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 #include <linux/slab.h>
  57 #ifdef CONFIG_SYSCTL
  58 #include <linux/sysctl.h>
  59 #endif
  60 #include <linux/capability.h>
  61 #include <linux/delay.h>
  62 #include <linux/notifier.h>
  63 #include <linux/string.h>
  64
  65 #include <net/net_namespace.h>
  66 #include <net/sock.h>
  67 #include <net/snmp.h>
  68
  69 #include <net/ipv6.h>
  70 #include <net/protocol.h>
  71 #include <net/ndisc.h>
  72 #include <net/ip6_route.h>
  73 #include <net/addrconf.h>
  74 #include <net/tcp.h>
  75 #include <net/ip.h>
  76 #include <net/netlink.h>
  77 #include <net/pkt_sched.h>
  78 #include <linux/if_tunnel.h>
  79 #include <linux/rtnetlink.h>
  80
  81 #ifdef CONFIG_IPV6_PRIVACY
  82 #include <linux/random.h>
  83 #endif
  84
  85 #include <linux/uaccess.h>
  86 #include <asm/unaligned.h>
  87
  88 #include <linux/proc_fs.h>
  89 #include <linux/seq_file.h>
  90 #include <linux/export.h>
  91
  92 /* Set to 3 to get tracing... */
  93 #define ACONF_DEBUG 2
  94
  95 #if ACONF_DEBUG >= 3
  96 #define ADBG(x) printk x
  97 #else
  98 #define ADBG(x)
  99 #endif
 100
 101 #define INFINITY_LIFE_TIME      0xFFFFFFFF
 102
 103 static inline u32 cstamp_delta(unsigned long cstamp)
 104 {
 105         return (cstamp - INITIAL_JIFFIES) * 100UL / HZ;
 106 }
 107
 108 #define ADDRCONF_TIMER_FUZZ_MINUS       (HZ > 50 ? HZ/50 : 1)
 109 #define ADDRCONF_TIMER_FUZZ             (HZ / 4)
 110 #define ADDRCONF_TIMER_FUZZ_MAX         (HZ)
 111
 112 #ifdef CONFIG_SYSCTL
 113 static void addrconf_sysctl_register(struct inet6_dev *idev);
 114 static void addrconf_sysctl_unregister(struct inet6_dev *idev);
 115 #else
 116 static inline void addrconf_sysctl_register(struct inet6_dev *idev)
 117 {
 118 }
 119
 120 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev)
 121 {
 122 }
 123 #endif
 124
 125 #ifdef CONFIG_IPV6_PRIVACY
 126 static int __ipv6_regen_rndid(struct inet6_dev *idev);
 127 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr);
 128 static void ipv6_regen_rndid(unsigned long data);
 129 #endif
 130
 131 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev);
 132 static int ipv6_count_addresses(struct inet6_dev *idev);
 133
 134 /*
 135  *      Configured unicast address hash table
 136  */
 137 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE];
 138 static DEFINE_SPINLOCK(addrconf_hash_lock);
 139
 140 static void addrconf_verify(unsigned long);
 141
 142 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0);
 143 static DEFINE_SPINLOCK(addrconf_verify_lock);
 144
 145 static void addrconf_join_anycast(struct inet6_ifaddr *ifp);
 146 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp);
 147
 148 static void addrconf_type_change(struct net_device *dev,
 149                                  unsigned long event);
 150 static int addrconf_ifdown(struct net_device *dev, int how);
 151
 152 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags);
 153 static void addrconf_dad_timer(unsigned long data);
 154 static void addrconf_dad_completed(struct inet6_ifaddr *ifp);
 155 static void addrconf_dad_run(struct inet6_dev *idev);
 156 static void addrconf_rs_timer(unsigned long data);
 157 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 158 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa);
 159
 160 static void inet6_prefix_notify(int event, struct inet6_dev *idev,
 161                                 struct prefix_info *pinfo);
 162 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
 163                                struct net_device *dev);
 164
 165 static ATOMIC_NOTIFIER_HEAD(inet6addr_chain);
 166
 167 static struct ipv6_devconf ipv6_devconf __read_mostly = {
 168         .forwarding             = 0,
 169         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 170         .mtu6                   = IPV6_MIN_MTU,
 171         .accept_ra              = 1,
 172         .accept_redirects       = 1,
 173         .autoconf               = 1,
 174         .force_mld_version      = 0,
 175         .dad_transmits          = 1,
 176         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 177         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 178         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 179 #ifdef CONFIG_IPV6_PRIVACY
 180         .use_tempaddr           = 0,
 181         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 182         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 183         .regen_max_retry        = REGEN_MAX_RETRY,
 184         .max_desync_factor      = MAX_DESYNC_FACTOR,
 185 #endif
 186         .max_addresses          = IPV6_MAX_ADDRESSES,
 187         .accept_ra_defrtr       = 1,
 188         .accept_ra_pinfo        = 1,
 189 #ifdef CONFIG_IPV6_ROUTER_PREF
 190         .accept_ra_rtr_pref     = 1,
 191         .rtr_probe_interval     = 60 * HZ,
 192 #ifdef CONFIG_IPV6_ROUTE_INFO
 193         .accept_ra_rt_info_max_plen = 0,
 194 #endif
 195 #endif
 196         .proxy_ndp              = 0,
 197         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 198         .disable_ipv6           = 0,
 199         .accept_dad             = 1,
 200 };
 201
 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = {
 203         .forwarding             = 0,
 204         .hop_limit              = IPV6_DEFAULT_HOPLIMIT,
 205         .mtu6                   = IPV6_MIN_MTU,
 206         .accept_ra              = 1,
 207         .accept_redirects       = 1,
 208         .autoconf               = 1,
 209         .dad_transmits          = 1,
 210         .rtr_solicits           = MAX_RTR_SOLICITATIONS,
 211         .rtr_solicit_interval   = RTR_SOLICITATION_INTERVAL,
 212         .rtr_solicit_delay      = MAX_RTR_SOLICITATION_DELAY,
 213 #ifdef CONFIG_IPV6_PRIVACY
 214         .use_tempaddr           = 0,
 215         .temp_valid_lft         = TEMP_VALID_LIFETIME,
 216         .temp_prefered_lft      = TEMP_PREFERRED_LIFETIME,
 217         .regen_max_retry        = REGEN_MAX_RETRY,
 218         .max_desync_factor      = MAX_DESYNC_FACTOR,
 219 #endif
 220         .max_addresses          = IPV6_MAX_ADDRESSES,
 221         .accept_ra_defrtr       = 1,
 222         .accept_ra_pinfo        = 1,
 223 #ifdef CONFIG_IPV6_ROUTER_PREF
 224         .accept_ra_rtr_pref     = 1,
 225         .rtr_probe_interval     = 60 * HZ,
 226 #ifdef CONFIG_IPV6_ROUTE_INFO
 227         .accept_ra_rt_info_max_plen = 0,
 228 #endif
 229 #endif
 230         .proxy_ndp              = 0,
 231         .accept_source_route    = 0,    /* we do not accept RH0 by default. */
 232         .disable_ipv6           = 0,
 233         .accept_dad             = 1,
 234 };
 235
 236 /* IPv6 Wildcard Address and Loopback Address defined by RFC2553 */
 237 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 238 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 239 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 240 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
 241
 242 /* Check if a valid qdisc is available */
 243 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
 244 {
 245         return !qdisc_tx_is_noop(dev);
 246 }
 247
 248 /* Check if a route is valid prefix route */
 249 static inline int addrconf_is_prefix_route(const struct rt6_info *rt)
 250 {
 251         return (rt->rt6i_flags & (RTF_GATEWAY | RTF_DEFAULT)) == 0;
 252 }
 253
 254 static void addrconf_del_timer(struct inet6_ifaddr *ifp)
 255 {
 256         if (del_timer(&ifp->timer))
 257                 __in6_ifa_put(ifp);
 258 }
 259
 260 enum addrconf_timer_t {
 261         AC_NONE,
 262         AC_DAD,
 263         AC_RS,
 264 };
 265
 266 static void addrconf_mod_timer(struct inet6_ifaddr *ifp,
 267                                enum addrconf_timer_t what,
 268                                unsigned long when)
 269 {
 270         if (!del_timer(&ifp->timer))
 271                 in6_ifa_hold(ifp);
 272
 273         switch (what) {
 274         case AC_DAD:
 275                 ifp->timer.function = addrconf_dad_timer;
 276                 break;
 277         case AC_RS:
 278                 ifp->timer.function = addrconf_rs_timer;
 279                 break;
 280         default:
 281                 break;
 282         }
 283         ifp->timer.expires = jiffies + when;
 284         add_timer(&ifp->timer);
 285 }
 286
 287 static int snmp6_alloc_dev(struct inet6_dev *idev)
 288 {
 289         if (snmp_mib_init((void __percpu **)idev->stats.ipv6,
 290                           sizeof(struct ipstats_mib),
 291                           __alignof__(struct ipstats_mib)) < 0)
 292                 goto err_ip;
 293         idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device),
 294                                         GFP_KERNEL);
 295         if (!idev->stats.icmpv6dev)
 296                 goto err_icmp;
 297         idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device),
 298                                            GFP_KERNEL);
 299         if (!idev->stats.icmpv6msgdev)
 300                 goto err_icmpmsg;
 301
 302         return 0;
 303
 304 err_icmpmsg:
 305         kfree(idev->stats.icmpv6dev);
 306 err_icmp:
 307         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 308 err_ip:
 309         return -ENOMEM;
 310 }
 311
 312 static void snmp6_free_dev(struct inet6_dev *idev)
 313 {
 314         kfree(idev->stats.icmpv6msgdev);
 315         kfree(idev->stats.icmpv6dev);
 316         snmp_mib_free((void __percpu **)idev->stats.ipv6);
 317 }
 318
 319 /* Nobody refers to this device, we may destroy it. */
 320
 321 void in6_dev_finish_destroy(struct inet6_dev *idev)
 322 {
 323         struct net_device *dev = idev->dev;
 324
 325         WARN_ON(!list_empty(&idev->addr_list));
 326         WARN_ON(idev->mc_list != NULL);
 327
 328 #ifdef NET_REFCNT_DEBUG
 329         printk(KERN_DEBUG "in6_dev_finish_destroy: %s\n", dev ? dev->name : "NIL");
 330 #endif
 331         dev_put(dev);
 332         if (!idev->dead) {
 333                 pr_warning("Freeing alive inet6 device %p\n", idev);
 334                 return;
 335         }
 336         snmp6_free_dev(idev);
 337         kfree_rcu(idev, rcu);
 338 }
 339
 340 EXPORT_SYMBOL(in6_dev_finish_destroy);
 341
 342 static struct inet6_dev * ipv6_add_dev(struct net_device *dev)
 343 {
 344         struct inet6_dev *ndev;
 345
 346         ASSERT_RTNL();
 347
 348         if (dev->mtu < IPV6_MIN_MTU)
 349                 return NULL;
 350
 351         ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
 352
 353         if (ndev == NULL)
 354                 return NULL;
 355
 356         rwlock_init(&ndev->lock);
 357         ndev->dev = dev;
 358         INIT_LIST_HEAD(&ndev->addr_list);
 359
 360         memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf));
 361         ndev->cnf.mtu6 = dev->mtu;
 362         ndev->cnf.sysctl = NULL;
 363         ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl);
 364         if (ndev->nd_parms == NULL) {
 365                 kfree(ndev);
 366                 return NULL;
 367         }
 368         if (ndev->cnf.forwarding)
 369                 dev_disable_lro(dev);
 370         /* We refer to the device */
 371         dev_hold(dev);
 372
 373         if (snmp6_alloc_dev(ndev) < 0) {
 374                 ADBG((KERN_WARNING
 375                         "%s(): cannot allocate memory for statistics; dev=%s.\n",
 376                         __func__, dev->name));
 377                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 378                 dev_put(dev);
 379                 kfree(ndev);
 380                 return NULL;
 381         }
 382
 383         if (snmp6_register_dev(ndev) < 0) {
 384                 ADBG((KERN_WARNING
 385                         "%s(): cannot create /proc/net/dev_snmp6/%s\n",
 386                         __func__, dev->name));
 387                 neigh_parms_release(&nd_tbl, ndev->nd_parms);
 388                 ndev->dead = 1;
 389                 in6_dev_finish_destroy(ndev);
 390                 return NULL;
 391         }
 392
 393         /* One reference from device.  We must do this before
 394          * we invoke __ipv6_regen_rndid().
