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, int dad_failed)
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                 if (dad_failed)
1381                         ifp->flags |= IFA_F_DADFAILED;
1382                 spin_unlock_bh(&ifp->lock);
1383                 in6_ifa_put(ifp);
1384 #ifdef CONFIG_IPV6_PRIVACY
1385         } else if (ifp->flags&IFA_F_TEMPORARY) {
1386                 struct inet6_ifaddr *ifpub;
1387                 spin_lock_bh(&ifp->lock);
1388                 ifpub = ifp->ifpub;
1389                 if (ifpub) {
1390                         in6_ifa_hold(ifpub);
1391                         spin_unlock_bh(&ifp->lock);
1392                         ipv6_create_tempaddr(ifpub, ifp);
1393                         in6_ifa_put(ifpub);
1394                 } else {
1395                         spin_unlock_bh(&ifp->lock);
1396                 }
1397                 ipv6_del_addr(ifp);
1398 #endif
1399         } else
1400                 ipv6_del_addr(ifp);
1401 }
1402
1403 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1404 {
1405         struct inet6_dev *idev = ifp->idev;
1406
1407         if (net_ratelimit())
1408                 printk(KERN_INFO "%s: IPv6 duplicate address detected!\n",
1409                         ifp->idev->dev->name);
1410
1411         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1412                 struct in6_addr addr;
1413
1414                 addr.s6_addr32[0] = htonl(0xfe800000);
1415                 addr.s6_addr32[1] = 0;
1416
1417                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1418                     ipv6_addr_equal(&ifp->addr, &addr)) {
1419                         /* DAD failed for link-local based on MAC address */
1420                         idev->cnf.disable_ipv6 = 1;
1421
1422                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1423                                 ifp->idev->dev->name);
1424                 }
1425         }
1426
1427         addrconf_dad_stop(ifp, 1);
1428 }
1429
1430 /* Join to solicited addr multicast group. */
1431
1432 void addrconf_join_solict(struct net_device *dev, struct in6_addr *addr)
1433 {
1434         struct in6_addr maddr;
1435
1436         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1437                 return;
1438
1439         addrconf_addr_solict_mult(addr, &maddr);
1440         ipv6_dev_mc_inc(dev, &maddr);
1441 }
1442
1443 void addrconf_leave_solict(struct inet6_dev *idev, struct in6_addr *addr)
1444 {
1445         struct in6_addr maddr;
1446
1447         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1448                 return;
1449
1450         addrconf_addr_solict_mult(addr, &maddr);
1451         __ipv6_dev_mc_dec(idev, &maddr);
1452 }
1453
1454 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1455 {
1456         struct in6_addr addr;
1457         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1458         if (ipv6_addr_any(&addr))
1459                 return;
1460         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1461 }
1462
1463 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1464 {
1465         struct in6_addr addr;
1466         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1467         if (ipv6_addr_any(&addr))
1468                 return;
1469         __ipv6_dev_ac_dec(ifp->idev, &addr);
1470 }
1471
1472 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1473 {
1474         if (dev->addr_len != ETH_ALEN)
1475                 return -1;
1476         memcpy(eui, dev->dev_addr, 3);
1477         memcpy(eui + 5, dev->dev_addr + 3, 3);
1478
1479         /*
1480          * The zSeries OSA network cards can be shared among various
1481          * OS instances, but the OSA cards have only one MAC address.
1482          * This leads to duplicate address conflicts in conjunction
1483          * with IPv6 if more than one instance uses the same card.
1484          *
1485          * The driver for these cards can deliver a unique 16-bit
1486          * identifier for each instance sharing the same card.  It is
1487          * placed instead of 0xFFFE in the interface identifier.  The
1488          * "u" bit of the interface identifier is not inverted in this
1489          * case.  Hence the resulting interface identifier has local
1490          * scope according to RFC2373.
1491          */
1492         if (dev->dev_id) {
1493                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1494                 eui[4] = dev->dev_id & 0xFF;
1495         } else {
1496                 eui[3] = 0xFF;
1497                 eui[4] = 0xFE;
1498                 eui[0] ^= 2;
1499         }
1500         return 0;
1501 }
1502
1503 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1504 {
1505         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1506         if (dev->addr_len != ARCNET_ALEN)
1507                 return -1;
1508         memset(eui, 0, 7);
1509         eui[7] = *(u8*)dev->dev_addr;
1510         return 0;
1511 }
1512
1513 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1514 {
1515         if (dev->addr_len != INFINIBAND_ALEN)
1516                 return -1;
1517         memcpy(eui, dev->dev_addr + 12, 8);
1518         eui[0] |= 2;
1519         return 0;
1520 }
1521
1522 int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1523 {
1524         if (addr == 0)
1525                 return -1;
1526         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1527                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1528                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1529                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1530                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1531                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1532         eui[1] = 0;
1533         eui[2] = 0x5E;
1534         eui[3] = 0xFE;
1535         memcpy(eui + 4, &addr, 4);
1536         return 0;
1537 }
1538 EXPORT_SYMBOL(__ipv6_isatap_ifid);
1539
1540 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1541 {
1542         if (dev->priv_flags & IFF_ISATAP)
1543                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1544         return -1;
1545 }
1546
1547 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1548 {
1549         switch (dev->type) {
1550         case ARPHRD_ETHER:
1551         case ARPHRD_FDDI:
1552         case ARPHRD_IEEE802_TR:
1553                 return addrconf_ifid_eui48(eui, dev);
1554         case ARPHRD_ARCNET:
1555                 return addrconf_ifid_arcnet(eui, dev);
1556         case ARPHRD_INFINIBAND:
1557                 return addrconf_ifid_infiniband(eui, dev);
1558         case ARPHRD_SIT:
1559                 return addrconf_ifid_sit(eui, dev);
1560         }
1561         return -1;
1562 }
1563
1564 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1565 {
1566         int err = -1;
1567         struct inet6_ifaddr *ifp;
1568
1569         read_lock_bh(&idev->lock);
1570         for (ifp=idev->addr_list; ifp; ifp=ifp->if_next) {
1571                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1572                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1573                         err = 0;
1574                         break;
1575                 }
1576         }
1577         read_unlock_bh(&idev->lock);
1578         return err;
1579 }
1580
1581 #ifdef CONFIG_IPV6_PRIVACY
1582 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1583 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1584 {
1585 regen:
1586         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1587         idev->rndid[0] &= ~0x02;
1588
1589         /*
1590          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1591          * check if generated address is not inappropriate
1592          *
1593          *  - Reserved subnet anycast (RFC 2526)
1594          *      11111101 11....11 1xxxxxxx
1595          *  - ISATAP (RFC4214) 6.1
1596          *      00-00-5E-FE-xx-xx-xx-xx
1597          *  - value 0
1598          *  - XXX: already assigned to an address on the device
1599          */
1600         if (idev->rndid[0] == 0xfd &&
1601             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1602             (idev->rndid[7]&0x80))
1603                 goto regen;
1604         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1605                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1606                         goto regen;
1607                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1608                         goto regen;
1609         }
1610
1611         return 0;
1612 }
1613
1614 static void ipv6_regen_rndid(unsigned long data)
1615 {
1616         struct inet6_dev *idev = (struct inet6_dev *) data;
1617         unsigned long expires;
1618
1619         rcu_read_lock_bh();
1620         write_lock_bh(&idev->lock);
1621
1622         if (idev->dead)
1623                 goto out;
1624
1625         if (__ipv6_regen_rndid(idev) < 0)
1626                 goto out;
1627
1628         expires = jiffies +
1629                 idev->cnf.temp_prefered_lft * HZ -
1630                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time - desync_factor;
1631         if (time_before(expires, jiffies)) {
1632                 printk(KERN_WARNING
1633                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1634                         idev->dev->name);
1635                 goto out;
1636         }
1637
1638         if (!mod_timer(&idev->regen_timer, expires))
1639                 in6_dev_hold(idev);
1640
1641 out:
1642         write_unlock_bh(&idev->lock);
1643         rcu_read_unlock_bh();
1644         in6_dev_put(idev);
1645 }
1646
1647 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1648         int ret = 0;
1649
1650         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1651                 ret = __ipv6_regen_rndid(idev);
1652         return ret;
1653 }
1654 #endif
1655
1656 /*
1657  *      Add prefix route.
1658  */
1659
1660 static void
1661 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1662                       unsigned long expires, u32 flags)
1663 {
1664         struct fib6_config cfg = {
1665                 .fc_table = RT6_TABLE_PREFIX,
1666                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1667                 .fc_ifindex = dev->ifindex,
1668                 .fc_expires = expires,
1669                 .fc_dst_len = plen,
1670                 .fc_flags = RTF_UP | flags,
1671                 .fc_nlinfo.nl_net = dev_net(dev),
1672                 .fc_protocol = RTPROT_KERNEL,
1673         };
1674
1675         ipv6_addr_copy(&cfg.fc_dst, pfx);
1676
1677         /* Prevent useless cloning on PtP SIT.
1678            This thing is done here expecting that the whole
1679            class of non-broadcast devices need not cloning.
1680          */
1681 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1682         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1683                 cfg.fc_flags |= RTF_NONEXTHOP;
1684 #endif
1685
1686         ip6_route_add(&cfg);
1687 }
1688
1689 /* Create "default" multicast route to the interface */
1690
1691 static void addrconf_add_mroute(struct net_device *dev)
1692 {
1693         struct fib6_config cfg = {
1694                 .fc_table = RT6_TABLE_LOCAL,
1695                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1696                 .fc_ifindex = dev->ifindex,
1697                 .fc_dst_len = 8,
1698                 .fc_flags = RTF_UP,
1699                 .fc_nlinfo.nl_net = dev_net(dev),
1700         };
1701
1702         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1703
1704         ip6_route_add(&cfg);
1705 }
1706
1707 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1708 static void sit_route_add(struct net_device *dev)
1709 {
1710         struct fib6_config cfg = {
1711                 .fc_table = RT6_TABLE_MAIN,
1712                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1713                 .fc_ifindex = dev->ifindex,
1714                 .fc_dst_len = 96,
1715                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1716                 .fc_nlinfo.nl_net = dev_net(dev),
1717         };
1718
1719         /* prefix length - 96 bits "::d.d.d.d" */
1720         ip6_route_add(&cfg);
1721 }
1722 #endif
1723
1724 static void addrconf_add_lroute(struct net_device *dev)
1725 {
1726         struct in6_addr addr;
1727
1728         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1729         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1730 }
1731
1732 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1733 {
1734         struct inet6_dev *idev;
1735
1736         ASSERT_RTNL();
1737
1738         if ((idev = ipv6_find_idev(dev)) == NULL)
1739                 return NULL;
1740
1741         /* Add default multicast route */
1742         addrconf_add_mroute(dev);
1743
1744         /* Add link local route */
1745         addrconf_add_lroute(dev);
1746         return idev;
1747 }
1748
1749 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1750 {
1751         struct prefix_info *pinfo;
1752         __u32 valid_lft;
1753         __u32 prefered_lft;
1754         int addr_type;
1755         struct inet6_dev *in6_dev;
1756         struct net *net = dev_net(dev);
1757
1758         pinfo = (struct prefix_info *) opt;
1759
1760         if (len < sizeof(struct prefix_info)) {
1761                 ADBG(("addrconf: prefix option too short\n"));
1762                 return;
1763         }
1764
1765         /*
1766          *      Validation checks ([ADDRCONF], page 19)
1767          */
1768
1769         addr_type = ipv6_addr_type(&pinfo->prefix);
1770
1771         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1772                 return;
1773
1774         valid_lft = ntohl(pinfo->valid);
1775         prefered_lft = ntohl(pinfo->prefered);
1776
1777         if (prefered_lft > valid_lft) {
1778                 if (net_ratelimit())
1779                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1780                 return;
1781         }
1782
1783         in6_dev = in6_dev_get(dev);
1784
1785         if (in6_dev == NULL) {
1786                 if (net_ratelimit())
1787                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1788                 return;
1789         }
1790
1791         /*
1792          *      Two things going on here:
1793          *      1) Add routes for on-link prefixes
1794          *      2) Configure prefixes with the auto flag set
1795          */
1796
1797         if (pinfo->onlink) {
1798                 struct rt6_info *rt;
1799                 unsigned long rt_expires;
1800
1801                 /* Avoid arithmetic overflow. Really, we could
1802                  * save rt_expires in seconds, likely valid_lft,
1803                  * but it would require division in fib gc, that it
1804                  * not good.
