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

   1 /*
   2  *      IPv6 Address [auto]configuration
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      This program is free software; you can redistribute it and/or
  10  *      modify it under the terms of the GNU General Public License
  11  *      as published by the Free Software Foundation; either version
  12  *      2 of the License, or (at your option) any later version.
  13  */
  14
  15 /*
  16  *      Changes:
  17  *
  18  *      Janos Farkas                    :       delete timer on ifdown
  19  *      <chexum@bankinf.banki.hu>
  20  *      Andi Kleen                      :       kill double kfree on module
  21  *                                              unload.
  22  *      Maciej W. Rozycki               :       FDDI support
  23  *      sekiya@USAGI                    :       Don't send too many RS
  24  *                                              packets.
  25  *      yoshfuji@USAGI                  :       Fixed interval between DAD
  26  *                                              packets.
  27  *      YOSHIFUJI Hideaki @USAGI        :       improved accuracy of
  28  *                                              address validation timer.
  29  *      YOSHIFUJI Hideaki @USAGI        :       Privacy Extensions (RFC3041)
  30  *                                              support.
  31  *      Yuji SEKIYA @USAGI              :       Don't assign a same IPv6
  32  *                                              address on a same interface.
  33  *      YOSHIFUJI Hideaki @USAGI        :       ARCnet support
  34  *      YOSHIFUJI Hideaki @USAGI        :       convert /proc/net/if_inet6 to
  35  *                                              seq_file.
  36  *      YOSHIFUJI Hideaki @USAGI        :       improved source address
  37  *                                              selection; consider scope,
  38  *                                              status etc.
  39  */
  40
  41 #include <linux/errno.h>
  42 #include <linux/types.h>
  43 #include <linux/kernel.h>
  44 #include <linux/socket.h>
  45 #include <linux/sockios.h>
  46 #include <linux/net.h>
  47 #include <linux/in6.h>
  48 #include <linux/netdevice.h>
  49 #include <linux/if_addr.h>
  50 #include <linux/if_arp.h>
  51 #include <linux/if_arcnet.h>
  52 #include <linux/if_infiniband.h>
  53 #include <linux/route.h>
  54 #include <linux/inetdevice.h>
  55 #include <linux/init.h>
  56 #ifdef CONFIG_SYSCTL
  57 #include <linux/sysctl.h>
  58 #endif
  59 #include <linux/capability.h>
  60 #include <linux/delay.h>
  61 #include <linux/notifier.h>
  62 #include <linux/string.h>
  63
  64 #include <net/net_namespace.h>
  65 #include <net/sock.h>
  66 #include <net/snmp.h>
  67
  68 #include <net/ipv6.h>
  69 #include <net/protocol.h>
  70 #include <net/ndisc.h>
  71 #include <net/ip6_route.h>
  72 #include <net/addrconf.h>
  73 #include <net/tcp.h>
  74 #include <net/ip.h>
  75 #include <net/netlink.h>
  76 #include <net/pkt_sched.h>
  77 #include <linux/if_tunnel.h>
  78 #include <linux/rtnetlink.h>
  79
  80 #ifdef CONFIG_IPV6_PRIVACY
  81 #include <linux/random.h>
  82 #endif
  83
  84 #include <asm/uaccess.h>
  85 #include <asm/unaligned.h>
  86
  87 #include <linux/proc_fs.h>
  88 #include <linux/seq_file.h>
  89
  90 /* Set to 3 to get tracing... */
  91 #define ACONF_DEBUG 2
  92
  93 #if ACONF_DEBUG >= 3
  94 #define ADBG(x) printk x
  95 #else
  96 #define ADBG(x)
  97 #endif
  98
  99 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 100 #define TIME_DELTA(a,b) ((unsigned long)((long)(a) - (long)(b)))
 101
 102 #ifdef CONFIG_SYSCTL
 103 static void addrconf_sysctl_register(struct inet6_dev *idev);
 104 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 105 #else
 106 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 107 {
 108 }
 109
 110 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 111 {
 112 }
 113 #endif
 114
 115 #ifdef CONFIG_IPV6_PRIVACY
 116 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 117 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 118 static void ipv6_regen_rndid(unsigned long data);
 119
 120 static int desync_factor = MAX_DESYNC_FACTOR * HZ;
 121 #endif
 122
 123 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 124 static int ipv6_count_addresses(struct inet6_dev *idev);
 125
 126 /*
 127  *      Configured unicast address hash table
 128  */
 129 static struct inet6_ifaddr              *inet6_addr_lst[IN6_ADDR_HSIZE];
 130 static DEFINE_RWLOCK(addrconf_hash_lock);
 131
 132 static void addrconf_verify(unsigned long);
 133
 134 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 135 static DEFINE_SPINLOCK(addrconf_verify_lock);
 136
 137 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 138 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 139
 140 static int addrconf_ifdown(struct net_device *dev, int how);
 141
 142 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 143 static void addrconf_dad_timer(unsigned long data);
 144 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 145 static void addrconf_dad_run(struct inet6_dev *idev);
 146 static void addrconf_rs_timer(unsigned long data);
 147 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 148 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 149
 150 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 151                                 struct prefix_info *pinfo);
 152 static int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 153                               struct net_device *dev);
 154
 155 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 156
 157 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 158         .forwarding             = 0,
 159         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 160         .mtu6                   = IPV6_MIN_MTU,
 161         .accept_ra              = 1,
 162         .accept_redirects       = 1,
 163         .autoconf               = 1,
 164         .force_mld_version      = 0,
 165         .dad_transmits          = 1,
 166         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 167         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 168         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 169 #ifdef CONFIG_IPV6_PRIVACY
 170         .use_tempaddr           = 0,
 171         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 172         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 173         .regen_max_retry        = REGEN_MAX_RETRY,
 174         .max_desync_factor      = MAX_DESYNC_FACTOR,
 175 #endif
 176         .max_addresses          = IPV6_MAX_ADDRESSES,
 177         .accept_ra_defrtr       = 1,
 178         .accept_ra_pinfo        = 1,
 179 #ifdef CONFIG_IPV6_ROUTER_PREF
 180         .accept_ra_rtr_pref     = 1,
 181         .rtr_probe_interval     = 60 * HZ,
 182 #ifdef CONFIG_IPV6_ROUTE_INFO
 183         .accept_ra_rt_info_max_plen = 0,
 184 #endif
 185 #endif
 186         .proxy_ndp              = 0,
 187         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 188         .disable_ipv6           = 0,
 189         .accept_dad             = 1,
 190 };
 191
 192 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 193         .forwarding             = 0,
 194         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 195         .mtu6                   = IPV6_MIN_MTU,
 196         .accept_ra              = 1,
 197         .accept_redirects       = 1,
 198         .autoconf               = 1,
 199         .dad_transmits          = 1,
 200         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 201         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 202         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 203 #ifdef CONFIG_IPV6_PRIVACY
 204         .use_tempaddr           = 0,
 205         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 206         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 207         .regen_max_retry        = REGEN_MAX_RETRY,
 208         .max_desync_factor      = MAX_DESYNC_FACTOR,
 209 #endif
 210         .max_addresses          = IPV6_MAX_ADDRESSES,
 211         .accept_ra_defrtr       = 1,
 212         .accept_ra_pinfo        = 1,
 213 #ifdef CONFIG_IPV6_ROUTER_PREF
 214         .accept_ra_rtr_pref     = 1,
 215         .rtr_probe_interval     = 60 * HZ,
 216 #ifdef CONFIG_IPV6_ROUTE_INFO
 217         .accept_ra_rt_info_max_plen = 0,
 218 #endif
 219 #endif
 220         .proxy_ndp              = 0,
 221         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 222         .disable_ipv6           = 0,
 223         .accept_dad             = 1,
 224 };
 225
 226 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 227 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 228 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 229 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 230 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 231
 232 /* Check if a valid qdisc is available */
 233 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 234 {
 235         return !qdisc_tx_is_noop(dev);
 236 }
 237
 238 /* Check if a route is valid prefix route */
 239 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 240 {
 241         return ((rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0);
 242 }
 243
 244 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 245 {
 246         if (del_timer(&ifp->timer))
 247                 __in6_ifa_put(ifp);
 248 }
 249
 250 enum addrconf_timer_t
 251 {
 252         AC_NONE,
 253         AC_DAD,
 254         AC_RS,
 255 };
 256
 257 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 258                                enum addrconf_timer_t what,
 259                                unsigned long when)
 260 {
 261         if (!del_timer(&ifp->timer))
 262                 in6_ifa_hold(ifp);
 263
 264         switch (what) {
 265         case AC_DAD:
 266                 ifp->timer.function = addrconf_dad_timer;
 267                 break;
 268         case AC_RS:
 269                 ifp->timer.function = addrconf_rs_timer;
 270                 break;
 271         default:;
 272         }
 273         ifp->timer.expires = jiffies + when;
 274         add_timer(&ifp->timer);
 275 }
 276
 277 static int snmp6_alloc_dev(struct inet6_dev *idev)
 278 {
 279         if (snmp_mib_init((void **)idev->stats.ipv6,
 280                           sizeof(struct ipstats_mib)) < 0)
 281                 goto err_ip;
 282         if (snmp_mib_init((void **)idev->stats.icmpv6,
 283                           sizeof(struct icmpv6_mib)) < 0)
 284                 goto err_icmp;
 285         if (snmp_mib_init((void **)idev->stats.icmpv6msg,
 286                           sizeof(struct icmpv6msg_mib)) < 0)
 287                 goto err_icmpmsg;
 288
 289         return 0;
 290
 291 err_icmpmsg:
 292         snmp_mib_free((void **)idev->stats.icmpv6);
 293 err_icmp:
 294         snmp_mib_free((void **)idev->stats.ipv6);
 295 err_ip:
 296         return -ENOMEM;
 297 }
 298
 299 static void snmp6_free_dev(struct inet6_dev *idev)
 300 {
 301         snmp_mib_free((void **)idev->stats.icmpv6msg);
 302         snmp_mib_free((void **)idev->stats.icmpv6);
 303         snmp_mib_free((void **)idev->stats.ipv6);
 304 }
 305
 306 /* Nobody refers to this device, we may destroy it. */
 307
 308 static void in6_dev_finish_destroy_rcu(struct rcu_head *head)
 309 {
 310         struct inet6_dev *idev = container_of(head, struct inet6_dev, rcu);
 311         kfree(idev);
 312 }
 313
 314 void in6_dev_finish_destroy(struct inet6_dev *idev)
 315 {
 316         struct net_device *dev = idev->dev;
 317
 318         WARN_ON(idev->addr_list != NULL);
 319         WARN_ON(idev->mc_list != NULL);
 320
 321 #ifdef NET_REFCNT_DEBUG
 322         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 323 #endif
 324         dev_put(dev);
 325         if (!idev->dead) {
 326                 printk("Freeing alive inet6 device %p\n", idev);
 327                 return;
 328         }
 329         snmp6_free_dev(idev);
 330         call_rcu(&idev->rcu, in6_dev_finish_destroy_rcu);
 331 }
 332
 333 EXPORT_SYMBOL(in6_dev_finish_destroy);
 334
 335 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 336 {
 337         struct inet6_dev *ndev;
 338
 339         ASSERT_RTNL();
 340
 341         if (dev->mtu < IPV6_MIN_MTU)
 342                 return NULL;
 343
 344         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 345
 346         if (ndev == NULL)
 347                 return NULL;
 348
 349         rwlock_init(&ndev->lock);
 350         ndev->dev = dev;
 351         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 352         ndev->cnf.mtu6 = dev->mtu;
 353         ndev->cnf.sysctl = NULL;
 354         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 355         if (ndev->nd_parms == NULL) {
 356                 kfree(ndev);
 357                 return NULL;
 358         }
 359         if (ndev->cnf.forwarding)
 360                 dev_disable_lro(dev);
 361         /* We refer to the device */
 362         dev_hold(dev);
 363
 364         if (snmp6_alloc_dev(ndev) < 0) {
 365                 ADBG((KERN_WARNING
 366                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 367                         __func__, dev->name));
 368                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 369                 ndev->dead = 1;
 370                 in6_dev_finish_destroy(ndev);
 371                 return NULL;
 372         }
 373
 374         if (snmp6_register_dev(ndev) < 0) {
 375                 ADBG((KERN_WARNING
 376                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 377                         __func__, dev->name));
 378                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 379                 ndev->dead = 1;
 380                 in6_dev_finish_destroy(ndev);
 381                 return NULL;
 382         }
 383
 384         /* One reference from device.  We must do this before
 385          * we invoke __ipv6_regen_rndid().