 395          */
 396         in6_dev_hold(ndev);
 397
 398         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 399                 ndev->cnf.accept_dad = -1;
 400
 401 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
 402         if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) {
 403                 printk(KERN_INFO
 404                        "%s: Disabled Multicast RS\n",
 405                        dev->name);
 406                 ndev->cnf.rtr_solicits = 0;
 407         }
 408 #endif
 409
 410 #ifdef CONFIG_IPV6_PRIVACY
 411         INIT_LIST_HEAD(&ndev->tempaddr_list);
 412         setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev);
 413         if ((dev->flags&IFF_LOOPBACK) ||
 414             dev->type == ARPHRD_TUNNEL ||
 415             dev->type == ARPHRD_TUNNEL6 ||
 416             dev->type == ARPHRD_SIT ||
 417             dev->type == ARPHRD_NONE) {
 418                 ndev->cnf.use_tempaddr = -1;
 419         } else {
 420                 in6_dev_hold(ndev);
 421                 ipv6_regen_rndid((unsigned long) ndev);
 422         }
 423 #endif
 424
 425         if (netif_running(dev) && addrconf_qdisc_ok(dev))
 426                 ndev->if_flags |= IF_READY;
 427
 428         ipv6_mc_init_dev(ndev);
 429         ndev->tstamp = jiffies;
 430         addrconf_sysctl_register(ndev);
 431         /* protected by rtnl_lock */
 432         rcu_assign_pointer(dev->ip6_ptr, ndev);
 433
 434         /* Join all-node multicast group */
 435         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes);
 436
 437         /* Join all-router multicast group if forwarding is set */
 438         if (ndev->cnf.forwarding && dev && (dev->flags & IFF_MULTICAST))
 439                 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 440
 441         return ndev;
 442 }
 443
 444 static struct inet6_dev * ipv6_find_idev(struct net_device *dev)
 445 {
 446         struct inet6_dev *idev;
 447
 448         ASSERT_RTNL();
 449
 450         idev = __in6_dev_get(dev);
 451         if (!idev) {
 452                 idev = ipv6_add_dev(dev);
 453                 if (!idev)
 454                         return NULL;
 455         }
 456
 457         if (dev->flags&IFF_UP)
 458                 ipv6_mc_up(idev);
 459         return idev;
 460 }
 461
 462 #ifdef CONFIG_SYSCTL
 463 static void dev_forward_change(struct inet6_dev *idev)
 464 {
 465         struct net_device *dev;
 466         struct inet6_ifaddr *ifa;
 467
 468         if (!idev)
 469                 return;
 470         dev = idev->dev;
 471         if (idev->cnf.forwarding)
 472                 dev_disable_lro(dev);
 473         if (dev && (dev->flags & IFF_MULTICAST)) {
 474                 if (idev->cnf.forwarding)
 475                         ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
 476                 else
 477                         ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
 478         }
 479
 480         list_for_each_entry(ifa, &idev->addr_list, if_list) {
 481                 if (ifa->flags&IFA_F_TENTATIVE)
 482                         continue;
 483                 if (idev->cnf.forwarding)
 484                         addrconf_join_anycast(ifa);
 485                 else
 486                         addrconf_leave_anycast(ifa);
 487         }
 488 }
 489
 490
 491 static void addrconf_forward_change(struct net *net, __s32 newf)
 492 {
 493         struct net_device *dev;
 494         struct inet6_dev *idev;
 495
 496         for_each_netdev(net, dev) {
 497                 idev = __in6_dev_get(dev);
 498                 if (idev) {
 499                         int changed = (!idev->cnf.forwarding) ^ (!newf);
 500                         idev->cnf.forwarding = newf;
 501                         if (changed)
 502                                 dev_forward_change(idev);
 503                 }
 504         }
 505 }
 506
 507 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int old)
 508 {
 509         struct net *net;
 510
 511         net = (struct net *)table->extra2;
 512         if (p == &net->ipv6.devconf_dflt->forwarding)
 513                 return 0;
 514
 515         if (!rtnl_trylock()) {
 516                 /* Restore the original values before restarting */
 517                 *p = old;
 518                 return restart_syscall();
 519         }
 520
 521         if (p == &net->ipv6.devconf_all->forwarding) {
 522                 __s32 newf = net->ipv6.devconf_all->forwarding;
 523                 net->ipv6.devconf_dflt->forwarding = newf;
 524                 addrconf_forward_change(net, newf);
 525         } else if ((!*p) ^ (!old))
 526                 dev_forward_change((struct inet6_dev *)table->extra1);
 527         rtnl_unlock();
 528
 529         if (*p)
 530                 rt6_purge_dflt_routers(net);
 531         return 1;
 532 }
 533 #endif
 534
 535 /* Nobody refers to this ifaddr, destroy it */
 536 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp)
 537 {
 538         WARN_ON(!hlist_unhashed(&ifp->addr_lst));
 539
 540 #ifdef NET_REFCNT_DEBUG
 541         printk(KERN_DEBUG "inet6_ifa_finish_destroy\n");
 542 #endif
 543
 544         in6_dev_put(ifp->idev);
 545
 546         if (del_timer(&ifp->timer))
 547                 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp);
 548
 549         if (ifp->state != INET6_IFADDR_STATE_DEAD) {
 550                 pr_warning("Freeing alive inet6 address %p\n", ifp);
 551                 return;
 552         }
 553         dst_release(&ifp->rt->dst);
 554
 555         kfree_rcu(ifp, rcu);
 556 }
 557
 558 static void
 559 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp)
 560 {
 561         struct list_head *p;
 562         int ifp_scope = ipv6_addr_src_scope(&ifp->addr);
 563
 564         /*
 565          * Each device address list is sorted in order of scope -
 566          * global before linklocal.
 567          */
 568         list_for_each(p, &idev->addr_list) {
 569                 struct inet6_ifaddr *ifa
 570                         = list_entry(p, struct inet6_ifaddr, if_list);
 571                 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr))
 572                         break;
 573         }
 574
 575         list_add_tail(&ifp->if_list, p);
 576 }
 577
 578 static u32 ipv6_addr_hash(const struct in6_addr *addr)
 579 {
 580         /*
 581          * We perform the hash function over the last 64 bits of the address
 582          * This will include the IEEE address token on links that support it.
 583          */
 584         return jhash_2words((__force u32)addr->s6_addr32[2],
 585                             (__force u32)addr->s6_addr32[3], 0)
 586                 & (IN6_ADDR_HSIZE - 1);
 587 }
 588
 589 /* On success it returns ifp with increased reference count */
 590
 591 static struct inet6_ifaddr *
 592 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, int pfxlen,
 593               int scope, u32 flags)
 594 {
 595         struct inet6_ifaddr *ifa = NULL;
 596         struct rt6_info *rt;
 597         unsigned int hash;
 598         int err = 0;
 599         int addr_type = ipv6_addr_type(addr);
 600
 601         if (addr_type == IPV6_ADDR_ANY ||
 602             addr_type & IPV6_ADDR_MULTICAST ||
 603             (!(idev->dev->flags & IFF_LOOPBACK) &&
 604              addr_type & IPV6_ADDR_LOOPBACK))
 605                 return ERR_PTR(-EADDRNOTAVAIL);
 606
 607         rcu_read_lock_bh();
 608         if (idev->dead) {
 609                 err = -ENODEV;                  /*XXX*/
 610                 goto out2;
 611         }
 612
 613         if (idev->cnf.disable_ipv6) {
 614                 err = -EACCES;
 615                 goto out2;
 616         }
 617
 618         spin_lock(&addrconf_hash_lock);
 619
 620         /* Ignore adding duplicate addresses on an interface */
 621         if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) {
 622                 ADBG(("ipv6_add_addr: already assigned\n"));
 623                 err = -EEXIST;
 624                 goto out;
 625         }
 626
 627         ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC);
 628
 629         if (ifa == NULL) {
 630                 ADBG(("ipv6_add_addr: malloc failed\n"));
 631                 err = -ENOBUFS;
 632                 goto out;
 633         }
 634
 635         rt = addrconf_dst_alloc(idev, addr, 0);
 636         if (IS_ERR(rt)) {
 637                 err = PTR_ERR(rt);
 638                 goto out;
 639         }
 640
 641         ipv6_addr_copy(&ifa->addr, addr);
 642
 643         spin_lock_init(&ifa->lock);
 644         spin_lock_init(&ifa->state_lock);
 645         init_timer(&ifa->timer);
 646         INIT_HLIST_NODE(&ifa->addr_lst);
 647         ifa->timer.data = (unsigned long) ifa;
 648         ifa->scope = scope;
 649         ifa->prefix_len = pfxlen;
 650         ifa->flags = flags | IFA_F_TENTATIVE;
 651         ifa->cstamp = ifa->tstamp = jiffies;
 652
 653         ifa->rt = rt;
 654
 655         /*
 656          * part one of RFC 4429, section 3.3
 657          * We should not configure an address as
 658          * optimistic if we do not yet know the link
 659          * layer address of our nexhop router
 660          */
 661
 662         if (dst_get_neighbour_raw(&rt->dst) == NULL)
 663                 ifa->flags &= ~IFA_F_OPTIMISTIC;
 664
 665         ifa->idev = idev;
 666         in6_dev_hold(idev);
 667         /* For caller */
 668         in6_ifa_hold(ifa);
 669
 670         /* Add to big hash table */
 671         hash = ipv6_addr_hash(addr);
 672
 673         hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]);
 674         spin_unlock(&addrconf_hash_lock);
 675
 676         write_lock(&idev->lock);
 677         /* Add to inet6_dev unicast addr list. */
 678         ipv6_link_dev_addr(idev, ifa);
 679
 680 #ifdef CONFIG_IPV6_PRIVACY
 681         if (ifa->flags&IFA_F_TEMPORARY) {
 682                 list_add(&ifa->tmp_list, &idev->tempaddr_list);
 683                 in6_ifa_hold(ifa);
 684         }
 685 #endif
 686
 687         in6_ifa_hold(ifa);
 688         write_unlock(&idev->lock);
 689 out2:
 690         rcu_read_unlock_bh();
 691
 692         if (likely(err == 0))
 693                 atomic_notifier_call_chain(&inet6addr_chain, NETDEV_UP, ifa);
 694         else {
 695                 kfree(ifa);
 696                 ifa = ERR_PTR(err);
 697         }
 698
 699         return ifa;
 700 out:
 701         spin_unlock(&addrconf_hash_lock);
 702         goto out2;
 703 }
 704
 705 /* This function wants to get referenced ifp and releases it before return */
 706
 707 static void ipv6_del_addr(struct inet6_ifaddr *ifp)
 708 {
 709         struct inet6_ifaddr *ifa, *ifn;
 710         struct inet6_dev *idev = ifp->idev;
 711         int state;
 712         int deleted = 0, onlink = 0;
 713         unsigned long expires = jiffies;
 714
 715         spin_lock_bh(&ifp->state_lock);
 716         state = ifp->state;
 717         ifp->state = INET6_IFADDR_STATE_DEAD;
 718         spin_unlock_bh(&ifp->state_lock);
 719
 720         if (state == INET6_IFADDR_STATE_DEAD)
 721                 goto out;
 722
 723         spin_lock_bh(&addrconf_hash_lock);
 724         hlist_del_init_rcu(&ifp->addr_lst);
 725         spin_unlock_bh(&addrconf_hash_lock);
 726
 727         write_lock_bh(&idev->lock);
 728 #ifdef CONFIG_IPV6_PRIVACY
 729         if (ifp->flags&IFA_F_TEMPORARY) {
 730                 list_del(&ifp->tmp_list);
 731                 if (ifp->ifpub) {
 732                         in6_ifa_put(ifp->ifpub);
 733                         ifp->ifpub = NULL;
 734                 }
 735                 __in6_ifa_put(ifp);
 736         }
 737 #endif
 738
 739         list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) {
 740                 if (ifa == ifp) {
 741                         list_del_init(&ifp->if_list);
 742                         __in6_ifa_put(ifp);
 743
 744                         if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0)
 745                                 break;
 746                         deleted = 1;
 747                         continue;
 748                 } else if (ifp->flags & IFA_F_PERMANENT) {
 749                         if (ipv6_prefix_equal(&ifa->addr, &ifp->addr,
 750                                               ifp->prefix_len)) {
 751                                 if (ifa->flags & IFA_F_PERMANENT) {
 752                                         onlink = 1;
 753                                         if (deleted)
 754                                                 break;
 755                                 } else {
 756                                         unsigned long lifetime;
 757
 758                                         if (!onlink)
 759                                                 onlink = -1;
 760
 761                                         spin_lock(&ifa->lock);
 762
 763                                         lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ);
 764                                         /*
 765                                          * Note: Because this address is
 766                                          * not permanent, lifetime <
 767                                          * LONG_MAX / HZ here.
 768                                          */
 769                                         if (time_before(expires,
 770                                                         ifa->tstamp + lifetime * HZ))
 771                                                 expires = ifa->tstamp + lifetime * HZ;
 772                                         spin_unlock(&ifa->lock);
 773                                 }
 774                         }
 775                 }
 776         }
 777         write_unlock_bh(&idev->lock);
 778
 779         addrconf_del_timer(ifp);
 780
 781         ipv6_ifa_notify(RTM_DELADDR, ifp);
 782
 783         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifp);
 784
 785         /*
 786          * Purge or update corresponding prefix
 787          *
 788          * 1) we don't purge prefix here if address was not permanent.
 789          *    prefix is managed by its own lifetime.
 790          * 2) if there're no addresses, delete prefix.
 791          * 3) if there're still other permanent address(es),
 792          *    corresponding prefix is still permanent.
 793          * 4) otherwise, update prefix lifetime to the
 794          *    longest valid lifetime among the corresponding
 795          *    addresses on the device.
 796          *    Note: subsequent RA will update lifetime.
