config cfg = {
1710                 .fc_table = RT6_TABLE_MAIN,
1711                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1712                 .fc_ifindex = dev->ifindex,
1713                 .fc_dst_len = 96,
1714                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1715                 .fc_nlinfo.nl_net = dev_net(dev),
1716         };
1717
1718         /* prefix length - 96 bits "::d.d.d.d" */
1719         ip6_route_add(&cfg);
1720 }
1721 #endif
1722
1723 static void addrconf_add_lroute(struct net_device *dev)
1724 {
1725         struct in6_addr addr;
1726
1727         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1728         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1729 }
1730
1731 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1732 {
1733         struct inet6_dev *idev;
1734
1735         ASSERT_RTNL();
1736
1737         if ((idev = ipv6_find_idev(dev)) == NULL)
1738                 return NULL;
1739
1740         /* Add default multicast route */
1741         addrconf_add_mroute(dev);
1742
1743         /* Add link local route */
1744         addrconf_add_lroute(dev);
1745         return idev;
1746 }
1747
1748 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1749 {
1750         struct prefix_info *pinfo;
1751         __u32 valid_lft;
1752         __u32 prefered_lft;
1753         int addr_type;
1754         struct inet6_dev *in6_dev;
1755
1756         pinfo = (struct prefix_info *) opt;
1757
1758         if (len < sizeof(struct prefix_info)) {
1759                 ADBG(("addrconf: prefix option too short\n"));
1760                 return;
1761         }
1762
1763         /*
1764          *      Validation checks ([ADDRCONF], page 19)
1765          */
1766
1767         addr_type = ipv6_addr_type(&pinfo->prefix);
1768
1769         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1770                 return;
1771
1772         valid_lft = ntohl(pinfo->valid);
1773         prefered_lft = ntohl(pinfo->prefered);
1774
1775         if (prefered_lft > valid_lft) {
1776                 if (net_ratelimit())
1777                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1778                 return;
1779         }
1780
1781         in6_dev = in6_dev_get(dev);
1782
1783         if (in6_dev == NULL) {
1784                 if (net_ratelimit())
1785                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1786                 return;
1787         }
1788
1789         /*
1790          *      Two things going on here:
1791          *      1) Add routes for on-link prefixes
1792          *      2) Configure prefixes with the auto flag set
1793          */
1794
1795         if (pinfo->onlink) {
1796                 struct rt6_info *rt;
1797                 unsigned long rt_expires;
1798
1799                 /* Avoid arithmetic overflow. Really, we could
1800                  * save rt_expires in seconds, likely valid_lft,
1801                  * but it would require division in fib gc, that it
1802                  * not good.
1803                  */
1804                 if (HZ > USER_HZ)
1805                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1806                 else
1807                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1808
1809                 if (addrconf_finite_timeout(rt_expires))
1810                         rt_expires *= HZ;
1811
1812                 rt = rt6_lookup(dev_net(dev), &pinfo->prefix, NULL,
1813                                 dev->ifindex, 1);
1814
1815                 if (rt && addrconf_is_prefix_route(rt)) {
1816                         /* Autoconf prefix route */
1817                         if (valid_lft == 0) {
1818                                 ip6_del_rt(rt);
1819                                 rt = NULL;
1820                         } else if (addrconf_finite_timeout(rt_expires)) {
1821                                 /* not infinity */
1822                                 rt->rt6i_expires = jiffies + rt_expires;
1823                                 rt->rt6i_flags |= RTF_EXPIRES;
1824                         } else {
1825                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1826                                 rt->rt6i_expires = 0;
1827                         }
1828                 } else if (valid_lft) {
1829                         clock_t expires = 0;
1830                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1831                         if (addrconf_finite_timeout(rt_expires)) {
1832                                 /* not infinity */
1833                                 flags |= RTF_EXPIRES;
1834                                 expires = jiffies_to_clock_t(rt_expires);
1835                         }
1836                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1837                                               dev, expires, flags);
1838                 }
1839                 if (rt)
1840                         dst_release(&rt->u.dst);
1841         }
1842
1843         /* Try to figure out our local address for this prefix */
1844
1845         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1846                 struct inet6_ifaddr * ifp;
1847                 struct in6_addr addr;
1848                 int create = 0, update_lft = 0;
1849                 struct net *net = dev_net(dev);
1850
1851                 if (pinfo->prefix_len == 64) {
1852                         memcpy(&addr, &pinfo->prefix, 8);
1853                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1854                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1855                                 in6_dev_put(in6_dev);
1856                                 return;
1857                         }
1858                         goto ok;
1859                 }
1860                 if (net_ratelimit())
1861                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1862                                pinfo->prefix_len);
1863                 in6_dev_put(in6_dev);
1864                 return;
1865
1866 ok:
1867
1868                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1869
1870                 if (ifp == NULL && valid_lft) {
1871                         int max_addresses = in6_dev->cnf.max_addresses;
1872                         u32 addr_flags = 0;
1873
1874 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1875                         if (in6_dev->cnf.optimistic_dad &&
1876                             !net->ipv6.devconf_all->forwarding)
1877                                 addr_flags = IFA_F_OPTIMISTIC;
1878 #endif
1879
1880                         /* Do not allow to create too much of autoconfigured
1881                          * addresses; this would be too easy way to crash kernel.
1882                          */
1883                         if (!max_addresses ||
1884                             ipv6_count_addresses(in6_dev) < max_addresses)
1885                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1886                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1887                                                     addr_flags);
1888
1889                         if (!ifp || IS_ERR(ifp)) {
1890                                 in6_dev_put(in6_dev);
1891                                 return;
1892                         }
1893
1894                         update_lft = create = 1;
1895                         ifp->cstamp = jiffies;
1896                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1897                 }
1898
1899                 if (ifp) {
1900                         int flags;
1901                         unsigned long now;
1902 #ifdef CONFIG_IPV6_PRIVACY
1903                         struct inet6_ifaddr *ift;
1904 #endif
1905                         u32 stored_lft;
1906
1907                         /* update lifetime (RFC2462 5.5.3 e) */
1908                         spin_lock(&ifp->lock);
1909                         now = jiffies;
1910                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1911                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1912                         else
1913                                 stored_lft = 0;
1914                         if (!update_lft && stored_lft) {
1915                                 if (valid_lft > MIN_VALID_LIFETIME ||
1916                                     valid_lft > stored_lft)
1917                                         update_lft = 1;
1918                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1919                                         /* valid_lft <= stored_lft is always true */
1920                                         /* XXX: IPsec */
1921                                         update_lft = 0;
1922                                 } else {
1923                                         valid_lft = MIN_VALID_LIFETIME;
1924                                         if (valid_lft < prefered_lft)
1925                                                 prefered_lft = valid_lft;
1926                                         update_lft = 1;
1927                                 }
1928                         }
1929
1930                         if (update_lft) {
1931                                 ifp->valid_lft = valid_lft;
1932                                 ifp->prefered_lft = prefered_lft;
1933                                 ifp->tstamp = now;
1934                                 flags = ifp->flags;
1935                                 ifp->flags &= ~IFA_F_DEPRECATED;
1936                                 spin_unlock(&ifp->lock);
1937
1938                                 if (!(flags&IFA_F_TENTATIVE))
1939                                         ipv6_ifa_notify(0, ifp);
1940                         } else
1941                                 spin_unlock(&ifp->lock);
1942
1943 #ifdef CONFIG_IPV6_PRIVACY
1944                         read_lock_bh(&in6_dev->lock);
1945                         /* update all temporary addresses in the list */
1946                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1947                                 /*
1948                                  * When adjusting the lifetimes of an existing
1949                                  * temporary address, only lower the lifetimes.
1950                                  * Implementations must not increase the
1951                                  * lifetimes of an existing temporary address
1952                                  * when processing a Prefix Information Option.
1953                                  */
1954                                 if (ifp != ift->ifpub)
1955                                         continue;
1956
1957                                 spin_lock(&ift->lock);
1958                                 flags = ift->flags;
1959                                 if (ift->valid_lft > valid_lft &&
1960                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1961                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1962                                 if (ift->prefered_lft > prefered_lft &&
1963                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1964                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1965                                 spin_unlock(&ift->lock);
1966                                 if (!(flags&IFA_F_TENTATIVE))
1967                                         ipv6_ifa_notify(0, ift);
1968                         }
1969
1970                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1971                                 /*
1972                                  * When a new public address is created as described in [ADDRCONF],
1973                                  * also create a new temporary address.
1974                                  */
1975                                 read_unlock_bh(&in6_dev->lock);
1976                                 ipv6_create_tempaddr(ifp, NULL);
1977                         } else {
1978                                 read_unlock_bh(&in6_dev->lock);
1979                         }
1980 #endif
1981                         in6_ifa_put(ifp);
1982                         addrconf_verify(0);
1983                 }
1984         }
1985         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
1986         in6_dev_put(in6_dev);
1987 }
1988
1989 /*
1990  *      Set destination address.
1991  *      Special case for SIT interfaces where we create a new "virtual"
1992  *      device.