 386          */
 387         in6_dev_hold(ndev);
 388
 389         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 390                 ndev->cnf.accept_dad = -1;
 391
 392 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 393         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 394                 printk(KERN_INFO
 395                        "%s: Disabled Multicast RS\n",
 396                        dev->name);
 397                 ndev->cnf.rtr_solicits = 0;
 398         }
 399 #endif
 400
 401 #ifdef CONFIG_IPV6_PRIVACY
 402         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 403         if ((dev->flags&IFF_LOOPBACK) ||
 404             dev->type == ARPHRD_TUNNEL ||
 405             dev->type == ARPHRD_TUNNEL6 ||
 406             dev->type == ARPHRD_SIT ||
 407             dev->type == ARPHRD_NONE) {
 408                 printk(KERN_INFO
 409                        "%s: Disabled Privacy Extensions\n",
 410                        dev->name);
 411                 ndev->cnf.use_tempaddr = -1;
 412         } else {
 413                 in6_dev_hold(ndev);
 414                 ipv6_regen_rndid((unsigned long) ndev);
 415         }
 416 #endif
 417
 418         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 419                 ndev->if_flags |= IF_READY;
 420
 421         ipv6_mc_init_dev(ndev);
 422         ndev->tstamp = jiffies;
 423         addrconf_sysctl_register(ndev);
 424         /* protected by rtnl_lock */
 425         rcu_assign_pointer(dev->ip6_ptr, ndev);
 426
 427         /* Join all-node multicast group */
 428         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 429
 430         return ndev;
 431 }
 432
 433 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 434 {
 435         struct inet6_dev *idev;
 436
 437         ASSERT_RTNL();
 438
 439         if ((idev = __in6_dev_get(dev)) == NULL) {
 440                 if ((idev = ipv6_add_dev(dev)) == NULL)
 441                         return NULL;
 442         }
 443
 444         if (dev->flags&IFF_UP)
 445                 ipv6_mc_up(idev);
 446         return idev;
 447 }
 448
 449 #ifdef CONFIG_SYSCTL
 450 static void dev_forward_change(struct inet6_dev *idev)
 451 {
 452         struct net_device *dev;
 453         struct inet6_ifaddr *ifa;
 454
 455         if (!idev)
 456                 return;
 457         dev = idev->dev;
 458         if (idev->cnf.forwarding)
 459                 dev_disable_lro(dev);
 460         if (dev && (dev->flags & IFF_MULTICAST)) {
 461                 if (idev->cnf.forwarding)
 462                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 463                 else
 464                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 465         }
 466         for (ifa=idev->addr_list; ifa; ifa=ifa->if_next) {
 467                 if (ifa->flags&IFA_F_TENTATIVE)
 468                         continue;
 469                 if (idev->cnf.forwarding)
 470                         addrconf_join_anycast(ifa);
 471                 else
 472                         addrconf_leave_anycast(ifa);
 473         }
 474 }
 475
 476
 477 static void addrconf_forward_change(struct net *net, __s32 newf)
 478 {
 479         struct net_device *dev;
 480         struct inet6_dev *idev;
 481
 482         read_lock(&dev_base_lock);
 483         for_each_netdev(net, dev) {
 484                 rcu_read_lock();
 485                 idev = __in6_dev_get(dev);
 486                 if (idev) {
 487                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 488                         idev->cnf.forwarding = newf;
 489                         if (changed)
 490                                 dev_forward_change(idev);
 491                 }
 492                 rcu_read_unlock();
 493         }
 494         read_unlock(&dev_base_lock);
 495 }
 496
 497 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 498 {
 499         struct net *net;
 500
 501         net = (struct net *)table->extra2;
 502         if (p == &net->ipv6.devconf_dflt->forwarding)
 503                 return 0;
 504
 505         if (!rtnl_trylock())
 506                 return restart_syscall();
 507
 508         if (p == &net->ipv6.devconf_all->forwarding) {
 509                 __s32 newf = net->ipv6.devconf_all->forwarding;
 510                 net->ipv6.devconf_dflt->forwarding = newf;
 511                 addrconf_forward_change(net, newf);
 512         } else if ((!*p) ^ (!old))
 513                 dev_forward_change((struct inet6_dev *)table->extra1);
 514         rtnl_unlock();
 515
 516         if (*p)
 517                 rt6_purge_dflt_routers(net);
 518         return 1;
 519 }
 520 #endif
 521
 522 /* Nobody refers to this ifaddr, destroy it */
 523
 524 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 525 {
 526         WARN_ON(ifp->if_next != NULL);
 527         WARN_ON(ifp->lst_next != NULL);
 528
 529 #ifdef NET_REFCNT_DEBUG
 530         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 531 #endif
 532
 533         in6_dev_put(ifp->idev);
 534
 535         if (del_timer(&ifp->timer))
 536                 printk("Timer is still running, when freeing ifa=%p\n", ifp);
 537
 538         if (!ifp->dead) {
 539                 printk("Freeing alive inet6 address %p\n", ifp);
 540                 return;
 541         }
 542         dst_release(&ifp->rt->u.dst);
 543
 544         kfree(ifp);
 545 }
 546
 547 static void
 548 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 549 {
 550         struct inet6_ifaddr *ifa, **ifap;
 551         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 552
 553         /*
 554          * Each device address list is sorted in order of scope -
 555          * global before linklocal.
 556          */
 557         for (ifap = &idev->addr_list; (ifa = *ifap) != NULL;
 558              ifap = &ifa->if_next) {
 559                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 560                         break;
 561         }
 562
 563         ifp->if_next = *ifap;
 564         *ifap = ifp;
 565 }
 566
 567 /*
 568  *      Hash function taken from net_alias.c
 569  */
 570 static u8 ipv6_addr_hash(const struct in6_addr *addr)
 571 {
 572         __u32 word;
 573
 574         /*
 575          * We perform the hash function over the last 64 bits of the address
 576          * This will include the IEEE address token on links that support it.
 577          */
 578
 579         word = (__force u32)(addr->s6_addr32[2] ^ addr->s6_addr32[3]);
 580         word ^= (word >> 16);
 581         word ^= (word >> 8);
 582
 583         return ((word ^ (word >> 4)) & 0x0f);
 584 }
 585
 586 /* On success it returns ifp with increased reference count */
 587
 588 static struct inet6_ifaddr *
 589 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 590               int scope, u32 flags)
 591 {
 592         struct inet6_ifaddr *ifa = NULL;
 593         struct rt6_info *rt;
 594         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         read_lock(&dev_base_lock);
1139         rcu_read_lock();
1140
1141         for_each_netdev(net, dev) {
1142                 struct inet6_dev *idev;
1143
1144                 /* Candidate Source Address (section 4)
1145                  *  - multicast and link-local destination address,
1146                  *    the set of candidate source address MUST only
1147                  *    include addresses assigned to interfaces
1148                  *    belonging to the same link as the outgoing
1149                  *    interface.
1150                  * (- For site-local destination addresses, the
1151                  *    set of candidate source addresses MUST only
1152                  *    include addresses assigned to interfaces
1153                  *    belonging to the same site as the outgoing
1154                  *    interface.)
1155                  */
1156                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1157                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1158                     dst.ifindex && dev->ifindex != dst.ifindex)
1159                         continue;
1160
1161                 idev = __in6_dev_get(dev);
1162                 if (!idev)
1163                         continue;
1164
1165                 read_lock_bh(&idev->lock);
1166                 for (score->ifa = idev->addr_list; score->ifa; score->ifa = score->ifa->if_next) {
1167                         int i;
1168
1169                         /*
1170                          * - Tentative Address (RFC2462 section 5.4)
1171                          *  - A tentative address is not considered
1172                          *    "assigned to an interface" in the traditional
1173                          *    sense, unless it is also flagged as optimistic.
1174                          * - Candidate Source Address (section 4)
1175                          *  - In any case, anycast addresses, multicast
1176                          *    addresses, and the unspecified address MUST
1177                          *    NOT be included in a candidate set.
1178                          */
1179                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1180                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1181                                 continue;
1182
1183                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1184
1185                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1186                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1187                                 LIMIT_NETDEBUG(KERN_DEBUG
1188                                                "ADDRCONF: unspecified / multicast address "
1189                                                "assigned as unicast address on %s",
1190                                                dev->name);
1191                                 continue;
1192                         }
1193
1194                         score->rule = -1;
1195                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1196
1197                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1198                                 int minihiscore, miniscore;
1199
1200                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1201                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1202
1203                                 if (minihiscore > miniscore) {
1204                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1205                                             score->scopedist > 0) {
1206                                                 /*
1207                                                  * special case:
1208                                                  * each remaining entry
1209                                                  * has too small (not enough)
1210                                                  * scope, because ifa entries
1211                                                  * are sorted by their scope
1212                                                  * values.