 797          *
 798          * --yoshfuji
 799          */
 800         if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) {
 801                 struct in6_addr prefix;
 802                 struct rt6_info *rt;
 803                 struct net *net = dev_net(ifp->idev->dev);
 804                 ipv6_addr_prefix(&prefix, &ifp->addr, ifp->prefix_len);
 805                 rt = rt6_lookup(net, &prefix, NULL, ifp->idev->dev->ifindex, 1);
 806
 807                 if (rt && addrconf_is_prefix_route(rt)) {
 808                         if (onlink == 0) {
 809                                 ip6_del_rt(rt);
 810                                 rt = NULL;
 811                         } else if (!(rt->rt6i_flags & RTF_EXPIRES)) {
 812                                 rt->rt6i_expires = expires;
 813                                 rt->rt6i_flags |= RTF_EXPIRES;
 814                         }
 815                 }
 816                 dst_release(&rt->dst);
 817         }
 818
 819         /* clean up prefsrc entries */
 820         rt6_remove_prefsrc(ifp);
 821 out:
 822         in6_ifa_put(ifp);
 823 }
 824
 825 #ifdef CONFIG_IPV6_PRIVACY
 826 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift)
 827 {
 828         struct inet6_dev *idev = ifp->idev;
 829         struct in6_addr addr, *tmpaddr;
 830         unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age;
 831         unsigned long regen_advance;
 832         int tmp_plen;
 833         int ret = 0;
 834         int max_addresses;
 835         u32 addr_flags;
 836         unsigned long now = jiffies;
 837
 838         write_lock(&idev->lock);
 839         if (ift) {
 840                 spin_lock_bh(&ift->lock);
 841                 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8);
 842                 spin_unlock_bh(&ift->lock);
 843                 tmpaddr = &addr;
 844         } else {
 845                 tmpaddr = NULL;
 846         }
 847 retry:
 848         in6_dev_hold(idev);
 849         if (idev->cnf.use_tempaddr <= 0) {
 850                 write_unlock(&idev->lock);
 851                 printk(KERN_INFO
 852                         "ipv6_create_tempaddr(): use_tempaddr is disabled.\n");
 853                 in6_dev_put(idev);
 854                 ret = -1;
 855                 goto out;
 856         }
 857         spin_lock_bh(&ifp->lock);
 858         if (ifp->regen_count++ >= idev->cnf.regen_max_retry) {
 859                 idev->cnf.use_tempaddr = -1;    /*XXX*/
 860                 spin_unlock_bh(&ifp->lock);
 861                 write_unlock(&idev->lock);
 862                 printk(KERN_WARNING
 863                         "ipv6_create_tempaddr(): regeneration time exceeded. disabled temporary address support.\n");
 864                 in6_dev_put(idev);
 865                 ret = -1;
 866                 goto out;
 867         }
 868         in6_ifa_hold(ifp);
 869         memcpy(addr.s6_addr, ifp->addr.s6_addr, 8);
 870         if (__ipv6_try_regen_rndid(idev, tmpaddr) < 0) {
 871                 spin_unlock_bh(&ifp->lock);
 872                 write_unlock(&idev->lock);
 873                 printk(KERN_WARNING
 874                         "ipv6_create_tempaddr(): regeneration of randomized interface id failed.\n");
 875                 in6_ifa_put(ifp);
 876                 in6_dev_put(idev);
 877                 ret = -1;
 878                 goto out;
 879         }
 880         memcpy(&addr.s6_addr[8], idev->rndid, 8);
 881         age = (now - ifp->tstamp) / HZ;
 882         tmp_valid_lft = min_t(__u32,
 883                               ifp->valid_lft,
 884                               idev->cnf.temp_valid_lft + age);
 885         tmp_prefered_lft = min_t(__u32,
 886                                  ifp->prefered_lft,
 887                                  idev->cnf.temp_prefered_lft + age -
 888                                  idev->cnf.max_desync_factor);
 889         tmp_plen = ifp->prefix_len;
 890         max_addresses = idev->cnf.max_addresses;
 891         tmp_tstamp = ifp->tstamp;
 892         spin_unlock_bh(&ifp->lock);
 893
 894         regen_advance = idev->cnf.regen_max_retry *
 895                         idev->cnf.dad_transmits *
 896                         idev->nd_parms->retrans_time / HZ;
 897         write_unlock(&idev->lock);
 898
 899         /* A temporary address is created only if this calculated Preferred
 900          * Lifetime is greater than REGEN_ADVANCE time units.  In particular,
 901          * an implementation must not create a temporary address with a zero
 902          * Preferred Lifetime.
 903          */
 904         if (tmp_prefered_lft <= regen_advance) {
 905                 in6_ifa_put(ifp);
 906                 in6_dev_put(idev);
 907                 ret = -1;
 908                 goto out;
 909         }
 910
 911         addr_flags = IFA_F_TEMPORARY;
 912         /* set in addrconf_prefix_rcv() */
 913         if (ifp->flags & IFA_F_OPTIMISTIC)
 914                 addr_flags |= IFA_F_OPTIMISTIC;
 915
 916         ift = !max_addresses ||
 917               ipv6_count_addresses(idev) < max_addresses ?
 918                 ipv6_add_addr(idev, &addr, tmp_plen,
 919                               ipv6_addr_type(&addr)&IPV6_ADDR_SCOPE_MASK,
 920                               addr_flags) : NULL;
 921         if (!ift || IS_ERR(ift)) {
 922                 in6_ifa_put(ifp);
 923                 in6_dev_put(idev);
 924                 printk(KERN_INFO
 925                         "ipv6_create_tempaddr(): retry temporary address regeneration.\n");
 926                 tmpaddr = &addr;
 927                 write_lock(&idev->lock);
 928                 goto retry;
 929         }
 930
 931         spin_lock_bh(&ift->lock);
 932         ift->ifpub = ifp;
 933         ift->valid_lft = tmp_valid_lft;
 934         ift->prefered_lft = tmp_prefered_lft;
 935         ift->cstamp = now;
 936         ift->tstamp = tmp_tstamp;
 937         spin_unlock_bh(&ift->lock);
 938
 939         addrconf_dad_start(ift, 0);
 940         in6_ifa_put(ift);
 941         in6_dev_put(idev);
 942 out:
 943         return ret;
 944 }
 945 #endif
 946
 947 /*
 948  *      Choose an appropriate source address (RFC3484)
 949  */
 950 enum {
 951         IPV6_SADDR_RULE_INIT = 0,
 952         IPV6_SADDR_RULE_LOCAL,
 953         IPV6_SADDR_RULE_SCOPE,
 954         IPV6_SADDR_RULE_PREFERRED,
 955 #ifdef CONFIG_IPV6_MIP6
 956         IPV6_SADDR_RULE_HOA,
 957 #endif
 958         IPV6_SADDR_RULE_OIF,
 959         IPV6_SADDR_RULE_LABEL,
 960 #ifdef CONFIG_IPV6_PRIVACY
 961         IPV6_SADDR_RULE_PRIVACY,
 962 #endif
 963         IPV6_SADDR_RULE_ORCHID,
 964         IPV6_SADDR_RULE_PREFIX,
 965         IPV6_SADDR_RULE_MAX
 966 };
 967
 968 struct ipv6_saddr_score {
 969         int                     rule;
 970         int                     addr_type;
 971         struct inet6_ifaddr     *ifa;
 972         DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX);
 973         int                     scopedist;
 974         int                     matchlen;
 975 };
 976
 977 struct ipv6_saddr_dst {
 978         const struct in6_addr *addr;
 979         int ifindex;
 980         int scope;
 981         int label;
 982         unsigned int prefs;
 983 };
 984
 985 static inline int ipv6_saddr_preferred(int type)
 986 {
 987         if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK))
 988                 return 1;
 989         return 0;
 990 }
 991
 992 static int ipv6_get_saddr_eval(struct net *net,
 993                                struct ipv6_saddr_score *score,
 994                                struct ipv6_saddr_dst *dst,
 995                                int i)
 996 {
 997         int ret;
 998
 999         if (i <= score->rule) {
1000                 switch (i) {
1001                 case IPV6_SADDR_RULE_SCOPE:
1002                         ret = score->scopedist;
1003                         break;
1004                 case IPV6_SADDR_RULE_PREFIX:
1005                         ret = score->matchlen;
1006                         break;
1007                 default:
1008                         ret = !!test_bit(i, score->scorebits);
1009                 }
1010                 goto out;
1011         }
1012
1013         switch (i) {
1014         case IPV6_SADDR_RULE_INIT:
1015                 /* Rule 0: remember if hiscore is not ready yet */
1016                 ret = !!score->ifa;
1017                 break;
1018         case IPV6_SADDR_RULE_LOCAL:
1019                 /* Rule 1: Prefer same address */
1020                 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr);
1021                 break;
1022         case IPV6_SADDR_RULE_SCOPE:
1023                 /* Rule 2: Prefer appropriate scope
1024                  *
1025                  *      ret
1026                  *       ^
1027                  *    -1 |  d 15
1028                  *    ---+--+-+---> scope
1029                  *       |
1030                  *       |             d is scope of the destination.
1031                  *  B-d  |  \
1032                  *       |   \      <- smaller scope is better if
1033                  *  B-15 |    \        if scope is enough for destinaion.
1034                  *       |             ret = B - scope (-1 <= scope >= d <= 15).
1035                  * d-C-1 | /
1036                  *       |/         <- greater is better
1037                  *   -C  /             if scope is not enough for destination.
1038                  *      /|             ret = scope - C (-1 <= d < scope <= 15).
1039                  *
1040                  * d - C - 1 < B -15 (for all -1 <= d <= 15).
1041                  * C > d + 14 - B >= 15 + 14 - B = 29 - B.
1042                  * Assume B = 0 and we get C > 29.
1043                  */
1044                 ret = __ipv6_addr_src_scope(score->addr_type);
1045                 if (ret >= dst->scope)
1046                         ret = -ret;
1047                 else
1048                         ret -= 128;     /* 30 is enough */
1049                 score->scopedist = ret;
1050                 break;
1051         case IPV6_SADDR_RULE_PREFERRED:
1052                 /* Rule 3: Avoid deprecated and optimistic addresses */
1053                 ret = ipv6_saddr_preferred(score->addr_type) ||
1054                       !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC));
1055                 break;
1056 #ifdef CONFIG_IPV6_MIP6
1057         case IPV6_SADDR_RULE_HOA:
1058             {
1059                 /* Rule 4: Prefer home address */
1060                 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA);
1061                 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome;
1062                 break;
1063             }
1064 #endif
1065         case IPV6_SADDR_RULE_OIF:
1066                 /* Rule 5: Prefer outgoing interface */
1067                 ret = (!dst->ifindex ||
1068                        dst->ifindex == score->ifa->idev->dev->ifindex);
1069                 break;
1070         case IPV6_SADDR_RULE_LABEL:
1071                 /* Rule 6: Prefer matching label */
1072                 ret = ipv6_addr_label(net,
1073                                       &score->ifa->addr, score->addr_type,
1074                                       score->ifa->idev->dev->ifindex) == dst->label;
1075                 break;
1076 #ifdef CONFIG_IPV6_PRIVACY
1077         case IPV6_SADDR_RULE_PRIVACY:
1078             {
1079                 /* Rule 7: Prefer public address
1080                  * Note: prefer temporary address if use_tempaddr >= 2
1081                  */
1082                 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ?
1083                                 !!(dst->prefs & IPV6_PREFER_SRC_TMP) :
1084                                 score->ifa->idev->cnf.use_tempaddr >= 2;
1085                 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp;
1086                 break;
1087             }
1088 #endif
1089         case IPV6_SADDR_RULE_ORCHID:
1090                 /* Rule 8-: Prefer ORCHID vs ORCHID or
1091                  *          non-ORCHID vs non-ORCHID
1092                  */
1093                 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^
1094                         ipv6_addr_orchid(dst->addr));
1095                 break;
1096         case IPV6_SADDR_RULE_PREFIX:
1097                 /* Rule 8: Use longest matching prefix */
1098                 score->matchlen = ret = ipv6_addr_diff(&score->ifa->addr,
1099                                                        dst->addr);
1100                 break;
1101         default:
1102                 ret = 0;
1103         }
1104
1105         if (ret)
1106                 __set_bit(i, score->scorebits);
1107         score->rule = i;
1108 out:
1109         return ret;
1110 }
1111
1112 int ipv6_dev_get_saddr(struct net *net, struct net_device *dst_dev,
1113                        const struct in6_addr *daddr, unsigned int prefs,
1114                        struct in6_addr *saddr)
1115 {
1116         struct ipv6_saddr_score scores[2],
1117                                 *score = &scores[0], *hiscore = &scores[1];
1118         struct ipv6_saddr_dst dst;
1119         struct net_device *dev;
1120         int dst_type;
1121
1122         dst_type = __ipv6_addr_type(daddr);
1123         dst.addr = daddr;
1124         dst.ifindex = dst_dev ? dst_dev->ifindex : 0;
1125         dst.scope = __ipv6_addr_src_scope(dst_type);
1126         dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex);
1127         dst.prefs = prefs;
1128
1129         hiscore->rule = -1;
1130         hiscore->ifa = NULL;
1131
1132         rcu_read_lock();
1133
1134         for_each_netdev_rcu(net, dev) {
1135                 struct inet6_dev *idev;
1136
1137                 /* Candidate Source Address (section 4)
1138                  *  - multicast and link-local destination address,
1139                  *    the set of candidate source address MUST only
1140                  *    include addresses assigned to interfaces
1141                  *    belonging to the same link as the outgoing
1142                  *    interface.
1143                  * (- For site-local destination addresses, the
1144                  *    set of candidate source addresses MUST only
1145                  *    include addresses assigned to interfaces
1146                  *    belonging to the same site as the outgoing
1147                  *    interface.)
1148                  */
1149                 if (((dst_type & IPV6_ADDR_MULTICAST) ||
1150                      dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) &&
1151                     dst.ifindex && dev->ifindex != dst.ifindex)
1152                         continue;
1153
1154                 idev = __in6_dev_get(dev);
1155                 if (!idev)
1156                         continue;
1157
1158                 read_lock_bh(&idev->lock);
1159                 list_for_each_entry(score->ifa, &idev->addr_list, if_list) {
1160                         int i;
1161
1162                         /*
1163                          * - Tentative Address (RFC2462 section 5.4)
1164                          *  - A tentative address is not considered
1165                          *    "assigned to an interface" in the traditional
1166                          *    sense, unless it is also flagged as optimistic.
