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 void addrconf_bonding_change(struct net_device *dev,
 141                                     unsigned long event);
 142 static int addrconf_ifdown(struct net_device *dev, int how);
 143
 144 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 145 static void addrconf_dad_timer(unsigned long data);
 146 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 147 static void addrconf_dad_run(struct inet6_dev *idev);
 148 static void addrconf_rs_timer(unsigned long data);
 149 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 150 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 151
 152 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 153                                 struct prefix_info *pinfo);
 154 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 155                               struct net_device *dev);
 156
 157 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 158
 159 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 160         .forwarding             = 0,
 161         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 162         .mtu6                   = IPV6_MIN_MTU,
 163         .accept_ra              = 1,
 164         .accept_redirects       = 1,
 165         .autoconf               = 1,
 166         .force_mld_version      = 0,
 167         .dad_transmits          = 1,
 168         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 169         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 170         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 171 #ifdef CONFIG_IPV6_PRIVACY
 172         .use_tempaddr           = 0,
 173         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 174         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 175         .regen_max_retry        = REGEN_MAX_RETRY,
 176         .max_desync_factor      = MAX_DESYNC_FACTOR,
 177 #endif
 178         .max_addresses          = IPV6_MAX_ADDRESSES,
 179         .accept_ra_defrtr       = 1,
 180         .accept_ra_pinfo        = 1,
 181 #ifdef CONFIG_IPV6_ROUTER_PREF
 182         .accept_ra_rtr_pref     = 1,
 183         .rtr_probe_interval     = 60 * HZ,
 184 #ifdef CONFIG_IPV6_ROUTE_INFO
 185         .accept_ra_rt_info_max_plen = 0,
 186 #endif
 187 #endif
 188         .proxy_ndp              = 0,
 189         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 190         .disable_ipv6           = 0,
 191         .accept_dad             = 1,
 192 };
 193
 194 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 195         .forwarding             = 0,
 196         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 197         .mtu6                   = IPV6_MIN_MTU,
 198         .accept_ra              = 1,
 199         .accept_redirects       = 1,
 200         .autoconf               = 1,
 201         .dad_transmits          = 1,
 202         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 203         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 204         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 205 #ifdef CONFIG_IPV6_PRIVACY
 206         .use_tempaddr           = 0,
 207         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 208         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 209         .regen_max_retry        = REGEN_MAX_RETRY,
 210         .max_desync_factor      = MAX_DESYNC_FACTOR,
 211 #endif
 212         .max_addresses          = IPV6_MAX_ADDRESSES,
 213         .accept_ra_defrtr       = 1,
 214         .accept_ra_pinfo        = 1,
 215 #ifdef CONFIG_IPV6_ROUTER_PREF
 216         .accept_ra_rtr_pref     = 1,
 217         .rtr_probe_interval     = 60 * HZ,
 218 #ifdef CONFIG_IPV6_ROUTE_INFO
 219         .accept_ra_rt_info_max_plen = 0,
 220 #endif
 221 #endif
 222         .proxy_ndp              = 0,
 223         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 224         .disable_ipv6           = 0,
 225         .accept_dad             = 1,
 226 };
 227
 228 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 229 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 230 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 231 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 232 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 233
 234 /* Check if a valid qdisc is available */
 235 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 236 {
 237         return !qdisc_tx_is_noop(dev);
 238 }
 239
 240 /* Check if a route is valid prefix route */
 241 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 242 {
 243         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 244 }
 245
 246 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 247 {
 248         if (del_timer(&ifp->timer))
 249                 __in6_ifa_put(ifp);
 250 }
 251
 252 enum addrconf_timer_t
 253 {
 254         AC_NONE,
 255         AC_DAD,
 256         AC_RS,
 257 };
 258
 259 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 260                                enum addrconf_timer_t what,
 261                                unsigned long when)
 262 {
 263         if (!del_timer(&ifp->timer))
 264                 in6_ifa_hold(ifp);
 265
 266         switch (what) {
 267         case AC_DAD:
 268                 ifp->timer.function = addrconf_dad_timer;
 269                 break;
 270         case AC_RS:
 271                 ifp->timer.function = addrconf_rs_timer;
 272                 break;
 273         default:;
 274         }
 275         ifp->timer.expires = jiffies + when;
 276         add_timer(&ifp->timer);
 277 }
 278
 279 static int snmp6_alloc_dev(struct inet6_dev *idev)
 280 {
 281         if (snmp_mib_init((void **)idev->stats.ipv6,
 282                           sizeof(struct ipstats_mib)) < 0)
 283                 goto err_ip;
 284         if (snmp_mib_init((void **)idev->stats.icmpv6,
 285                           sizeof(struct icmpv6_mib)) < 0)
 286                 goto err_icmp;
 287         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 288                           sizeof(struct icmpv6msg_mib)) < 0)
 289                 goto err_icmpmsg;
 290
 291         return 0;
 292
 293 err_icmpmsg:
 294         snmp_mib_free((void **)idev->stats.icmpv6);
 295 err_icmp:
 296         snmp_mib_free((void **)idev->stats.ipv6);
 297 err_ip:
 298         return -ENOMEM;
 299 }
 300
 301 static void snmp6_free_dev(struct inet6_dev *idev)
 302 {
 303         snmp_mib_free((void **)idev->stats.icmpv6msg);
 304         snmp_mib_free((void **)idev->stats.icmpv6);
 305         snmp_mib_free((void **)idev->stats.ipv6);
 306 }
 307
 308 /* Nobody refers to this device, we may destroy it. */
 309
 310 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 311 {
 312         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 313         kfree(idev);
 314 }
 315
 316 void in6_dev_finish_destroy(struct inet6_dev *idev)
 317 {
 318         struct net_device *dev = idev->dev;
 319
 320         WARN_ON(idev->addr_list != NULL);
 321         WARN_ON(idev->mc_list != NULL);
 322
 323 #ifdef NET_REFCNT_DEBUG
 324         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 325 #endif
 326         dev_put(dev);
 327         if (!idev->dead) {
 328                 printk("Freeing alive inet6 device %p\n", idev);
 329                 return;
 330         }
 331         snmp6_free_dev(idev);
 332         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 333 }
 334
 335 EXPORT_SYMBOL(in6_dev_finish_destroy);
 336
 337 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 338 {
 339         struct inet6_dev *ndev;
 340
 341         ASSERT_RTNL();
 342
 343         if (dev->mtu < IPV6_MIN_MTU)
 344                 return NULL;
 345
 346         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 347
 348         if (ndev == NULL)
 349                 return NULL;
 350
 351         rwlock_init(&ndev->lock);
 352         ndev->dev = dev;
 353         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 354         ndev->cnf.mtu6 = dev->mtu;
 355         ndev->cnf.sysctl = NULL;
 356         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 357         if (ndev->nd_parms == NULL) {
 358                 kfree(ndev);
 359                 return NULL;
 360         }
 361         if (ndev->cnf.forwarding)
 362                 dev_disable_lro(dev);
 363         /* We refer to the device */
 364         dev_hold(dev);
 365
 366         if (snmp6_alloc_dev(ndev) < 0) {
 367                 ADBG((KERN_WARNING
 368                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 369                         __func__, dev->name));
 370                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 371                 ndev->dead = 1;
 372                 in6_dev_finish_destroy(ndev);
 373                 return NULL;
 374         }
 375
 376         if (snmp6_register_dev(ndev) < 0) {
 377                 ADBG((KERN_WARNING
 378                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 379                         __func__, dev->name));
 380                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 381                 ndev->dead = 1;
 382                 in6_dev_finish_destroy(ndev);
 383                 return NULL;
 384         }
 385
 386         /* One reference from device.  We must do this before
 387          * we invoke __ipv6_regen_rndid().
 388          */
 389         in6_dev_hold(ndev);
 390
 391         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 392                 ndev->cnf.accept_dad = -1;
 393
 394 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 395         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 396                 printk(KERN_INFO
 397                        "%s: Disabled Multicast RS\n",
 398                        dev->name);
 399                 ndev->cnf.rtr_solicits = 0;
 400         }
 401 #endif
 402
 403 #ifdef CONFIG_IPV6_PRIVACY
 404         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 405         if ((dev->flags&IFF_LOOPBACK) ||
 406             dev->type == ARPHRD_TUNNEL ||
 407             dev->type == ARPHRD_TUNNEL6 ||
 408             dev->type == ARPHRD_SIT ||
 409             dev->type == ARPHRD_NONE) {
 410                 printk(KERN_INFO
 411                        "%s: Disabled Privacy Extensions\n",
 412                        dev->name);
 413                 ndev->cnf.use_tempaddr = -1;
 414         } else {
 415                 in6_dev_hold(ndev);
 416                 ipv6_regen_rndid((unsigned long) ndev);
 417         }
 418 #endif
 419
 420         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 421                 ndev->if_flags |= IF_READY;
 422
 423         ipv6_mc_init_dev(ndev);
 424         ndev->tstamp = jiffies;
 425         addrconf_sysctl_register(ndev);
 426         /* protected by rtnl_lock */
 427         rcu_assign_pointer(dev->ip6_ptr, ndev);
 428
 429         /* Join all-node multicast group */
 430         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 431
 432         return ndev;
 433 }
 434
 435 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 436 {
 437         struct inet6_dev *idev;
 438
 439         ASSERT_RTNL();
 440
 441         if ((idev = __in6_dev_get(dev)) == NULL) {
 442                 if ((idev = ipv6_add_dev(dev)) == NULL)
 443                         return NULL;
 444         }
 445
 446         if (dev->flags&IFF_UP)
 447                 ipv6_mc_up(idev);
 448         return idev;
 449 }
 450
 451 #ifdef CONFIG_SYSCTL
 452 static void dev_forward_change(struct inet6_dev *idev)
 453 {
 454         struct net_device *dev;
 455         struct inet6_ifaddr *ifa;
 456
 457         if (!idev)
 458                 return;
 459         dev = idev->dev;
 460         if (idev->cnf.forwarding)
 461                 dev_disable_lro(dev);
 462         if (dev && (dev->flags & IFF_MULTICAST)) {
 463                 if (idev->cnf.forwarding)
 464                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 465                 else
 466                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 467         }
 468         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 469                 if (ifa->flags&IFA_F_TENTATIVE)
 470                         continue;
 471                 if (idev->cnf.forwarding)
 472                         addrconf_join_anycast(ifa);
 473                 else
 474                         addrconf_leave_anycast(ifa);
 475         }
 476 }
 477
 478
 479 static void addrconf_forward_change(struct net *net, __s32 newf)
 480 {
 481         struct net_device *dev;
 482         struct inet6_dev *idev;
 483
 484         rcu_read_lock();
 485         for_each_netdev_rcu(net, dev) {
 486                 idev = __in6_dev_get(dev);
 487                 if (idev) {
 488                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 489                         idev->cnf.forwarding = newf;
 490                         if (changed)
 491                                 dev_forward_change(idev);
 492                 }
 493         }
 494         rcu_read_unlock();
 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         int hash;
 595         int err = 0;
 596         int addr_type = ipv6_addr_type(addr);
 597
 598         if (addr_type == IPV6_ADDR_ANY ||
 599             addr_type & IPV6_ADDR_MULTICAST ||
 600             (!(idev->dev->flags & IFF_LOOPBACK) &&
 601              addr_type & IPV6_ADDR_LOOPBACK))
 602                 return ERR_PTR(-EADDRNOTAVAIL);
 603
 604         rcu_read_lock_bh();
 605         if (idev->dead) {
 606                 err = -ENODEV;                  /*XXX*/
 607                 goto out2;
 608         }
 609
 610         if (idev->cnf.disable_ipv6) {
 611                 err = -EACCES;
 612                 goto out2;
 613         }
 614
 615         write_lock(&addrconf_hash_lock);
 616
 617         /* Ignore adding duplicate addresses on an interface */
 618         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 619                 ADBG(("ipv6_add_addr: already assigned\n"));
 620                 err = -EEXIST;
 621                 goto out;
 622         }
 623
 624         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 625
 626         if (ifa == NULL) {
 627                 ADBG(("ipv6_add_addr: malloc failed\n"));
 628                 err = -ENOBUFS;
 629                 goto out;
 630         }
 631
 632         rt = addrconf_dst_alloc(idev, addr, 0);
 633         if (IS_ERR(rt)) {
 634                 err = PTR_ERR(rt);
 635                 goto out;
 636         }
 637
 638         ipv6_addr_copy(&ifa->addr, addr);
 639
 640         spin_lock_init(&ifa->lock);
 641         init_timer(&ifa->timer);
 642         ifa->timer.data = (unsigned long) ifa;
 643         ifa->scope = scope;
 644         ifa->prefix_len = pfxlen;
 645         ifa->flags = flags | IFA_F_TENTATIVE;
 646         ifa->cstamp = ifa->tstamp = jiffies;
 647
 648         ifa->rt = rt;
 649
 650         /*
 651          * part one of RFC 4429, section 3.3
 652          * We should not configure an address as
 653          * optimistic if we do not yet know the link
 654          * layer address of our nexhop router
 655          */
 656
 657         if (rt->rt6i_nexthop == NULL)
 658                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 659
 660         ifa->idev = idev;
 661         in6_dev_hold(idev);
 662         /* For caller */
 663         in6_ifa_hold(ifa);
 664
 665         /* Add to big hash table */
 666         hash = ipv6_addr_hash(addr);
 667
 668         ifa->lst_next = inet6_addr_lst[hash];
 669         inet6_addr_lst[hash] = ifa;
 670         in6_ifa_hold(ifa);
 671         write_unlock(&addrconf_hash_lock);
 672
 673         write_lock(&idev->lock);
 674         /* Add to inet6_dev unicast addr list. */
 675         ipv6_link_dev_addr(idev, ifa);
 676
 677 #ifdef CONFIG_IPV6_PRIVACY
 678         if (ifa->flags&IFA_F_TEMPORARY) {
 679                 ifa->tmp_next = idev->tempaddr_list;
 680                 idev->tempaddr_list = ifa;
 681                 in6_ifa_hold(ifa);
 682         }
 683 #endif
 684
 685         in6_ifa_hold(ifa);
 686         write_unlock(&idev->lock);
 687 out2:
 688         rcu_read_unlock_bh();
 689
 690         if (likely(err == 0))
 691                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 692         else {
 693                 kfree(ifa);
 694                 ifa = ERR_PTR(err);
 695         }
 696
 697         return ifa;
 698 out:
 699         write_unlock(&addrconf_hash_lock);
 700         goto out2;
 701 }
 702
 703 /* This function wants to get referenced ifp and releases it before return */
 704
 705 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 706 {
 707         struct inet6_ifaddr *ifa, **ifap;
 708         struct inet6_dev *idev = ifp->idev;
 709         int hash;
 710         int deleted = 0, onlink = 0;
 711         unsigned long expires = jiffies;
 712
 713         hash = ipv6_addr_hash(&ifp->addr);
 714
 715         ifp->dead = 1;
 716
 717         write_lock_bh(&addrconf_hash_lock);
 718         for (ifap = &inet6_addr_lst[hash]; (ifa=*ifap) != NULL;
 719              ifap = &ifa->lst_next) {
 720                 if (ifa == ifp) {
 721                         *ifap = ifa->lst_next;
 722                         __in6_ifa_put(ifp);
 723                         ifa->lst_next = NULL;
 724                         break;
 725                 }
 726         }
 727         write_unlock_bh(&addrconf_hash_lock);
 728
 729         write_lock_bh(&idev->lock);
 730 #ifdef CONFIG_IPV6_PRIVACY
 731         if (ifp->flags&IFA_F_TEMPORARY) {
 732                 for (ifap = &idev->tempaddr_list; (ifa=*ifap) != NULL;
 733                      ifap = &ifa->tmp_next) {
 734                         if (ifa == ifp) {
 735                                 *ifap = ifa->tmp_next;
 736                                 if (ifp->ifpub) {
 737                                         in6_ifa_put(ifp->ifpub);
 738                                         ifp->ifpub = NULL;
 739                                 }
 740                                 __in6_ifa_put(ifp);
 741                                 ifa->tmp_next = NULL;
 742                                 break;
 743                         }
 744                 }
 745         }
 746 #endif
 747
 748         for (ifap = &idev->addr_list; (ifa=*ifap) != NULL;) {
 749                 if (ifa == ifp) {
 750                         *ifap = ifa->if_next;
 751                         __in6_ifa_put(ifp);
 752                         ifa->if_next = NULL;
 753                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 754                                 break;
 755                         deleted = 1;
 756                         continue;
 757                 } else if (ifp->flags & IFA_F_PERMANENT) {
 758                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 759                                               ifp->prefix_len)) {
 760                                 if (ifa->flags & IFA_F_PERMANENT) {
 761                                         onlink = 1;
 762                                         if (deleted)
 763                                                 break;
 764                                 } else {
 765                                         unsigned long lifetime;
 766
 767                                         if (!onlink)
 768                                                 onlink = -1;
 769
 770                                         spin_lock(&ifa->lock);
 771
 772                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 773                                         /*
 774                                          * Note: Because this address is
 775                                          * not permanent, lifetime <
 776                                          * LONG_MAX / HZ here.