1213                                                  */
1214                                                 goto try_nextdev;
1215                                         }
1216                                         break;
1217                                 } else if (minihiscore < miniscore) {
1218                                         if (hiscore->ifa)
1219                                                 in6_ifa_put(hiscore->ifa);
1220
1221                                         in6_ifa_hold(score->ifa);
1222
1223                                         swap(hiscore, score);
1224
1225                                         /* restore our iterator */
1226                                         score->ifa = hiscore->ifa;
1227
1228                                         break;
1229                                 }
1230                         }
1231                 }
1232 try_nextdev:
1233                 read_unlock_bh(&idev->lock);
1234         }
1235         rcu_read_unlock();
1236         read_unlock(&dev_base_lock);
1237
1238         if (!hiscore->ifa)
1239                 return -EADDRNOTAVAIL;
1240
1241         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1242         in6_ifa_put(hiscore->ifa);
1243         return 0;
1244 }
1245
1246 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1247
1248 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1249                     unsigned char banned_flags)
1250 {
1251         struct inet6_dev *idev;
1252         int err = -EADDRNOTAVAIL;
1253
1254         rcu_read_lock();
1255         if ((idev = __in6_dev_get(dev)) != NULL) {
1256                 struct inet6_ifaddr *ifp;
1257
1258                 read_lock_bh(&idev->lock);
1259                 for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1260                         if (ifp->scope == IFA_LINK && !(ifp->flags & banned_flags)) {
1261                                 ipv6_addr_copy(addr, &ifp->addr);
1262                                 err = 0;
1263                                 break;
1264                         }
1265                 }
1266                 read_unlock_bh(&idev->lock);
1267         }
1268         rcu_read_unlock();
1269         return err;
1270 }
1271
1272 static int ipv6_count_addresses(struct inet6_dev *idev)
1273 {
1274         int cnt = 0;
1275         struct inet6_ifaddr *ifp;
1276
1277         read_lock_bh(&idev->lock);
1278         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next)
1279                 cnt++;
1280         read_unlock_bh(&idev->lock);
1281         return cnt;
1282 }
1283
1284 int ipv6_chk_addr(struct net *net, struct in6_addr *addr,
1285                   struct net_device *dev, int strict)
1286 {
1287         struct inet6_ifaddr * ifp;
1288         u8 hash = ipv6_addr_hash(addr);
1289
1290         read_lock_bh(&addrconf_hash_lock);
1291         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1292                 if (!net_eq(dev_net(ifp->idev->dev), net))
1293                         continue;
1294                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1295                     !(ifp->flags&IFA_F_TENTATIVE)) {
1296                         if (dev == NULL || ifp->idev->dev == dev ||
1297                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))
1298                                 break;
1299                 }
1300         }
1301         read_unlock_bh(&addrconf_hash_lock);
1302         return ifp != NULL;
1303 }
1304 EXPORT_SYMBOL(ipv6_chk_addr);
1305
1306 static
1307 int ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1308                        struct net_device *dev)
1309 {
1310         struct inet6_ifaddr * ifp;
1311         u8 hash = ipv6_addr_hash(addr);
1312
1313         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1314                 if (!net_eq(dev_net(ifp->idev->dev), net))
1315                         continue;
1316                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1317                         if (dev == NULL || ifp->idev->dev == dev)
1318                                 break;
1319                 }
1320         }
1321         return ifp != NULL;
1322 }
1323
1324 int ipv6_chk_prefix(struct in6_addr *addr, struct net_device *dev)
1325 {
1326         struct inet6_dev *idev;
1327         struct inet6_ifaddr *ifa;
1328         int     onlink;
1329
1330         onlink = 0;
1331         rcu_read_lock();
1332         idev = __in6_dev_get(dev);
1333         if (idev) {
1334                 read_lock_bh(&idev->lock);
1335                 for (ifa = idev->addr_list; ifa; ifa = ifa->if_next) {
1336                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1337                                                    ifa->prefix_len);
1338                         if (onlink)
1339                                 break;
1340                 }
1341                 read_unlock_bh(&idev->lock);
1342         }
1343         rcu_read_unlock();
1344         return onlink;
1345 }
1346
1347 EXPORT_SYMBOL(ipv6_chk_prefix);
1348
1349 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1350                                      struct net_device *dev, int strict)
1351 {
1352         struct inet6_ifaddr * ifp;
1353         u8 hash = ipv6_addr_hash(addr);
1354
1355         read_lock_bh(&addrconf_hash_lock);
1356         for(ifp = inet6_addr_lst[hash]; ifp; ifp=ifp->lst_next) {
1357                 if (!net_eq(dev_net(ifp->idev->dev), net))
1358                         continue;
1359                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1360                         if (dev == NULL || ifp->idev->dev == dev ||
1361                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1362                                 in6_ifa_hold(ifp);
1363                                 break;
1364                         }
1365                 }
1366         }
1367         read_unlock_bh(&addrconf_hash_lock);
1368
1369         return ifp;
1370 }
1371
1372 /* Gets referenced address, destroys ifaddr */
1373
1374 static void addrconf_dad_stop(struct inet6_ifaddr *ifp)
1375 {
1376         if (ifp->flags&IFA_F_PERMANENT) {
1377                 spin_lock_bh(&ifp->lock);
1378                 addrconf_del_timer(ifp);
1379                 ifp->flags |= IFA_F_TENTATIVE;
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 detected!\n",
1407                         ifp->idev->dev->name);
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);
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         }
2584
2585         return NOTIFY_OK;
2586 }
2587
2588 /*
2589  *      addrconf module should be notified of a device going up
2590  */
2591 static struct notifier_block ipv6_dev_notf = {
2592         .notifier_call = addrconf_notify,
2593         .priority = 0
2594 };
2595
2596 static int addrconf_ifdown(struct net_device *dev, int how)
2597 {
2598         struct inet6_dev *idev;
2599         struct inet6_ifaddr *ifa, **bifa;
2600         struct net *net = dev_net(dev);
2601         int i;
2602
2603         ASSERT_RTNL();
2604
2605         rt6_ifdown(net, dev);
2606         neigh_ifdown(&nd_tbl, dev);
2607
2608         idev = __in6_dev_get(dev);
2609         if (idev == NULL)
2610                 return -ENODEV;
2611
2612         /* Step 1: remove reference to ipv6 device from parent device.
2613                    Do not dev_put!
2614          */
2615         if (how) {
2616                 idev->dead = 1;
2617
2618                 /* protected by rtnl_lock */
2619                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2620
2621                 /* Step 1.5: remove snmp6 entry */
2622                 snmp6_unregister_dev(idev);
2623
2624         }
2625
2626         /* Step 2: clear hash table */
2627         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2628                 bifa = &inet6_addr_lst[i];
2629
2630                 write_lock_bh(&addrconf_hash_lock);
2631                 while ((ifa = *bifa) != NULL) {
2632                         if (ifa->idev == idev) {
2633                                 *bifa = ifa->lst_next;
2634                                 ifa->lst_next = NULL;
2635                                 addrconf_del_timer(ifa);
2636                                 in6_ifa_put(ifa);
2637                                 continue;
2638                         }
2639                         bifa = &ifa->lst_next;
2640                 }
2641                 write_unlock_bh(&addrconf_hash_lock);
2642         }
2643
2644         write_lock_bh(&idev->lock);
2645
2646         /* Step 3: clear flags for stateless addrconf */
2647         if (!how)
2648                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2649
2650         /* Step 4: clear address list */
2651 #ifdef CONFIG_IPV6_PRIVACY
2652         if (how && del_timer(&idev->regen_timer))
2653                 in6_dev_put(idev);
2654
2655         /* clear tempaddr list */
2656         while ((ifa = idev->tempaddr_list) != NULL) {
2657                 idev->tempaddr_list = ifa->tmp_next;
2658                 ifa->tmp_next = NULL;
2659                 ifa->dead = 1;
2660                 write_unlock_bh(&idev->lock);
2661                 spin_lock_bh(&ifa->lock);
2662
2663                 if (ifa->ifpub) {
2664                         in6_ifa_put(ifa->ifpub);
2665                         ifa->ifpub = NULL;
2666                 }
2667                 spin_unlock_bh(&ifa->lock);
2668                 in6_ifa_put(ifa);
2669                 write_lock_bh(&idev->lock);
2670         }
2671 #endif
2672         while ((ifa = idev->addr_list) != NULL) {
2673                 idev->addr_list = ifa->if_next;
2674                 ifa->if_next = NULL;
2675                 ifa->dead = 1;
2676                 addrconf_del_timer(ifa);
2677                 write_unlock_bh(&idev->lock);
2678
2679                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2680                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2681                 in6_ifa_put(ifa);
2682
2683                 write_lock_bh(&idev->lock);
2684         }
2685         write_unlock_bh(&idev->lock);
2686
2687         /* Step 5: Discard multicast list */
2688
2689         if (how)
2690                 ipv6_mc_destroy_dev(idev);
2691         else
2692                 ipv6_mc_down(idev);
2693
2694         idev->tstamp = jiffies;
2695
2696         /* Shot the device (if unregistered) */
2697
2698         if (how) {
2699                 addrconf_sysctl_unregister(idev);
2700                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2701                 neigh_ifdown(&nd_tbl, dev);
2702                 in6_dev_put(idev);
2703         }
2704         return 0;
2705 }
2706
2707 static void addrconf_rs_timer(unsigned long data)
2708 {
2709         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2710
2711         if (ifp->idev->cnf.forwarding)
2712                 goto out;
2713
2714         if (ifp->idev->if_flags & IF_RA_RCVD) {
2715                 /*
2716                  *      Announcement received after solicitation
2717                  *      was sent
2718                  */
2719                 goto out;
2720         }
2721
2722         spin_lock(&ifp->lock);
2723         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2724                 /* The wait after the last probe can be shorter */
2725                 addrconf_mod_timer(ifp, AC_RS,
2726                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2727                                    ifp->idev->cnf.rtr_solicit_delay :
2728                                    ifp->idev->cnf.rtr_solicit_interval);
2729                 spin_unlock(&ifp->lock);
2730
2731                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2732         } else {
2733                 spin_unlock(&ifp->lock);
2734                 /*
2735                  * Note: we do not support deprecated "all on-link"
2736                  * assumption any longer.
2737                  */
2738                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2739                        ifp->idev->dev->name);
2740         }
2741
2742 out:
2743         in6_ifa_put(ifp);
2744 }
2745
2746 /*
2747  *      Duplicate Address Detection
2748  */
2749 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2750 {
2751         unsigned long rand_num;
2752         struct inet6_dev *idev = ifp->idev;
2753
2754         if (ifp->flags & IFA_F_OPTIMISTIC)
2755                 rand_num = 0;
2756         else
2757                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2758
2759         ifp->probes = idev->cnf.dad_transmits;
2760         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2761 }
2762
2763 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2764 {
2765         struct inet6_dev *idev = ifp->idev;
2766         struct net_device *dev = idev->dev;
2767
2768         addrconf_join_solict(dev, &ifp->addr);
2769
2770         net_srandom(ifp->addr.s6_addr32[3]);
2771
2772         read_lock_bh(&idev->lock);
2773         if (ifp->dead)
2774                 goto out;
2775         spin_lock_bh(&ifp->lock);
2776
2777         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2778             idev->cnf.accept_dad < 1 ||
2779             !(ifp->flags&IFA_F_TENTATIVE) ||
2780             ifp->flags & IFA_F_NODAD) {
2781                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2782                 spin_unlock_bh(&ifp->lock);
2783                 read_unlock_bh(&idev->lock);
2784
2785                 addrconf_dad_completed(ifp);
2786                 return;
2787         }
2788
2789         if (!(idev->if_flags & IF_READY)) {
2790                 spin_unlock_bh(&ifp->lock);
2791                 read_unlock_bh(&idev->lock);
2792                 /*
2793                  * If the device is not ready:
2794                  * - keep it tentative if it is a permanent address.