1167                          * - Candidate Source Address (section 4)
1168                          *  - In any case, anycast addresses, multicast
1169                          *    addresses, and the unspecified address MUST
1170                          *    NOT be included in a candidate set.
1171                          */
1172                         if ((score->ifa->flags & IFA_F_TENTATIVE) &&
1173                             (!(score->ifa->flags & IFA_F_OPTIMISTIC)))
1174                                 continue;
1175
1176                         score->addr_type = __ipv6_addr_type(&score->ifa->addr);
1177
1178                         if (unlikely(score->addr_type == IPV6_ADDR_ANY ||
1179                                      score->addr_type & IPV6_ADDR_MULTICAST)) {
1180                                 LIMIT_NETDEBUG(KERN_DEBUG
1181                                                "ADDRCONF: unspecified / multicast address "
1182                                                "assigned as unicast address on %s",
1183                                                dev->name);
1184                                 continue;
1185                         }
1186
1187                         score->rule = -1;
1188                         bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX);
1189
1190                         for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) {
1191                                 int minihiscore, miniscore;
1192
1193                                 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i);
1194                                 miniscore = ipv6_get_saddr_eval(net, score, &dst, i);
1195
1196                                 if (minihiscore > miniscore) {
1197                                         if (i == IPV6_SADDR_RULE_SCOPE &&
1198                                             score->scopedist > 0) {
1199                                                 /*
1200                                                  * special case:
1201                                                  * each remaining entry
1202                                                  * has too small (not enough)
1203                                                  * scope, because ifa entries
1204                                                  * are sorted by their scope
1205                                                  * values.
1206                                                  */
1207                                                 goto try_nextdev;
1208                                         }
1209                                         break;
1210                                 } else if (minihiscore < miniscore) {
1211                                         if (hiscore->ifa)
1212                                                 in6_ifa_put(hiscore->ifa);
1213
1214                                         in6_ifa_hold(score->ifa);
1215
1216                                         swap(hiscore, score);
1217
1218                                         /* restore our iterator */
1219                                         score->ifa = hiscore->ifa;
1220
1221                                         break;
1222                                 }
1223                         }
1224                 }
1225 try_nextdev:
1226                 read_unlock_bh(&idev->lock);
1227         }
1228         rcu_read_unlock();
1229
1230         if (!hiscore->ifa)
1231                 return -EADDRNOTAVAIL;
1232
1233         ipv6_addr_copy(saddr, &hiscore->ifa->addr);
1234         in6_ifa_put(hiscore->ifa);
1235         return 0;
1236 }
1237 EXPORT_SYMBOL(ipv6_dev_get_saddr);
1238
1239 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr,
1240                       unsigned char banned_flags)
1241 {
1242         struct inet6_ifaddr *ifp;
1243         int err = -EADDRNOTAVAIL;
1244
1245         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1246                 if (ifp->scope == IFA_LINK &&
1247                     !(ifp->flags & banned_flags)) {
1248                         ipv6_addr_copy(addr, &ifp->addr);
1249                         err = 0;
1250                         break;
1251                 }
1252         }
1253         return err;
1254 }
1255
1256 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr,
1257                     unsigned char banned_flags)
1258 {
1259         struct inet6_dev *idev;
1260         int err = -EADDRNOTAVAIL;
1261
1262         rcu_read_lock();
1263         idev = __in6_dev_get(dev);
1264         if (idev) {
1265                 read_lock_bh(&idev->lock);
1266                 err = __ipv6_get_lladdr(idev, addr, banned_flags);
1267                 read_unlock_bh(&idev->lock);
1268         }
1269         rcu_read_unlock();
1270         return err;
1271 }
1272
1273 static int ipv6_count_addresses(struct inet6_dev *idev)
1274 {
1275         int cnt = 0;
1276         struct inet6_ifaddr *ifp;
1277
1278         read_lock_bh(&idev->lock);
1279         list_for_each_entry(ifp, &idev->addr_list, if_list)
1280                 cnt++;
1281         read_unlock_bh(&idev->lock);
1282         return cnt;
1283 }
1284
1285 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr,
1286                   struct net_device *dev, int strict)
1287 {
1288         struct inet6_ifaddr *ifp;
1289         struct hlist_node *node;
1290         unsigned int hash = ipv6_addr_hash(addr);
1291
1292         rcu_read_lock_bh();
1293         hlist_for_each_entry_rcu(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1294                 if (!net_eq(dev_net(ifp->idev->dev), net))
1295                         continue;
1296                 if (ipv6_addr_equal(&ifp->addr, addr) &&
1297                     !(ifp->flags&IFA_F_TENTATIVE) &&
1298                     (dev == NULL || ifp->idev->dev == dev ||
1299                      !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) {
1300                         rcu_read_unlock_bh();
1301                         return 1;
1302                 }
1303         }
1304
1305         rcu_read_unlock_bh();
1306         return 0;
1307 }
1308 EXPORT_SYMBOL(ipv6_chk_addr);
1309
1310 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr,
1311                                struct net_device *dev)
1312 {
1313         unsigned int hash = ipv6_addr_hash(addr);
1314         struct inet6_ifaddr *ifp;
1315         struct hlist_node *node;
1316
1317         hlist_for_each_entry(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1318                 if (!net_eq(dev_net(ifp->idev->dev), net))
1319                         continue;
1320                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1321                         if (dev == NULL || ifp->idev->dev == dev)
1322                                 return true;
1323                 }
1324         }
1325         return false;
1326 }
1327
1328 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev)
1329 {
1330         struct inet6_dev *idev;
1331         struct inet6_ifaddr *ifa;
1332         int     onlink;
1333
1334         onlink = 0;
1335         rcu_read_lock();
1336         idev = __in6_dev_get(dev);
1337         if (idev) {
1338                 read_lock_bh(&idev->lock);
1339                 list_for_each_entry(ifa, &idev->addr_list, if_list) {
1340                         onlink = ipv6_prefix_equal(addr, &ifa->addr,
1341                                                    ifa->prefix_len);
1342                         if (onlink)
1343                                 break;
1344                 }
1345                 read_unlock_bh(&idev->lock);
1346         }
1347         rcu_read_unlock();
1348         return onlink;
1349 }
1350
1351 EXPORT_SYMBOL(ipv6_chk_prefix);
1352
1353 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr,
1354                                      struct net_device *dev, int strict)
1355 {
1356         struct inet6_ifaddr *ifp, *result = NULL;
1357         unsigned int hash = ipv6_addr_hash(addr);
1358         struct hlist_node *node;
1359
1360         rcu_read_lock_bh();
1361         hlist_for_each_entry_rcu_bh(ifp, node, &inet6_addr_lst[hash], addr_lst) {
1362                 if (!net_eq(dev_net(ifp->idev->dev), net))
1363                         continue;
1364                 if (ipv6_addr_equal(&ifp->addr, addr)) {
1365                         if (dev == NULL || ifp->idev->dev == dev ||
1366                             !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) {
1367                                 result = ifp;
1368                                 in6_ifa_hold(ifp);
1369                                 break;
1370                         }
1371                 }
1372         }
1373         rcu_read_unlock_bh();
1374
1375         return result;
1376 }
1377
1378 /* Gets referenced address, destroys ifaddr */
1379
1380 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed)
1381 {
1382         if (ifp->flags&IFA_F_PERMANENT) {
1383                 spin_lock_bh(&ifp->lock);
1384                 addrconf_del_timer(ifp);
1385                 ifp->flags |= IFA_F_TENTATIVE;
1386                 if (dad_failed)
1387                         ifp->flags |= IFA_F_DADFAILED;
1388                 spin_unlock_bh(&ifp->lock);
1389                 if (dad_failed)
1390                         ipv6_ifa_notify(0, ifp);
1391                 in6_ifa_put(ifp);
1392 #ifdef CONFIG_IPV6_PRIVACY
1393         } else if (ifp->flags&IFA_F_TEMPORARY) {
1394                 struct inet6_ifaddr *ifpub;
1395                 spin_lock_bh(&ifp->lock);
1396                 ifpub = ifp->ifpub;
1397                 if (ifpub) {
1398                         in6_ifa_hold(ifpub);
1399                         spin_unlock_bh(&ifp->lock);
1400                         ipv6_create_tempaddr(ifpub, ifp);
1401                         in6_ifa_put(ifpub);
1402                 } else {
1403                         spin_unlock_bh(&ifp->lock);
1404                 }
1405                 ipv6_del_addr(ifp);
1406 #endif
1407         } else
1408                 ipv6_del_addr(ifp);
1409 }
1410
1411 static int addrconf_dad_end(struct inet6_ifaddr *ifp)
1412 {
1413         int err = -ENOENT;
1414
1415         spin_lock(&ifp->state_lock);
1416         if (ifp->state == INET6_IFADDR_STATE_DAD) {
1417                 ifp->state = INET6_IFADDR_STATE_POSTDAD;
1418                 err = 0;
1419         }
1420         spin_unlock(&ifp->state_lock);
1421
1422         return err;
1423 }
1424
1425 void addrconf_dad_failure(struct inet6_ifaddr *ifp)
1426 {
1427         struct inet6_dev *idev = ifp->idev;
1428
1429         if (addrconf_dad_end(ifp)) {
1430                 in6_ifa_put(ifp);
1431                 return;
1432         }
1433
1434         if (net_ratelimit())
1435                 printk(KERN_INFO "%s: IPv6 duplicate address %pI6c detected!\n",
1436                         ifp->idev->dev->name, &ifp->addr);
1437
1438         if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) {
1439                 struct in6_addr addr;
1440
1441                 addr.s6_addr32[0] = htonl(0xfe800000);
1442                 addr.s6_addr32[1] = 0;
1443
1444                 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) &&
1445                     ipv6_addr_equal(&ifp->addr, &addr)) {
1446                         /* DAD failed for link-local based on MAC address */
1447                         idev->cnf.disable_ipv6 = 1;
1448
1449                         printk(KERN_INFO "%s: IPv6 being disabled!\n",
1450                                 ifp->idev->dev->name);
1451                 }
1452         }
1453
1454         addrconf_dad_stop(ifp, 1);
1455 }
1456
1457 /* Join to solicited addr multicast group. */
1458
1459 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr)
1460 {
1461         struct in6_addr maddr;
1462
1463         if (dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1464                 return;
1465
1466         addrconf_addr_solict_mult(addr, &maddr);
1467         ipv6_dev_mc_inc(dev, &maddr);
1468 }
1469
1470 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr)
1471 {
1472         struct in6_addr maddr;
1473
1474         if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP))
1475                 return;
1476
1477         addrconf_addr_solict_mult(addr, &maddr);
1478         __ipv6_dev_mc_dec(idev, &maddr);
1479 }
1480
1481 static void addrconf_join_anycast(struct inet6_ifaddr *ifp)
1482 {
1483         struct in6_addr addr;
1484         if (ifp->prefix_len == 127) /* RFC 6164 */
1485                 return;
1486         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1487         if (ipv6_addr_any(&addr))
1488                 return;
1489         ipv6_dev_ac_inc(ifp->idev->dev, &addr);
1490 }
1491
1492 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp)
1493 {
1494         struct in6_addr addr;
1495         if (ifp->prefix_len == 127) /* RFC 6164 */
1496                 return;
1497         ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len);
1498         if (ipv6_addr_any(&addr))
1499                 return;
1500         __ipv6_dev_ac_dec(ifp->idev, &addr);
1501 }
1502
1503 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev)
1504 {
1505         if (dev->addr_len != ETH_ALEN)
1506                 return -1;
1507         memcpy(eui, dev->dev_addr, 3);
1508         memcpy(eui + 5, dev->dev_addr + 3, 3);
1509
1510         /*
1511          * The zSeries OSA network cards can be shared among various
1512          * OS instances, but the OSA cards have only one MAC address.
1513          * This leads to duplicate address conflicts in conjunction
1514          * with IPv6 if more than one instance uses the same card.
1515          *
1516          * The driver for these cards can deliver a unique 16-bit
1517          * identifier for each instance sharing the same card.  It is
1518          * placed instead of 0xFFFE in the interface identifier.  The
1519          * "u" bit of the interface identifier is not inverted in this
1520          * case.  Hence the resulting interface identifier has local
1521          * scope according to RFC2373.