 777                                          */
 778                                         if (time_before(expires,
 779                                                         ifa->tstamp + lifetime * HZ))
 780                                                 expires = ifa->tstamp + lifetime * HZ;
 781                                         spin_unlock(&ifa->lock);
 782                                 }
 783                         }
 784                 }
 785                 ifap = &ifa->if_next;
 786         }
 787         write_unlock_bh(&idev->lock);
 788
 789         addrconf_del_timer(ifp);
 790
 791         ipv6_ifa_notify(RTM_DELADDR, ifp);
 792
 793         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 794
 795         /*
 796          * Purge or update corresponding prefix
 797          *
 798          * 1) we don't purge prefix here if address was not permanent.
 799          *    prefix is managed by its own lifetime.
 800          * 2) if there're no addresses, delete prefix.
 801          * 3) if there're still other permanent address(es),
 802          *    corresponding prefix is still permanent.
 803          * 4) otherwise, update prefix lifetime to the
 804          *    longest valid lifetime among the corresponding
 805          *    addresses on the device.
 806          *    Note: subsequent RA will update lifetime.
 807          *
 808          * --yoshfuji
 809          */
 810         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 811                 struct in6_addr prefix;
 812                 struct rt6_info *rt;
 813                 struct net *net = dev_net(ifp->idev->dev);
 814                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 815                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 816
 817                 if (rt && addrconf_is_prefix_route(rt)) {
 818                         if (onlink == 0) {
 819                                 ip6_del_rt(rt);
 820                                 rt = NULL;
 821                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 822                                 rt->rt6i_expires = expires;
 823                                 rt->rt6i_flags |= RTF_EXPIRES;
 824                         }
 825                 }
 826                 dst_release(&rt->u.dst);
 827         }
 828
 829         in6_ifa_put(ifp);
 830 }
 831
 832 #ifdef CONFIG_IPV6_PRIVACY
 833 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 834 {
 835         struct inet6_dev *idev = ifp->idev;
 836         struct in6_addr addr, *tmpaddr;
 837         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_cstamp, tmp_tstamp;
 838         unsigned long regen_advance;
 839         int tmp_plen;
 840         int ret = 0;
 841         int max_addresses;
 842         u32 addr_flags;
 843
 844         write_lock(&idev->lock);
 845         if (ift) {
 846                 spin_lock_bh(&ift->lock);
 847                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 848                 spin_unlock_bh(&ift->lock);
 849                 tmpaddr = &addr;
 850         } else {
 851                 tmpaddr = NULL;
 852         }
 853 retry:
 854         in6_dev_hold(idev);
 855         if (idev->cnf.use_tempaddr <= 0) {
 856                 write_unlock(&idev->lock);
 857                 printk(KERN_INFO
 858                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 859                 in6_dev_put(idev);
 860                 ret = -1;
 861                 goto out;
 862         }
 863         spin_lock_bh(&ifp->lock);
 864         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 865                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 866                 spin_unlock_bh(&ifp->lock);
 867                 write_unlock(&idev->lock);
 868                 printk(KERN_WARNING
 869                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 870                 in6_dev_put(idev);
 871                 ret = -1;
 872                 goto out;
 873         }
 874         in6_ifa_hold(ifp);
 875         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 876         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 877                 spin_unlock_bh(&ifp->lock);
 878                 write_unlock(&idev->lock);
 879                 printk(KERN_WARNING
 880                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 881                 in6_ifa_put(ifp);
 882                 in6_dev_put(idev);
 883                 ret = -1;
 884                 goto out;
 885         }
 886         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 887         tmp_valid_lft = min_t(__u32,
 888                               ifp->valid_lft,
 889                               idev->cnf.temp_valid_lft);
 890         tmp_prefered_lft = min_t(__u32,
 891                                  ifp->prefered_lft,
 892                                  idev->cnf.temp_prefered_lft - desync_factor / HZ);
 893         tmp_plen = ifp->prefix_len;
 894         max_addresses = idev->cnf.max_addresses;
 895         tmp_cstamp = ifp->cstamp;
 896         tmp_tstamp = ifp->tstamp;
 897         spin_unlock_bh(&ifp->lock);
 898
 899         regen_advance = idev->cnf.regen_max_retry *
 900                         idev->cnf.dad_transmits *
 901                         idev->nd_parms->retrans_time / HZ;
 902         write_unlock(&idev->lock);
 903
 904         /* A temporary address is created only if this calculated Preferred
 905          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 906          * an implementation must not create a temporary address with a zero
 907          * Preferred Lifetime.
 908          */
 909         if (tmp_prefered_lft <= regen_advance) {
 910                 in6_ifa_put(ifp);
 911                 in6_dev_put(idev);
 912                 ret = -1;
 913                 goto out;
 914         }
 915
 916         addr_flags = IFA_F_TEMPORARY;
 917         /* set in addrconf_prefix_rcv() */
 918         if (ifp->flags & IFA_F_OPTIMISTIC)
 919                 addr_flags |= IFA_F_OPTIMISTIC;
 920
 921         ift = !max_addresses ||
 922               ipv6_count_addresses(idev) < max_addresses ?
 923                 ipv6_add_addr(idev, &addr, tmp_plen,
 924                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 925                               addr_flags) : NULL;
 926         if (!ift || IS_ERR(ift)) {
 927                 in6_ifa_put(ifp);
 928                 in6_dev_put(idev);
 929                 printk(KERN_INFO
 930                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 931                 tmpaddr = &addr;
 932                 write_lock(&idev->lock);
 933                 goto retry;
 934         }
 935
 936         spin_lock_bh(&ift->lock);
 937         ift->ifpub = ifp;
 938         ift->valid_lft = tmp_valid_lft;
 939         ift->prefered_lft = tmp_prefered_lft;
 940         ift->cstamp = tmp_cstamp;
 941         ift->tstamp = tmp_tstamp;
 942         spin_unlock_bh(&ift->lock);
 943
 944         addrconf_dad_start(ift, 0);
 945         in6_ifa_put(ift);
 946         in6_dev_put(idev);
 947 out:
 948         return ret;
 949 }
 950 #endif
 951
 952 /*
 953  *      Choose an appropriate source address (RFC3484)
 954  */
 955 enum {
 956         IPV6_SADDR_RULE_INIT = 0,
 957         IPV6_SADDR_RULE_LOCAL,
 958         IPV6_SADDR_RULE_SCOPE,
 959         IPV6_SADDR_RULE_PREFERRED,
 960 #ifdef CONFIG_IPV6_MIP6
 961         IPV6_SADDR_RULE_HOA,
 962 #endif
 963         IPV6_SADDR_RULE_OIF,
 964         IPV6_SADDR_RULE_LABEL,
 965 #ifdef CONFIG_IPV6_PRIVACY
 966         IPV6_SADDR_RULE_PRIVACY,
 967 #endif
 968         IPV6_SADDR_RULE_ORCHID,
 969         IPV6_SADDR_RULE_PREFIX,
 970         IPV6_SADDR_RULE_MAX
 971 };
 972
 973 struct ipv6_saddr_score {
 974         int                     rule;
 975         int                     addr_type;
 976         struct inet6_ifaddr     *ifa;
 977         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 978         int                     scopedist;
 979         int                     matchlen;
 980 };
 981
 982 struct ipv6_saddr_dst {
 983         const struct in6_addr *addr;
 984         int ifindex;
 985         int scope;
 986         int label;
 987         unsigned int prefs;
 988 };
 989
 990 static inline int ipv6_saddr_preferred(int type)
 991 {
 992         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|
 993                     IPV6_ADDR_LOOPBACK|IPV6_ADDR_RESERVED))
 994                 return 1;
 995         return 0;
 996 }
 997
 998 static int ipv6_get_saddr_eval(struct net *net,
 999                                struct ipv6_saddr_score *score,
1000                                struct ipv6_saddr_dst *dst,
1001                                int i)
1002 {
1003         int ret;
1004
1005         if (i <= score->rule) {
1006                 switch (i) {
1007                 case IPV6_SADDR_RULE_SCOPE:
1008                         ret = score->scopedist;
1009                         break;
1010                 case IPV6_SADDR_RULE_PREFIX:
1011                         ret = score->matchlen;
1012                         break;
1013                 default:
1014                         ret = !!test_bit(i, score->scorebits);
1015                 }
1016                 goto out;
1017         }
1018
1019         switch (i) {
1020         case IPV6_SADDR_RULE_INIT:
1021                 /* Rule 0: remember if hiscore is not ready yet */
1022                 ret = !!score->ifa;
1023                 break;
1024         case IPV6_SADDR_RULE_LOCAL:
1025                 /* Rule 1: Prefer same address */
1026                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1027                 break;
1028         case IPV6_SADDR_RULE_SCOPE:
1029                 /* Rule 2: Prefer appropriate scope
1030                  *
1031                  *      ret
1032                  *       ^
1033                  *    -1 |  d 15
1034                  *    ---+--+-+---> scope
1035                  *       |
1036                  *       |             d is scope of the destination.
1037                  *  B-d  |  \
1038                  *       |   \      <- smaller scope is better if
1039                  *  B-15 |    \        if scope is enough for destinaion.
1040                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1041                  * d-C-1 | /
1042                  *       |/         <- greater is better
1043                  *   -C  /             if scope is not enough for destination.
1044                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1045                  *
1046                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1047                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1048                  * Assume B = 0 and we get C > 29.
1049                  */
1050                 ret = __ipv6_addr_src_scope(score->addr_type);
1051                 if (ret >= dst->scope)
1052                         ret = -ret;
1053                 else
1054                         ret -= 128;     /* 30 is enough */
1055                 score->scopedist = ret;
1056                 break;
1057         case IPV6_SADDR_RULE_PREFERRED:
1058                 /* Rule 3: Avoid deprecated and optimistic addresses */
1059                 ret = ipv6_saddr_preferred(score->addr_type) ||
1060                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1061                 break;
1062 #ifdef CONFIG_IPV6_MIP6
1063         case IPV6_SADDR_RULE_HOA:
1064             {
1065                 /* Rule 4: Prefer home address */
1066                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1067                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1068                 break;
1069             }
1070 #endif
1071         case IPV6_SADDR_RULE_OIF:
1072                 /* Rule 5: Prefer outgoing interface */
1073                 ret = (!dst->ifindex ||
1074                        dst->ifindex == score->ifa->idev->dev->ifindex);
1075                 break;
1076         case IPV6_SADDR_RULE_LABEL:
1077                 /* Rule 6: Prefer matching label */
1078                 ret = ipv6_addr_label(net,
1079                                       &score->ifa->addr, score->addr_type,
1080                                       score->ifa->idev->dev->ifindex) == dst->label;
1081                 break;
1082 #ifdef CONFIG_IPV6_PRIVACY
1083         case IPV6_SADDR_RULE_PRIVACY:
1084             {
1085                 /* Rule 7: Prefer public address
1086                  * Note: prefer temprary address if use_tempaddr >= 2
1087                  */
1088                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1089                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1090                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1091                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1092                 break;
1093             }
1094 #endif
1095         case IPV6_SADDR_RULE_ORCHID:
1096                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1097                  *          non-ORCHID vs non-ORCHID
1098                  */
1099                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1100                         ipv6_addr_orchid(dst->addr));
1101                 break;
1102         case IPV6_SADDR_RULE_PREFIX:
1103                 /* Rule 8: Use longest matching prefix */
1104                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1105                                                        dst->addr);
1106                 break;
1107         default:
1108                 ret = 0;
1109         }
1110
1111         if (ret)
1112                 __set_bit(i, score->scorebits);
1113         score->rule = i;
1114 out:
1115         return ret;
1116 }
1117
1118 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1119                        const struct in6_addr *daddr, unsigned int prefs,
1120                        struct in6_addr *saddr)
1121 {
1122         struct ipv6_saddr_score scores[2],
1123                                 *score = &scores[0], *hiscore = &scores[1];
1124         struct ipv6_saddr_dst dst;
1125         struct net_device *dev;
1126         int dst_type;
1127
1128         dst_type = __ipv6_addr_type(daddr);
1129         dst.addr = daddr;
1130         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1131         dst.scope = __ipv6_addr_src_scope(dst_type);
1132         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1133         dst.prefs = prefs;
1134
1135         hiscore->rule = -1;
1136         hiscore->ifa = NULL;
1137
1138         rcu_read_lock();
1139
1140         for_each_netdev_rcu(net, dev) {
1141                 struct inet6_dev *idev;
1142
1143                 /* Candidate Source Address (section 4)
1144                  *  - multicast and link-local destination address,
1145                  *    the set of candidate source address MUST only
1146                  *    include addresses assigned to interfaces
1147                  *    belonging to the same link as the outgoing
1148                  *    interface.
1149                  * (- For site-local destination addresses, the
1150                  *    set of candidate source addresses MUST only
1151                  *    include addresses assigned to interfaces
1152                  *    belonging to the same site as the outgoing
1153                  *    interface.)
1154                  */
1155                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1156                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1157                     dst.ifindex && dev->ifindex != dst.ifindex)
1158                         continue;
1159
1160                 idev = __in6_dev_get(dev);
1161                 if (!idev)
1162                         continue;
1163
1164                 read_lock_bh(&idev->lock);
1165                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1166                         int i;
1167
1168                         /*
1169                          * - Tentative Address (RFC2462 section 5.4)
1170                          *  - A tentative address is not considered
1171                          *    "assigned to an interface" in the traditional
1172                          *    sense, unless it is also flagged as optimistic.
1173                          * - Candidate Source Address (section 4)
1174                          *  - In any case, anycast addresses, multicast
1175                          *    addresses, and the unspecified address MUST
1176                          *    NOT be included in a candidate set.
1177                          */
1178                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1179                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1180                                 continue;
1181
1182                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1183
1184                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1185                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1186                                 LIMIT_NETDEBUG(KERN_DEBUG
1187                                                "ADDRCONF: unspecified / multicast address "
1188                                                "assigned as unicast address on %s",
1189                                                dev->name);
1190                                 continue;
1191                         }
1192
1193                         score->rule = -1;
1194                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1195
1196                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1197                                 int minihiscore, miniscore;
1198
1199                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1200                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1201
1202                                 if (minihiscore > miniscore) {
1203                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1204                                             score->scopedist > 0) {
1205                                                 /*
1206                                                  * special case:
1207                                                  * each remaining entry
1208                                                  * has too small (not enough)
1209                                                  * scope, because ifa entries
1210                                                  * are sorted by their scope
1211                                                  * values.