1993  */
1994 int addrconf_set_dstaddr(struct net *net, void __user *arg)
1995 {
1996         struct in6_ifreq ireq;
1997         struct net_device *dev;
1998         int err = -EINVAL;
1999
2000         rtnl_lock();
2001
2002         err = -EFAULT;
2003         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2004                 goto err_exit;
2005
2006         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2007
2008         err = -ENODEV;
2009         if (dev == NULL)
2010                 goto err_exit;
2011
2012 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2013         if (dev->type == ARPHRD_SIT) {
2014                 const struct net_device_ops *ops = dev->netdev_ops;
2015                 struct ifreq ifr;
2016                 struct ip_tunnel_parm p;
2017
2018                 err = -EADDRNOTAVAIL;
2019                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2020                         goto err_exit;
2021
2022                 memset(&p, 0, sizeof(p));
2023                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2024                 p.iph.saddr = 0;
2025                 p.iph.version = 4;
2026                 p.iph.ihl = 5;
2027                 p.iph.protocol = IPPROTO_IPV6;
2028                 p.iph.ttl = 64;
2029                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2030
2031                 if (ops->ndo_do_ioctl) {
2032                         mm_segment_t oldfs = get_fs();
2033
2034                         set_fs(KERNEL_DS);
2035                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2036                         set_fs(oldfs);
2037                 } else
2038                         err = -EOPNOTSUPP;
2039
2040                 if (err == 0) {
2041                         err = -ENOBUFS;
2042                         dev = __dev_get_by_name(net, p.name);
2043                         if (!dev)
2044                                 goto err_exit;
2045                         err = dev_open(dev);
2046                 }
2047         }
2048 #endif
2049
2050 err_exit:
2051         rtnl_unlock();
2052         return err;
2053 }
2054
2055 /*
2056  *      Manual configuration of address on an interface
2057  */
2058 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2059                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2060                           __u32 valid_lft)
2061 {
2062         struct inet6_ifaddr *ifp;
2063         struct inet6_dev *idev;
2064         struct net_device *dev;
2065         int scope;
2066         u32 flags;
2067         clock_t expires;
2068         unsigned long timeout;
2069
2070         ASSERT_RTNL();
2071
2072         if (plen > 128)
2073                 return -EINVAL;
2074
2075         /* check the lifetime */
2076         if (!valid_lft || prefered_lft > valid_lft)
2077                 return -EINVAL;
2078
2079         dev = __dev_get_by_index(net, ifindex);
2080         if (!dev)
2081                 return -ENODEV;
2082
2083         if ((idev = addrconf_add_dev(dev)) == NULL)
2084                 return -ENOBUFS;
2085
2086         scope = ipv6_addr_scope(pfx);
2087
2088         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2089         if (addrconf_finite_timeout(timeout)) {
2090                 expires = jiffies_to_clock_t(timeout * HZ);
2091                 valid_lft = timeout;
2092                 flags = RTF_EXPIRES;
2093         } else {
2094                 expires = 0;
2095                 flags = 0;
2096                 ifa_flags |= IFA_F_PERMANENT;
2097         }
2098
2099         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2100         if (addrconf_finite_timeout(timeout)) {
2101                 if (timeout == 0)
2102                         ifa_flags |= IFA_F_DEPRECATED;
2103                 prefered_lft = timeout;
2104         }
2105
2106         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2107
2108         if (!IS_ERR(ifp)) {
2109                 spin_lock_bh(&ifp->lock);
2110                 ifp->valid_lft = valid_lft;
2111                 ifp->prefered_lft = prefered_lft;
2112                 ifp->tstamp = jiffies;
2113                 spin_unlock_bh(&ifp->lock);
2114
2115                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2116                                       expires, flags);
2117                 /*
2118                  * Note that section 3.1 of RFC 4429 indicates
2119                  * that the Optimistic flag should not be set for
2120                  * manually configured addresses
2121                  */
2122                 addrconf_dad_start(ifp, 0);
2123                 in6_ifa_put(ifp);
2124                 addrconf_verify(0);
2125                 return 0;
2126         }
2127
2128         return PTR_ERR(ifp);
2129 }
2130
2131 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2132                           unsigned int plen)
2133 {
2134         struct inet6_ifaddr *ifp;
2135         struct inet6_dev *idev;
2136         struct net_device *dev;
2137
2138         if (plen > 128)
2139                 return -EINVAL;
2140
2141         dev = __dev_get_by_index(net, ifindex);
2142         if (!dev)
2143                 return -ENODEV;
2144
2145         if ((idev = __in6_dev_get(dev)) == NULL)
2146                 return -ENXIO;
2147
2148         read_lock_bh(&idev->lock);
2149         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2150                 if (ifp->prefix_len == plen &&
2151                     ipv6_addr_equal(pfx, &ifp->addr)) {
2152                         in6_ifa_hold(ifp);
2153                         read_unlock_bh(&idev->lock);
2154
2155                         ipv6_del_addr(ifp);
2156
2157                         /* If the last address is deleted administratively,
2158                            disable IPv6 on this interface.
2159                          */
2160                         if (idev->addr_list == NULL)
2161                                 addrconf_ifdown(idev->dev, 1);
2162                         return 0;
2163                 }
2164         }
2165         read_unlock_bh(&idev->lock);
2166         return -EADDRNOTAVAIL;
2167 }
2168
2169
2170 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2171 {
2172         struct in6_ifreq ireq;
2173         int err;
2174
2175         if (!capable(CAP_NET_ADMIN))
2176                 return -EPERM;
2177
2178         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2179                 return -EFAULT;
2180
2181         rtnl_lock();
2182         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2183                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2184                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2185         rtnl_unlock();
2186         return err;
2187 }
2188
2189 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2190 {
2191         struct in6_ifreq ireq;
2192         int err;
2193
2194         if (!capable(CAP_NET_ADMIN))
2195                 return -EPERM;
2196
2197         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2198                 return -EFAULT;
2199
2200         rtnl_lock();
2201         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2202                              ireq.ifr6_prefixlen);
2203         rtnl_unlock();
2204         return err;
2205 }
2206
2207 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2208                      int plen, int scope)
2209 {
2210         struct inet6_ifaddr *ifp;
2211
2212         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2213         if (!IS_ERR(ifp)) {
2214                 spin_lock_bh(&ifp->lock);
2215                 ifp->flags &= ~IFA_F_TENTATIVE;
2216                 spin_unlock_bh(&ifp->lock);
2217                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2218                 in6_ifa_put(ifp);
2219         }
2220 }
2221
2222 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2223 static void sit_add_v4_addrs(struct inet6_dev *idev)
2224 {
2225         struct in6_addr addr;
2226         struct net_device *dev;
2227         struct net *net = dev_net(idev->dev);
2228         int scope;
2229
2230         ASSERT_RTNL();
2231
2232         memset(&addr, 0, sizeof(struct in6_addr));
2233         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2234
2235         if (idev->dev->flags&IFF_POINTOPOINT) {
2236                 addr.s6_addr32[0] = htonl(0xfe800000);
2237                 scope = IFA_LINK;
2238         } else {
2239                 scope = IPV6_ADDR_COMPATv4;
2240         }
2241
2242         if (addr.s6_addr32[3]) {
2243                 add_addr(idev, &addr, 128, scope);
2244                 return;
2245         }
2246
2247         for_each_netdev(net, dev) {
2248                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2249                 if (in_dev && (dev->flags & IFF_UP)) {
2250                         struct in_ifaddr * ifa;
2251
2252                         int flag = scope;
2253
2254                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2255                                 int plen;
2256
2257                                 addr.s6_addr32[3] = ifa->ifa_local;
2258
2259                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2260                                         continue;
2261                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2262                                         if (idev->dev->flags&IFF_POINTOPOINT)
2263                                                 continue;
2264                                         flag |= IFA_HOST;
2265                                 }
2266                                 if (idev->dev->flags&IFF_POINTOPOINT)
2267                                         plen = 64;
2268                                 else
2269                                         plen = 96;
2270
2271                                 add_addr(idev, &addr, plen, flag);
2272                         }
2273                 }
2274         }
2275 }
2276 #endif
2277
2278 static void init_loopback(struct net_device *dev)
2279 {
2280         struct inet6_dev  *idev;
2281
2282         /* ::1 */
2283
2284         ASSERT_RTNL();
2285
2286         if ((idev = ipv6_find_idev(dev)) == NULL) {
2287                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2288                 return;
2289         }
2290
2291         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2292 }
2293
2294 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2295 {
2296         struct inet6_ifaddr * ifp;
2297         u32 addr_flags = IFA_F_PERMANENT;
2298
2299 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2300         if (idev->cnf.optimistic_dad &&
2301             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2302                 addr_flags |= IFA_F_OPTIMISTIC;
2303 #endif
2304
2305
2306         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2307         if (!IS_ERR(ifp)) {
2308                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2309                 addrconf_dad_start(ifp, 0);
2310                 in6_ifa_put(ifp);
2311         }
2312 }
2313
2314 static void addrconf_dev_config(struct net_device *dev)
2315 {
2316         struct in6_addr addr;
2317         struct inet6_dev    * idev;
2318
2319         ASSERT_RTNL();
2320
2321         if ((dev->type != ARPHRD_ETHER) &&
2322             (dev->type != ARPHRD_FDDI) &&
2323             (dev->type != ARPHRD_IEEE802_TR) &&
2324             (dev->type != ARPHRD_ARCNET) &&
2325             (dev->type != ARPHRD_INFINIBAND)) {
2326                 /* Alas, we support only Ethernet autoconfiguration. */
2327                 return;
2328         }
2329
2330         idev = addrconf_add_dev(dev);
2331         if (idev == NULL)
2332                 return;
2333
2334         memset(&addr, 0, sizeof(struct in6_addr));
2335         addr.s6_addr32[0] = htonl(0xFE800000);
2336
2337         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2338                 addrconf_add_linklocal(idev, &addr);
2339 }
2340
2341 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2342 static void addrconf_sit_config(struct net_device *dev)
2343 {
2344         struct inet6_dev *idev;
2345
2346         ASSERT_RTNL();
2347
2348         /*
2349          * Configure the tunnel with one of our IPv4
2350          * addresses... we should configure all of
2351          * our v4 addrs in the tunnel
2352          */
2353
2354         if ((idev = ipv6_find_idev(dev)) == NULL) {
2355                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2356                 return;
2357         }
2358
2359         if (dev->priv_flags & IFF_ISATAP) {
2360                 struct in6_addr addr;
2361
2362                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2363                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2364                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2365                         addrconf_add_linklocal(idev, &addr);
2366                 return;
2367         }
2368
2369         sit_add_v4_addrs(idev);
2370
2371         if (dev->flags&IFF_POINTOPOINT) {
2372                 addrconf_add_mroute(dev);
2373                 addrconf_add_lroute(dev);
2374         } else
2375                 sit_route_add(dev);
2376 }
2377 #endif
2378
2379 static inline int
2380 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2381 {
2382         struct in6_addr lladdr;
2383
2384         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2385                 addrconf_add_linklocal(idev, &lladdr);
2386                 return 0;
2387         }
2388         return -1;
2389 }
2390
2391 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2392 {
2393         struct net_device *link_dev;
2394         struct net *net = dev_net(idev->dev);
2395
2396         /* first try to inherit the link-local address from the link device */
2397         if (idev->dev->iflink &&
2398             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2399                 if (!ipv6_inherit_linklocal(idev, link_dev))
2400                         return;
2401         }
2402         /* then try to inherit it from any device */
2403         for_each_netdev(net, link_dev) {
2404                 if (!ipv6_inherit_linklocal(idev, link_dev))
2405                         return;
2406         }
2407         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2408 }
2409
2410 /*
2411  * Autoconfigure tunnel with a link-local address so routing protocols,
2412  * DHCPv6, MLD etc. can be run over the virtual link
2413  */
2414
2415 static void addrconf_ip6_tnl_config(struct net_device *dev)
2416 {
2417         struct inet6_dev *idev;
2418
2419         ASSERT_RTNL();
2420
2421         if ((idev = addrconf_add_dev(dev)) == NULL) {
2422                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2423                 return;
2424         }
2425         ip6_tnl_add_linklocal(idev);
2426 }
2427
2428 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2429                            void * data)
2430 {
2431         struct net_device *dev = (struct net_device *) data;
2432         struct inet6_dev *idev = __in6_dev_get(dev);
2433         int run_pending = 0;
2434         int err;
2435
2436         switch(event) {
2437         case NETDEV_REGISTER:
2438                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2439                         idev = ipv6_add_dev(dev);
2440                         if (!idev)
2441                                 return notifier_from_errno(-ENOMEM);
2442                 }
2443                 break;
2444         case NETDEV_UP:
2445         case NETDEV_CHANGE:
2446                 if (dev->flags & IFF_SLAVE)
2447                         break;
2448
2449                 if (event == NETDEV_UP) {
2450                         if (!addrconf_qdisc_ok(dev)) {
2451                                 /* device is not ready yet. */
2452                                 printk(KERN_INFO
2453                                         "ADDRCONF(NETDEV_UP): %s: "
2454                                         "link is not ready\n",
2455                                         dev->name);
2456                                 break;
2457                         }
2458
2459                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2460                                 idev = ipv6_add_dev(dev);
2461
2462                         if (idev) {
2463                                 idev->if_flags |= IF_READY;
2464                                 run_pending = 1;
2465                         }
2466                 } else {
2467                         if (!addrconf_qdisc_ok(dev)) {
2468                                 /* device is still not ready. */
2469                                 break;
2470                         }
2471
2472                         if (idev) {
2473                                 if (idev->if_flags & IF_READY) {
2474                                         /* device is already configured. */
2475                                         break;
2476                                 }
2477                                 idev->if_flags |= IF_READY;
2478                         }
2479
2480                         printk(KERN_INFO
2481                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2482                                         "link becomes ready\n",
2483                                         dev->name);
2484
2485                         run_pending = 1;
2486                 }
2487
2488                 switch(dev->type) {
2489 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2490                 case ARPHRD_SIT:
2491                         addrconf_sit_config(dev);
2492                         break;
2493 #endif
2494                 case ARPHRD_TUNNEL6:
2495                         addrconf_ip6_tnl_config(dev);
2496                         break;
2497                 case ARPHRD_LOOPBACK:
2498                         init_loopback(dev);
2499                         break;
2500
2501                 default:
2502                         addrconf_dev_config(dev);
2503                         break;
2504                 }
2505                 if (idev) {
2506                         if (run_pending)
2507                                 addrconf_dad_run(idev);
2508
2509                         /* If the MTU changed during the interface down, when the
2510                            interface up, the changed MTU must be reflected in the
2511                            idev as well as routers.