2795                  * - otherwise, kill it.
2796                  */
2797                 in6_ifa_hold(ifp);
2798                 addrconf_dad_stop(ifp);
2799                 return;
2800         }
2801
2802         /*
2803          * Optimistic nodes can start receiving
2804          * Frames right away
2805          */
2806         if(ifp->flags & IFA_F_OPTIMISTIC)
2807                 ip6_ins_rt(ifp->rt);
2808
2809         addrconf_dad_kick(ifp);
2810         spin_unlock_bh(&ifp->lock);
2811 out:
2812         read_unlock_bh(&idev->lock);
2813 }
2814
2815 static void addrconf_dad_timer(unsigned long data)
2816 {
2817         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2818         struct inet6_dev *idev = ifp->idev;
2819         struct in6_addr mcaddr;
2820
2821         read_lock_bh(&idev->lock);
2822         if (idev->dead) {
2823                 read_unlock_bh(&idev->lock);
2824                 goto out;
2825         }
2826         spin_lock_bh(&ifp->lock);
2827         if (ifp->probes == 0) {
2828                 /*
2829                  * DAD was successful
2830                  */
2831
2832                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC);
2833                 spin_unlock_bh(&ifp->lock);
2834                 read_unlock_bh(&idev->lock);
2835
2836                 addrconf_dad_completed(ifp);
2837
2838                 goto out;
2839         }
2840
2841         ifp->probes--;
2842         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2843         spin_unlock_bh(&ifp->lock);
2844         read_unlock_bh(&idev->lock);
2845
2846         /* send a neighbour solicitation for our addr */
2847         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2848         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2849 out:
2850         in6_ifa_put(ifp);
2851 }
2852
2853 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2854 {
2855         struct net_device *     dev = ifp->idev->dev;
2856
2857         /*
2858          *      Configure the address for reception. Now it is valid.
2859          */
2860
2861         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2862
2863         /* If added prefix is link local and forwarding is off,
2864            start sending router solicitations.
2865          */
2866
2867         if (ifp->idev->cnf.forwarding == 0 &&
2868             ifp->idev->cnf.rtr_solicits > 0 &&
2869             (dev->flags&IFF_LOOPBACK) == 0 &&
2870             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2871                 /*
2872                  *      If a host as already performed a random delay
2873                  *      [...] as part of DAD [...] there is no need
2874                  *      to delay again before sending the first RS
2875                  */
2876                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2877
2878                 spin_lock_bh(&ifp->lock);
2879                 ifp->probes = 1;
2880                 ifp->idev->if_flags |= IF_RS_SENT;
2881                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2882                 spin_unlock_bh(&ifp->lock);
2883         }
2884 }
2885
2886 static void addrconf_dad_run(struct inet6_dev *idev) {
2887         struct inet6_ifaddr *ifp;
2888
2889         read_lock_bh(&idev->lock);
2890         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2891                 spin_lock_bh(&ifp->lock);
2892                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2893                         spin_unlock_bh(&ifp->lock);
2894                         continue;
2895                 }
2896                 spin_unlock_bh(&ifp->lock);
2897                 addrconf_dad_kick(ifp);
2898         }
2899         read_unlock_bh(&idev->lock);
2900 }
2901
2902 #ifdef CONFIG_PROC_FS
2903 struct if6_iter_state {
2904         struct seq_net_private p;
2905         int bucket;
2906 };
2907
2908 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2909 {
2910         struct inet6_ifaddr *ifa = NULL;
2911         struct if6_iter_state *state = seq->private;
2912         struct net *net = seq_file_net(seq);
2913
2914         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2915                 ifa = inet6_addr_lst[state->bucket];
2916
2917                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2918                         ifa = ifa->lst_next;
2919                 if (ifa)
2920                         break;
2921         }
2922         return ifa;
2923 }
2924
2925 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2926 {
2927         struct if6_iter_state *state = seq->private;
2928         struct net *net = seq_file_net(seq);
2929
2930         ifa = ifa->lst_next;
2931 try_again:
2932         if (ifa) {
2933                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2934                         ifa = ifa->lst_next;
2935                         goto try_again;
2936                 }
2937         }
2938
2939         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2940                 ifa = inet6_addr_lst[state->bucket];
2941                 goto try_again;
2942         }
2943
2944         return ifa;
2945 }
2946
2947 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2948 {
2949         struct inet6_ifaddr *ifa = if6_get_first(seq);
2950
2951         if (ifa)
2952                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2953                         --pos;
2954         return pos ? NULL : ifa;
2955 }
2956
2957 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2958         __acquires(addrconf_hash_lock)
2959 {
2960         read_lock_bh(&addrconf_hash_lock);
2961         return if6_get_idx(seq, *pos);
2962 }
2963
2964 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2965 {
2966         struct inet6_ifaddr *ifa;
2967
2968         ifa = if6_get_next(seq, v);
2969         ++*pos;
2970         return ifa;
2971 }
2972
2973 static void if6_seq_stop(struct seq_file *seq, void *v)
2974         __releases(addrconf_hash_lock)
2975 {
2976         read_unlock_bh(&addrconf_hash_lock);
2977 }
2978
2979 static int if6_seq_show(struct seq_file *seq, void *v)
2980 {
2981         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2982         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2983                    &ifp->addr,
2984                    ifp->idev->dev->ifindex,
2985                    ifp->prefix_len,
2986                    ifp->scope,
2987                    ifp->flags,
2988                    ifp->idev->dev->name);
2989         return 0;
2990 }
2991
2992 static const struct seq_operations if6_seq_ops = {
2993         .start  = if6_seq_start,
2994         .next   = if6_seq_next,
2995         .show   = if6_seq_show,
2996         .stop   = if6_seq_stop,
2997 };
2998
2999 static int if6_seq_open(struct inode *inode, struct file *file)
3000 {
3001         return seq_open_net(inode, file, &if6_seq_ops,
3002                             sizeof(struct if6_iter_state));
3003 }
3004
3005 static const struct file_operations if6_fops = {
3006         .owner          = THIS_MODULE,
3007         .open           = if6_seq_open,
3008         .read           = seq_read,
3009         .llseek         = seq_lseek,
3010         .release        = seq_release_net,
3011 };
3012
3013 static int if6_proc_net_init(struct net *net)
3014 {
3015         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3016                 return -ENOMEM;
3017         return 0;
3018 }
3019
3020 static void if6_proc_net_exit(struct net *net)
3021 {
3022        proc_net_remove(net, "if_inet6");
3023 }
3024
3025 static struct pernet_operations if6_proc_net_ops = {
3026        .init = if6_proc_net_init,
3027        .exit = if6_proc_net_exit,
3028 };
3029
3030 int __init if6_proc_init(void)
3031 {
3032         return register_pernet_subsys(&if6_proc_net_ops);
3033 }
3034
3035 void if6_proc_exit(void)
3036 {
3037         unregister_pernet_subsys(&if6_proc_net_ops);
3038 }
3039 #endif  /* CONFIG_PROC_FS */
3040
3041 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3042 /* Check if address is a home address configured on any interface. */
3043 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3044 {
3045         int ret = 0;
3046         struct inet6_ifaddr * ifp;
3047         u8 hash = ipv6_addr_hash(addr);
3048         read_lock_bh(&addrconf_hash_lock);
3049         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3050                 if (!net_eq(dev_net(ifp->idev->dev), net))
3051                         continue;
3052                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3053                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3054                         ret = 1;
3055                         break;
3056                 }
3057         }
3058         read_unlock_bh(&addrconf_hash_lock);
3059         return ret;
3060 }
3061 #endif
3062
3063 /*
3064  *      Periodic address status verification
3065  */
3066
3067 static void addrconf_verify(unsigned long foo)
3068 {
3069         struct inet6_ifaddr *ifp;
3070         unsigned long now, next;
3071         int i;
3072
3073         spin_lock_bh(&addrconf_verify_lock);
3074         now = jiffies;
3075         next = now + ADDR_CHECK_FREQUENCY;
3076
3077         del_timer(&addr_chk_timer);
3078
3079         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3080
3081 restart:
3082                 read_lock(&addrconf_hash_lock);
3083                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3084                         unsigned long age;
3085 #ifdef CONFIG_IPV6_PRIVACY
3086                         unsigned long regen_advance;
3087 #endif
3088
3089                         if (ifp->flags & IFA_F_PERMANENT)
3090                                 continue;
3091
3092                         spin_lock(&ifp->lock);
3093                         age = (now - ifp->tstamp) / HZ;
3094
3095 #ifdef CONFIG_IPV6_PRIVACY
3096                         regen_advance = ifp->idev->cnf.regen_max_retry *
3097                                         ifp->idev->cnf.dad_transmits *
3098                                         ifp->idev->nd_parms->retrans_time / HZ;
3099 #endif
3100
3101                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3102                             age >= ifp->valid_lft) {
3103                                 spin_unlock(&ifp->lock);
3104                                 in6_ifa_hold(ifp);
3105                                 read_unlock(&addrconf_hash_lock);
3106                                 ipv6_del_addr(ifp);
3107                                 goto restart;
3108                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3109                                 spin_unlock(&ifp->lock);
3110                                 continue;
3111                         } else if (age >= ifp->prefered_lft) {
3112                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3113                                 int deprecate = 0;
3114
3115                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3116                                         deprecate = 1;
3117                                         ifp->flags |= IFA_F_DEPRECATED;
3118                                 }
3119
3120                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3121                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3122
3123                                 spin_unlock(&ifp->lock);
3124
3125                                 if (deprecate) {
3126                                         in6_ifa_hold(ifp);
3127                                         read_unlock(&addrconf_hash_lock);
3128
3129                                         ipv6_ifa_notify(0, ifp);
3130                                         in6_ifa_put(ifp);
3131                                         goto restart;
3132                                 }
3133 #ifdef CONFIG_IPV6_PRIVACY
3134                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3135                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3136                                 if (age >= ifp->prefered_lft - regen_advance) {
3137                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3138                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3139                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3140                                         if (!ifp->regen_count && ifpub) {
3141                                                 ifp->regen_count++;
3142                                                 in6_ifa_hold(ifp);
3143                                                 in6_ifa_hold(ifpub);
3144                                                 spin_unlock(&ifp->lock);
3145                                                 read_unlock(&addrconf_hash_lock);
3146                                                 spin_lock(&ifpub->lock);
3147                                                 ifpub->regen_count = 0;
3148                                                 spin_unlock(&ifpub->lock);
3149                                                 ipv6_create_tempaddr(ifpub, ifp);
3150                                                 in6_ifa_put(ifpub);
3151                                                 in6_ifa_put(ifp);
3152                                                 goto restart;
3153                                         }
3154                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3155                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3156                                 spin_unlock(&ifp->lock);
3157 #endif
3158                         } else {
3159                                 /* ifp->prefered_lft <= ifp->valid_lft */
3160                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3161                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3162                                 spin_unlock(&ifp->lock);
3163                         }
3164                 }
3165                 read_unlock(&addrconf_hash_lock);
3166         }
3167
3168         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3169         add_timer(&addr_chk_timer);
3170         spin_unlock_bh(&addrconf_verify_lock);
3171 }
3172
3173 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3174 {
3175         struct in6_addr *pfx = NULL;
3176
3177         if (addr)
3178                 pfx = nla_data(addr);