1212                                                  */
1213                                                 goto try_nextdev;
1214                                         }
1215                                         break;
1216                                 } else if (minihiscore < miniscore) {
1217                                         if (hiscore->ifa)
1218                                                 in6_ifa_put(hiscore->ifa);
1219
1220                                         in6_ifa_hold(score->ifa);
1221
1222                                         swap(hiscore, score);
1223
1224                                         /* restore our iterator */
1225                                         score->ifa = hiscore->ifa;
1226
1227                                         break;
1228                                 }
1229                         }
1230                 }
1231 try_nextdev:
1232                 read_unlock_bh(&idev->lock);
1233         }
1234         rcu_read_unlock();
1235
1236         if (!hiscore->ifa)
1237                 return -EADDRNOTAVAIL;
1238
1239         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1240         in6_ifa_put(hiscore->ifa);
1241         return 0;
1242 }
1243
1244 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1245
1246 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1247                     unsigned char banned_flags)
1248 {
1249         struct inet6_dev *idev;
1250         int err = -EADDRNOTAVAIL;
1251
1252         rcu_read_lock();
1253         if ((idev = __in6_dev_get(dev)) != NULL) {
1254                 struct inet6_ifaddr *ifp;
1255
1256                 read_lock_bh(&idev->lock);
1257                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1258                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1259                                 ipv6_addr_copy(addr, &ifp->addr);
1260                                 err = 0;
1261                                 break;
1262                         }
1263                 }
1264                 read_unlock_bh(&idev->lock);
1265         }
1266         rcu_read_unlock();
1267         return err;
1268 }
1269
1270 static int ipv6_count_addresses(struct inet6_dev *idev)
1271 {
1272         int cnt = 0;
1273         struct inet6_ifaddr *ifp;
1274
1275         read_lock_bh(&idev->lock);
1276         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1277                 cnt++;
1278         read_unlock_bh(&idev->lock);
1279         return cnt;
1280 }
1281
1282 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1283                   struct net_device *dev, int strict)
1284 {
1285         struct inet6_ifaddr * ifp;
1286         u8 hash = ipv6_addr_hash(addr);
1287
1288         read_lock_bh(&addrconf_hash_lock);
1289         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1290                 if (!net_eq(dev_net(ifp->idev->dev), net))
1291                         continue;
1292                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1293                     !(ifp->flags&IFA_F_TENTATIVE)) {
1294                         if (dev == NULL || ifp->idev->dev == dev ||
1295                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1296                                 break;
1297                 }
1298         }
1299         read_unlock_bh(&addrconf_hash_lock);
1300         return ifp != NULL;
1301 }
1302 EXPORT_SYMBOL(ipv6_chk_addr);
1303
1304 static
1305 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1306                        struct net_device *dev)
1307 {
1308         struct inet6_ifaddr * ifp;
1309         u8 hash = ipv6_addr_hash(addr);
1310
1311         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1312                 if (!net_eq(dev_net(ifp->idev->dev), net))
1313                         continue;
1314                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1315                         if (dev == NULL || ifp->idev->dev == dev)
1316                                 break;
1317                 }
1318         }
1319         return ifp != NULL;
1320 }
1321
1322 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1323 {
1324         struct inet6_dev *idev;
1325         struct inet6_ifaddr *ifa;
1326         int     onlink;
1327
1328         onlink = 0;
1329         rcu_read_lock();
1330         idev = __in6_dev_get(dev);
1331         if (idev) {
1332                 read_lock_bh(&idev->lock);
1333                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1334                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1335                                                    ifa->prefix_len);
1336                         if (onlink)
1337                                 break;
1338                 }
1339                 read_unlock_bh(&idev->lock);
1340         }
1341         rcu_read_unlock();
1342         return onlink;
1343 }
1344
1345 EXPORT_SYMBOL(ipv6_chk_prefix);
1346
1347 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1348                                      struct net_device *dev, int strict)
1349 {
1350         struct inet6_ifaddr * ifp;
1351         u8 hash = ipv6_addr_hash(addr);
1352
1353         read_lock_bh(&addrconf_hash_lock);
1354         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1355                 if (!net_eq(dev_net(ifp->idev->dev), net))
1356                         continue;
1357                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1358                         if (dev == NULL || ifp->idev->dev == dev ||
1359                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1360                                 in6_ifa_hold(ifp);
1361                                 break;
1362                         }
1363                 }
1364         }
1365         read_unlock_bh(&addrconf_hash_lock);
1366
1367         return ifp;
1368 }
1369
1370 /* Gets referenced address, destroys ifaddr */
1371
1372 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1373 {
1374         if (ifp->flags&IFA_F_PERMANENT) {
1375                 spin_lock_bh(&ifp->lock);
1376                 addrconf_del_timer(ifp);
1377                 ifp->flags |= IFA_F_TENTATIVE;
1378                 if (dad_failed)
1379                         ifp->flags |= IFA_F_DADFAILED;
1380                 spin_unlock_bh(&ifp->lock);
1381                 in6_ifa_put(ifp);
1382 #ifdef CONFIG_IPV6_PRIVACY
1383         } else if (ifp->flags&IFA_F_TEMPORARY) {
1384                 struct inet6_ifaddr *ifpub;
1385                 spin_lock_bh(&ifp->lock);
1386                 ifpub = ifp->ifpub;
1387                 if (ifpub) {
1388                         in6_ifa_hold(ifpub);
1389                         spin_unlock_bh(&ifp->lock);
1390                         ipv6_create_tempaddr(ifpub, ifp);
1391                         in6_ifa_put(ifpub);
1392                 } else {
1393                         spin_unlock_bh(&ifp->lock);
1394                 }
1395                 ipv6_del_addr(ifp);
1396 #endif
1397         } else
1398                 ipv6_del_addr(ifp);
1399 }
1400
1401 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1402 {
1403         struct inet6_dev *idev = ifp->idev;
1404
1405         if (net_ratelimit())
1406                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1407                         ifp->idev->dev->name, &ifp->addr);
1408
1409         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1410                 struct in6_addr addr;
1411
1412                 addr.s6_addr32[0] = htonl(0xfe800000);
1413                 addr.s6_addr32[1] = 0;
1414
1415                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1416                     ipv6_addr_equal(&ifp->addr, &addr)) {
1417                         /* DAD failed for link-local based on MAC address */
1418                         idev->cnf.disable_ipv6 = 1;
1419
1420                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1421                                 ifp->idev->dev->name);
1422                 }
1423         }
1424
1425         addrconf_dad_stop(ifp, 1);
1426 }
1427
1428 /* Join to solicited addr multicast group. */
1429
1430 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1431 {
1432         struct in6_addr maddr;
1433
1434         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1435                 return;
1436
1437         addrconf_addr_solict_mult(addr, &maddr);
1438         ipv6_dev_mc_inc(dev, &maddr);
1439 }
1440
1441 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1442 {
1443         struct in6_addr maddr;
1444
1445         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1446                 return;
1447
1448         addrconf_addr_solict_mult(addr, &maddr);
1449         __ipv6_dev_mc_dec(idev, &maddr);
1450 }
1451
1452 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1453 {
1454         struct in6_addr addr;
1455         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1456         if (ipv6_addr_any(&addr))
1457                 return;
1458         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1459 }
1460
1461 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1462 {
1463         struct in6_addr addr;
1464         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1465         if (ipv6_addr_any(&addr))
1466                 return;
1467         __ipv6_dev_ac_dec(ifp->idev, &addr);
1468 }
1469
1470 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1471 {
1472         if (dev->addr_len != ETH_ALEN)
1473                 return -1;
1474         memcpy(eui, dev->dev_addr, 3);
1475         memcpy(eui + 5, dev->dev_addr + 3, 3);
1476
1477         /*
1478          * The zSeries OSA network cards can be shared among various
1479          * OS instances, but the OSA cards have only one MAC address.
1480          * This leads to duplicate address conflicts in conjunction
1481          * with IPv6 if more than one instance uses the same card.
1482          *
1483          * The driver for these cards can deliver a unique 16-bit
1484          * identifier for each instance sharing the same card.  It is
1485          * placed instead of 0xFFFE in the interface identifier.  The
1486          * "u" bit of the interface identifier is not inverted in this
1487          * case.  Hence the resulting interface identifier has local
1488          * scope according to RFC2373.
1489          */
1490         if (dev->dev_id) {
1491                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1492                 eui[4] = dev->dev_id & 0xFF;
1493         } else {
1494                 eui[3] = 0xFF;
1495                 eui[4] = 0xFE;
1496                 eui[0] ^= 2;
1497         }
1498         return 0;
1499 }
1500
1501 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1502 {
1503         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1504         if (dev->addr_len != ARCNET_ALEN)
1505                 return -1;
1506         memset(eui, 0, 7);
1507         eui[7] = *(u8*)dev->dev_addr;
1508         return 0;
1509 }
1510
1511 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1512 {
1513         if (dev->addr_len != INFINIBAND_ALEN)
1514                 return -1;
1515         memcpy(eui, dev->dev_addr + 12, 8);
1516         eui[0] |= 2;
1517         return 0;
1518 }
1519
1520 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1521 {
1522         if (addr == 0)
1523                 return -1;
1524         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1525                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1526                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1527                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1528                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1529                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1530         eui[1] = 0;
1531         eui[2] = 0x5E;
1532         eui[3] = 0xFE;
1533         memcpy(eui + 4, &addr, 4);
1534         return 0;
1535 }
1536 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1537
1538 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1539 {
1540         if (dev->priv_flags & IFF_ISATAP)
1541                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1542         return -1;
1543 }
1544
1545 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1546 {
1547         switch (dev->type) {
1548         case ARPHRD_ETHER:
1549         case ARPHRD_FDDI:
1550         case ARPHRD_IEEE802_TR:
1551                 return addrconf_ifid_eui48(eui, dev);
1552         case ARPHRD_ARCNET:
1553                 return addrconf_ifid_arcnet(eui, dev);
1554         case ARPHRD_INFINIBAND:
1555                 return addrconf_ifid_infiniband(eui, dev);
1556         case ARPHRD_SIT:
1557                 return addrconf_ifid_sit(eui, dev);
1558         }
1559         return -1;
1560 }
1561
1562 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1563 {
1564         int err = -1;
1565         struct inet6_ifaddr *ifp;
1566
1567         read_lock_bh(&idev->lock);
1568         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1569                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1570                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1571                         err = 0;
1572                         break;
1573                 }
1574         }
1575         read_unlock_bh(&idev->lock);
1576         return err;
1577 }
1578
1579 #ifdef CONFIG_IPV6_PRIVACY
1580 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1581 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1582 {
1583 regen:
1584         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1585         idev->rndid[0] &= ~0x02;
1586
1587         /*
1588          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1589          * check if generated address is not inappropriate
1590          *
1591          *  - Reserved subnet anycast (RFC 2526)
1592          *      11111101 11....11 1xxxxxxx
1593          *  - ISATAP (RFC4214) 6.1
1594          *      00-00-5E-FE-xx-xx-xx-xx
1595          *  - value 0
1596          *  - XXX: already assigned to an address on the device
1597          */
1598         if (idev->rndid[0] == 0xfd &&
1599             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1600             (idev->rndid[7]&0x80))
1601                 goto regen;
1602         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1603                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1604                         goto regen;
1605                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1606                         goto regen;
1607         }
1608
1609         return 0;
1610 }
1611
1612 static void ipv6_regen_rndid(unsigned long data)
1613 {
1614         struct inet6_dev *idev = (struct inet6_dev *) data;
1615         unsigned long expires;
1616
1617         rcu_read_lock_bh();
1618         write_lock_bh(&idev->lock);
1619
1620         if (idev->dead)
1621                 goto out;
1622
1623         if (__ipv6_regen_rndid(idev) < 0)
1624                 goto out;
1625
1626         expires = jiffies +
1627                 idev->cnf.temp_prefered_lft * HZ -
1628                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1629         if (time_before(expires, jiffies)) {
1630                 printk(KERN_WARNING
1631                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1632                         idev->dev->name);
1633                 goto out;
1634         }
1635
1636         if (!mod_timer(&idev->regen_timer, expires))
1637                 in6_dev_hold(idev);
1638
1639 out:
1640         write_unlock_bh(&idev->lock);
1641         rcu_read_unlock_bh();
1642         in6_dev_put(idev);
1643 }
1644
1645 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1646         int ret = 0;
1647
1648         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1649                 ret = __ipv6_regen_rndid(idev);
1650         return ret;
1651 }
1652 #endif
1653
1654 /*
1655  *      Add prefix route.
1656  */
1657
1658 static void
1659 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1660                       unsigned long expires, u32 flags)
1661 {
1662         struct fib6_config cfg = {
1663                 .fc_table = RT6_TABLE_PREFIX,
1664                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1665                 .fc_ifindex = dev->ifindex,
1666                 .fc_expires = expires,
1667                 .fc_dst_len = plen,
1668                 .fc_flags = RTF_UP | flags,
1669                 .fc_nlinfo.nl_net = dev_net(dev),
1670                 .fc_protocol = RTPROT_KERNEL,
1671         };
1672
1673         ipv6_addr_copy(&cfg.fc_dst, pfx);
1674
1675         /* Prevent useless cloning on PtP SIT.
1676            This thing is done here expecting that the whole
1677            class of non-broadcast devices need not cloning.
1678          */
1679 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1680         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1681                 cfg.fc_flags |= RTF_NONEXTHOP;
1682 #endif
1683
1684         ip6_route_add(&cfg);
1685 }
1686
1687 /* Create "default" multicast route to the interface */
1688
1689 static void addrconf_add_mroute(struct net_device *dev)
1690 {
1691         struct fib6_config cfg = {
1692                 .fc_table = RT6_TABLE_LOCAL,
1693                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1694                 .fc_ifindex = dev->ifindex,
1695                 .fc_dst_len = 8,
1696                 .fc_flags = RTF_UP,
1697                 .fc_nlinfo.nl_net = dev_net(dev),
1698         };
1699
1700         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1701
1702         ip6_route_add(&cfg);
1703 }
1704
1705 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1706 static void sit_route_add(struct net_device *dev)
1707 {
1708         struct fib6_config cfg = {
1709                 .fc_table = RT6_TABLE_MAIN,
1710                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1711                 .fc_ifindex = dev->ifindex,
1712                 .fc_dst_len = 96,
1713                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1714                 .fc_nlinfo.nl_net = dev_net(dev),
1715         };
1716
1717         /* prefix length - 96 bits "::d.d.d.d" */
1718         ip6_route_add(&cfg);
1719 }
1720 #endif
1721
1722 static void addrconf_add_lroute(struct net_device *dev)
1723 {
1724         struct in6_addr addr;
1725
1726         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1727         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1728 }
1729
1730 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1731 {
1732         struct inet6_dev *idev;
1733
1734         ASSERT_RTNL();
1735
1736         if ((idev = ipv6_find_idev(dev)) == NULL)
1737                 return NULL;
1738
1739         /* Add default multicast route */
1740         addrconf_add_mroute(dev);
1741
1742         /* Add link local route */
1743         addrconf_add_lroute(dev);
1744         return idev;
1745 }
1746
1747 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1748 {
1749         struct prefix_info *pinfo;
1750         __u32 valid_lft;
1751         __u32 prefered_lft;
1752         int addr_type;
1753         struct inet6_dev *in6_dev;
1754         struct net *net = dev_net(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(net, &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
1850                 if (pinfo->prefix_len == 64) {
1851                         memcpy(&addr, &pinfo->prefix, 8);
1852                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1853                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1854                                 in6_dev_put(in6_dev);
1855                                 return;
1856                         }
1857                         goto ok;
1858                 }
1859                 if (net_ratelimit())
1860                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1861                                pinfo->prefix_len);
1862                 in6_dev_put(in6_dev);
1863                 return;
1864
1865 ok:
1866
1867                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1868
1869                 if (ifp == NULL && valid_lft) {
1870                         int max_addresses = in6_dev->cnf.max_addresses;
1871                         u32 addr_flags = 0;
1872
1873 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1874                         if (in6_dev->cnf.optimistic_dad &&
1875                             !net->ipv6.devconf_all->forwarding)
1876                                 addr_flags = IFA_F_OPTIMISTIC;
1877 #endif
1878
1879                         /* Do not allow to create too much of autoconfigured
1880                          * addresses; this would be too easy way to crash kernel.