1522          */
1523         if (dev->dev_id) {
1524                 eui[3] = (dev->dev_id >> 8) & 0xFF;
1525                 eui[4] = dev->dev_id & 0xFF;
1526         } else {
1527                 eui[3] = 0xFF;
1528                 eui[4] = 0xFE;
1529                 eui[0] ^= 2;
1530         }
1531         return 0;
1532 }
1533
1534 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev)
1535 {
1536         /* XXX: inherit EUI-64 from other interface -- yoshfuji */
1537         if (dev->addr_len != ARCNET_ALEN)
1538                 return -1;
1539         memset(eui, 0, 7);
1540         eui[7] = *(u8*)dev->dev_addr;
1541         return 0;
1542 }
1543
1544 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev)
1545 {
1546         if (dev->addr_len != INFINIBAND_ALEN)
1547                 return -1;
1548         memcpy(eui, dev->dev_addr + 12, 8);
1549         eui[0] |= 2;
1550         return 0;
1551 }
1552
1553 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr)
1554 {
1555         if (addr == 0)
1556                 return -1;
1557         eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) ||
1558                   ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) ||
1559                   ipv4_is_private_172(addr) || ipv4_is_test_192(addr) ||
1560                   ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) ||
1561                   ipv4_is_test_198(addr) || ipv4_is_multicast(addr) ||
1562                   ipv4_is_lbcast(addr)) ? 0x00 : 0x02;
1563         eui[1] = 0;
1564         eui[2] = 0x5E;
1565         eui[3] = 0xFE;
1566         memcpy(eui + 4, &addr, 4);
1567         return 0;
1568 }
1569
1570 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev)
1571 {
1572         if (dev->priv_flags & IFF_ISATAP)
1573                 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1574         return -1;
1575 }
1576
1577 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev)
1578 {
1579         return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr);
1580 }
1581
1582 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev)
1583 {
1584         switch (dev->type) {
1585         case ARPHRD_ETHER:
1586         case ARPHRD_FDDI:
1587         case ARPHRD_IEEE802_TR:
1588                 return addrconf_ifid_eui48(eui, dev);
1589         case ARPHRD_ARCNET:
1590                 return addrconf_ifid_arcnet(eui, dev);
1591         case ARPHRD_INFINIBAND:
1592                 return addrconf_ifid_infiniband(eui, dev);
1593         case ARPHRD_SIT:
1594                 return addrconf_ifid_sit(eui, dev);
1595         case ARPHRD_IPGRE:
1596                 return addrconf_ifid_gre(eui, dev);
1597         }
1598         return -1;
1599 }
1600
1601 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev)
1602 {
1603         int err = -1;
1604         struct inet6_ifaddr *ifp;
1605
1606         read_lock_bh(&idev->lock);
1607         list_for_each_entry(ifp, &idev->addr_list, if_list) {
1608                 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) {
1609                         memcpy(eui, ifp->addr.s6_addr+8, 8);
1610                         err = 0;
1611                         break;
1612                 }
1613         }
1614         read_unlock_bh(&idev->lock);
1615         return err;
1616 }
1617
1618 #ifdef CONFIG_IPV6_PRIVACY
1619 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */
1620 static int __ipv6_regen_rndid(struct inet6_dev *idev)
1621 {
1622 regen:
1623         get_random_bytes(idev->rndid, sizeof(idev->rndid));
1624         idev->rndid[0] &= ~0x02;
1625
1626         /*
1627          * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>:
1628          * check if generated address is not inappropriate
1629          *
1630          *  - Reserved subnet anycast (RFC 2526)
1631          *      11111101 11....11 1xxxxxxx
1632          *  - ISATAP (RFC4214) 6.1
1633          *      00-00-5E-FE-xx-xx-xx-xx
1634          *  - value 0
1635          *  - XXX: already assigned to an address on the device
1636          */
1637         if (idev->rndid[0] == 0xfd &&
1638             (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff &&
1639             (idev->rndid[7]&0x80))
1640                 goto regen;
1641         if ((idev->rndid[0]|idev->rndid[1]) == 0) {
1642                 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe)
1643                         goto regen;
1644                 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00)
1645                         goto regen;
1646         }
1647
1648         return 0;
1649 }
1650
1651 static void ipv6_regen_rndid(unsigned long data)
1652 {
1653         struct inet6_dev *idev = (struct inet6_dev *) data;
1654         unsigned long expires;
1655
1656         rcu_read_lock_bh();
1657         write_lock_bh(&idev->lock);
1658
1659         if (idev->dead)
1660                 goto out;
1661
1662         if (__ipv6_regen_rndid(idev) < 0)
1663                 goto out;
1664
1665         expires = jiffies +
1666                 idev->cnf.temp_prefered_lft * HZ -
1667                 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * idev->nd_parms->retrans_time -
1668                 idev->cnf.max_desync_factor * HZ;
1669         if (time_before(expires, jiffies)) {
1670                 printk(KERN_WARNING
1671                         "ipv6_regen_rndid(): too short regeneration interval; timer disabled for %s.\n",
1672                         idev->dev->name);
1673                 goto out;
1674         }
1675
1676         if (!mod_timer(&idev->regen_timer, expires))
1677                 in6_dev_hold(idev);
1678
1679 out:
1680         write_unlock_bh(&idev->lock);
1681         rcu_read_unlock_bh();
1682         in6_dev_put(idev);
1683 }
1684
1685 static int __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) {
1686         int ret = 0;
1687
1688         if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0)
1689                 ret = __ipv6_regen_rndid(idev);
1690         return ret;
1691 }
1692 #endif
1693
1694 /*
1695  *      Add prefix route.
1696  */
1697
1698 static void
1699 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev,
1700                       unsigned long expires, u32 flags)
1701 {
1702         struct fib6_config cfg = {
1703                 .fc_table = RT6_TABLE_PREFIX,
1704                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1705                 .fc_ifindex = dev->ifindex,
1706                 .fc_expires = expires,
1707                 .fc_dst_len = plen,
1708                 .fc_flags = RTF_UP | flags,
1709                 .fc_nlinfo.nl_net = dev_net(dev),
1710                 .fc_protocol = RTPROT_KERNEL,
1711         };
1712
1713         ipv6_addr_copy(&cfg.fc_dst, pfx);
1714
1715         /* Prevent useless cloning on PtP SIT.
1716            This thing is done here expecting that the whole
1717            class of non-broadcast devices need not cloning.
1718          */
1719 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1720         if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT))
1721                 cfg.fc_flags |= RTF_NONEXTHOP;
1722 #endif
1723
1724         ip6_route_add(&cfg);
1725 }
1726
1727
1728 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx,
1729                                                   int plen,
1730                                                   const struct net_device *dev,
1731                                                   u32 flags, u32 noflags)
1732 {
1733         struct fib6_node *fn;
1734         struct rt6_info *rt = NULL;
1735         struct fib6_table *table;
1736
1737         table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX);
1738         if (table == NULL)
1739                 return NULL;
1740
1741         write_lock_bh(&table->tb6_lock);
1742         fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0);
1743         if (!fn)
1744                 goto out;
1745         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1746                 if (rt->rt6i_dev->ifindex != dev->ifindex)
1747                         continue;
1748                 if ((rt->rt6i_flags & flags) != flags)
1749                         continue;
1750                 if ((rt->rt6i_flags & noflags) != 0)
1751                         continue;
1752                 dst_hold(&rt->dst);
1753                 break;
1754         }
1755 out:
1756         write_unlock_bh(&table->tb6_lock);
1757         return rt;
1758 }
1759
1760
1761 /* Create "default" multicast route to the interface */
1762
1763 static void addrconf_add_mroute(struct net_device *dev)
1764 {
1765         struct fib6_config cfg = {
1766                 .fc_table = RT6_TABLE_LOCAL,
1767                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1768                 .fc_ifindex = dev->ifindex,
1769                 .fc_dst_len = 8,
1770                 .fc_flags = RTF_UP,
1771                 .fc_nlinfo.nl_net = dev_net(dev),
1772         };
1773
1774         ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0);
1775
1776         ip6_route_add(&cfg);
1777 }
1778
1779 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
1780 static void sit_route_add(struct net_device *dev)
1781 {
1782         struct fib6_config cfg = {
1783                 .fc_table = RT6_TABLE_MAIN,
1784                 .fc_metric = IP6_RT_PRIO_ADDRCONF,
1785                 .fc_ifindex = dev->ifindex,
1786                 .fc_dst_len = 96,
1787                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
1788                 .fc_nlinfo.nl_net = dev_net(dev),
1789         };
1790
1791         /* prefix length - 96 bits "::d.d.d.d" */
1792         ip6_route_add(&cfg);
1793 }
1794 #endif
1795
1796 static void addrconf_add_lroute(struct net_device *dev)
1797 {
1798         struct in6_addr addr;
1799
1800         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
1801         addrconf_prefix_route(&addr, 64, dev, 0, 0);
1802 }
1803
1804 static struct inet6_dev *addrconf_add_dev(struct net_device *dev)
1805 {
1806         struct inet6_dev *idev;
1807
1808         ASSERT_RTNL();
1809
1810         idev = ipv6_find_idev(dev);
1811         if (!idev)
1812                 return ERR_PTR(-ENOBUFS);
1813
1814         if (idev->cnf.disable_ipv6)
1815                 return ERR_PTR(-EACCES);
1816
1817         /* Add default multicast route */
1818         if (!(dev->flags & IFF_LOOPBACK))
1819                 addrconf_add_mroute(dev);
1820
1821         /* Add link local route */
1822         addrconf_add_lroute(dev);
1823         return idev;
1824 }
1825
1826 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len)
1827 {
1828         struct prefix_info *pinfo;
1829         __u32 valid_lft;
1830         __u32 prefered_lft;
1831         int addr_type;
1832         struct inet6_dev *in6_dev;
1833         struct net *net = dev_net(dev);
1834
1835         pinfo = (struct prefix_info *) opt;
1836
1837         if (len < sizeof(struct prefix_info)) {
1838                 ADBG(("addrconf: prefix option too short\n"));
1839                 return;
1840         }
1841
1842         /*
1843          *      Validation checks ([ADDRCONF], page 19)
1844          */
1845
1846         addr_type = ipv6_addr_type(&pinfo->prefix);
1847
1848         if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL))
1849                 return;
1850
1851         valid_lft = ntohl(pinfo->valid);
1852         prefered_lft = ntohl(pinfo->prefered);
1853
1854         if (prefered_lft > valid_lft) {
1855                 if (net_ratelimit())
1856                         printk(KERN_WARNING "addrconf: prefix option has invalid lifetime\n");
1857                 return;
1858         }
1859
1860         in6_dev = in6_dev_get(dev);
1861
1862         if (in6_dev == NULL) {
1863                 if (net_ratelimit())
1864                         printk(KERN_DEBUG "addrconf: device %s not configured\n", dev->name);
1865                 return;
1866         }
1867
1868         /*
1869          *      Two things going on here:
1870          *      1) Add routes for on-link prefixes
1871          *      2) Configure prefixes with the auto flag set
1872          */
1873
1874         if (pinfo->onlink) {
1875                 struct rt6_info *rt;
1876                 unsigned long rt_expires;
1877
1878                 /* Avoid arithmetic overflow. Really, we could
1879                  * save rt_expires in seconds, likely valid_lft,
1880                  * but it would require division in fib gc, that it
1881                  * not good.
1882                  */
1883                 if (HZ > USER_HZ)
1884                         rt_expires = addrconf_timeout_fixup(valid_lft, HZ);
1885                 else
1886                         rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ);
1887
1888                 if (addrconf_finite_timeout(rt_expires))
1889                         rt_expires *= HZ;
1890
1891                 rt = addrconf_get_prefix_route(&pinfo->prefix,
1892                                                pinfo->prefix_len,
1893                                                dev,
1894                                                RTF_ADDRCONF | RTF_PREFIX_RT,
1895                                                RTF_GATEWAY | RTF_DEFAULT);
1896
1897                 if (rt) {
1898                         /* Autoconf prefix route */
1899                         if (valid_lft == 0) {
1900                                 ip6_del_rt(rt);
1901                                 rt = NULL;
1902                         } else if (addrconf_finite_timeout(rt_expires)) {
1903                                 /* not infinity */
1904                                 rt->rt6i_expires = jiffies + rt_expires;
1905                                 rt->rt6i_flags |= RTF_EXPIRES;
1906                         } else {
1907                                 rt->rt6i_flags &= ~RTF_EXPIRES;
1908                                 rt->rt6i_expires = 0;
1909                         }
1910                 } else if (valid_lft) {
1911                         clock_t expires = 0;
1912                         int flags = RTF_ADDRCONF | RTF_PREFIX_RT;
1913                         if (addrconf_finite_timeout(rt_expires)) {
1914                                 /* not infinity */
1915                                 flags |= RTF_EXPIRES;
1916                                 expires = jiffies_to_clock_t(rt_expires);
1917                         }
1918                         addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len,
1919                                               dev, expires, flags);
1920                 }
1921                 if (rt)
1922                         dst_release(&rt->dst);
1923         }
1924
1925         /* Try to figure out our local address for this prefix */
1926
1927         if (pinfo->autoconf && in6_dev->cnf.autoconf) {
1928                 struct inet6_ifaddr * ifp;
1929                 struct in6_addr addr;
1930                 int create = 0, update_lft = 0;
1931
1932                 if (pinfo->prefix_len == 64) {
1933                         memcpy(&addr, &pinfo->prefix, 8);
1934                         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) &&
1935                             ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) {
1936                                 in6_dev_put(in6_dev);
1937                                 return;
1938                         }
1939                         goto ok;
1940                 }
1941                 if (net_ratelimit())
1942                         printk(KERN_DEBUG "IPv6 addrconf: prefix with wrong length %d\n",
1943                                pinfo->prefix_len);
1944                 in6_dev_put(in6_dev);
1945                 return;
1946
1947 ok:
1948
1949                 ifp = ipv6_get_ifaddr(net, &addr, dev, 1);
1950
1951                 if (ifp == NULL && valid_lft) {
1952                         int max_addresses = in6_dev->cnf.max_addresses;
1953                         u32 addr_flags = 0;
1954
1955 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
1956                         if (in6_dev->cnf.optimistic_dad &&
1957                             !net->ipv6.devconf_all->forwarding)
1958                                 addr_flags = IFA_F_OPTIMISTIC;
1959 #endif
1960
1961                         /* Do not allow to create too much of autoconfigured
1962                          * addresses; this would be too easy way to crash kernel.