2512                          */
2513                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2514                                 rt6_mtu_change(dev, dev->mtu);
2515                                 idev->cnf.mtu6 = dev->mtu;
2516                         }
2517                         idev->tstamp = jiffies;
2518                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2519                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2520                            stop IPv6 on this interface.
2521                          */
2522                         if (dev->mtu < IPV6_MIN_MTU)
2523                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2524                 }
2525                 break;
2526
2527         case NETDEV_CHANGEMTU:
2528                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2529                         rt6_mtu_change(dev, dev->mtu);
2530                         idev->cnf.mtu6 = dev->mtu;
2531                         break;
2532                 }
2533
2534                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2535                         idev = ipv6_add_dev(dev);
2536                         if (idev)
2537                                 break;
2538                 }
2539
2540                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2541
2542         case NETDEV_DOWN:
2543         case NETDEV_UNREGISTER:
2544                 /*
2545                  *      Remove all addresses from this interface.
2546                  */
2547                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2548                 break;
2549
2550         case NETDEV_CHANGENAME:
2551                 if (idev) {
2552                         snmp6_unregister_dev(idev);
2553                         addrconf_sysctl_unregister(idev);
2554                         addrconf_sysctl_register(idev);
2555                         err = snmp6_register_dev(idev);
2556                         if (err)
2557                                 return notifier_from_errno(err);
2558                 }
2559                 break;
2560         }
2561
2562         return NOTIFY_OK;
2563 }
2564
2565 /*
2566  *      addrconf module should be notified of a device going up
2567  */
2568 static struct notifier_block ipv6_dev_notf = {
2569         .notifier_call = addrconf_notify,
2570         .priority = 0
2571 };
2572
2573 static int addrconf_ifdown(struct net_device *dev, int how)
2574 {
2575         struct inet6_dev *idev;
2576         struct inet6_ifaddr *ifa, **bifa;
2577         struct net *net = dev_net(dev);
2578         int i;
2579
2580         ASSERT_RTNL();
2581
2582         rt6_ifdown(net, dev);
2583         neigh_ifdown(&nd_tbl, dev);
2584
2585         idev = __in6_dev_get(dev);
2586         if (idev == NULL)
2587                 return -ENODEV;
2588
2589         /* Step 1: remove reference to ipv6 device from parent device.
2590                    Do not dev_put!
2591          */
2592         if (how) {
2593                 idev->dead = 1;
2594
2595                 /* protected by rtnl_lock */
2596                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2597
2598                 /* Step 1.5: remove snmp6 entry */
2599                 snmp6_unregister_dev(idev);
2600
2601         }
2602
2603         /* Step 2: clear hash table */
2604         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2605                 bifa = &inet6_addr_lst[i];
2606
2607                 write_lock_bh(&addrconf_hash_lock);
2608                 while ((ifa = *bifa) != NULL) {
2609                         if (ifa->idev == idev) {
2610                                 *bifa = ifa->lst_next;
2611                                 ifa->lst_next = NULL;
2612                                 addrconf_del_timer(ifa);
2613                                 in6_ifa_put(ifa);
2614                                 continue;
2615                         }
2616                         bifa = &ifa->lst_next;
2617                 }
2618                 write_unlock_bh(&addrconf_hash_lock);
2619         }
2620
2621         write_lock_bh(&idev->lock);
2622
2623         /* Step 3: clear flags for stateless addrconf */
2624         if (!how)
2625                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2626
2627         /* Step 4: clear address list */
2628 #ifdef CONFIG_IPV6_PRIVACY
2629         if (how && del_timer(&idev->regen_timer))
2630                 in6_dev_put(idev);
2631
2632         /* clear tempaddr list */
2633         while ((ifa = idev->tempaddr_list) != NULL) {
2634                 idev->tempaddr_list = ifa->tmp_next;
2635                 ifa->tmp_next = NULL;
2636                 ifa->dead = 1;
2637                 write_unlock_bh(&idev->lock);
2638                 spin_lock_bh(&ifa->lock);
2639
2640                 if (ifa->ifpub) {
2641                         in6_ifa_put(ifa->ifpub);
2642                         ifa->ifpub = NULL;
2643                 }
2644                 spin_unlock_bh(&ifa->lock);
2645                 in6_ifa_put(ifa);
2646                 write_lock_bh(&idev->lock);
2647         }
2648 #endif
2649         while ((ifa = idev->addr_list) != NULL) {
2650                 idev->addr_list = ifa->if_next;
2651                 ifa->if_next = NULL;
2652                 ifa->dead = 1;
2653                 addrconf_del_timer(ifa);
2654                 write_unlock_bh(&idev->lock);
2655
2656                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2657                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2658                 in6_ifa_put(ifa);
2659
2660                 write_lock_bh(&idev->lock);
2661         }
2662         write_unlock_bh(&idev->lock);
2663
2664         /* Step 5: Discard multicast list */
2665
2666         if (how)
2667                 ipv6_mc_destroy_dev(idev);
2668         else
2669                 ipv6_mc_down(idev);
2670
2671         idev->tstamp = jiffies;
2672
2673         /* Shot the device (if unregistered) */
2674
2675         if (how) {
2676                 addrconf_sysctl_unregister(idev);
2677                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2678                 neigh_ifdown(&nd_tbl, dev);
2679                 in6_dev_put(idev);
2680         }
2681         return 0;
2682 }
2683
2684 static void addrconf_rs_timer(unsigned long data)
2685 {
2686         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2687
2688         if (ifp->idev->cnf.forwarding)
2689                 goto out;
2690
2691         if (ifp->idev->if_flags & IF_RA_RCVD) {
2692                 /*
2693                  *      Announcement received after solicitation
2694                  *      was sent
2695                  */
2696                 goto out;
2697         }
2698
2699         spin_lock(&ifp->lock);
2700         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2701                 /* The wait after the last probe can be shorter */
2702                 addrconf_mod_timer(ifp, AC_RS,
2703                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2704                                    ifp->idev->cnf.rtr_solicit_delay :
2705                                    ifp->idev->cnf.rtr_solicit_interval);
2706                 spin_unlock(&ifp->lock);
2707
2708                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2709         } else {
2710                 spin_unlock(&ifp->lock);
2711                 /*
2712                  * Note: we do not support deprecated "all on-link"
2713                  * assumption any longer.
2714                  */
2715                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2716                        ifp->idev->dev->name);
2717         }
2718
2719 out:
2720         in6_ifa_put(ifp);
2721 }
2722
2723 /*
2724  *      Duplicate Address Detection
2725  */
2726 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2727 {
2728         unsigned long rand_num;
2729         struct inet6_dev *idev = ifp->idev;
2730
2731         if (ifp->flags & IFA_F_OPTIMISTIC)
2732                 rand_num = 0;
2733         else
2734                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2735
2736         ifp->probes = idev->cnf.dad_transmits;
2737         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2738 }
2739
2740 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2741 {
2742         struct inet6_dev *idev = ifp->idev;
2743         struct net_device *dev = idev->dev;
2744
2745         addrconf_join_solict(dev, &ifp->addr);
2746
2747         net_srandom(ifp->addr.s6_addr32[3]);
2748
2749         read_lock_bh(&idev->lock);
2750         if (ifp->dead)
2751                 goto out;
2752         spin_lock_bh(&ifp->lock);
2753
2754         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2755             idev->cnf.accept_dad < 1 ||
2756             !(ifp->flags&IFA_F_TENTATIVE) ||
2757             ifp->flags & IFA_F_NODAD) {
2758                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2759                 spin_unlock_bh(&ifp->lock);
2760                 read_unlock_bh(&idev->lock);
2761
2762                 addrconf_dad_completed(ifp);
2763                 return;
2764         }
2765
2766         if (!(idev->if_flags & IF_READY)) {
2767                 spin_unlock_bh(&ifp->lock);
2768                 read_unlock_bh(&idev->lock);
2769                 /*
2770                  * If the defice is not ready:
2771                  * - keep it tentative if it is a permanent address.