1881                          */
1882                         if (!max_addresses ||
1883                             ipv6_count_addresses(in6_dev) < max_addresses)
1884                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1885                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1886                                                     addr_flags);
1887
1888                         if (!ifp || IS_ERR(ifp)) {
1889                                 in6_dev_put(in6_dev);
1890                                 return;
1891                         }
1892
1893                         update_lft = create = 1;
1894                         ifp->cstamp = jiffies;
1895                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1896                 }
1897
1898                 if (ifp) {
1899                         int flags;
1900                         unsigned long now;
1901 #ifdef CONFIG_IPV6_PRIVACY
1902                         struct inet6_ifaddr *ift;
1903 #endif
1904                         u32 stored_lft;
1905
1906                         /* update lifetime (RFC2462 5.5.3 e) */
1907                         spin_lock(&ifp->lock);
1908                         now = jiffies;
1909                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1910                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1911                         else
1912                                 stored_lft = 0;
1913                         if (!update_lft && stored_lft) {
1914                                 if (valid_lft > MIN_VALID_LIFETIME ||
1915                                     valid_lft > stored_lft)
1916                                         update_lft = 1;
1917                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1918                                         /* valid_lft <= stored_lft is always true */
1919                                         /*
1920                                          * RFC 4862 Section 5.5.3e:
1921                                          * "Note that the preferred lifetime of
1922                                          *  the corresponding address is always
1923                                          *  reset to the Preferred Lifetime in
1924                                          *  the received Prefix Information
1925                                          *  option, regardless of whether the
1926                                          *  valid lifetime is also reset or
1927                                          *  ignored."
1928                                          *
1929                                          *  So if the preferred lifetime in
1930                                          *  this advertisement is different
1931                                          *  than what we have stored, but the
1932                                          *  valid lifetime is invalid, just
1933                                          *  reset prefered_lft.
1934                                          *
1935                                          *  We must set the valid lifetime
1936                                          *  to the stored lifetime since we'll
1937                                          *  be updating the timestamp below,
1938                                          *  else we'll set it back to the
1939                                          *  minumum.
1940                                          */
1941                                         if (prefered_lft != ifp->prefered_lft) {
1942                                                 valid_lft = stored_lft;
1943                                                 update_lft = 1;
1944                                         }
1945                                 } else {
1946                                         valid_lft = MIN_VALID_LIFETIME;
1947                                         if (valid_lft < prefered_lft)
1948                                                 prefered_lft = valid_lft;
1949                                         update_lft = 1;
1950                                 }
1951                         }
1952
1953                         if (update_lft) {
1954                                 ifp->valid_lft = valid_lft;
1955                                 ifp->prefered_lft = prefered_lft;
1956                                 ifp->tstamp = now;
1957                                 flags = ifp->flags;
1958                                 ifp->flags &= ~IFA_F_DEPRECATED;
1959                                 spin_unlock(&ifp->lock);
1960
1961                                 if (!(flags&IFA_F_TENTATIVE))
1962                                         ipv6_ifa_notify(0, ifp);
1963                         } else
1964                                 spin_unlock(&ifp->lock);
1965
1966 #ifdef CONFIG_IPV6_PRIVACY
1967                         read_lock_bh(&in6_dev->lock);
1968                         /* update all temporary addresses in the list */
1969                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1970                                 /*
1971                                  * When adjusting the lifetimes of an existing
1972                                  * temporary address, only lower the lifetimes.
1973                                  * Implementations must not increase the
1974                                  * lifetimes of an existing temporary address
1975                                  * when processing a Prefix Information Option.
1976                                  */
1977                                 if (ifp != ift->ifpub)
1978                                         continue;
1979
1980                                 spin_lock(&ift->lock);
1981                                 flags = ift->flags;
1982                                 if (ift->valid_lft > valid_lft &&
1983                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1984                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1985                                 if (ift->prefered_lft > prefered_lft &&
1986                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1987                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1988                                 spin_unlock(&ift->lock);
1989                                 if (!(flags&IFA_F_TENTATIVE))
1990                                         ipv6_ifa_notify(0, ift);
1991                         }
1992
1993                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1994                                 /*
1995                                  * When a new public address is created as described in [ADDRCONF],
1996                                  * also create a new temporary address.
1997                                  */
1998                                 read_unlock_bh(&in6_dev->lock);
1999                                 ipv6_create_tempaddr(ifp, NULL);
2000                         } else {
2001                                 read_unlock_bh(&in6_dev->lock);
2002                         }
2003 #endif
2004                         in6_ifa_put(ifp);
2005                         addrconf_verify(0);
2006                 }
2007         }
2008         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2009         in6_dev_put(in6_dev);
2010 }
2011
2012 /*
2013  *      Set destination address.
2014  *      Special case for SIT interfaces where we create a new "virtual"
2015  *      device.
2016  */
2017 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2018 {
2019         struct in6_ifreq ireq;
2020         struct net_device *dev;
2021         int err = -EINVAL;
2022
2023         rtnl_lock();
2024
2025         err = -EFAULT;
2026         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2027                 goto err_exit;
2028
2029         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2030
2031         err = -ENODEV;
2032         if (dev == NULL)
2033                 goto err_exit;
2034
2035 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2036         if (dev->type == ARPHRD_SIT) {
2037                 const struct net_device_ops *ops = dev->netdev_ops;
2038                 struct ifreq ifr;
2039                 struct ip_tunnel_parm p;
2040
2041                 err = -EADDRNOTAVAIL;
2042                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2043                         goto err_exit;
2044
2045                 memset(&p, 0, sizeof(p));
2046                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2047                 p.iph.saddr = 0;
2048                 p.iph.version = 4;
2049                 p.iph.ihl = 5;
2050                 p.iph.protocol = IPPROTO_IPV6;
2051                 p.iph.ttl = 64;
2052                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2053
2054                 if (ops->ndo_do_ioctl) {
2055                         mm_segment_t oldfs = get_fs();
2056
2057                         set_fs(KERNEL_DS);
2058                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2059                         set_fs(oldfs);
2060                 } else
2061                         err = -EOPNOTSUPP;
2062
2063                 if (err == 0) {
2064                         err = -ENOBUFS;
2065                         dev = __dev_get_by_name(net, p.name);
2066                         if (!dev)
2067                                 goto err_exit;
2068                         err = dev_open(dev);
2069                 }
2070         }
2071 #endif
2072
2073 err_exit:
2074         rtnl_unlock();
2075         return err;
2076 }
2077
2078 /*
2079  *      Manual configuration of address on an interface
2080  */
2081 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2082                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2083                           __u32 valid_lft)
2084 {
2085         struct inet6_ifaddr *ifp;
2086         struct inet6_dev *idev;
2087         struct net_device *dev;
2088         int scope;
2089         u32 flags;
2090         clock_t expires;
2091         unsigned long timeout;
2092
2093         ASSERT_RTNL();
2094
2095         if (plen > 128)
2096                 return -EINVAL;
2097
2098         /* check the lifetime */
2099         if (!valid_lft || prefered_lft > valid_lft)
2100                 return -EINVAL;
2101
2102         dev = __dev_get_by_index(net, ifindex);
2103         if (!dev)
2104                 return -ENODEV;
2105
2106         if ((idev = addrconf_add_dev(dev)) == NULL)
2107                 return -ENOBUFS;
2108
2109         scope = ipv6_addr_scope(pfx);
2110
2111         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2112         if (addrconf_finite_timeout(timeout)) {
2113                 expires = jiffies_to_clock_t(timeout * HZ);
2114                 valid_lft = timeout;
2115                 flags = RTF_EXPIRES;
2116         } else {
2117                 expires = 0;
2118                 flags = 0;
2119                 ifa_flags |= IFA_F_PERMANENT;
2120         }
2121
2122         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2123         if (addrconf_finite_timeout(timeout)) {
2124                 if (timeout == 0)
2125                         ifa_flags |= IFA_F_DEPRECATED;
2126                 prefered_lft = timeout;
2127         }
2128
2129         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2130
2131         if (!IS_ERR(ifp)) {
2132                 spin_lock_bh(&ifp->lock);
2133                 ifp->valid_lft = valid_lft;
2134                 ifp->prefered_lft = prefered_lft;
2135                 ifp->tstamp = jiffies;
2136                 spin_unlock_bh(&ifp->lock);
2137
2138                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2139                                       expires, flags);
2140                 /*
2141                  * Note that section 3.1 of RFC 4429 indicates
2142                  * that the Optimistic flag should not be set for
2143                  * manually configured addresses
2144                  */
2145                 addrconf_dad_start(ifp, 0);
2146                 in6_ifa_put(ifp);
2147                 addrconf_verify(0);
2148                 return 0;
2149         }
2150
2151         return PTR_ERR(ifp);
2152 }
2153
2154 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2155                           unsigned int plen)
2156 {
2157         struct inet6_ifaddr *ifp;
2158         struct inet6_dev *idev;
2159         struct net_device *dev;
2160
2161         if (plen > 128)
2162                 return -EINVAL;
2163
2164         dev = __dev_get_by_index(net, ifindex);
2165         if (!dev)
2166                 return -ENODEV;
2167
2168         if ((idev = __in6_dev_get(dev)) == NULL)
2169                 return -ENXIO;
2170
2171         read_lock_bh(&idev->lock);
2172         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2173                 if (ifp->prefix_len == plen &&
2174                     ipv6_addr_equal(pfx, &ifp->addr)) {
2175                         in6_ifa_hold(ifp);
2176                         read_unlock_bh(&idev->lock);
2177
2178                         ipv6_del_addr(ifp);
2179
2180                         /* If the last address is deleted administratively,
2181                            disable IPv6 on this interface.
2182                          */
2183                         if (idev->addr_list == NULL)
2184                                 addrconf_ifdown(idev->dev, 1);
2185                         return 0;
2186                 }
2187         }
2188         read_unlock_bh(&idev->lock);
2189         return -EADDRNOTAVAIL;
2190 }
2191
2192
2193 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2194 {
2195         struct in6_ifreq ireq;
2196         int err;
2197
2198         if (!capable(CAP_NET_ADMIN))
2199                 return -EPERM;
2200
2201         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2202                 return -EFAULT;
2203
2204         rtnl_lock();
2205         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2206                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2207                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2208         rtnl_unlock();
2209         return err;
2210 }
2211
2212 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2213 {
2214         struct in6_ifreq ireq;
2215         int err;
2216
2217         if (!capable(CAP_NET_ADMIN))
2218                 return -EPERM;
2219
2220         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2221                 return -EFAULT;
2222
2223         rtnl_lock();
2224         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2225                              ireq.ifr6_prefixlen);
2226         rtnl_unlock();
2227         return err;
2228 }
2229
2230 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2231                      int plen, int scope)
2232 {
2233         struct inet6_ifaddr *ifp;
2234
2235         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2236         if (!IS_ERR(ifp)) {
2237                 spin_lock_bh(&ifp->lock);
2238                 ifp->flags &= ~IFA_F_TENTATIVE;
2239                 spin_unlock_bh(&ifp->lock);
2240                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2241                 in6_ifa_put(ifp);
2242         }
2243 }
2244
2245 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2246 static void sit_add_v4_addrs(struct inet6_dev *idev)
2247 {
2248         struct in6_addr addr;
2249         struct net_device *dev;
2250         struct net *net = dev_net(idev->dev);
2251         int scope;
2252
2253         ASSERT_RTNL();
2254
2255         memset(&addr, 0, sizeof(struct in6_addr));
2256         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2257
2258         if (idev->dev->flags&IFF_POINTOPOINT) {
2259                 addr.s6_addr32[0] = htonl(0xfe800000);
2260                 scope = IFA_LINK;
2261         } else {
2262                 scope = IPV6_ADDR_COMPATv4;
2263         }
2264
2265         if (addr.s6_addr32[3]) {
2266                 add_addr(idev, &addr, 128, scope);
2267                 return;
2268         }
2269
2270         for_each_netdev(net, dev) {
2271                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2272                 if (in_dev && (dev->flags & IFF_UP)) {
2273                         struct in_ifaddr * ifa;
2274
2275                         int flag = scope;
2276
2277                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2278                                 int plen;
2279
2280                                 addr.s6_addr32[3] = ifa->ifa_local;
2281
2282                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2283                                         continue;
2284                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2285                                         if (idev->dev->flags&IFF_POINTOPOINT)
2286                                                 continue;
2287                                         flag |= IFA_HOST;
2288                                 }
2289                                 if (idev->dev->flags&IFF_POINTOPOINT)
2290                                         plen = 64;
2291                                 else
2292                                         plen = 96;
2293
2294                                 add_addr(idev, &addr, plen, flag);
2295                         }
2296                 }
2297         }
2298 }
2299 #endif
2300
2301 static void init_loopback(struct net_device *dev)
2302 {
2303         struct inet6_dev  *idev;
2304
2305         /* ::1 */
2306
2307         ASSERT_RTNL();
2308
2309         if ((idev = ipv6_find_idev(dev)) == NULL) {
2310                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2311                 return;
2312         }
2313
2314         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2315 }
2316
2317 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2318 {
2319         struct inet6_ifaddr * ifp;
2320         u32 addr_flags = IFA_F_PERMANENT;
2321
2322 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2323         if (idev->cnf.optimistic_dad &&
2324             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2325                 addr_flags |= IFA_F_OPTIMISTIC;
2326 #endif
2327
2328
2329         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2330         if (!IS_ERR(ifp)) {
2331                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2332                 addrconf_dad_start(ifp, 0);
2333                 in6_ifa_put(ifp);
2334         }
2335 }
2336
2337 static void addrconf_dev_config(struct net_device *dev)
2338 {
2339         struct in6_addr addr;
2340         struct inet6_dev    * idev;
2341
2342         ASSERT_RTNL();
2343
2344         if ((dev->type != ARPHRD_ETHER) &&
2345             (dev->type != ARPHRD_FDDI) &&
2346             (dev->type != ARPHRD_IEEE802_TR) &&
2347             (dev->type != ARPHRD_ARCNET) &&
2348             (dev->type != ARPHRD_INFINIBAND)) {
2349                 /* Alas, we support only Ethernet autoconfiguration. */
2350                 return;
2351         }
2352
2353         idev = addrconf_add_dev(dev);
2354         if (idev == NULL)
2355                 return;
2356
2357         memset(&addr, 0, sizeof(struct in6_addr));
2358         addr.s6_addr32[0] = htonl(0xFE800000);
2359
2360         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2361                 addrconf_add_linklocal(idev, &addr);
2362 }
2363
2364 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2365 static void addrconf_sit_config(struct net_device *dev)
2366 {
2367         struct inet6_dev *idev;
2368
2369         ASSERT_RTNL();
2370
2371         /*
2372          * Configure the tunnel with one of our IPv4
2373          * addresses... we should configure all of
2374          * our v4 addrs in the tunnel
2375          */
2376
2377         if ((idev = ipv6_find_idev(dev)) == NULL) {
2378                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2379                 return;
2380         }
2381
2382         if (dev->priv_flags & IFF_ISATAP) {
2383                 struct in6_addr addr;
2384
2385                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2386                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2387                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2388                         addrconf_add_linklocal(idev, &addr);
2389                 return;
2390         }
2391
2392         sit_add_v4_addrs(idev);
2393
2394         if (dev->flags&IFF_POINTOPOINT) {
2395                 addrconf_add_mroute(dev);
2396                 addrconf_add_lroute(dev);
2397         } else
2398                 sit_route_add(dev);
2399 }
2400 #endif
2401
2402 static inline int
2403 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2404 {
2405         struct in6_addr lladdr;
2406
2407         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2408                 addrconf_add_linklocal(idev, &lladdr);
2409                 return 0;
2410         }
2411         return -1;
2412 }
2413
2414 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2415 {
2416         struct net_device *link_dev;
2417         struct net *net = dev_net(idev->dev);
2418
2419         /* first try to inherit the link-local address from the link device */
2420         if (idev->dev->iflink &&
2421             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2422                 if (!ipv6_inherit_linklocal(idev, link_dev))
2423                         return;
2424         }
2425         /* then try to inherit it from any device */
2426         for_each_netdev(net, link_dev) {
2427                 if (!ipv6_inherit_linklocal(idev, link_dev))
2428                         return;
2429         }
2430         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2431 }
2432
2433 /*
2434  * Autoconfigure tunnel with a link-local address so routing protocols,
2435  * DHCPv6, MLD etc. can be run over the virtual link
2436  */
2437
2438 static void addrconf_ip6_tnl_config(struct net_device *dev)
2439 {
2440         struct inet6_dev *idev;
2441
2442         ASSERT_RTNL();
2443
2444         if ((idev = addrconf_add_dev(dev)) == NULL) {
2445                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2446                 return;
2447         }
2448         ip6_tnl_add_linklocal(idev);
2449 }
2450
2451 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2452                            void * data)
2453 {
2454         struct net_device *dev = (struct net_device *) data;
2455         struct inet6_dev *idev = __in6_dev_get(dev);
2456         int run_pending = 0;
2457         int err;
2458
2459         switch(event) {
2460         case NETDEV_REGISTER:
2461                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2462                         idev = ipv6_add_dev(dev);
2463                         if (!idev)
2464                                 return notifier_from_errno(-ENOMEM);
2465                 }
2466                 break;
2467         case NETDEV_UP:
2468         case NETDEV_CHANGE:
2469                 if (dev->flags & IFF_SLAVE)
2470                         break;
2471
2472                 if (event == NETDEV_UP) {
2473                         if (!addrconf_qdisc_ok(dev)) {
2474                                 /* device is not ready yet. */
2475                                 printk(KERN_INFO
2476                                         "ADDRCONF(NETDEV_UP): %s: "
2477                                         "link is not ready\n",
2478                                         dev->name);
2479                                 break;
2480                         }
2481
2482                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2483                                 idev = ipv6_add_dev(dev);
2484
2485                         if (idev) {
2486                                 idev->if_flags |= IF_READY;
2487                                 run_pending = 1;
2488                         }
2489                 } else {
2490                         if (!addrconf_qdisc_ok(dev)) {
2491                                 /* device is still not ready. */
2492                                 break;
2493                         }
2494
2495                         if (idev) {
2496                                 if (idev->if_flags & IF_READY) {
2497                                         /* device is already configured. */
2498                                         break;
2499                                 }
2500                                 idev->if_flags |= IF_READY;
2501                         }
2502
2503                         printk(KERN_INFO
2504                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2505                                         "link becomes ready\n",
2506                                         dev->name);
2507
2508                         run_pending = 1;
2509                 }
2510
2511                 switch(dev->type) {
2512 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2513                 case ARPHRD_SIT:
2514                         addrconf_sit_config(dev);
2515                         break;
2516 #endif
2517                 case ARPHRD_TUNNEL6:
2518                         addrconf_ip6_tnl_config(dev);
2519                         break;
2520                 case ARPHRD_LOOPBACK:
2521                         init_loopback(dev);
2522                         break;
2523
2524                 default:
2525                         addrconf_dev_config(dev);
2526                         break;
2527                 }
2528                 if (idev) {
2529                         if (run_pending)
2530                                 addrconf_dad_run(idev);
2531
2532                         /* If the MTU changed during the interface down, when the
2533                            interface up, the changed MTU must be reflected in the
2534                            idev as well as routers.