1805                  */
1806                 if (HZ > USER_HZ)
1807                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1808                 else
1809                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1810
1811                 if (addrconf_finite_timeout(rt_expires))
1812                         rt_expires *= HZ;
1813
1814                 rt = rt6_lookup(net, &pinfo->prefix, NULL,
1815                                 dev->ifindex, 1);
1816
1817                 if (rt && addrconf_is_prefix_route(rt)) {
1818                         /* Autoconf prefix route */
1819                         if (valid_lft == 0) {
1820                                 ip6_del_rt(rt);
1821                                 rt = NULL;
1822                         } else if (addrconf_finite_timeout(rt_expires)) {
1823                                 /* not infinity */
1824                                 rt->rt6i_expires = jiffies + rt_expires;
1825                                 rt->rt6i_flags |= RTF_EXPIRES;
1826                         } else {
1827                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1828                                 rt->rt6i_expires = 0;
1829                         }
1830                 } else if (valid_lft) {
1831                         clock_t expires = 0;
1832                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1833                         if (addrconf_finite_timeout(rt_expires)) {
1834                                 /* not infinity */
1835                                 flags |= RTF_EXPIRES;
1836                                 expires = jiffies_to_clock_t(rt_expires);
1837                         }
1838                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1839                                               dev, expires, flags);
1840                 }
1841                 if (rt)
1842                         dst_release(&rt->u.dst);
1843         }
1844
1845         /* Try to figure out our local address for this prefix */
1846
1847         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1848                 struct inet6_ifaddr * ifp;
1849                 struct in6_addr addr;
1850                 int create = 0, update_lft = 0;
1851
1852                 if (pinfo->prefix_len == 64) {
1853                         memcpy(&addr, &pinfo->prefix, 8);
1854                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1855                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1856                                 in6_dev_put(in6_dev);
1857                                 return;
1858                         }
1859                         goto ok;
1860                 }
1861                 if (net_ratelimit())
1862                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1863                                pinfo->prefix_len);
1864                 in6_dev_put(in6_dev);
1865                 return;
1866
1867 ok:
1868
1869                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1870
1871                 if (ifp == NULL && valid_lft) {
1872                         int max_addresses = in6_dev->cnf.max_addresses;
1873                         u32 addr_flags = 0;
1874
1875 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1876                         if (in6_dev->cnf.optimistic_dad &&
1877                             !net->ipv6.devconf_all->forwarding)
1878                                 addr_flags = IFA_F_OPTIMISTIC;
1879 #endif
1880
1881                         /* Do not allow to create too much of autoconfigured
1882                          * addresses; this would be too easy way to crash kernel.
1883                          */
1884                         if (!max_addresses ||
1885                             ipv6_count_addresses(in6_dev) < max_addresses)
1886                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1887                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1888                                                     addr_flags);
1889
1890                         if (!ifp || IS_ERR(ifp)) {
1891                                 in6_dev_put(in6_dev);
1892                                 return;
1893                         }
1894
1895                         update_lft = create = 1;
1896                         ifp->cstamp = jiffies;
1897                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1898                 }
1899
1900                 if (ifp) {
1901                         int flags;
1902                         unsigned long now;
1903 #ifdef CONFIG_IPV6_PRIVACY
1904                         struct inet6_ifaddr *ift;
1905 #endif
1906                         u32 stored_lft;
1907
1908                         /* update lifetime (RFC2462 5.5.3 e) */
1909                         spin_lock(&ifp->lock);
1910                         now = jiffies;
1911                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1912                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1913                         else
1914                                 stored_lft = 0;
1915                         if (!update_lft && stored_lft) {
1916                                 if (valid_lft > MIN_VALID_LIFETIME ||
1917                                     valid_lft > stored_lft)
1918                                         update_lft = 1;
1919                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
1920                                         /* valid_lft <= stored_lft is always true */
1921                                         /*
1922                                          * RFC 4862 Section 5.5.3e:
1923                                          * "Note that the preferred lifetime of
1924                                          *  the corresponding address is always
1925                                          *  reset to the Preferred Lifetime in
1926                                          *  the received Prefix Information
1927                                          *  option, regardless of whether the
1928                                          *  valid lifetime is also reset or
1929                                          *  ignored."
1930                                          *
1931                                          *  So if the preferred lifetime in
1932                                          *  this advertisement is different
1933                                          *  than what we have stored, but the
1934                                          *  valid lifetime is invalid, just
1935                                          *  reset prefered_lft.
1936                                          *
1937                                          *  We must set the valid lifetime
1938                                          *  to the stored lifetime since we'll
1939                                          *  be updating the timestamp below,
1940                                          *  else we'll set it back to the
1941                                          *  minumum.
1942                                          */
1943                                         if (prefered_lft != ifp->prefered_lft) {
1944                                                 valid_lft = stored_lft;
1945                                                 update_lft = 1;
1946                                         }
1947                                 } else {
1948                                         valid_lft = MIN_VALID_LIFETIME;
1949                                         if (valid_lft < prefered_lft)
1950                                                 prefered_lft = valid_lft;
1951                                         update_lft = 1;
1952                                 }
1953                         }
1954
1955                         if (update_lft) {
1956                                 ifp->valid_lft = valid_lft;
1957                                 ifp->prefered_lft = prefered_lft;
1958                                 ifp->tstamp = now;
1959                                 flags = ifp->flags;
1960                                 ifp->flags &= ~IFA_F_DEPRECATED;
1961                                 spin_unlock(&ifp->lock);
1962
1963                                 if (!(flags&IFA_F_TENTATIVE))
1964                                         ipv6_ifa_notify(0, ifp);
1965                         } else
1966                                 spin_unlock(&ifp->lock);
1967
1968 #ifdef CONFIG_IPV6_PRIVACY
1969                         read_lock_bh(&in6_dev->lock);
1970                         /* update all temporary addresses in the list */
1971                         for (ift=in6_dev->tempaddr_list; ift; ift=ift->tmp_next) {
1972                                 /*
1973                                  * When adjusting the lifetimes of an existing
1974                                  * temporary address, only lower the lifetimes.
1975                                  * Implementations must not increase the
1976                                  * lifetimes of an existing temporary address
1977                                  * when processing a Prefix Information Option.
1978                                  */
1979                                 if (ifp != ift->ifpub)
1980                                         continue;
1981
1982                                 spin_lock(&ift->lock);
1983                                 flags = ift->flags;
1984                                 if (ift->valid_lft > valid_lft &&
1985                                     ift->valid_lft - valid_lft > (jiffies - ift->tstamp) / HZ)
1986                                         ift->valid_lft = valid_lft + (jiffies - ift->tstamp) / HZ;
1987                                 if (ift->prefered_lft > prefered_lft &&
1988                                     ift->prefered_lft - prefered_lft > (jiffies - ift->tstamp) / HZ)
1989                                         ift->prefered_lft = prefered_lft + (jiffies - ift->tstamp) / HZ;
1990                                 spin_unlock(&ift->lock);
1991                                 if (!(flags&IFA_F_TENTATIVE))
1992                                         ipv6_ifa_notify(0, ift);
1993                         }
1994
1995                         if (create && in6_dev->cnf.use_tempaddr > 0) {
1996                                 /*
1997                                  * When a new public address is created as described in [ADDRCONF],
1998                                  * also create a new temporary address.
1999                                  */
2000                                 read_unlock_bh(&in6_dev->lock);
2001                                 ipv6_create_tempaddr(ifp, NULL);
2002                         } else {
2003                                 read_unlock_bh(&in6_dev->lock);
2004                         }
2005 #endif
2006                         in6_ifa_put(ifp);
2007                         addrconf_verify(0);
2008                 }
2009         }
2010         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2011         in6_dev_put(in6_dev);
2012 }
2013
2014 /*
2015  *      Set destination address.
2016  *      Special case for SIT interfaces where we create a new "virtual"
2017  *      device.
2018  */
2019 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2020 {
2021         struct in6_ifreq ireq;
2022         struct net_device *dev;
2023         int err = -EINVAL;
2024
2025         rtnl_lock();
2026
2027         err = -EFAULT;
2028         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2029                 goto err_exit;
2030
2031         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2032
2033         err = -ENODEV;
2034         if (dev == NULL)
2035                 goto err_exit;
2036
2037 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2038         if (dev->type == ARPHRD_SIT) {
2039                 const struct net_device_ops *ops = dev->netdev_ops;
2040                 struct ifreq ifr;
2041                 struct ip_tunnel_parm p;
2042
2043                 err = -EADDRNOTAVAIL;
2044                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2045                         goto err_exit;
2046
2047                 memset(&p, 0, sizeof(p));
2048                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2049                 p.iph.saddr = 0;
2050                 p.iph.version = 4;
2051                 p.iph.ihl = 5;
2052                 p.iph.protocol = IPPROTO_IPV6;
2053                 p.iph.ttl = 64;
2054                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2055
2056                 if (ops->ndo_do_ioctl) {
2057                         mm_segment_t oldfs = get_fs();
2058
2059                         set_fs(KERNEL_DS);
2060                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2061                         set_fs(oldfs);
2062                 } else
2063                         err = -EOPNOTSUPP;
2064
2065                 if (err == 0) {
2066                         err = -ENOBUFS;
2067                         dev = __dev_get_by_name(net, p.name);
2068                         if (!dev)
2069                                 goto err_exit;
2070                         err = dev_open(dev);
2071                 }
2072         }
2073 #endif
2074
2075 err_exit:
2076         rtnl_unlock();
2077         return err;
2078 }
2079
2080 /*
2081  *      Manual configuration of address on an interface
2082  */
2083 static int inet6_addr_add(struct net *net, int ifindex, struct in6_addr *pfx,
2084                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2085                           __u32 valid_lft)
2086 {
2087         struct inet6_ifaddr *ifp;
2088         struct inet6_dev *idev;
2089         struct net_device *dev;
2090         int scope;
2091         u32 flags;
2092         clock_t expires;
2093         unsigned long timeout;
2094
2095         ASSERT_RTNL();
2096
2097         if (plen > 128)
2098                 return -EINVAL;
2099
2100         /* check the lifetime */
2101         if (!valid_lft || prefered_lft > valid_lft)
2102                 return -EINVAL;
2103
2104         dev = __dev_get_by_index(net, ifindex);
2105         if (!dev)
2106                 return -ENODEV;
2107
2108         if ((idev = addrconf_add_dev(dev)) == NULL)
2109                 return -ENOBUFS;
2110
2111         scope = ipv6_addr_scope(pfx);
2112
2113         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2114         if (addrconf_finite_timeout(timeout)) {
2115                 expires = jiffies_to_clock_t(timeout * HZ);
2116                 valid_lft = timeout;
2117                 flags = RTF_EXPIRES;
2118         } else {
2119                 expires = 0;
2120                 flags = 0;
2121                 ifa_flags |= IFA_F_PERMANENT;
2122         }
2123
2124         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2125         if (addrconf_finite_timeout(timeout)) {
2126                 if (timeout == 0)
2127                         ifa_flags |= IFA_F_DEPRECATED;
2128                 prefered_lft = timeout;
2129         }
2130
2131         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2132
2133         if (!IS_ERR(ifp)) {
2134                 spin_lock_bh(&ifp->lock);
2135                 ifp->valid_lft = valid_lft;
2136                 ifp->prefered_lft = prefered_lft;
2137                 ifp->tstamp = jiffies;
2138                 spin_unlock_bh(&ifp->lock);
2139
2140                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2141                                       expires, flags);
2142                 /*
2143                  * Note that section 3.1 of RFC 4429 indicates
2144                  * that the Optimistic flag should not be set for
2145                  * manually configured addresses
2146                  */
2147                 addrconf_dad_start(ifp, 0);
2148                 in6_ifa_put(ifp);
2149                 addrconf_verify(0);
2150                 return 0;
2151         }
2152
2153         return PTR_ERR(ifp);
2154 }
2155
2156 static int inet6_addr_del(struct net *net, int ifindex, struct in6_addr *pfx,
2157                           unsigned int plen)
2158 {
2159         struct inet6_ifaddr *ifp;
2160         struct inet6_dev *idev;
2161         struct net_device *dev;
2162
2163         if (plen > 128)
2164                 return -EINVAL;
2165
2166         dev = __dev_get_by_index(net, ifindex);
2167         if (!dev)
2168                 return -ENODEV;
2169
2170         if ((idev = __in6_dev_get(dev)) == NULL)
2171                 return -ENXIO;
2172
2173         read_lock_bh(&idev->lock);
2174         for (ifp = idev->addr_list; ifp; ifp=ifp->if_next) {
2175                 if (ifp->prefix_len == plen &&
2176                     ipv6_addr_equal(pfx, &ifp->addr)) {
2177                         in6_ifa_hold(ifp);
2178                         read_unlock_bh(&idev->lock);
2179
2180                         ipv6_del_addr(ifp);
2181
2182                         /* If the last address is deleted administratively,
2183                            disable IPv6 on this interface.