2772                  * - otherwise, kill it.
2773                  */
2774                 in6_ifa_hold(ifp);
2775                 addrconf_dad_stop(ifp);
2776                 return;
2777         }
2778
2779         /*
2780          * Optimistic nodes can start receiving
2781          * Frames right away
2782          */
2783         if(ifp->flags & IFA_F_OPTIMISTIC)
2784                 ip6_ins_rt(ifp->rt);
2785
2786         addrconf_dad_kick(ifp);
2787         spin_unlock_bh(&ifp->lock);
2788 out:
2789         read_unlock_bh(&idev->lock);
2790 }
2791
2792 static void addrconf_dad_timer(unsigned long data)
2793 {
2794         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2795         struct inet6_dev *idev = ifp->idev;
2796         struct in6_addr mcaddr;
2797
2798         read_lock_bh(&idev->lock);
2799         if (idev->dead) {
2800                 read_unlock_bh(&idev->lock);
2801                 goto out;
2802         }
2803         spin_lock_bh(&ifp->lock);
2804         if (ifp->probes == 0) {
2805                 /*
2806                  * DAD was successful
2807                  */
2808
2809                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2810                 spin_unlock_bh(&ifp->lock);
2811                 read_unlock_bh(&idev->lock);
2812
2813                 addrconf_dad_completed(ifp);
2814
2815                 goto out;
2816         }
2817
2818         ifp->probes--;
2819         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2820         spin_unlock_bh(&ifp->lock);
2821         read_unlock_bh(&idev->lock);
2822
2823         /* send a neighbour solicitation for our addr */
2824         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2825         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2826 out:
2827         in6_ifa_put(ifp);
2828 }
2829
2830 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2831 {
2832         struct net_device *     dev = ifp->idev->dev;
2833
2834         /*
2835          *      Configure the address for reception. Now it is valid.
2836          */
2837
2838         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2839
2840         /* If added prefix is link local and forwarding is off,
2841            start sending router solicitations.
2842          */
2843
2844         if (ifp->idev->cnf.forwarding == 0 &&
2845             ifp->idev->cnf.rtr_solicits > 0 &&
2846             (dev->flags&IFF_LOOPBACK) == 0 &&
2847             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2848                 /*
2849                  *      If a host as already performed a random delay
2850                  *      [...] as part of DAD [...] there is no need
2851                  *      to delay again before sending the first RS
2852                  */
2853                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2854
2855                 spin_lock_bh(&ifp->lock);
2856                 ifp->probes = 1;
2857                 ifp->idev->if_flags |= IF_RS_SENT;
2858                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2859                 spin_unlock_bh(&ifp->lock);
2860         }
2861 }
2862
2863 static void addrconf_dad_run(struct inet6_dev *idev) {
2864         struct inet6_ifaddr *ifp;
2865
2866         read_lock_bh(&idev->lock);
2867         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2868                 spin_lock_bh(&ifp->lock);
2869                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2870                         spin_unlock_bh(&ifp->lock);
2871                         continue;
2872                 }
2873                 spin_unlock_bh(&ifp->lock);
2874                 addrconf_dad_kick(ifp);
2875         }
2876         read_unlock_bh(&idev->lock);
2877 }
2878
2879 #ifdef CONFIG_PROC_FS
2880 struct if6_iter_state {
2881         struct seq_net_private p;
2882         int bucket;
2883 };
2884
2885 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2886 {
2887         struct inet6_ifaddr *ifa = NULL;
2888         struct if6_iter_state *state = seq->private;
2889         struct net *net = seq_file_net(seq);
2890
2891         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2892                 ifa = inet6_addr_lst[state->bucket];
2893
2894                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2895                         ifa = ifa->lst_next;
2896                 if (ifa)
2897                         break;
2898         }
2899         return ifa;
2900 }
2901
2902 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2903 {
2904         struct if6_iter_state *state = seq->private;
2905         struct net *net = seq_file_net(seq);
2906
2907         ifa = ifa->lst_next;
2908 try_again:
2909         if (ifa) {
2910                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2911                         ifa = ifa->lst_next;
2912                         goto try_again;
2913                 }
2914         }
2915
2916         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2917                 ifa = inet6_addr_lst[state->bucket];
2918                 goto try_again;
2919         }
2920
2921         return ifa;
2922 }
2923
2924 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2925 {
2926         struct inet6_ifaddr *ifa = if6_get_first(seq);
2927
2928         if (ifa)
2929                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2930                         --pos;
2931         return pos ? NULL : ifa;
2932 }
2933
2934 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2935         __acquires(addrconf_hash_lock)
2936 {
2937         read_lock_bh(&addrconf_hash_lock);
2938         return if6_get_idx(seq, *pos);
2939 }
2940
2941 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2942 {
2943         struct inet6_ifaddr *ifa;
2944
2945         ifa = if6_get_next(seq, v);
2946         ++*pos;
2947         return ifa;
2948 }
2949
2950 static void if6_seq_stop(struct seq_file *seq, void *v)
2951         __releases(addrconf_hash_lock)
2952 {
2953         read_unlock_bh(&addrconf_hash_lock);
2954 }
2955
2956 static int if6_seq_show(struct seq_file *seq, void *v)
2957 {
2958         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2959         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2960                    &ifp->addr,
2961                    ifp->idev->dev->ifindex,
2962                    ifp->prefix_len,
2963                    ifp->scope,
2964                    ifp->flags,
2965                    ifp->idev->dev->name);
2966         return 0;
2967 }
2968
2969 static const struct seq_operations if6_seq_ops = {
2970         .start  = if6_seq_start,
2971         .next   = if6_seq_next,
2972         .show   = if6_seq_show,
2973         .stop   = if6_seq_stop,
2974 };
2975
2976 static int if6_seq_open(struct inode *inode, struct file *file)
2977 {
2978         return seq_open_net(inode, file, &if6_seq_ops,
2979                             sizeof(struct if6_iter_state));
2980 }
2981
2982 static const struct file_operations if6_fops = {
2983         .owner          = THIS_MODULE,
2984         .open           = if6_seq_open,
2985         .read           = seq_read,
2986         .llseek         = seq_lseek,
2987         .release        = seq_release_net,
2988 };
2989
2990 static int if6_proc_net_init(struct net *net)
2991 {
2992         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
2993                 return -ENOMEM;
2994         return 0;
2995 }
2996
2997 static void if6_proc_net_exit(struct net *net)
2998 {
2999        proc_net_remove(net, "if_inet6");
3000 }
3001
3002 static struct pernet_operations if6_proc_net_ops = {
3003        .init = if6_proc_net_init,
3004        .exit = if6_proc_net_exit,
3005 };
3006
3007 int __init if6_proc_init(void)
3008 {
3009         return register_pernet_subsys(&if6_proc_net_ops);
3010 }
3011
3012 void if6_proc_exit(void)
3013 {
3014         unregister_pernet_subsys(&if6_proc_net_ops);
3015 }
3016 #endif  /* CONFIG_PROC_FS */
3017
3018 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3019 /* Check if address is a home address configured on any interface. */
3020 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3021 {
3022         int ret = 0;
3023         struct inet6_ifaddr * ifp;
3024         u8 hash = ipv6_addr_hash(addr);
3025         read_lock_bh(&addrconf_hash_lock);
3026         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3027                 if (!net_eq(dev_net(ifp->idev->dev), net))
3028                         continue;
3029                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3030                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3031                         ret = 1;
3032                         break;
3033                 }
3034         }
3035         read_unlock_bh(&addrconf_hash_lock);
3036         return ret;
3037 }
3038 #endif
3039
3040 /*
3041  *      Periodic address status verification
3042  */
3043
3044 static void addrconf_verify(unsigned long foo)
3045 {
3046         struct inet6_ifaddr *ifp;
3047         unsigned long now, next;
3048         int i;
3049
3050         spin_lock_bh(&addrconf_verify_lock);
3051         now = jiffies;
3052         next = now + ADDR_CHECK_FREQUENCY;
3053
3054         del_timer(&addr_chk_timer);
3055
3056         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3057
3058 restart:
3059                 read_lock(&addrconf_hash_lock);
3060                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3061                         unsigned long age;
3062 #ifdef CONFIG_IPV6_PRIVACY
3063                         unsigned long regen_advance;
3064 #endif
3065
3066                         if (ifp->flags & IFA_F_PERMANENT)
3067                                 continue;
3068
3069                         spin_lock(&ifp->lock);
3070                         age = (now - ifp->tstamp) / HZ;
3071
3072 #ifdef CONFIG_IPV6_PRIVACY
3073                         regen_advance = ifp->idev->cnf.regen_max_retry *
3074                                         ifp->idev->cnf.dad_transmits *
3075                                         ifp->idev->nd_parms->retrans_time / HZ;
3076 #endif
3077
3078                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3079                             age >= ifp->valid_lft) {
3080                                 spin_unlock(&ifp->lock);
3081                                 in6_ifa_hold(ifp);
3082                                 read_unlock(&addrconf_hash_lock);
3083                                 ipv6_del_addr(ifp);
3084                                 goto restart;
3085                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3086                                 spin_unlock(&ifp->lock);
3087                                 continue;
3088                         } else if (age >= ifp->prefered_lft) {
3089                                 /* jiffies - ifp->tsamp > age >= ifp->prefered_lft */
3090                                 int deprecate = 0;
3091
3092                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3093                                         deprecate = 1;
3094                                         ifp->flags |= IFA_F_DEPRECATED;
3095                                 }
3096
3097                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3098                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3099
3100                                 spin_unlock(&ifp->lock);
3101
3102                                 if (deprecate) {
3103                                         in6_ifa_hold(ifp);
3104                                         read_unlock(&addrconf_hash_lock);
3105
3106                                         ipv6_ifa_notify(0, ifp);
3107                                         in6_ifa_put(ifp);
3108                                         goto restart;
3109                                 }
3110 #ifdef CONFIG_IPV6_PRIVACY
3111                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3112                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3113                                 if (age >= ifp->prefered_lft - regen_advance) {
3114                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3115                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3116                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3117                                         if (!ifp->regen_count && ifpub) {
3118                                                 ifp->regen_count++;
3119                                                 in6_ifa_hold(ifp);
3120                                                 in6_ifa_hold(ifpub);
3121                                                 spin_unlock(&ifp->lock);
3122                                                 read_unlock(&addrconf_hash_lock);
3123                                                 spin_lock(&ifpub->lock);
3124                                                 ifpub->regen_count = 0;
3125                                                 spin_unlock(&ifpub->lock);
3126                                                 ipv6_create_tempaddr(ifpub, ifp);
3127                                                 in6_ifa_put(ifpub);
3128                                                 in6_ifa_put(ifp);
3129                                                 goto restart;
3130                                         }
3131                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3132                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3133                                 spin_unlock(&ifp->lock);
3134 #endif
3135                         } else {
3136                                 /* ifp->prefered_lft <= ifp->valid_lft */
3137                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3138                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3139                                 spin_unlock(&ifp->lock);