3179
3180         if (local) {
3181                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3182                         pfx = NULL;
3183                 else
3184                         pfx = nla_data(local);
3185         }
3186
3187         return pfx;
3188 }
3189
3190 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3191         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3192         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3193         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3194 };
3195
3196 static int
3197 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3198 {
3199         struct net *net = sock_net(skb->sk);
3200         struct ifaddrmsg *ifm;
3201         struct nlattr *tb[IFA_MAX+1];
3202         struct in6_addr *pfx;
3203         int err;
3204
3205         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3206         if (err < 0)
3207                 return err;
3208
3209         ifm = nlmsg_data(nlh);
3210         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3211         if (pfx == NULL)
3212                 return -EINVAL;
3213
3214         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3215 }
3216
3217 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3218                              u32 prefered_lft, u32 valid_lft)
3219 {
3220         u32 flags;
3221         clock_t expires;
3222         unsigned long timeout;
3223
3224         if (!valid_lft || (prefered_lft > valid_lft))
3225                 return -EINVAL;
3226
3227         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3228         if (addrconf_finite_timeout(timeout)) {
3229                 expires = jiffies_to_clock_t(timeout * HZ);
3230                 valid_lft = timeout;
3231                 flags = RTF_EXPIRES;
3232         } else {
3233                 expires = 0;
3234                 flags = 0;
3235                 ifa_flags |= IFA_F_PERMANENT;
3236         }
3237
3238         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3239         if (addrconf_finite_timeout(timeout)) {
3240                 if (timeout == 0)
3241                         ifa_flags |= IFA_F_DEPRECATED;
3242                 prefered_lft = timeout;
3243         }
3244
3245         spin_lock_bh(&ifp->lock);
3246         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3247         ifp->tstamp = jiffies;
3248         ifp->valid_lft = valid_lft;
3249         ifp->prefered_lft = prefered_lft;
3250
3251         spin_unlock_bh(&ifp->lock);
3252         if (!(ifp->flags&IFA_F_TENTATIVE))
3253                 ipv6_ifa_notify(0, ifp);
3254
3255         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3256                               expires, flags);
3257         addrconf_verify(0);
3258
3259         return 0;
3260 }
3261
3262 static int
3263 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3264 {
3265         struct net *net = sock_net(skb->sk);
3266         struct ifaddrmsg *ifm;
3267         struct nlattr *tb[IFA_MAX+1];
3268         struct in6_addr *pfx;
3269         struct inet6_ifaddr *ifa;
3270         struct net_device *dev;
3271         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3272         u8 ifa_flags;
3273         int err;
3274
3275         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3276         if (err < 0)
3277                 return err;
3278
3279         ifm = nlmsg_data(nlh);
3280         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3281         if (pfx == NULL)
3282                 return -EINVAL;
3283
3284         if (tb[IFA_CACHEINFO]) {
3285                 struct ifa_cacheinfo *ci;
3286
3287                 ci = nla_data(tb[IFA_CACHEINFO]);
3288                 valid_lft = ci->ifa_valid;
3289                 preferred_lft = ci->ifa_prefered;
3290         } else {
3291                 preferred_lft = INFINITY_LIFE_TIME;
3292                 valid_lft = INFINITY_LIFE_TIME;
3293         }
3294
3295         dev =  __dev_get_by_index(net, ifm->ifa_index);
3296         if (dev == NULL)
3297                 return -ENODEV;
3298
3299         /* We ignore other flags so far. */
3300         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3301
3302         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3303         if (ifa == NULL) {
3304                 /*
3305                  * It would be best to check for !NLM_F_CREATE here but
3306                  * userspace alreay relies on not having to provide this.
3307                  */
3308                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3309                                       ifm->ifa_prefixlen, ifa_flags,
3310                                       preferred_lft, valid_lft);
3311         }
3312
3313         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3314             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3315                 err = -EEXIST;
3316         else
3317                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3318
3319         in6_ifa_put(ifa);
3320
3321         return err;
3322 }
3323
3324 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3325                           u8 scope, int ifindex)
3326 {
3327         struct ifaddrmsg *ifm;
3328
3329         ifm = nlmsg_data(nlh);
3330         ifm->ifa_family = AF_INET6;
3331         ifm->ifa_prefixlen = prefixlen;
3332         ifm->ifa_flags = flags;
3333         ifm->ifa_scope = scope;
3334         ifm->ifa_index = ifindex;
3335 }
3336
3337 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3338                          unsigned long tstamp, u32 preferred, u32 valid)
3339 {
3340         struct ifa_cacheinfo ci;
3341
3342         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3343                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3344         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3345                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3346         ci.ifa_prefered = preferred;
3347         ci.ifa_valid = valid;
3348
3349         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3350 }
3351
3352 static inline int rt_scope(int ifa_scope)
3353 {
3354         if (ifa_scope & IFA_HOST)
3355                 return RT_SCOPE_HOST;
3356         else if (ifa_scope & IFA_LINK)
3357                 return RT_SCOPE_LINK;
3358         else if (ifa_scope & IFA_SITE)
3359                 return RT_SCOPE_SITE;
3360         else
3361                 return RT_SCOPE_UNIVERSE;
3362 }
3363
3364 static inline int inet6_ifaddr_msgsize(void)
3365 {
3366         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3367                + nla_total_size(16) /* IFA_ADDRESS */
3368                + nla_total_size(sizeof(struct ifa_cacheinfo));
3369 }
3370
3371 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3372                              u32 pid, u32 seq, int event, unsigned int flags)
3373 {
3374         struct nlmsghdr  *nlh;
3375         u32 preferred, valid;
3376
3377         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3378         if (nlh == NULL)
3379                 return -EMSGSIZE;
3380
3381         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3382                       ifa->idev->dev->ifindex);
3383
3384         if (!(ifa->flags&IFA_F_PERMANENT)) {
3385                 preferred = ifa->prefered_lft;
3386                 valid = ifa->valid_lft;
3387                 if (preferred != INFINITY_LIFE_TIME) {
3388                         long tval = (jiffies - ifa->tstamp)/HZ;
3389                         if (preferred > tval)
3390                                 preferred -= tval;
3391                         else
3392                                 preferred = 0;
3393                         if (valid != INFINITY_LIFE_TIME)
3394                                 valid -= tval;
3395                 }
3396         } else {
3397                 preferred = INFINITY_LIFE_TIME;
3398                 valid = INFINITY_LIFE_TIME;
3399         }
3400
3401         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3402             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3403                 nlmsg_cancel(skb, nlh);
3404                 return -EMSGSIZE;
3405         }
3406
3407         return nlmsg_end(skb, nlh);
3408 }
3409
3410 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3411                                 u32 pid, u32 seq, int event, u16 flags)
3412 {
3413         struct nlmsghdr  *nlh;
3414         u8 scope = RT_SCOPE_UNIVERSE;
3415         int ifindex = ifmca->idev->dev->ifindex;
3416
3417         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3418                 scope = RT_SCOPE_SITE;
3419
3420         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3421         if (nlh == NULL)
3422                 return -EMSGSIZE;
3423
3424         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3425         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3426             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3427                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3428                 nlmsg_cancel(skb, nlh);
3429                 return -EMSGSIZE;
3430         }
3431
3432         return nlmsg_end(skb, nlh);
3433 }
3434
3435 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3436                                 u32 pid, u32 seq, int event, unsigned int flags)
3437 {
3438         struct nlmsghdr  *nlh;
3439         u8 scope = RT_SCOPE_UNIVERSE;
3440         int ifindex = ifaca->aca_idev->dev->ifindex;
3441
3442         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3443                 scope = RT_SCOPE_SITE;
3444
3445         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3446         if (nlh == NULL)
3447                 return -EMSGSIZE;
3448
3449         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3450         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3451             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3452                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3453                 nlmsg_cancel(skb, nlh);
3454                 return -EMSGSIZE;
3455         }
3456
3457         return nlmsg_end(skb, nlh);
3458 }
3459
3460 enum addr_type_t
3461 {
3462         UNICAST_ADDR,
3463         MULTICAST_ADDR,
3464         ANYCAST_ADDR,
3465 };
3466
3467 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3468                            enum addr_type_t type)
3469 {
3470         int idx, ip_idx;
3471         int s_idx, s_ip_idx;
3472         int err = 1;
3473         struct net_device *dev;
3474         struct inet6_dev *idev = NULL;
3475         struct inet6_ifaddr *ifa;
3476         struct ifmcaddr6 *ifmca;
3477         struct ifacaddr6 *ifaca;
3478         struct net *net = sock_net(skb->sk);
3479
3480         s_idx = cb->args[0];
3481         s_ip_idx = ip_idx = cb->args[1];
3482
3483         idx = 0;
3484         for_each_netdev(net, dev) {
3485                 if (idx < s_idx)
3486                         goto cont;
3487                 if (idx > s_idx)
3488                         s_ip_idx = 0;
3489                 ip_idx = 0;
3490                 if ((idev = in6_dev_get(dev)) == NULL)
3491                         goto cont;
3492                 read_lock_bh(&idev->lock);
3493                 switch (type) {
3494                 case UNICAST_ADDR:
3495                         /* unicast address incl. temp addr */
3496                         for (ifa = idev->addr_list; ifa;
3497                              ifa = ifa->if_next, ip_idx++) {
3498                                 if (ip_idx < s_ip_idx)
3499                                         continue;
3500                                 err = inet6_fill_ifaddr(skb, ifa,
3501                                                         NETLINK_CB(cb->skb).pid,
3502                                                         cb->nlh->nlmsg_seq,
3503                                                         RTM_NEWADDR,
3504                                                         NLM_F_MULTI);
3505                         }
3506                         break;
3507                 case MULTICAST_ADDR:
3508                         /* multicast address */
3509                         for (ifmca = idev->mc_list; ifmca;
3510                              ifmca = ifmca->next, ip_idx++) {
3511                                 if (ip_idx < s_ip_idx)
3512                                         continue;
3513                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3514                                                           NETLINK_CB(cb->skb).pid,
3515                                                           cb->nlh->nlmsg_seq,
3516                                                           RTM_GETMULTICAST,
3517                                                           NLM_F_MULTI);
3518                         }
3519                         break;
3520                 case ANYCAST_ADDR:
3521                         /* anycast address */
3522                         for (ifaca = idev->ac_list; ifaca;
3523                              ifaca = ifaca->aca_next, ip_idx++) {
3524                                 if (ip_idx < s_ip_idx)
3525                                         continue;
3526                                 err = inet6_fill_ifacaddr(skb, ifaca,
3527                                                           NETLINK_CB(cb->skb).pid,
3528                                                           cb->nlh->nlmsg_seq,
3529                                                           RTM_GETANYCAST,
3530                                                           NLM_F_MULTI);
3531                         }
3532                         break;
3533                 default:
3534                         break;
3535                 }
3536                 read_unlock_bh(&idev->lock);
3537                 in6_dev_put(idev);
3538
3539                 if (err <= 0)
3540                         break;
3541 cont:
3542                 idx++;
3543         }
3544         cb->args[0] = idx;
3545         cb->args[1] = ip_idx;
3546         return skb->len;
3547 }
3548
3549 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3550 {
3551         enum addr_type_t type = UNICAST_ADDR;
3552
3553         return inet6_dump_addr(skb, cb, type);
3554 }
3555
3556 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3557 {