2184                          */
2185                         if (idev->addr_list == NULL)
2186                                 addrconf_ifdown(idev->dev, 1);
2187                         return 0;
2188                 }
2189         }
2190         read_unlock_bh(&idev->lock);
2191         return -EADDRNOTAVAIL;
2192 }
2193
2194
2195 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2196 {
2197         struct in6_ifreq ireq;
2198         int err;
2199
2200         if (!capable(CAP_NET_ADMIN))
2201                 return -EPERM;
2202
2203         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2204                 return -EFAULT;
2205
2206         rtnl_lock();
2207         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2208                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2209                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2210         rtnl_unlock();
2211         return err;
2212 }
2213
2214 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2215 {
2216         struct in6_ifreq ireq;
2217         int err;
2218
2219         if (!capable(CAP_NET_ADMIN))
2220                 return -EPERM;
2221
2222         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2223                 return -EFAULT;
2224
2225         rtnl_lock();
2226         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2227                              ireq.ifr6_prefixlen);
2228         rtnl_unlock();
2229         return err;
2230 }
2231
2232 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2233                      int plen, int scope)
2234 {
2235         struct inet6_ifaddr *ifp;
2236
2237         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2238         if (!IS_ERR(ifp)) {
2239                 spin_lock_bh(&ifp->lock);
2240                 ifp->flags &= ~IFA_F_TENTATIVE;
2241                 spin_unlock_bh(&ifp->lock);
2242                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2243                 in6_ifa_put(ifp);
2244         }
2245 }
2246
2247 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2248 static void sit_add_v4_addrs(struct inet6_dev *idev)
2249 {
2250         struct in6_addr addr;
2251         struct net_device *dev;
2252         struct net *net = dev_net(idev->dev);
2253         int scope;
2254
2255         ASSERT_RTNL();
2256
2257         memset(&addr, 0, sizeof(struct in6_addr));
2258         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2259
2260         if (idev->dev->flags&IFF_POINTOPOINT) {
2261                 addr.s6_addr32[0] = htonl(0xfe800000);
2262                 scope = IFA_LINK;
2263         } else {
2264                 scope = IPV6_ADDR_COMPATv4;
2265         }
2266
2267         if (addr.s6_addr32[3]) {
2268                 add_addr(idev, &addr, 128, scope);
2269                 return;
2270         }
2271
2272         for_each_netdev(net, dev) {
2273                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2274                 if (in_dev && (dev->flags & IFF_UP)) {
2275                         struct in_ifaddr * ifa;
2276
2277                         int flag = scope;
2278
2279                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2280                                 int plen;
2281
2282                                 addr.s6_addr32[3] = ifa->ifa_local;
2283
2284                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2285                                         continue;
2286                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2287                                         if (idev->dev->flags&IFF_POINTOPOINT)
2288                                                 continue;
2289                                         flag |= IFA_HOST;
2290                                 }
2291                                 if (idev->dev->flags&IFF_POINTOPOINT)
2292                                         plen = 64;
2293                                 else
2294                                         plen = 96;
2295
2296                                 add_addr(idev, &addr, plen, flag);
2297                         }
2298                 }
2299         }
2300 }
2301 #endif
2302
2303 static void init_loopback(struct net_device *dev)
2304 {
2305         struct inet6_dev  *idev;
2306
2307         /* ::1 */
2308
2309         ASSERT_RTNL();
2310
2311         if ((idev = ipv6_find_idev(dev)) == NULL) {
2312                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2313                 return;
2314         }
2315
2316         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2317 }
2318
2319 static void addrconf_add_linklocal(struct inet6_dev *idev, struct in6_addr *addr)
2320 {
2321         struct inet6_ifaddr * ifp;
2322         u32 addr_flags = IFA_F_PERMANENT;
2323
2324 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2325         if (idev->cnf.optimistic_dad &&
2326             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2327                 addr_flags |= IFA_F_OPTIMISTIC;
2328 #endif
2329
2330
2331         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2332         if (!IS_ERR(ifp)) {
2333                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2334                 addrconf_dad_start(ifp, 0);
2335                 in6_ifa_put(ifp);
2336         }
2337 }
2338
2339 static void addrconf_dev_config(struct net_device *dev)
2340 {
2341         struct in6_addr addr;
2342         struct inet6_dev    * idev;
2343
2344         ASSERT_RTNL();
2345
2346         if ((dev->type != ARPHRD_ETHER) &&
2347             (dev->type != ARPHRD_FDDI) &&
2348             (dev->type != ARPHRD_IEEE802_TR) &&
2349             (dev->type != ARPHRD_ARCNET) &&
2350             (dev->type != ARPHRD_INFINIBAND)) {
2351                 /* Alas, we support only Ethernet autoconfiguration. */
2352                 return;
2353         }
2354
2355         idev = addrconf_add_dev(dev);
2356         if (idev == NULL)
2357                 return;
2358
2359         memset(&addr, 0, sizeof(struct in6_addr));
2360         addr.s6_addr32[0] = htonl(0xFE800000);
2361
2362         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2363                 addrconf_add_linklocal(idev, &addr);
2364 }
2365
2366 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2367 static void addrconf_sit_config(struct net_device *dev)
2368 {
2369         struct inet6_dev *idev;
2370
2371         ASSERT_RTNL();
2372
2373         /*
2374          * Configure the tunnel with one of our IPv4
2375          * addresses... we should configure all of
2376          * our v4 addrs in the tunnel
2377          */
2378
2379         if ((idev = ipv6_find_idev(dev)) == NULL) {
2380                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2381                 return;
2382         }
2383
2384         if (dev->priv_flags & IFF_ISATAP) {
2385                 struct in6_addr addr;
2386
2387                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2388                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2389                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2390                         addrconf_add_linklocal(idev, &addr);
2391                 return;
2392         }
2393
2394         sit_add_v4_addrs(idev);
2395
2396         if (dev->flags&IFF_POINTOPOINT) {
2397                 addrconf_add_mroute(dev);
2398                 addrconf_add_lroute(dev);
2399         } else
2400                 sit_route_add(dev);
2401 }
2402 #endif
2403
2404 static inline int
2405 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2406 {
2407         struct in6_addr lladdr;
2408
2409         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2410                 addrconf_add_linklocal(idev, &lladdr);
2411                 return 0;
2412         }
2413         return -1;
2414 }
2415
2416 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2417 {
2418         struct net_device *link_dev;
2419         struct net *net = dev_net(idev->dev);
2420
2421         /* first try to inherit the link-local address from the link device */
2422         if (idev->dev->iflink &&
2423             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2424                 if (!ipv6_inherit_linklocal(idev, link_dev))
2425                         return;
2426         }
2427         /* then try to inherit it from any device */
2428         for_each_netdev(net, link_dev) {
2429                 if (!ipv6_inherit_linklocal(idev, link_dev))
2430                         return;
2431         }
2432         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2433 }
2434
2435 /*
2436  * Autoconfigure tunnel with a link-local address so routing protocols,
2437  * DHCPv6, MLD etc. can be run over the virtual link
2438  */
2439
2440 static void addrconf_ip6_tnl_config(struct net_device *dev)
2441 {
2442         struct inet6_dev *idev;
2443
2444         ASSERT_RTNL();
2445
2446         if ((idev = addrconf_add_dev(dev)) == NULL) {
2447                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2448                 return;
2449         }
2450         ip6_tnl_add_linklocal(idev);
2451 }
2452
2453 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2454                            void * data)
2455 {
2456         struct net_device *dev = (struct net_device *) data;
2457         struct inet6_dev *idev = __in6_dev_get(dev);
2458         int run_pending = 0;
2459         int err;
2460
2461         switch(event) {
2462         case NETDEV_REGISTER:
2463                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2464                         idev = ipv6_add_dev(dev);
2465                         if (!idev)
2466                                 return notifier_from_errno(-ENOMEM);
2467                 }
2468                 break;
2469         case NETDEV_UP:
2470         case NETDEV_CHANGE:
2471                 if (dev->flags & IFF_SLAVE)
2472                         break;
2473
2474                 if (event == NETDEV_UP) {
2475                         if (!addrconf_qdisc_ok(dev)) {
2476                                 /* device is not ready yet. */
2477                                 printk(KERN_INFO
2478                                         "ADDRCONF(NETDEV_UP): %s: "
2479                                         "link is not ready\n",
2480                                         dev->name);
2481                                 break;
2482                         }
2483
2484                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2485                                 idev = ipv6_add_dev(dev);
2486
2487                         if (idev) {
2488                                 idev->if_flags |= IF_READY;
2489                                 run_pending = 1;
2490                         }
2491                 } else {
2492                         if (!addrconf_qdisc_ok(dev)) {
2493                                 /* device is still not ready. */
2494                                 break;
2495                         }
2496
2497                         if (idev) {
2498                                 if (idev->if_flags & IF_READY) {
2499                                         /* device is already configured. */
2500                                         break;
2501                                 }
2502                                 idev->if_flags |= IF_READY;
2503                         }
2504
2505                         printk(KERN_INFO
2506                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2507                                         "link becomes ready\n",
2508                                         dev->name);
2509
2510                         run_pending = 1;
2511                 }
2512
2513                 switch(dev->type) {
2514 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2515                 case ARPHRD_SIT:
2516                         addrconf_sit_config(dev);
2517                         break;
2518 #endif
2519                 case ARPHRD_TUNNEL6:
2520                         addrconf_ip6_tnl_config(dev);
2521                         break;
2522                 case ARPHRD_LOOPBACK:
2523                         init_loopback(dev);
2524                         break;
2525
2526                 default:
2527                         addrconf_dev_config(dev);
2528                         break;
2529                 }
2530                 if (idev) {
2531                         if (run_pending)
2532                                 addrconf_dad_run(idev);
2533
2534                         /* If the MTU changed during the interface down, when the
2535                            interface up, the changed MTU must be reflected in the
2536                            idev as well as routers.
2537                          */
2538                         if (idev->cnf.mtu6 != dev->mtu && dev->mtu >= IPV6_MIN_MTU) {
2539                                 rt6_mtu_change(dev, dev->mtu);
2540                                 idev->cnf.mtu6 = dev->mtu;
2541                         }
2542                         idev->tstamp = jiffies;
2543                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2544                         /* If the changed mtu during down is lower than IPV6_MIN_MTU
2545                            stop IPv6 on this interface.
2546                          */
2547                         if (dev->mtu < IPV6_MIN_MTU)
2548                                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2549                 }
2550                 break;
2551
2552         case NETDEV_CHANGEMTU:
2553                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2554                         rt6_mtu_change(dev, dev->mtu);
2555                         idev->cnf.mtu6 = dev->mtu;
2556                         break;
2557                 }
2558
2559                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2560                         idev = ipv6_add_dev(dev);
2561                         if (idev)
2562                                 break;
2563                 }
2564
2565                 /* MTU falled under IPV6_MIN_MTU. Stop IPv6 on this interface. */
2566
2567         case NETDEV_DOWN:
2568         case NETDEV_UNREGISTER:
2569                 /*
2570                  *      Remove all addresses from this interface.
2571                  */
2572                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2573                 break;
2574
2575         case NETDEV_CHANGENAME:
2576                 if (idev) {
2577                         snmp6_unregister_dev(idev);
2578                         addrconf_sysctl_unregister(idev);
2579                         addrconf_sysctl_register(idev);
2580                         err = snmp6_register_dev(idev);
2581                         if (err)
2582                                 return notifier_from_errno(err);
2583                 }
2584                 break;
2585         }
2586
2587         return NOTIFY_OK;
2588 }
2589
2590 /*
2591  *      addrconf module should be notified of a device going up
2592  */
2593 static struct notifier_block ipv6_dev_notf = {
2594         .notifier_call = addrconf_notify,
2595         .priority = 0
2596 };
2597
2598 static int addrconf_ifdown(struct net_device *dev, int how)
2599 {
2600         struct inet6_dev *idev;
2601         struct inet6_ifaddr *ifa, **bifa;
2602         struct net *net = dev_net(dev);
2603         int i;
2604
2605         ASSERT_RTNL();
2606
2607         rt6_ifdown(net, dev);
2608         neigh_ifdown(&nd_tbl, dev);
2609
2610         idev = __in6_dev_get(dev);
2611         if (idev == NULL)
2612                 return -ENODEV;
2613
2614         /* Step 1: remove reference to ipv6 device from parent device.
2615                    Do not dev_put!
2616          */
2617         if (how) {
2618                 idev->dead = 1;
2619
2620                 /* protected by rtnl_lock */
2621                 rcu_assign_pointer(dev->ip6_ptr, NULL);
2622
2623                 /* Step 1.5: remove snmp6 entry */
2624                 snmp6_unregister_dev(idev);
2625
2626         }
2627
2628         /* Step 2: clear hash table */
2629         for (i=0; i<IN6_ADDR_HSIZE; i++) {
2630                 bifa = &inet6_addr_lst[i];
2631
2632                 write_lock_bh(&addrconf_hash_lock);
2633                 while ((ifa = *bifa) != NULL) {
2634                         if (ifa->idev == idev) {
2635                                 *bifa = ifa->lst_next;
2636                                 ifa->lst_next = NULL;
2637                                 addrconf_del_timer(ifa);
2638                                 in6_ifa_put(ifa);
2639                                 continue;
2640                         }
2641                         bifa = &ifa->lst_next;
2642                 }
2643                 write_unlock_bh(&addrconf_hash_lock);
2644         }
2645
2646         write_lock_bh(&idev->lock);
2647
2648         /* Step 3: clear flags for stateless addrconf */
2649         if (!how)
2650                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2651
2652         /* Step 4: clear address list */
2653 #ifdef CONFIG_IPV6_PRIVACY
2654         if (how && del_timer(&idev->regen_timer))
2655                 in6_dev_put(idev);
2656
2657         /* clear tempaddr list */
2658         while ((ifa = idev->tempaddr_list) != NULL) {
2659                 idev->tempaddr_list = ifa->tmp_next;
2660                 ifa->tmp_next = NULL;
2661                 ifa->dead = 1;
2662                 write_unlock_bh(&idev->lock);
2663                 spin_lock_bh(&ifa->lock);
2664
2665                 if (ifa->ifpub) {
2666                         in6_ifa_put(ifa->ifpub);
2667                         ifa->ifpub = NULL;
2668                 }
2669                 spin_unlock_bh(&ifa->lock);
2670                 in6_ifa_put(ifa);
2671                 write_lock_bh(&idev->lock);
2672         }
2673 #endif
2674         while ((ifa = idev->addr_list) != NULL) {
2675                 idev->addr_list = ifa->if_next;
2676                 ifa->if_next = NULL;
2677                 ifa->dead = 1;
2678                 addrconf_del_timer(ifa);
2679                 write_unlock_bh(&idev->lock);
2680
2681                 __ipv6_ifa_notify(RTM_DELADDR, ifa);
2682                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2683                 in6_ifa_put(ifa);
2684
2685                 write_lock_bh(&idev->lock);
2686         }
2687         write_unlock_bh(&idev->lock);
2688
2689         /* Step 5: Discard multicast list */
2690
2691         if (how)
2692                 ipv6_mc_destroy_dev(idev);
2693         else
2694                 ipv6_mc_down(idev);
2695
2696         idev->tstamp = jiffies;
2697
2698         /* Shot the device (if unregistered) */
2699
2700         if (how) {
2701                 addrconf_sysctl_unregister(idev);
2702                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2703                 neigh_ifdown(&nd_tbl, dev);
2704                 in6_dev_put(idev);
2705         }
2706         return 0;
2707 }
2708
2709 static void addrconf_rs_timer(unsigned long data)
2710 {
2711         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2712
2713         if (ifp->idev->cnf.forwarding)
2714                 goto out;
2715
2716         if (ifp->idev->if_flags & IF_RA_RCVD) {
2717                 /*
2718                  *      Announcement received after solicitation
2719                  *      was sent
2720                  */
2721                 goto out;
2722         }
2723
2724         spin_lock(&ifp->lock);
2725         if (ifp->probes++ < ifp->idev->cnf.rtr_solicits) {
2726                 /* The wait after the last probe can be shorter */
2727                 addrconf_mod_timer(ifp, AC_RS,
2728                                    (ifp->probes == ifp->idev->cnf.rtr_solicits) ?