3140                         }
3141                 }
3142                 read_unlock(&addrconf_hash_lock);
3143         }
3144
3145         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3146         add_timer(&addr_chk_timer);
3147         spin_unlock_bh(&addrconf_verify_lock);
3148 }
3149
3150 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3151 {
3152         struct in6_addr *pfx = NULL;
3153
3154         if (addr)
3155                 pfx = nla_data(addr);
3156
3157         if (local) {
3158                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3159                         pfx = NULL;
3160                 else
3161                         pfx = nla_data(local);
3162         }
3163
3164         return pfx;
3165 }
3166
3167 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3168         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3169         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3170         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3171 };
3172
3173 static int
3174 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3175 {
3176         struct net *net = sock_net(skb->sk);
3177         struct ifaddrmsg *ifm;
3178         struct nlattr *tb[IFA_MAX+1];
3179         struct in6_addr *pfx;
3180         int err;
3181
3182         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3183         if (err < 0)
3184                 return err;
3185
3186         ifm = nlmsg_data(nlh);
3187         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3188         if (pfx == NULL)
3189                 return -EINVAL;
3190
3191         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3192 }
3193
3194 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3195                              u32 prefered_lft, u32 valid_lft)
3196 {
3197         u32 flags;
3198         clock_t expires;
3199         unsigned long timeout;
3200
3201         if (!valid_lft || (prefered_lft > valid_lft))
3202                 return -EINVAL;
3203
3204         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3205         if (addrconf_finite_timeout(timeout)) {
3206                 expires = jiffies_to_clock_t(timeout * HZ);
3207                 valid_lft = timeout;
3208                 flags = RTF_EXPIRES;
3209         } else {
3210                 expires = 0;
3211                 flags = 0;
3212                 ifa_flags |= IFA_F_PERMANENT;
3213         }
3214
3215         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3216         if (addrconf_finite_timeout(timeout)) {
3217                 if (timeout == 0)
3218                         ifa_flags |= IFA_F_DEPRECATED;
3219                 prefered_lft = timeout;
3220         }
3221
3222         spin_lock_bh(&ifp->lock);
3223         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3224         ifp->tstamp = jiffies;
3225         ifp->valid_lft = valid_lft;
3226         ifp->prefered_lft = prefered_lft;
3227
3228         spin_unlock_bh(&ifp->lock);
3229         if (!(ifp->flags&IFA_F_TENTATIVE))
3230                 ipv6_ifa_notify(0, ifp);
3231
3232         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3233                               expires, flags);
3234         addrconf_verify(0);
3235
3236         return 0;
3237 }
3238
3239 static int
3240 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3241 {
3242         struct net *net = sock_net(skb->sk);
3243         struct ifaddrmsg *ifm;
3244         struct nlattr *tb[IFA_MAX+1];
3245         struct in6_addr *pfx;
3246         struct inet6_ifaddr *ifa;
3247         struct net_device *dev;
3248         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3249         u8 ifa_flags;
3250         int err;
3251
3252         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3253         if (err < 0)
3254                 return err;
3255
3256         ifm = nlmsg_data(nlh);
3257         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3258         if (pfx == NULL)
3259                 return -EINVAL;
3260
3261         if (tb[IFA_CACHEINFO]) {
3262                 struct ifa_cacheinfo *ci;
3263
3264                 ci = nla_data(tb[IFA_CACHEINFO]);
3265                 valid_lft = ci->ifa_valid;
3266                 preferred_lft = ci->ifa_prefered;
3267         } else {
3268                 preferred_lft = INFINITY_LIFE_TIME;
3269                 valid_lft = INFINITY_LIFE_TIME;
3270         }
3271
3272         dev =  __dev_get_by_index(net, ifm->ifa_index);
3273         if (dev == NULL)
3274                 return -ENODEV;
3275
3276         /* We ignore other flags so far. */
3277         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3278
3279         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3280         if (ifa == NULL) {
3281                 /*
3282                  * It would be best to check for !NLM_F_CREATE here but
3283                  * userspace alreay relies on not having to provide this.
3284                  */
3285                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3286                                       ifm->ifa_prefixlen, ifa_flags,
3287                                       preferred_lft, valid_lft);
3288         }
3289
3290         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3291             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3292                 err = -EEXIST;
3293         else
3294                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3295
3296         in6_ifa_put(ifa);
3297
3298         return err;
3299 }
3300
3301 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3302                           u8 scope, int ifindex)
3303 {
3304         struct ifaddrmsg *ifm;
3305
3306         ifm = nlmsg_data(nlh);
3307         ifm->ifa_family = AF_INET6;
3308         ifm->ifa_prefixlen = prefixlen;
3309         ifm->ifa_flags = flags;
3310         ifm->ifa_scope = scope;
3311         ifm->ifa_index = ifindex;
3312 }
3313
3314 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3315                          unsigned long tstamp, u32 preferred, u32 valid)
3316 {
3317         struct ifa_cacheinfo ci;
3318
3319         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3320                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3321         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3322                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3323         ci.ifa_prefered = preferred;
3324         ci.ifa_valid = valid;
3325
3326         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3327 }
3328
3329 static inline int rt_scope(int ifa_scope)
3330 {
3331         if (ifa_scope & IFA_HOST)
3332                 return RT_SCOPE_HOST;
3333         else if (ifa_scope & IFA_LINK)
3334                 return RT_SCOPE_LINK;
3335         else if (ifa_scope & IFA_SITE)
3336                 return RT_SCOPE_SITE;
3337         else
3338                 return RT_SCOPE_UNIVERSE;
3339 }
3340
3341 static inline int inet6_ifaddr_msgsize(void)
3342 {
3343         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3344                + nla_total_size(16) /* IFA_ADDRESS */
3345                + nla_total_size(sizeof(struct ifa_cacheinfo));
3346 }
3347
3348 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3349                              u32 pid, u32 seq, int event, unsigned int flags)
3350 {
3351         struct nlmsghdr  *nlh;
3352         u32 preferred, valid;
3353
3354         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3355         if (nlh == NULL)
3356                 return -EMSGSIZE;
3357
3358         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3359                       ifa->idev->dev->ifindex);
3360
3361         if (!(ifa->flags&IFA_F_PERMANENT)) {
3362                 preferred = ifa->prefered_lft;
3363                 valid = ifa->valid_lft;
3364                 if (preferred != INFINITY_LIFE_TIME) {
3365                         long tval = (jiffies - ifa->tstamp)/HZ;
3366                         preferred -= tval;
3367                         if (valid != INFINITY_LIFE_TIME)
3368                                 valid -= tval;
3369                 }
3370         } else {
3371                 preferred = INFINITY_LIFE_TIME;
3372                 valid = INFINITY_LIFE_TIME;
3373         }
3374
3375         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3376             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3377                 nlmsg_cancel(skb, nlh);
3378                 return -EMSGSIZE;
3379         }
3380
3381         return nlmsg_end(skb, nlh);
3382 }
3383
3384 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3385                                 u32 pid, u32 seq, int event, u16 flags)
3386 {
3387         struct nlmsghdr  *nlh;
3388         u8 scope = RT_SCOPE_UNIVERSE;
3389         int ifindex = ifmca->idev->dev->ifindex;
3390
3391         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3392                 scope = RT_SCOPE_SITE;
3393
3394         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3395         if (nlh == NULL)
3396                 return -EMSGSIZE;
3397
3398         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3399         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3400             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3401                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3402                 nlmsg_cancel(skb, nlh);
3403                 return -EMSGSIZE;
3404         }
3405
3406         return nlmsg_end(skb, nlh);
3407 }
3408
3409 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3410                                 u32 pid, u32 seq, int event, unsigned int flags)
3411 {
3412         struct nlmsghdr  *nlh;
3413         u8 scope = RT_SCOPE_UNIVERSE;
3414         int ifindex = ifaca->aca_idev->dev->ifindex;
3415
3416         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3417                 scope = RT_SCOPE_SITE;
3418
3419         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3420         if (nlh == NULL)
3421                 return -EMSGSIZE;
3422
3423         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3424         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3425             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3426                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3427                 nlmsg_cancel(skb, nlh);
3428                 return -EMSGSIZE;
3429         }
3430
3431         return nlmsg_end(skb, nlh);
3432 }
3433
3434 enum addr_type_t
3435 {
3436         UNICAST_ADDR,
3437         MULTICAST_ADDR,
3438         ANYCAST_ADDR,
3439 };
3440
3441 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3442                            enum addr_type_t type)
3443 {
3444         int idx, ip_idx;
3445         int s_idx, s_ip_idx;
3446         int err = 1;
3447         struct net_device *dev;
3448         struct inet6_dev *idev = NULL;
3449         struct inet6_ifaddr *ifa;
3450         struct ifmcaddr6 *ifmca;
3451         struct ifacaddr6 *ifaca;
3452         struct net *net = sock_net(skb->sk);
3453
3454         s_idx = cb->args[0];
3455         s_ip_idx = ip_idx = cb->args[1];
3456
3457         idx = 0;
3458         for_each_netdev(net, dev) {
3459                 if (idx < s_idx)
3460                         goto cont;
3461                 if (idx > s_idx)
3462                         s_ip_idx = 0;
3463                 ip_idx = 0;
3464                 if ((idev = in6_dev_get(dev)) == NULL)
3465                         goto cont;
3466                 read_lock_bh(&idev->lock);
3467                 switch (type) {
3468                 case UNICAST_ADDR:
3469                         /* unicast address incl. temp addr */
3470                         for (ifa = idev->addr_list; ifa;
3471                              ifa = ifa->if_next, ip_idx++) {
3472                                 if (ip_idx < s_ip_idx)
3473                                         continue;
3474                                 err = inet6_fill_ifaddr(skb, ifa,
3475                                                         NETLINK_CB(cb->skb).pid,
3476                                                         cb->nlh->nlmsg_seq,
3477                                                         RTM_NEWADDR,
3478                                                         NLM_F_MULTI);
3479                         }
3480                         break;
3481                 case MULTICAST_ADDR:
3482                         /* multicast address */
3483                         for (ifmca = idev->mc_list; ifmca;
3484                              ifmca = ifmca->next, ip_idx++) {
3485                                 if (ip_idx < s_ip_idx)
3486                                         continue;
3487                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3488                                                           NETLINK_CB(cb->skb).pid,
3489                                                           cb->nlh->nlmsg_seq,
3490                                                           RTM_GETMULTICAST,
3491                                                           NLM_F_MULTI);
3492                         }
3493                         break;
3494                 case ANYCAST_ADDR:
3495                         /* anycast address */
3496                         for (ifaca = idev->ac_list; ifaca;
3497                              ifaca = ifaca->aca_next, ip_idx++) {
3498                                 if (ip_idx < s_ip_idx)
3499                                         continue;
3500                                 err = inet6_fill_ifacaddr(skb, ifaca,
3501                                                           NETLINK_CB(cb->skb).pid,