1963                          */
1964                         if (!max_addresses ||
1965                             ipv6_count_addresses(in6_dev) < max_addresses)
1966                                 ifp = ipv6_add_addr(in6_dev, &addr, pinfo->prefix_len,
1967                                                     addr_type&IPV6_ADDR_SCOPE_MASK,
1968                                                     addr_flags);
1969
1970                         if (!ifp || IS_ERR(ifp)) {
1971                                 in6_dev_put(in6_dev);
1972                                 return;
1973                         }
1974
1975                         update_lft = create = 1;
1976                         ifp->cstamp = jiffies;
1977                         addrconf_dad_start(ifp, RTF_ADDRCONF|RTF_PREFIX_RT);
1978                 }
1979
1980                 if (ifp) {
1981                         int flags;
1982                         unsigned long now;
1983 #ifdef CONFIG_IPV6_PRIVACY
1984                         struct inet6_ifaddr *ift;
1985 #endif
1986                         u32 stored_lft;
1987
1988                         /* update lifetime (RFC2462 5.5.3 e) */
1989                         spin_lock(&ifp->lock);
1990                         now = jiffies;
1991                         if (ifp->valid_lft > (now - ifp->tstamp) / HZ)
1992                                 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ;
1993                         else
1994                                 stored_lft = 0;
1995                         if (!update_lft && stored_lft) {
1996                                 if (valid_lft > MIN_VALID_LIFETIME ||
1997                                     valid_lft > stored_lft)
1998                                         update_lft = 1;
1999                                 else if (stored_lft <= MIN_VALID_LIFETIME) {
2000                                         /* valid_lft <= stored_lft is always true */
2001                                         /*
2002                                          * RFC 4862 Section 5.5.3e:
2003                                          * "Note that the preferred lifetime of
2004                                          *  the corresponding address is always
2005                                          *  reset to the Preferred Lifetime in
2006                                          *  the received Prefix Information
2007                                          *  option, regardless of whether the
2008                                          *  valid lifetime is also reset or
2009                                          *  ignored."
2010                                          *
2011                                          *  So if the preferred lifetime in
2012                                          *  this advertisement is different
2013                                          *  than what we have stored, but the
2014                                          *  valid lifetime is invalid, just
2015                                          *  reset prefered_lft.
2016                                          *
2017                                          *  We must set the valid lifetime
2018                                          *  to the stored lifetime since we'll
2019                                          *  be updating the timestamp below,
2020                                          *  else we'll set it back to the
2021                                          *  minimum.
2022                                          */
2023                                         if (prefered_lft != ifp->prefered_lft) {
2024                                                 valid_lft = stored_lft;
2025                                                 update_lft = 1;
2026                                         }
2027                                 } else {
2028                                         valid_lft = MIN_VALID_LIFETIME;
2029                                         if (valid_lft < prefered_lft)
2030                                                 prefered_lft = valid_lft;
2031                                         update_lft = 1;
2032                                 }
2033                         }
2034
2035                         if (update_lft) {
2036                                 ifp->valid_lft = valid_lft;
2037                                 ifp->prefered_lft = prefered_lft;
2038                                 ifp->tstamp = now;
2039                                 flags = ifp->flags;
2040                                 ifp->flags &= ~IFA_F_DEPRECATED;
2041                                 spin_unlock(&ifp->lock);
2042
2043                                 if (!(flags&IFA_F_TENTATIVE))
2044                                         ipv6_ifa_notify(0, ifp);
2045                         } else
2046                                 spin_unlock(&ifp->lock);
2047
2048 #ifdef CONFIG_IPV6_PRIVACY
2049                         read_lock_bh(&in6_dev->lock);
2050                         /* update all temporary addresses in the list */
2051                         list_for_each_entry(ift, &in6_dev->tempaddr_list,
2052                                             tmp_list) {
2053                                 int age, max_valid, max_prefered;
2054
2055                                 if (ifp != ift->ifpub)
2056                                         continue;
2057
2058                                 /*
2059                                  * RFC 4941 section 3.3:
2060                                  * If a received option will extend the lifetime
2061                                  * of a public address, the lifetimes of
2062                                  * temporary addresses should be extended,
2063                                  * subject to the overall constraint that no
2064                                  * temporary addresses should ever remain
2065                                  * "valid" or "preferred" for a time longer than
2066                                  * (TEMP_VALID_LIFETIME) or
2067                                  * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR),
2068                                  * respectively.
2069                                  */
2070                                 age = (now - ift->cstamp) / HZ;
2071                                 max_valid = in6_dev->cnf.temp_valid_lft - age;
2072                                 if (max_valid < 0)
2073                                         max_valid = 0;
2074
2075                                 max_prefered = in6_dev->cnf.temp_prefered_lft -
2076                                                in6_dev->cnf.max_desync_factor -
2077                                                age;
2078                                 if (max_prefered < 0)
2079                                         max_prefered = 0;
2080
2081                                 if (valid_lft > max_valid)
2082                                         valid_lft = max_valid;
2083
2084                                 if (prefered_lft > max_prefered)
2085                                         prefered_lft = max_prefered;
2086
2087                                 spin_lock(&ift->lock);
2088                                 flags = ift->flags;
2089                                 ift->valid_lft = valid_lft;
2090                                 ift->prefered_lft = prefered_lft;
2091                                 ift->tstamp = now;
2092                                 if (prefered_lft > 0)
2093                                         ift->flags &= ~IFA_F_DEPRECATED;
2094
2095                                 spin_unlock(&ift->lock);
2096                                 if (!(flags&IFA_F_TENTATIVE))
2097                                         ipv6_ifa_notify(0, ift);
2098                         }
2099
2100                         if ((create || list_empty(&in6_dev->tempaddr_list)) && in6_dev->cnf.use_tempaddr > 0) {
2101                                 /*
2102                                  * When a new public address is created as
2103                                  * described in [ADDRCONF], also create a new
2104                                  * temporary address. Also create a temporary
2105                                  * address if it's enabled but no temporary
2106                                  * address currently exists.
2107                                  */
2108                                 read_unlock_bh(&in6_dev->lock);
2109                                 ipv6_create_tempaddr(ifp, NULL);
2110                         } else {
2111                                 read_unlock_bh(&in6_dev->lock);
2112                         }
2113 #endif
2114                         in6_ifa_put(ifp);
2115                         addrconf_verify(0);
2116                 }
2117         }
2118         inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo);
2119         in6_dev_put(in6_dev);
2120 }
2121
2122 /*
2123  *      Set destination address.
2124  *      Special case for SIT interfaces where we create a new "virtual"
2125  *      device.
2126  */
2127 int addrconf_set_dstaddr(struct net *net, void __user *arg)
2128 {
2129         struct in6_ifreq ireq;
2130         struct net_device *dev;
2131         int err = -EINVAL;
2132
2133         rtnl_lock();
2134
2135         err = -EFAULT;
2136         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2137                 goto err_exit;
2138
2139         dev = __dev_get_by_index(net, ireq.ifr6_ifindex);
2140
2141         err = -ENODEV;
2142         if (dev == NULL)
2143                 goto err_exit;
2144
2145 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2146         if (dev->type == ARPHRD_SIT) {
2147                 const struct net_device_ops *ops = dev->netdev_ops;
2148                 struct ifreq ifr;
2149                 struct ip_tunnel_parm p;
2150
2151                 err = -EADDRNOTAVAIL;
2152                 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4))
2153                         goto err_exit;
2154
2155                 memset(&p, 0, sizeof(p));
2156                 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3];
2157                 p.iph.saddr = 0;
2158                 p.iph.version = 4;
2159                 p.iph.ihl = 5;
2160                 p.iph.protocol = IPPROTO_IPV6;
2161                 p.iph.ttl = 64;
2162                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
2163
2164                 if (ops->ndo_do_ioctl) {
2165                         mm_segment_t oldfs = get_fs();
2166
2167                         set_fs(KERNEL_DS);
2168                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
2169                         set_fs(oldfs);
2170                 } else
2171                         err = -EOPNOTSUPP;
2172
2173                 if (err == 0) {
2174                         err = -ENOBUFS;
2175                         dev = __dev_get_by_name(net, p.name);
2176                         if (!dev)
2177                                 goto err_exit;
2178                         err = dev_open(dev);
2179                 }
2180         }
2181 #endif
2182
2183 err_exit:
2184         rtnl_unlock();
2185         return err;
2186 }
2187
2188 /*
2189  *      Manual configuration of address on an interface
2190  */
2191 static int inet6_addr_add(struct net *net, int ifindex, const struct in6_addr *pfx,
2192                           unsigned int plen, __u8 ifa_flags, __u32 prefered_lft,
2193                           __u32 valid_lft)
2194 {
2195         struct inet6_ifaddr *ifp;
2196         struct inet6_dev *idev;
2197         struct net_device *dev;
2198         int scope;
2199         u32 flags;
2200         clock_t expires;
2201         unsigned long timeout;
2202
2203         ASSERT_RTNL();
2204
2205         if (plen > 128)
2206                 return -EINVAL;
2207
2208         /* check the lifetime */
2209         if (!valid_lft || prefered_lft > valid_lft)
2210                 return -EINVAL;
2211
2212         dev = __dev_get_by_index(net, ifindex);
2213         if (!dev)
2214                 return -ENODEV;
2215
2216         idev = addrconf_add_dev(dev);
2217         if (IS_ERR(idev))
2218                 return PTR_ERR(idev);
2219
2220         scope = ipv6_addr_scope(pfx);
2221
2222         timeout = addrconf_timeout_fixup(valid_lft, HZ);
2223         if (addrconf_finite_timeout(timeout)) {
2224                 expires = jiffies_to_clock_t(timeout * HZ);
2225                 valid_lft = timeout;
2226                 flags = RTF_EXPIRES;
2227         } else {
2228                 expires = 0;
2229                 flags = 0;
2230                 ifa_flags |= IFA_F_PERMANENT;
2231         }
2232
2233         timeout = addrconf_timeout_fixup(prefered_lft, HZ);
2234         if (addrconf_finite_timeout(timeout)) {
2235                 if (timeout == 0)
2236                         ifa_flags |= IFA_F_DEPRECATED;
2237                 prefered_lft = timeout;
2238         }
2239
2240         ifp = ipv6_add_addr(idev, pfx, plen, scope, ifa_flags);
2241
2242         if (!IS_ERR(ifp)) {
2243                 spin_lock_bh(&ifp->lock);
2244                 ifp->valid_lft = valid_lft;
2245                 ifp->prefered_lft = prefered_lft;
2246                 ifp->tstamp = jiffies;
2247                 spin_unlock_bh(&ifp->lock);
2248
2249                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev,
2250                                       expires, flags);
2251                 /*
2252                  * Note that section 3.1 of RFC 4429 indicates
2253                  * that the Optimistic flag should not be set for
2254                  * manually configured addresses
2255                  */
2256                 addrconf_dad_start(ifp, 0);
2257                 in6_ifa_put(ifp);
2258                 addrconf_verify(0);
2259                 return 0;
2260         }
2261
2262         return PTR_ERR(ifp);
2263 }
2264
2265 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx,
2266                           unsigned int plen)
2267 {
2268         struct inet6_ifaddr *ifp;
2269         struct inet6_dev *idev;
2270         struct net_device *dev;
2271
2272         if (plen > 128)
2273                 return -EINVAL;
2274
2275         dev = __dev_get_by_index(net, ifindex);
2276         if (!dev)
2277                 return -ENODEV;
2278
2279         if ((idev = __in6_dev_get(dev)) == NULL)
2280                 return -ENXIO;
2281
2282         read_lock_bh(&idev->lock);
2283         list_for_each_entry(ifp, &idev->addr_list, if_list) {
2284                 if (ifp->prefix_len == plen &&
2285                     ipv6_addr_equal(pfx, &ifp->addr)) {
2286                         in6_ifa_hold(ifp);
2287                         read_unlock_bh(&idev->lock);
2288
2289                         ipv6_del_addr(ifp);
2290
2291                         /* If the last address is deleted administratively,
2292                            disable IPv6 on this interface.