2729                                    ifp->idev->cnf.rtr_solicit_delay :
2730                                    ifp->idev->cnf.rtr_solicit_interval);
2731                 spin_unlock(&ifp->lock);
2732
2733                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2734         } else {
2735                 spin_unlock(&ifp->lock);
2736                 /*
2737                  * Note: we do not support deprecated "all on-link"
2738                  * assumption any longer.
2739                  */
2740                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2741                        ifp->idev->dev->name);
2742         }
2743
2744 out:
2745         in6_ifa_put(ifp);
2746 }
2747
2748 /*
2749  *      Duplicate Address Detection
2750  */
2751 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2752 {
2753         unsigned long rand_num;
2754         struct inet6_dev *idev = ifp->idev;
2755
2756         if (ifp->flags & IFA_F_OPTIMISTIC)
2757                 rand_num = 0;
2758         else
2759                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2760
2761         ifp->probes = idev->cnf.dad_transmits;
2762         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2763 }
2764
2765 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2766 {
2767         struct inet6_dev *idev = ifp->idev;
2768         struct net_device *dev = idev->dev;
2769
2770         addrconf_join_solict(dev, &ifp->addr);
2771
2772         net_srandom(ifp->addr.s6_addr32[3]);
2773
2774         read_lock_bh(&idev->lock);
2775         if (ifp->dead)
2776                 goto out;
2777         spin_lock_bh(&ifp->lock);
2778
2779         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2780             idev->cnf.accept_dad < 1 ||
2781             !(ifp->flags&IFA_F_TENTATIVE) ||
2782             ifp->flags & IFA_F_NODAD) {
2783                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2784                 spin_unlock_bh(&ifp->lock);
2785                 read_unlock_bh(&idev->lock);
2786
2787                 addrconf_dad_completed(ifp);
2788                 return;
2789         }
2790
2791         if (!(idev->if_flags & IF_READY)) {
2792                 spin_unlock_bh(&ifp->lock);
2793                 read_unlock_bh(&idev->lock);
2794                 /*
2795                  * If the device is not ready:
2796                  * - keep it tentative if it is a permanent address.
2797                  * - otherwise, kill it.
2798                  */
2799                 in6_ifa_hold(ifp);
2800                 addrconf_dad_stop(ifp, 0);
2801                 return;
2802         }
2803
2804         /*
2805          * Optimistic nodes can start receiving
2806          * Frames right away
2807          */
2808         if(ifp->flags & IFA_F_OPTIMISTIC)
2809                 ip6_ins_rt(ifp->rt);
2810
2811         addrconf_dad_kick(ifp);
2812         spin_unlock_bh(&ifp->lock);
2813 out:
2814         read_unlock_bh(&idev->lock);
2815 }
2816
2817 static void addrconf_dad_timer(unsigned long data)
2818 {
2819         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2820         struct inet6_dev *idev = ifp->idev;
2821         struct in6_addr mcaddr;
2822
2823         read_lock_bh(&idev->lock);
2824         if (idev->dead) {
2825                 read_unlock_bh(&idev->lock);
2826                 goto out;
2827         }
2828         spin_lock_bh(&ifp->lock);
2829         if (ifp->probes == 0) {
2830                 /*
2831                  * DAD was successful
2832                  */
2833
2834                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2835                 spin_unlock_bh(&ifp->lock);
2836                 read_unlock_bh(&idev->lock);
2837
2838                 addrconf_dad_completed(ifp);
2839
2840                 goto out;
2841         }
2842
2843         ifp->probes--;
2844         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
2845         spin_unlock_bh(&ifp->lock);
2846         read_unlock_bh(&idev->lock);
2847
2848         /* send a neighbour solicitation for our addr */
2849         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
2850         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
2851 out:
2852         in6_ifa_put(ifp);
2853 }
2854
2855 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
2856 {
2857         struct net_device *     dev = ifp->idev->dev;
2858
2859         /*
2860          *      Configure the address for reception. Now it is valid.
2861          */
2862
2863         ipv6_ifa_notify(RTM_NEWADDR, ifp);
2864
2865         /* If added prefix is link local and forwarding is off,
2866            start sending router solicitations.
2867          */
2868
2869         if (ifp->idev->cnf.forwarding == 0 &&
2870             ifp->idev->cnf.rtr_solicits > 0 &&
2871             (dev->flags&IFF_LOOPBACK) == 0 &&
2872             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
2873                 /*
2874                  *      If a host as already performed a random delay
2875                  *      [...] as part of DAD [...] there is no need
2876                  *      to delay again before sending the first RS
2877                  */
2878                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2879
2880                 spin_lock_bh(&ifp->lock);
2881                 ifp->probes = 1;
2882                 ifp->idev->if_flags |= IF_RS_SENT;
2883                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
2884                 spin_unlock_bh(&ifp->lock);
2885         }
2886 }
2887
2888 static void addrconf_dad_run(struct inet6_dev *idev) {
2889         struct inet6_ifaddr *ifp;
2890
2891         read_lock_bh(&idev->lock);
2892         for (ifp = idev->addr_list; ifp; ifp = ifp->if_next) {
2893                 spin_lock_bh(&ifp->lock);
2894                 if (!(ifp->flags & IFA_F_TENTATIVE)) {
2895                         spin_unlock_bh(&ifp->lock);
2896                         continue;
2897                 }
2898                 spin_unlock_bh(&ifp->lock);
2899                 addrconf_dad_kick(ifp);
2900         }
2901         read_unlock_bh(&idev->lock);
2902 }
2903
2904 #ifdef CONFIG_PROC_FS
2905 struct if6_iter_state {
2906         struct seq_net_private p;
2907         int bucket;
2908 };
2909
2910 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
2911 {
2912         struct inet6_ifaddr *ifa = NULL;
2913         struct if6_iter_state *state = seq->private;
2914         struct net *net = seq_file_net(seq);
2915
2916         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
2917                 ifa = inet6_addr_lst[state->bucket];
2918
2919                 while (ifa && !net_eq(dev_net(ifa->idev->dev), net))
2920                         ifa = ifa->lst_next;
2921                 if (ifa)
2922                         break;
2923         }
2924         return ifa;
2925 }
2926
2927 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, struct inet6_ifaddr *ifa)
2928 {
2929         struct if6_iter_state *state = seq->private;
2930         struct net *net = seq_file_net(seq);
2931
2932         ifa = ifa->lst_next;
2933 try_again:
2934         if (ifa) {
2935                 if (!net_eq(dev_net(ifa->idev->dev), net)) {
2936                         ifa = ifa->lst_next;
2937                         goto try_again;
2938                 }
2939         }
2940
2941         if (!ifa && ++state->bucket < IN6_ADDR_HSIZE) {
2942                 ifa = inet6_addr_lst[state->bucket];
2943                 goto try_again;
2944         }
2945
2946         return ifa;
2947 }
2948
2949 static struct inet6_ifaddr *if6_get_idx(struct seq_file *seq, loff_t pos)
2950 {
2951         struct inet6_ifaddr *ifa = if6_get_first(seq);
2952
2953         if (ifa)
2954                 while(pos && (ifa = if6_get_next(seq, ifa)) != NULL)
2955                         --pos;
2956         return pos ? NULL : ifa;
2957 }
2958
2959 static void *if6_seq_start(struct seq_file *seq, loff_t *pos)
2960         __acquires(addrconf_hash_lock)
2961 {
2962         read_lock_bh(&addrconf_hash_lock);
2963         return if6_get_idx(seq, *pos);
2964 }
2965
2966 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2967 {
2968         struct inet6_ifaddr *ifa;
2969
2970         ifa = if6_get_next(seq, v);
2971         ++*pos;
2972         return ifa;
2973 }
2974
2975 static void if6_seq_stop(struct seq_file *seq, void *v)
2976         __releases(addrconf_hash_lock)
2977 {
2978         read_unlock_bh(&addrconf_hash_lock);
2979 }
2980
2981 static int if6_seq_show(struct seq_file *seq, void *v)
2982 {
2983         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v;
2984         seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n",
2985                    &ifp->addr,
2986                    ifp->idev->dev->ifindex,
2987                    ifp->prefix_len,
2988                    ifp->scope,
2989                    ifp->flags,
2990                    ifp->idev->dev->name);
2991         return 0;
2992 }
2993
2994 static const struct seq_operations if6_seq_ops = {
2995         .start  = if6_seq_start,
2996         .next   = if6_seq_next,
2997         .show   = if6_seq_show,
2998         .stop   = if6_seq_stop,
2999 };
3000
3001 static int if6_seq_open(struct inode *inode, struct file *file)
3002 {
3003         return seq_open_net(inode, file, &if6_seq_ops,
3004                             sizeof(struct if6_iter_state));
3005 }
3006
3007 static const struct file_operations if6_fops = {
3008         .owner          = THIS_MODULE,
3009         .open           = if6_seq_open,
3010         .read           = seq_read,
3011         .llseek         = seq_lseek,
3012         .release        = seq_release_net,
3013 };
3014
3015 static int if6_proc_net_init(struct net *net)
3016 {
3017         if (!proc_net_fops_create(net, "if_inet6", S_IRUGO, &if6_fops))
3018                 return -ENOMEM;
3019         return 0;
3020 }
3021
3022 static void if6_proc_net_exit(struct net *net)
3023 {
3024        proc_net_remove(net, "if_inet6");
3025 }
3026
3027 static struct pernet_operations if6_proc_net_ops = {
3028        .init = if6_proc_net_init,
3029        .exit = if6_proc_net_exit,
3030 };
3031
3032 int __init if6_proc_init(void)
3033 {
3034         return register_pernet_subsys(&if6_proc_net_ops);
3035 }
3036
3037 void if6_proc_exit(void)
3038 {
3039         unregister_pernet_subsys(&if6_proc_net_ops);
3040 }
3041 #endif  /* CONFIG_PROC_FS */
3042
3043 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE)
3044 /* Check if address is a home address configured on any interface. */
3045 int ipv6_chk_home_addr(struct net *net, struct in6_addr *addr)
3046 {
3047         int ret = 0;
3048         struct inet6_ifaddr * ifp;
3049         u8 hash = ipv6_addr_hash(addr);
3050         read_lock_bh(&addrconf_hash_lock);
3051         for (ifp = inet6_addr_lst[hash]; ifp; ifp = ifp->lst_next) {
3052                 if (!net_eq(dev_net(ifp->idev->dev), net))
3053                         continue;
3054                 if (ipv6_addr_equal(&ifp->addr, addr) &&
3055                     (ifp->flags & IFA_F_HOMEADDRESS)) {
3056                         ret = 1;
3057                         break;
3058                 }
3059         }
3060         read_unlock_bh(&addrconf_hash_lock);
3061         return ret;
3062 }
3063 #endif
3064
3065 /*
3066  *      Periodic address status verification
3067  */
3068
3069 static void addrconf_verify(unsigned long foo)
3070 {
3071         struct inet6_ifaddr *ifp;
3072         unsigned long now, next;
3073         int i;
3074
3075         spin_lock_bh(&addrconf_verify_lock);
3076         now = jiffies;
3077         next = now + ADDR_CHECK_FREQUENCY;
3078
3079         del_timer(&addr_chk_timer);
3080
3081         for (i=0; i < IN6_ADDR_HSIZE; i++) {
3082
3083 restart:
3084                 read_lock(&addrconf_hash_lock);
3085                 for (ifp=inet6_addr_lst[i]; ifp; ifp=ifp->lst_next) {
3086                         unsigned long age;
3087 #ifdef CONFIG_IPV6_PRIVACY
3088                         unsigned long regen_advance;
3089 #endif
3090
3091                         if (ifp->flags & IFA_F_PERMANENT)
3092                                 continue;
3093
3094                         spin_lock(&ifp->lock);
3095                         age = (now - ifp->tstamp) / HZ;
3096
3097 #ifdef CONFIG_IPV6_PRIVACY
3098                         regen_advance = ifp->idev->cnf.regen_max_retry *
3099                                         ifp->idev->cnf.dad_transmits *
3100                                         ifp->idev->nd_parms->retrans_time / HZ;
3101 #endif
3102
3103                         if (ifp->valid_lft != INFINITY_LIFE_TIME &&
3104                             age >= ifp->valid_lft) {
3105                                 spin_unlock(&ifp->lock);
3106                                 in6_ifa_hold(ifp);
3107                                 read_unlock(&addrconf_hash_lock);
3108                                 ipv6_del_addr(ifp);
3109                                 goto restart;
3110                         } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) {
3111                                 spin_unlock(&ifp->lock);
3112                                 continue;
3113                         } else if (age >= ifp->prefered_lft) {
3114                                 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */
3115                                 int deprecate = 0;
3116
3117                                 if (!(ifp->flags&IFA_F_DEPRECATED)) {
3118                                         deprecate = 1;
3119                                         ifp->flags |= IFA_F_DEPRECATED;
3120                                 }
3121
3122                                 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))
3123                                         next = ifp->tstamp + ifp->valid_lft * HZ;
3124
3125                                 spin_unlock(&ifp->lock);
3126
3127                                 if (deprecate) {
3128                                         in6_ifa_hold(ifp);
3129                                         read_unlock(&addrconf_hash_lock);
3130
3131                                         ipv6_ifa_notify(0, ifp);
3132                                         in6_ifa_put(ifp);
3133                                         goto restart;
3134                                 }
3135 #ifdef CONFIG_IPV6_PRIVACY
3136                         } else if ((ifp->flags&IFA_F_TEMPORARY) &&
3137                                    !(ifp->flags&IFA_F_TENTATIVE)) {
3138                                 if (age >= ifp->prefered_lft - regen_advance) {
3139                                         struct inet6_ifaddr *ifpub = ifp->ifpub;
3140                                         if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3141                                                 next = ifp->tstamp + ifp->prefered_lft * HZ;
3142                                         if (!ifp->regen_count && ifpub) {
3143                                                 ifp->regen_count++;
3144                                                 in6_ifa_hold(ifp);
3145                                                 in6_ifa_hold(ifpub);