3502                                                           cb->nlh->nlmsg_seq,
3503                                                           RTM_GETANYCAST,
3504                                                           NLM_F_MULTI);
3505                         }
3506                         break;
3507                 default:
3508                         break;
3509                 }
3510                 read_unlock_bh(&idev->lock);
3511                 in6_dev_put(idev);
3512
3513                 if (err <= 0)
3514                         break;
3515 cont:
3516                 idx++;
3517         }
3518         cb->args[0] = idx;
3519         cb->args[1] = ip_idx;
3520         return skb->len;
3521 }
3522
3523 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3524 {
3525         enum addr_type_t type = UNICAST_ADDR;
3526
3527         return inet6_dump_addr(skb, cb, type);
3528 }
3529
3530 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3531 {
3532         enum addr_type_t type = MULTICAST_ADDR;
3533
3534         return inet6_dump_addr(skb, cb, type);
3535 }
3536
3537
3538 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3539 {
3540         enum addr_type_t type = ANYCAST_ADDR;
3541
3542         return inet6_dump_addr(skb, cb, type);
3543 }
3544
3545 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3546                              void *arg)
3547 {
3548         struct net *net = sock_net(in_skb->sk);
3549         struct ifaddrmsg *ifm;
3550         struct nlattr *tb[IFA_MAX+1];
3551         struct in6_addr *addr = NULL;
3552         struct net_device *dev = NULL;
3553         struct inet6_ifaddr *ifa;
3554         struct sk_buff *skb;
3555         int err;
3556
3557         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3558         if (err < 0)
3559                 goto errout;
3560
3561         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3562         if (addr == NULL) {
3563                 err = -EINVAL;
3564                 goto errout;
3565         }
3566
3567         ifm = nlmsg_data(nlh);
3568         if (ifm->ifa_index)
3569                 dev = __dev_get_by_index(net, ifm->ifa_index);
3570
3571         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3572                 err = -EADDRNOTAVAIL;
3573                 goto errout;
3574         }
3575
3576         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3577                 err = -ENOBUFS;
3578                 goto errout_ifa;
3579         }
3580
3581         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3582                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3583         if (err < 0) {
3584                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3585                 WARN_ON(err == -EMSGSIZE);
3586                 kfree_skb(skb);
3587                 goto errout_ifa;
3588         }
3589         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3590 errout_ifa:
3591         in6_ifa_put(ifa);
3592 errout:
3593         return err;
3594 }
3595
3596 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3597 {
3598         struct sk_buff *skb;
3599         struct net *net = dev_net(ifa->idev->dev);
3600         int err = -ENOBUFS;
3601
3602         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3603         if (skb == NULL)
3604                 goto errout;
3605
3606         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3607         if (err < 0) {
3608                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3609                 WARN_ON(err == -EMSGSIZE);
3610                 kfree_skb(skb);
3611                 goto errout;
3612         }
3613         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3614         return;
3615 errout:
3616         if (err < 0)
3617                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3618 }
3619
3620 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3621                                 __s32 *array, int bytes)
3622 {
3623         BUG_ON(bytes < (DEVCONF_MAX * 4));
3624
3625         memset(array, 0, bytes);
3626         array[DEVCONF_FORWARDING] = cnf->forwarding;
3627         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3628         array[DEVCONF_MTU6] = cnf->mtu6;
3629         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3630         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3631         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3632         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3633         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3634         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3635         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3636         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3637 #ifdef CONFIG_IPV6_PRIVACY
3638         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3639         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3640         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3641         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3642         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3643 #endif
3644         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3645         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3646         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3647 #ifdef CONFIG_IPV6_ROUTER_PREF
3648         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3649         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3650 #ifdef CONFIG_IPV6_ROUTE_INFO
3651         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3652 #endif
3653 #endif
3654         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3655         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3656 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3657         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3658 #endif
3659 #ifdef CONFIG_IPV6_MROUTE
3660         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3661 #endif
3662         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3663         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3664 }
3665
3666 static inline size_t inet6_if_nlmsg_size(void)
3667 {
3668         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3669                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3670                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3671                + nla_total_size(4) /* IFLA_MTU */
3672                + nla_total_size(4) /* IFLA_LINK */
3673                + nla_total_size( /* IFLA_PROTINFO */
3674                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3675                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3676                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3677                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3678                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3679                  );
3680 }
3681
3682 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3683                                       int bytes)
3684 {
3685         int i;
3686         int pad = bytes - sizeof(u64) * items;
3687         BUG_ON(pad < 0);
3688
3689         /* Use put_unaligned() because stats may not be aligned for u64. */
3690         put_unaligned(items, &stats[0]);
3691         for (i = 1; i < items; i++)
3692                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3693
3694         memset(&stats[items], 0, pad);
3695 }
3696
3697 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3698                              int bytes)
3699 {
3700         switch(attrtype) {
3701         case IFLA_INET6_STATS:
3702                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3703                 break;
3704         case IFLA_INET6_ICMP6STATS:
3705                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3706                 break;
3707         }
3708 }
3709
3710 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3711                              u32 pid, u32 seq, int event, unsigned int flags)
3712 {
3713         struct net_device *dev = idev->dev;
3714         struct nlattr *nla;
3715         struct ifinfomsg *hdr;
3716         struct nlmsghdr *nlh;
3717         void *protoinfo;
3718         struct ifla_cacheinfo ci;
3719
3720         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3721         if (nlh == NULL)
3722                 return -EMSGSIZE;
3723
3724         hdr = nlmsg_data(nlh);
3725         hdr->ifi_family = AF_INET6;
3726         hdr->__ifi_pad = 0;
3727         hdr->ifi_type = dev->type;
3728         hdr->ifi_index = dev->ifindex;
3729         hdr->ifi_flags = dev_get_flags(dev);
3730         hdr->ifi_change = 0;
3731
3732         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3733
3734         if (dev->addr_len)
3735                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3736
3737         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3738         if (dev->ifindex != dev->iflink)
3739                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3740
3741         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3742         if (protoinfo == NULL)
3743                 goto nla_put_failure;
3744
3745         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3746
3747         ci.max_reasm_len = IPV6_MAXPLEN;
3748         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3749                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3750         ci.reachable_time = idev->nd_parms->reachable_time;
3751         ci.retrans_time = idev->nd_parms->retrans_time;
3752         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3753
3754         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3755         if (nla == NULL)
3756                 goto nla_put_failure;
3757         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3758
3759         /* XXX - MC not implemented */
3760
3761         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3762         if (nla == NULL)
3763                 goto nla_put_failure;
3764         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3765
3766         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3767         if (nla == NULL)
3768                 goto nla_put_failure;
3769         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3770
3771         nla_nest_end(skb, protoinfo);
3772         return nlmsg_end(skb, nlh);
3773
3774 nla_put_failure:
3775         nlmsg_cancel(skb, nlh);
3776         return -EMSGSIZE;
3777 }
3778
3779 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3780 {
3781         struct net *net = sock_net(skb->sk);
3782         int idx, err;
3783         int s_idx = cb->args[0];
3784         struct net_device *dev;
3785         struct inet6_dev *idev;
3786
3787         read_lock(&dev_base_lock);
3788         idx = 0;
3789         for_each_netdev(net, dev) {
3790                 if (idx < s_idx)
3791                         goto cont;
3792                 if ((idev = in6_dev_get(dev)) == NULL)
3793                         goto cont;
3794                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3795                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3796                 in6_dev_put(idev);
3797                 if (err <= 0)
3798                         break;
3799 cont:
3800                 idx++;
3801         }
3802         read_unlock(&dev_base_lock);
3803         cb->args[0] = idx;
3804
3805         return skb->len;
3806 }
3807
3808 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3809 {
3810         struct sk_buff *skb;
3811         struct net *net = dev_net(idev->dev);
3812         int err = -ENOBUFS;
3813
3814         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3815         if (skb == NULL)
3816                 goto errout;
3817
3818         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3819         if (err < 0) {
3820                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3821                 WARN_ON(err == -EMSGSIZE);
3822                 kfree_skb(skb);
3823                 goto errout;
3824         }
3825         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3826         return;
3827 errout:
3828         if (err < 0)
3829                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3830 }
3831
3832 static inline size_t inet6_prefix_nlmsg_size(void)
3833 {
3834         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3835                + nla_total_size(sizeof(struct in6_addr))
3836                + nla_total_size(sizeof(struct prefix_cacheinfo));
3837 }
3838
3839 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3840                              struct prefix_info *pinfo, u32 pid, u32 seq,
3841                              int event, unsigned int flags)
3842 {
3843         struct prefixmsg *pmsg;
3844         struct nlmsghdr *nlh;
3845         struct prefix_cacheinfo ci;
3846
3847         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3848         if (nlh == NULL)
3849                 return -EMSGSIZE;
3850
3851         pmsg = nlmsg_data(nlh);
3852         pmsg->prefix_family = AF_INET6;
3853         pmsg->prefix_pad1 = 0;
3854         pmsg->prefix_pad2 = 0;
3855         pmsg->prefix_ifindex = idev->dev->ifindex;
3856         pmsg->prefix_len = pinfo->prefix_len;
3857         pmsg->prefix_type = pinfo->type;
3858         pmsg->prefix_pad3 = 0;
3859         pmsg->prefix_flags = 0;
3860         if (pinfo->onlink)
3861                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3862         if (pinfo->autoconf)
3863                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3864
3865         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3866