3558         enum addr_type_t type = MULTICAST_ADDR;
3559
3560         return inet6_dump_addr(skb, cb, type);
3561 }
3562
3563
3564 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3565 {
3566         enum addr_type_t type = ANYCAST_ADDR;
3567
3568         return inet6_dump_addr(skb, cb, type);
3569 }
3570
3571 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3572                              void *arg)
3573 {
3574         struct net *net = sock_net(in_skb->sk);
3575         struct ifaddrmsg *ifm;
3576         struct nlattr *tb[IFA_MAX+1];
3577         struct in6_addr *addr = NULL;
3578         struct net_device *dev = NULL;
3579         struct inet6_ifaddr *ifa;
3580         struct sk_buff *skb;
3581         int err;
3582
3583         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3584         if (err < 0)
3585                 goto errout;
3586
3587         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3588         if (addr == NULL) {
3589                 err = -EINVAL;
3590                 goto errout;
3591         }
3592
3593         ifm = nlmsg_data(nlh);
3594         if (ifm->ifa_index)
3595                 dev = __dev_get_by_index(net, ifm->ifa_index);
3596
3597         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3598                 err = -EADDRNOTAVAIL;
3599                 goto errout;
3600         }
3601
3602         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3603                 err = -ENOBUFS;
3604                 goto errout_ifa;
3605         }
3606
3607         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3608                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3609         if (err < 0) {
3610                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3611                 WARN_ON(err == -EMSGSIZE);
3612                 kfree_skb(skb);
3613                 goto errout_ifa;
3614         }
3615         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3616 errout_ifa:
3617         in6_ifa_put(ifa);
3618 errout:
3619         return err;
3620 }
3621
3622 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3623 {
3624         struct sk_buff *skb;
3625         struct net *net = dev_net(ifa->idev->dev);
3626         int err = -ENOBUFS;
3627
3628         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3629         if (skb == NULL)
3630                 goto errout;
3631
3632         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3633         if (err < 0) {
3634                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3635                 WARN_ON(err == -EMSGSIZE);
3636                 kfree_skb(skb);
3637                 goto errout;
3638         }
3639         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3640         return;
3641 errout:
3642         if (err < 0)
3643                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3644 }
3645
3646 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3647                                 __s32 *array, int bytes)
3648 {
3649         BUG_ON(bytes < (DEVCONF_MAX * 4));
3650
3651         memset(array, 0, bytes);
3652         array[DEVCONF_FORWARDING] = cnf->forwarding;
3653         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3654         array[DEVCONF_MTU6] = cnf->mtu6;
3655         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3656         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3657         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3658         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3659         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3660         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3661         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3662         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3663 #ifdef CONFIG_IPV6_PRIVACY
3664         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3665         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3666         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3667         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3668         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3669 #endif
3670         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3671         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3672         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3673 #ifdef CONFIG_IPV6_ROUTER_PREF
3674         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3675         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3676 #ifdef CONFIG_IPV6_ROUTE_INFO
3677         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3678 #endif
3679 #endif
3680         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3681         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3682 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3683         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3684 #endif
3685 #ifdef CONFIG_IPV6_MROUTE
3686         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3687 #endif
3688         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3689         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3690 }
3691
3692 static inline size_t inet6_if_nlmsg_size(void)
3693 {
3694         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3695                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3696                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3697                + nla_total_size(4) /* IFLA_MTU */
3698                + nla_total_size(4) /* IFLA_LINK */
3699                + nla_total_size( /* IFLA_PROTINFO */
3700                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3701                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3702                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3703                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3704                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3705                  );
3706 }
3707
3708 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3709                                       int bytes)
3710 {
3711         int i;
3712         int pad = bytes - sizeof(u64) * items;
3713         BUG_ON(pad < 0);
3714
3715         /* Use put_unaligned() because stats may not be aligned for u64. */
3716         put_unaligned(items, &stats[0]);
3717         for (i = 1; i < items; i++)
3718                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3719
3720         memset(&stats[items], 0, pad);
3721 }
3722
3723 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3724                              int bytes)
3725 {
3726         switch(attrtype) {
3727         case IFLA_INET6_STATS:
3728                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3729                 break;
3730         case IFLA_INET6_ICMP6STATS:
3731                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3732                 break;
3733         }
3734 }
3735
3736 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3737                              u32 pid, u32 seq, int event, unsigned int flags)
3738 {
3739         struct net_device *dev = idev->dev;
3740         struct nlattr *nla;
3741         struct ifinfomsg *hdr;
3742         struct nlmsghdr *nlh;
3743         void *protoinfo;
3744         struct ifla_cacheinfo ci;
3745
3746         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3747         if (nlh == NULL)
3748                 return -EMSGSIZE;
3749
3750         hdr = nlmsg_data(nlh);
3751         hdr->ifi_family = AF_INET6;
3752         hdr->__ifi_pad = 0;
3753         hdr->ifi_type = dev->type;
3754         hdr->ifi_index = dev->ifindex;
3755         hdr->ifi_flags = dev_get_flags(dev);
3756         hdr->ifi_change = 0;
3757
3758         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3759
3760         if (dev->addr_len)
3761                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3762
3763         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3764         if (dev->ifindex != dev->iflink)
3765                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3766
3767         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3768         if (protoinfo == NULL)
3769                 goto nla_put_failure;
3770
3771         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3772
3773         ci.max_reasm_len = IPV6_MAXPLEN;
3774         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3775                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3776         ci.reachable_time = idev->nd_parms->reachable_time;
3777         ci.retrans_time = idev->nd_parms->retrans_time;
3778         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3779
3780         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3781         if (nla == NULL)
3782                 goto nla_put_failure;
3783         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3784
3785         /* XXX - MC not implemented */
3786
3787         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3788         if (nla == NULL)
3789                 goto nla_put_failure;
3790         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3791
3792         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3793         if (nla == NULL)
3794                 goto nla_put_failure;
3795         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3796
3797         nla_nest_end(skb, protoinfo);
3798         return nlmsg_end(skb, nlh);
3799
3800 nla_put_failure:
3801         nlmsg_cancel(skb, nlh);
3802         return -EMSGSIZE;
3803 }
3804
3805 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3806 {
3807         struct net *net = sock_net(skb->sk);
3808         int idx, err;
3809         int s_idx = cb->args[0];
3810         struct net_device *dev;
3811         struct inet6_dev *idev;
3812
3813         read_lock(&dev_base_lock);
3814         idx = 0;
3815         for_each_netdev(net, dev) {
3816                 if (idx < s_idx)
3817                         goto cont;
3818                 if ((idev = in6_dev_get(dev)) == NULL)
3819                         goto cont;
3820                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3821                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3822                 in6_dev_put(idev);
3823                 if (err <= 0)
3824                         break;
3825 cont:
3826                 idx++;
3827         }
3828         read_unlock(&dev_base_lock);
3829         cb->args[0] = idx;
3830
3831         return skb->len;
3832 }
3833
3834 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3835 {
3836         struct sk_buff *skb;
3837         struct net *net = dev_net(idev->dev);
3838         int err = -ENOBUFS;
3839
3840         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3841         if (skb == NULL)
3842                 goto errout;
3843
3844         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3845         if (err < 0) {
3846                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3847                 WARN_ON(err == -EMSGSIZE);
3848                 kfree_skb(skb);
3849                 goto errout;
3850         }
3851         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3852         return;
3853 errout:
3854         if (err < 0)
3855                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3856 }
3857
3858 static inline size_t inet6_prefix_nlmsg_size(void)
3859 {
3860         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3861                + nla_total_size(sizeof(struct in6_addr))
3862                + nla_total_size(sizeof(struct prefix_cacheinfo));
3863 }
3864
3865 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3866                              struct prefix_info *pinfo, u32 pid, u32 seq,
3867                              int event, unsigned int flags)
3868 {
3869         struct prefixmsg *pmsg;
3870         struct nlmsghdr *nlh;
3871         struct prefix_cacheinfo ci;
3872
3873         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3874         if (nlh == NULL)
3875                 return -EMSGSIZE;
3876
3877         pmsg = nlmsg_data(nlh);
3878         pmsg->prefix_family = AF_INET6;
3879         pmsg->prefix_pad1 = 0;
3880         pmsg->prefix_pad2 = 0;
3881         pmsg->prefix_ifindex = idev->dev->ifindex;
3882         pmsg->prefix_len = pinfo->prefix_len;
3883         pmsg->prefix_type = pinfo->type;
3884         pmsg->prefix_pad3 = 0;
3885         pmsg->prefix_flags = 0;
3886         if (pinfo->onlink)
3887                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3888         if (pinfo->autoconf)
3889                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3890
3891         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3892
3893         ci.preferred_time = ntohl(pinfo->prefered);
3894         ci.valid_time = ntohl(pinfo->valid);
3895         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3896
3897         return nlmsg_end(skb, nlh);
3898
3899 nla_put_failure:
3900         nlmsg_cancel(skb, nlh);
3901         return -EMSGSIZE;
3902 }
3903
3904 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3905                          struct prefix_info *pinfo)
3906 {
3907         struct sk_buff *skb;
3908         struct net *net = dev_net(idev->dev);
3909         int err = -ENOBUFS;
3910
3911         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3912         if (skb == NULL)
3913                 goto errout;
3914
3915         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3916         if (err < 0) {
3917                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3918                 WARN_ON(err == -EMSGSIZE);
3919                 kfree_skb(skb);
3920                 goto errout;
3921         }
3922         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3923         return;
3924 errout:
3925         if (err < 0)
3926                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3927 }
3928
3929 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3930 {
3931         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3932
3933         switch (event) {
3934         case RTM_NEWADDR:
3935                 /*
3936                  * If the address was optimistic
3937                  * we inserted the route at the start of