3146                                                 spin_unlock(&ifp->lock);
3147                                                 read_unlock(&addrconf_hash_lock);
3148                                                 spin_lock(&ifpub->lock);
3149                                                 ifpub->regen_count = 0;
3150                                                 spin_unlock(&ifpub->lock);
3151                                                 ipv6_create_tempaddr(ifpub, ifp);
3152                                                 in6_ifa_put(ifpub);
3153                                                 in6_ifa_put(ifp);
3154                                                 goto restart;
3155                                         }
3156                                 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next))
3157                                         next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ;
3158                                 spin_unlock(&ifp->lock);
3159 #endif
3160                         } else {
3161                                 /* ifp->prefered_lft <= ifp->valid_lft */
3162                                 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next))
3163                                         next = ifp->tstamp + ifp->prefered_lft * HZ;
3164                                 spin_unlock(&ifp->lock);
3165                         }
3166                 }
3167                 read_unlock(&addrconf_hash_lock);
3168         }
3169
3170         addr_chk_timer.expires = time_before(next, jiffies + HZ) ? jiffies + HZ : next;
3171         add_timer(&addr_chk_timer);
3172         spin_unlock_bh(&addrconf_verify_lock);
3173 }
3174
3175 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local)
3176 {
3177         struct in6_addr *pfx = NULL;
3178
3179         if (addr)
3180                 pfx = nla_data(addr);
3181
3182         if (local) {
3183                 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx)))
3184                         pfx = NULL;
3185                 else
3186                         pfx = nla_data(local);
3187         }
3188
3189         return pfx;
3190 }
3191
3192 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = {
3193         [IFA_ADDRESS]           = { .len = sizeof(struct in6_addr) },
3194         [IFA_LOCAL]             = { .len = sizeof(struct in6_addr) },
3195         [IFA_CACHEINFO]         = { .len = sizeof(struct ifa_cacheinfo) },
3196 };
3197
3198 static int
3199 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3200 {
3201         struct net *net = sock_net(skb->sk);
3202         struct ifaddrmsg *ifm;
3203         struct nlattr *tb[IFA_MAX+1];
3204         struct in6_addr *pfx;
3205         int err;
3206
3207         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3208         if (err < 0)
3209                 return err;
3210
3211         ifm = nlmsg_data(nlh);
3212         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3213         if (pfx == NULL)
3214                 return -EINVAL;
3215
3216         return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen);
3217 }
3218
3219 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u8 ifa_flags,
3220                              u32 prefered_lft, u32 valid_lft)
3221 {
3222         u32 flags;
3223         clock_t expires;
3224         unsigned long timeout;
3225
3226         if (!valid_lft || (prefered_lft > valid_lft))
3227                 return -EINVAL;
3228
3229         timeout = addrconf_timeout_fixup(valid_lft, HZ);
3230         if (addrconf_finite_timeout(timeout)) {
3231                 expires = jiffies_to_clock_t(timeout * HZ);
3232                 valid_lft = timeout;
3233                 flags = RTF_EXPIRES;
3234         } else {
3235                 expires = 0;
3236                 flags = 0;
3237                 ifa_flags |= IFA_F_PERMANENT;
3238         }
3239
3240         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
3241         if (addrconf_finite_timeout(timeout)) {
3242                 if (timeout == 0)
3243                         ifa_flags |= IFA_F_DEPRECATED;
3244                 prefered_lft = timeout;
3245         }
3246
3247         spin_lock_bh(&ifp->lock);
3248         ifp->flags = (ifp->flags & ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | IFA_F_HOMEADDRESS)) | ifa_flags;
3249         ifp->tstamp = jiffies;
3250         ifp->valid_lft = valid_lft;
3251         ifp->prefered_lft = prefered_lft;
3252
3253         spin_unlock_bh(&ifp->lock);
3254         if (!(ifp->flags&IFA_F_TENTATIVE))
3255                 ipv6_ifa_notify(0, ifp);
3256
3257         addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev,
3258                               expires, flags);
3259         addrconf_verify(0);
3260
3261         return 0;
3262 }
3263
3264 static int
3265 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, void *arg)
3266 {
3267         struct net *net = sock_net(skb->sk);
3268         struct ifaddrmsg *ifm;
3269         struct nlattr *tb[IFA_MAX+1];
3270         struct in6_addr *pfx;
3271         struct inet6_ifaddr *ifa;
3272         struct net_device *dev;
3273         u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME;
3274         u8 ifa_flags;
3275         int err;
3276
3277         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3278         if (err < 0)
3279                 return err;
3280
3281         ifm = nlmsg_data(nlh);
3282         pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3283         if (pfx == NULL)
3284                 return -EINVAL;
3285
3286         if (tb[IFA_CACHEINFO]) {
3287                 struct ifa_cacheinfo *ci;
3288
3289                 ci = nla_data(tb[IFA_CACHEINFO]);
3290                 valid_lft = ci->ifa_valid;
3291                 preferred_lft = ci->ifa_prefered;
3292         } else {
3293                 preferred_lft = INFINITY_LIFE_TIME;
3294                 valid_lft = INFINITY_LIFE_TIME;
3295         }
3296
3297         dev =  __dev_get_by_index(net, ifm->ifa_index);
3298         if (dev == NULL)
3299                 return -ENODEV;
3300
3301         /* We ignore other flags so far. */
3302         ifa_flags = ifm->ifa_flags & (IFA_F_NODAD | IFA_F_HOMEADDRESS);
3303
3304         ifa = ipv6_get_ifaddr(net, pfx, dev, 1);
3305         if (ifa == NULL) {
3306                 /*
3307                  * It would be best to check for !NLM_F_CREATE here but
3308                  * userspace alreay relies on not having to provide this.
3309                  */
3310                 return inet6_addr_add(net, ifm->ifa_index, pfx,
3311                                       ifm->ifa_prefixlen, ifa_flags,
3312                                       preferred_lft, valid_lft);
3313         }
3314
3315         if (nlh->nlmsg_flags & NLM_F_EXCL ||
3316             !(nlh->nlmsg_flags & NLM_F_REPLACE))
3317                 err = -EEXIST;
3318         else
3319                 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft);
3320
3321         in6_ifa_put(ifa);
3322
3323         return err;
3324 }
3325
3326 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u8 flags,
3327                           u8 scope, int ifindex)
3328 {
3329         struct ifaddrmsg *ifm;
3330
3331         ifm = nlmsg_data(nlh);
3332         ifm->ifa_family = AF_INET6;
3333         ifm->ifa_prefixlen = prefixlen;
3334         ifm->ifa_flags = flags;
3335         ifm->ifa_scope = scope;
3336         ifm->ifa_index = ifindex;
3337 }
3338
3339 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp,
3340                          unsigned long tstamp, u32 preferred, u32 valid)
3341 {
3342         struct ifa_cacheinfo ci;
3343
3344         ci.cstamp = (u32)(TIME_DELTA(cstamp, INITIAL_JIFFIES) / HZ * 100
3345                         + TIME_DELTA(cstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3346         ci.tstamp = (u32)(TIME_DELTA(tstamp, INITIAL_JIFFIES) / HZ * 100
3347                         + TIME_DELTA(tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3348         ci.ifa_prefered = preferred;
3349         ci.ifa_valid = valid;
3350
3351         return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci);
3352 }
3353
3354 static inline int rt_scope(int ifa_scope)
3355 {
3356         if (ifa_scope & IFA_HOST)
3357                 return RT_SCOPE_HOST;
3358         else if (ifa_scope & IFA_LINK)
3359                 return RT_SCOPE_LINK;
3360         else if (ifa_scope & IFA_SITE)
3361                 return RT_SCOPE_SITE;
3362         else
3363                 return RT_SCOPE_UNIVERSE;
3364 }
3365
3366 static inline int inet6_ifaddr_msgsize(void)
3367 {
3368         return NLMSG_ALIGN(sizeof(struct ifaddrmsg))
3369                + nla_total_size(16) /* IFA_ADDRESS */
3370                + nla_total_size(sizeof(struct ifa_cacheinfo));
3371 }
3372
3373 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa,
3374                              u32 pid, u32 seq, int event, unsigned int flags)
3375 {
3376         struct nlmsghdr  *nlh;
3377         u32 preferred, valid;
3378
3379         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3380         if (nlh == NULL)
3381                 return -EMSGSIZE;
3382
3383         put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope),
3384                       ifa->idev->dev->ifindex);
3385
3386         if (!(ifa->flags&IFA_F_PERMANENT)) {
3387                 preferred = ifa->prefered_lft;
3388                 valid = ifa->valid_lft;
3389                 if (preferred != INFINITY_LIFE_TIME) {
3390                         long tval = (jiffies - ifa->tstamp)/HZ;
3391                         if (preferred > tval)
3392                                 preferred -= tval;
3393                         else
3394                                 preferred = 0;
3395                         if (valid != INFINITY_LIFE_TIME)
3396                                 valid -= tval;
3397                 }
3398         } else {
3399                 preferred = INFINITY_LIFE_TIME;
3400                 valid = INFINITY_LIFE_TIME;
3401         }
3402
3403         if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0 ||
3404             put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) {
3405                 nlmsg_cancel(skb, nlh);
3406                 return -EMSGSIZE;
3407         }
3408
3409         return nlmsg_end(skb, nlh);
3410 }
3411
3412 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca,
3413                                 u32 pid, u32 seq, int event, u16 flags)
3414 {
3415         struct nlmsghdr  *nlh;
3416         u8 scope = RT_SCOPE_UNIVERSE;
3417         int ifindex = ifmca->idev->dev->ifindex;
3418
3419         if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE)
3420                 scope = RT_SCOPE_SITE;
3421
3422         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3423         if (nlh == NULL)
3424                 return -EMSGSIZE;
3425
3426         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3427         if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 ||
3428             put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp,
3429                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3430                 nlmsg_cancel(skb, nlh);
3431                 return -EMSGSIZE;
3432         }
3433
3434         return nlmsg_end(skb, nlh);
3435 }
3436
3437 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca,
3438                                 u32 pid, u32 seq, int event, unsigned int flags)
3439 {
3440         struct nlmsghdr  *nlh;
3441         u8 scope = RT_SCOPE_UNIVERSE;
3442         int ifindex = ifaca->aca_idev->dev->ifindex;
3443
3444         if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE)
3445                 scope = RT_SCOPE_SITE;
3446
3447         nlh = nlmsg_put(skb, pid, seq, event, sizeof(struct ifaddrmsg), flags);
3448         if (nlh == NULL)
3449                 return -EMSGSIZE;
3450
3451         put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex);
3452         if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 ||
3453             put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp,
3454                           INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) {
3455                 nlmsg_cancel(skb, nlh);
3456                 return -EMSGSIZE;
3457         }
3458
3459         return nlmsg_end(skb, nlh);
3460 }
3461
3462 enum addr_type_t
3463 {
3464         UNICAST_ADDR,
3465         MULTICAST_ADDR,
3466         ANYCAST_ADDR,
3467 };
3468
3469 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb,
3470                            enum addr_type_t type)
3471 {
3472         int idx, ip_idx;
3473         int s_idx, s_ip_idx;
3474         int err = 1;
3475         struct net_device *dev;
3476         struct inet6_dev *idev = NULL;
3477         struct inet6_ifaddr *ifa;
3478         struct ifmcaddr6 *ifmca;
3479         struct ifacaddr6 *ifaca;
3480         struct net *net = sock_net(skb->sk);
3481
3482         s_idx = cb->args[0];
3483         s_ip_idx = ip_idx = cb->args[1];
3484
3485         idx = 0;
3486         for_each_netdev(net, dev) {
3487                 if (idx < s_idx)
3488                         goto cont;
3489                 if (idx > s_idx)
3490                         s_ip_idx = 0;
3491                 ip_idx = 0;
3492                 if ((idev = in6_dev_get(dev)) == NULL)
3493                         goto cont;
3494                 read_lock_bh(&idev->lock);
3495                 switch (type) {
3496                 case UNICAST_ADDR:
3497                         /* unicast address incl. temp addr */
3498                         for (ifa = idev->addr_list; ifa;
3499                              ifa = ifa->if_next, ip_idx++) {
3500                                 if (ip_idx < s_ip_idx)
3501                                         continue;
3502                                 err = inet6_fill_ifaddr(skb, ifa,
3503                                                         NETLINK_CB(cb->skb).pid,
3504                                                         cb->nlh->nlmsg_seq,
3505                                                         RTM_NEWADDR,
3506                                                         NLM_F_MULTI);
3507                         }
3508                         break;
3509                 case MULTICAST_ADDR:
3510                         /* multicast address */
3511                         for (ifmca = idev->mc_list; ifmca;
3512                              ifmca = ifmca->next, ip_idx++) {
3513                                 if (ip_idx < s_ip_idx)
3514                                         continue;
3515                                 err = inet6_fill_ifmcaddr(skb, ifmca,
3516                                                           NETLINK_CB(cb->skb).pid,
3517                                                           cb->nlh->nlmsg_seq,
3518                                                           RTM_GETMULTICAST,
3519                                                           NLM_F_MULTI);
3520                         }
3521                         break;
3522                 case ANYCAST_ADDR:
3523                         /* anycast address */