2535                          */
2536                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2537                                 rt6_mtu_change(dev, dev->mtu);
2538                                 idev->cnf.mtu6 = dev->mtu;
2539                         }
2540                         idev->tstamp = jiffies;
2541                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2542                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2543                            stop IPv6 on this interface.
2544                          */
2545                         if (dev->mtu < IPV6_MIN_MTU)
2546                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2547                 }
2548                 break;
2549
2550         case NETDEV_CHANGEMTU:
2551                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2552                         rt6_mtu_change(dev, dev->mtu);
2553                         idev->cnf.mtu6 = dev->mtu;
2554                         break;
2555                 }
2556
2557                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2558                         idev = ipv6_add_dev(dev);
2559                         if (idev)
2560                                 break;
2561                 }
2562
2563                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2564
2565         case NETDEV_DOWN:
2566         case NETDEV_UNREGISTER:
2567                 /*
2568                  *      Remove all addresses from this interface.
2569                  */
2570                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2571                 break;
2572
2573         case NETDEV_CHANGENAME:
2574                 if (idev) {
2575                         snmp6_unregister_dev(idev);
2576                         addrconf_sysctl_unregister(idev);
2577                         addrconf_sysctl_register(idev);
2578                         err = snmp6_register_dev(idev);
2579                         if (err)
2580                                 return notifier_from_errno(err);
2581                 }
2582                 break;
2583         case NETDEV_BONDING_OLDTYPE:
2584         case NETDEV_BONDING_NEWTYPE:
2585                 addrconf_bonding_change(dev, event);
2586                 break;
2587         }
2588
2589         return NOTIFY_OK;
2590 }
2591
2592 /*
2593  *      addrconf module should be notified of a device going up
2594  */
2595 static struct notifier_block ipv6_dev_notf = {
2596         .notifier_call = addrconf_notify,
2597         .priority = 0
2598 };
2599
2600 static void addrconf_bonding_change(struct net_device *dev, unsigned long event)
2601 {
2602         struct inet6_dev *idev;
2603         ASSERT_RTNL();
2604
2605         idev = __in6_dev_get(dev);
2606
2607         if (event == NETDEV_BONDING_NEWTYPE)
2608                 ipv6_mc_remap(idev);
2609         else if (event == NETDEV_BONDING_OLDTYPE)
2610                 ipv6_mc_unmap(idev);
2611 }
2612
2613 static int addrconf_ifdown(struct net_device *dev, int how)
2614 {
2615         struct inet6_dev *idev;
2616         struct inet6_ifaddr *ifa, **bifa;
2617         struct net *net = dev_net(dev);
2618         int i;
2619
2620         ASSERT_RTNL();
2621
2622         rt6_ifdown(net, dev);
2623         neigh_ifdown(&nd_tbl, dev);
2624
2625         idev = __in6_dev_get(dev);
2626         if (idev == NULL)
2627                 return -ENODEV;
2628
2629         /* Step 1: remove reference to ipv6 device from parent device.
2630                    Do not dev_put!
2631          */
2632         if (how) {
2633                 idev->dead = 1;
2634
2635                 /* protected by rtnl_lock */
2636                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2637
2638                 /* Step 1.5: remove snmp6 entry */
2639                 snmp6_unregister_dev(idev);
2640
2641         }
2642
2643         /* Step 2: clear hash table */
2644         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2645                 bifa = &inet6_addr_lst[i];
2646
2647                 write_lock_bh(&addrconf_hash_lock);
2648                 while ((ifa = *bifa) != NULL) {
2649                         if (ifa->idev == idev &&
2650                             (how || !(ifa->flags&IFA_F_PERMANENT))) {
2651                                 *bifa = ifa->lst_next;
2652                                 ifa->lst_next = NULL;
2653                                 addrconf_del_timer(ifa);
2654                                 in6_ifa_put(ifa);
2655                                 continue;
2656                         }
2657                         bifa = &ifa->lst_next;
2658                 }
2659                 write_unlock_bh(&addrconf_hash_lock);
2660         }
2661
2662         write_lock_bh(&idev->lock);
2663
2664         /* Step 3: clear flags for stateless addrconf */
2665         if (!how)
2666                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2667
2668         /* Step 4: clear address list */
2669 #ifdef CONFIG_IPV6_PRIVACY
2670         if (how && del_timer(&idev->regen_timer))
2671                 in6_dev_put(idev);
2672
2673         /* clear tempaddr list */
2674         while ((ifa = idev->tempaddr_list) != NULL) {
2675                 idev->tempaddr_list = ifa->tmp_next;
2676                 ifa->tmp_next = NULL;
2677                 ifa->dead = 1;
2678                 write_unlock_bh(&idev->lock);
2679                 spin_lock_bh(&ifa->lock);
2680
2681                 if (ifa->ifpub) {
2682                         in6_ifa_put(ifa->ifpub);
2683                         ifa->ifpub = NULL;
2684                 }
2685                 spin_unlock_bh(&ifa->lock);
2686                 in6_ifa_put(ifa);
2687                 write_lock_bh(&idev->lock);
2688         }
2689 #endif
2690         bifa = &idev->addr_list;
2691         while ((ifa = *bifa) != NULL) {
2692                 if (how == 0 && (ifa->flags&IFA_F_PERMANENT)) {
2693                         /* Retain permanent address on admin down */
2694                         bifa = &ifa->if_next;
2695
2696                         /* Restart DAD if needed when link comes back up */
2697                         if ( !((dev->flags&(IFF_NOARP|IFF_LOOPBACK)) ||
2698                                idev->cnf.accept_dad <= 0 ||
2699                                (ifa->flags & IFA_F_NODAD)))
2700                                 ifa->flags |= IFA_F_TENTATIVE;
2701                 } else {
2702                         *bifa = ifa->if_next;
2703                         ifa->if_next = NULL;
2704
2705                         ifa->dead = 1;
2706                         write_unlock_bh(&idev->lock);
2707
2708                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2709                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2710                         in6_ifa_put(ifa);
2711
2712                         write_lock_bh(&idev->lock);
2713                 }
2714         }
2715         write_unlock_bh(&idev->lock);
2716
2717         /* Step 5: Discard multicast list */
2718
2719         if (how)
2720                 ipv6_mc_destroy_dev(idev);
2721         else
2722                 ipv6_mc_down(idev);
2723
2724         idev->tstamp = jiffies;
2725
2726         /* Shot the device (if unregistered) */
2727
2728         if (how) {
2729                 addrconf_sysctl_unregister(idev);
2730                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2731                 neigh_ifdown(&nd_tbl, dev);
2732                 in6_dev_put(idev);
2733         }
2734         return 0;
2735 }
2736
2737 static void addrconf_rs_timer(unsigned long data)
2738 {
2739         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2740
2741         if (ifp->idev->cnf.forwarding)
2742                 goto out;
2743
2744         if (ifp->idev->if_flags & IF_RA_RCVD) {
2745                 /*
2746                  *      Announcement received after solicitation
2747                  *      was sent
2748                  */
2749                 goto out;
2750         }
2751
2752         spin_lock(&ifp->lock);
2753         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2754                 /* The wait after the last probe can be shorter */
2755                 addrconf_mod_timer(ifp, AC_RS,
2756                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2757                                    ifp->idev->cnf.rtr_solicit_delay :
2758                                    ifp->idev->cnf.rtr_solicit_interval);
2759                 spin_unlock(&ifp->lock);
2760
2761                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2762         } else {
2763                 spin_unlock(&ifp->lock);
2764                 /*
2765                  * Note: we do not support deprecated "all on-link"
2766                  * assumption any longer.
2767                  */
2768                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2769                        ifp->idev->dev->name);
2770         }
2771
2772 out:
2773         in6_ifa_put(ifp);
2774 }
2775
2776 /*
2777  *      Duplicate Address Detection
2778  */
2779 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2780 {
2781         unsigned long rand_num;
2782         struct inet6_dev *idev = ifp->idev;
2783
2784         if (ifp->flags & IFA_F_OPTIMISTIC)
2785                 rand_num = 0;
2786         else
2787                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2788
2789         ifp->probes = idev->cnf.dad_transmits;
2790         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2791 }
2792
2793 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2794 {
2795         struct inet6_dev *idev = ifp->idev;
2796         struct net_device *dev = idev->dev;
2797
2798         addrconf_join_solict(dev, &ifp->addr);
2799
2800         net_srandom(ifp->addr.s6_addr32[3]);
2801
2802         read_lock_bh(&idev->lock);
2803         if (ifp->dead)
2804                 goto out;
2805
2806         spin_lock(&ifp->lock);
2807         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2808             idev->cnf.accept_dad < 1 ||
2809             !(ifp->flags&IFA_F_TENTATIVE) ||
2810             ifp->flags & IFA_F_NODAD) {
2811                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2812                 spin_unlock(&ifp->lock);
2813                 read_unlock_bh(&idev->lock);
2814
2815                 addrconf_dad_completed(ifp);
2816                 return;
2817         }
2818
2819         if (!(idev->if_flags & IF_READY)) {
2820                 spin_unlock(&ifp->lock);
2821                 read_unlock_bh(&idev->lock);
2822                 /*
2823                  * If the device is not ready:
2824                  * - keep it tentative if it is a permanent address.
2825                  * - otherwise, kill it.
2826                  */
2827                 in6_ifa_hold(ifp);
2828                 addrconf_dad_stop(ifp, 0);
2829                 return;
2830         }
2831
2832         /*
2833          * Optimistic nodes can start receiving
2834          * Frames right away
2835          */
2836         if(ifp->flags & IFA_F_OPTIMISTIC)
2837                 ip6_ins_rt(ifp->rt);
2838
2839         addrconf_dad_kick(ifp);
2840         spin_unlock(&ifp->lock);
2841 out:
2842         read_unlock_bh(&idev->lock);
2843 }
2844
2845 static void addrconf_dad_timer(unsigned long data)
2846 {
2847         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2848         struct inet6_dev *idev = ifp->idev;
2849         struct in6_addr mcaddr;
2850
2851         read_lock_bh(&idev->lock);
2852         if (idev->dead) {
2853                 read_unlock_bh(&idev->lock);
2854                 goto out;
2855         }
2856
2857         spin_lock(&ifp->lock);
2858         if (ifp->probes == 0) {
2859                 /*
2860                  * DAD was successful
2861                  */
2862
2863                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2864                 spin_unlock(&ifp->lock);
2865                 read_unlock_bh(&idev->lock);
2866
2867                 addrconf_dad_completed(ifp);
2868
2869                 goto out;
2870         }
2871
2872         ifp->probes--;
2873         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2874         spin_unlock(&ifp->lock);
2875         read_unlock_bh(&idev->lock);
2876
2877         /* send a neighbour solicitation for our addr */
2878         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2879         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2880 out:
2881         in6_ifa_put(ifp);
2882 }
2883
2884 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2885 {
2886         struct net_device *     dev = ifp->idev->dev;
2887
2888         /*
2889          *      Configure the address for reception. Now it is valid.
2890          */
2891
2892         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2893
2894         /* If added prefix is link local and forwarding is off,
2895            start sending router solicitations.