3938                  * our DAD process, so we don't need
3939                  * to do it again
3940                  */
3941                 if (!(ifp->rt->rt6i_node))
3942                         ip6_ins_rt(ifp->rt);
3943                 if (ifp->idev->cnf.forwarding)
3944                         addrconf_join_anycast(ifp);
3945                 break;
3946         case RTM_DELADDR:
3947                 if (ifp->idev->cnf.forwarding)
3948                         addrconf_leave_anycast(ifp);
3949                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3950                 dst_hold(&ifp->rt->u.dst);
3951                 if (ip6_del_rt(ifp->rt))
3952                         dst_free(&ifp->rt->u.dst);
3953                 break;
3954         }
3955 }
3956
3957 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3958 {
3959         rcu_read_lock_bh();
3960         if (likely(ifp->idev->dead == 0))
3961                 __ipv6_ifa_notify(event, ifp);
3962         rcu_read_unlock_bh();
3963 }
3964
3965 #ifdef CONFIG_SYSCTL
3966
3967 static
3968 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3969                            void __user *buffer, size_t *lenp, loff_t *ppos)
3970 {
3971         int *valp = ctl->data;
3972         int val = *valp;
3973         int ret;
3974
3975         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3976
3977         if (write)
3978                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3979         return ret;
3980 }
3981
3982 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3983                                             void __user *oldval,
3984                                             size_t __user *oldlenp,
3985                                             void __user *newval, size_t newlen)
3986 {
3987         int *valp = table->data;
3988         int val = *valp;
3989         int new;
3990
3991         if (!newval || !newlen)
3992                 return 0;
3993         if (newlen != sizeof(int))
3994                 return -EINVAL;
3995         if (get_user(new, (int __user *)newval))
3996                 return -EFAULT;
3997         if (new == *valp)
3998                 return 0;
3999         if (oldval && oldlenp) {
4000                 size_t len;
4001                 if (get_user(len, oldlenp))
4002                         return -EFAULT;
4003                 if (len) {
4004                         if (len > table->maxlen)
4005                                 len = table->maxlen;
4006                         if (copy_to_user(oldval, valp, len))
4007                                 return -EFAULT;
4008                         if (put_user(len, oldlenp))
4009                                 return -EFAULT;
4010                 }
4011         }
4012
4013         *valp = new;
4014         return addrconf_fixup_forwarding(table, valp, val);
4015 }
4016
4017 static void dev_disable_change(struct inet6_dev *idev)
4018 {
4019         if (!idev || !idev->dev)
4020                 return;
4021
4022         if (idev->cnf.disable_ipv6)
4023                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4024         else
4025                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4026 }
4027
4028 static void addrconf_disable_change(struct net *net, __s32 newf)
4029 {
4030         struct net_device *dev;
4031         struct inet6_dev *idev;
4032
4033         read_lock(&dev_base_lock);
4034         for_each_netdev(net, dev) {
4035                 rcu_read_lock();
4036                 idev = __in6_dev_get(dev);
4037                 if (idev) {
4038                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4039                         idev->cnf.disable_ipv6 = newf;
4040                         if (changed)
4041                                 dev_disable_change(idev);
4042                 }
4043                 rcu_read_unlock();
4044         }
4045         read_unlock(&dev_base_lock);
4046 }
4047
4048 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4049 {
4050         struct net *net;
4051
4052         net = (struct net *)table->extra2;
4053
4054         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4055                 return 0;
4056
4057         if (!rtnl_trylock())
4058                 return restart_syscall();
4059
4060         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4061                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4062                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4063                 addrconf_disable_change(net, newf);
4064         } else if ((!*p) ^ (!old))
4065                 dev_disable_change((struct inet6_dev *)table->extra1);
4066
4067         rtnl_unlock();
4068         return 0;
4069 }
4070
4071 static
4072 int addrconf_sysctl_disable(ctl_table *ctl, int write, struct file * filp,
4073                             void __user *buffer, size_t *lenp, loff_t *ppos)
4074 {
4075         int *valp = ctl->data;
4076         int val = *valp;
4077         int ret;
4078
4079         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
4080
4081         if (write)
4082                 ret = addrconf_disable_ipv6(ctl, valp, val);
4083         return ret;
4084 }
4085
4086 static struct addrconf_sysctl_table
4087 {
4088         struct ctl_table_header *sysctl_header;
4089         ctl_table addrconf_vars[DEVCONF_MAX+1];
4090         char *dev_name;
4091 } addrconf_sysctl __read_mostly = {
4092         .sysctl_header = NULL,
4093         .addrconf_vars = {
4094                 {
4095                         .ctl_name       =       NET_IPV6_FORWARDING,
4096                         .procname       =       "forwarding",
4097                         .data           =       &ipv6_devconf.forwarding,
4098                         .maxlen         =       sizeof(int),
4099                         .mode           =       0644,
4100                         .proc_handler   =       addrconf_sysctl_forward,
4101                         .strategy       =       addrconf_sysctl_forward_strategy,
4102                 },
4103                 {
4104                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4105                         .procname       =       "hop_limit",
4106                         .data           =       &ipv6_devconf.hop_limit,
4107                         .maxlen         =       sizeof(int),
4108                         .mode           =       0644,
4109                         .proc_handler   =       proc_dointvec,
4110                 },
4111                 {
4112                         .ctl_name       =       NET_IPV6_MTU,
4113                         .procname       =       "mtu",
4114                         .data           =       &ipv6_devconf.mtu6,
4115                         .maxlen         =       sizeof(int),
4116                         .mode           =       0644,
4117                         .proc_handler   =       proc_dointvec,
4118                 },
4119                 {
4120                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4121                         .procname       =       "accept_ra",
4122                         .data           =       &ipv6_devconf.accept_ra,
4123                         .maxlen         =       sizeof(int),
4124                         .mode           =       0644,
4125                         .proc_handler   =       proc_dointvec,
4126                 },
4127                 {
4128                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4129                         .procname       =       "accept_redirects",
4130                         .data           =       &ipv6_devconf.accept_redirects,
4131                         .maxlen         =       sizeof(int),
4132                         .mode           =       0644,
4133                         .proc_handler   =       proc_dointvec,
4134                 },
4135                 {
4136                         .ctl_name       =       NET_IPV6_AUTOCONF,
4137                         .procname       =       "autoconf",
4138                         .data           =       &ipv6_devconf.autoconf,
4139                         .maxlen         =       sizeof(int),
4140                         .mode           =       0644,
4141                         .proc_handler   =       proc_dointvec,
4142                 },
4143                 {
4144                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4145                         .procname       =       "dad_transmits",
4146                         .data           =       &ipv6_devconf.dad_transmits,
4147                         .maxlen         =       sizeof(int),
4148                         .mode           =       0644,
4149                         .proc_handler   =       proc_dointvec,
4150                 },
4151                 {
4152                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4153                         .procname       =       "router_solicitations",
4154                         .data           =       &ipv6_devconf.rtr_solicits,
4155                         .maxlen         =       sizeof(int),
4156                         .mode           =       0644,
4157                         .proc_handler   =       proc_dointvec,
4158                 },
4159                 {
4160                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4161                         .procname       =       "router_solicitation_interval",
4162                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4163                         .maxlen         =       sizeof(int),
4164                         .mode           =       0644,
4165                         .proc_handler   =       proc_dointvec_jiffies,
4166                         .strategy       =       sysctl_jiffies,
4167                 },
4168                 {
4169                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4170                         .procname       =       "router_solicitation_delay",
4171                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4172                         .maxlen         =       sizeof(int),
4173                         .mode           =       0644,
4174                         .proc_handler   =       proc_dointvec_jiffies,
4175                         .strategy       =       sysctl_jiffies,
4176                 },
4177                 {
4178                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4179                         .procname       =       "force_mld_version",
4180                         .data           =       &ipv6_devconf.force_mld_version,
4181                         .maxlen         =       sizeof(int),
4182                         .mode           =       0644,
4183                         .proc_handler   =       proc_dointvec,
4184                 },
4185 #ifdef CONFIG_IPV6_PRIVACY
4186                 {
4187                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4188                         .procname       =       "use_tempaddr",
4189                         .data           =       &ipv6_devconf.use_tempaddr,
4190                         .maxlen         =       sizeof(int),
4191                         .mode           =       0644,
4192                         .proc_handler   =       proc_dointvec,
4193                 },
4194                 {
4195                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4196                         .procname       =       "temp_valid_lft",
4197                         .data           =       &ipv6_devconf.temp_valid_lft,
4198                         .maxlen         =       sizeof(int),
4199                         .mode           =       0644,
4200                         .proc_handler   =       proc_dointvec,
4201                 },
4202                 {
4203                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4204                         .procname       =       "temp_prefered_lft",
4205                         .data           =       &ipv6_devconf.temp_prefered_lft,
4206                         .maxlen         =       sizeof(int),
4207                         .mode           =       0644,
4208                         .proc_handler   =       proc_dointvec,
4209                 },
4210                 {
4211                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4212                         .procname       =       "regen_max_retry",
4213                         .data           =       &ipv6_devconf.regen_max_retry,
4214                         .maxlen         =       sizeof(int),
4215                         .mode           =       0644,
4216                         .proc_handler   =       proc_dointvec,
4217                 },
4218                 {
4219                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4220                         .procname       =       "max_desync_factor",
4221                         .data           =       &ipv6_devconf.max_desync_factor,
4222                         .maxlen         =       sizeof(int),
4223                         .mode           =       0644,
4224                         .proc_handler   =       proc_dointvec,
4225                 },
4226 #endif
4227                 {
4228                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4229                         .procname       =       "max_addresses",
4230                         .data           =       &ipv6_devconf.max_addresses,
4231                         .maxlen         =       sizeof(int),
4232                         .mode           =       0644,
4233                         .proc_handler   =       proc_dointvec,
4234                 },
4235                 {
4236                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4237                         .procname       =       "accept_ra_defrtr",
4238                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4239                         .maxlen         =       sizeof(int),
4240                         .mode           =       0644,
4241                         .proc_handler   =       proc_dointvec,
4242                 },
4243                 {
4244                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4245                         .procname       =       "accept_ra_pinfo",
4246                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4247                         .maxlen         =       sizeof(int),