3524                         for (ifaca = idev->ac_list; ifaca;
3525                              ifaca = ifaca->aca_next, ip_idx++) {
3526                                 if (ip_idx < s_ip_idx)
3527                                         continue;
3528                                 err = inet6_fill_ifacaddr(skb, ifaca,
3529                                                           NETLINK_CB(cb->skb).pid,
3530                                                           cb->nlh->nlmsg_seq,
3531                                                           RTM_GETANYCAST,
3532                                                           NLM_F_MULTI);
3533                         }
3534                         break;
3535                 default:
3536                         break;
3537                 }
3538                 read_unlock_bh(&idev->lock);
3539                 in6_dev_put(idev);
3540
3541                 if (err <= 0)
3542                         break;
3543 cont:
3544                 idx++;
3545         }
3546         cb->args[0] = idx;
3547         cb->args[1] = ip_idx;
3548         return skb->len;
3549 }
3550
3551 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb)
3552 {
3553         enum addr_type_t type = UNICAST_ADDR;
3554
3555         return inet6_dump_addr(skb, cb, type);
3556 }
3557
3558 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb)
3559 {
3560         enum addr_type_t type = MULTICAST_ADDR;
3561
3562         return inet6_dump_addr(skb, cb, type);
3563 }
3564
3565
3566 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb)
3567 {
3568         enum addr_type_t type = ANYCAST_ADDR;
3569
3570         return inet6_dump_addr(skb, cb, type);
3571 }
3572
3573 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr* nlh,
3574                              void *arg)
3575 {
3576         struct net *net = sock_net(in_skb->sk);
3577         struct ifaddrmsg *ifm;
3578         struct nlattr *tb[IFA_MAX+1];
3579         struct in6_addr *addr = NULL;
3580         struct net_device *dev = NULL;
3581         struct inet6_ifaddr *ifa;
3582         struct sk_buff *skb;
3583         int err;
3584
3585         err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy);
3586         if (err < 0)
3587                 goto errout;
3588
3589         addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL]);
3590         if (addr == NULL) {
3591                 err = -EINVAL;
3592                 goto errout;
3593         }
3594
3595         ifm = nlmsg_data(nlh);
3596         if (ifm->ifa_index)
3597                 dev = __dev_get_by_index(net, ifm->ifa_index);
3598
3599         if ((ifa = ipv6_get_ifaddr(net, addr, dev, 1)) == NULL) {
3600                 err = -EADDRNOTAVAIL;
3601                 goto errout;
3602         }
3603
3604         if ((skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL)) == NULL) {
3605                 err = -ENOBUFS;
3606                 goto errout_ifa;
3607         }
3608
3609         err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).pid,
3610                                 nlh->nlmsg_seq, RTM_NEWADDR, 0);
3611         if (err < 0) {
3612                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3613                 WARN_ON(err == -EMSGSIZE);
3614                 kfree_skb(skb);
3615                 goto errout_ifa;
3616         }
3617         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid);
3618 errout_ifa:
3619         in6_ifa_put(ifa);
3620 errout:
3621         return err;
3622 }
3623
3624 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa)
3625 {
3626         struct sk_buff *skb;
3627         struct net *net = dev_net(ifa->idev->dev);
3628         int err = -ENOBUFS;
3629
3630         skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC);
3631         if (skb == NULL)
3632                 goto errout;
3633
3634         err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0);
3635         if (err < 0) {
3636                 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */
3637                 WARN_ON(err == -EMSGSIZE);
3638                 kfree_skb(skb);
3639                 goto errout;
3640         }
3641         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3642         return;
3643 errout:
3644         if (err < 0)
3645                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3646 }
3647
3648 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf,
3649                                 __s32 *array, int bytes)
3650 {
3651         BUG_ON(bytes < (DEVCONF_MAX * 4));
3652
3653         memset(array, 0, bytes);
3654         array[DEVCONF_FORWARDING] = cnf->forwarding;
3655         array[DEVCONF_HOPLIMIT] = cnf->hop_limit;
3656         array[DEVCONF_MTU6] = cnf->mtu6;
3657         array[DEVCONF_ACCEPT_RA] = cnf->accept_ra;
3658         array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects;
3659         array[DEVCONF_AUTOCONF] = cnf->autoconf;
3660         array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits;
3661         array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits;
3662         array[DEVCONF_RTR_SOLICIT_INTERVAL] = cnf->rtr_solicit_interval;
3663         array[DEVCONF_RTR_SOLICIT_DELAY] = cnf->rtr_solicit_delay;
3664         array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version;
3665 #ifdef CONFIG_IPV6_PRIVACY
3666         array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr;
3667         array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft;
3668         array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft;
3669         array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry;
3670         array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor;
3671 #endif
3672         array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses;
3673         array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr;
3674         array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo;
3675 #ifdef CONFIG_IPV6_ROUTER_PREF
3676         array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref;
3677         array[DEVCONF_RTR_PROBE_INTERVAL] = cnf->rtr_probe_interval;
3678 #ifdef CONFIG_IPV6_ROUTE_INFO
3679         array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen;
3680 #endif
3681 #endif
3682         array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp;
3683         array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route;
3684 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
3685         array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad;
3686 #endif
3687 #ifdef CONFIG_IPV6_MROUTE
3688         array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding;
3689 #endif
3690         array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6;
3691         array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad;
3692 }
3693
3694 static inline size_t inet6_if_nlmsg_size(void)
3695 {
3696         return NLMSG_ALIGN(sizeof(struct ifinfomsg))
3697                + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */
3698                + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */
3699                + nla_total_size(4) /* IFLA_MTU */
3700                + nla_total_size(4) /* IFLA_LINK */
3701                + nla_total_size( /* IFLA_PROTINFO */
3702                         nla_total_size(4) /* IFLA_INET6_FLAGS */
3703                         + nla_total_size(sizeof(struct ifla_cacheinfo))
3704                         + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */
3705                         + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */
3706                         + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */
3707                  );
3708 }
3709
3710 static inline void __snmp6_fill_stats(u64 *stats, void **mib, int items,
3711                                       int bytes)
3712 {
3713         int i;
3714         int pad = bytes - sizeof(u64) * items;
3715         BUG_ON(pad < 0);
3716
3717         /* Use put_unaligned() because stats may not be aligned for u64. */
3718         put_unaligned(items, &stats[0]);
3719         for (i = 1; i < items; i++)
3720                 put_unaligned(snmp_fold_field(mib, i), &stats[i]);
3721
3722         memset(&stats[items], 0, pad);
3723 }
3724
3725 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype,
3726                              int bytes)
3727 {
3728         switch(attrtype) {
3729         case IFLA_INET6_STATS:
3730                 __snmp6_fill_stats(stats, (void **)idev->stats.ipv6, IPSTATS_MIB_MAX, bytes);
3731                 break;
3732         case IFLA_INET6_ICMP6STATS:
3733                 __snmp6_fill_stats(stats, (void **)idev->stats.icmpv6, ICMP6_MIB_MAX, bytes);
3734                 break;
3735         }
3736 }
3737
3738 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev,
3739                              u32 pid, u32 seq, int event, unsigned int flags)
3740 {
3741         struct net_device *dev = idev->dev;
3742         struct nlattr *nla;
3743         struct ifinfomsg *hdr;
3744         struct nlmsghdr *nlh;
3745         void *protoinfo;
3746         struct ifla_cacheinfo ci;
3747
3748         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*hdr), flags);
3749         if (nlh == NULL)
3750                 return -EMSGSIZE;
3751
3752         hdr = nlmsg_data(nlh);
3753         hdr->ifi_family = AF_INET6;
3754         hdr->__ifi_pad = 0;
3755         hdr->ifi_type = dev->type;
3756         hdr->ifi_index = dev->ifindex;
3757         hdr->ifi_flags = dev_get_flags(dev);
3758         hdr->ifi_change = 0;
3759
3760         NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
3761
3762         if (dev->addr_len)
3763                 NLA_PUT(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr);
3764
3765         NLA_PUT_U32(skb, IFLA_MTU, dev->mtu);
3766         if (dev->ifindex != dev->iflink)
3767                 NLA_PUT_U32(skb, IFLA_LINK, dev->iflink);
3768
3769         protoinfo = nla_nest_start(skb, IFLA_PROTINFO);
3770         if (protoinfo == NULL)
3771                 goto nla_put_failure;
3772
3773         NLA_PUT_U32(skb, IFLA_INET6_FLAGS, idev->if_flags);
3774
3775         ci.max_reasm_len = IPV6_MAXPLEN;
3776         ci.tstamp = (__u32)(TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) / HZ * 100
3777                     + TIME_DELTA(idev->tstamp, INITIAL_JIFFIES) % HZ * 100 / HZ);
3778         ci.reachable_time = idev->nd_parms->reachable_time;
3779         ci.retrans_time = idev->nd_parms->retrans_time;
3780         NLA_PUT(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci);
3781
3782         nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32));
3783         if (nla == NULL)
3784                 goto nla_put_failure;
3785         ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla));
3786
3787         /* XXX - MC not implemented */
3788
3789         nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64));
3790         if (nla == NULL)
3791                 goto nla_put_failure;
3792         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla));
3793
3794         nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64));
3795         if (nla == NULL)
3796                 goto nla_put_failure;
3797         snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla));
3798
3799         nla_nest_end(skb, protoinfo);
3800         return nlmsg_end(skb, nlh);
3801
3802 nla_put_failure:
3803         nlmsg_cancel(skb, nlh);
3804         return -EMSGSIZE;
3805 }
3806
3807 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb)
3808 {
3809         struct net *net = sock_net(skb->sk);
3810         int idx, err;
3811         int s_idx = cb->args[0];
3812         struct net_device *dev;
3813         struct inet6_dev *idev;
3814
3815         read_lock(&dev_base_lock);
3816         idx = 0;
3817         for_each_netdev(net, dev) {
3818                 if (idx < s_idx)
3819                         goto cont;
3820                 if ((idev = in6_dev_get(dev)) == NULL)
3821                         goto cont;
3822                 err = inet6_fill_ifinfo(skb, idev, NETLINK_CB(cb->skb).pid,
3823                                 cb->nlh->nlmsg_seq, RTM_NEWLINK, NLM_F_MULTI);
3824                 in6_dev_put(idev);
3825                 if (err <= 0)
3826                         break;
3827 cont:
3828                 idx++;
3829         }
3830         read_unlock(&dev_base_lock);
3831         cb->args[0] = idx;
3832
3833         return skb->len;
3834 }
3835
3836 void inet6_ifinfo_notify(int event, struct inet6_dev *idev)
3837 {
3838         struct sk_buff *skb;
3839         struct net *net = dev_net(idev->dev);
3840         int err = -ENOBUFS;
3841
3842         skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC);
3843         if (skb == NULL)
3844                 goto errout;
3845
3846         err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0);
3847         if (err < 0) {
3848                 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */
3849                 WARN_ON(err == -EMSGSIZE);
3850                 kfree_skb(skb);
3851                 goto errout;
3852         }
3853         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC);
3854         return;
3855 errout:
3856         if (err < 0)
3857                 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err);
3858 }
3859
3860 static inline size_t inet6_prefix_nlmsg_size(void)
3861 {
3862         return NLMSG_ALIGN(sizeof(struct prefixmsg))
3863                + nla_total_size(sizeof(struct in6_addr))
3864                + nla_total_size(sizeof(struct prefix_cacheinfo));
3865 }
3866
3867 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev,
3868                              struct prefix_info *pinfo, u32 pid, u32 seq,
3869                              int event, unsigned int flags)
3870 {
3871         struct prefixmsg *pmsg;
3872         struct nlmsghdr *nlh;
3873         struct prefix_cacheinfo ci;
3874
3875         nlh = nlmsg_put(skb, pid, seq, event, sizeof(*pmsg), flags);
3876         if (nlh == NULL)
3877                 return -EMSGSIZE;
3878
3879         pmsg = nlmsg_data(nlh);
3880         pmsg->prefix_family = AF_INET6;
3881         pmsg->prefix_pad1 = 0;
3882         pmsg->prefix_pad2 = 0;
3883         pmsg->prefix_ifindex = idev->dev->ifindex;
3884         pmsg->prefix_len = pinfo->prefix_len;
3885         pmsg->prefix_type = pinfo->type;
3886         pmsg->prefix_pad3 = 0;
3887         pmsg->prefix_flags = 0;
3888         if (pinfo->onlink)
3889                 pmsg->prefix_flags |= IF_PREFIX_ONLINK;
3890         if (pinfo->autoconf)
3891                 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF;
3892
3893         NLA_PUT(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix);
3894
3895         ci.preferred_time = ntohl(pinfo->prefered);
3896         ci.valid_time = ntohl(pinfo->valid);
3897         NLA_PUT(skb, PREFIX_CACHEINFO, sizeof(ci), &ci);
3898
3899         return nlmsg_end(skb, nlh);
3900
3901 nla_put_failure:
3902         nlmsg_cancel(skb, nlh);
3903         return -EMSGSIZE;
3904 }
3905
3906 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
3907                          struct prefix_info *pinfo)
3908 {