2293                          */
2294                         if (list_empty(&idev->addr_list))
2295                                 addrconf_ifdown(idev->dev, 1);
2296                         return 0;
2297                 }
2298         }
2299         read_unlock_bh(&idev->lock);
2300         return -EADDRNOTAVAIL;
2301 }
2302
2303
2304 int addrconf_add_ifaddr(struct net *net, void __user *arg)
2305 {
2306         struct in6_ifreq ireq;
2307         int err;
2308
2309         if (!capable(CAP_NET_ADMIN))
2310                 return -EPERM;
2311
2312         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2313                 return -EFAULT;
2314
2315         rtnl_lock();
2316         err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2317                              ireq.ifr6_prefixlen, IFA_F_PERMANENT,
2318                              INFINITY_LIFE_TIME, INFINITY_LIFE_TIME);
2319         rtnl_unlock();
2320         return err;
2321 }
2322
2323 int addrconf_del_ifaddr(struct net *net, void __user *arg)
2324 {
2325         struct in6_ifreq ireq;
2326         int err;
2327
2328         if (!capable(CAP_NET_ADMIN))
2329                 return -EPERM;
2330
2331         if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq)))
2332                 return -EFAULT;
2333
2334         rtnl_lock();
2335         err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr,
2336                              ireq.ifr6_prefixlen);
2337         rtnl_unlock();
2338         return err;
2339 }
2340
2341 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr,
2342                      int plen, int scope)
2343 {
2344         struct inet6_ifaddr *ifp;
2345
2346         ifp = ipv6_add_addr(idev, addr, plen, scope, IFA_F_PERMANENT);
2347         if (!IS_ERR(ifp)) {
2348                 spin_lock_bh(&ifp->lock);
2349                 ifp->flags &= ~IFA_F_TENTATIVE;
2350                 spin_unlock_bh(&ifp->lock);
2351                 ipv6_ifa_notify(RTM_NEWADDR, ifp);
2352                 in6_ifa_put(ifp);
2353         }
2354 }
2355
2356 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2357 static void sit_add_v4_addrs(struct inet6_dev *idev)
2358 {
2359         struct in6_addr addr;
2360         struct net_device *dev;
2361         struct net *net = dev_net(idev->dev);
2362         int scope;
2363
2364         ASSERT_RTNL();
2365
2366         memset(&addr, 0, sizeof(struct in6_addr));
2367         memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4);
2368
2369         if (idev->dev->flags&IFF_POINTOPOINT) {
2370                 addr.s6_addr32[0] = htonl(0xfe800000);
2371                 scope = IFA_LINK;
2372         } else {
2373                 scope = IPV6_ADDR_COMPATv4;
2374         }
2375
2376         if (addr.s6_addr32[3]) {
2377                 add_addr(idev, &addr, 128, scope);
2378                 return;
2379         }
2380
2381         for_each_netdev(net, dev) {
2382                 struct in_device * in_dev = __in_dev_get_rtnl(dev);
2383                 if (in_dev && (dev->flags & IFF_UP)) {
2384                         struct in_ifaddr * ifa;
2385
2386                         int flag = scope;
2387
2388                         for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2389                                 int plen;
2390
2391                                 addr.s6_addr32[3] = ifa->ifa_local;
2392
2393                                 if (ifa->ifa_scope == RT_SCOPE_LINK)
2394                                         continue;
2395                                 if (ifa->ifa_scope >= RT_SCOPE_HOST) {
2396                                         if (idev->dev->flags&IFF_POINTOPOINT)
2397                                                 continue;
2398                                         flag |= IFA_HOST;
2399                                 }
2400                                 if (idev->dev->flags&IFF_POINTOPOINT)
2401                                         plen = 64;
2402                                 else
2403                                         plen = 96;
2404
2405                                 add_addr(idev, &addr, plen, flag);
2406                         }
2407                 }
2408         }
2409 }
2410 #endif
2411
2412 static void init_loopback(struct net_device *dev)
2413 {
2414         struct inet6_dev  *idev;
2415         struct net_device *sp_dev;
2416         struct inet6_ifaddr *sp_ifa;
2417         struct rt6_info *sp_rt;
2418
2419         /* ::1 */
2420
2421         ASSERT_RTNL();
2422
2423         if ((idev = ipv6_find_idev(dev)) == NULL) {
2424                 printk(KERN_DEBUG "init loopback: add_dev failed\n");
2425                 return;
2426         }
2427
2428         add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
2429
2430         /* Add routes to other interface's IPv6 addresses */
2431         for_each_netdev(dev_net(dev), sp_dev) {
2432                 if (!strcmp(sp_dev->name, dev->name))
2433                         continue;
2434
2435                 idev = __in6_dev_get(sp_dev);
2436                 if (!idev)
2437                         continue;
2438
2439                 read_lock_bh(&idev->lock);
2440                 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) {
2441
2442                         if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE))
2443                                 continue;
2444
2445                         if (sp_ifa->rt)
2446                                 continue;
2447
2448                         sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, 0);
2449
2450                         /* Failure cases are ignored */
2451                         if (!IS_ERR(sp_rt)) {
2452                                 sp_ifa->rt = sp_rt;
2453                                 ip6_ins_rt(sp_rt);
2454                         }
2455                 }
2456                 read_unlock_bh(&idev->lock);
2457         }
2458 }
2459
2460 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
2461 {
2462         struct inet6_ifaddr * ifp;
2463         u32 addr_flags = IFA_F_PERMANENT;
2464
2465 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD
2466         if (idev->cnf.optimistic_dad &&
2467             !dev_net(idev->dev)->ipv6.devconf_all->forwarding)
2468                 addr_flags |= IFA_F_OPTIMISTIC;
2469 #endif
2470
2471
2472         ifp = ipv6_add_addr(idev, addr, 64, IFA_LINK, addr_flags);
2473         if (!IS_ERR(ifp)) {
2474                 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0);
2475                 addrconf_dad_start(ifp, 0);
2476                 in6_ifa_put(ifp);
2477         }
2478 }
2479
2480 static void addrconf_dev_config(struct net_device *dev)
2481 {
2482         struct in6_addr addr;
2483         struct inet6_dev    * idev;
2484
2485         ASSERT_RTNL();
2486
2487         if ((dev->type != ARPHRD_ETHER) &&
2488             (dev->type != ARPHRD_FDDI) &&
2489             (dev->type != ARPHRD_IEEE802_TR) &&
2490             (dev->type != ARPHRD_ARCNET) &&
2491             (dev->type != ARPHRD_INFINIBAND)) {
2492                 /* Alas, we support only Ethernet autoconfiguration. */
2493                 return;
2494         }
2495
2496         idev = addrconf_add_dev(dev);
2497         if (IS_ERR(idev))
2498                 return;
2499
2500         memset(&addr, 0, sizeof(struct in6_addr));
2501         addr.s6_addr32[0] = htonl(0xFE800000);
2502
2503         if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0)
2504                 addrconf_add_linklocal(idev, &addr);
2505 }
2506
2507 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2508 static void addrconf_sit_config(struct net_device *dev)
2509 {
2510         struct inet6_dev *idev;
2511
2512         ASSERT_RTNL();
2513
2514         /*
2515          * Configure the tunnel with one of our IPv4
2516          * addresses... we should configure all of
2517          * our v4 addrs in the tunnel
2518          */
2519
2520         if ((idev = ipv6_find_idev(dev)) == NULL) {
2521                 printk(KERN_DEBUG "init sit: add_dev failed\n");
2522                 return;
2523         }
2524
2525         if (dev->priv_flags & IFF_ISATAP) {
2526                 struct in6_addr addr;
2527
2528                 ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2529                 addrconf_prefix_route(&addr, 64, dev, 0, 0);
2530                 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2531                         addrconf_add_linklocal(idev, &addr);
2532                 return;
2533         }
2534
2535         sit_add_v4_addrs(idev);
2536
2537         if (dev->flags&IFF_POINTOPOINT) {
2538                 addrconf_add_mroute(dev);
2539                 addrconf_add_lroute(dev);
2540         } else
2541                 sit_route_add(dev);
2542 }
2543 #endif
2544
2545 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2546 static void addrconf_gre_config(struct net_device *dev)
2547 {
2548         struct inet6_dev *idev;
2549         struct in6_addr addr;
2550
2551         pr_info("ipv6: addrconf_gre_config(%s)\n", dev->name);
2552
2553         ASSERT_RTNL();
2554
2555         if ((idev = ipv6_find_idev(dev)) == NULL) {
2556                 printk(KERN_DEBUG "init gre: add_dev failed\n");
2557                 return;
2558         }
2559
2560         ipv6_addr_set(&addr,  htonl(0xFE800000), 0, 0, 0);
2561         addrconf_prefix_route(&addr, 64, dev, 0, 0);
2562
2563         if (!ipv6_generate_eui64(addr.s6_addr + 8, dev))
2564                 addrconf_add_linklocal(idev, &addr);
2565 }
2566 #endif
2567
2568 static inline int
2569 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev)
2570 {
2571         struct in6_addr lladdr;
2572
2573         if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) {
2574                 addrconf_add_linklocal(idev, &lladdr);
2575                 return 0;
2576         }
2577         return -1;
2578 }
2579
2580 static void ip6_tnl_add_linklocal(struct inet6_dev *idev)
2581 {
2582         struct net_device *link_dev;
2583         struct net *net = dev_net(idev->dev);
2584
2585         /* first try to inherit the link-local address from the link device */
2586         if (idev->dev->iflink &&
2587             (link_dev = __dev_get_by_index(net, idev->dev->iflink))) {
2588                 if (!ipv6_inherit_linklocal(idev, link_dev))
2589                         return;
2590         }
2591         /* then try to inherit it from any device */
2592         for_each_netdev(net, link_dev) {
2593                 if (!ipv6_inherit_linklocal(idev, link_dev))
2594                         return;
2595         }
2596         printk(KERN_DEBUG "init ip6-ip6: add_linklocal failed\n");
2597 }
2598
2599 /*
2600  * Autoconfigure tunnel with a link-local address so routing protocols,
2601  * DHCPv6, MLD etc. can be run over the virtual link
2602  */
2603
2604 static void addrconf_ip6_tnl_config(struct net_device *dev)
2605 {
2606         struct inet6_dev *idev;
2607
2608         ASSERT_RTNL();
2609
2610         idev = addrconf_add_dev(dev);
2611         if (IS_ERR(idev)) {
2612                 printk(KERN_DEBUG "init ip6-ip6: add_dev failed\n");
2613                 return;
2614         }
2615         ip6_tnl_add_linklocal(idev);
2616 }
2617
2618 static int addrconf_notify(struct notifier_block *this, unsigned long event,
2619                            void * data)
2620 {
2621         struct net_device *dev = (struct net_device *) data;
2622         struct inet6_dev *idev = __in6_dev_get(dev);
2623         int run_pending = 0;
2624         int err;
2625
2626         switch (event) {
2627         case NETDEV_REGISTER:
2628                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2629                         idev = ipv6_add_dev(dev);
2630                         if (!idev)
2631                                 return notifier_from_errno(-ENOMEM);
2632                 }
2633                 break;
2634
2635         case NETDEV_UP:
2636         case NETDEV_CHANGE:
2637                 if (dev->flags & IFF_SLAVE)
2638                         break;
2639
2640                 if (event == NETDEV_UP) {
2641                         if (!addrconf_qdisc_ok(dev)) {
2642                                 /* device is not ready yet. */
2643                                 printk(KERN_INFO
2644                                         "ADDRCONF(NETDEV_UP): %s: "
2645                                         "link is not ready\n",
2646                                         dev->name);
2647                                 break;
2648                         }
2649
2650                         if (!idev && dev->mtu >= IPV6_MIN_MTU)
2651                                 idev = ipv6_add_dev(dev);
2652
2653                         if (idev) {
2654                                 idev->if_flags |= IF_READY;
2655                                 run_pending = 1;
2656                         }
2657                 } else {
2658                         if (!addrconf_qdisc_ok(dev)) {
2659                                 /* device is still not ready. */
2660                                 break;
2661                         }
2662
2663                         if (idev) {
2664                                 if (idev->if_flags & IF_READY)
2665                                         /* device is already configured. */
2666                                         break;
2667                                 idev->if_flags |= IF_READY;
2668                         }
2669
2670                         printk(KERN_INFO
2671                                         "ADDRCONF(NETDEV_CHANGE): %s: "
2672                                         "link becomes ready\n",
2673                                         dev->name);
2674
2675                         run_pending = 1;
2676                 }
2677
2678                 switch (dev->type) {
2679 #if defined(CONFIG_IPV6_SIT) || defined(CONFIG_IPV6_SIT_MODULE)
2680                 case ARPHRD_SIT:
2681                         addrconf_sit_config(dev);
2682                         break;
2683 #endif
2684 #if defined(CONFIG_NET_IPGRE) || defined(CONFIG_NET_IPGRE_MODULE)
2685                 case ARPHRD_IPGRE:
2686                         addrconf_gre_config(dev);
2687                         break;
2688 #endif
2689                 case ARPHRD_TUNNEL6:
2690                         addrconf_ip6_tnl_config(dev);
2691                         break;
2692                 case ARPHRD_LOOPBACK:
2693                         init_loopback(dev);
2694                         break;
2695
2696                 default:
2697                         addrconf_dev_config(dev);
2698                         break;
2699                 }
2700
2701                 if (idev) {
2702                         if (run_pending)
2703                                 addrconf_dad_run(idev);
2704
2705                         /*
2706                          * If the MTU changed during the interface down,
2707                          * when the interface up, the changed MTU must be
2708                          * reflected in the idev as well as routers.
2709                          */
2710                         if (idev->cnf.mtu6 != dev->mtu &&
2711                             dev->mtu >= IPV6_MIN_MTU) {
2712                                 rt6_mtu_change(dev, dev->mtu);
2713                                 idev->cnf.mtu6 = dev->mtu;
2714                         }
2715                         idev->tstamp = jiffies;
2716                         inet6_ifinfo_notify(RTM_NEWLINK, idev);
2717
2718                         /*
2719                          * If the changed mtu during down is lower than
2720                          * IPV6_MIN_MTU stop IPv6 on this interface.
2721                          */
2722                         if (dev->mtu < IPV6_MIN_MTU)
2723                                 addrconf_ifdown(dev, 1);
2724                 }
2725                 break;
2726
2727         case NETDEV_CHANGEMTU:
2728                 if (idev && dev->mtu >= IPV6_MIN_MTU) {
2729                         rt6_mtu_change(dev, dev->mtu);
2730                         idev->cnf.mtu6 = dev->mtu;
2731                         break;
2732                 }
2733
2734                 if (!idev && dev->mtu >= IPV6_MIN_MTU) {
2735                         idev = ipv6_add_dev(dev);
2736                         if (idev)
2737                                 break;
2738                 }
2739
2740                 /*
2741                  * MTU falled under IPV6_MIN_MTU.
2742                  * Stop IPv6 on this interface.
2743                  */
2744
2745         case NETDEV_DOWN:
2746         case NETDEV_UNREGISTER:
2747                 /*
2748                  *      Remove all addresses from this interface.