3867         ci.preferred_time = ntohl(pinfo->prefered);
3868         ci.valid_time = ntohl(pinfo->valid);
3869         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3870
3871         return nlmsg_end(skb, nlh);
3872
3873 nla_put_failure:
3874         nlmsg_cancel(skb, nlh);
3875         return -EMSGSIZE;
3876 }
3877
3878 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3879                          struct prefix_info *pinfo)
3880 {
3881         struct sk_buff *skb;
3882         struct net *net = dev_net(idev->dev);
3883         int err = -ENOBUFS;
3884
3885         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3886         if (skb == NULL)
3887                 goto errout;
3888
3889         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3890         if (err < 0) {
3891                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3892                 WARN_ON(err == -EMSGSIZE);
3893                 kfree_skb(skb);
3894                 goto errout;
3895         }
3896         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3897         return;
3898 errout:
3899         if (err < 0)
3900                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3901 }
3902
3903 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3904 {
3905         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3906
3907         switch (event) {
3908         case RTM_NEWADDR:
3909                 /*
3910                  * If the address was optimistic
3911                  * we inserted the route at the start of
3912                  * our DAD process, so we don't need
3913                  * to do it again
3914                  */
3915                 if (!(ifp->rt->rt6i_node))
3916                         ip6_ins_rt(ifp->rt);
3917                 if (ifp->idev->cnf.forwarding)
3918                         addrconf_join_anycast(ifp);
3919                 break;
3920         case RTM_DELADDR:
3921                 if (ifp->idev->cnf.forwarding)
3922                         addrconf_leave_anycast(ifp);
3923                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3924                 dst_hold(&ifp->rt->u.dst);
3925                 if (ip6_del_rt(ifp->rt))
3926                         dst_free(&ifp->rt->u.dst);
3927                 break;
3928         }
3929 }
3930
3931 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3932 {
3933         rcu_read_lock_bh();
3934         if (likely(ifp->idev->dead == 0))
3935                 __ipv6_ifa_notify(event, ifp);
3936         rcu_read_unlock_bh();
3937 }
3938
3939 #ifdef CONFIG_SYSCTL
3940
3941 static
3942 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3943                            void __user *buffer, size_t *lenp, loff_t *ppos)
3944 {
3945         int *valp = ctl->data;
3946         int val = *valp;
3947         int ret;
3948
3949         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3950
3951         if (write)
3952                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3953         return ret;
3954 }
3955
3956 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3957                                             void __user *oldval,
3958                                             size_t __user *oldlenp,
3959                                             void __user *newval, size_t newlen)
3960 {
3961         int *valp = table->data;
3962         int val = *valp;
3963         int new;
3964
3965         if (!newval || !newlen)
3966                 return 0;
3967         if (newlen != sizeof(int))
3968                 return -EINVAL;
3969         if (get_user(new, (int __user *)newval))
3970                 return -EFAULT;
3971         if (new == *valp)
3972                 return 0;
3973         if (oldval && oldlenp) {
3974                 size_t len;
3975                 if (get_user(len, oldlenp))
3976                         return -EFAULT;
3977                 if (len) {
3978                         if (len > table->maxlen)
3979                                 len = table->maxlen;
3980                         if (copy_to_user(oldval, valp, len))
3981                                 return -EFAULT;
3982                         if (put_user(len, oldlenp))
3983                                 return -EFAULT;
3984                 }
3985         }
3986
3987         *valp = new;
3988         return addrconf_fixup_forwarding(table, valp, val);
3989 }
3990
3991 static struct addrconf_sysctl_table
3992 {
3993         struct ctl_table_header *sysctl_header;
3994         ctl_table addrconf_vars[DEVCONF_MAX+1];
3995         char *dev_name;
3996 } addrconf_sysctl __read_mostly = {
3997         .sysctl_header = NULL,
3998         .addrconf_vars = {
3999                 {
4000                         .ctl_name       =       NET_IPV6_FORWARDING,
4001                         .procname       =       "forwarding",
4002                         .data           =       &ipv6_devconf.forwarding,
4003                         .maxlen         =       sizeof(int),
4004                         .mode           =       0644,
4005                         .proc_handler   =       addrconf_sysctl_forward,
4006                         .strategy       =       addrconf_sysctl_forward_strategy,
4007                 },
4008                 {
4009                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4010                         .procname       =       "hop_limit",
4011                         .data           =       &ipv6_devconf.hop_limit,
4012                         .maxlen         =       sizeof(int),
4013                         .mode           =       0644,
4014                         .proc_handler   =       proc_dointvec,
4015                 },
4016                 {
4017                         .ctl_name       =       NET_IPV6_MTU,
4018                         .procname       =       "mtu",
4019                         .data           =       &ipv6_devconf.mtu6,
4020                         .maxlen         =       sizeof(int),
4021                         .mode           =       0644,
4022                         .proc_handler   =       proc_dointvec,
4023                 },
4024                 {
4025                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4026                         .procname       =       "accept_ra",
4027                         .data           =       &ipv6_devconf.accept_ra,
4028                         .maxlen         =       sizeof(int),
4029                         .mode           =       0644,
4030                         .proc_handler   =       proc_dointvec,
4031                 },
4032                 {
4033                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4034                         .procname       =       "accept_redirects",
4035                         .data           =       &ipv6_devconf.accept_redirects,
4036                         .maxlen         =       sizeof(int),
4037                         .mode           =       0644,
4038                         .proc_handler   =       proc_dointvec,
4039                 },
4040                 {
4041                         .ctl_name       =       NET_IPV6_AUTOCONF,
4042                         .procname       =       "autoconf",
4043                         .data           =       &ipv6_devconf.autoconf,
4044                         .maxlen         =       sizeof(int),
4045                         .mode           =       0644,
4046                         .proc_handler   =       proc_dointvec,
4047                 },
4048                 {
4049                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4050                         .procname       =       "dad_transmits",
4051                         .data           =       &ipv6_devconf.dad_transmits,
4052                         .maxlen         =       sizeof(int),
4053                         .mode           =       0644,
4054                         .proc_handler   =       proc_dointvec,
4055                 },
4056                 {
4057                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4058                         .procname       =       "router_solicitations",
4059                         .data           =       &ipv6_devconf.rtr_solicits,
4060                         .maxlen         =       sizeof(int),
4061                         .mode           =       0644,
4062                         .proc_handler   =       proc_dointvec,
4063                 },
4064                 {
4065                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4066                         .procname       =       "router_solicitation_interval",
4067                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4068                         .maxlen         =       sizeof(int),
4069                         .mode           =       0644,
4070                         .proc_handler   =       proc_dointvec_jiffies,
4071                         .strategy       =       sysctl_jiffies,
4072                 },
4073                 {
4074                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4075                         .procname       =       "router_solicitation_delay",
4076                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4077                         .maxlen         =       sizeof(int),
4078                         .mode           =       0644,
4079                         .proc_handler   =       proc_dointvec_jiffies,
4080                         .strategy       =       sysctl_jiffies,
4081                 },
4082                 {
4083                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4084                         .procname       =       "force_mld_version",
4085                         .data           =       &ipv6_devconf.force_mld_version,
4086                         .maxlen         =       sizeof(int),
4087                         .mode           =       0644,
4088                         .proc_handler   =       proc_dointvec,
4089                 },
4090 #ifdef CONFIG_IPV6_PRIVACY
4091                 {
4092                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4093                         .procname       =       "use_tempaddr",
4094                         .data           =       &ipv6_devconf.use_tempaddr,
4095                         .maxlen         =       sizeof(int),
4096                         .mode           =       0644,
4097                         .proc_handler   =       proc_dointvec,
4098                 },
4099                 {
4100                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4101                         .procname       =       "temp_valid_lft",
4102                         .data           =       &ipv6_devconf.temp_valid_lft,
4103                         .maxlen         =       sizeof(int),
4104                         .mode           =       0644,
4105                         .proc_handler   =       proc_dointvec,
4106                 },
4107                 {
4108                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4109                         .procname       =       "temp_prefered_lft",
4110                         .data           =       &ipv6_devconf.temp_prefered_lft,
4111                         .maxlen         =       sizeof(int),
4112                         .mode           =       0644,
4113                         .proc_handler   =       proc_dointvec,
4114                 },
4115                 {
4116                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4117                         .procname       =       "regen_max_retry",
4118                         .data           =       &ipv6_devconf.regen_max_retry,
4119                         .maxlen         =       sizeof(int),
4120                         .mode           =       0644,
4121                         .proc_handler   =       proc_dointvec,
4122                 },
4123                 {
4124                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4125                         .procname       =       "max_desync_factor",
4126                         .data           =       &ipv6_devconf.max_desync_factor,
4127                         .maxlen         =       sizeof(int),
4128                         .mode           =       0644,
4129                         .proc_handler   =       proc_dointvec,
4130                 },
4131 #endif
4132                 {
4133                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4134                         .procname       =       "max_addresses",
4135                         .data           =       &ipv6_devconf.max_addresses,
4136                         .maxlen         =       sizeof(int),
4137                         .mode           =       0644,
4138                         .proc_handler   =       proc_dointvec,
4139                 },
4140                 {
4141                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4142                         .procname       =       "accept_ra_defrtr",
4143                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4144                         .maxlen         =       sizeof(int),
4145                         .mode           =       0644,
4146                         .proc_handler   =       proc_dointvec,
4147                 },
4148                 {
4149                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4150                         .procname       =       "accept_ra_pinfo",
4151                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4152                         .maxlen         =       sizeof(int),
4153                         .mode           =       0644,
4154                         .proc_handler   =       proc_dointvec,
4155                 },
4156 #ifdef CONFIG_IPV6_ROUTER_PREF
4157                 {