2896          */
2897
2898         if (ifp->idev->cnf.forwarding == 0 &&
2899             ifp->idev->cnf.rtr_solicits > 0 &&
2900             (dev->flags&IFF_LOOPBACK) == 0 &&
2901             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2902                 /*
2903                  *      If a host as already performed a random delay
2904                  *      [...] as part of DAD [...] there is no need
2905                  *      to delay again before sending the first RS
2906                  */
2907                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2908
2909                 spin_lock_bh(&ifp->lock);
2910                 ifp->probes = 1;
2911                 ifp->idev->if_flags |= IF_RS_SENT;
2912                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2913                 spin_unlock_bh(&ifp->lock);
2914         }
2915 }
2916
2917 static void addrconf_dad_run(struct inet6_dev *idev) {
2918         struct inet6_ifaddr *ifp;
2919
2920         read_lock_bh(&idev->lock);
2921         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2922                 spin_lock(&ifp->lock);
2923                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2924                         spin_unlock(&ifp->lock);
2925                         continue;
2926                 }
2927                 spin_unlock(&ifp->lock);
2928                 addrconf_dad_kick(ifp);
2929         }
2930         read_unlock_bh(&idev->lock);
2931 }
2932
2933 #ifdef CONFIG_PROC_FS
2934 struct if6_iter_state {
2935         struct seq_net_private p;
2936         int bucket;
2937 };
2938
2939 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2940 {
2941         struct inet6_ifaddr *ifa = NULL;
2942         struct if6_iter_state *state = seq->private;
2943         struct net *net = seq_file_net(seq);
2944
2945         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2946                 ifa = inet6_addr_lst[state->bucket];
2947
2948                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2949                         ifa = ifa->lst_next;
2950                 if (ifa)
2951                         break;
2952         }
2953         return ifa;
2954 }
2955
2956 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2957 {
2958         struct if6_iter_state *state = seq->private;
2959         struct net *net = seq_file_net(seq);
2960
2961         ifa = ifa->lst_next;
2962 try_again:
2963         if (ifa) {
2964                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2965                         ifa = ifa->lst_next;
2966                         goto try_again;
2967                 }
2968         }
2969
2970         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2971                 ifa = inet6_addr_lst[state->bucket];
2972                 goto try_again;
2973         }
2974
2975         return ifa;
2976 }
2977
2978 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2979 {
2980         struct inet6_ifaddr *ifa = if6_get_first(seq);
2981
2982         if (ifa)
2983                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2984                         --pos;
2985         return pos ? NULL : ifa;
2986 }
2987
2988 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2989         __acquires(addrconf_hash_lock)
2990 {
2991         read_lock_bh(&addrconf_hash_lock);
2992         return if6_get_idx(seq, *pos);
2993 }
2994
2995 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2996 {
2997         struct inet6_ifaddr *ifa;
2998
2999         ifa = if6_get_next(seq, v);
3000         ++*pos;
3001         return ifa;
3002 }
3003
3004 static void if6_seq_stop(struct seq_file *seq, void *v)
3005         __releases(addrconf_hash_lock)
3006 {
3007         read_unlock_bh(&addrconf_hash_lock);
3008 }
3009
3010 static int if6_seq_show(struct seq_file *seq, void *v)
3011 {
3012         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
3013         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
3014                    &ifp->addr,
3015                    ifp->idev->dev->ifindex,
3016                    ifp->prefix_len,
3017                    ifp->scope,
3018                    ifp->flags,
3019                    ifp->idev->dev->name);
3020         return 0;
3021 }
3022
3023 static const struct seq_operations if6_seq_ops = {
3024         .start  = if6_seq_start,
3025         .next   = if6_seq_next,
3026         .show   = if6_seq_show,
3027         .stop   = if6_seq_stop,
3028 };
3029
3030 static int if6_seq_open(struct inode *inode, struct file *file)
3031 {
3032         return seq_open_net(inode, file, &if6_seq_ops,
3033                             sizeof(struct if6_iter_state));
3034 }
3035
3036 static const struct file_operations if6_fops = {
3037         .owner          = THIS_MODULE,
3038         .open           = if6_seq_open,
3039         .read           = seq_read,
3040         .llseek         = seq_lseek,
3041         .release        = seq_release_net,
3042 };
3043
3044 static int __net_init if6_proc_net_init(struct net *net)
3045 {
3046         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3047                 return -ENOMEM;
3048         return 0;
3049 }
3050
3051 static void __net_exit if6_proc_net_exit(struct net *net)
3052 {
3053        proc_net_remove(net, "if_inet6");
3054 }
3055
3056 static struct pernet_operations if6_proc_net_ops = {
3057        .init = if6_proc_net_init,
3058        .exit = if6_proc_net_exit,
3059 };
3060
3061 int __init if6_proc_init(void)
3062 {
3063         return register_pernet_subsys(&if6_proc_net_ops);
3064 }
3065
3066 void if6_proc_exit(void)
3067 {
3068         unregister_pernet_subsys(&if6_proc_net_ops);
3069 }
3070 #endif  /* CONFIG_PROC_FS */
3071
3072 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3073 /* Check if address is a home address configured on any interface. */
3074 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3075 {
3076         int ret = 0;
3077         struct inet6_ifaddr * ifp;
3078         u8 hash = ipv6_addr_hash(addr);
3079         read_lock_bh(&addrconf_hash_lock);
3080         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3081                 if (!net_eq(dev_net(ifp->idev->dev), net))
3082                         continue;
3083                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3084                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3085                         ret = 1;
3086                         break;
3087                 }
3088         }
3089         read_unlock_bh(&addrconf_hash_lock);
3090         return ret;
3091 }
3092 #endif
3093
3094 /*
3095  *      Periodic address status verification
3096  */
3097
3098 static void addrconf_verify(unsigned long foo)
3099 {
3100         struct inet6_ifaddr *ifp;
3101         unsigned long now, next;
3102         int i;
3103
3104         spin_lock_bh(&addrconf_verify_lock);
3105         now = jiffies;
3106         next = now + ADDR_CHECK_FREQUENCY;
3107
3108         del_timer(&addr_chk_timer);
3109
3110         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3111
3112 restart:
3113                 read_lock(&addrconf_hash_lock);
3114                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3115                         unsigned long age;
3116 #ifdef CONFIG_IPV6_PRIVACY
3117                         unsigned long regen_advance;
3118 #endif
3119
3120                         if (ifp->flags & IFA_F_PERMANENT)
3121                                 continue;
3122
3123                         spin_lock(&ifp->lock);
3124                         age = (now - ifp->tstamp) / HZ;
3125
3126 #ifdef CONFIG_IPV6_PRIVACY
3127                         regen_advance = ifp->idev->cnf.regen_max_retry *
3128                                         ifp->idev->cnf.dad_transmits *
3129                                         ifp->idev->nd_parms->retrans_time / HZ;
3130 #endif
3131
3132                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3133                             age >= ifp->valid_lft) {
3134                                 spin_unlock(&ifp->lock);
3135                                 in6_ifa_hold(ifp);
3136                                 read_unlock(&addrconf_hash_lock);
3137                                 ipv6_del_addr(ifp);
3138                                 goto restart;
3139                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3140                                 spin_unlock(&ifp->lock);
3141                                 continue;
3142                         } else if (age >= ifp->prefered_lft) {
3143                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3144                                 int deprecate = 0;
3145
3146                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3147                                         deprecate = 1;
3148                                         ifp->flags |= IFA_F_DEPRECATED;
3149                                 }
3150
3151                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3152                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3153
3154                                 spin_unlock(&ifp->lock);
3155
3156                                 if (deprecate) {
3157                                         in6_ifa_hold(ifp);
3158                                         read_unlock(&addrconf_hash_lock);
3159
3160                                         ipv6_ifa_notify(0, ifp);
3161                                         in6_ifa_put(ifp);
3162                                         goto restart;
3163                                 }
3164 #ifdef CONFIG_IPV6_PRIVACY
3165                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3166                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3167                                 if (age >= ifp->prefered_lft - regen_advance) {
3168                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3169                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3170                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3171                                         if (!ifp->regen_count && ifpub) {
3172                                                 ifp->regen_count++;
3173                                                 in6_ifa_hold(ifp);
3174                                                 in6_ifa_hold(ifpub);
3175                                                 spin_unlock(&ifp->lock);
3176                                                 read_unlock(&addrconf_hash_lock);
3177                                                 spin_lock(&ifpub->lock);
3178                                                 ifpub->regen_count = 0;
3179                                                 spin_unlock(&ifpub->lock);
3180                                                 ipv6_create_tempaddr(ifpub, ifp);
3181                                                 in6_ifa_put(ifpub);
3182                                                 in6_ifa_put(ifp);
3183                                                 goto restart;
3184                                         }
3185                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3186                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3187                                 spin_unlock(&ifp->lock);
3188 #endif
3189                         } else {
3190                                 /* ifp->prefered_lft <= ifp->valid_lft */
3191                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3192                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3193                                 spin_unlock(&ifp->lock);
3194                         }
3195                 }
3196                 read_unlock(&addrconf_hash_lock);
3197         }
3198
3199         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3200         add_timer(&addr_chk_timer);
3201         spin_unlock_bh(&addrconf_verify_lock);
3202 }
3203
3204 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3205 {
3206         struct in6_addr *pfx = NULL;
3207
3208         if (addr)
3209                 pfx = nla_data(addr);
3210
3211         if (local) {
3212                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3213                         pfx = NULL;
3214                 else
3215                         pfx = nla_data(local);
3216         }
3217
3218         return pfx;
3219 }
3220
3221 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3222         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3223         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3224         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3225 };
3226
3227 static int
3228 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3229 {
3230         struct net *net = sock_net(skb->sk);
3231         struct ifaddrmsg *ifm;
3232         struct nlattr *tb[IFA_MAX+1];
3233         struct in6_addr *pfx;
3234         int err;
3235
3236         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3237         if (err < 0)
3238                 return err;
3239
3240         ifm = nlmsg_data(nlh);
3241         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3242         if (pfx == NULL)
3243                 return -EINVAL;
3244
3245         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3246 }
3247
3248 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3249                              u32 prefered_lft, u32 valid_lft)
3250 {
3251         u32 flags;
3252         clock_t expires;
3253         unsigned long timeout;
3254
3255         if (!valid_lft || (prefered_lft > valid_lft))
3256                 return -EINVAL;
3257
3258         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3259         if (addrconf_finite_timeout(timeout)) {
3260                 expires = jiffies_to_clock_t(timeout * HZ);
3261                 valid_lft = timeout;
3262                 flags = RTF_EXPIRES;
3263         } else {
3264                 expires = 0;
3265                 flags = 0;
3266                 ifa_flags |= IFA_F_PERMANENT;
3267         }
3268
3269         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3270         if (addrconf_finite_timeout(timeout)) {
3271                 if (timeout == 0)
3272                         ifa_flags |= IFA_F_DEPRECATED;
3273                 prefered_lft = timeout;
3274         }
3275
3276         spin_lock_bh(&ifp->lock);
3277         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3278         ifp->tstamp = jiffies;
3279         ifp->valid_lft = valid_lft;
3280         ifp->prefered_lft = prefered_lft;
3281
3282         spin_unlock_bh(&ifp->lock);
3283         if (!(ifp->flags&IFA_F_TENTATIVE))
3284                 ipv6_ifa_notify(0, ifp);
3285
3286         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3287                               expires, flags);
3288         addrconf_verify(0);
3289
3290         return 0;
3291 }
3292
3293 static int
3294 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3295 {
3296         struct net *net = sock_net(skb->sk);
3297         struct ifaddrmsg *ifm;
3298         struct nlattr *tb[IFA_MAX+1];
3299         struct in6_addr *pfx;
3300         struct inet6_ifaddr *ifa;
3301         struct net_device *dev;
3302         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3303         u8 ifa_flags;
3304         int err;
3305
3306         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3307         if (err < 0)
3308                 return err;
3309
3310         ifm = nlmsg_data(nlh);
3311         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3312         if (pfx == NULL)
3313                 return -EINVAL;
3314
3315         if (tb[IFA_CACHEINFO]) {
3316                 struct ifa_cacheinfo *ci;
3317
3318                 ci = nla_data(tb[IFA_CACHEINFO]);
3319                 valid_lft = ci->ifa_valid;
3320                 preferred_lft = ci->ifa_prefered;
3321         } else {
3322                 preferred_lft = INFINITY_LIFE_TIME;
3323                 valid_lft = INFINITY_LIFE_TIME;
3324         }
3325
3326         dev =  __dev_get_by_index(net, ifm->ifa_index);
3327         if (dev == NULL)
3328                 return -ENODEV;
3329
3330         /* We ignore other flags so far. */
3331         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3332
3333         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3334         if (ifa == NULL) {
3335                 /*
3336                  * It would be best to check for !NLM_F_CREATE here but
3337                  * userspace alreay relies on not having to provide this.
3338                  */
3339                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3340                                       ifm->ifa_prefixlen, ifa_flags,
3341                                       preferred_lft, valid_lft);
3342         }
3343
3344         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3345             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3346                 err = -EEXIST;
3347         else
3348                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3349
3350         in6_ifa_put(ifa);
3351
3352         return err;
3353 }
3354
3355 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3356                           u8 scope, int ifindex)
3357 {
3358         struct ifaddrmsg *ifm;
3359
3360         ifm = nlmsg_data(nlh);
3361         ifm->ifa_family = AF_INET6;
3362         ifm->ifa_prefixlen = prefixlen;
3363         ifm->ifa_flags = flags;
3364         ifm->ifa_scope = scope;
3365         ifm->ifa_index = ifindex;
3366 }
3367
3368 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3369                          unsigned long tstamp, u32 preferred, u32 valid)
3370 {
3371         struct ifa_cacheinfo ci;
3372
3373         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3374                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3375         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3376                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3377         ci.ifa_prefered = preferred;
3378         ci.ifa_valid = valid;
3379
3380         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3381 }
3382
3383 static inline int rt_scope(int ifa_scope)
3384 {
3385         if (ifa_scope & IFA_HOST)
3386                 return RT_SCOPE_HOST;
3387         else if (ifa_scope & IFA_LINK)
3388                 return RT_SCOPE_LINK;
3389         else if (ifa_scope & IFA_SITE)
3390                 return RT_SCOPE_SITE;
3391         else
3392                 return RT_SCOPE_UNIVERSE;
3393 }
3394
3395 static inline int inet6_ifaddr_msgsize(void)
3396 {
3397         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3398                + nla_total_size(16) /* IFA_ADDRESS */
3399                + nla_total_size(sizeof(struct ifa_cacheinfo));
3400 }
3401
3402 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3403                              u32 pid, u32 seq, int event, unsigned int flags)
3404 {
3405         struct nlmsghdr  *nlh;
3406         u32 preferred, valid;
3407
3408         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3409         if (nlh == NULL)
3410                 return -EMSGSIZE;
3411
3412         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3413                       ifa->idev->dev->ifindex);
3414
3415         if (!(ifa->flags&IFA_F_PERMANENT)) {
3416                 preferred = ifa->prefered_lft;
3417                 valid = ifa->valid_lft;
3418                 if (preferred != INFINITY_LIFE_TIME) {
3419                         long tval = (jiffies - ifa->tstamp)/HZ;
3420                         if (preferred > tval)
3421                                 preferred -= tval;
3422                         else
3423                                 preferred = 0;
3424                         if (valid != INFINITY_LIFE_TIME)
3425                                 valid -= tval;
3426                 }
3427         } else {
3428                 preferred = INFINITY_LIFE_TIME;
3429                 valid = INFINITY_LIFE_TIME;
3430         }
3431
3432         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3433             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3434                 nlmsg_cancel(skb, nlh);
3435                 return -EMSGSIZE;
3436         }
3437
3438         return nlmsg_end(skb, nlh);
3439 }
3440
3441 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3442                                 u32 pid, u32 seq, int event, u16 flags)
3443 {
3444         struct nlmsghdr  *nlh;
3445         u8 scope = RT_SCOPE_UNIVERSE;
3446         int ifindex = ifmca->idev->dev->ifindex;
3447
3448         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3449                 scope = RT_SCOPE_SITE;
3450
3451         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3452         if (nlh == NULL)
3453                 return -EMSGSIZE;
3454
3455         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3456         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3457             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3458                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3459                 nlmsg_cancel(skb, nlh);
3460                 return -EMSGSIZE;
3461         }
3462
3463         return nlmsg_end(skb, nlh);
3464 }
3465
3466 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3467                                 u32 pid, u32 seq, int event, unsigned int flags)
3468 {
3469         struct nlmsghdr  *nlh;
3470         u8 scope = RT_SCOPE_UNIVERSE;
3471         int ifindex = ifaca->aca_idev->dev->ifindex;
3472
3473         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3474                 scope = RT_SCOPE_SITE;
3475
3476         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3477         if (nlh == NULL)
3478                 return -EMSGSIZE;
3479
3480         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3481         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3482             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3483                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3484                 nlmsg_cancel(skb, nlh);
3485                 return -EMSGSIZE;
3486         }
3487
3488         return nlmsg_end(skb, nlh);
3489 }
3490
3491 enum addr_type_t
3492 {
3493         UNICAST_ADDR,