4248                         .mode           =       0644,
4249                         .proc_handler   =       proc_dointvec,
4250                 },
4251 #ifdef CONFIG_IPV6_ROUTER_PREF
4252                 {
4253                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4254                         .procname       =       "accept_ra_rtr_pref",
4255                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4256                         .maxlen         =       sizeof(int),
4257                         .mode           =       0644,
4258                         .proc_handler   =       proc_dointvec,
4259                 },
4260                 {
4261                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4262                         .procname       =       "router_probe_interval",
4263                         .data           =       &ipv6_devconf.rtr_probe_interval,
4264                         .maxlen         =       sizeof(int),
4265                         .mode           =       0644,
4266                         .proc_handler   =       proc_dointvec_jiffies,
4267                         .strategy       =       sysctl_jiffies,
4268                 },
4269 #ifdef CONFIG_IPV6_ROUTE_INFO
4270                 {
4271                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4272                         .procname       =       "accept_ra_rt_info_max_plen",
4273                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4274                         .maxlen         =       sizeof(int),
4275                         .mode           =       0644,
4276                         .proc_handler   =       proc_dointvec,
4277                 },
4278 #endif
4279 #endif
4280                 {
4281                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4282                         .procname       =       "proxy_ndp",
4283                         .data           =       &ipv6_devconf.proxy_ndp,
4284                         .maxlen         =       sizeof(int),
4285                         .mode           =       0644,
4286                         .proc_handler   =       proc_dointvec,
4287                 },
4288                 {
4289                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4290                         .procname       =       "accept_source_route",
4291                         .data           =       &ipv6_devconf.accept_source_route,
4292                         .maxlen         =       sizeof(int),
4293                         .mode           =       0644,
4294                         .proc_handler   =       proc_dointvec,
4295                 },
4296 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4297                 {
4298                         .ctl_name       =       CTL_UNNUMBERED,
4299                         .procname       =       "optimistic_dad",
4300                         .data           =       &ipv6_devconf.optimistic_dad,
4301                         .maxlen         =       sizeof(int),
4302                         .mode           =       0644,
4303                         .proc_handler   =       proc_dointvec,
4304
4305                 },
4306 #endif
4307 #ifdef CONFIG_IPV6_MROUTE
4308                 {
4309                         .ctl_name       =       CTL_UNNUMBERED,
4310                         .procname       =       "mc_forwarding",
4311                         .data           =       &ipv6_devconf.mc_forwarding,
4312                         .maxlen         =       sizeof(int),
4313                         .mode           =       0444,
4314                         .proc_handler   =       proc_dointvec,
4315                 },
4316 #endif
4317                 {
4318                         .ctl_name       =       CTL_UNNUMBERED,
4319                         .procname       =       "disable_ipv6",
4320                         .data           =       &ipv6_devconf.disable_ipv6,
4321                         .maxlen         =       sizeof(int),
4322                         .mode           =       0644,
4323                         .proc_handler   =       addrconf_sysctl_disable,
4324                         .strategy       =       sysctl_intvec,
4325                 },
4326                 {
4327                         .ctl_name       =       CTL_UNNUMBERED,
4328                         .procname       =       "accept_dad",
4329                         .data           =       &ipv6_devconf.accept_dad,
4330                         .maxlen         =       sizeof(int),
4331                         .mode           =       0644,
4332                         .proc_handler   =       proc_dointvec,
4333                 },
4334                 {
4335                         .ctl_name       =       0,      /* sentinel */
4336                 }
4337         },
4338 };
4339
4340 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4341                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4342 {
4343         int i;
4344         struct addrconf_sysctl_table *t;
4345
4346 #define ADDRCONF_CTL_PATH_DEV   3
4347
4348         struct ctl_path addrconf_ctl_path[] = {
4349                 { .procname = "net", .ctl_name = CTL_NET, },
4350                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4351                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4352                 { /* to be set */ },
4353                 { },
4354         };
4355
4356
4357         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4358         if (t == NULL)
4359                 goto out;
4360
4361         for (i=0; t->addrconf_vars[i].data; i++) {
4362                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4363                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4364                 t->addrconf_vars[i].extra2 = net;
4365         }
4366
4367         /*
4368          * Make a copy of dev_name, because '.procname' is regarded as const
4369          * by sysctl and we wouldn't want anyone to change it under our feet
4370          * (see SIOCSIFNAME).
4371          */
4372         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4373         if (!t->dev_name)
4374                 goto free;
4375
4376         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4377         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4378
4379         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4380                         t->addrconf_vars);
4381         if (t->sysctl_header == NULL)
4382                 goto free_procname;
4383
4384         p->sysctl = t;
4385         return 0;
4386
4387 free_procname:
4388         kfree(t->dev_name);
4389 free:
4390         kfree(t);
4391 out:
4392         return -ENOBUFS;
4393 }
4394
4395 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4396 {
4397         struct addrconf_sysctl_table *t;
4398
4399         if (p->sysctl == NULL)
4400                 return;
4401
4402         t = p->sysctl;
4403         p->sysctl = NULL;
4404         unregister_sysctl_table(t->sysctl_header);
4405         kfree(t->dev_name);
4406         kfree(t);
4407 }
4408
4409 static void addrconf_sysctl_register(struct inet6_dev *idev)
4410 {
4411         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4412                               NET_IPV6_NEIGH, "ipv6",
4413                               &ndisc_ifinfo_sysctl_change,
4414                               ndisc_ifinfo_sysctl_strategy);
4415         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4416                         idev->dev->ifindex, idev, &idev->cnf);
4417 }
4418
4419 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4420 {
4421         __addrconf_sysctl_unregister(&idev->cnf);
4422         neigh_sysctl_unregister(idev->nd_parms);
4423 }
4424
4425
4426 #endif
4427
4428 static int addrconf_init_net(struct net *net)
4429 {
4430         int err;
4431         struct ipv6_devconf *all, *dflt;
4432
4433         err = -ENOMEM;
4434         all = &ipv6_devconf;
4435         dflt = &ipv6_devconf_dflt;
4436
4437         if (net != &init_net) {
4438                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4439                 if (all == NULL)
4440                         goto err_alloc_all;
4441
4442                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4443                 if (dflt == NULL)
4444                         goto err_alloc_dflt;
4445         } else {
4446                 /* these will be inherited by all namespaces */
4447                 dflt->autoconf = ipv6_defaults.autoconf;
4448                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4449         }
4450
4451         net->ipv6.devconf_all = all;
4452         net->ipv6.devconf_dflt = dflt;
4453
4454 #ifdef CONFIG_SYSCTL
4455         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4456                         NULL, all);
4457         if (err < 0)
4458                 goto err_reg_all;
4459
4460         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4461                         NULL, dflt);
4462         if (err < 0)
4463                 goto err_reg_dflt;
4464 #endif
4465         return 0;
4466
4467 #ifdef CONFIG_SYSCTL
4468 err_reg_dflt:
4469         __addrconf_sysctl_unregister(all);
4470 err_reg_all:
4471         kfree(dflt);
4472 #endif
4473 err_alloc_dflt:
4474         kfree(all);
4475 err_alloc_all:
4476         return err;
4477 }
4478
4479 static void addrconf_exit_net(struct net *net)
4480 {
4481 #ifdef CONFIG_SYSCTL
4482         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4483         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4484 #endif
4485         if (net != &init_net) {
4486                 kfree(net->ipv6.devconf_dflt);
4487                 kfree(net->ipv6.devconf_all);
4488         }
4489 }
4490
4491 static struct pernet_operations addrconf_ops = {
4492         .init = addrconf_init_net,
4493         .exit = addrconf_exit_net,
4494 };
4495
4496 /*
4497  *      Device notifier
4498  */
4499
4500 int register_inet6addr_notifier(struct notifier_block *nb)
4501 {
4502         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4503 }
4504
4505 EXPORT_SYMBOL(register_inet6addr_notifier);
4506
4507 int unregister_inet6addr_notifier(struct notifier_block *nb)
4508 {
4509         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4510 }
4511
4512 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4513
4514 /*
4515  *      Init / cleanup code
4516  */
4517
4518 int __init addrconf_init(void)
4519 {
4520         int err;
4521
4522         if ((err = ipv6_addr_label_init()) < 0) {
4523                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4524                         err);
4525                 return err;
4526         }
4527
4528         register_pernet_subsys(&addrconf_ops);
4529
4530         /* The addrconf netdev notifier requires that loopback_dev
4531          * has it's ipv6 private information allocated and setup
4532          * before it can bring up and give link-local addresses
4533          * to other devices which are up.
4534          *
4535          * Unfortunately, loopback_dev is not necessarily the first
4536          * entry in the global dev_base list of net devices.  In fact,
4537          * it is likely to be the very last entry on that list.
4538          * So this causes the notifier registry below to try and
4539          * give link-local addresses to all devices besides loopback_dev
4540          * first, then loopback_dev, which cases all the non-loopback_dev
4541          * devices to fail to get a link-local address.
4542          *
4543          * So, as a temporary fix, allocate the ipv6 structure for
4544          * loopback_dev first by hand.
4545          * Longer term, all of the dependencies ipv6 has upon the loopback
4546          * device and it being up should be removed.
4547          */
4548         rtnl_lock();
4549         if (!ipv6_add_dev(init_net.loopback_dev))
4550                 err = -ENOMEM;
4551         rtnl_unlock();
4552         if (err)
4553                 goto errlo;
4554
4555         register_netdevice_notifier(&ipv6_dev_notf);
4556
4557         addrconf_verify(0);
4558
4559         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4560         if (err < 0)
4561                 goto errout;
4562
4563         /* Only the first call to __rtnl_register can fail */
4564         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4565         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4566         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4567         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4568         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4569
4570         ipv6_addr_label_rtnl_register();
4571
4572         return 0;
4573 errout:
4574         unregister_netdevice_notifier(&ipv6_dev_notf);
4575 errlo:
4576         unregister_pernet_subsys(&addrconf_ops);
4577
4578         return err;
4579 }
4580
4581 void addrconf_cleanup(void)
4582 {
4583         struct inet6_ifaddr *ifa;
4584         struct net_device *dev;
4585         int i;
4586
4587         unregister_netdevice_notifier(&ipv6_dev_notf);
4588         unregister_pernet_subsys(&addrconf_ops);
4589
4590         rtnl_lock();
4591
4592         /* clean dev list */
4593         for_each_netdev(&init_net, dev) {
4594                 if (__in6_dev_get(dev) == NULL)
4595                         continue;
4596                 addrconf_ifdown(dev, 1);
4597         }
4598         addrconf_ifdown(init_net.loopback_dev, 2);
4599
4600         /*
4601          *      Check hash table.
4602          */
4603         write_lock_bh(&addrconf_hash_lock);
4604         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4605                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4606                         struct inet6_ifaddr *bifa;
4607
4608                         bifa = ifa;
4609                         ifa = ifa->lst_next;
4610                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4611                         /* Do not free it; something is wrong.
4612                            Now we can investigate it with debugger.
4613                          */
4614                 }
4615         }
4616         write_unlock_bh(&addrconf_hash_lock);
4617
4618         del_timer(&addr_chk_timer);
4619         rtnl_unlock();
4620 }