4158                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4159                         .procname       =       "accept_ra_rtr_pref",
4160                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4161                         .maxlen         =       sizeof(int),
4162                         .mode           =       0644,
4163                         .proc_handler   =       proc_dointvec,
4164                 },
4165                 {
4166                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4167                         .procname       =       "router_probe_interval",
4168                         .data           =       &ipv6_devconf.rtr_probe_interval,
4169                         .maxlen         =       sizeof(int),
4170                         .mode           =       0644,
4171                         .proc_handler   =       proc_dointvec_jiffies,
4172                         .strategy       =       sysctl_jiffies,
4173                 },
4174 #ifdef CONFIG_IPV6_ROUTE_INFO
4175                 {
4176                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4177                         .procname       =       "accept_ra_rt_info_max_plen",
4178                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4179                         .maxlen         =       sizeof(int),
4180                         .mode           =       0644,
4181                         .proc_handler   =       proc_dointvec,
4182                 },
4183 #endif
4184 #endif
4185                 {
4186                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4187                         .procname       =       "proxy_ndp",
4188                         .data           =       &ipv6_devconf.proxy_ndp,
4189                         .maxlen         =       sizeof(int),
4190                         .mode           =       0644,
4191                         .proc_handler   =       proc_dointvec,
4192                 },
4193                 {
4194                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4195                         .procname       =       "accept_source_route",
4196                         .data           =       &ipv6_devconf.accept_source_route,
4197                         .maxlen         =       sizeof(int),
4198                         .mode           =       0644,
4199                         .proc_handler   =       proc_dointvec,
4200                 },
4201 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4202                 {
4203                         .ctl_name       =       CTL_UNNUMBERED,
4204                         .procname       =       "optimistic_dad",
4205                         .data           =       &ipv6_devconf.optimistic_dad,
4206                         .maxlen         =       sizeof(int),
4207                         .mode           =       0644,
4208                         .proc_handler   =       proc_dointvec,
4209
4210                 },
4211 #endif
4212 #ifdef CONFIG_IPV6_MROUTE
4213                 {
4214                         .ctl_name       =       CTL_UNNUMBERED,
4215                         .procname       =       "mc_forwarding",
4216                         .data           =       &ipv6_devconf.mc_forwarding,
4217                         .maxlen         =       sizeof(int),
4218                         .mode           =       0444,
4219                         .proc_handler   =       proc_dointvec,
4220                 },
4221 #endif
4222                 {
4223                         .ctl_name       =       CTL_UNNUMBERED,
4224                         .procname       =       "disable_ipv6",
4225                         .data           =       &ipv6_devconf.disable_ipv6,
4226                         .maxlen         =       sizeof(int),
4227                         .mode           =       0644,
4228                         .proc_handler   =       proc_dointvec,
4229                 },
4230                 {
4231                         .ctl_name       =       CTL_UNNUMBERED,
4232                         .procname       =       "accept_dad",
4233                         .data           =       &ipv6_devconf.accept_dad,
4234                         .maxlen         =       sizeof(int),
4235                         .mode           =       0644,
4236                         .proc_handler   =       proc_dointvec,
4237                 },
4238                 {
4239                         .ctl_name       =       0,      /* sentinel */
4240                 }
4241         },
4242 };
4243
4244 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4245                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4246 {
4247         int i;
4248         struct addrconf_sysctl_table *t;
4249
4250 #define ADDRCONF_CTL_PATH_DEV   3
4251
4252         struct ctl_path addrconf_ctl_path[] = {
4253                 { .procname = "net", .ctl_name = CTL_NET, },
4254                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4255                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4256                 { /* to be set */ },
4257                 { },
4258         };
4259
4260
4261         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4262         if (t == NULL)
4263                 goto out;
4264
4265         for (i=0; t->addrconf_vars[i].data; i++) {
4266                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4267                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4268                 t->addrconf_vars[i].extra2 = net;
4269         }
4270
4271         /*
4272          * Make a copy of dev_name, because '.procname' is regarded as const
4273          * by sysctl and we wouldn't want anyone to change it under our feet
4274          * (see SIOCSIFNAME).
4275          */
4276         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4277         if (!t->dev_name)
4278                 goto free;
4279
4280         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4281         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4282
4283         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4284                         t->addrconf_vars);
4285         if (t->sysctl_header == NULL)
4286                 goto free_procname;
4287
4288         p->sysctl = t;
4289         return 0;
4290
4291 free_procname:
4292         kfree(t->dev_name);
4293 free:
4294         kfree(t);
4295 out:
4296         return -ENOBUFS;
4297 }
4298
4299 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4300 {
4301         struct addrconf_sysctl_table *t;
4302
4303         if (p->sysctl == NULL)
4304                 return;
4305
4306         t = p->sysctl;
4307         p->sysctl = NULL;
4308         unregister_sysctl_table(t->sysctl_header);
4309         kfree(t->dev_name);
4310         kfree(t);
4311 }
4312
4313 static void addrconf_sysctl_register(struct inet6_dev *idev)
4314 {
4315         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4316                               NET_IPV6_NEIGH, "ipv6",
4317                               &ndisc_ifinfo_sysctl_change,
4318                               ndisc_ifinfo_sysctl_strategy);
4319         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4320                         idev->dev->ifindex, idev, &idev->cnf);
4321 }
4322
4323 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4324 {
4325         __addrconf_sysctl_unregister(&idev->cnf);
4326         neigh_sysctl_unregister(idev->nd_parms);
4327 }
4328
4329
4330 #endif
4331
4332 static int addrconf_init_net(struct net *net)
4333 {
4334         int err;
4335         struct ipv6_devconf *all, *dflt;
4336
4337         err = -ENOMEM;
4338         all = &ipv6_devconf;
4339         dflt = &ipv6_devconf_dflt;
4340
4341         if (net != &init_net) {
4342                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4343                 if (all == NULL)
4344                         goto err_alloc_all;
4345
4346                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4347                 if (dflt == NULL)
4348                         goto err_alloc_dflt;
4349         }
4350
4351         net->ipv6.devconf_all = all;
4352         net->ipv6.devconf_dflt = dflt;
4353
4354 #ifdef CONFIG_SYSCTL
4355         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4356                         NULL, all);
4357         if (err < 0)
4358                 goto err_reg_all;
4359
4360         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4361                         NULL, dflt);
4362         if (err < 0)
4363                 goto err_reg_dflt;
4364 #endif
4365         return 0;
4366
4367 #ifdef CONFIG_SYSCTL
4368 err_reg_dflt:
4369         __addrconf_sysctl_unregister(all);
4370 err_reg_all:
4371         kfree(dflt);
4372 #endif
4373 err_alloc_dflt:
4374         kfree(all);
4375 err_alloc_all:
4376         return err;
4377 }
4378
4379 static void addrconf_exit_net(struct net *net)
4380 {
4381 #ifdef CONFIG_SYSCTL
4382         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4383         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4384 #endif
4385         if (net != &init_net) {
4386                 kfree(net->ipv6.devconf_dflt);
4387                 kfree(net->ipv6.devconf_all);
4388         }
4389 }
4390
4391 static struct pernet_operations addrconf_ops = {
4392         .init = addrconf_init_net,
4393         .exit = addrconf_exit_net,
4394 };
4395
4396 /*
4397  *      Device notifier
4398  */
4399
4400 int register_inet6addr_notifier(struct notifier_block *nb)
4401 {
4402         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4403 }
4404
4405 EXPORT_SYMBOL(register_inet6addr_notifier);
4406
4407 int unregister_inet6addr_notifier(struct notifier_block *nb)
4408 {
4409         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4410 }
4411
4412 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4413
4414 /*
4415  *      Init / cleanup code
4416  */
4417
4418 int __init addrconf_init(void)
4419 {
4420         int err;
4421
4422         if ((err = ipv6_addr_label_init()) < 0) {
4423                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4424                         err);
4425                 return err;
4426         }
4427
4428         register_pernet_subsys(&addrconf_ops);
4429
4430         /* The addrconf netdev notifier requires that loopback_dev
4431          * has it's ipv6 private information allocated and setup
4432          * before it can bring up and give link-local addresses
4433          * to other devices which are up.
4434          *
4435          * Unfortunately, loopback_dev is not necessarily the first
4436          * entry in the global dev_base list of net devices.  In fact,
4437          * it is likely to be the very last entry on that list.
4438          * So this causes the notifier registry below to try and
4439          * give link-local addresses to all devices besides loopback_dev
4440          * first, then loopback_dev, which cases all the non-loopback_dev
4441          * devices to fail to get a link-local address.
4442          *
4443          * So, as a temporary fix, allocate the ipv6 structure for
4444          * loopback_dev first by hand.
4445          * Longer term, all of the dependencies ipv6 has upon the loopback
4446          * device and it being up should be removed.
4447          */
4448         rtnl_lock();
4449         if (!ipv6_add_dev(init_net.loopback_dev))
4450                 err = -ENOMEM;
4451         rtnl_unlock();
4452         if (err)
4453                 goto errlo;
4454
4455         register_netdevice_notifier(&ipv6_dev_notf);
4456
4457         addrconf_verify(0);
4458
4459         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4460         if (err < 0)
4461                 goto errout;
4462
4463         /* Only the first call to __rtnl_register can fail */
4464         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4465         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4466         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4467         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4468         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4469
4470         ipv6_addr_label_rtnl_register();
4471
4472         return 0;
4473 errout:
4474         unregister_netdevice_notifier(&ipv6_dev_notf);
4475 errlo:
4476         unregister_pernet_subsys(&addrconf_ops);
4477
4478         return err;
4479 }
4480
4481 void addrconf_cleanup(void)
4482 {
4483         struct inet6_ifaddr *ifa;
4484         struct net_device *dev;
4485         int i;
4486
4487         unregister_netdevice_notifier(&ipv6_dev_notf);
4488         unregister_pernet_subsys(&addrconf_ops);
4489
4490         rtnl_lock();
4491
4492         /* clean dev list */
4493         for_each_netdev(&init_net, dev) {
4494                 if (__in6_dev_get(dev) == NULL)
4495                         continue;
4496                 addrconf_ifdown(dev, 1);
4497         }
4498         addrconf_ifdown(init_net.loopback_dev, 2);
4499
4500         /*
4501          *      Check hash table.
4502          */
4503         write_lock_bh(&addrconf_hash_lock);
4504         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4505                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4506                         struct inet6_ifaddr *bifa;
4507
4508                         bifa = ifa;
4509                         ifa = ifa->lst_next;
4510                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4511                         /* Do not free it; something is wrong.
4512                            Now we can investigate it with debugger.
4513                          */
4514                 }
4515         }
4516         write_unlock_bh(&addrconf_hash_lock);
4517
4518         del_timer(&addr_chk_timer);
4519         rtnl_unlock();
4520 }