3494         MULTICAST_ADDR,
3495         ANYCAST_ADDR,
3496 };
3497
3498 /* called with rcu_read_lock() */
3499 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb,
3500                           struct netlink_callback *cb, enum addr_type_t type,
3501                           int s_ip_idx, int *p_ip_idx)
3502 {
3503         struct inet6_ifaddr *ifa;
3504         struct ifmcaddr6 *ifmca;
3505         struct ifacaddr6 *ifaca;
3506         int err = 1;
3507         int ip_idx = *p_ip_idx;
3508
3509         read_lock_bh(&idev->lock);
3510         switch (type) {
3511         case UNICAST_ADDR:
3512                 /* unicast address incl. temp addr */
3513                 for (ifa = idev->addr_list; ifa;
3514                      ifa = ifa->if_next, ip_idx++) {
3515                         if (ip_idx < s_ip_idx)
3516                                 continue;
3517                         err = inet6_fill_ifaddr(skb, ifa,
3518                                                 NETLINK_CB(cb->skb).pid,
3519                                                 cb->nlh->nlmsg_seq,
3520                                                 RTM_NEWADDR,
3521                                                 NLM_F_MULTI);
3522                         if (err <= 0)
3523                                 break;
3524                 }
3525                 break;
3526         case MULTICAST_ADDR:
3527                 /* multicast address */
3528                 for (ifmca = idev->mc_list; ifmca;
3529                      ifmca = ifmca->next, ip_idx++) {
3530                         if (ip_idx < s_ip_idx)
3531                                 continue;
3532                         err = inet6_fill_ifmcaddr(skb, ifmca,
3533                                                   NETLINK_CB(cb->skb).pid,
3534                                                   cb->nlh->nlmsg_seq,
3535                                                   RTM_GETMULTICAST,
3536                                                   NLM_F_MULTI);
3537                         if (err <= 0)
3538                                 break;
3539                 }
3540                 break;
3541         case ANYCAST_ADDR:
3542                 /* anycast address */
3543                 for (ifaca = idev->ac_list; ifaca;
3544                      ifaca = ifaca->aca_next, ip_idx++) {
3545                         if (ip_idx < s_ip_idx)
3546                                 continue;
3547                         err = inet6_fill_ifacaddr(skb, ifaca,
3548                                                   NETLINK_CB(cb->skb).pid,
3549                                                   cb->nlh->nlmsg_seq,
3550                                                   RTM_GETANYCAST,
3551                                                   NLM_F_MULTI);
3552                         if (err <= 0)
3553                                 break;
3554                 }
3555                 break;
3556         default:
3557                 break;
3558         }
3559         read_unlock_bh(&idev->lock);
3560         *p_ip_idx = ip_idx;
3561         return err;
3562 }
3563
3564 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3565                            enum addr_type_t type)
3566 {
3567         struct net *net = sock_net(skb->sk);
3568         int h, s_h;
3569         int idx, ip_idx;
3570         int s_idx, s_ip_idx;
3571         struct net_device *dev;
3572         struct inet6_dev *idev;
3573         struct hlist_head *head;
3574         struct hlist_node *node;
3575
3576         s_h = cb->args[0];
3577         s_idx = idx = cb->args[1];
3578         s_ip_idx = ip_idx = cb->args[2];
3579
3580         rcu_read_lock();
3581         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3582                 idx = 0;
3583                 head = &net->dev_index_head[h];
3584                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3585                         if (idx < s_idx)
3586                                 goto cont;
3587                         if (idx > s_idx)
3588                                 s_ip_idx = 0;
3589                         ip_idx = 0;
3590                         if ((idev = __in6_dev_get(dev)) == NULL)
3591                                 goto cont;
3592
3593                         if (in6_dump_addrs(idev, skb, cb, type,
3594                                            s_ip_idx, &ip_idx) <= 0)
3595                                 goto done;
3596 cont:
3597                         idx++;
3598                 }
3599         }
3600 done:
3601         rcu_read_unlock();
3602         cb->args[0] = h;
3603         cb->args[1] = idx;
3604         cb->args[2] = ip_idx;
3605
3606         return skb->len;
3607 }
3608
3609 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3610 {
3611         enum addr_type_t type = UNICAST_ADDR;
3612
3613         return inet6_dump_addr(skb, cb, type);
3614 }
3615
3616 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3617 {
3618         enum addr_type_t type = MULTICAST_ADDR;
3619
3620         return inet6_dump_addr(skb, cb, type);
3621 }
3622
3623
3624 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3625 {
3626         enum addr_type_t type = ANYCAST_ADDR;
3627
3628         return inet6_dump_addr(skb, cb, type);
3629 }
3630
3631 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3632                              void *arg)
3633 {
3634         struct net *net = sock_net(in_skb->sk);
3635         struct ifaddrmsg *ifm;
3636         struct nlattr *tb[IFA_MAX+1];
3637         struct in6_addr *addr = NULL;
3638         struct net_device *dev = NULL;
3639         struct inet6_ifaddr *ifa;
3640         struct sk_buff *skb;
3641         int err;
3642
3643         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3644         if (err < 0)
3645                 goto errout;
3646
3647         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3648         if (addr == NULL) {
3649                 err = -EINVAL;
3650                 goto errout;
3651         }
3652
3653         ifm = nlmsg_data(nlh);
3654         if (ifm->ifa_index)
3655                 dev = __dev_get_by_index(net, ifm->ifa_index);
3656
3657         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3658                 err = -EADDRNOTAVAIL;
3659                 goto errout;
3660         }
3661
3662         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3663                 err = -ENOBUFS;
3664                 goto errout_ifa;
3665         }
3666
3667         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3668                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3669         if (err < 0) {
3670                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3671                 WARN_ON(err == -EMSGSIZE);
3672                 kfree_skb(skb);
3673                 goto errout_ifa;
3674         }
3675         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3676 errout_ifa:
3677         in6_ifa_put(ifa);
3678 errout:
3679         return err;
3680 }
3681
3682 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3683 {
3684         struct sk_buff *skb;
3685         struct net *net = dev_net(ifa->idev->dev);
3686         int err = -ENOBUFS;
3687
3688         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3689         if (skb == NULL)
3690                 goto errout;
3691
3692         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3693         if (err < 0) {
3694                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3695                 WARN_ON(err == -EMSGSIZE);
3696                 kfree_skb(skb);
3697                 goto errout;
3698         }
3699         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3700         return;
3701 errout:
3702         if (err < 0)
3703                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3704 }
3705
3706 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3707                                 __s32 *array, int bytes)
3708 {
3709         BUG_ON(bytes < (DEVCONF_MAX * 4));
3710
3711         memset(array, 0, bytes);
3712         array[DEVCONF_FORWARDING] = cnf->forwarding;
3713         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3714         array[DEVCONF_MTU6] = cnf->mtu6;
3715         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3716         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3717         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3718         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3719         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3720         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3721         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3722         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3723 #ifdef CONFIG_IPV6_PRIVACY
3724         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3725         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3726         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3727         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3728         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3729 #endif
3730         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3731         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3732         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3733 #ifdef CONFIG_IPV6_ROUTER_PREF
3734         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3735         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3736 #ifdef CONFIG_IPV6_ROUTE_INFO
3737         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3738 #endif
3739 #endif
3740         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3741         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3742 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3743         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3744 #endif
3745 #ifdef CONFIG_IPV6_MROUTE
3746         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3747 #endif
3748         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3749         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3750         array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao;
3751 }
3752
3753 static inline size_t inet6_if_nlmsg_size(void)
3754 {
3755         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3756                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3757                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3758                + nla_total_size(4) /* IFLA_MTU */
3759                + nla_total_size(4) /* IFLA_LINK */
3760                + nla_total_size( /* IFLA_PROTINFO */
3761                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3762                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3763                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3764                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3765                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3766                  );
3767 }
3768
3769 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3770                                       int bytes)
3771 {
3772         int i;
3773         int pad = bytes - sizeof(u64) * items;
3774         BUG_ON(pad < 0);
3775
3776         /* Use put_unaligned() because stats may not be aligned for u64. */
3777         put_unaligned(items, &stats[0]);
3778         for (i = 1; i < items; i++)
3779                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3780
3781         memset(&stats[items], 0, pad);
3782 }
3783
3784 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3785                              int bytes)
3786 {
3787         switch(attrtype) {
3788         case IFLA_INET6_STATS:
3789                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3790                 break;
3791         case IFLA_INET6_ICMP6STATS:
3792                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3793                 break;
3794         }
3795 }
3796
3797 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3798                              u32 pid, u32 seq, int event, unsigned int flags)
3799 {
3800         struct net_device *dev = idev->dev;
3801         struct nlattr *nla;
3802         struct ifinfomsg *hdr;
3803         struct nlmsghdr *nlh;
3804         void *protoinfo;
3805         struct ifla_cacheinfo ci;
3806
3807         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3808         if (nlh == NULL)
3809                 return -EMSGSIZE;
3810
3811         hdr = nlmsg_data(nlh);
3812         hdr->ifi_family = AF_INET6;
3813         hdr->__ifi_pad = 0;
3814         hdr->ifi_type = dev->type;
3815         hdr->ifi_index = dev->ifindex;
3816         hdr->ifi_flags = dev_get_flags(dev);
3817         hdr->ifi_change = 0;
3818
3819         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3820
3821         if (dev->addr_len)
3822                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3823
3824         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3825         if (dev->ifindex != dev->iflink)
3826                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3827
3828         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3829         if (protoinfo == NULL)
3830                 goto nla_put_failure;
3831
3832         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3833
3834         ci.max_reasm_len = IPV6_MAXPLEN;
3835         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3836                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3837         ci.reachable_time = idev->nd_parms->reachable_time;
3838         ci.retrans_time = idev->nd_parms->retrans_time;
3839         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3840
3841         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3842         if (nla == NULL)
3843                 goto nla_put_failure;
3844         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3845
3846         /* XXX - MC not implemented */
3847
3848         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3849         if (nla == NULL)
3850                 goto nla_put_failure;
3851         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3852
3853         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3854         if (nla == NULL)
3855                 goto nla_put_failure;
3856         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3857
3858         nla_nest_end(skb, protoinfo);
3859         return nlmsg_end(skb, nlh);
3860
3861 nla_put_failure:
3862         nlmsg_cancel(skb, nlh);
3863         return -EMSGSIZE;
3864 }
3865
3866 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3867 {
3868         struct net *net = sock_net(skb->sk);
3869         int h, s_h;
3870         int idx = 0, s_idx;
3871         struct net_device *dev;
3872         struct inet6_dev *idev;
3873         struct hlist_head *head;
3874         struct hlist_node *node;
3875
3876         s_h = cb->args[0];
3877         s_idx = cb->args[1];
3878
3879         rcu_read_lock();
3880         for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) {
3881                 idx = 0;
3882                 head = &net->dev_index_head[h];
3883                 hlist_for_each_entry_rcu(dev, node, head, index_hlist) {
3884                         if (idx < s_idx)
3885                                 goto cont;
3886                         idev = __in6_dev_get(dev);
3887                         if (!idev)
3888                                 goto cont;
3889                         if (inet6_fill_ifinfo(skb, idev,
3890                                               NETLINK_CB(cb->skb).pid,
3891                                               cb->nlh->nlmsg_seq,
3892                                               RTM_NEWLINK, NLM_F_MULTI) <= 0)
3893                                 goto out;
3894 cont:
3895                         idx++;
3896                 }
3897         }
3898 out:
3899         rcu_read_unlock();
3900         cb->args[1] = idx;
3901         cb->args[0] = h;
3902
3903         return skb->len;
3904 }
3905
3906 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3907 {
3908         struct sk_buff *skb;
3909         struct net *net = dev_net(idev->dev);
3910         int err = -ENOBUFS;
3911
3912         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3913         if (skb == NULL)
3914                 goto errout;
3915
3916         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3917         if (err < 0) {
3918                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3919                 WARN_ON(err == -EMSGSIZE);
3920                 kfree_skb(skb);
3921                 goto errout;
3922         }
3923         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3924         return;
3925 errout:
3926         if (err < 0)
3927                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3928 }
3929
3930 static inline size_t inet6_prefix_nlmsg_size(void)
3931 {
3932         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3933                + nla_total_size(sizeof(struct in6_addr))
3934                + nla_total_size(sizeof(struct prefix_cacheinfo));
3935 }
3936
3937 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3938                              struct prefix_info *pinfo, u32 pid, u32 seq,
3939                              int event, unsigned int flags)
3940 {
3941         struct prefixmsg *pmsg;
3942         struct nlmsghdr *nlh;
3943         struct prefix_cacheinfo ci;
3944
3945         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3946         if (nlh == NULL)
3947                 return -EMSGSIZE;
3948
3949         pmsg = nlmsg_data(nlh);
3950         pmsg->prefix_family = AF_INET6;
3951         pmsg->prefix_pad1 = 0;
3952         pmsg->prefix_pad2 = 0;
3953         pmsg->prefix_ifindex = idev->dev->ifindex;
3954         pmsg->prefix_len = pinfo->prefix_len;
3955         pmsg->prefix_type = pinfo->type;
3956         pmsg->prefix_pad3 = 0;
3957         pmsg->prefix_flags = 0;
3958         if (pinfo->onlink)
3959                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3960         if (pinfo->autoconf)
3961                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3962
3963         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3964
3965         ci.preferred_time = ntohl(pinfo->prefered);
3966         ci.valid_time = ntohl(pinfo->valid);
3967         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3968
3969         return nlmsg_end(skb, nlh);
3970
3971 nla_put_failure:
3972         nlmsg_cancel(skb, nlh);
3973         return -EMSGSIZE;
3974 }
3975
3976 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3977                          struct prefix_info *pinfo)
3978 {
3979         struct sk_buff *skb;
3980         struct net *net = dev_net(idev->dev);
3981         int err = -ENOBUFS;
3982
3983         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3984         if (skb == NULL)
3985                 goto errout;
3986
3987         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3988         if (err < 0) {
3989                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3990                 WARN_ON(err == -EMSGSIZE);
3991                 kfree_skb(skb);
3992                 goto errout;
3993         }
3994         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3995         return;
3996 errout:
3997         if (err < 0)
3998                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3999 }
4000
4001 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4002 {
4003         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
4004
4005         switch (event) {
4006         case RTM_NEWADDR:
4007                 /*
4008                  * If the address was optimistic
4009                  * we inserted the route at the start of
4010                  * our DAD process, so we don't need
4011                  * to do it again
4012                  */
4013                 if (!(ifp->rt->rt6i_node))
4014                         ip6_ins_rt(ifp->rt);
4015                 if (ifp->idev->cnf.forwarding)
4016                         addrconf_join_anycast(ifp);
4017                 break;
4018         case RTM_DELADDR:
4019                 if (ifp->idev->cnf.forwarding)
4020                         addrconf_leave_anycast(ifp);
4021                 addrconf_leave_solict(ifp->idev, &ifp->addr);
4022                 dst_hold(&ifp->rt->u.dst);
4023                 if (ip6_del_rt(ifp->rt))
4024                         dst_free(&ifp->rt->u.dst);
4025                 break;
4026         }
4027 }
4028
4029 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
4030 {
4031         rcu_read_lock_bh();
4032         if (likely(ifp->idev->dead == 0))
4033                 __ipv6_ifa_notify(event, ifp);
4034         rcu_read_unlock_bh();
4035 }
4036
4037 #ifdef CONFIG_SYSCTL
4038
4039 static
4040 int addrconf_sysctl_forward(ctl_table *ctl, int write,
4041                            void __user *buffer, size_t *lenp, loff_t *ppos)
4042 {
4043         int *valp = ctl->data;
4044         int val = *valp;
4045         int ret;
4046
4047         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4048
4049         if (write)
4050                 ret = addrconf_fixup_forwarding(ctl, valp, val);
4051         return ret;
4052 }
4053
4054 static void dev_disable_change(struct inet6_dev *idev)
4055 {
4056         if (!idev || !idev->dev)
4057                 return;
4058
4059         if (idev->cnf.disable_ipv6)
4060                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4061         else
4062                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4063 }
4064
4065 static void addrconf_disable_change(struct net *net, __s32 newf)
4066 {
4067         struct net_device *dev;
4068         struct inet6_dev *idev;
4069
4070         rcu_read_lock();
4071         for_each_netdev_rcu(net, dev) {
4072                 idev = __in6_dev_get(dev);
4073                 if (idev) {
4074                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4075                         idev->cnf.disable_ipv6 = newf;
4076                         if (changed)
4077                                 dev_disable_change(idev);
4078                 }
4079         }
4080         rcu_read_unlock();
4081 }