3909         struct sk_buff *skb;
3910         struct net *net = dev_net(idev->dev);
3911         int err = -ENOBUFS;
3912
3913         skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC);
3914         if (skb == NULL)
3915                 goto errout;
3916
3917         err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0);
3918         if (err < 0) {
3919                 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */
3920                 WARN_ON(err == -EMSGSIZE);
3921                 kfree_skb(skb);
3922                 goto errout;
3923         }
3924         rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC);
3925         return;
3926 errout:
3927         if (err < 0)
3928                 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err);
3929 }
3930
3931 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3932 {
3933         inet6_ifa_notify(event ? : RTM_NEWADDR, ifp);
3934
3935         switch (event) {
3936         case RTM_NEWADDR:
3937                 /*
3938                  * If the address was optimistic
3939                  * we inserted the route at the start of
3940                  * our DAD process, so we don't need
3941                  * to do it again
3942                  */
3943                 if (!(ifp->rt->rt6i_node))
3944                         ip6_ins_rt(ifp->rt);
3945                 if (ifp->idev->cnf.forwarding)
3946                         addrconf_join_anycast(ifp);
3947                 break;
3948         case RTM_DELADDR:
3949                 if (ifp->idev->cnf.forwarding)
3950                         addrconf_leave_anycast(ifp);
3951                 addrconf_leave_solict(ifp->idev, &ifp->addr);
3952                 dst_hold(&ifp->rt->u.dst);
3953                 if (ip6_del_rt(ifp->rt))
3954                         dst_free(&ifp->rt->u.dst);
3955                 break;
3956         }
3957 }
3958
3959 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp)
3960 {
3961         rcu_read_lock_bh();
3962         if (likely(ifp->idev->dead == 0))
3963                 __ipv6_ifa_notify(event, ifp);
3964         rcu_read_unlock_bh();
3965 }
3966
3967 #ifdef CONFIG_SYSCTL
3968
3969 static
3970 int addrconf_sysctl_forward(ctl_table *ctl, int write, struct file * filp,
3971                            void __user *buffer, size_t *lenp, loff_t *ppos)
3972 {
3973         int *valp = ctl->data;
3974         int val = *valp;
3975         int ret;
3976
3977         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
3978
3979         if (write)
3980                 ret = addrconf_fixup_forwarding(ctl, valp, val);
3981         return ret;
3982 }
3983
3984 static int addrconf_sysctl_forward_strategy(ctl_table *table,
3985                                             void __user *oldval,
3986                                             size_t __user *oldlenp,
3987                                             void __user *newval, size_t newlen)
3988 {
3989         int *valp = table->data;
3990         int val = *valp;
3991         int new;
3992
3993         if (!newval || !newlen)
3994                 return 0;
3995         if (newlen != sizeof(int))
3996                 return -EINVAL;
3997         if (get_user(new, (int __user *)newval))
3998                 return -EFAULT;
3999         if (new == *valp)
4000                 return 0;
4001         if (oldval && oldlenp) {
4002                 size_t len;
4003                 if (get_user(len, oldlenp))
4004                         return -EFAULT;
4005                 if (len) {
4006                         if (len > table->maxlen)
4007                                 len = table->maxlen;
4008                         if (copy_to_user(oldval, valp, len))
4009                                 return -EFAULT;
4010                         if (put_user(len, oldlenp))
4011                                 return -EFAULT;
4012                 }
4013         }
4014
4015         *valp = new;
4016         return addrconf_fixup_forwarding(table, valp, val);
4017 }
4018
4019 static void dev_disable_change(struct inet6_dev *idev)
4020 {
4021         if (!idev || !idev->dev)
4022                 return;
4023
4024         if (idev->cnf.disable_ipv6)
4025                 addrconf_notify(NULL, NETDEV_DOWN, idev->dev);
4026         else
4027                 addrconf_notify(NULL, NETDEV_UP, idev->dev);
4028 }
4029
4030 static void addrconf_disable_change(struct net *net, __s32 newf)
4031 {
4032         struct net_device *dev;
4033         struct inet6_dev *idev;
4034
4035         read_lock(&dev_base_lock);
4036         for_each_netdev(net, dev) {
4037                 rcu_read_lock();
4038                 idev = __in6_dev_get(dev);
4039                 if (idev) {
4040                         int changed = (!idev->cnf.disable_ipv6) ^ (!newf);
4041                         idev->cnf.disable_ipv6 = newf;
4042                         if (changed)
4043                                 dev_disable_change(idev);
4044                 }
4045                 rcu_read_unlock();
4046         }
4047         read_unlock(&dev_base_lock);
4048 }
4049
4050 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int old)
4051 {
4052         struct net *net;
4053
4054         net = (struct net *)table->extra2;
4055
4056         if (p == &net->ipv6.devconf_dflt->disable_ipv6)
4057                 return 0;
4058
4059         if (!rtnl_trylock())
4060                 return restart_syscall();
4061
4062         if (p == &net->ipv6.devconf_all->disable_ipv6) {
4063                 __s32 newf = net->ipv6.devconf_all->disable_ipv6;
4064                 net->ipv6.devconf_dflt->disable_ipv6 = newf;
4065                 addrconf_disable_change(net, newf);
4066         } else if ((!*p) ^ (!old))
4067                 dev_disable_change((struct inet6_dev *)table->extra1);
4068
4069         rtnl_unlock();
4070         return 0;
4071 }
4072
4073 static
4074 int addrconf_sysctl_disable(ctl_table *ctl, int write, struct file * filp,
4075                             void __user *buffer, size_t *lenp, loff_t *ppos)
4076 {
4077         int *valp = ctl->data;
4078         int val = *valp;
4079         int ret;
4080
4081         ret = proc_dointvec(ctl, write, filp, buffer, lenp, ppos);
4082
4083         if (write)
4084                 ret = addrconf_disable_ipv6(ctl, valp, val);
4085         return ret;
4086 }
4087
4088 static struct addrconf_sysctl_table
4089 {
4090         struct ctl_table_header *sysctl_header;
4091         ctl_table addrconf_vars[DEVCONF_MAX+1];
4092         char *dev_name;
4093 } addrconf_sysctl __read_mostly = {
4094         .sysctl_header = NULL,
4095         .addrconf_vars = {
4096                 {
4097                         .ctl_name       =       NET_IPV6_FORWARDING,
4098                         .procname       =       "forwarding",
4099                         .data           =       &ipv6_devconf.forwarding,
4100                         .maxlen         =       sizeof(int),
4101                         .mode           =       0644,
4102                         .proc_handler   =       addrconf_sysctl_forward,
4103                         .strategy       =       addrconf_sysctl_forward_strategy,
4104                 },
4105                 {
4106                         .ctl_name       =       NET_IPV6_HOP_LIMIT,
4107                         .procname       =       "hop_limit",
4108                         .data           =       &ipv6_devconf.hop_limit,
4109                         .maxlen         =       sizeof(int),
4110                         .mode           =       0644,
4111                         .proc_handler   =       proc_dointvec,
4112                 },
4113                 {
4114                         .ctl_name       =       NET_IPV6_MTU,
4115                         .procname       =       "mtu",
4116                         .data           =       &ipv6_devconf.mtu6,
4117                         .maxlen         =       sizeof(int),
4118                         .mode           =       0644,
4119                         .proc_handler   =       proc_dointvec,
4120                 },
4121                 {
4122                         .ctl_name       =       NET_IPV6_ACCEPT_RA,
4123                         .procname       =       "accept_ra",
4124                         .data           =       &ipv6_devconf.accept_ra,
4125                         .maxlen         =       sizeof(int),
4126                         .mode           =       0644,
4127                         .proc_handler   =       proc_dointvec,
4128                 },
4129                 {
4130                         .ctl_name       =       NET_IPV6_ACCEPT_REDIRECTS,
4131                         .procname       =       "accept_redirects",
4132                         .data           =       &ipv6_devconf.accept_redirects,
4133                         .maxlen         =       sizeof(int),
4134                         .mode           =       0644,
4135                         .proc_handler   =       proc_dointvec,
4136                 },
4137                 {
4138                         .ctl_name       =       NET_IPV6_AUTOCONF,
4139                         .procname       =       "autoconf",
4140                         .data           =       &ipv6_devconf.autoconf,
4141                         .maxlen         =       sizeof(int),
4142                         .mode           =       0644,
4143                         .proc_handler   =       proc_dointvec,
4144                 },
4145                 {
4146                         .ctl_name       =       NET_IPV6_DAD_TRANSMITS,
4147                         .procname       =       "dad_transmits",
4148                         .data           =       &ipv6_devconf.dad_transmits,
4149                         .maxlen         =       sizeof(int),
4150                         .mode           =       0644,
4151                         .proc_handler   =       proc_dointvec,
4152                 },
4153                 {
4154                         .ctl_name       =       NET_IPV6_RTR_SOLICITS,
4155                         .procname       =       "router_solicitations",
4156                         .data           =       &ipv6_devconf.rtr_solicits,
4157                         .maxlen         =       sizeof(int),
4158                         .mode           =       0644,
4159                         .proc_handler   =       proc_dointvec,
4160                 },
4161                 {
4162                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_INTERVAL,
4163                         .procname       =       "router_solicitation_interval",
4164                         .data           =       &ipv6_devconf.rtr_solicit_interval,
4165                         .maxlen         =       sizeof(int),
4166                         .mode           =       0644,
4167                         .proc_handler   =       proc_dointvec_jiffies,
4168                         .strategy       =       sysctl_jiffies,
4169                 },
4170                 {
4171                         .ctl_name       =       NET_IPV6_RTR_SOLICIT_DELAY,
4172                         .procname       =       "router_solicitation_delay",
4173                         .data           =       &ipv6_devconf.rtr_solicit_delay,
4174                         .maxlen         =       sizeof(int),
4175                         .mode           =       0644,
4176                         .proc_handler   =       proc_dointvec_jiffies,
4177                         .strategy       =       sysctl_jiffies,
4178                 },
4179                 {
4180                         .ctl_name       =       NET_IPV6_FORCE_MLD_VERSION,
4181                         .procname       =       "force_mld_version",
4182                         .data           =       &ipv6_devconf.force_mld_version,
4183                         .maxlen         =       sizeof(int),
4184                         .mode           =       0644,
4185                         .proc_handler   =       proc_dointvec,
4186                 },
4187 #ifdef CONFIG_IPV6_PRIVACY
4188                 {
4189                         .ctl_name       =       NET_IPV6_USE_TEMPADDR,
4190                         .procname       =       "use_tempaddr",
4191                         .data           =       &ipv6_devconf.use_tempaddr,
4192                         .maxlen         =       sizeof(int),
4193                         .mode           =       0644,
4194                         .proc_handler   =       proc_dointvec,
4195                 },
4196                 {
4197                         .ctl_name       =       NET_IPV6_TEMP_VALID_LFT,
4198                         .procname       =       "temp_valid_lft",
4199                         .data           =       &ipv6_devconf.temp_valid_lft,
4200                         .maxlen         =       sizeof(int),
4201                         .mode           =       0644,
4202                         .proc_handler   =       proc_dointvec,
4203                 },
4204                 {
4205                         .ctl_name       =       NET_IPV6_TEMP_PREFERED_LFT,
4206                         .procname       =       "temp_prefered_lft",
4207                         .data           =       &ipv6_devconf.temp_prefered_lft,
4208                         .maxlen         =       sizeof(int),
4209                         .mode           =       0644,
4210                         .proc_handler   =       proc_dointvec,
4211                 },
4212                 {
4213                         .ctl_name       =       NET_IPV6_REGEN_MAX_RETRY,
4214                         .procname       =       "regen_max_retry",
4215                         .data           =       &ipv6_devconf.regen_max_retry,
4216                         .maxlen         =       sizeof(int),
4217                         .mode           =       0644,
4218                         .proc_handler   =       proc_dointvec,
4219                 },
4220                 {
4221                         .ctl_name       =       NET_IPV6_MAX_DESYNC_FACTOR,
4222                         .procname       =       "max_desync_factor",
4223                         .data           =       &ipv6_devconf.max_desync_factor,
4224                         .maxlen         =       sizeof(int),
4225                         .mode           =       0644,
4226                         .proc_handler   =       proc_dointvec,
4227                 },
4228 #endif
4229                 {
4230                         .ctl_name       =       NET_IPV6_MAX_ADDRESSES,
4231                         .procname       =       "max_addresses",
4232                         .data           =       &ipv6_devconf.max_addresses,
4233                         .maxlen         =       sizeof(int),
4234                         .mode           =       0644,
4235                         .proc_handler   =       proc_dointvec,
4236                 },
4237                 {
4238                         .ctl_name       =       NET_IPV6_ACCEPT_RA_DEFRTR,
4239                         .procname       =       "accept_ra_defrtr",
4240                         .data           =       &ipv6_devconf.accept_ra_defrtr,
4241                         .maxlen         =       sizeof(int),