2749                  */
2750                 addrconf_ifdown(dev, event != NETDEV_DOWN);
2751                 break;
2752
2753         case NETDEV_CHANGENAME:
2754                 if (idev) {
2755                         snmp6_unregister_dev(idev);
2756                         addrconf_sysctl_unregister(idev);
2757                         addrconf_sysctl_register(idev);
2758                         err = snmp6_register_dev(idev);
2759                         if (err)
2760                                 return notifier_from_errno(err);
2761                 }
2762                 break;
2763
2764         case NETDEV_PRE_TYPE_CHANGE:
2765         case NETDEV_POST_TYPE_CHANGE:
2766                 addrconf_type_change(dev, event);
2767                 break;
2768         }
2769
2770         return NOTIFY_OK;
2771 }
2772
2773 /*
2774  *      addrconf module should be notified of a device going up
2775  */
2776 static struct notifier_block ipv6_dev_notf = {
2777         .notifier_call = addrconf_notify,
2778 };
2779
2780 static void addrconf_type_change(struct net_device *dev, unsigned long event)
2781 {
2782         struct inet6_dev *idev;
2783         ASSERT_RTNL();
2784
2785         idev = __in6_dev_get(dev);
2786
2787         if (event == NETDEV_POST_TYPE_CHANGE)
2788                 ipv6_mc_remap(idev);
2789         else if (event == NETDEV_PRE_TYPE_CHANGE)
2790                 ipv6_mc_unmap(idev);
2791 }
2792
2793 static int addrconf_ifdown(struct net_device *dev, int how)
2794 {
2795         struct net *net = dev_net(dev);
2796         struct inet6_dev *idev;
2797         struct inet6_ifaddr *ifa;
2798         int state, i;
2799
2800         ASSERT_RTNL();
2801
2802         rt6_ifdown(net, dev);
2803         neigh_ifdown(&nd_tbl, dev);
2804
2805         idev = __in6_dev_get(dev);
2806         if (idev == NULL)
2807                 return -ENODEV;
2808
2809         /*
2810          * Step 1: remove reference to ipv6 device from parent device.
2811          *         Do not dev_put!
2812          */
2813         if (how) {
2814                 idev->dead = 1;
2815
2816                 /* protected by rtnl_lock */
2817                 RCU_INIT_POINTER(dev->ip6_ptr, NULL);
2818
2819                 /* Step 1.5: remove snmp6 entry */
2820                 snmp6_unregister_dev(idev);
2821
2822         }
2823
2824         /* Step 2: clear hash table */
2825         for (i = 0; i < IN6_ADDR_HSIZE; i++) {
2826                 struct hlist_head *h = &inet6_addr_lst[i];
2827                 struct hlist_node *n;
2828
2829                 spin_lock_bh(&addrconf_hash_lock);
2830         restart:
2831                 hlist_for_each_entry_rcu(ifa, n, h, addr_lst) {
2832                         if (ifa->idev == idev) {
2833                                 hlist_del_init_rcu(&ifa->addr_lst);
2834                                 addrconf_del_timer(ifa);
2835                                 goto restart;
2836                         }
2837                 }
2838                 spin_unlock_bh(&addrconf_hash_lock);
2839         }
2840
2841         write_lock_bh(&idev->lock);
2842
2843         /* Step 2: clear flags for stateless addrconf */
2844         if (!how)
2845                 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY);
2846
2847 #ifdef CONFIG_IPV6_PRIVACY
2848         if (how && del_timer(&idev->regen_timer))
2849                 in6_dev_put(idev);
2850
2851         /* Step 3: clear tempaddr list */
2852         while (!list_empty(&idev->tempaddr_list)) {
2853                 ifa = list_first_entry(&idev->tempaddr_list,
2854                                        struct inet6_ifaddr, tmp_list);
2855                 list_del(&ifa->tmp_list);
2856                 write_unlock_bh(&idev->lock);
2857                 spin_lock_bh(&ifa->lock);
2858
2859                 if (ifa->ifpub) {
2860                         in6_ifa_put(ifa->ifpub);
2861                         ifa->ifpub = NULL;
2862                 }
2863                 spin_unlock_bh(&ifa->lock);
2864                 in6_ifa_put(ifa);
2865                 write_lock_bh(&idev->lock);
2866         }
2867 #endif
2868
2869         while (!list_empty(&idev->addr_list)) {
2870                 ifa = list_first_entry(&idev->addr_list,
2871                                        struct inet6_ifaddr, if_list);
2872                 addrconf_del_timer(ifa);
2873
2874                 list_del(&ifa->if_list);
2875
2876                 write_unlock_bh(&idev->lock);
2877
2878                 spin_lock_bh(&ifa->state_lock);
2879                 state = ifa->state;
2880                 ifa->state = INET6_IFADDR_STATE_DEAD;
2881                 spin_unlock_bh(&ifa->state_lock);
2882
2883                 if (state != INET6_IFADDR_STATE_DEAD) {
2884                         __ipv6_ifa_notify(RTM_DELADDR, ifa);
2885                         atomic_notifier_call_chain(&inet6addr_chain, NETDEV_DOWN, ifa);
2886                 }
2887                 in6_ifa_put(ifa);
2888
2889                 write_lock_bh(&idev->lock);
2890         }
2891
2892         write_unlock_bh(&idev->lock);
2893
2894         /* Step 5: Discard multicast list */
2895         if (how)
2896                 ipv6_mc_destroy_dev(idev);
2897         else
2898                 ipv6_mc_down(idev);
2899
2900         idev->tstamp = jiffies;
2901
2902         /* Last: Shot the device (if unregistered) */
2903         if (how) {
2904                 addrconf_sysctl_unregister(idev);
2905                 neigh_parms_release(&nd_tbl, idev->nd_parms);
2906                 neigh_ifdown(&nd_tbl, dev);
2907                 in6_dev_put(idev);
2908         }
2909         return 0;
2910 }
2911
2912 static void addrconf_rs_timer(unsigned long data)
2913 {
2914         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
2915         struct inet6_dev *idev = ifp->idev;
2916
2917         read_lock(&idev->lock);
2918         if (idev->dead || !(idev->if_flags & IF_READY))
2919                 goto out;
2920
2921         if (idev->cnf.forwarding)
2922                 goto out;
2923
2924         /* Announcement received after solicitation was sent */
2925         if (idev->if_flags & IF_RA_RCVD)
2926                 goto out;
2927
2928         spin_lock(&ifp->lock);
2929         if (ifp->probes++ < idev->cnf.rtr_solicits) {
2930                 /* The wait after the last probe can be shorter */
2931                 addrconf_mod_timer(ifp, AC_RS,
2932                                    (ifp->probes == idev->cnf.rtr_solicits) ?
2933                                    idev->cnf.rtr_solicit_delay :
2934                                    idev->cnf.rtr_solicit_interval);
2935                 spin_unlock(&ifp->lock);
2936
2937                 ndisc_send_rs(idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
2938         } else {
2939                 spin_unlock(&ifp->lock);
2940                 /*
2941                  * Note: we do not support deprecated "all on-link"
2942                  * assumption any longer.
2943                  */
2944                 printk(KERN_DEBUG "%s: no IPv6 routers present\n",
2945                        idev->dev->name);
2946         }
2947
2948 out:
2949         read_unlock(&idev->lock);
2950         in6_ifa_put(ifp);
2951 }
2952
2953 /*
2954  *      Duplicate Address Detection
2955  */
2956 static void addrconf_dad_kick(struct inet6_ifaddr *ifp)
2957 {
2958         unsigned long rand_num;
2959         struct inet6_dev *idev = ifp->idev;
2960
2961         if (ifp->flags & IFA_F_OPTIMISTIC)
2962                 rand_num = 0;
2963         else
2964                 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1);
2965
2966         ifp->probes = idev->cnf.dad_transmits;
2967         addrconf_mod_timer(ifp, AC_DAD, rand_num);
2968 }
2969
2970 static void addrconf_dad_start(struct inet6_ifaddr *ifp, u32 flags)
2971 {
2972         struct inet6_dev *idev = ifp->idev;
2973         struct net_device *dev = idev->dev;
2974
2975         addrconf_join_solict(dev, &ifp->addr);
2976
2977         net_srandom(ifp->addr.s6_addr32[3]);
2978
2979         read_lock_bh(&idev->lock);
2980         spin_lock(&ifp->lock);
2981         if (ifp->state == INET6_IFADDR_STATE_DEAD)
2982                 goto out;
2983
2984         if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) ||
2985             idev->cnf.accept_dad < 1 ||
2986             !(ifp->flags&IFA_F_TENTATIVE) ||
2987             ifp->flags & IFA_F_NODAD) {
2988                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
2989                 spin_unlock(&ifp->lock);
2990                 read_unlock_bh(&idev->lock);
2991
2992                 addrconf_dad_completed(ifp);
2993                 return;
2994         }
2995
2996         if (!(idev->if_flags & IF_READY)) {
2997                 spin_unlock(&ifp->lock);
2998                 read_unlock_bh(&idev->lock);
2999                 /*
3000                  * If the device is not ready:
3001                  * - keep it tentative if it is a permanent address.
3002                  * - otherwise, kill it.
3003                  */
3004                 in6_ifa_hold(ifp);
3005                 addrconf_dad_stop(ifp, 0);
3006                 return;
3007         }
3008
3009         /*
3010          * Optimistic nodes can start receiving
3011          * Frames right away
3012          */
3013         if (ifp->flags & IFA_F_OPTIMISTIC)
3014                 ip6_ins_rt(ifp->rt);
3015
3016         addrconf_dad_kick(ifp);
3017 out:
3018         spin_unlock(&ifp->lock);
3019         read_unlock_bh(&idev->lock);
3020 }
3021
3022 static void addrconf_dad_timer(unsigned long data)
3023 {
3024         struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data;
3025         struct inet6_dev *idev = ifp->idev;
3026         struct in6_addr mcaddr;
3027
3028         if (!ifp->probes && addrconf_dad_end(ifp))
3029                 goto out;
3030
3031         read_lock(&idev->lock);
3032         if (idev->dead || !(idev->if_flags & IF_READY)) {
3033                 read_unlock(&idev->lock);
3034                 goto out;
3035         }
3036
3037         spin_lock(&ifp->lock);
3038         if (ifp->state == INET6_IFADDR_STATE_DEAD) {
3039                 spin_unlock(&ifp->lock);
3040                 read_unlock(&idev->lock);
3041                 goto out;
3042         }
3043
3044         if (ifp->probes == 0) {
3045                 /*
3046                  * DAD was successful
3047                  */
3048
3049                 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED);
3050                 spin_unlock(&ifp->lock);
3051                 read_unlock(&idev->lock);
3052
3053                 addrconf_dad_completed(ifp);
3054
3055                 goto out;
3056         }
3057
3058         ifp->probes--;
3059         addrconf_mod_timer(ifp, AC_DAD, ifp->idev->nd_parms->retrans_time);
3060         spin_unlock(&ifp->lock);
3061         read_unlock(&idev->lock);
3062
3063         /* send a neighbour solicitation for our addr */
3064         addrconf_addr_solict_mult(&ifp->addr, &mcaddr);
3065         ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any);
3066 out:
3067         in6_ifa_put(ifp);
3068 }
3069
3070 static void addrconf_dad_completed(struct inet6_ifaddr *ifp)
3071 {
3072         struct net_device *dev = ifp->idev->dev;
3073
3074         /*
3075          *      Configure the address for reception. Now it is valid.
3076          */
3077
3078         ipv6_ifa_notify(RTM_NEWADDR, ifp);
3079
3080         /* If added prefix is link local and we are prepared to process
3081            router advertisements, start sending router solicitations.
3082          */
3083
3084         if (((ifp->idev->cnf.accept_ra == 1 && !ifp->idev->cnf.forwarding) ||
3085              ifp->idev->cnf.accept_ra == 2) &&
3086             ifp->idev->cnf.rtr_solicits > 0 &&
3087             (dev->flags&IFF_LOOPBACK) == 0 &&
3088             (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) {
3089                 /*
3090                  *      If a host as already performed a random delay
3091                  *      [...] as part of DAD [...] there is no need
3092                  *      to delay again before sending the first RS
3093                  */
3094                 ndisc_send_rs(ifp->idev->dev, &ifp->addr, &in6addr_linklocal_allrouters);
3095
3096                 spin_lock_bh(&ifp->lock);
3097                 ifp->probes = 1;
3098                 ifp->idev->if_flags |= IF_RS_SENT;
3099                 addrconf_mod_timer(ifp, AC_RS, ifp->idev->cnf.rtr_solicit_interval);
3100                 spin_unlock_bh(&ifp->lock);
3101         }
3102 }
3103
3104 static void addrconf_dad_run(struct inet6_dev *idev)
3105 {
3106         struct inet6_ifaddr *ifp;
3107
3108         read_lock_bh(&idev->lock);
3109         list_for_each_entry(ifp, &idev->addr_list, if_list) {
3110                 spin_lock(&ifp->lock);
3111                 if (ifp->flags & IFA_F_TENTATIVE &&
3112                     ifp->state == INET6_IFADDR_STATE_DAD)
3113                         addrconf_dad_kick(ifp);
3114                 spin_unlock(&ifp->lock);
3115         }
3116         read_unlock_bh(&idev->lock);
3117 }
3118
3119 #ifdef CONFIG_PROC_FS
3120 struct if6_iter_state {
3121         struct seq_net_private p;
3122         int bucket;
3123 };
3124
3125 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq)
3126 {
3127         struct inet6_ifaddr *ifa = NULL;
3128         struct if6_iter_state *state = seq->private;
3129         struct net *net = seq_file_net(seq);
3130
3131         for (state->bucket = 0; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) {
3132                 struct hlist_node *n;
3133                 hlist_for_each_entry_rcu_bh(ifa, n, &inet6_addr_lst[state->bucket],
3134                                          addr_lst)
3135                         if (net_eq(dev_net(ifa->idev->dev), net))
3136                                 return ifa;