4082
4083 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4084 {
4085         struct net *net;
4086
4087         net = (struct net *)table->extra2;
4088
4089         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4090                 return 0;
4091
4092         if (!rtnl_trylock())
4093                 return restart_syscall();
4094
4095         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4096                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4097                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4098                 addrconf_disable_change(net, newf);
4099         } else if ((!*p) ^ (!old))
4100                 dev_disable_change((struct inet6_dev *)table->extra1);
4101
4102         rtnl_unlock();
4103         return 0;
4104 }
4105
4106 static
4107 int addrconf_sysctl_disable(ctl_table *ctl, int write,
4108                             void __user *buffer, size_t *lenp, loff_t *ppos)
4109 {
4110         int *valp = ctl->data;
4111         int val = *valp;
4112         int ret;
4113
4114         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
4115
4116         if (write)
4117                 ret = addrconf_disable_ipv6(ctl, valp, val);
4118         return ret;
4119 }
4120
4121 static struct addrconf_sysctl_table
4122 {
4123         struct ctl_table_header *sysctl_header;
4124         ctl_table addrconf_vars[DEVCONF_MAX+1];
4125         char *dev_name;
4126 } addrconf_sysctl __read_mostly = {
4127         .sysctl_header = NULL,
4128         .addrconf_vars = {
4129                 {
4130                         .procname       =       "forwarding",
4131                         .data           =       &ipv6_devconf.forwarding,
4132                         .maxlen         =       sizeof(int),
4133                         .mode           =       0644,
4134                         .proc_handler   =       addrconf_sysctl_forward,
4135                 },
4136                 {
4137                         .procname       =       "hop_limit",
4138                         .data           =       &ipv6_devconf.hop_limit,
4139                         .maxlen         =       sizeof(int),
4140                         .mode           =       0644,
4141                         .proc_handler   =       proc_dointvec,
4142                 },
4143                 {
4144                         .procname       =       "mtu",
4145                         .data           =       &ipv6_devconf.mtu6,
4146                         .maxlen         =       sizeof(int),
4147                         .mode           =       0644,
4148                         .proc_handler   =       proc_dointvec,
4149                 },
4150                 {
4151                         .procname       =       "accept_ra",
4152                         .data           =       &ipv6_devconf.accept_ra,
4153                         .maxlen         =       sizeof(int),
4154                         .mode           =       0644,
4155                         .proc_handler   =       proc_dointvec,
4156                 },
4157                 {
4158                         .procname       =       "accept_redirects",
4159                         .data           =       &ipv6_devconf.accept_redirects,
4160                         .maxlen         =       sizeof(int),
4161                         .mode           =       0644,
4162                         .proc_handler   =       proc_dointvec,
4163                 },
4164                 {
4165                         .procname       =       "autoconf",
4166                         .data           =       &ipv6_devconf.autoconf,
4167                         .maxlen         =       sizeof(int),
4168                         .mode           =       0644,
4169                         .proc_handler   =       proc_dointvec,
4170                 },
4171                 {
4172                         .procname       =       "dad_transmits",
4173                         .data           =       &ipv6_devconf.dad_transmits,
4174                         .maxlen         =       sizeof(int),
4175                         .mode           =       0644,
4176                         .proc_handler   =       proc_dointvec,
4177                 },
4178                 {
4179                         .procname       =       "router_solicitations",
4180                         .data           =       &ipv6_devconf.rtr_solicits,
4181                         .maxlen         =       sizeof(int),
4182                         .mode           =       0644,
4183                         .proc_handler   =       proc_dointvec,
4184                 },
4185                 {
4186                         .procname       =       "router_solicitation_interval",
4187                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4188                         .maxlen         =       sizeof(int),
4189                         .mode           =       0644,
4190                         .proc_handler   =       proc_dointvec_jiffies,
4191                 },
4192                 {
4193                         .procname       =       "router_solicitation_delay",
4194                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4195                         .maxlen         =       sizeof(int),
4196                         .mode           =       0644,
4197                         .proc_handler   =       proc_dointvec_jiffies,
4198                 },
4199                 {
4200                         .procname       =       "force_mld_version",
4201                         .data           =       &ipv6_devconf.force_mld_version,
4202                         .maxlen         =       sizeof(int),
4203                         .mode           =       0644,
4204                         .proc_handler   =       proc_dointvec,
4205                 },
4206 #ifdef CONFIG_IPV6_PRIVACY
4207                 {
4208                         .procname       =       "use_tempaddr",
4209                         .data           =       &ipv6_devconf.use_tempaddr,
4210                         .maxlen         =       sizeof(int),
4211                         .mode           =       0644,
4212                         .proc_handler   =       proc_dointvec,
4213                 },
4214                 {
4215                         .procname       =       "temp_valid_lft",
4216                         .data           =       &ipv6_devconf.temp_valid_lft,
4217                         .maxlen         =       sizeof(int),
4218                         .mode           =       0644,
4219                         .proc_handler   =       proc_dointvec,
4220                 },
4221                 {
4222                         .procname       =       "temp_prefered_lft",
4223                         .data           =       &ipv6_devconf.temp_prefered_lft,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       proc_dointvec,
4227                 },
4228                 {
4229                         .procname       =       "regen_max_retry",
4230                         .data           =       &ipv6_devconf.regen_max_retry,
4231                         .maxlen         =       sizeof(int),
4232                         .mode           =       0644,
4233                         .proc_handler   =       proc_dointvec,
4234                 },
4235                 {
4236                         .procname       =       "max_desync_factor",
4237                         .data           =       &ipv6_devconf.max_desync_factor,
4238                         .maxlen         =       sizeof(int),
4239                         .mode           =       0644,
4240                         .proc_handler   =       proc_dointvec,
4241                 },
4242 #endif
4243                 {
4244                         .procname       =       "max_addresses",
4245                         .data           =       &ipv6_devconf.max_addresses,
4246                         .maxlen         =       sizeof(int),
4247                         .mode           =       0644,
4248                         .proc_handler   =       proc_dointvec,
4249                 },
4250                 {
4251                         .procname       =       "accept_ra_defrtr",
4252                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4253                         .maxlen         =       sizeof(int),
4254                         .mode           =       0644,
4255                         .proc_handler   =       proc_dointvec,
4256                 },
4257                 {
4258                         .procname       =       "accept_ra_pinfo",
4259                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4260                         .maxlen         =       sizeof(int),
4261                         .mode           =       0644,
4262                         .proc_handler   =       proc_dointvec,
4263                 },
4264 #ifdef CONFIG_IPV6_ROUTER_PREF
4265                 {
4266                         .procname       =       "accept_ra_rtr_pref",
4267                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4268                         .maxlen         =       sizeof(int),
4269                         .mode           =       0644,
4270                         .proc_handler   =       proc_dointvec,
4271                 },
4272                 {
4273                         .procname       =       "router_probe_interval",
4274                         .data           =       &ipv6_devconf.rtr_probe_interval,
4275                         .maxlen         =       sizeof(int),
4276                         .mode           =       0644,
4277                         .proc_handler   =       proc_dointvec_jiffies,
4278                 },
4279 #ifdef CONFIG_IPV6_ROUTE_INFO
4280                 {
4281                         .procname       =       "accept_ra_rt_info_max_plen",
4282                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4283                         .maxlen         =       sizeof(int),
4284                         .mode           =       0644,
4285                         .proc_handler   =       proc_dointvec,
4286                 },
4287 #endif
4288 #endif
4289                 {
4290                         .procname       =       "proxy_ndp",
4291                         .data           =       &ipv6_devconf.proxy_ndp,
4292                         .maxlen         =       sizeof(int),
4293                         .mode           =       0644,
4294                         .proc_handler   =       proc_dointvec,
4295                 },
4296                 {
4297                         .procname       =       "accept_source_route",
4298                         .data           =       &ipv6_devconf.accept_source_route,
4299                         .maxlen         =       sizeof(int),
4300                         .mode           =       0644,
4301                         .proc_handler   =       proc_dointvec,
4302                 },
4303 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4304                 {
4305                         .procname       =       "optimistic_dad",
4306                         .data           =       &ipv6_devconf.optimistic_dad,
4307                         .maxlen         =       sizeof(int),
4308                         .mode           =       0644,
4309                         .proc_handler   =       proc_dointvec,
4310
4311                 },
4312 #endif
4313 #ifdef CONFIG_IPV6_MROUTE
4314                 {
4315                         .procname       =       "mc_forwarding",
4316                         .data           =       &ipv6_devconf.mc_forwarding,
4317                         .maxlen         =       sizeof(int),
4318                         .mode           =       0444,
4319                         .proc_handler   =       proc_dointvec,
4320                 },
4321 #endif
4322                 {
4323                         .procname       =       "disable_ipv6",
4324                         .data           =       &ipv6_devconf.disable_ipv6,
4325                         .maxlen         =       sizeof(int),
4326                         .mode           =       0644,
4327                         .proc_handler   =       addrconf_sysctl_disable,
4328                 },
4329                 {
4330                         .procname       =       "accept_dad",
4331                         .data           =       &ipv6_devconf.accept_dad,
4332                         .maxlen         =       sizeof(int),
4333                         .mode           =       0644,
4334                         .proc_handler   =       proc_dointvec,
4335                 },
4336                 {
4337                         .procname       = "force_tllao",
4338                         .data           = &ipv6_devconf.force_tllao,
4339                         .maxlen         = sizeof(int),
4340                         .mode           = 0644,
4341                         .proc_handler   = proc_dointvec
4342                 },
4343                 {
4344                         /* sentinel */
4345                 }
4346         },
4347 };
4348
4349 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4350                 struct inet6_dev *idev, struct ipv6_devconf *p)
4351 {
4352         int i;
4353         struct addrconf_sysctl_table *t;
4354
4355 #define ADDRCONF_CTL_PATH_DEV   3
4356
4357         struct ctl_path addrconf_ctl_path[] = {
4358                 { .procname = "net", },
4359                 { .procname = "ipv6", },
4360                 { .procname = "conf", },
4361                 { /* to be set */ },
4362                 { },
4363         };
4364
4365
4366         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4367         if (t == NULL)
4368                 goto out;
4369
4370         for (i=0; t->addrconf_vars[i].data; i++) {
4371                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4372                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4373                 t->addrconf_vars[i].extra2 = net;
4374         }
4375
4376         /*
4377          * Make a copy of dev_name, because '.procname' is regarded as const
4378          * by sysctl and we wouldn't want anyone to change it under our feet
4379          * (see SIOCSIFNAME).
4380          */
4381         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4382         if (!t->dev_name)
4383                 goto free;
4384
4385         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4386
4387         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4388                         t->addrconf_vars);
4389         if (t->sysctl_header == NULL)
4390                 goto free_procname;
4391
4392         p->sysctl = t;
4393         return 0;
4394
4395 free_procname:
4396         kfree(t->dev_name);
4397 free:
4398         kfree(t);
4399 out:
4400         return -ENOBUFS;
4401 }
4402
4403 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4404 {
4405         struct addrconf_sysctl_table *t;
4406
4407         if (p->sysctl == NULL)
4408                 return;
4409
4410         t = p->sysctl;
4411         p->sysctl = NULL;
4412         unregister_sysctl_table(t->sysctl_header);
4413         kfree(t->dev_name);
4414         kfree(t);
4415 }
4416
4417 static void addrconf_sysctl_register(struct inet6_dev *idev)
4418 {
4419         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4420                               NET_IPV6_NEIGH, "ipv6",
4421                               &ndisc_ifinfo_sysctl_change);
4422         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4423                                         idev, &idev->cnf);
4424 }
4425
4426 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4427 {
4428         __addrconf_sysctl_unregister(&idev->cnf);
4429         neigh_sysctl_unregister(idev->nd_parms);
4430 }
4431
4432
4433 #endif
4434
4435 static int __net_init addrconf_init_net(struct net *net)
4436 {
4437         int err;
4438         struct ipv6_devconf *all, *dflt;
4439
4440         err = -ENOMEM;
4441         all = &ipv6_devconf;
4442         dflt = &ipv6_devconf_dflt;
4443
4444         if (!net_eq(net, &init_net)) {
4445                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4446                 if (all == NULL)
4447                         goto err_alloc_all;
4448
4449                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4450                 if (dflt == NULL)
4451                         goto err_alloc_dflt;
4452         } else {
4453                 /* these will be inherited by all namespaces */
4454                 dflt->autoconf = ipv6_defaults.autoconf;
4455                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4456         }
4457
4458         net->ipv6.devconf_all = all;
4459         net->ipv6.devconf_dflt = dflt;
4460
4461 #ifdef CONFIG_SYSCTL
4462         err = __addrconf_sysctl_register(net, "all", NULL, all);
4463         if (err < 0)
4464                 goto err_reg_all;
4465
4466         err = __addrconf_sysctl_register(net, "default", NULL, dflt);
4467         if (err < 0)
4468                 goto err_reg_dflt;
4469 #endif
4470         return 0;
4471
4472 #ifdef CONFIG_SYSCTL
4473 err_reg_dflt:
4474         __addrconf_sysctl_unregister(all);
4475 err_reg_all:
4476         kfree(dflt);
4477 #endif
4478 err_alloc_dflt:
4479         kfree(all);
4480 err_alloc_all:
4481         return err;
4482 }
4483
4484 static void __net_exit addrconf_exit_net(struct net *net)
4485 {
4486 #ifdef CONFIG_SYSCTL
4487         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4488         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4489 #endif
4490         if (!net_eq(net, &init_net)) {
4491                 kfree(net->ipv6.devconf_dflt);
4492                 kfree(net->ipv6.devconf_all);
4493         }
4494 }
4495
4496 static struct pernet_operations addrconf_ops = {
4497         .init = addrconf_init_net,
4498         .exit = addrconf_exit_net,
4499 };
4500
4501 /*
4502  *      Device notifier
4503  */
4504
4505 int register_inet6addr_notifier(struct notifier_block *nb)
4506 {
4507         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4508 }
4509
4510 EXPORT_SYMBOL(register_inet6addr_notifier);
4511
4512 int unregister_inet6addr_notifier(struct notifier_block *nb)
4513 {
4514         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4515 }
4516
4517 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4518
4519 /*
4520  *      Init / cleanup code
4521  */
4522
4523 int __init addrconf_init(void)
4524 {
4525         int err;
4526
4527         if ((err = ipv6_addr_label_init()) < 0) {
4528                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4529                         err);
4530                 return err;
4531         }
4532
4533         register_pernet_subsys(&addrconf_ops);
4534
4535         /* The addrconf netdev notifier requires that loopback_dev
4536          * has it's ipv6 private information allocated and setup
4537          * before it can bring up and give link-local addresses
4538          * to other devices which are up.
4539          *
4540          * Unfortunately, loopback_dev is not necessarily the first
4541          * entry in the global dev_base list of net devices.  In fact,
4542          * it is likely to be the very last entry on that list.
4543          * So this causes the notifier registry below to try and
4544          * give link-local addresses to all devices besides loopback_dev
4545          * first, then loopback_dev, which cases all the non-loopback_dev
4546          * devices to fail to get a link-local address.
4547          *
4548          * So, as a temporary fix, allocate the ipv6 structure for
4549          * loopback_dev first by hand.
4550          * Longer term, all of the dependencies ipv6 has upon the loopback
4551          * device and it being up should be removed.
4552          */
4553         rtnl_lock();
4554         if (!ipv6_add_dev(init_net.loopback_dev))
4555                 err = -ENOMEM;
4556         rtnl_unlock();
4557         if (err)
4558                 goto errlo;
4559
4560         register_netdevice_notifier(&ipv6_dev_notf);
4561
4562         addrconf_verify(0);
4563
4564         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4565         if (err < 0)
4566                 goto errout;
4567
4568         /* Only the first call to __rtnl_register can fail */
4569         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4570         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4571         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4572         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4573         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4574
4575         ipv6_addr_label_rtnl_register();
4576
4577         return 0;
4578 errout:
4579         unregister_netdevice_notifier(&ipv6_dev_notf);
4580 errlo:
4581         unregister_pernet_subsys(&addrconf_ops);
4582
4583         return err;
4584 }
4585
4586 void addrconf_cleanup(void)
4587 {
4588         struct inet6_ifaddr *ifa;
4589         struct net_device *dev;
4590         int i;
4591
4592         unregister_netdevice_notifier(&ipv6_dev_notf);
4593         unregister_pernet_subsys(&addrconf_ops);
4594
4595         rtnl_lock();
4596
4597         /* clean dev list */
4598         for_each_netdev(&init_net, dev) {
4599                 if (__in6_dev_get(dev) == NULL)
4600                         continue;
4601                 addrconf_ifdown(dev, 1);
4602         }
4603         addrconf_ifdown(init_net.loopback_dev, 2);
4604
4605         /*
4606          *      Check hash table.
4607          */
4608         write_lock_bh(&addrconf_hash_lock);
4609         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4610                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4611                         struct inet6_ifaddr *bifa;
4612
4613                         bifa = ifa;
4614                         ifa = ifa->lst_next;
4615                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4616                         /* Do not free it; something is wrong.
4617                            Now we can investigate it with debugger.
4618                          */
4619                 }
4620         }
4621         write_unlock_bh(&addrconf_hash_lock);
4622
4623         del_timer(&addr_chk_timer);
4624         rtnl_unlock();
4625 }