4242                         .mode           =       0644,
4243                         .proc_handler   =       proc_dointvec,
4244                 },
4245                 {
4246                         .ctl_name       =       NET_IPV6_ACCEPT_RA_PINFO,
4247                         .procname       =       "accept_ra_pinfo",
4248                         .data           =       &ipv6_devconf.accept_ra_pinfo,
4249                         .maxlen         =       sizeof(int),
4250                         .mode           =       0644,
4251                         .proc_handler   =       proc_dointvec,
4252                 },
4253 #ifdef CONFIG_IPV6_ROUTER_PREF
4254                 {
4255                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RTR_PREF,
4256                         .procname       =       "accept_ra_rtr_pref",
4257                         .data           =       &ipv6_devconf.accept_ra_rtr_pref,
4258                         .maxlen         =       sizeof(int),
4259                         .mode           =       0644,
4260                         .proc_handler   =       proc_dointvec,
4261                 },
4262                 {
4263                         .ctl_name       =       NET_IPV6_RTR_PROBE_INTERVAL,
4264                         .procname       =       "router_probe_interval",
4265                         .data           =       &ipv6_devconf.rtr_probe_interval,
4266                         .maxlen         =       sizeof(int),
4267                         .mode           =       0644,
4268                         .proc_handler   =       proc_dointvec_jiffies,
4269                         .strategy       =       sysctl_jiffies,
4270                 },
4271 #ifdef CONFIG_IPV6_ROUTE_INFO
4272                 {
4273                         .ctl_name       =       NET_IPV6_ACCEPT_RA_RT_INFO_MAX_PLEN,
4274                         .procname       =       "accept_ra_rt_info_max_plen",
4275                         .data           =       &ipv6_devconf.accept_ra_rt_info_max_plen,
4276                         .maxlen         =       sizeof(int),
4277                         .mode           =       0644,
4278                         .proc_handler   =       proc_dointvec,
4279                 },
4280 #endif
4281 #endif
4282                 {
4283                         .ctl_name       =       NET_IPV6_PROXY_NDP,
4284                         .procname       =       "proxy_ndp",
4285                         .data           =       &ipv6_devconf.proxy_ndp,
4286                         .maxlen         =       sizeof(int),
4287                         .mode           =       0644,
4288                         .proc_handler   =       proc_dointvec,
4289                 },
4290                 {
4291                         .ctl_name       =       NET_IPV6_ACCEPT_SOURCE_ROUTE,
4292                         .procname       =       "accept_source_route",
4293                         .data           =       &ipv6_devconf.accept_source_route,
4294                         .maxlen         =       sizeof(int),
4295                         .mode           =       0644,
4296                         .proc_handler   =       proc_dointvec,
4297                 },
4298 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
4299                 {
4300                         .ctl_name       =       CTL_UNNUMBERED,
4301                         .procname       =       "optimistic_dad",
4302                         .data           =       &ipv6_devconf.optimistic_dad,
4303                         .maxlen         =       sizeof(int),
4304                         .mode           =       0644,
4305                         .proc_handler   =       proc_dointvec,
4306
4307                 },
4308 #endif
4309 #ifdef CONFIG_IPV6_MROUTE
4310                 {
4311                         .ctl_name       =       CTL_UNNUMBERED,
4312                         .procname       =       "mc_forwarding",
4313                         .data           =       &ipv6_devconf.mc_forwarding,
4314                         .maxlen         =       sizeof(int),
4315                         .mode           =       0444,
4316                         .proc_handler   =       proc_dointvec,
4317                 },
4318 #endif
4319                 {
4320                         .ctl_name       =       CTL_UNNUMBERED,
4321                         .procname       =       "disable_ipv6",
4322                         .data           =       &ipv6_devconf.disable_ipv6,
4323                         .maxlen         =       sizeof(int),
4324                         .mode           =       0644,
4325                         .proc_handler   =       addrconf_sysctl_disable,
4326                         .strategy       =       sysctl_intvec,
4327                 },
4328                 {
4329                         .ctl_name       =       CTL_UNNUMBERED,
4330                         .procname       =       "accept_dad",
4331                         .data           =       &ipv6_devconf.accept_dad,
4332                         .maxlen         =       sizeof(int),
4333                         .mode           =       0644,
4334                         .proc_handler   =       proc_dointvec,
4335                 },
4336                 {
4337                         .ctl_name       =       0,      /* sentinel */
4338                 }
4339         },
4340 };
4341
4342 static int __addrconf_sysctl_register(struct net *net, char *dev_name,
4343                 int ctl_name, struct inet6_dev *idev, struct ipv6_devconf *p)
4344 {
4345         int i;
4346         struct addrconf_sysctl_table *t;
4347
4348 #define ADDRCONF_CTL_PATH_DEV   3
4349
4350         struct ctl_path addrconf_ctl_path[] = {
4351                 { .procname = "net", .ctl_name = CTL_NET, },
4352                 { .procname = "ipv6", .ctl_name = NET_IPV6, },
4353                 { .procname = "conf", .ctl_name = NET_IPV6_CONF, },
4354                 { /* to be set */ },
4355                 { },
4356         };
4357
4358
4359         t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL);
4360         if (t == NULL)
4361                 goto out;
4362
4363         for (i=0; t->addrconf_vars[i].data; i++) {
4364                 t->addrconf_vars[i].data += (char*)p - (char*)&ipv6_devconf;
4365                 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */
4366                 t->addrconf_vars[i].extra2 = net;
4367         }
4368
4369         /*
4370          * Make a copy of dev_name, because '.procname' is regarded as const
4371          * by sysctl and we wouldn't want anyone to change it under our feet
4372          * (see SIOCSIFNAME).
4373          */
4374         t->dev_name = kstrdup(dev_name, GFP_KERNEL);
4375         if (!t->dev_name)
4376                 goto free;
4377
4378         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].procname = t->dev_name;
4379         addrconf_ctl_path[ADDRCONF_CTL_PATH_DEV].ctl_name = ctl_name;
4380
4381         t->sysctl_header = register_net_sysctl_table(net, addrconf_ctl_path,
4382                         t->addrconf_vars);
4383         if (t->sysctl_header == NULL)
4384                 goto free_procname;
4385
4386         p->sysctl = t;
4387         return 0;
4388
4389 free_procname:
4390         kfree(t->dev_name);
4391 free:
4392         kfree(t);
4393 out:
4394         return -ENOBUFS;
4395 }
4396
4397 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p)
4398 {
4399         struct addrconf_sysctl_table *t;
4400
4401         if (p->sysctl == NULL)
4402                 return;
4403
4404         t = p->sysctl;
4405         p->sysctl = NULL;
4406         unregister_sysctl_table(t->sysctl_header);
4407         kfree(t->dev_name);
4408         kfree(t);
4409 }
4410
4411 static void addrconf_sysctl_register(struct inet6_dev *idev)
4412 {
4413         neigh_sysctl_register(idev->dev, idev->nd_parms, NET_IPV6,
4414                               NET_IPV6_NEIGH, "ipv6",
4415                               &ndisc_ifinfo_sysctl_change,
4416                               ndisc_ifinfo_sysctl_strategy);
4417         __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name,
4418                         idev->dev->ifindex, idev, &idev->cnf);
4419 }
4420
4421 static void addrconf_sysctl_unregister(struct inet6_dev *idev)
4422 {
4423         __addrconf_sysctl_unregister(&idev->cnf);
4424         neigh_sysctl_unregister(idev->nd_parms);
4425 }
4426
4427
4428 #endif
4429
4430 static int addrconf_init_net(struct net *net)
4431 {
4432         int err;
4433         struct ipv6_devconf *all, *dflt;
4434
4435         err = -ENOMEM;
4436         all = &ipv6_devconf;
4437         dflt = &ipv6_devconf_dflt;
4438
4439         if (net != &init_net) {
4440                 all = kmemdup(all, sizeof(ipv6_devconf), GFP_KERNEL);
4441                 if (all == NULL)
4442                         goto err_alloc_all;
4443
4444                 dflt = kmemdup(dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL);
4445                 if (dflt == NULL)
4446                         goto err_alloc_dflt;
4447         } else {
4448                 /* these will be inherited by all namespaces */
4449                 dflt->autoconf = ipv6_defaults.autoconf;
4450                 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6;
4451         }
4452
4453         net->ipv6.devconf_all = all;
4454         net->ipv6.devconf_dflt = dflt;
4455
4456 #ifdef CONFIG_SYSCTL
4457         err = __addrconf_sysctl_register(net, "all", NET_PROTO_CONF_ALL,
4458                         NULL, all);
4459         if (err < 0)
4460                 goto err_reg_all;
4461
4462         err = __addrconf_sysctl_register(net, "default", NET_PROTO_CONF_DEFAULT,
4463                         NULL, dflt);
4464         if (err < 0)
4465                 goto err_reg_dflt;
4466 #endif
4467         return 0;
4468
4469 #ifdef CONFIG_SYSCTL
4470 err_reg_dflt:
4471         __addrconf_sysctl_unregister(all);
4472 err_reg_all:
4473         kfree(dflt);
4474 #endif
4475 err_alloc_dflt:
4476         kfree(all);
4477 err_alloc_all:
4478         return err;
4479 }
4480
4481 static void addrconf_exit_net(struct net *net)
4482 {
4483 #ifdef CONFIG_SYSCTL
4484         __addrconf_sysctl_unregister(net->ipv6.devconf_dflt);
4485         __addrconf_sysctl_unregister(net->ipv6.devconf_all);
4486 #endif
4487         if (net != &init_net) {
4488                 kfree(net->ipv6.devconf_dflt);
4489                 kfree(net->ipv6.devconf_all);
4490         }
4491 }
4492
4493 static struct pernet_operations addrconf_ops = {
4494         .init = addrconf_init_net,
4495         .exit = addrconf_exit_net,
4496 };
4497
4498 /*
4499  *      Device notifier
4500  */
4501
4502 int register_inet6addr_notifier(struct notifier_block *nb)
4503 {
4504         return atomic_notifier_chain_register(&inet6addr_chain, nb);
4505 }
4506
4507 EXPORT_SYMBOL(register_inet6addr_notifier);
4508
4509 int unregister_inet6addr_notifier(struct notifier_block *nb)
4510 {
4511         return atomic_notifier_chain_unregister(&inet6addr_chain,nb);
4512 }
4513
4514 EXPORT_SYMBOL(unregister_inet6addr_notifier);
4515
4516 /*
4517  *      Init / cleanup code
4518  */
4519
4520 int __init addrconf_init(void)
4521 {
4522         int err;
4523
4524         if ((err = ipv6_addr_label_init()) < 0) {
4525                 printk(KERN_CRIT "IPv6 Addrconf: cannot initialize default policy table: %d.\n",
4526                         err);
4527                 return err;
4528         }
4529
4530         register_pernet_subsys(&addrconf_ops);
4531
4532         /* The addrconf netdev notifier requires that loopback_dev
4533          * has it's ipv6 private information allocated and setup
4534          * before it can bring up and give link-local addresses
4535          * to other devices which are up.
4536          *
4537          * Unfortunately, loopback_dev is not necessarily the first
4538          * entry in the global dev_base list of net devices.  In fact,
4539          * it is likely to be the very last entry on that list.
4540          * So this causes the notifier registry below to try and
4541          * give link-local addresses to all devices besides loopback_dev
4542          * first, then loopback_dev, which cases all the non-loopback_dev
4543          * devices to fail to get a link-local address.
4544          *
4545          * So, as a temporary fix, allocate the ipv6 structure for
4546          * loopback_dev first by hand.
4547          * Longer term, all of the dependencies ipv6 has upon the loopback
4548          * device and it being up should be removed.
4549          */
4550         rtnl_lock();
4551         if (!ipv6_add_dev(init_net.loopback_dev))
4552                 err = -ENOMEM;
4553         rtnl_unlock();
4554         if (err)
4555                 goto errlo;
4556
4557         register_netdevice_notifier(&ipv6_dev_notf);
4558
4559         addrconf_verify(0);
4560
4561         err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo);
4562         if (err < 0)
4563                 goto errout;
4564
4565         /* Only the first call to __rtnl_register can fail */
4566         __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL);
4567         __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL);
4568         __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, inet6_dump_ifaddr);
4569         __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, inet6_dump_ifmcaddr);
4570         __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, inet6_dump_ifacaddr);
4571
4572         ipv6_addr_label_rtnl_register();
4573
4574         return 0;
4575 errout:
4576         unregister_netdevice_notifier(&ipv6_dev_notf);
4577 errlo:
4578         unregister_pernet_subsys(&addrconf_ops);
4579
4580         return err;
4581 }
4582
4583 void addrconf_cleanup(void)
4584 {
4585         struct inet6_ifaddr *ifa;
4586         struct net_device *dev;
4587         int i;
4588
4589         unregister_netdevice_notifier(&ipv6_dev_notf);
4590         unregister_pernet_subsys(&addrconf_ops);
4591
4592         rtnl_lock();
4593
4594         /* clean dev list */
4595         for_each_netdev(&init_net, dev) {
4596                 if (__in6_dev_get(dev) == NULL)
4597                         continue;
4598                 addrconf_ifdown(dev, 1);
4599         }
4600         addrconf_ifdown(init_net.loopback_dev, 2);
4601
4602         /*
4603          *      Check hash table.
4604          */
4605         write_lock_bh(&addrconf_hash_lock);
4606         for (i=0; i < IN6_ADDR_HSIZE; i++) {
4607                 for (ifa=inet6_addr_lst[i]; ifa; ) {
4608                         struct inet6_ifaddr *bifa;
4609
4610                         bifa = ifa;
4611                         ifa = ifa->lst_next;
4612                         printk(KERN_DEBUG "bug: IPv6 address leakage detected: ifa=%p\n", bifa);
4613                         /* Do not free it; something is wrong.
4614                            Now we can investigate it with debugger.
4615                          */
4616                 }
4617         }
4618         write_unlock_bh(&addrconf_hash_lock);
4619
4620         del_timer(&addr_chk_timer);
4621         rtnl_unlock();
4622 }